From 0a30315eb57ef6445663921ec9e4e25a88acfde1 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Thu, 1 Sep 2022 00:29:22 +0800
Subject: [PATCH 01/13] feat(stdio): add stdio without wide api

---
 doc/stdio.md                                 |  97 +++
 src/Makefile                                 |   6 +-
 src/arch/generic/bits/errno.h                |   5 +-
 src/arch/riscv64/include/bits/types.h        |   2 +
 src/arch/riscv64/include/limits.h            |   5 +
 src/arch/x86/include/bits/types.h            |   4 +-
 src/arch/x86/include/limits.h                |   5 +
 src/ctype/alpha.h                            | 172 +++++
 src/ctype/casemap.h                          | 297 ++++++++
 src/ctype/iswalnum.c                         |   6 +
 src/ctype/iswalpha.c                         |  14 +
 src/ctype/iswblank.c                         |   7 +
 src/ctype/iswcntrl.c                         |   9 +
 src/ctype/iswctype.c                         |  62 ++
 src/ctype/iswdigit.c                         |   8 +
 src/ctype/iswgraph.c                         |   7 +
 src/ctype/iswlower.c                         |   6 +
 src/ctype/iswprint.c                         |  19 +
 src/ctype/iswpunct.c                         |  12 +
 src/ctype/iswspace.c                         |  17 +
 src/ctype/iswupper.c                         |   6 +
 src/ctype/iswxdigit.c                        |   6 +
 src/ctype/nonspacing.h                       |  89 +++
 src/ctype/punct.h                            | 141 ++++
 src/ctype/towctrans.c                        |  71 ++
 src/ctype/wcswidth.c                         |   8 +
 src/ctype/wctrans.c                          |  16 +
 src/ctype/wcwidth.c                          |  29 +
 src/ctype/wide.h                             |  65 ++
 src/{crt/start.c => env/__libc_start_main.c} |  11 +
 src/include/features.h                       |   4 +
 src/include/locale.h                         |  60 ++
 src/include/math.h                           |  27 +
 src/include/stdio.h                          | 139 ++++
 src/include/stdlib.h                         |  31 +
 src/include/string.h                         |   4 +
 src/include/uchar.h                          |  27 +
 src/include/wchar.h                          | 129 ++++
 src/include/wctype.h                         |  47 ++
 src/internal/floatscan.c                     | 507 ++++++++++++++
 src/internal/floatscan.h                     |   8 +
 src/internal/intscan.c                       | 100 +++
 src/internal/intscan.h                       |   8 +
 src/internal/libc.c                          |   9 +
 src/internal/libc.h                          |  30 +
 src/internal/shgetc.c                        |  37 +
 src/internal/shgetc.h                        |  32 +
 src/internal/stdio_impl.h                    |  84 +++
 src/internal/version.c                       |   4 +
 src/internal/version.h                       |   6 +
 src/locale/localeconv.c                      |  34 +
 src/locale/setlocale.c                       |   8 +
 src/math/__signbit.c                         |  13 +
 src/math/__signbitf.c                        |  11 +
 src/math/__signbitl.c                        |  14 +
 src/math/copysignl.c                         |  16 +
 src/math/fabsl.c                             |  15 +
 src/math/fdiml.c                             |  18 +
 src/math/floorl.c                            |  34 +
 src/math/fmaxl.c                             |  21 +
 src/math/fminl.c                             |  21 +
 src/math/fmodl.c                             | 105 +++
 src/math/frexpl.c                            |  29 +
 src/math/scalblnl.c                          |  19 +
 src/math/scalbnl.c                           |  36 +
 src/multibyte/wcrtomb.c                      |  37 +
 src/stdio/__fclose_ca.c                      |   6 +
 src/stdio/__fdopen.c                         |  61 ++
 src/stdio/__fmodeflags.c                     |  16 +
 src/stdio/__fopen_rb_ca.c                    |  21 +
 src/stdio/__init_stdio.c                     |  19 +
 src/stdio/__lockfile.c                       |  10 +
 src/stdio/__overflow.c                       |  10 +
 src/stdio/__stdio_close.c                    |  13 +
 src/stdio/__stdio_exit.c                     |  25 +
 src/stdio/__stdio_read.c                     |  50 ++
 src/stdio/__stdio_seek.c                     |   7 +
 src/stdio/__stdio_write.c                    |  60 ++
 src/stdio/__stdout_write.c                   |   9 +
 src/stdio/__toread.c                         |  19 +
 src/stdio/__towrite.c                        |  24 +
 src/stdio/__uflow.c                          |  11 +
 src/stdio/asprintf.c_                        |  13 +
 src/stdio/clearerr.c                         |  10 +
 src/stdio/dprintf.c_                         |  12 +
 src/stdio/ext.c_                             |  57 ++
 src/stdio/ext2.c_                            |  24 +
 src/stdio/fclose.c                           |  38 +
 src/stdio/feof.c                             |  14 +
 src/stdio/ferror.c                           |  14 +
 src/stdio/fflush.c                           |  47 ++
 src/stdio/fgetc.c                            |   7 +
 src/stdio/fgetln.c_                          |  21 +
 src/stdio/fgetpos.c                          |  11 +
 src/stdio/fgets.c                            |  48 ++
 src/stdio/fgetwc.c_                          |  68 ++
 src/stdio/fgetws.c_                          |  33 +
 src/stdio/fileno.c_                          |  16 +
 src/stdio/flockfile.c_                       |   9 +
 src/stdio/fmemopen.c_                        | 127 ++++
 src/stdio/fopen.c                            |  30 +
 src/stdio/fopencookie.c_                     | 135 ++++
 src/stdio/fprintf.c                          |  12 +
 src/stdio/fputc.c                            |   7 +
 src/stdio/fputs.c                            |  10 +
 src/stdio/fputwc.c_                          |  40 ++
 src/stdio/fputws.c_                          |  29 +
 src/stdio/fread.c                            |  38 +
 src/stdio/freopen.c                          |  48 ++
 src/stdio/fscanf.c                           |  14 +
 src/stdio/fseek.c                            |  43 ++
 src/stdio/fsetpos.c                          |   8 +
 src/stdio/ftell.c                            |  41 ++
 src/stdio/ftrylockfile.c_                    |  46 ++
 src/stdio/funlockfile.c_                     |  13 +
 src/stdio/fwide.c_                           |  16 +
 src/stdio/fwprintf.c_                        |  13 +
 src/stdio/fwrite.c                           |  38 +
 src/stdio/fwscanf.c_                         |  15 +
 src/stdio/getc.c                             |   9 +
 src/stdio/getc.h                             |  15 +
 src/stdio/getc_unlocked.c_                   |   9 +
 src/stdio/getchar.c                          |   7 +
 src/stdio/getchar_unlocked.c_                |   6 +
 src/stdio/getdelim.c_                        |  81 +++
 src/stdio/getline.c_                         |   6 +
 src/stdio/gets.c                             |  15 +
 src/stdio/getw.c_                            |   8 +
 src/stdio/getwc.c_                           |   7 +
 src/stdio/getwchar.c_                        |   9 +
 src/stdio/ofl.c                              |  14 +
 src/stdio/ofl_add.c                          |  11 +
 src/stdio/open_memstream.c_                  |  99 +++
 src/stdio/open_wmemstream.c_                 | 102 +++
 src/stdio/pclose.c_                          |  13 +
 src/stdio/perror.c                           |  30 +
 src/stdio/popen.c_                           |  73 ++
 src/stdio/printf.c                           |  12 +
 src/stdio/putc.c                             |   9 +
 src/stdio/putc.h                             |  17 +
 src/stdio/putc_unlocked.c_                   |   9 +
 src/stdio/putchar.c                          |   7 +
 src/stdio/putchar_unlocked.c_                |   6 +
 src/stdio/puts.c                             |  10 +
 src/stdio/putw.c_                            |   7 +
 src/stdio/putwc.c_                           |   7 +
 src/stdio/putwchar.c_                        |   9 +
 src/stdio/remove.c                           |   8 +
 src/stdio/rename.c                           |   7 +
 src/stdio/rewind.c                           |   9 +
 src/stdio/scanf.c                            |  14 +
 src/stdio/setbuf.c                           |   6 +
 src/stdio/setbuffer.c_                       |   7 +
 src/stdio/setlinebuf.c_                      |   7 +
 src/stdio/setvbuf.c                          |  29 +
 src/stdio/snprintf.c                         |  13 +
 src/stdio/sprintf.c                          |  12 +
 src/stdio/sscanf.c                           |  14 +
 src/stdio/stderr.c                           |  18 +
 src/stdio/stdin.c                            |  17 +
 src/stdio/stdout.c                           |  18 +
 src/stdio/swprintf.c_                        |  13 +
 src/stdio/swscanf.c_                         |  14 +
 src/stdio/tempnam.c_                         |  49 ++
 src/stdio/tmpfile.c                          |  27 +
 src/stdio/tmpnam.c                           |  22 +
 src/stdio/ungetc.c                           |  20 +
 src/stdio/ungetwc.c_                         |  35 +
 src/stdio/vasprintf.c_                       |  15 +
 src/stdio/vdprintf.c_                        |  11 +
 src/stdio/vfprintf.c                         | 699 +++++++++++++++++++
 src/stdio/vfscanf.c                          | 340 +++++++++
 src/stdio/vfwprintf.c_                       | 365 ++++++++++
 src/stdio/vfwscanf.c_                        | 332 +++++++++
 src/stdio/vprintf.c                          |   6 +
 src/stdio/vscanf.c                           |   9 +
 src/stdio/vsnprintf.c                        |  55 ++
 src/stdio/vsprintf.c                         |   7 +
 src/stdio/vsscanf.c                          |  27 +
 src/stdio/vswprintf.c_                       |  60 ++
 src/stdio/vswscanf.c_                        |  38 +
 src/stdio/vwprintf.c_                        |   7 +
 src/stdio/vwscanf.c_                         |  10 +
 src/stdio/wprintf.c_                         |  13 +
 src/stdio/wscanf.c_                          |  15 +
 src/string/strnlen.c                         |   7 +
 src/string/wcschr.c                          |   8 +
 src/string/wcslen.c                          |   8 +
 src/string/wcsrchr.c                         |   8 +
 src/temp/__randname.c                        |  20 +
 190 files changed, 7475 insertions(+), 3 deletions(-)
 create mode 100644 doc/stdio.md
 create mode 100644 src/ctype/alpha.h
 create mode 100644 src/ctype/casemap.h
 create mode 100644 src/ctype/iswalnum.c
 create mode 100644 src/ctype/iswalpha.c
 create mode 100644 src/ctype/iswblank.c
 create mode 100644 src/ctype/iswcntrl.c
 create mode 100644 src/ctype/iswctype.c
 create mode 100644 src/ctype/iswdigit.c
 create mode 100644 src/ctype/iswgraph.c
 create mode 100644 src/ctype/iswlower.c
 create mode 100644 src/ctype/iswprint.c
 create mode 100644 src/ctype/iswpunct.c
 create mode 100644 src/ctype/iswspace.c
 create mode 100644 src/ctype/iswupper.c
 create mode 100644 src/ctype/iswxdigit.c
 create mode 100644 src/ctype/nonspacing.h
 create mode 100644 src/ctype/punct.h
 create mode 100644 src/ctype/towctrans.c
 create mode 100644 src/ctype/wcswidth.c
 create mode 100644 src/ctype/wctrans.c
 create mode 100644 src/ctype/wcwidth.c
 create mode 100644 src/ctype/wide.h
 rename src/{crt/start.c => env/__libc_start_main.c} (70%)
 create mode 100644 src/include/locale.h
 create mode 100644 src/include/uchar.h
 create mode 100644 src/include/wctype.h
 create mode 100644 src/internal/floatscan.c
 create mode 100644 src/internal/floatscan.h
 create mode 100644 src/internal/intscan.c
 create mode 100644 src/internal/intscan.h
 create mode 100644 src/internal/libc.c
 create mode 100644 src/internal/libc.h
 create mode 100644 src/internal/shgetc.c
 create mode 100644 src/internal/shgetc.h
 create mode 100644 src/internal/stdio_impl.h
 create mode 100644 src/internal/version.c
 create mode 100644 src/internal/version.h
 create mode 100644 src/locale/localeconv.c
 create mode 100644 src/locale/setlocale.c
 create mode 100644 src/math/__signbit.c
 create mode 100644 src/math/__signbitf.c
 create mode 100644 src/math/__signbitl.c
 create mode 100644 src/math/copysignl.c
 create mode 100644 src/math/fabsl.c
 create mode 100644 src/math/fdiml.c
 create mode 100644 src/math/floorl.c
 create mode 100644 src/math/fmaxl.c
 create mode 100644 src/math/fminl.c
 create mode 100644 src/math/fmodl.c
 create mode 100644 src/math/frexpl.c
 create mode 100644 src/math/scalblnl.c
 create mode 100644 src/math/scalbnl.c
 create mode 100644 src/multibyte/wcrtomb.c
 create mode 100644 src/stdio/__fclose_ca.c
 create mode 100644 src/stdio/__fdopen.c
 create mode 100644 src/stdio/__fmodeflags.c
 create mode 100644 src/stdio/__fopen_rb_ca.c
 create mode 100644 src/stdio/__init_stdio.c
 create mode 100644 src/stdio/__lockfile.c
 create mode 100644 src/stdio/__overflow.c
 create mode 100644 src/stdio/__stdio_close.c
 create mode 100644 src/stdio/__stdio_exit.c
 create mode 100644 src/stdio/__stdio_read.c
 create mode 100644 src/stdio/__stdio_seek.c
 create mode 100644 src/stdio/__stdio_write.c
 create mode 100644 src/stdio/__stdout_write.c
 create mode 100644 src/stdio/__toread.c
 create mode 100644 src/stdio/__towrite.c
 create mode 100644 src/stdio/__uflow.c
 create mode 100644 src/stdio/asprintf.c_
 create mode 100644 src/stdio/clearerr.c
 create mode 100644 src/stdio/dprintf.c_
 create mode 100644 src/stdio/ext.c_
 create mode 100644 src/stdio/ext2.c_
 create mode 100644 src/stdio/fclose.c
 create mode 100644 src/stdio/feof.c
 create mode 100644 src/stdio/ferror.c
 create mode 100644 src/stdio/fflush.c
 create mode 100644 src/stdio/fgetc.c
 create mode 100644 src/stdio/fgetln.c_
 create mode 100644 src/stdio/fgetpos.c
 create mode 100644 src/stdio/fgets.c
 create mode 100644 src/stdio/fgetwc.c_
 create mode 100644 src/stdio/fgetws.c_
 create mode 100644 src/stdio/fileno.c_
 create mode 100644 src/stdio/flockfile.c_
 create mode 100644 src/stdio/fmemopen.c_
 create mode 100644 src/stdio/fopen.c
 create mode 100644 src/stdio/fopencookie.c_
 create mode 100644 src/stdio/fprintf.c
 create mode 100644 src/stdio/fputc.c
 create mode 100644 src/stdio/fputs.c
 create mode 100644 src/stdio/fputwc.c_
 create mode 100644 src/stdio/fputws.c_
 create mode 100644 src/stdio/fread.c
 create mode 100644 src/stdio/freopen.c
 create mode 100644 src/stdio/fscanf.c
 create mode 100644 src/stdio/fseek.c
 create mode 100644 src/stdio/fsetpos.c
 create mode 100644 src/stdio/ftell.c
 create mode 100644 src/stdio/ftrylockfile.c_
 create mode 100644 src/stdio/funlockfile.c_
 create mode 100644 src/stdio/fwide.c_
 create mode 100644 src/stdio/fwprintf.c_
 create mode 100644 src/stdio/fwrite.c
 create mode 100644 src/stdio/fwscanf.c_
 create mode 100644 src/stdio/getc.c
 create mode 100644 src/stdio/getc.h
 create mode 100644 src/stdio/getc_unlocked.c_
 create mode 100644 src/stdio/getchar.c
 create mode 100644 src/stdio/getchar_unlocked.c_
 create mode 100644 src/stdio/getdelim.c_
 create mode 100644 src/stdio/getline.c_
 create mode 100644 src/stdio/gets.c
 create mode 100644 src/stdio/getw.c_
 create mode 100644 src/stdio/getwc.c_
 create mode 100644 src/stdio/getwchar.c_
 create mode 100644 src/stdio/ofl.c
 create mode 100644 src/stdio/ofl_add.c
 create mode 100644 src/stdio/open_memstream.c_
 create mode 100644 src/stdio/open_wmemstream.c_
 create mode 100644 src/stdio/pclose.c_
 create mode 100644 src/stdio/perror.c
 create mode 100644 src/stdio/popen.c_
 create mode 100644 src/stdio/printf.c
 create mode 100644 src/stdio/putc.c
 create mode 100644 src/stdio/putc.h
 create mode 100644 src/stdio/putc_unlocked.c_
 create mode 100644 src/stdio/putchar.c
 create mode 100644 src/stdio/putchar_unlocked.c_
 create mode 100644 src/stdio/puts.c
 create mode 100644 src/stdio/putw.c_
 create mode 100644 src/stdio/putwc.c_
 create mode 100644 src/stdio/putwchar.c_
 create mode 100644 src/stdio/remove.c
 create mode 100644 src/stdio/rename.c
 create mode 100644 src/stdio/rewind.c
 create mode 100644 src/stdio/scanf.c
 create mode 100644 src/stdio/setbuf.c
 create mode 100644 src/stdio/setbuffer.c_
 create mode 100644 src/stdio/setlinebuf.c_
 create mode 100644 src/stdio/setvbuf.c
 create mode 100644 src/stdio/snprintf.c
 create mode 100644 src/stdio/sprintf.c
 create mode 100644 src/stdio/sscanf.c
 create mode 100644 src/stdio/stderr.c
 create mode 100644 src/stdio/stdin.c
 create mode 100644 src/stdio/stdout.c
 create mode 100644 src/stdio/swprintf.c_
 create mode 100644 src/stdio/swscanf.c_
 create mode 100644 src/stdio/tempnam.c_
 create mode 100644 src/stdio/tmpfile.c
 create mode 100644 src/stdio/tmpnam.c
 create mode 100644 src/stdio/ungetc.c
 create mode 100644 src/stdio/ungetwc.c_
 create mode 100644 src/stdio/vasprintf.c_
 create mode 100644 src/stdio/vdprintf.c_
 create mode 100644 src/stdio/vfprintf.c
 create mode 100644 src/stdio/vfscanf.c
 create mode 100644 src/stdio/vfwprintf.c_
 create mode 100644 src/stdio/vfwscanf.c_
 create mode 100644 src/stdio/vprintf.c
 create mode 100644 src/stdio/vscanf.c
 create mode 100644 src/stdio/vsnprintf.c
 create mode 100644 src/stdio/vsprintf.c
 create mode 100644 src/stdio/vsscanf.c
 create mode 100644 src/stdio/vswprintf.c_
 create mode 100644 src/stdio/vswscanf.c_
 create mode 100644 src/stdio/vwprintf.c_
 create mode 100644 src/stdio/vwscanf.c_
 create mode 100644 src/stdio/wprintf.c_
 create mode 100644 src/stdio/wscanf.c_
 create mode 100644 src/string/strnlen.c
 create mode 100644 src/string/wcschr.c
 create mode 100644 src/string/wcslen.c
 create mode 100644 src/string/wcsrchr.c
 create mode 100644 src/temp/__randname.c

diff --git a/doc/stdio.md b/doc/stdio.md
new file mode 100644
index 0000000..e2be7bb
--- /dev/null
+++ b/doc/stdio.md
@@ -0,0 +1,97 @@
+# stdio （标准输入输出）
+
+| 类型 |  支持情况 |
+| ------- |  ---------- |
+| FILE | N/A |
+| fpos_t | N/A |
+
+| 标准流 |  支持情况 |
+| ------- |  ---------- |
+| stdin | N/A |
+| stdout | N/A |
+| stderr | N/A |
+
+| 接口 |  支持情况 |
+| ------- |  ---------- |
+| fopen | N/A |
+| freopen | N/A |
+| fclose | N/A |
+| fflush | N/A |
+| setbuf | N/A |
+| setvbuf | N/A |
+| fwide | N/A |
+| fread | N/A |
+| fwrite | N/A |
+| fgetc | N/A |
+| getc | N/A |
+| fgets | N/A |
+| fputc | N/A |
+| putc | N/A |
+| fputs | N/A |
+| getchar | N/A |
+| gets | N/A |
+| putchar | N/A |
+| puts | N/A |
+| ungetc | N/A |
+| fgetwc | N/A |
+| getwc | N/A |
+| fgetws | N/A |
+| fputwc | N/A |
+| putwc | N/A |
+| fputws | N/A |
+| getwchar | N/A |
+| putwchar | N/A |
+| ungetwc | N/A |
+| scanf | N/A |
+| fscanf | N/A |
+| sscanf | N/A |
+| vscanf | N/A |
+| vfscanf | N/A |
+| vsscanf | N/A |
+| printf | N/A |
+| fprintf | N/A |
+| sprintf | N/A |
+| snprintf | N/A |
+| vprintf | N/A |
+| vfprintf | N/A |
+| vsprintf | N/A |
+| vsnprintf | N/A |
+| wscanf | N/A |
+| fwscanf | N/A |
+| swscanf | N/A |
+| vwscanf | N/A |
+| vfwscanf | N/A |
+| wprintf | N/A |
+| fwprintf | N/A |
+| swprintf | N/A |
+| vwprintf | N/A |
+| vfprintf | N/A |
+| vswprintf | N/A |
+| ftell | N/A |
+| fgetpos | N/A |
+| fseek | N/A |
+| fsetpos | N/A |
+| rewind | N/A |
+| clearerr | N/A |
+| feof | N/A |
+| ferror | N/A |
+| perror | N/A |
+| remove | N/A |
+| rename | N/A |
+| tmpfile | N/A |
+| tmpnam | N/A |
+
+| 宏定义 |  支持情况 |
+| ------- |  ---------- |
+| EOF | N/A |
+| FOPEN_MAX | N/A |
+| FILENAME_MAX | N/A |
+| BUFSIZ | N/A |
+| _IOFBF | N/A |
+| _IOLBF | N/A |
+| _IONBF | N/A |
+| SEEK_SET | N/A |
+| SEEK_CUR | N/A |
+| SEEK_END | N/A |
+| TMP_MAX | N/A |
+| L_tmpnam | N/A |
diff --git a/src/Makefile b/src/Makefile
index e8ca307..ce4073e 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -43,9 +43,13 @@ X_LIBDIRS	:= $(LIBS_DIR)
 # we must link nxbase lib.
 X_LIBS		+= libnxbase.a
 
+SRC			+= stdio/
+SRC			+= locale/
+SRC			+= temp/
+SRC			+= internal/
 SRC			+= exit/
 SRC			+= ctype/
-SRC			+= crt/
+SRC			+= env/
 SRC			+= errno/
 SRC			+= string/
 SRC			+= math/
diff --git a/src/arch/generic/bits/errno.h b/src/arch/generic/bits/errno.h
index e4ce380..fb7901d 100644
--- a/src/arch/generic/bits/errno.h
+++ b/src/arch/generic/bits/errno.h
@@ -1,4 +1,7 @@
 /* ANSI C errno value */
 #define EDOM            1 /* Mathematics argument out of domain of function */
 #define ERANGE          2 /* Result too large */
-#define EILSEQ          3 /* Illegal byte sequence */
\ No newline at end of file
+#define EILSEQ          3 /* Illegal byte sequence */
+
+#define EINVAL          4 /* Invalid value */
+#define EOVERFLOW       5 /* Overflow */
\ No newline at end of file
diff --git a/src/arch/riscv64/include/bits/types.h b/src/arch/riscv64/include/bits/types.h
index d771648..eab49eb 100644
--- a/src/arch/riscv64/include/bits/types.h
+++ b/src/arch/riscv64/include/bits/types.h
@@ -23,3 +23,5 @@ typedef signed long	long		ssize_t;
 typedef signed int				bool_t;
 
 typedef int wchar_t;   
+typedef unsigned wint_t;
+typedef unsigned long wctype_t;
diff --git a/src/arch/riscv64/include/limits.h b/src/arch/riscv64/include/limits.h
index e3789a2..4908d22 100644
--- a/src/arch/riscv64/include/limits.h
+++ b/src/arch/riscv64/include/limits.h
@@ -54,6 +54,11 @@ extern "C" {
 /* Maximum value an 'max uint' can hold (Minimum is 0) */
 #define UINTMAX_MAX		ULLONG_MAX
 
+#define NL_ARGMAX 9
+#define NL_MSGMAX 32767
+#define NL_SETMAX 255
+#define NL_TEXTMAX 2048
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/arch/x86/include/bits/types.h b/src/arch/x86/include/bits/types.h
index 7eba42d..1aadc7f 100644
--- a/src/arch/x86/include/bits/types.h
+++ b/src/arch/x86/include/bits/types.h
@@ -25,4 +25,6 @@ typedef signed long long		loff_t;
 
 typedef signed int				bool_t;
 
-typedef unsigned short wchar_t;   
+typedef unsigned short wchar_t;
+typedef unsigned wint_t;
+typedef unsigned long wctype_t;
diff --git a/src/arch/x86/include/limits.h b/src/arch/x86/include/limits.h
index 0807c03..637d049 100644
--- a/src/arch/x86/include/limits.h
+++ b/src/arch/x86/include/limits.h
@@ -54,6 +54,11 @@ extern "C" {
 /* Maximum value an 'max uint' can hold (Minimum is 0) */
 #define UINTMAX_MAX		ULLONG_MAX
 
+#define NL_ARGMAX 9
+#define NL_MSGMAX 32767
+#define NL_SETMAX 255
+#define NL_TEXTMAX 2048
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/ctype/alpha.h b/src/ctype/alpha.h
new file mode 100644
index 0000000..4167f38
--- /dev/null
+++ b/src/ctype/alpha.h
@@ -0,0 +1,172 @@
+18,17,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,17,34,35,36,17,37,38,39,40,
+41,42,43,44,17,45,46,47,16,16,48,16,16,16,16,16,16,16,49,50,51,16,52,53,16,16,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,54,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,55,17,17,17,17,56,17,57,58,59,60,61,62,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,63,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,64,65,17,66,67,
+68,69,70,71,72,73,74,17,75,76,77,78,79,80,81,16,82,83,84,85,86,87,88,89,90,91,
+92,93,16,94,95,96,16,17,17,17,97,98,99,16,16,16,16,16,16,16,16,16,16,17,17,17,
+17,100,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,17,101,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,17,17,102,103,16,16,104,105,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,106,17,17,107,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,
+108,109,16,16,16,16,16,16,16,16,16,110,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,111,112,113,114,16,16,16,16,16,16,16,16,115,116,
+117,16,16,16,16,16,118,119,16,16,16,16,120,16,16,121,16,16,16,16,16,16,16,16,
+16,16,16,16,16,
+16,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,254,255,255,7,254,
+255,255,7,0,0,0,0,0,4,32,4,255,255,127,255,255,255,127,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,195,255,3,0,31,80,0,0,0,0,0,0,0,0,0,0,32,0,0,0,0,0,223,188,64,215,255,255,
+251,255,255,255,255,255,255,255,255,255,191,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,3,252,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,254,255,255,255,127,2,255,255,255,
+255,255,1,0,0,0,0,255,191,182,0,255,255,255,135,7,0,0,0,255,7,255,255,255,255,
+255,255,255,254,255,195,255,255,255,255,255,255,255,255,255,255,255,255,239,
+31,254,225,255,
+159,0,0,255,255,255,255,255,255,0,224,255,255,255,255,255,255,255,255,255,255,
+255,255,3,0,255,255,255,255,255,7,48,4,255,255,255,252,255,31,0,0,255,255,255,
+1,255,7,0,0,0,0,0,0,255,255,223,63,0,0,240,255,248,3,255,255,255,255,255,255,
+255,255,255,239,255,223,225,255,207,255,254,255,239,159,249,255,255,253,197,
+227,159,89,128,176,207,255,3,16,238,135,249,255,255,253,109,195,135,25,2,94,
+192,255,63,0,238,191,251,255,255,253,237,227,191,27,1,0,207,255,0,30,238,159,
+249,255,255,253,237,227,159,25,192,176,207,255,2,0,236,199,61,214,24,199,255,
+195,199,29,129,0,192,255,0,0,239,223,253,255,255,253,255,227,223,29,96,7,207,
+255,0,0,239,223,253,255,255,253,239,227,223,29,96,64,207,255,6,0,239,223,253,
+255,255,255,255,231,223,93,240,128,207,255,0,252,236,255,127,252,255,255,251,
+47,127,128,95,255,192,255,12,0,254,255,255,255,255,127,255,7,63,32,255,3,0,0,
+0,0,214,247,255,255,175,255,255,59,95,32,255,243,0,0,0,
+0,1,0,0,0,255,3,0,0,255,254,255,255,255,31,254,255,3,255,255,254,255,255,255,
+31,0,0,0,0,0,0,0,0,255,255,255,255,255,255,127,249,255,3,255,255,255,255,255,
+255,255,255,255,63,255,255,255,255,191,32,255,255,255,255,255,247,255,255,255,
+255,255,255,255,255,255,61,127,61,255,255,255,255,255,61,255,255,255,255,61,
+127,61,255,127,255,255,255,255,255,255,255,61,255,255,255,255,255,255,255,255,
+7,0,0,0,0,255,255,0,0,255,255,255,255,255,255,255,255,255,255,63,63,254,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,159,255,255,254,255,255,7,255,255,255,255,255,255,255,255,
+255,199,255,1,255,223,15,0,255,255,15,0,255,255,15,0,255,223,13,0,255,255,255,
+255,255,255,207,255,255,1,128,16,255,3,0,0,0,0,255,3,255,255,255,255,255,255,
+255,255,255,255,255,1,255,255,255,255,255,7,255,255,255,255,255,255,255,255,
+63,
+0,255,255,255,127,255,15,255,1,192,255,255,255,255,63,31,0,255,255,255,255,
+255,15,255,255,255,3,255,3,0,0,0,0,255,255,255,15,255,255,255,255,255,255,255,
+127,254,255,31,0,255,3,255,3,128,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,
+255,239,255,239,15,255,3,0,0,0,0,255,255,255,255,255,243,255,255,255,255,255,
+255,191,255,3,0,255,255,255,255,255,255,127,0,255,227,255,255,255,255,255,63,
+255,1,255,255,255,255,255,231,0,0,0,0,0,222,111,4,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,
+128,255,31,0,255,255,63,63,255,255,255,255,63,63,255,170,255,255,255,63,255,
+255,255,255,255,255,223,95,220,31,207,15,255,31,220,31,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,2,128,0,0,255,31,0,0,0,0,0,0,0,0,0,0,0,0,132,252,47,62,80,189,255,243,
+224,67,0,0,255,255,255,255,255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,192,255,255,255,255,255,255,3,0,
+0,255,255,255,255,255,127,255,255,255,255,255,127,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,31,120,12,0,255,255,255,255,191,32,255,
+255,255,255,255,255,255,128,0,0,255,255,127,0,127,127,127,127,127,127,127,127,
+255,255,255,255,0,0,0,0,0,128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,224,0,0,0,254,3,62,31,254,255,255,255,255,255,255,255,255,255,127,224,254,
+255,255,255,255,255,255,255,255,255,255,247,224,255,255,255,255,255,254,255,
+255,255,255,255,255,255,255,255,255,127,0,0,255,255,255,7,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,63,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,
+0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,31,0,0,
+0,0,0,0,0,0,255,255,255,255,255,63,255,31,255,255,255,15,0,0,255,255,255,255,
+255,127,240,143,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,
+0,128,255,252,255,255,255,255,255,255,255,255,255,255,255,255,249,255,255,255,
+255,255,255,124,0,0,0,0,0,128,255,191,255,255,255,255,0,0,0,255,255,255,255,
+255,255,15,0,255,255,255,255,255,255,255,255,47,0,255,3,0,0,252,232,255,255,
+255,255,255,7,255,255,255,255,7,0,255,255,255,31,255,255,255,255,255,255,247,
+255,0,128,255,3,255,255,255,127,255,255,255,255,255,255,127,0,255,63,255,3,
+255,255,127,252,255,255,255,255,255,255,255,127,5,0,0,56,255,255,60,0,126,126,
+126,0,127,127,255,255,255,255,255,247,255,0,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,7,255,3,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,15,0,255,255,127,248,255,255,255,255,
+255,
+15,255,255,255,255,255,255,255,255,255,255,255,255,255,63,255,255,255,255,255,
+255,255,255,255,255,255,255,255,3,0,0,0,0,127,0,248,224,255,253,127,95,219,
+255,255,255,255,255,255,255,255,255,255,255,255,255,3,0,0,0,248,255,255,255,
+255,255,255,255,255,255,255,255,255,63,0,0,255,255,255,255,255,255,255,255,
+252,255,255,255,255,255,255,0,0,0,0,0,255,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,223,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,31,0,0,255,3,
+254,255,255,7,254,255,255,7,192,255,255,255,255,255,255,255,255,255,255,127,
+252,252,252,28,0,0,0,0,255,239,255,255,127,255,255,183,255,63,255,63,0,0,0,0,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,7,0,0,0,0,0,0,0,0,
+255,255,255,255,255,255,31,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,255,255,255,31,255,255,255,255,255,255,1,0,0,0,0,
+0,255,255,255,255,0,224,255,255,255,7,255,255,255,255,255,7,255,255,255,63,
+255,255,255,255,15,255,62,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,63,255,3,255,255,255,255,15,255,255,255,
+255,15,255,255,255,255,255,0,255,255,255,255,255,255,15,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,255,255,255,255,255,255,127,0,255,255,63,0,255,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,63,253,255,255,255,255,191,145,255,255,63,0,255,255,
+127,0,255,255,255,127,0,0,0,0,0,0,0,0,255,255,55,0,255,255,63,0,255,255,255,3,
+0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,192,0,0,0,0,0,0,0,0,111,240,239,
+254,255,255,63,0,0,0,0,0,255,255,255,31,255,255,255,31,0,0,0,0,255,254,255,
+255,31,0,0,0,255,255,255,255,255,255,63,0,255,255,63,0,255,255,7,0,255,255,3,
+0,0,0,0,0,0,0,0,0,0,0,0,
+0,255,255,255,255,255,255,255,255,255,1,0,0,0,0,0,0,255,255,255,255,255,255,7,
+0,255,255,255,255,255,255,7,0,255,255,255,255,255,0,255,3,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,31,128,0,255,255,63,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,255,255,127,0,255,255,255,255,255,255,255,255,63,0,0,0,
+192,255,0,0,252,255,255,255,255,255,255,1,0,0,255,255,255,1,255,3,255,255,255,
+255,255,255,199,255,112,0,255,255,255,255,71,0,255,255,255,255,255,255,255,
+255,30,0,255,23,0,0,0,0,255,255,251,255,255,255,159,64,0,0,0,0,0,0,0,0,127,
+189,255,191,255,1,255,255,255,255,255,255,255,1,255,3,239,159,249,255,255,253,
+237,227,159,25,129,224,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,
+255,255,255,255,255,187,7,255,131,0,0,0,0,255,255,255,255,255,255,255,255,179,
+0,255,3,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,63,127,0,0,0,63,0,0,
+0,0,255,255,255,255,255,255,255,127,17,0,255,3,0,0,0,0,255,255,255,255,255,
+255,63,1,255,3,0,0,0,0,0,0,255,255,255,231,255,7,255,3,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,1,0,0,0,0,0,0,0,0,0,0,0,
+0,255,255,255,255,255,255,255,255,255,3,0,128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,255,252,255,255,255,255,255,252,26,0,0,0,255,255,255,255,255,255,231,
+127,0,0,255,255,255,255,255,255,255,255,255,32,0,0,0,0,255,255,255,255,255,
+255,255,1,255,253,255,255,255,255,127,127,1,0,255,3,0,0,252,255,255,255,252,
+255,255,254,127,0,0,0,0,0,0,0,0,0,127,251,255,255,255,255,127,180,203,0,255,3,
+191,253,255,255,255,127,123,1,255,3,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,127,0,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,3,0,0,
+0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,127,0,
+0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,
+255,255,255,127,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,127,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,
+255,255,255,255,255,255,1,255,255,255,127,255,3,0,0,0,0,0,0,0,0,0,0,0,0,255,
+255,255,63,0,0,255,255,255,255,255,255,0,0,15,0,255,3,248,255,255,224,255,255,
+0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,255,255,255,255,255,255,255,255,255,135,255,255,255,255,255,255,255,128,
+255,255,0,0,0,0,0,0,0,0,11,0,0,0,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,7,0,255,255,255,127,0,0,0,0,0,
+0,7,0,240,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,15,255,255,255,255,
+255,255,255,255,255,255,255,255,255,7,255,31,255,1,255,67,0,0,0,0,0,0,0,0,0,0,
+0,0,255,255,255,255,255,255,255,255,255,255,223,255,255,255,255,255,255,255,
+255,223,100,222,255,235,239,255,255,255,255,255,255,
+255,191,231,223,223,255,255,255,123,95,252,253,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,63,255,255,255,
+253,255,255,247,255,255,255,247,255,255,223,255,255,255,223,255,255,127,255,
+255,255,127,255,255,255,253,255,255,255,253,255,255,247,207,255,255,255,255,
+255,255,127,255,255,249,219,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,255,255,255,255,255,31,128,63,255,67,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,
+15,255,3,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,31,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,
+143,8,255,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,239,255,255,255,150,254,247,10,132,234,150,170,150,247,247,94,255,251,255,
+15,238,251,255,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,3,255,255,255,3,255,
+255,255,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
diff --git a/src/ctype/casemap.h b/src/ctype/casemap.h
new file mode 100644
index 0000000..6ee1209
--- /dev/null
+++ b/src/ctype/casemap.h
@@ -0,0 +1,297 @@
+static const unsigned char tab[] = {
+	7, 8, 9, 10, 11, 12, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	13, 6, 6, 14, 6, 6, 6, 6, 6, 6, 6, 6, 15, 16, 17, 18,
+	6, 19, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 20, 21, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 22, 23, 6, 6, 6, 24, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 25,
+	6, 6, 6, 6, 26, 6, 6, 6, 6, 6, 6, 6, 27, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 28, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 29, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 30, 6, 6, 6, 6, 6, 6,
+	6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 36,
+	43, 43, 43, 43, 43, 43, 43, 43, 1, 0, 84, 86, 86, 86, 86, 86,
+	86, 86, 86, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 24, 0, 0, 0, 43, 43, 43, 43, 43, 43,
+	43, 7, 43, 43, 91, 86, 86, 86, 86, 86, 86, 86, 74, 86, 86, 5,
+	49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80,
+	36, 80, 121, 49, 80, 49, 80, 49, 56, 80, 49, 80, 49, 80, 49, 80,
+	49, 80, 49, 80, 49, 80, 49, 80, 78, 49, 2, 78, 13, 13, 78, 3,
+	78, 0, 36, 110, 0, 78, 49, 38, 110, 81, 78, 36, 80, 78, 57, 20,
+	129, 27, 29, 29, 83, 49, 80, 49, 80, 13, 49, 80, 49, 80, 49, 80,
+	27, 83, 36, 80, 49, 2, 92, 123, 92, 123, 92, 123, 92, 123, 92, 123,
+	20, 121, 92, 123, 92, 123, 92, 45, 43, 73, 3, 72, 3, 120, 92, 123,
+	20, 0, 150, 10, 1, 43, 40, 6, 6, 0, 42, 6, 42, 42, 43, 7,
+	187, 181, 43, 30, 0, 43, 7, 43, 43, 43, 1, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 1, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 42, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 205, 70, 205, 43, 0, 37, 43, 7, 1, 6, 1, 85, 86, 86, 86,
+	86, 86, 85, 86, 86, 2, 36, 129, 129, 129, 129, 129, 21, 129, 129, 129,
+	0, 0, 43, 0, 178, 209, 178, 209, 178, 209, 178, 209, 0, 0, 205, 204,
+	1, 0, 215, 215, 215, 215, 215, 131, 129, 129, 129, 129, 129, 129, 129, 129,
+	129, 129, 172, 172, 172, 172, 172, 172, 172, 172, 172, 172, 28, 0, 0, 0,
+	0, 0, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 2, 0, 0,
+	49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80,
+	49, 80, 78, 49, 80, 49, 80, 78, 49, 80, 49, 80, 49, 80, 49, 80,
+	49, 80, 49, 80, 49, 80, 49, 2, 135, 166, 135, 166, 135, 166, 135, 166,
+	135, 166, 135, 166, 135, 166, 135, 166, 42, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 0, 0, 0, 84, 86, 86, 86, 86, 86, 86, 86,
+	86, 86, 86, 86, 86, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 84, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86,
+	12, 0, 12, 42, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 7, 42, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 42, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 86, 86, 108, 129, 21, 0, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 7, 108, 3, 65, 43, 43, 86, 86, 86, 86, 86, 86,
+	86, 86, 86, 86, 86, 86, 86, 86, 44, 86, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 1,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 12, 108, 0, 0, 0, 0, 0, 6,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37,
+	6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37,
+	6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37,
+	6, 37, 6, 37, 6, 37, 6, 37, 86, 122, 158, 38, 6, 37, 6, 37,
+	6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 37,
+	6, 37, 6, 37, 6, 37, 6, 37, 6, 37, 6, 1, 43, 43, 79, 86,
+	86, 44, 43, 127, 86, 86, 57, 43, 43, 85, 86, 86, 43, 43, 79, 86,
+	86, 44, 43, 127, 86, 86, 129, 55, 117, 91, 123, 92, 43, 43, 79, 86,
+	86, 2, 172, 4, 0, 0, 57, 43, 43, 85, 86, 86, 43, 43, 79, 86,
+	86, 44, 43, 43, 86, 86, 50, 19, 129, 87, 0, 111, 129, 126, 201, 215,
+	126, 45, 129, 129, 14, 126, 57, 127, 111, 87, 0, 129, 129, 126, 21, 0,
+	126, 3, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 7, 43,
+	36, 43, 151, 43, 43, 43, 43, 43, 43, 43, 43, 43, 42, 43, 43, 43,
+	43, 43, 86, 86, 86, 86, 86, 128, 129, 129, 129, 129, 57, 187, 42, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 1, 129, 129, 129, 129, 129, 129, 129, 129,
+	129, 129, 129, 129, 129, 129, 129, 201, 172, 172, 172, 172, 172, 172, 172, 172,
+	172, 172, 172, 172, 172, 172, 172, 208, 13, 0, 78, 49, 2, 180, 193, 193,
+	215, 215, 36, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80,
+	49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80, 49, 80,
+	49, 80, 49, 80, 215, 215, 83, 193, 71, 212, 215, 215, 215, 5, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 7, 1, 0, 1, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 78, 49, 80, 49, 80, 49, 80,
+	49, 80, 49, 80, 49, 80, 49, 80, 13, 0, 0, 0, 0, 0, 36, 80,
+	49, 80, 49, 80, 49, 80, 49, 80, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 121, 92, 123, 92, 123, 79, 123, 92, 123, 92, 123,
+	92, 123, 92, 123, 92, 123, 92, 123, 92, 123, 92, 123, 92, 123, 92, 45,
+	43, 43, 121, 20, 92, 123, 92, 45, 121, 42, 92, 39, 92, 123, 92, 123,
+	92, 123, 164, 0, 10, 180, 92, 123, 92, 123, 79, 3, 42, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 1,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 72, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 42, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 43, 43, 43, 43, 43, 43, 43, 43, 7, 0, 72, 86, 86, 86, 86,
+	86, 86, 86, 86, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 43, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 85, 86, 86, 86, 86, 86, 86,
+	86, 86, 86, 86, 86, 86, 14, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 36, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+	43, 43, 7, 0, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 36, 43, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 7, 0, 0,
+	0, 0, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 86,
+	86, 86, 86, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 42, 43, 43,
+	43, 43, 43, 43, 43, 43, 43, 43, 86, 86, 86, 86, 86, 86, 86, 86,
+	86, 86, 14, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 42, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 86, 86,
+	86, 86, 86, 86, 86, 86, 86, 86, 14, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 85,
+	86, 86, 86, 86, 86, 86, 86, 86, 86, 86, 14, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+static const int rules[] = {
+	0x0, 0x2001, -0x2000, 0x1dbf00, 0x2e700, 0x7900,
+	0x2402, 0x101, -0x100, 0x0, 0x201, -0x200,
+	-0xc6ff, -0xe800, -0x78ff, -0x12c00, 0xc300, 0xd201,
+	0xce01, 0xcd01, 0x4f01, 0xca01, 0xcb01, 0xcf01,
+	0x6100, 0xd301, 0xd101, 0xa300, 0xd501, 0x8200,
+	0xd601, 0xda01, 0xd901, 0xdb01, 0x3800, 0x3,
+	-0x4f00, -0x60ff, -0x37ff, 0x242802, 0x0, 0x101,
+	-0x100, -0xcd00, -0xda00, -0x81ff, 0x2a2b01, -0xa2ff,
+	0x2a2801, 0x2a3f00, -0xc2ff, 0x4501, 0x4701, 0x2a1f00,
+	0x2a1c00, 0x2a1e00, -0xd200, -0xce00, -0xca00, -0xcb00,
+	0xa54f00, 0xa54b00, -0xcf00, 0xa52800, 0xa54400, -0xd100,
+	-0xd300, 0x29f700, 0xa54100, 0x29fd00, -0xd500, -0xd600,
+	0x29e700, 0xa54300, 0xa52a00, -0x4500, -0xd900, -0x4700,
+	-0xdb00, 0xa51500, 0xa51200, 0x4c2402, 0x0, 0x2001,
+	-0x2000, 0x101, -0x100, 0x5400, 0x7401, 0x2601,
+	0x2501, 0x4001, 0x3f01, -0x2600, -0x2500, -0x1f00,
+	-0x4000, -0x3f00, 0x801, -0x3e00, -0x3900, -0x2f00,
+	-0x3600, -0x800, -0x5600, -0x5000, 0x700, -0x7400,
+	-0x3bff, -0x6000, -0x6ff, 0x701a02, 0x101, -0x100,
+	0x2001, -0x2000, 0x5001, 0xf01, -0xf00, 0x0,
+	0x3001, -0x3000, 0x101, -0x100, 0x0, 0xbc000,
+	0x1c6001, 0x0, 0x97d001, 0x801, -0x800, 0x8a0502,
+	0x0, -0xbbfff, -0x186200, 0x89c200, -0x182500, -0x186e00,
+	-0x186d00, -0x186400, -0x186300, -0x185c00, 0x0, 0x8a3800,
+	0x8a0400, 0xee600, 0x101, -0x100, 0x0, -0x3b00,
+	-0x1dbeff, 0x8f1d02, 0x800, -0x7ff, 0x0, 0x5600,
+	-0x55ff, 0x4a00, 0x6400, 0x8000, 0x7000, 0x7e00,
+	0x900, -0x49ff, -0x8ff, -0x1c2500, -0x63ff, -0x6fff,
+	-0x7fff, -0x7dff, 0xac0502, 0x0, 0x1001, -0x1000,
+	0x1c01, 0x101, -0x1d5cff, -0x20beff, -0x2045ff, -0x1c00,
+	0xb10b02, 0x101, -0x100, 0x3001, -0x3000, 0x0,
+	-0x29f6ff, -0xee5ff, -0x29e6ff, -0x2a2b00, -0x2a2800, -0x2a1bff,
+	-0x29fcff, -0x2a1eff, -0x2a1dff, -0x2a3eff, 0x0, -0x1c6000,
+	0x0, 0x101, -0x100, 0xbc0c02, 0x0, 0x101,
+	-0x100, -0xa543ff, 0x3a001, -0x8a03ff, -0xa527ff, 0x3000,
+	-0xa54eff, -0xa54aff, -0xa540ff, -0xa511ff, -0xa529ff, -0xa514ff,
+	-0x2fff, -0xa542ff, -0x8a37ff, 0x0, -0x97d000, -0x3a000,
+	0x0, 0x2001, -0x2000, 0x0, 0x2801, -0x2800,
+	0x0, 0x4001, -0x4000, 0x0, 0x2001, -0x2000,
+	0x0, 0x2001, -0x2000, 0x0, 0x2201, -0x2200,
+};
+static const unsigned char rulebases[] = {
+	0, 6, 39, 81, 111, 119, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	124, 0, 0, 127, 0, 0, 0, 0, 0, 0, 0, 0, 131, 142, 146, 151,
+	0, 170, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 180, 196, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 198, 201, 0, 0, 0, 219, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 222,
+	0, 0, 0, 0, 225, 0, 0, 0, 0, 0, 0, 0, 228, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 231, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 234, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 237, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+static const unsigned char exceptions[][2] = {
+	{ 48, 12 }, { 49, 13 }, { 120, 14 }, { 127, 15 },
+	{ 128, 16 }, { 129, 17 }, { 134, 18 }, { 137, 19 },
+	{ 138, 19 }, { 142, 20 }, { 143, 21 }, { 144, 22 },
+	{ 147, 19 }, { 148, 23 }, { 149, 24 }, { 150, 25 },
+	{ 151, 26 }, { 154, 27 }, { 156, 25 }, { 157, 28 },
+	{ 158, 29 }, { 159, 30 }, { 166, 31 }, { 169, 31 },
+	{ 174, 31 }, { 177, 32 }, { 178, 32 }, { 183, 33 },
+	{ 191, 34 }, { 197, 35 }, { 200, 35 }, { 203, 35 },
+	{ 221, 36 }, { 242, 35 }, { 246, 37 }, { 247, 38 },
+	{ 32, 45 }, { 58, 46 }, { 61, 47 }, { 62, 48 },
+	{ 63, 49 }, { 64, 49 }, { 67, 50 }, { 68, 51 },
+	{ 69, 52 }, { 80, 53 }, { 81, 54 }, { 82, 55 },
+	{ 83, 56 }, { 84, 57 }, { 89, 58 }, { 91, 59 },
+	{ 92, 60 }, { 97, 61 }, { 99, 62 }, { 101, 63 },
+	{ 102, 64 }, { 104, 65 }, { 105, 66 }, { 106, 64 },
+	{ 107, 67 }, { 108, 68 }, { 111, 66 }, { 113, 69 },
+	{ 114, 70 }, { 117, 71 }, { 125, 72 }, { 130, 73 },
+	{ 135, 74 }, { 137, 75 }, { 138, 76 }, { 139, 76 },
+	{ 140, 77 }, { 146, 78 }, { 157, 79 }, { 158, 80 },
+	{ 69, 87 }, { 123, 29 }, { 124, 29 }, { 125, 29 },
+	{ 127, 88 }, { 134, 89 }, { 136, 90 }, { 137, 90 },
+	{ 138, 90 }, { 140, 91 }, { 142, 92 }, { 143, 92 },
+	{ 172, 93 }, { 173, 94 }, { 174, 94 }, { 175, 94 },
+	{ 194, 95 }, { 204, 96 }, { 205, 97 }, { 206, 97 },
+	{ 207, 98 }, { 208, 99 }, { 209, 100 }, { 213, 101 },
+	{ 214, 102 }, { 215, 103 }, { 240, 104 }, { 241, 105 },
+	{ 242, 106 }, { 243, 107 }, { 244, 108 }, { 245, 109 },
+	{ 249, 110 }, { 253, 45 }, { 254, 45 }, { 255, 45 },
+	{ 80, 105 }, { 81, 105 }, { 82, 105 }, { 83, 105 },
+	{ 84, 105 }, { 85, 105 }, { 86, 105 }, { 87, 105 },
+	{ 88, 105 }, { 89, 105 }, { 90, 105 }, { 91, 105 },
+	{ 92, 105 }, { 93, 105 }, { 94, 105 }, { 95, 105 },
+	{ 130, 0 }, { 131, 0 }, { 132, 0 }, { 133, 0 },
+	{ 134, 0 }, { 135, 0 }, { 136, 0 }, { 137, 0 },
+	{ 192, 117 }, { 207, 118 }, { 128, 137 }, { 129, 138 },
+	{ 130, 139 }, { 133, 140 }, { 134, 141 }, { 112, 157 },
+	{ 113, 157 }, { 118, 158 }, { 119, 158 }, { 120, 159 },
+	{ 121, 159 }, { 122, 160 }, { 123, 160 }, { 124, 161 },
+	{ 125, 161 }, { 179, 162 }, { 186, 163 }, { 187, 163 },
+	{ 188, 164 }, { 190, 165 }, { 195, 162 }, { 204, 164 },
+	{ 218, 166 }, { 219, 166 }, { 229, 106 }, { 234, 167 },
+	{ 235, 167 }, { 236, 110 }, { 243, 162 }, { 248, 168 },
+	{ 249, 168 }, { 250, 169 }, { 251, 169 }, { 252, 164 },
+	{ 38, 176 }, { 42, 177 }, { 43, 178 }, { 78, 179 },
+	{ 132, 8 }, { 98, 186 }, { 99, 187 }, { 100, 188 },
+	{ 101, 189 }, { 102, 190 }, { 109, 191 }, { 110, 192 },
+	{ 111, 193 }, { 112, 194 }, { 126, 195 }, { 127, 195 },
+	{ 125, 207 }, { 141, 208 }, { 148, 209 }, { 171, 210 },
+	{ 172, 211 }, { 173, 212 }, { 176, 213 }, { 177, 214 },
+	{ 178, 215 }, { 196, 216 }, { 197, 217 }, { 198, 218 },
+};
diff --git a/src/ctype/iswalnum.c b/src/ctype/iswalnum.c
new file mode 100644
index 0000000..5842991
--- /dev/null
+++ b/src/ctype/iswalnum.c
@@ -0,0 +1,6 @@
+#include <wctype.h>
+
+int iswalnum(wint_t wc)
+{
+	return iswdigit(wc) || iswalpha(wc);
+}
diff --git a/src/ctype/iswalpha.c b/src/ctype/iswalpha.c
new file mode 100644
index 0000000..d558fae
--- /dev/null
+++ b/src/ctype/iswalpha.c
@@ -0,0 +1,14 @@
+#include <wctype.h>
+
+static const unsigned char table[] = {
+#include "alpha.h"
+};
+
+int iswalpha(wint_t wc)
+{
+	if (wc<0x20000U)
+		return (table[table[wc>>8]*32+((wc&255)>>3)]>>(wc&7))&1;
+	if (wc<0x2fffeU)
+		return 1;
+	return 0;
+}
diff --git a/src/ctype/iswblank.c b/src/ctype/iswblank.c
new file mode 100644
index 0000000..0a66843
--- /dev/null
+++ b/src/ctype/iswblank.c
@@ -0,0 +1,7 @@
+#include <wctype.h>
+#include <ctype.h>
+
+int iswblank(wint_t wc)
+{
+	return isblank(wc);
+}
diff --git a/src/ctype/iswcntrl.c b/src/ctype/iswcntrl.c
new file mode 100644
index 0000000..6f6edf9
--- /dev/null
+++ b/src/ctype/iswcntrl.c
@@ -0,0 +1,9 @@
+#include <wctype.h>
+
+int iswcntrl(wint_t wc)
+{
+	return (unsigned)wc < 32
+	    || (unsigned)(wc-0x7f) < 33
+	    || (unsigned)(wc-0x2028) < 2
+	    || (unsigned)(wc-0xfff9) < 3;
+}
diff --git a/src/ctype/iswctype.c b/src/ctype/iswctype.c
new file mode 100644
index 0000000..a431c90
--- /dev/null
+++ b/src/ctype/iswctype.c
@@ -0,0 +1,62 @@
+#include <wctype.h>
+#include <string.h>
+
+#define WCTYPE_ALNUM  1
+#define WCTYPE_ALPHA  2
+#define WCTYPE_BLANK  3
+#define WCTYPE_CNTRL  4
+#define WCTYPE_DIGIT  5
+#define WCTYPE_GRAPH  6
+#define WCTYPE_LOWER  7
+#define WCTYPE_PRINT  8
+#define WCTYPE_PUNCT  9
+#define WCTYPE_SPACE  10
+#define WCTYPE_UPPER  11
+#define WCTYPE_XDIGIT 12
+
+int iswctype(wint_t wc, wctype_t type)
+{
+	switch (type) {
+	case WCTYPE_ALNUM:
+		return iswalnum(wc);
+	case WCTYPE_ALPHA:
+		return iswalpha(wc);
+	case WCTYPE_BLANK:
+		return iswblank(wc);
+	case WCTYPE_CNTRL:
+		return iswcntrl(wc);
+	case WCTYPE_DIGIT:
+		return iswdigit(wc);
+	case WCTYPE_GRAPH:
+		return iswgraph(wc);
+	case WCTYPE_LOWER:
+		return iswlower(wc);
+	case WCTYPE_PRINT:
+		return iswprint(wc);
+	case WCTYPE_PUNCT:
+		return iswpunct(wc);
+	case WCTYPE_SPACE:
+		return iswspace(wc);
+	case WCTYPE_UPPER:
+		return iswupper(wc);
+	case WCTYPE_XDIGIT:
+		return iswxdigit(wc);
+	}
+	return 0;
+}
+
+wctype_t wctype(const char *s)
+{
+	int i;
+	const char *p;
+	/* order must match! */
+	static const char names[] =
+		"alnum\0" "alpha\0" "blank\0"
+		"cntrl\0" "digit\0" "graph\0"
+		"lower\0" "print\0" "punct\0"
+		"space\0" "upper\0" "xdigit";
+	for (i=1, p=names; *p; i++, p+=6)
+		if (*s == *p && !strcmp(s, p))
+			return i;
+	return 0;
+}
diff --git a/src/ctype/iswdigit.c b/src/ctype/iswdigit.c
new file mode 100644
index 0000000..b1d9fa9
--- /dev/null
+++ b/src/ctype/iswdigit.c
@@ -0,0 +1,8 @@
+#include <wctype.h>
+
+#undef iswdigit
+
+int iswdigit(wint_t wc)
+{
+	return (unsigned)wc-'0' < 10;
+}
diff --git a/src/ctype/iswgraph.c b/src/ctype/iswgraph.c
new file mode 100644
index 0000000..fdc9785
--- /dev/null
+++ b/src/ctype/iswgraph.c
@@ -0,0 +1,7 @@
+#include <wctype.h>
+
+int iswgraph(wint_t wc)
+{
+	/* ISO C defines this function as: */
+	return !iswspace(wc) && iswprint(wc);
+}
diff --git a/src/ctype/iswlower.c b/src/ctype/iswlower.c
new file mode 100644
index 0000000..0a568e7
--- /dev/null
+++ b/src/ctype/iswlower.c
@@ -0,0 +1,6 @@
+#include <wctype.h>
+
+int iswlower(wint_t wc)
+{
+	return towupper(wc) != wc;
+}
diff --git a/src/ctype/iswprint.c b/src/ctype/iswprint.c
new file mode 100644
index 0000000..333f19c
--- /dev/null
+++ b/src/ctype/iswprint.c
@@ -0,0 +1,19 @@
+#include <wctype.h>
+
+/* Consider all legal codepoints as printable except for:
+ * - C0 and C1 control characters
+ * - U+2028 and U+2029 (line/para break)
+ * - U+FFF9 through U+FFFB (interlinear annotation controls)
+ * The following code is optimized heavily to make hot paths for the
+ * expected printable characters. */
+
+int iswprint(wint_t wc)
+{
+	if (wc < 0xffU)
+		return (wc+1 & 0x7f) >= 0x21;
+	if (wc < 0x2028U || wc-0x202aU < 0xd800-0x202a || wc-0xe000U < 0xfff9-0xe000)
+		return 1;
+	if (wc-0xfffcU > 0x10ffff-0xfffc || (wc&0xfffe)==0xfffe)
+		return 0;
+	return 1;
+}
diff --git a/src/ctype/iswpunct.c b/src/ctype/iswpunct.c
new file mode 100644
index 0000000..16e8703
--- /dev/null
+++ b/src/ctype/iswpunct.c
@@ -0,0 +1,12 @@
+#include <wctype.h>
+
+static const unsigned char table[] = {
+#include "punct.h"
+};
+
+int iswpunct(wint_t wc)
+{
+	if (wc<0x20000U)
+		return (table[table[wc>>8]*32+((wc&255)>>3)]>>(wc&7))&1;
+	return 0;
+}
diff --git a/src/ctype/iswspace.c b/src/ctype/iswspace.c
new file mode 100644
index 0000000..c3b20f3
--- /dev/null
+++ b/src/ctype/iswspace.c
@@ -0,0 +1,17 @@
+#include <wchar.h>
+#include <wctype.h>
+
+/* Our definition of whitespace is the Unicode White_Space property,
+ * minus non-breaking spaces (U+00A0, U+2007, and U+202F) and script-
+ * specific characters with non-blank glyphs (U+1680 and U+180E). */
+
+int iswspace(wint_t wc)
+{
+	static const wchar_t spaces[] = {
+		' ', '\t', '\n', '\r', 11, 12,  0x0085,
+		0x2000, 0x2001, 0x2002, 0x2003, 0x2004, 0x2005,
+		0x2006, 0x2008, 0x2009, 0x200a,
+		0x2028, 0x2029, 0x205f, 0x3000, 0
+	};
+	return wc && wcschr(spaces, wc);
+}
diff --git a/src/ctype/iswupper.c b/src/ctype/iswupper.c
new file mode 100644
index 0000000..eae59a7
--- /dev/null
+++ b/src/ctype/iswupper.c
@@ -0,0 +1,6 @@
+#include <wctype.h>
+
+int iswupper(wint_t wc)
+{
+	return towlower(wc) != wc;
+}
diff --git a/src/ctype/iswxdigit.c b/src/ctype/iswxdigit.c
new file mode 100644
index 0000000..4e0bede
--- /dev/null
+++ b/src/ctype/iswxdigit.c
@@ -0,0 +1,6 @@
+#include <wctype.h>
+
+int iswxdigit(wint_t wc)
+{
+	return (unsigned)(wc-'0') < 10 || (unsigned)((wc|32)-'a') < 6;
+}
diff --git a/src/ctype/nonspacing.h b/src/ctype/nonspacing.h
new file mode 100644
index 0000000..5d05a3d
--- /dev/null
+++ b/src/ctype/nonspacing.h
@@ -0,0 +1,89 @@
+16,16,16,18,19,20,21,22,23,24,25,26,27,28,29,30,31,16,16,32,16,16,16,33,34,35,
+36,37,38,39,16,16,40,16,16,16,16,16,16,16,16,16,16,16,41,42,16,16,43,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,44,16,45,46,47,48,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,49,16,16,50,
+51,16,52,53,54,16,16,16,16,16,16,55,16,16,56,16,57,58,59,60,61,62,63,64,65,66,
+67,68,16,69,70,71,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,72,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,73,74,16,16,16,75,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,76,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,77,78,16,16,16,16,16,16,16,79,16,16,16,16,16,80,81,82,16,16,16,16,16,83,
+84,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,248,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,254,255,255,255,255,191,182,0,0,0,0,0,0,0,63,0,255,23,0,0,0,0,0,248,255,
+255,0,0,1,0,0,0,0,0,0,0,0,0,0,0,192,191,159,61,0,0,0,128,2,0,0,0,255,255,255,
+7,0,0,0,0,0,0,0,0,0,0,192,255,1,0,0,0,0,0,0,248,15,32,0,0,192,251,239,62,0,0,
+0,0,0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0,248,255,255,255,255,
+255,7,0,0,0,0,0,0,20,254,33,254,0,12,0,0,0,2,0,0,0,0,0,0,16,30,32,0,0,12,0,0,
+64,6,0,0,0,0,0,0,16,134,57,2,0,0,0,35,0,6,0,0,0,0,0,0,16,190,33,0,0,12,0,0,
+252,2,0,0,0,0,0,0,144,30,32,64,0,12,0,0,0,4,0,0,0,0,0,0,0,1,32,0,0,0,0,0,0,17,
+0,0,0,0,0,0,192,193,61,96,0,12,0,0,0,2,0,0,0,0,0,0,144,64,48,0,0,12,0,0,0,3,0,
+0,0,0,0,0,24,30,32,0,0,12,0,0,0,0,0,0,0,0,0,0,0,0,4,92,0,0,0,0,0,0,0,0,0,0,0,
+242,7,128,127,0,0,0,0,0,0,0,0,0,0,0,0,242,31,0,63,0,0,0,0,0,0,0,0,0,3,0,0,160,
+2,0,0,0,0,0,0,254,127,223,224,255,254,255,255,255,31,64,0,0,0,0,0,0,0,0,0,0,0,
+0,224,253,102,0,0,0,195,1,0,30,0,100,32,0,32,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,224,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,28,0,
+0,0,28,0,0,0,12,0,0,0,12,0,0,0,0,0,0,0,176,63,64,254,15,32,0,0,0,0,0,120,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,96,0,0,0,0,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,135,1,4,14,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,9,0,0,0,0,0,0,64,127,
+229,31,248,159,0,0,0,0,0,0,255,127,0,0,0,0,0,0,0,0,15,0,0,0,0,0,208,23,4,0,0,
+0,0,248,15,0,3,0,0,0,60,59,0,0,0,0,0,0,64,163,3,0,0,0,0,0,0,240,207,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,247,255,253,33,16,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,
+251,0,248,0,0,0,124,0,0,0,0,0,0,223,255,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,
+255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,3,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,0,0,0,0,
+0,60,0,0,0,0,0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,128,247,63,0,0,0,192,0,0,0,0,0,0,0,0,0,0,3,0,68,8,0,0,96,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,48,0,0,0,255,255,3,128,0,0,0,0,192,63,0,0,128,255,3,0,
+0,0,0,0,7,0,0,0,0,0,200,51,0,0,0,0,32,0,0,0,0,0,0,0,0,126,102,0,8,16,0,0,0,0,
+0,16,0,0,0,0,0,0,157,193,2,0,0,0,0,48,64,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,32,33,0,0,0,0,0,64,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,0,0,255,255,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,32,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,192,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,110,240,0,0,0,0,0,135,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,96,0,0,
+0,0,0,0,0,240,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,192,255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,0,0,0,0,255,
+127,0,0,0,0,0,0,128,3,0,0,0,0,0,120,38,0,32,0,0,0,0,0,0,7,0,0,0,128,239,31,0,
+0,0,0,0,0,0,8,0,3,0,0,0,0,0,192,127,0,30,0,0,0,0,0,0,0,0,0,0,0,128,211,64,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,248,7,0,0,3,0,0,0,0,0,0,24,1,0,0,0,192,
+31,31,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,92,0,0,64,0,0,0,0,0,
+0,0,0,0,0,248,133,13,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,60,176,1,0,0,48,0,0,0,
+0,0,0,0,0,0,0,248,167,1,0,0,0,0,0,0,0,0,0,0,0,0,40,191,0,0,0,0,0,0,0,0,0,0,0,
+0,224,188,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+128,255,6,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,240,12,1,0,0,0,254,7,0,0,0,0,248,121,128,0,
+126,14,0,0,0,0,0,252,127,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,127,191,0,0,0,
+0,0,0,0,0,0,0,252,255,255,252,109,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,126,180,191,0,
+0,0,0,0,0,0,0,0,163,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,24,
+0,0,0,0,0,0,0,255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,31,0,0,0,0,0,0,0,127,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,0,
+0,128,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,96,15,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,3,248,255,231,15,0,0,0,60,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,28,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,255,255,255,255,255,255,127,248,255,255,255,255,255,31,32,0,16,0,0,248,
+254,255,0,0,0,0,0,0,0,0,0,
+0,127,255,255,249,219,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,127,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,240,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,127,0,0,0,0,0,0,0,0,0,0,0,0,0,240,7,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,
diff --git a/src/ctype/punct.h b/src/ctype/punct.h
new file mode 100644
index 0000000..6792947
--- /dev/null
+++ b/src/ctype/punct.h
@@ -0,0 +1,141 @@
+18,16,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,16,16,34,35,16,36,37,38,39,
+40,41,42,43,16,44,45,46,17,17,47,17,17,17,17,17,17,48,49,50,51,52,53,54,55,17,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,56,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,57,16,58,59,60,61,62,63,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,64,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,65,16,16,66,16,67,68,
+69,16,70,71,72,16,73,16,16,74,75,76,77,78,16,79,80,81,82,83,84,85,86,87,88,89,
+90,91,16,92,93,94,95,16,16,16,16,96,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,97,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,98,99,16,16,100,101,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,102,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,103,104,105,106,16,16,107,108,17,17,109,16,16,16,16,16,16,110,111,16,
+16,16,16,16,112,113,16,16,114,115,116,16,117,118,119,17,17,17,120,121,122,123,
+124,16,16,16,16,
+16,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,254,255,0,252,1,0,0,248,1,
+0,0,120,0,0,0,0,255,251,223,251,0,0,128,0,0,0,128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,60,0,252,255,224,175,255,255,255,255,255,255,255,255,
+255,255,223,255,255,255,255,255,32,64,176,0,0,0,0,0,0,0,0,0,0,0,0,0,64,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,252,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,252,0,0,0,0,0,230,254,255,255,255,0,64,73,0,0,0,0,0,24,0,255,255,0,216,
+0,0,0,0,0,0,0,1,0,60,0,0,0,0,0,0,0,0,0,0,0,0,16,224,1,30,0,
+96,255,191,0,0,0,0,0,0,255,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,248,207,
+227,0,0,0,3,0,32,255,127,0,0,0,78,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8,0,7,252,0,0,0,
+0,0,0,0,0,0,16,0,32,30,0,48,0,1,0,0,0,0,0,0,0,0,16,0,32,0,0,0,0,252,111,0,0,0,
+0,0,0,0,16,0,32,0,0,0,0,64,0,0,0,0,0,0,0,0,16,0,32,0,0,0,0,3,224,0,0,0,0,0,0,
+0,16,0,32,0,0,0,0,253,0,0,0,0,0,0,0,0,0,0,32,0,0,0,0,255,7,16,0,0,0,0,0,0,0,0,
+32,0,0,0,0,128,255,16,0,0,0,0,0,0,16,0,32,0,0,0,0,0,0,0,0,0,0,0,0,0,24,0,160,
+0,127,0,0,255,3,0,0,0,0,0,0,0,0,0,4,0,0,0,0,16,0,0,0,0,0,0,128,0,128,192,223,
+0,12,0,0,0,0,0,0,0,0,0,0,0,4,0,31,0,0,0,0,0,
+0,254,255,255,255,0,252,255,255,0,0,0,0,0,0,0,0,252,0,0,0,0,0,0,192,255,223,
+255,7,0,0,0,0,0,0,0,0,0,0,128,6,0,252,0,0,0,0,0,0,0,0,0,192,0,0,0,0,0,0,0,0,0,
+0,0,8,0,0,0,0,0,0,0,0,0,0,0,224,255,255,255,31,0,0,255,3,0,0,0,0,0,0,0,0,0,0,
+0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,96,0,0,1,0,0,24,0,0,0,0,0,0,0,0,0,56,0,0,0,0,16,0,0,0,112,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,48,0,0,254,127,47,0,0,255,3,255,127,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,14,49,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,196,255,255,255,
+255,0,0,0,192,0,0,0,0,0,0,0,0,1,0,224,159,0,0,0,0,127,63,255,127,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,16,0,16,0,0,252,255,255,255,31,0,0,0,0,0,12,0,0,0,0,0,0,64,0,
+12,240,0,0,0,0,0,0,128,248,0,0,0,0,0,0,0,192,0,0,0,0,0,0,0,0,255,0,255,255,
+255,33,144,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,
+127,0,224,251,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,160,3,224,0,224,0,
+224,0,96,128,248,255,255,255,252,255,255,255,255,255,127,223,255,241,127,255,
+127,0,0,255,255,255,255,0,0,255,255,255,255,1,0,123,3,208,193,175,66,0,12,31,
+188,255,255,0,0,0,0,0,14,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,127,0,0,0,255,7,0,0,255,255,255,255,255,255,255,255,255,
+255,63,0,0,0,0,0,0,252,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,207,255,255,255,
+63,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,224,135,3,254,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,
+128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,127,255,255,255,255,0,
+0,0,0,0,0,255,255,255,251,255,255,255,255,255,255,255,255,255,255,15,0,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,63,0,0,0,255,15,30,255,255,255,1,252,193,224,0,0,0,0,
+0,0,0,0,0,0,0,30,1,0,0,0,0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+255,255,0,0,0,0,255,255,255,255,15,0,0,0,255,255,255,127,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,
+255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,
+255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,127,0,0,0,
+0,0,0,192,0,224,0,0,0,0,0,0,0,0,0,0,0,128,15,112,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+255,0,255,255,127,0,3,0,0,0,0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+64,0,0,0,0,15,255,3,0,0,0,0,0,0,240,0,0,0,0,0,0,0,0,0,16,192,0,0,255,255,3,23,
+0,0,0,0,0,248,0,0,0,0,8,128,0,0,0,0,0,0,0,0,0,0,8,0,255,63,0,192,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,240,0,0,128,3,0,0,0,0,0,0,0,128,2,0,0,192,0,0,67,0,0,0,0,0,
+0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,56,0,
+0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,2,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,252,255,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,192,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,48,255,255,255,3,255,255,255,255,255,255,247,
+255,127,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,254,255,0,252,1,0,0,248,1,0,
+0,248,63,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,127,127,0,48,135,255,255,255,255,255,
+143,255,0,0,0,0,0,0,224,255,255,127,255,15,1,0,0,0,0,0,255,255,255,255,255,63,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,
+15,0,0,0,0,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+128,255,0,0,128,255,0,0,0,0,128,255,0,0,0,0,0,0,0,0,0,248,0,0,192,143,0,0,0,
+128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,48,255,255,252,255,255,255,255,255,0,0,0,0,
+0,0,0,135,255,1,255,1,0,0,0,224,0,0,0,224,0,0,0,0,0,1,0,0,96,248,127,0,0,0,0,
+0,0,0,0,254,0,0,0,255,0,0,0,255,0,0,0,30,0,254,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,252,0,0,0,0,0,0,0,0,0,0,0,
+0,255,255,255,127,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,224,127,0,0,0,192,255,255,3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,192,63,252,255,63,0,0,128,3,0,0,0,0,0,0,254,3,32,0,0,0,0,0,0,0,
+0,0,0,0,0,24,0,15,0,0,0,0,0,56,0,0,0,0,0,0,0,0,0,225,63,0,232,254,255,31,0,0,
+0,0,0,0,0,96,63,0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,0,0,0,0,0,6,0,0,0,0,0,0,0,0,0,
+24,0,32,0,0,192,31,31,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,68,
+248,0,104,0,0,0,0,0,0,0,0,0,0,0,0,76,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,128,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,128,14,0,0,0,255,
+31,0,0,0,0,0,0,0,0,192,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,8,0,252,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,252,7,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,24,128,255,0,0,0,0,0,
+0,0,0,0,0,223,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,62,0,0,252,255,31,3,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,52,0,0,0,0,0,0,0,0,0,128,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,128,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,
+255,3,
+128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,31,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,192,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,63,0,0,0,0,0,0,0,255,255,48,0,0,248,
+3,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,
+255,255,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,4,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,176,15,0,0,0,0,0,0,
+0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,63,
+0,255,255,255,255,127,254,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,1,0,0,255,255,255,255,255,255,255,255,
+63,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,15,0,255,255,255,255,255,255,
+255,255,255,255,127,0,255,255,255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,8,0,0,0,8,0,0,32,0,0,0,32,0,0,128,
+0,0,0,128,0,0,0,2,0,0,0,2,0,0,8,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,15,0,248,254,255,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,127,0,0,128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,240,0,
+128,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,255,127,0,0,0,0,0,0,0,
+0,0,0,0,0,0,112,7,0,192,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,254,255,255,255,255,255,255,255,31,0,0,0,0,0,0,0,0,0,254,255,
+255,255,255,255,255,63,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,3,0,255,255,255,255,255,
+15,255,255,255,255,255,255,255,255,255,255,255,255,15,0,255,127,254,255,254,
+255,254,255,255,255,63,0,255,31,255,255,255,255,0,0,0,252,0,0,0,28,0,0,0,252,
+255,255,255,31,0,0,0,0,0,0,192,255,255,255,7,0,255,255,255,255,255,15,255,1,3,
+0,63,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,63,0,255,31,255,7,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,15,0,255,255,255,255,255,255,255,255,255,255,255,1,
+255,15,0,0,255,15,255,255,255,255,255,255,255,0,255,3,255,255,255,255,255,0,
+255,255,255,63,0,0,0,0,0,0,0,0,0,0,255,239,255,255,255,255,255,255,255,255,
+255,255,255,255,123,252,255,255,255,255,231,199,255,255,255,231,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,15,0,255,63,15,7,7,0,63,0,
+0,0,0,0,0,0,0,0,0,0,0,0,
diff --git a/src/ctype/towctrans.c b/src/ctype/towctrans.c
new file mode 100644
index 0000000..7645362
--- /dev/null
+++ b/src/ctype/towctrans.c
@@ -0,0 +1,71 @@
+#include <wctype.h>
+
+static const unsigned char tab[];
+
+static const unsigned char rulebases[512];
+static const int rules[];
+
+static const unsigned char exceptions[][2];
+
+#include "casemap.h"
+
+static int casemap(unsigned c, int dir)
+{
+	unsigned b, x, y, v, rt, xb, xn;
+	int r, rd, c0 = c;
+
+	if (c >= 0x20000) return c;
+
+	b = c>>8;
+	c &= 255;
+	x = c/3;
+	y = c%3;
+
+	/* lookup entry in two-level base-6 table */
+	v = tab[tab[b]*86+x];
+	static const int mt[] = { 2048, 342, 57 };
+	v = (v*mt[y]>>11)%6;
+
+	/* use the bit vector out of the tables as an index into
+	 * a block-specific set of rules and decode the rule into
+	 * a type and a case-mapping delta. */
+	r = rules[rulebases[b]+v];
+	rt = r & 255;
+	rd = r >> 8;
+
+	/* rules 0/1 are simple lower/upper case with a delta.
+	 * apply according to desired mapping direction. */
+	if (rt < 2) return c0 + (rd & -(rt^dir));
+
+	/* binary search. endpoints of the binary search for
+	 * this block are stored in the rule delta field. */
+	xn = rd & 0xff;
+	xb = (unsigned)rd >> 8;
+	while (xn) {
+		unsigned try = exceptions[xb+xn/2][0];
+		if (try == c) {
+			r = rules[exceptions[xb+xn/2][1]];
+			rt = r & 255;
+			rd = r >> 8;
+			if (rt < 2) return c0 + (rd & -(rt^dir));
+			/* Hard-coded for the four exceptional titlecase */
+			return c0 + (dir ? -1 : 1);
+		} else if (try > c) {
+			xn /= 2;
+		} else {
+			xb += xn/2;
+			xn -= xn/2;
+		}
+	}
+	return c0;
+}
+
+wint_t towlower(wint_t wc)
+{
+	return casemap(wc, 0);
+}
+
+wint_t towupper(wint_t wc)
+{
+	return casemap(wc, 1);
+}
diff --git a/src/ctype/wcswidth.c b/src/ctype/wcswidth.c
new file mode 100644
index 0000000..5c8a5a4
--- /dev/null
+++ b/src/ctype/wcswidth.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+int wcswidth(const wchar_t *wcs, size_t n)
+{
+	int l=0, k=0;
+	for (; n-- && *wcs && (k = wcwidth(*wcs)) >= 0; l+=k, wcs++);
+	return (k < 0) ? k : l;
+}
diff --git a/src/ctype/wctrans.c b/src/ctype/wctrans.c
new file mode 100644
index 0000000..739869d
--- /dev/null
+++ b/src/ctype/wctrans.c
@@ -0,0 +1,16 @@
+#include <wctype.h>
+#include <string.h>
+
+wctrans_t wctrans(const char *class)
+{
+	if (!strcmp(class, "toupper")) return (wctrans_t)1;
+	if (!strcmp(class, "tolower")) return (wctrans_t)2;
+	return 0;
+}
+
+wint_t towctrans(wint_t wc, wctrans_t trans)
+{
+	if (trans == (wctrans_t)1) return towupper(wc);
+	if (trans == (wctrans_t)2) return towlower(wc);
+	return wc;
+}
diff --git a/src/ctype/wcwidth.c b/src/ctype/wcwidth.c
new file mode 100644
index 0000000..36256a5
--- /dev/null
+++ b/src/ctype/wcwidth.c
@@ -0,0 +1,29 @@
+#include <wchar.h>
+
+static const unsigned char table[] = {
+#include "nonspacing.h"
+};
+
+static const unsigned char wtable[] = {
+#include "wide.h"
+};
+
+int wcwidth(wchar_t wc)
+{
+	if (wc < 0xffU)
+		return (wc+1 & 0x7f) >= 0x21 ? 1 : wc ? -1 : 0;
+	if ((wc & 0xfffeffffU) < 0xfffe) {
+		if ((table[table[wc>>8]*32+((wc&255)>>3)]>>(wc&7))&1)
+			return 0;
+		if ((wtable[wtable[wc>>8]*32+((wc&255)>>3)]>>(wc&7))&1)
+			return 2;
+		return 1;
+	}
+	if ((wc & 0xfffe) == 0xfffe)
+		return -1;
+	if (wc-0x20000U < 0x20000)
+		return 2;
+	if (wc == 0xe0001 || wc-0xe0020U < 0x5f || wc-0xe0100U < 0xef)
+		return 0;
+	return 1;
+}
diff --git a/src/ctype/wide.h b/src/ctype/wide.h
new file mode 100644
index 0000000..e403c9a
--- /dev/null
+++ b/src/ctype/wide.h
@@ -0,0 +1,65 @@
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,18,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,19,16,20,21,22,16,16,16,23,16,16,24,25,26,27,28,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,29,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,30,16,16,16,16,31,16,16,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,17,32,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,17,16,16,16,33,
+34,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,35,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,
+17,17,17,17,17,17,36,17,17,37,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,38,39,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,
+16,16,16,16,16,16,16,40,41,42,43,44,45,46,47,16,48,49,16,16,16,16,
+16,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,12,0,6,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,30,9,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,96,0,0,48,0,0,0,0,0,0,255,15,0,0,0,0,128,0,0,8,
+0,2,12,0,96,48,64,16,0,0,4,44,36,32,12,0,0,0,1,0,0,0,80,184,0,0,0,0,0,0,0,224,
+0,0,0,1,128,0,0,0,0,0,0,0,0,0,0,0,24,0,0,0,0,0,0,33,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,251,255,255,255,255,255,255,255,
+255,255,255,15,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,63,0,0,0,255,15,255,255,255,255,
+255,255,255,127,254,255,255,255,255,255,255,255,255,255,127,254,255,255,255,
+255,255,255,255,255,255,255,255,255,224,255,255,255,255,255,254,255,255,255,
+255,255,255,255,255,255,255,127,255,255,255,255,255,7,255,255,255,255,15,0,
+255,255,255,255,255,127,255,255,255,255,255,0,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,
+0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,31,255,255,255,255,255,255,127,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,
+255,255,31,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,15,0,0,0,0,0,0,0,0,0,0,0,0,0,255,3,0,0,255,255,255,255,247,255,127,15,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,254,255,255,255,255,255,255,255,255,255,255,
+255,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,127,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,15,0,0,0,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,7,0,255,255,255,127,0,0,0,0,0,
+0,7,0,240,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+15,16,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,128,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,0,64,254,7,0,0,0,0,0,0,0,0,0,0,0,0,7,0,255,255,255,
+255,255,15,255,1,3,0,63,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,
+1,224,191,255,255,255,255,255,255,255,255,223,255,255,15,0,255,255,255,255,
+255,135,15,0,255,255,17,255,255,255,255,255,255,255,255,127,253,255,255,255,
+255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
+159,255,255,255,255,255,255,255,63,0,120,255,255,255,0,0,4,0,0,96,0,16,0,0,0,
+0,0,0,0,0,0,0,248,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,
+255,255,255,255,255,255,63,16,39,0,0,24,240,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,0,0,0,0,0,0,0,0,0,0,0,255,15,0,
+0,0,224,255,255,255,255,255,255,255,255,255,255,255,255,123,252,255,255,255,
+255,231,199,255,255,255,231,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,
+0,15,7,7,0,63,0,0,0,0,0,0,0,0,0,0,0,0,0,
diff --git a/src/crt/start.c b/src/env/__libc_start_main.c
similarity index 70%
rename from src/crt/start.c
rename to src/env/__libc_start_main.c
index 0079bfd..3109e61 100644
--- a/src/crt/start.c
+++ b/src/env/__libc_start_main.c
@@ -1,10 +1,20 @@
 #include <nxbase.h>
 #include <stdlib.h>
+#include <assert.h>
+#include "stdio_impl.h"
+#include "libc.h"
 
 #define MAX_ARGS 1024
 
 extern int main(int argc, char *argv[]);
 
+void __init_libc(char **envp, char *pn)
+{
+    libc.page_size = 4096;
+
+    __init_stdio();
+}
+
 NX_Error NX_WEAK_SYM NX_Main(char * cmdline, char * envline)
 {
     /* build cmdline and envline */
@@ -14,6 +24,7 @@ NX_Error NX_WEAK_SYM NX_Main(char * cmdline, char * envline)
 
     argc = NX_CmdToArray(cmdline, argv, MAX_ARGS);
 
+    __init_libc(NX_NULL, NX_NULL);
     ret = main(argc, argv);
     if (ret == EXIT_SUCCESS)
     {
diff --git a/src/include/features.h b/src/include/features.h
index c108bb9..46f7f81 100644
--- a/src/include/features.h
+++ b/src/include/features.h
@@ -22,4 +22,8 @@
 
 #define __REDIR(x,y) __typeof__(x) x __asm__(#y)
 
+#define hidden __attribute__((__visibility__("hidden")))
+#define weak_alias(old, new) \
+	extern __typeof(old) new __attribute__((__weak__, __alias__(#old)))
+
 #endif
diff --git a/src/include/locale.h b/src/include/locale.h
new file mode 100644
index 0000000..d6f3b14
--- /dev/null
+++ b/src/include/locale.h
@@ -0,0 +1,60 @@
+#ifndef	_LOCALE_H
+#define	_LOCALE_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <features.h>
+
+#ifdef __cplusplus
+#define NULL 0L
+#else
+#define NULL ((void*)0)
+#endif
+
+#define LC_CTYPE    0
+#define LC_NUMERIC  1
+#define LC_TIME     2
+#define LC_COLLATE  3
+#define LC_MONETARY 4
+#define LC_MESSAGES 5
+#define LC_ALL      6
+
+struct lconv {
+	char *decimal_point;
+	char *thousands_sep;
+	char *grouping;
+
+	char *int_curr_symbol;
+	char *currency_symbol;
+	char *mon_decimal_point;
+	char *mon_thousands_sep;
+	char *mon_grouping;
+	char *positive_sign;
+	char *negative_sign;
+	char int_frac_digits;
+	char frac_digits;
+	char p_cs_precedes;
+	char p_sep_by_space;
+	char n_cs_precedes;
+	char n_sep_by_space;
+	char p_sign_posn;
+	char n_sign_posn;
+	char int_p_cs_precedes;
+	char int_p_sep_by_space;
+	char int_n_cs_precedes;
+	char int_n_sep_by_space;
+	char int_p_sign_posn;
+	char int_n_sign_posn;
+};
+
+
+char *setlocale (int, const char *);
+struct lconv *localeconv(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/include/math.h b/src/include/math.h
index 443bf99..a21cd8e 100644
--- a/src/include/math.h
+++ b/src/include/math.h
@@ -94,6 +94,15 @@ do {														\
 	(d) = __u.f;											\
 } while (0)
 
+int __signbit(double);
+int __signbitf(float);
+int __signbitl(long double);
+
+#define signbit(x) ( \
+	sizeof(x) == sizeof(float) ? (int)(__FLOAT_BITS(x)>>31) : \
+	sizeof(x) == sizeof(double) ? (int)(__DOUBLE_BITS(x)>>63) : \
+	__signbitl(x) )
+
 #define FP_NAN					0
 #define FP_INFINITE				1
 #define FP_ZERO					2
@@ -196,14 +205,23 @@ double	expm1(double);
 float	expm1f(float);
 double	fabs(double);
 float	fabsf(float);
+long double fabsl(long double);
 double	fdim(double, double);
 float	fdimf(float, float);
+long double fdiml(long double, long double);
+
 double	floor(double);
 float	floorf(float);
+long double floorl(long double);
+
 double	fmod(double, double);
 float	fmodf(float, float);
+long double fmodl(long double, long double);
+
 double	frexp(double, int *);
 float	frexpf(float, int *);
+long double frexpl(long double, int *);
+
 double	hypot(double, double);
 float	hypotf(float, float);
 double	ldexp(double, int);
@@ -230,8 +248,12 @@ float	roundf(float);
 long double roundl(long double x);
 double	scalbn(double, int);
 float	scalbnf(float, int);
+long double scalblnl(long double x, long n);
+
 double	scalbln(double, long);
 float	scalblnf(float, long);
+long double scalbnl(long double x, int n);
+
 double	sin(double);
 float	sinf(float);
 double	sinh(double);
@@ -246,10 +268,15 @@ double	trunc(double);
 float	truncf(float);
 double fmax(double, double);
 float fmaxf(float, float);
+long double fmaxl(long double, long double);
+
 double fmin(double, double);
 float fminf(float, float);
+long double fminl(long double, long double);
+
 float copysignf(float x, float y);
 double copysign(double x, double y);
+long double copysignl(long double x, long double y);
 
 /*
  * libm kernel functions
diff --git a/src/include/stdio.h b/src/include/stdio.h
index 3ea8373..1a4aff8 100644
--- a/src/include/stdio.h
+++ b/src/include/stdio.h
@@ -5,11 +5,150 @@
 
 #include <nxbase.h> // NX_Printf
 
+#include <bits/types.h>
+#include <stdarg.h>
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
+#if __STDC_VERSION__ < 201112L
+#define __NEED_struct__IO_FILE
+#endif
+
+#ifdef __cplusplus
+#define NULL 0L
+#else
+#define NULL ((void*)0)
+#endif
+
+#undef EOF
+#define EOF (-1)
+
+#undef SEEK_SET
+#undef SEEK_CUR
+#undef SEEK_END
+#define SEEK_SET 0
+#define SEEK_CUR 1
+#define SEEK_END 2
+
+#define _IOFBF 0
+#define _IOLBF 1
+#define _IONBF 2
+
+#define BUFSIZ 1024
+#define FILENAME_MAX 4096
+#define FOPEN_MAX 1000
+#define TMP_MAX 10000
+#define L_tmpnam 20
+
+typedef loff_t fpos_t;
+
+typedef struct _IO_FILE FILE;
+struct _IO_FILE {
+	unsigned flags;
+	unsigned char *rpos, *rend;
+	int (*close)(FILE *);
+	unsigned char *wend, *wpos;
+	unsigned char *mustbezero_1;
+	unsigned char *wbase;
+	size_t (*read)(FILE *, unsigned char *, size_t);
+	size_t (*write)(FILE *, const unsigned char *, size_t);
+	off_t (*seek)(FILE *, off_t, int);
+	unsigned char *buf;
+	size_t buf_size;
+	FILE *prev, *next;
+	int fd;
+	int pipe_pid;
+	long lockcount;
+	int mode;
+	volatile int lock;
+	int lbf;
+	void *cookie;
+	off_t off;
+	char *getln_buf;
+	void *mustbezero_2;
+	unsigned char *shend;
+	off_t shlim, shcnt;
+	FILE *prev_locked, *next_locked;
+	struct __locale_struct *locale;
+};
+
+extern FILE *const stdin;
+extern FILE *const stdout;
+extern FILE *const stderr;
+
+#define stdin  (stdin)
+#define stdout (stdout)
+#define stderr (stderr)
+
+FILE *fopen(const char *__restrict, const char *__restrict);
+FILE *freopen(const char *__restrict, const char *__restrict, FILE *__restrict);
+int fclose(FILE *);
+
+int remove(const char *);
+int rename(const char *, const char *);
+
+int feof(FILE *);
+int ferror(FILE *);
+int fflush(FILE *);
+void clearerr(FILE *);
+
+int fseek(FILE *, long, int);
+long ftell(FILE *);
+void rewind(FILE *);
+
+int fgetpos(FILE *__restrict, fpos_t *__restrict);
+int fsetpos(FILE *, const fpos_t *);
+
+size_t fread(void *__restrict, size_t, size_t, FILE *__restrict);
+size_t fwrite(const void *__restrict, size_t, size_t, FILE *__restrict);
+
+int fgetc(FILE *);
+int getc(FILE *);
+int getchar(void);
+int ungetc(int, FILE *);
+
+int fputc(int, FILE *);
+int putc(int, FILE *);
+int putchar(int);
+
+char *fgets(char *__restrict, int, FILE *__restrict);
+#if __STDC_VERSION__ < 201112L
+char *gets(char *);
+#endif
+
+int fputs(const char *__restrict, FILE *__restrict);
+int puts(const char *);
+
+int printf(const char *__restrict, ...);
+int fprintf(FILE *__restrict, const char *__restrict, ...);
+int sprintf(char *__restrict, const char *__restrict, ...);
+int snprintf(char *__restrict, size_t, const char *__restrict, ...);
+
+int vprintf(const char *__restrict, va_list);
+int vfprintf(FILE *__restrict, const char *__restrict, va_list);
+int vsprintf(char *__restrict, const char *__restrict, va_list);
+int vsnprintf(char *__restrict, size_t, const char *__restrict, va_list);
+
+int scanf(const char *__restrict, ...);
+int fscanf(FILE *__restrict, const char *__restrict, ...);
+int sscanf(const char *__restrict, const char *__restrict, ...);
+int vscanf(const char *__restrict, va_list);
+int vfscanf(FILE *__restrict, const char *__restrict, va_list);
+int vsscanf(const char *__restrict, const char *__restrict, va_list);
+
+void perror(const char *);
+
+int setvbuf(FILE *__restrict, char *__restrict, int, size_t);
+void setbuf(FILE *__restrict, char *__restrict);
+
+char *tmpnam(char *);
+FILE *tmpfile(void);
+
+#if 0
 #define printf(...) NX_Printf(__VA_ARGS__)
+#endif
 
 #ifdef __cplusplus
 }
diff --git a/src/include/stdlib.h b/src/include/stdlib.h
index 4107a6e..e70f7e7 100644
--- a/src/include/stdlib.h
+++ b/src/include/stdlib.h
@@ -1,8 +1,11 @@
 #ifndef _STDLIB_H
 #define _STDLIB_H
 
+#include <nxbase.h>
 #include <features.h>
 
+#include <bits/types.h>
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -13,6 +16,34 @@ extern "C" {
 _Noreturn void exit (int);
 _Noreturn void abort(void);
 
+static inline void* malloc( size_t size )
+{
+    return (void *)NX_MemAlloc(size);
+}
+
+static inline void *realloc( void *ptr, size_t new_size )
+{
+    return (void *)NX_MemReAlloc(ptr, new_size);
+}
+
+static inline void* calloc( size_t num, size_t size )
+{
+    return (void *)NX_MemAlloc(num * size);
+}
+
+static inline void free( void* ptr )
+{
+    NX_MemFree(ptr);
+}
+
+int mblen (const char *, size_t);
+int mbtowc (wchar_t *__restrict, const char *__restrict, size_t);
+int wctomb (char *, wchar_t);
+size_t mbstowcs (wchar_t *__restrict, const char *__restrict, size_t);
+size_t wcstombs (char *__restrict, const wchar_t *__restrict, size_t);
+
+hidden char *__randname(char *);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/include/string.h b/src/include/string.h
index 83a273b..e6ad245 100644
--- a/src/include/string.h
+++ b/src/include/string.h
@@ -47,6 +47,10 @@ size_t strlen (const char *);
 
 char *strerror (int);
 
+/* None-ANSI */
+size_t strnlen (const char *, size_t);
+
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/include/uchar.h b/src/include/uchar.h
new file mode 100644
index 0000000..49ffa45
--- /dev/null
+++ b/src/include/uchar.h
@@ -0,0 +1,27 @@
+#ifndef _UCHAR_H
+#define _UCHAR_H
+
+#include <wchar.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if __cplusplus < 201103L
+typedef unsigned short char16_t;
+typedef unsigned char32_t;
+#endif
+
+#include <features.h>
+
+size_t c16rtomb(char *__restrict, char16_t, mbstate_t *__restrict);
+size_t mbrtoc16(char16_t *__restrict, const char *__restrict, size_t, mbstate_t *__restrict);
+
+size_t c32rtomb(char *__restrict, char32_t, mbstate_t *__restrict);
+size_t mbrtoc32(char32_t *__restrict, const char *__restrict, size_t, mbstate_t *__restrict);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/include/wchar.h b/src/include/wchar.h
index a982546..83390c6 100644
--- a/src/include/wchar.h
+++ b/src/include/wchar.h
@@ -1,10 +1,139 @@
 #ifndef	_WCHAR_H
 #define	_WCHAR_H
 
+#include <bits/types.h>
+#include <stdio.h>
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
+typedef int mbstate_t;
+
+#if L'\0'-1 > 0
+#define WCHAR_MAX (0xffffffffu+L'\0')
+#define WCHAR_MIN (0+L'\0')
+#else
+#define WCHAR_MAX (0x7fffffff+L'\0')
+#define WCHAR_MIN (-1-0x7fffffff+L'\0')
+#endif
+
+#ifdef __cplusplus
+#define NULL 0L
+#else
+#define NULL ((void*)0)
+#endif
+
+#undef WEOF
+#define WEOF 0xffffffffU
+
+wchar_t *wcscpy (wchar_t *__restrict, const wchar_t *__restrict);
+wchar_t *wcsncpy (wchar_t *__restrict, const wchar_t *__restrict, size_t);
+
+wchar_t *wcscat (wchar_t *__restrict, const wchar_t *__restrict);
+wchar_t *wcsncat (wchar_t *__restrict, const wchar_t *__restrict, size_t);
+
+int wcscmp (const wchar_t *, const wchar_t *);
+int wcsncmp (const wchar_t *, const wchar_t *, size_t);
+
+int wcscoll(const wchar_t *, const wchar_t *);
+size_t wcsxfrm (wchar_t *__restrict, const wchar_t *__restrict, size_t);
+
+wchar_t *wcschr (const wchar_t *, wchar_t);
+wchar_t *wcsrchr (const wchar_t *, wchar_t);
+
+size_t wcscspn (const wchar_t *, const wchar_t *);
+size_t wcsspn (const wchar_t *, const wchar_t *);
+wchar_t *wcspbrk (const wchar_t *, const wchar_t *);
+
+wchar_t *wcstok (wchar_t *__restrict, const wchar_t *__restrict, wchar_t **__restrict);
+
+size_t wcslen (const wchar_t *);
+
+wchar_t *wcsstr (const wchar_t *__restrict, const wchar_t *__restrict);
+wchar_t *wcswcs (const wchar_t *, const wchar_t *);
+
+wchar_t *wmemchr (const wchar_t *, wchar_t, size_t);
+int wmemcmp (const wchar_t *, const wchar_t *, size_t);
+wchar_t *wmemcpy (wchar_t *__restrict, const wchar_t *__restrict, size_t);
+wchar_t *wmemmove (wchar_t *, const wchar_t *, size_t);
+wchar_t *wmemset (wchar_t *, wchar_t, size_t);
+
+wint_t btowc (int);
+int wctob (wint_t);
+
+int mbsinit (const mbstate_t *);
+size_t mbrtowc (wchar_t *__restrict, const char *__restrict, size_t, mbstate_t *__restrict);
+size_t wcrtomb (char *__restrict, wchar_t, mbstate_t *__restrict);
+
+size_t mbrlen (const char *__restrict, size_t, mbstate_t *__restrict);
+
+size_t mbsrtowcs (wchar_t *__restrict, const char **__restrict, size_t, mbstate_t *__restrict);
+size_t wcsrtombs (char *__restrict, const wchar_t **__restrict, size_t, mbstate_t *__restrict);
+
+float wcstof (const wchar_t *__restrict, wchar_t **__restrict);
+double wcstod (const wchar_t *__restrict, wchar_t **__restrict);
+long double wcstold (const wchar_t *__restrict, wchar_t **__restrict);
+
+long wcstol (const wchar_t *__restrict, wchar_t **__restrict, int);
+unsigned long wcstoul (const wchar_t *__restrict, wchar_t **__restrict, int);
+
+long long wcstoll (const wchar_t *__restrict, wchar_t **__restrict, int);
+unsigned long long wcstoull (const wchar_t *__restrict, wchar_t **__restrict, int);
+
+int fwide (FILE *, int);
+
+int wprintf (const wchar_t *__restrict, ...);
+int fwprintf (FILE *__restrict, const wchar_t *__restrict, ...);
+int swprintf (wchar_t *__restrict, size_t, const wchar_t *__restrict, ...);
+
+int vwprintf (const wchar_t *__restrict, va_list);
+int vfwprintf (FILE *__restrict, const wchar_t *__restrict, va_list);
+int vswprintf (wchar_t *__restrict, size_t, const wchar_t *__restrict, va_list);
+
+int wscanf (const wchar_t *__restrict, ...);
+int fwscanf (FILE *__restrict, const wchar_t *__restrict, ...);
+int swscanf (const wchar_t *__restrict, const wchar_t *__restrict, ...);
+
+int vwscanf (const wchar_t *__restrict, va_list);
+int vfwscanf (FILE *__restrict, const wchar_t *__restrict, va_list);
+int vswscanf (const wchar_t *__restrict, const wchar_t *__restrict, va_list);
+
+wint_t fgetwc (FILE *);
+wint_t getwc (FILE *);
+wint_t getwchar (void);
+
+wint_t fputwc (wchar_t, FILE *);
+wint_t putwc (wchar_t, FILE *);
+wint_t putwchar (wchar_t);
+
+wchar_t *fgetws (wchar_t *__restrict, int, FILE *__restrict);
+int fputws (const wchar_t *__restrict, FILE *__restrict);
+
+wint_t ungetwc (wint_t, FILE *);
+
+int mbsinit (const mbstate_t *);
+size_t mbrtowc (wchar_t *__restrict, const char *__restrict, size_t, mbstate_t *__restrict);
+size_t wcrtomb (char *__restrict, wchar_t, mbstate_t *__restrict);
+
+size_t mbrlen (const char *__restrict, size_t, mbstate_t *__restrict);
+
+size_t mbsrtowcs (wchar_t *__restrict, const char **__restrict, size_t, mbstate_t *__restrict);
+size_t wcsrtombs (char *__restrict, const wchar_t **__restrict, size_t, mbstate_t *__restrict);
+
+float wcstof (const wchar_t *__restrict, wchar_t **__restrict);
+double wcstod (const wchar_t *__restrict, wchar_t **__restrict);
+long double wcstold (const wchar_t *__restrict, wchar_t **__restrict);
+
+long wcstol (const wchar_t *__restrict, wchar_t **__restrict, int);
+unsigned long wcstoul (const wchar_t *__restrict, wchar_t **__restrict, int);
+
+long long wcstoll (const wchar_t *__restrict, wchar_t **__restrict, int);
+unsigned long long wcstoull (const wchar_t *__restrict, wchar_t **__restrict, int);
+
+struct tm;
+size_t wcsftime (wchar_t *__restrict, size_t, const wchar_t *__restrict, const struct tm *__restrict);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/include/wctype.h b/src/include/wctype.h
new file mode 100644
index 0000000..2d78ee3
--- /dev/null
+++ b/src/include/wctype.h
@@ -0,0 +1,47 @@
+#ifndef _WCTYPE_H
+#define _WCTYPE_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <features.h>
+
+#include <bits/types.h>
+
+typedef const int * wctrans_t;
+
+#undef WEOF
+#define WEOF 0xffffffffU
+
+#undef iswdigit
+
+int       iswalnum(wint_t);
+int       iswalpha(wint_t);
+int       iswblank(wint_t);
+int       iswcntrl(wint_t);
+int       iswdigit(wint_t);
+int       iswgraph(wint_t);
+int       iswlower(wint_t);
+int       iswprint(wint_t);
+int       iswpunct(wint_t);
+int       iswspace(wint_t);
+int       iswupper(wint_t);
+int       iswxdigit(wint_t);
+int       iswctype(wint_t, wctype_t);
+wint_t    towctrans(wint_t, wctrans_t);
+wint_t    towlower(wint_t);
+wint_t    towupper(wint_t);
+wctrans_t wctrans(const char *);
+wctype_t  wctype(const char *);
+
+#ifndef __cplusplus
+#undef iswdigit
+#define iswdigit(a) (0 ? iswdigit(a) : ((unsigned)(a)-'0') < 10)
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/internal/floatscan.c b/src/internal/floatscan.c
new file mode 100644
index 0000000..8c0828f
--- /dev/null
+++ b/src/internal/floatscan.c
@@ -0,0 +1,507 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <math.h>
+#include <float.h>
+#include <limits.h>
+#include <errno.h>
+#include <ctype.h>
+
+#include "shgetc.h"
+#include "floatscan.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+
+#define LD_B1B_DIG 2
+#define LD_B1B_MAX 9007199, 254740991
+#define KMAX 128
+
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+
+#define LD_B1B_DIG 3
+#define LD_B1B_MAX 18, 446744073, 709551615
+#define KMAX 2048
+
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+
+#define LD_B1B_DIG 4
+#define LD_B1B_MAX 10384593, 717069655, 257060992, 658440191
+#define KMAX 2048
+
+#else
+#error Unsupported long double representation
+#endif
+
+#define MASK (KMAX-1)
+
+static long long scanexp(FILE *f, int pok)
+{
+	int c;
+	int x;
+	long long y;
+	int neg = 0;
+	
+	c = shgetc(f);
+	if (c=='+' || c=='-') {
+		neg = (c=='-');
+		c = shgetc(f);
+		if (c-'0'>=10U && pok) shunget(f);
+	}
+	if (c-'0'>=10U) {
+		shunget(f);
+		return LLONG_MIN;
+	}
+	for (x=0; c-'0'<10U && x<INT_MAX/10; c = shgetc(f))
+		x = 10*x + c-'0';
+	for (y=x; c-'0'<10U && y<LLONG_MAX/100; c = shgetc(f))
+		y = 10*y + c-'0';
+	for (; c-'0'<10U; c = shgetc(f));
+	shunget(f);
+	return neg ? -y : y;
+}
+
+
+static long double decfloat(FILE *f, int c, int bits, int emin, int sign, int pok)
+{
+	uint32_t x[KMAX];
+	static const uint32_t th[] = { LD_B1B_MAX };
+	int i, j, k, a, z;
+	long long lrp=0, dc=0;
+	long long e10=0;
+	int lnz = 0;
+	int gotdig = 0, gotrad = 0;
+	int rp;
+	int e2;
+	int emax = -emin-bits+3;
+	int denormal = 0;
+	long double y;
+	long double frac=0;
+	long double bias=0;
+	static const int p10s[] = { 10, 100, 1000, 10000,
+		100000, 1000000, 10000000, 100000000 };
+
+	j=0;
+	k=0;
+
+	/* Don't let leading zeros consume buffer space */
+	for (; c=='0'; c = shgetc(f)) gotdig=1;
+	if (c=='.') {
+		gotrad = 1;
+		for (c = shgetc(f); c=='0'; c = shgetc(f)) gotdig=1, lrp--;
+	}
+
+	x[0] = 0;
+	for (; c-'0'<10U || c=='.'; c = shgetc(f)) {
+		if (c == '.') {
+			if (gotrad) break;
+			gotrad = 1;
+			lrp = dc;
+		} else if (k < KMAX-3) {
+			dc++;
+			if (c!='0') lnz = dc;
+			if (j) x[k] = x[k]*10 + c-'0';
+			else x[k] = c-'0';
+			if (++j==9) {
+				k++;
+				j=0;
+			}
+			gotdig=1;
+		} else {
+			dc++;
+			if (c!='0') {
+				lnz = (KMAX-4)*9;
+				x[KMAX-4] |= 1;
+			}
+		}
+	}
+	if (!gotrad) lrp=dc;
+
+	if (gotdig && (c|32)=='e') {
+		e10 = scanexp(f, pok);
+		if (e10 == LLONG_MIN) {
+			if (pok) {
+				shunget(f);
+			} else {
+				shlim(f, 0);
+				return 0;
+			}
+			e10 = 0;
+		}
+		lrp += e10;
+	} else if (c>=0) {
+		shunget(f);
+	}
+	if (!gotdig) {
+		errno = EINVAL;
+		shlim(f, 0);
+		return 0;
+	}
+
+	/* Handle zero specially to avoid nasty special cases later */
+	if (!x[0]) return sign * 0.0;
+
+	/* Optimize small integers (w/no exponent) and over/under-flow */
+	if (lrp==dc && dc<10 && (bits>30 || x[0]>>bits==0))
+		return sign * (long double)x[0];
+	if (lrp > -emin/2) {
+		errno = ERANGE;
+		return sign * LDBL_MAX * LDBL_MAX;
+	}
+	if (lrp < emin-2*LDBL_MANT_DIG) {
+		errno = ERANGE;
+		return sign * LDBL_MIN * LDBL_MIN;
+	}
+
+	/* Align incomplete final B1B digit */
+	if (j) {
+		for (; j<9; j++) x[k]*=10;
+		k++;
+		j=0;
+	}
+
+	a = 0;
+	z = k;
+	e2 = 0;
+	rp = lrp;
+
+	/* Optimize small to mid-size integers (even in exp. notation) */
+	if (lnz<9 && lnz<=rp && rp < 18) {
+		if (rp == 9) return sign * (long double)x[0];
+		if (rp < 9) return sign * (long double)x[0] / p10s[8-rp];
+		int bitlim = bits-3*(int)(rp-9);
+		if (bitlim>30 || x[0]>>bitlim==0)
+			return sign * (long double)x[0] * p10s[rp-10];
+	}
+
+	/* Drop trailing zeros */
+	for (; !x[z-1]; z--);
+
+	/* Align radix point to B1B digit boundary */
+	if (rp % 9) {
+		int rpm9 = rp>=0 ? rp%9 : rp%9+9;
+		int p10 = p10s[8-rpm9];
+		uint32_t carry = 0;
+		for (k=a; k!=z; k++) {
+			uint32_t tmp = x[k] % p10;
+			x[k] = x[k]/p10 + carry;
+			carry = 1000000000/p10 * tmp;
+			if (k==a && !x[k]) {
+				a = (a+1 & MASK);
+				rp -= 9;
+			}
+		}
+		if (carry) x[z++] = carry;
+		rp += 9-rpm9;
+	}
+
+	/* Upscale until desired number of bits are left of radix point */
+	while (rp < 9*LD_B1B_DIG || (rp == 9*LD_B1B_DIG && x[a]<th[0])) {
+		uint32_t carry = 0;
+		e2 -= 29;
+		for (k=(z-1 & MASK); ; k=(k-1 & MASK)) {
+			uint64_t tmp = ((uint64_t)x[k] << 29) + carry;
+			if (tmp > 1000000000) {
+				carry = tmp / 1000000000;
+				x[k] = tmp % 1000000000;
+			} else {
+				carry = 0;
+				x[k] = tmp;
+			}
+			if (k==(z-1 & MASK) && k!=a && !x[k]) z = k;
+			if (k==a) break;
+		}
+		if (carry) {
+			rp += 9;
+			a = (a-1 & MASK);
+			if (a == z) {
+				z = (z-1 & MASK);
+				x[z-1 & MASK] |= x[z];
+			}
+			x[a] = carry;
+		}
+	}
+
+	/* Downscale until exactly number of bits are left of radix point */
+	for (;;) {
+		uint32_t carry = 0;
+		int sh = 1;
+		for (i=0; i<LD_B1B_DIG; i++) {
+			k = (a+i & MASK);
+			if (k == z || x[k] < th[i]) {
+				i=LD_B1B_DIG;
+				break;
+			}
+			if (x[a+i & MASK] > th[i]) break;
+		}
+		if (i==LD_B1B_DIG && rp==9*LD_B1B_DIG) break;
+		/* FIXME: find a way to compute optimal sh */
+		if (rp > 9+9*LD_B1B_DIG) sh = 9;
+		e2 += sh;
+		for (k=a; k!=z; k=(k+1 & MASK)) {
+			uint32_t tmp = x[k] & (1<<sh)-1;
+			x[k] = (x[k]>>sh) + carry;
+			carry = (1000000000>>sh) * tmp;
+			if (k==a && !x[k]) {
+				a = (a+1 & MASK);
+				i--;
+				rp -= 9;
+			}
+		}
+		if (carry) {
+			if ((z+1 & MASK) != a) {
+				x[z] = carry;
+				z = (z+1 & MASK);
+			} else x[z-1 & MASK] |= 1;
+		}
+	}
+
+	/* Assemble desired bits into floating point variable */
+	for (y=i=0; i<LD_B1B_DIG; i++) {
+		if ((a+i & MASK)==z) x[(z=(z+1 & MASK))-1] = 0;
+		y = 1000000000.0L * y + x[a+i & MASK];
+	}
+
+	y *= sign;
+
+	/* Limit precision for denormal results */
+	if (bits > LDBL_MANT_DIG+e2-emin) {
+		bits = LDBL_MANT_DIG+e2-emin;
+		if (bits<0) bits=0;
+		denormal = 1;
+	}
+
+	/* Calculate bias term to force rounding, move out lower bits */
+	if (bits < LDBL_MANT_DIG) {
+		bias = copysignl(scalbn(1, 2*LDBL_MANT_DIG-bits-1), y);
+		frac = fmodl(y, scalbn(1, LDBL_MANT_DIG-bits));
+		y -= frac;
+		y += bias;
+	}
+
+	/* Process tail of decimal input so it can affect rounding */
+	if ((a+i & MASK) != z) {
+		uint32_t t = x[a+i & MASK];
+		if (t < 500000000 && (t || (a+i+1 & MASK) != z))
+			frac += 0.25*sign;
+		else if (t > 500000000)
+			frac += 0.75*sign;
+		else if (t == 500000000) {
+			if ((a+i+1 & MASK) == z)
+				frac += 0.5*sign;
+			else
+				frac += 0.75*sign;
+		}
+		if (LDBL_MANT_DIG-bits >= 2 && !fmodl(frac, 1))
+			frac++;
+	}
+
+	y += frac;
+	y -= bias;
+
+	if ((e2+LDBL_MANT_DIG & INT_MAX) > emax-5) {
+		if (fabsl(y) >= 2/LDBL_EPSILON) {
+			if (denormal && bits==LDBL_MANT_DIG+e2-emin)
+				denormal = 0;
+			y *= 0.5;
+			e2++;
+		}
+		if (e2+LDBL_MANT_DIG>emax || (denormal && frac))
+			errno = ERANGE;
+	}
+
+	return scalbnl(y, e2);
+}
+
+static long double hexfloat(FILE *f, int bits, int emin, int sign, int pok)
+{
+	uint32_t x = 0;
+	long double y = 0;
+	long double scale = 1;
+	long double bias = 0;
+	int gottail = 0, gotrad = 0, gotdig = 0;
+	long long rp = 0;
+	long long dc = 0;
+	long long e2 = 0;
+	int d;
+	int c;
+
+	c = shgetc(f);
+
+	/* Skip leading zeros */
+	for (; c=='0'; c = shgetc(f)) gotdig = 1;
+
+	if (c=='.') {
+		gotrad = 1;
+		c = shgetc(f);
+		/* Count zeros after the radix point before significand */
+		for (rp=0; c=='0'; c = shgetc(f), rp--) gotdig = 1;
+	}
+
+	for (; c-'0'<10U || (c|32)-'a'<6U || c=='.'; c = shgetc(f)) {
+		if (c=='.') {
+			if (gotrad) break;
+			rp = dc;
+			gotrad = 1;
+		} else {
+			gotdig = 1;
+			if (c > '9') d = (c|32)+10-'a';
+			else d = c-'0';
+			if (dc<8) {
+				x = x*16 + d;
+			} else if (dc < LDBL_MANT_DIG/4+1) {
+				y += d*(scale/=16);
+			} else if (d && !gottail) {
+				y += 0.5*scale;
+				gottail = 1;
+			}
+			dc++;
+		}
+	}
+	if (!gotdig) {
+		shunget(f);
+		if (pok) {
+			shunget(f);
+			if (gotrad) shunget(f);
+		} else {
+			shlim(f, 0);
+		}
+		return sign * 0.0;
+	}
+	if (!gotrad) rp = dc;
+	while (dc<8) x *= 16, dc++;
+	if ((c|32)=='p') {
+		e2 = scanexp(f, pok);
+		if (e2 == LLONG_MIN) {
+			if (pok) {
+				shunget(f);
+			} else {
+				shlim(f, 0);
+				return 0;
+			}
+			e2 = 0;
+		}
+	} else {
+		shunget(f);
+	}
+	e2 += 4*rp - 32;
+
+	if (!x) return sign * 0.0;
+	if (e2 > -emin) {
+		errno = ERANGE;
+		return sign * LDBL_MAX * LDBL_MAX;
+	}
+	if (e2 < emin-2*LDBL_MANT_DIG) {
+		errno = ERANGE;
+		return sign * LDBL_MIN * LDBL_MIN;
+	}
+
+	while (x < 0x80000000) {
+		if (y>=0.5) {
+			x += x + 1;
+			y += y - 1;
+		} else {
+			x += x;
+			y += y;
+		}
+		e2--;
+	}
+
+	if (bits > 32+e2-emin) {
+		bits = 32+e2-emin;
+		if (bits<0) bits=0;
+	}
+
+	if (bits < LDBL_MANT_DIG)
+		bias = copysignl(scalbn(1, 32+LDBL_MANT_DIG-bits-1), sign);
+
+	if (bits<32 && y && !(x&1)) x++, y=0;
+
+	y = bias + sign*(long double)x + sign*y;
+	y -= bias;
+
+	if (!y) errno = ERANGE;
+
+	return scalbnl(y, e2);
+}
+
+long double __floatscan(FILE *f, int prec, int pok)
+{
+	int sign = 1;
+	size_t i;
+	int bits;
+	int emin;
+	int c;
+
+	switch (prec) {
+	case 0:
+		bits = FLT_MANT_DIG;
+		emin = FLT_MIN_EXP-bits;
+		break;
+	case 1:
+		bits = DBL_MANT_DIG;
+		emin = DBL_MIN_EXP-bits;
+		break;
+	case 2:
+		bits = LDBL_MANT_DIG;
+		emin = LDBL_MIN_EXP-bits;
+		break;
+	default:
+		return 0;
+	}
+
+	while (isspace((c=shgetc(f))));
+
+	if (c=='+' || c=='-') {
+		sign -= 2*(c=='-');
+		c = shgetc(f);
+	}
+
+	for (i=0; i<8 && (c|32)=="infinity"[i]; i++)
+		if (i<7) c = shgetc(f);
+	if (i==3 || i==8 || (i>3 && pok)) {
+		if (i!=8) {
+			shunget(f);
+			if (pok) for (; i>3; i--) shunget(f);
+		}
+		return sign * INFINITY;
+	}
+	if (!i) for (i=0; i<3 && (c|32)=="nan"[i]; i++)
+		if (i<2) c = shgetc(f);
+	if (i==3) {
+		if (shgetc(f) != '(') {
+			shunget(f);
+			return NAN;
+		}
+		for (i=1; ; i++) {
+			c = shgetc(f);
+			if (c-'0'<10U || c-'A'<26U || c-'a'<26U || c=='_')
+				continue;
+			if (c==')') return NAN;
+			shunget(f);
+			if (!pok) {
+				errno = EINVAL;
+				shlim(f, 0);
+				return 0;
+			}
+			while (i--) shunget(f);
+			return NAN;
+		}
+		return NAN;
+	}
+
+	if (i) {
+		shunget(f);
+		errno = EINVAL;
+		shlim(f, 0);
+		return 0;
+	}
+
+	if (c=='0') {
+		c = shgetc(f);
+		if ((c|32) == 'x')
+			return hexfloat(f, bits, emin, sign, pok);
+		shunget(f);
+		c = '0';
+	}
+
+	return decfloat(f, c, bits, emin, sign, pok);
+}
diff --git a/src/internal/floatscan.h b/src/internal/floatscan.h
new file mode 100644
index 0000000..f2b1dcf
--- /dev/null
+++ b/src/internal/floatscan.h
@@ -0,0 +1,8 @@
+#ifndef FLOATSCAN_H
+#define FLOATSCAN_H
+
+#include <stdio.h>
+
+hidden long double __floatscan(FILE *, int, int);
+
+#endif
diff --git a/src/internal/intscan.c b/src/internal/intscan.c
new file mode 100644
index 0000000..a4a5ae8
--- /dev/null
+++ b/src/internal/intscan.c
@@ -0,0 +1,100 @@
+#include <limits.h>
+#include <errno.h>
+#include <ctype.h>
+#include "shgetc.h"
+
+/* Lookup table for digit values. -1==255>=36 -> invalid */
+static const unsigned char table[] = { -1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,-1,-1,-1,-1,-1,-1,
+-1,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,
+25,26,27,28,29,30,31,32,33,34,35,-1,-1,-1,-1,-1,
+-1,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,
+25,26,27,28,29,30,31,32,33,34,35,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
+};
+
+unsigned long long __intscan(FILE *f, unsigned base, int pok, unsigned long long lim)
+{
+	const unsigned char *val = table+1;
+	int c, neg=0;
+	unsigned x;
+	unsigned long long y;
+	if (base > 36 || base == 1) {
+		errno = EINVAL;
+		return 0;
+	}
+	while (isspace((c=shgetc(f))));
+	if (c=='+' || c=='-') {
+		neg = -(c=='-');
+		c = shgetc(f);
+	}
+	if ((base == 0 || base == 16) && c=='0') {
+		c = shgetc(f);
+		if ((c|32)=='x') {
+			c = shgetc(f);
+			if (val[c]>=16) {
+				shunget(f);
+				if (pok) shunget(f);
+				else shlim(f, 0);
+				return 0;
+			}
+			base = 16;
+		} else if (base == 0) {
+			base = 8;
+		}
+	} else {
+		if (base == 0) base = 10;
+		if (val[c] >= base) {
+			shunget(f);
+			shlim(f, 0);
+			errno = EINVAL;
+			return 0;
+		}
+	}
+	if (base == 10) {
+		for (x=0; c-'0'<10U && x<=UINT_MAX/10-1; c=shgetc(f))
+			x = x*10 + (c-'0');
+		for (y=x; c-'0'<10U && y<=ULLONG_MAX/10 && 10*y<=ULLONG_MAX-(c-'0'); c=shgetc(f))
+			y = y*10 + (c-'0');
+		if (c-'0'>=10U) goto done;
+	} else if (!(base & base-1)) {
+		int bs = "\0\1\2\4\7\3\6\5"[(0x17*base)>>5&7];
+		for (x=0; val[c]<base && x<=UINT_MAX/32; c=shgetc(f))
+			x = x<<bs | val[c];
+		for (y=x; val[c]<base && y<=ULLONG_MAX>>bs; c=shgetc(f))
+			y = y<<bs | val[c];
+	} else {
+		for (x=0; val[c]<base && x<=UINT_MAX/36-1; c=shgetc(f))
+			x = x*base + val[c];
+		for (y=x; val[c]<base && y<=ULLONG_MAX/base && base*y<=ULLONG_MAX-val[c]; c=shgetc(f))
+			y = y*base + val[c];
+	}
+	if (val[c]<base) {
+		for (; val[c]<base; c=shgetc(f));
+		errno = ERANGE;
+		y = lim;
+		if (lim&1) neg = 0;
+	}
+done:
+	shunget(f);
+	if (y>=lim) {
+		if (!(lim&1) && !neg) {
+			errno = ERANGE;
+			return lim-1;
+		} else if (y>lim) {
+			errno = ERANGE;
+			return lim;
+		}
+	}
+	return (y^neg)-neg;
+}
diff --git a/src/internal/intscan.h b/src/internal/intscan.h
new file mode 100644
index 0000000..ccf9f11
--- /dev/null
+++ b/src/internal/intscan.h
@@ -0,0 +1,8 @@
+#ifndef INTSCAN_H
+#define INTSCAN_H
+
+#include <stdio.h>
+
+hidden unsigned long long __intscan(FILE *, unsigned, int, unsigned long long);
+
+#endif
diff --git a/src/internal/libc.c b/src/internal/libc.c
new file mode 100644
index 0000000..cb05181
--- /dev/null
+++ b/src/internal/libc.c
@@ -0,0 +1,9 @@
+#include "libc.h"
+
+struct __libc __libc;
+
+size_t __hwcap;
+char *__progname=0, *__progname_full=0;
+
+weak_alias(__progname, program_invocation_short_name);
+weak_alias(__progname_full, program_invocation_name);
diff --git a/src/internal/libc.h b/src/internal/libc.h
new file mode 100644
index 0000000..c43ef96
--- /dev/null
+++ b/src/internal/libc.h
@@ -0,0 +1,30 @@
+#ifndef LIBC_H
+#define LIBC_H
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <limits.h>
+
+struct __locale_map;
+
+struct __locale_struct {
+	const struct __locale_map *cat[6];
+};
+
+struct __libc {
+	size_t page_size;
+	struct __locale_struct global_locale;
+};
+
+#ifndef PAGE_SIZE
+#define PAGE_SIZE libc.page_size
+#endif
+
+extern hidden struct __libc __libc;
+#define libc __libc
+
+hidden void __init_libc(char **, char *);
+
+extern hidden const char __libc_version[];
+
+#endif
diff --git a/src/internal/shgetc.c b/src/internal/shgetc.c
new file mode 100644
index 0000000..7455d2f
--- /dev/null
+++ b/src/internal/shgetc.c
@@ -0,0 +1,37 @@
+#include "shgetc.h"
+
+/* The shcnt field stores the number of bytes read so far, offset by
+ * the value of buf-rpos at the last function call (__shlim or __shgetc),
+ * so that between calls the inline shcnt macro can add rpos-buf to get
+ * the actual count. */
+
+void __shlim(FILE *f, off_t lim)
+{
+	f->shlim = lim;
+	f->shcnt = f->buf - f->rpos;
+	/* If lim is nonzero, rend must be a valid pointer. */
+	if (lim && f->rend - f->rpos > lim)
+		f->shend = f->rpos + lim;
+	else
+		f->shend = f->rend;
+}
+
+int __shgetc(FILE *f)
+{
+	int c;
+	off_t cnt = shcnt(f);
+	if (f->shlim && cnt >= f->shlim || (c=__uflow(f)) < 0) {
+		f->shcnt = f->buf - f->rpos + cnt;
+		f->shend = f->rpos;
+		f->shlim = -1;
+		return EOF;
+	}
+	cnt++;
+	if (f->shlim && f->rend - f->rpos > f->shlim - cnt)
+		f->shend = f->rpos + (f->shlim - cnt);
+	else
+		f->shend = f->rend;
+	f->shcnt = f->buf - f->rpos + cnt;
+	if (f->rpos <= f->buf) f->rpos[-1] = c;
+	return c;
+}
diff --git a/src/internal/shgetc.h b/src/internal/shgetc.h
new file mode 100644
index 0000000..9435381
--- /dev/null
+++ b/src/internal/shgetc.h
@@ -0,0 +1,32 @@
+#include "stdio_impl.h"
+
+/* Scan helper "stdio" functions for use by scanf-family and strto*-family
+ * functions. These accept either a valid stdio FILE, or a minimal pseudo
+ * FILE whose buffer pointers point into a null-terminated string. In the
+ * latter case, the sh_fromstring macro should be used to setup the FILE;
+ * the rest of the structure can be left uninitialized.
+ *
+ * To begin using these functions, shlim must first be called on the FILE
+ * to set a field width limit, or 0 for no limit. For string pseudo-FILEs,
+ * a nonzero limit is not valid and produces undefined behavior. After that,
+ * shgetc, shunget, and shcnt are valid as long as no other stdio functions
+ * are called on the stream.
+ *
+ * When used with a real FILE object, shunget has only one byte of pushback
+ * available. Further shunget (up to a limit of the stdio UNGET buffer size)
+ * will adjust the position but will not restore the data to be read again.
+ * This functionality is needed for the wcsto*-family functions, where it's
+ * okay because the FILE will be discarded immediately anyway. When used
+ * with string pseudo-FILEs, shunget has unlimited pushback, back to the
+ * beginning of the string. */
+
+hidden void __shlim(FILE *, off_t);
+hidden int __shgetc(FILE *);
+
+#define shcnt(f) ((f)->shcnt + ((f)->rpos - (f)->buf))
+#define shlim(f, lim) __shlim((f), (lim))
+#define shgetc(f) (((f)->rpos != (f)->shend) ? *(f)->rpos++ : __shgetc(f))
+#define shunget(f) ((f)->shlim>=0 ? (void)(f)->rpos-- : (void)0)
+
+#define sh_fromstring(f, s) \
+	((f)->buf = (f)->rpos = (void *)(s), (f)->rend = (void*)-1)
diff --git a/src/internal/stdio_impl.h b/src/internal/stdio_impl.h
new file mode 100644
index 0000000..28e8114
--- /dev/null
+++ b/src/internal/stdio_impl.h
@@ -0,0 +1,84 @@
+#ifndef _STDIO_IMPL_H
+#define _STDIO_IMPL_H
+
+#include <stdio.h>
+
+#define UNGET 8
+
+#define FFINALLOCK(f) ((f)->lock>=0 ? __lockfile((f)) : 0)
+#define FLOCK(f) int __need_unlock = ((f)->lock>=0 ? __lockfile((f)) : 0)
+#define FUNLOCK(f) do { if (__need_unlock) __unlockfile((f)); } while (0)
+
+#define F_PERM 1
+#define F_NORD 4
+#define F_NOWR 8
+#define F_EOF 16
+#define F_ERR 32
+#define F_SVB 64
+#define F_APP 128
+
+extern hidden FILE *volatile __stdin_used;
+extern hidden FILE *volatile __stdout_used;
+extern hidden FILE *volatile __stderr_used;
+
+hidden int __lockfile(FILE *);
+hidden void __unlockfile(FILE *);
+
+hidden size_t __stdio_read(FILE *, unsigned char *, size_t);
+hidden size_t __stdio_write(FILE *, const unsigned char *, size_t);
+hidden size_t __stdout_write(FILE *, const unsigned char *, size_t);
+hidden off_t __stdio_seek(FILE *, off_t, int);
+hidden int __stdio_close(FILE *);
+
+hidden int __toread(FILE *);
+hidden int __towrite(FILE *);
+
+hidden void __stdio_exit(void);
+hidden void __stdio_exit_needed(void);
+
+#if defined(__PIC__) && (100*__GNUC__+__GNUC_MINOR__ >= 303)
+__attribute__((visibility("protected")))
+#endif
+int __overflow(FILE *, int), __uflow(FILE *);
+
+hidden int __fseeko(FILE *, off_t, int);
+hidden int __fseeko_unlocked(FILE *, off_t, int);
+hidden off_t __ftello(FILE *);
+hidden off_t __ftello_unlocked(FILE *);
+hidden size_t __fwritex(const unsigned char *, size_t, FILE *);
+hidden int __putc_unlocked(int, FILE *);
+
+hidden FILE *__fdopen(int, const char *);
+hidden int __fmodeflags(const char *);
+
+hidden FILE *__ofl_add(FILE *f);
+hidden FILE **__ofl_lock(void);
+hidden void __ofl_unlock(void);
+
+struct __pthread;
+hidden void __register_locked_file(FILE *, struct __pthread *);
+hidden void __unlist_locked_file(FILE *);
+hidden void __do_orphaned_stdio_locks(void);
+
+#define MAYBE_WAITERS 0x40000000
+
+hidden void __getopt_msg(const char *, const char *, const char *, size_t);
+
+#define feof(f) ((f)->flags & F_EOF)
+#define ferror(f) ((f)->flags & F_ERR)
+
+#define getc_unlocked(f) \
+	( ((f)->rpos != (f)->rend) ? *(f)->rpos++ : __uflow((f)) )
+
+#define putc_unlocked(c, f) \
+	( (((unsigned char)(c)!=(f)->lbf && (f)->wpos!=(f)->wend)) \
+	? *(f)->wpos++ = (unsigned char)(c) \
+	: __overflow((f),(unsigned char)(c)) )
+
+/* Caller-allocated FILE * operations */
+hidden FILE *__fopen_rb_ca(const char *, FILE *, unsigned char *, size_t);
+hidden int __fclose_ca(FILE *);
+
+void __init_stdio(void);
+
+#endif
diff --git a/src/internal/version.c b/src/internal/version.c
new file mode 100644
index 0000000..08bbf5b
--- /dev/null
+++ b/src/internal/version.c
@@ -0,0 +1,4 @@
+#include "version.h"
+#include "libc.h"
+
+const char __libc_version[] = VERSION;
diff --git a/src/internal/version.h b/src/internal/version.h
new file mode 100644
index 0000000..e3639f1
--- /dev/null
+++ b/src/internal/version.h
@@ -0,0 +1,6 @@
+#ifndef _VERSION_H
+#define _VERSION_H
+
+#define VERSION "1.2.2"
+
+#endif
diff --git a/src/locale/localeconv.c b/src/locale/localeconv.c
new file mode 100644
index 0000000..4cbb9dc
--- /dev/null
+++ b/src/locale/localeconv.c
@@ -0,0 +1,34 @@
+#include <locale.h>
+#include <limits.h>
+
+static const struct lconv posix_lconv = {
+	.decimal_point = ".",
+	.thousands_sep = "",
+	.grouping = "",
+	.int_curr_symbol = "",
+	.currency_symbol = "",
+	.mon_decimal_point = "",
+	.mon_thousands_sep = "",
+	.mon_grouping = "",
+	.positive_sign = "",
+	.negative_sign = "",
+	.int_frac_digits = CHAR_MAX,
+	.frac_digits = CHAR_MAX,
+	.p_cs_precedes = CHAR_MAX,
+	.p_sep_by_space = CHAR_MAX,
+	.n_cs_precedes = CHAR_MAX,
+	.n_sep_by_space = CHAR_MAX,
+	.p_sign_posn = CHAR_MAX,
+	.n_sign_posn = CHAR_MAX,
+	.int_p_cs_precedes = CHAR_MAX,
+	.int_p_sep_by_space = CHAR_MAX,
+	.int_n_cs_precedes = CHAR_MAX,
+	.int_n_sep_by_space = CHAR_MAX,
+	.int_p_sign_posn = CHAR_MAX,
+	.int_n_sign_posn = CHAR_MAX,
+};
+
+struct lconv *localeconv(void)
+{
+	return (void *)&posix_lconv;
+}
diff --git a/src/locale/setlocale.c b/src/locale/setlocale.c
new file mode 100644
index 0000000..cb31b01
--- /dev/null
+++ b/src/locale/setlocale.c
@@ -0,0 +1,8 @@
+#include <locale.h>
+#include <stdlib.h>
+#include <string.h>
+
+char *setlocale(int cat, const char *name)
+{
+	return "C.UTF-8";
+}
diff --git a/src/math/__signbit.c b/src/math/__signbit.c
new file mode 100644
index 0000000..e700b6b
--- /dev/null
+++ b/src/math/__signbit.c
@@ -0,0 +1,13 @@
+#include "libm.h"
+
+// FIXME: macro in math.h
+int __signbit(double x)
+{
+	union {
+		double d;
+		uint64_t i;
+	} y = { x };
+	return y.i>>63;
+}
+
+
diff --git a/src/math/__signbitf.c b/src/math/__signbitf.c
new file mode 100644
index 0000000..40ad3cf
--- /dev/null
+++ b/src/math/__signbitf.c
@@ -0,0 +1,11 @@
+#include "libm.h"
+
+// FIXME: macro in math.h
+int __signbitf(float x)
+{
+	union {
+		float f;
+		uint32_t i;
+	} y = { x };
+	return y.i>>31;
+}
diff --git a/src/math/__signbitl.c b/src/math/__signbitl.c
new file mode 100644
index 0000000..63b3dc5
--- /dev/null
+++ b/src/math/__signbitl.c
@@ -0,0 +1,14 @@
+#include "libm.h"
+
+#if (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+int __signbitl(long double x)
+{
+	union ldshape u = {x};
+	return u.i.se >> 15;
+}
+#elif LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+int __signbitl(long double x)
+{
+	return __signbit(x);
+}
+#endif
diff --git a/src/math/copysignl.c b/src/math/copysignl.c
new file mode 100644
index 0000000..9dd933c
--- /dev/null
+++ b/src/math/copysignl.c
@@ -0,0 +1,16 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double copysignl(long double x, long double y)
+{
+	return copysign(x, y);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double copysignl(long double x, long double y)
+{
+	union ldshape ux = {x}, uy = {y};
+	ux.i.se &= 0x7fff;
+	ux.i.se |= uy.i.se & 0x8000;
+	return ux.f;
+}
+#endif
diff --git a/src/math/fabsl.c b/src/math/fabsl.c
new file mode 100644
index 0000000..c4f36ec
--- /dev/null
+++ b/src/math/fabsl.c
@@ -0,0 +1,15 @@
+#include "libm.h"
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double fabsl(long double x)
+{
+	return fabs(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double fabsl(long double x)
+{
+	union ldshape u = {x};
+
+	u.i.se &= 0x7fff;
+	return u.f;
+}
+#endif
diff --git a/src/math/fdiml.c b/src/math/fdiml.c
new file mode 100644
index 0000000..62e29b7
--- /dev/null
+++ b/src/math/fdiml.c
@@ -0,0 +1,18 @@
+#include <math.h>
+#include <float.h>
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double fdiml(long double x, long double y)
+{
+	return fdim(x, y);
+}
+#else
+long double fdiml(long double x, long double y)
+{
+	if (isnan(x))
+		return x;
+	if (isnan(y))
+		return y;
+	return x > y ? x - y : 0;
+}
+#endif
diff --git a/src/math/floorl.c b/src/math/floorl.c
new file mode 100644
index 0000000..16aaec4
--- /dev/null
+++ b/src/math/floorl.c
@@ -0,0 +1,34 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double floorl(long double x)
+{
+	return floor(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+
+static const long double toint = 1/LDBL_EPSILON;
+
+long double floorl(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	long double y;
+
+	if (e >= 0x3fff+LDBL_MANT_DIG-1 || x == 0)
+		return x;
+	/* y = int(x) - x, where int(x) is an integer neighbor of x */
+	if (u.i.se >> 15)
+		y = x - toint + toint - x;
+	else
+		y = x + toint - toint - x;
+	/* special case because of non-nearest rounding modes */
+	if (e <= 0x3fff-1) {
+		FORCE_EVAL(y);
+		return u.i.se >> 15 ? -1 : 0;
+	}
+	if (y > 0)
+		return x + y - 1;
+	return x + y;
+}
+#endif
diff --git a/src/math/fmaxl.c b/src/math/fmaxl.c
new file mode 100644
index 0000000..4b03158
--- /dev/null
+++ b/src/math/fmaxl.c
@@ -0,0 +1,21 @@
+#include <math.h>
+#include <float.h>
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double fmaxl(long double x, long double y)
+{
+	return fmax(x, y);
+}
+#else
+long double fmaxl(long double x, long double y)
+{
+	if (isnan(x))
+		return y;
+	if (isnan(y))
+		return x;
+	/* handle signed zeros, see C99 Annex F.9.9.2 */
+	if (signbit(x) != signbit(y))
+		return signbit(x) ? y : x;
+	return x < y ? y : x;
+}
+#endif
diff --git a/src/math/fminl.c b/src/math/fminl.c
new file mode 100644
index 0000000..69bc24a
--- /dev/null
+++ b/src/math/fminl.c
@@ -0,0 +1,21 @@
+#include <math.h>
+#include <float.h>
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double fminl(long double x, long double y)
+{
+	return fmin(x, y);
+}
+#else
+long double fminl(long double x, long double y)
+{
+	if (isnan(x))
+		return y;
+	if (isnan(y))
+		return x;
+	/* handle signed zeros, see C99 Annex F.9.9.2 */
+	if (signbit(x) != signbit(y))
+		return signbit(x) ? x : y;
+	return x < y ? x : y;
+}
+#endif
diff --git a/src/math/fmodl.c b/src/math/fmodl.c
new file mode 100644
index 0000000..9f5b873
--- /dev/null
+++ b/src/math/fmodl.c
@@ -0,0 +1,105 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double fmodl(long double x, long double y)
+{
+	return fmod(x, y);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double fmodl(long double x, long double y)
+{
+	union ldshape ux = {x}, uy = {y};
+	int ex = ux.i.se & 0x7fff;
+	int ey = uy.i.se & 0x7fff;
+	int sx = ux.i.se & 0x8000;
+
+	if (y == 0 || isnan(y) || ex == 0x7fff)
+		return (x*y)/(x*y);
+	ux.i.se = ex;
+	uy.i.se = ey;
+	if (ux.f <= uy.f) {
+		if (ux.f == uy.f)
+			return 0*x;
+		return x;
+	}
+
+	/* normalize x and y */
+	if (!ex) {
+		ux.f *= 0x1p120f;
+		ex = ux.i.se - 120;
+	}
+	if (!ey) {
+		uy.f *= 0x1p120f;
+		ey = uy.i.se - 120;
+	}
+
+	/* x mod y */
+#if LDBL_MANT_DIG == 64
+	uint64_t i, mx, my;
+	mx = ux.i.m;
+	my = uy.i.m;
+	for (; ex > ey; ex--) {
+		i = mx - my;
+		if (mx >= my) {
+			if (i == 0)
+				return 0*x;
+			mx = 2*i;
+		} else if (2*mx < mx) {
+			mx = 2*mx - my;
+		} else {
+			mx = 2*mx;
+		}
+	}
+	i = mx - my;
+	if (mx >= my) {
+		if (i == 0)
+			return 0*x;
+		mx = i;
+	}
+	for (; mx >> 63 == 0; mx *= 2, ex--);
+	ux.i.m = mx;
+#elif LDBL_MANT_DIG == 113
+	uint64_t hi, lo, xhi, xlo, yhi, ylo;
+	xhi = (ux.i2.hi & -1ULL>>16) | 1ULL<<48;
+	yhi = (uy.i2.hi & -1ULL>>16) | 1ULL<<48;
+	xlo = ux.i2.lo;
+	ylo = uy.i2.lo;
+	for (; ex > ey; ex--) {
+		hi = xhi - yhi;
+		lo = xlo - ylo;
+		if (xlo < ylo)
+			hi -= 1;
+		if (hi >> 63 == 0) {
+			if ((hi|lo) == 0)
+				return 0*x;
+			xhi = 2*hi + (lo>>63);
+			xlo = 2*lo;
+		} else {
+			xhi = 2*xhi + (xlo>>63);
+			xlo = 2*xlo;
+		}
+	}
+	hi = xhi - yhi;
+	lo = xlo - ylo;
+	if (xlo < ylo)
+		hi -= 1;
+	if (hi >> 63 == 0) {
+		if ((hi|lo) == 0)
+			return 0*x;
+		xhi = hi;
+		xlo = lo;
+	}
+	for (; xhi >> 48 == 0; xhi = 2*xhi + (xlo>>63), xlo = 2*xlo, ex--);
+	ux.i2.hi = xhi;
+	ux.i2.lo = xlo;
+#endif
+
+	/* scale result */
+	if (ex <= 0) {
+		ux.i.se = (ex+120)|sx;
+		ux.f *= 0x1p-120f;
+	} else
+		ux.i.se = ex|sx;
+	return ux.f;
+}
+#endif
diff --git a/src/math/frexpl.c b/src/math/frexpl.c
new file mode 100644
index 0000000..3c1b553
--- /dev/null
+++ b/src/math/frexpl.c
@@ -0,0 +1,29 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double frexpl(long double x, int *e)
+{
+	return frexp(x, e);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double frexpl(long double x, int *e)
+{
+	union ldshape u = {x};
+	int ee = u.i.se & 0x7fff;
+
+	if (!ee) {
+		if (x) {
+			x = frexpl(x*0x1p120, e);
+			*e -= 120;
+		} else *e = 0;
+		return x;
+	} else if (ee == 0x7fff) {
+		return x;
+	}
+
+	*e = ee - 0x3ffe;
+	u.i.se &= 0x8000;
+	u.i.se |= 0x3ffe;
+	return u.f;
+}
+#endif
diff --git a/src/math/scalblnl.c b/src/math/scalblnl.c
new file mode 100644
index 0000000..854c51c
--- /dev/null
+++ b/src/math/scalblnl.c
@@ -0,0 +1,19 @@
+#include <limits.h>
+#include <math.h>
+#include <float.h>
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double scalblnl(long double x, long n)
+{
+	return scalbln(x, n);
+}
+#else
+long double scalblnl(long double x, long n)
+{
+	if (n > INT_MAX)
+		n = INT_MAX;
+	else if (n < INT_MIN)
+		n = INT_MIN;
+	return scalbnl(x, n);
+}
+#endif
diff --git a/src/math/scalbnl.c b/src/math/scalbnl.c
new file mode 100644
index 0000000..db44dab
--- /dev/null
+++ b/src/math/scalbnl.c
@@ -0,0 +1,36 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double scalbnl(long double x, int n)
+{
+	return scalbn(x, n);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double scalbnl(long double x, int n)
+{
+	union ldshape u;
+
+	if (n > 16383) {
+		x *= 0x1p16383L;
+		n -= 16383;
+		if (n > 16383) {
+			x *= 0x1p16383L;
+			n -= 16383;
+			if (n > 16383)
+				n = 16383;
+		}
+	} else if (n < -16382) {
+		x *= 0x1p-16382L * 0x1p113L;
+		n += 16382 - 113;
+		if (n < -16382) {
+			x *= 0x1p-16382L * 0x1p113L;
+			n += 16382 - 113;
+			if (n < -16382)
+				n = -16382;
+		}
+	}
+	u.f = 1.0;
+	u.i.se = 0x3fff + n;
+	return x * u.f;
+}
+#endif
diff --git a/src/multibyte/wcrtomb.c b/src/multibyte/wcrtomb.c
new file mode 100644
index 0000000..8e34926
--- /dev/null
+++ b/src/multibyte/wcrtomb.c
@@ -0,0 +1,37 @@
+#include <stdlib.h>
+#include <wchar.h>
+#include <errno.h>
+#include "internal.h"
+
+size_t wcrtomb(char *restrict s, wchar_t wc, mbstate_t *restrict st)
+{
+	if (!s) return 1;
+	if ((unsigned)wc < 0x80) {
+		*s = wc;
+		return 1;
+	} else if (MB_CUR_MAX == 1) {
+		if (!IS_CODEUNIT(wc)) {
+			errno = EILSEQ;
+			return -1;
+		}
+		*s = wc;
+		return 1;
+	} else if ((unsigned)wc < 0x800) {
+		*s++ = 0xc0 | (wc>>6);
+		*s = 0x80 | (wc&0x3f);
+		return 2;
+	} else if ((unsigned)wc < 0xd800 || (unsigned)wc-0xe000 < 0x2000) {
+		*s++ = 0xe0 | (wc>>12);
+		*s++ = 0x80 | ((wc>>6)&0x3f);
+		*s = 0x80 | (wc&0x3f);
+		return 3;
+	} else if ((unsigned)wc-0x10000 < 0x100000) {
+		*s++ = 0xf0 | (wc>>18);
+		*s++ = 0x80 | ((wc>>12)&0x3f);
+		*s++ = 0x80 | ((wc>>6)&0x3f);
+		*s = 0x80 | (wc&0x3f);
+		return 4;
+	}
+	errno = EILSEQ;
+	return -1;
+}
diff --git a/src/stdio/__fclose_ca.c b/src/stdio/__fclose_ca.c
new file mode 100644
index 0000000..e0b12a1
--- /dev/null
+++ b/src/stdio/__fclose_ca.c
@@ -0,0 +1,6 @@
+#include "stdio_impl.h"
+
+int __fclose_ca(FILE *f)
+{
+	return f->close(f);
+}
diff --git a/src/stdio/__fdopen.c b/src/stdio/__fdopen.c
new file mode 100644
index 0000000..1b82b92
--- /dev/null
+++ b/src/stdio/__fdopen.c
@@ -0,0 +1,61 @@
+#include "stdio_impl.h"
+#include <stdlib.h>
+#include <errno.h>
+#include <string.h>
+#include "libc.h"
+
+FILE *__fdopen(int fd, const char *mode)
+{
+	FILE *f;
+
+	/* Check for valid initial mode character */
+	if (!strchr("rwa", *mode)) {
+		errno = EINVAL;
+		return 0;
+	}
+
+	/* Allocate FILE+buffer or fail */
+	if (!(f=malloc(sizeof *f + UNGET + BUFSIZ))) return 0;
+
+	/* Zero-fill only the struct, not the buffer */
+	memset(f, 0, sizeof *f);
+
+	/* Impose mode restrictions */
+	if (!strchr(mode, '+')) f->flags = (*mode == 'r') ? F_NOWR : F_NORD;
+
+	/* Apply close-on-exec flag */
+	// FIXME: if (strchr(mode, 'e')) __syscall(SYS_fcntl, fd, F_SETFD, FD_CLOEXEC);
+
+	/* Set append mode on fd if opened for append */
+	if (*mode == 'a') {
+        /* FIXME: 
+		int flags = __syscall(SYS_fcntl, fd, F_GETFL);
+		if (!(flags & O_APPEND))
+			__syscall(SYS_fcntl, fd, F_SETFL, flags | O_APPEND);
+		*/
+        f->flags |= F_APP;
+	}
+
+	f->fd = fd;
+	f->buf = (unsigned char *)f + sizeof *f + UNGET;
+	f->buf_size = BUFSIZ;
+
+	/* Activate line buffered mode for terminals */
+	f->lbf = EOF;
+	
+    if (!(f->flags & F_NOWR))
+		f->lbf = '\n';
+
+	/* Initialize op ptrs. No problem if some are unneeded. */
+	f->read = __stdio_read;
+	f->write = __stdio_write;
+	f->seek = __stdio_seek;
+	f->close = __stdio_close;
+
+	f->lock = -1;
+
+	/* Add new FILE to open file list */
+	return __ofl_add(f);
+}
+
+weak_alias(__fdopen, fdopen);
diff --git a/src/stdio/__fmodeflags.c b/src/stdio/__fmodeflags.c
new file mode 100644
index 0000000..3832dc6
--- /dev/null
+++ b/src/stdio/__fmodeflags.c
@@ -0,0 +1,16 @@
+#include <nxbase.h>
+#include <string.h>
+
+int __fmodeflags(const char *mode)
+{
+	int flags;
+	if (strchr(mode, '+')) flags = NX_VFS_O_RDWR;
+	else if (*mode == 'r') flags = NX_VFS_O_RDONLY;
+	else flags = NX_VFS_O_WRONLY;
+	if (strchr(mode, 'x')) flags |= NX_VFS_O_EXCL;
+	//FIXME: if (strchr(mode, 'e')) flags |= NX_VFS_O_CLOEXEC;
+	if (*mode != 'r') flags |= NX_VFS_O_CREAT;
+	if (*mode == 'w') flags |= NX_VFS_O_TRUNC;
+	if (*mode == 'a') flags |= NX_VFS_O_APPEND;
+	return flags;
+}
diff --git a/src/stdio/__fopen_rb_ca.c b/src/stdio/__fopen_rb_ca.c
new file mode 100644
index 0000000..6075abd
--- /dev/null
+++ b/src/stdio/__fopen_rb_ca.c
@@ -0,0 +1,21 @@
+#include "stdio_impl.h"
+#include <string.h>
+#include <nxbase.h>
+
+FILE *__fopen_rb_ca(const char *filename, FILE *f, unsigned char *buf, size_t len)
+{
+	memset(f, 0, sizeof *f);
+
+	f->fd = NX_FileOpen(filename, NX_VFS_O_RDONLY, 0666);
+	if (f->fd < 0) return 0;
+	
+	f->flags = F_NOWR | F_PERM;
+	f->buf = buf + UNGET;
+	f->buf_size = len - UNGET;
+	f->read = __stdio_read;
+	f->seek = __stdio_seek;
+	f->close = __stdio_close;
+	f->lock = -1;
+
+	return f;
+}
diff --git a/src/stdio/__init_stdio.c b/src/stdio/__init_stdio.c
new file mode 100644
index 0000000..8634d09
--- /dev/null
+++ b/src/stdio/__init_stdio.c
@@ -0,0 +1,19 @@
+#include <nxbase.h>
+#include <assert.h>
+#include "stdio_impl.h"
+
+void __init_stdio(void)
+{
+    /* open stdio device */
+    stdin->fd = NX_FileOpen("/dev/console", NX_VFS_O_RDONLY, 0666);
+    assert(stdin->fd >= 0);
+
+    stdout->fd = NX_FileOpen("/dev/console", NX_VFS_O_WRONLY, 0666);
+    assert(stdout->fd >= 0);
+    stderr->fd = NX_FileOpen("/dev/console", NX_VFS_O_WRONLY, 0666);
+    assert(stderr->fd >= 0);
+
+    /* no buffer for stdin & stdout */
+    setbuf(stdin, NULL);
+    setbuf(stdout, NULL);
+}
diff --git a/src/stdio/__lockfile.c b/src/stdio/__lockfile.c
new file mode 100644
index 0000000..b2b667b
--- /dev/null
+++ b/src/stdio/__lockfile.c
@@ -0,0 +1,10 @@
+#include "stdio_impl.h"
+
+int __lockfile(FILE *f)
+{
+	return 1;
+}
+
+void __unlockfile(FILE *f)
+{
+}
diff --git a/src/stdio/__overflow.c b/src/stdio/__overflow.c
new file mode 100644
index 0000000..e65a594
--- /dev/null
+++ b/src/stdio/__overflow.c
@@ -0,0 +1,10 @@
+#include "stdio_impl.h"
+
+int __overflow(FILE *f, int _c)
+{
+	unsigned char c = _c;
+	if (!f->wend && __towrite(f)) return EOF;
+	if (f->wpos != f->wend && c != f->lbf) return *f->wpos++ = c;
+	if (f->write(f, &c, 1)!=1) return EOF;
+	return c;
+}
diff --git a/src/stdio/__stdio_close.c b/src/stdio/__stdio_close.c
new file mode 100644
index 0000000..6958367
--- /dev/null
+++ b/src/stdio/__stdio_close.c
@@ -0,0 +1,13 @@
+#include "stdio_impl.h"
+
+static int dummy(int fd)
+{
+	return fd;
+}
+
+weak_alias(dummy, __aio_close);
+
+int __stdio_close(FILE *f)
+{
+    return NX_FileClose(__aio_close(f->fd)) == NX_EOK ? 0 : -1;
+}
diff --git a/src/stdio/__stdio_exit.c b/src/stdio/__stdio_exit.c
new file mode 100644
index 0000000..a5e42c6
--- /dev/null
+++ b/src/stdio/__stdio_exit.c
@@ -0,0 +1,25 @@
+#include "stdio_impl.h"
+
+static FILE *volatile dummy_file = 0;
+weak_alias(dummy_file, __stdin_used);
+weak_alias(dummy_file, __stdout_used);
+weak_alias(dummy_file, __stderr_used);
+
+static void close_file(FILE *f)
+{
+	if (!f) return;
+	FFINALLOCK(f);
+	if (f->wpos != f->wbase) f->write(f, 0, 0);
+	if (f->rpos != f->rend) f->seek(f, f->rpos-f->rend, SEEK_CUR);
+}
+
+void __stdio_exit(void)
+{
+	FILE *f;
+	for (f=*__ofl_lock(); f; f=f->next) close_file(f);
+	close_file(__stdin_used);
+	close_file(__stdout_used);
+	close_file(__stderr_used);
+}
+
+weak_alias(__stdio_exit, __stdio_exit_needed);
diff --git a/src/stdio/__stdio_read.c b/src/stdio/__stdio_read.c
new file mode 100644
index 0000000..62da921
--- /dev/null
+++ b/src/stdio/__stdio_read.c
@@ -0,0 +1,50 @@
+#include "stdio_impl.h"
+#include <nxbase.h>
+
+struct iovec {
+    void      *iov_base;      /* starting address of buffer */
+    size_t    iov_len;        /* size of buffer */
+};
+
+static ssize_t __readv(int fd, struct iovec * vec, size_t nr)
+{
+    ssize_t len;
+    int i;
+
+    len = 0;
+    for (i = 0; i < nr; i++)
+    {
+        if (vec[i].iov_base && vec[i].iov_len > 0)
+        {
+            ssize_t err = NX_FileRead(fd, vec[i].iov_base, vec[i].iov_len);
+            len += err;
+            if (!err)
+            {
+                return len;
+            }
+        }
+    }
+    return len;
+}
+
+size_t __stdio_read(FILE *f, unsigned char *buf, size_t len)
+{
+	struct iovec iov[2] = {
+		{ .iov_base = buf, .iov_len = len - !!f->buf_size },
+		{ .iov_base = f->buf, .iov_len = f->buf_size }
+	};
+	ssize_t cnt;
+
+	cnt = iov[0].iov_len ? __readv(f->fd, iov, 2)
+		: NX_FileRead(f->fd, iov[1].iov_base, iov[1].iov_len);
+	if (cnt <= 0) {
+		f->flags |= cnt ? F_ERR : F_EOF;
+		return 0;
+	}
+	if (cnt <= iov[0].iov_len) return cnt;
+	cnt -= iov[0].iov_len;
+	f->rpos = f->buf;
+	f->rend = f->buf + cnt;
+	if (f->buf_size) buf[len-1] = *f->rpos++;
+	return len;
+}
diff --git a/src/stdio/__stdio_seek.c b/src/stdio/__stdio_seek.c
new file mode 100644
index 0000000..00f8dc1
--- /dev/null
+++ b/src/stdio/__stdio_seek.c
@@ -0,0 +1,7 @@
+#include "stdio_impl.h"
+#include <nxbase.h>
+
+off_t __stdio_seek(FILE *f, off_t off, int whence)
+{
+    return NX_FileSeek(f->fd, off, whence);
+}
diff --git a/src/stdio/__stdio_write.c b/src/stdio/__stdio_write.c
new file mode 100644
index 0000000..f9fa660
--- /dev/null
+++ b/src/stdio/__stdio_write.c
@@ -0,0 +1,60 @@
+#include "stdio_impl.h"
+#include <nxbase.h>
+
+struct iovec {
+    void      *iov_base;      /* starting address of buffer */
+    size_t    iov_len;        /* size of buffer */
+};
+
+static ssize_t __writev(int fd, struct iovec * vec, size_t nr)
+{
+    ssize_t len;
+    int i;
+
+    len = 0;
+    for (i = 0; i < nr; i++)
+    {
+        if (vec[i].iov_base && vec[i].iov_len > 0)
+        {
+            ssize_t err = NX_FileWrite(fd, vec[i].iov_base, vec[i].iov_len);
+            len += err;
+            if (!err)
+            {
+                return len;
+            }
+        }
+    }
+    return len;
+}
+
+size_t __stdio_write(FILE *f, const unsigned char *buf, size_t len)
+{
+	struct iovec iovs[2] = {
+		{ .iov_base = f->wbase, .iov_len = f->wpos-f->wbase },
+		{ .iov_base = (void *)buf, .iov_len = len }
+	};
+	struct iovec *iov = iovs;
+	size_t rem = iov[0].iov_len + iov[1].iov_len;
+	int iovcnt = 2;
+	ssize_t cnt;
+	for (;;) {
+		cnt = __writev(f->fd, iov, iovcnt);
+		if (cnt == rem) {
+			f->wend = f->buf + f->buf_size;
+			f->wpos = f->wbase = f->buf;
+			return len;
+		}
+		if (cnt < 0) {
+			f->wpos = f->wbase = f->wend = 0;
+			f->flags |= F_ERR;
+			return iovcnt == 2 ? 0 : len-iov[0].iov_len;
+		}
+		rem -= cnt;
+		if (cnt > iov[0].iov_len) {
+			cnt -= iov[0].iov_len;
+			iov++; iovcnt--;
+		}
+		iov[0].iov_base = (char *)iov[0].iov_base + cnt;
+		iov[0].iov_len -= cnt;
+	}
+}
diff --git a/src/stdio/__stdout_write.c b/src/stdio/__stdout_write.c
new file mode 100644
index 0000000..3df7956
--- /dev/null
+++ b/src/stdio/__stdout_write.c
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+
+size_t __stdout_write(FILE *f, const unsigned char *buf, size_t len)
+{
+	f->write = __stdio_write;
+	if (!(f->flags & F_SVB))
+		f->lbf = -1;
+	return __stdio_write(f, buf, len);
+}
diff --git a/src/stdio/__toread.c b/src/stdio/__toread.c
new file mode 100644
index 0000000..f142ff0
--- /dev/null
+++ b/src/stdio/__toread.c
@@ -0,0 +1,19 @@
+#include <stdio_impl.h>
+
+int __toread(FILE *f)
+{
+	f->mode |= f->mode-1;
+	if (f->wpos != f->wbase) f->write(f, 0, 0);
+	f->wpos = f->wbase = f->wend = 0;
+	if (f->flags & F_NORD) {
+		f->flags |= F_ERR;
+		return EOF;
+	}
+	f->rpos = f->rend = f->buf + f->buf_size;
+	return (f->flags & F_EOF) ? EOF : 0;
+}
+
+hidden void __toread_needs_stdio_exit()
+{
+	__stdio_exit_needed();
+}
diff --git a/src/stdio/__towrite.c b/src/stdio/__towrite.c
new file mode 100644
index 0000000..5ce1970
--- /dev/null
+++ b/src/stdio/__towrite.c
@@ -0,0 +1,24 @@
+#include "stdio_impl.h"
+#include <nxbase.h>
+
+int __towrite(FILE *f)
+{
+	f->mode |= f->mode-1;
+	if (f->flags & F_NOWR) {
+		f->flags |= F_ERR;
+		return EOF;
+	}
+	/* Clear read buffer (easier than summoning nasal demons) */
+	f->rpos = f->rend = 0;
+
+	/* Activate write through the buffer. */
+	f->wpos = f->wbase = f->buf;
+	f->wend = f->buf + f->buf_size;
+
+	return 0;
+}
+
+hidden void __towrite_needs_stdio_exit()
+{
+	__stdio_exit_needed();
+}
diff --git a/src/stdio/__uflow.c b/src/stdio/__uflow.c
new file mode 100644
index 0000000..2a88bca
--- /dev/null
+++ b/src/stdio/__uflow.c
@@ -0,0 +1,11 @@
+#include "stdio_impl.h"
+
+/* This function assumes it will never be called if there is already
+ * data buffered for reading. */
+
+int __uflow(FILE *f)
+{
+	unsigned char c;
+	if (!__toread(f) && f->read(f, &c, 1)==1) return c;
+	return EOF;
+}
diff --git a/src/stdio/asprintf.c_ b/src/stdio/asprintf.c_
new file mode 100644
index 0000000..4ec8353
--- /dev/null
+++ b/src/stdio/asprintf.c_
@@ -0,0 +1,13 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdarg.h>
+
+int asprintf(char **s, const char *fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vasprintf(s, fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/clearerr.c b/src/stdio/clearerr.c
new file mode 100644
index 0000000..3bf94d3
--- /dev/null
+++ b/src/stdio/clearerr.c
@@ -0,0 +1,10 @@
+#include "stdio_impl.h"
+
+void clearerr(FILE *f)
+{
+	FLOCK(f);
+	f->flags &= ~(F_EOF|F_ERR);
+	FUNLOCK(f);
+}
+
+weak_alias(clearerr, clearerr_unlocked);
diff --git a/src/stdio/dprintf.c_ b/src/stdio/dprintf.c_
new file mode 100644
index 0000000..93082ee
--- /dev/null
+++ b/src/stdio/dprintf.c_
@@ -0,0 +1,12 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int dprintf(int fd, const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vdprintf(fd, fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/ext.c_ b/src/stdio/ext.c_
new file mode 100644
index 0000000..1fd9549
--- /dev/null
+++ b/src/stdio/ext.c_
@@ -0,0 +1,57 @@
+#define _GNU_SOURCE
+#include "stdio_impl.h"
+#include <stdio_ext.h>
+
+void _flushlbf(void)
+{
+	fflush(0);
+}
+
+int __fsetlocking(FILE *f, int type)
+{
+	return 0;
+}
+
+int __fwriting(FILE *f)
+{
+	return (f->flags & F_NORD) || f->wend;
+}
+
+int __freading(FILE *f)
+{
+	return (f->flags & F_NOWR) || f->rend;
+}
+
+int __freadable(FILE *f)
+{
+	return !(f->flags & F_NORD);
+}
+
+int __fwritable(FILE *f)
+{
+	return !(f->flags & F_NOWR);
+}
+
+int __flbf(FILE *f)
+{
+	return f->lbf >= 0;
+}
+
+size_t __fbufsize(FILE *f)
+{
+	return f->buf_size;
+}
+
+size_t __fpending(FILE *f)
+{
+	return f->wend ? f->wpos - f->wbase : 0;
+}
+
+int __fpurge(FILE *f)
+{
+	f->wpos = f->wbase = f->wend = 0;
+	f->rpos = f->rend = 0;
+	return 0;
+}
+
+weak_alias(__fpurge, fpurge);
diff --git a/src/stdio/ext2.c_ b/src/stdio/ext2.c_
new file mode 100644
index 0000000..3416278
--- /dev/null
+++ b/src/stdio/ext2.c_
@@ -0,0 +1,24 @@
+#include "stdio_impl.h"
+#include <stdio_ext.h>
+
+size_t __freadahead(FILE *f)
+{
+	return f->rend ? f->rend - f->rpos : 0;
+}
+
+const char *__freadptr(FILE *f, size_t *sizep)
+{
+	if (f->rpos == f->rend) return 0;
+	*sizep = f->rend - f->rpos;
+	return (const char *)f->rpos;
+}
+
+void __freadptrinc(FILE *f, size_t inc)
+{
+	f->rpos += inc;
+}
+
+void __fseterr(FILE *f)
+{
+	f->flags |= F_ERR;
+}
diff --git a/src/stdio/fclose.c b/src/stdio/fclose.c
new file mode 100644
index 0000000..d594532
--- /dev/null
+++ b/src/stdio/fclose.c
@@ -0,0 +1,38 @@
+#include "stdio_impl.h"
+#include <stdlib.h>
+
+static void dummy(FILE *f) { }
+weak_alias(dummy, __unlist_locked_file);
+
+int fclose(FILE *f)
+{
+	int r;
+	
+	FLOCK(f);
+	r = fflush(f);
+	r |= f->close(f);
+	FUNLOCK(f);
+
+	/* Past this point, f is closed and any further explict access
+	 * to it is undefined. However, it still exists as an entry in
+	 * the open file list and possibly in the thread's locked files
+	 * list, if it was closed while explicitly locked. Functions
+	 * which process these lists must tolerate dead FILE objects
+	 * (which necessarily have inactive buffer pointers) without
+	 * producing any side effects. */
+
+	if (f->flags & F_PERM) return r;
+
+	__unlist_locked_file(f);
+
+	FILE **head = __ofl_lock();
+	if (f->prev) f->prev->next = f->next;
+	if (f->next) f->next->prev = f->prev;
+	if (*head == f) *head = f->next;
+	__ofl_unlock();
+
+	free(f->getln_buf);
+	free(f);
+
+	return r;
+}
diff --git a/src/stdio/feof.c b/src/stdio/feof.c
new file mode 100644
index 0000000..56da6b9
--- /dev/null
+++ b/src/stdio/feof.c
@@ -0,0 +1,14 @@
+#include "stdio_impl.h"
+
+#undef feof
+
+int feof(FILE *f)
+{
+	FLOCK(f);
+	int ret = !!(f->flags & F_EOF);
+	FUNLOCK(f);
+	return ret;
+}
+
+weak_alias(feof, feof_unlocked);
+weak_alias(feof, _IO_feof_unlocked);
diff --git a/src/stdio/ferror.c b/src/stdio/ferror.c
new file mode 100644
index 0000000..d692eed
--- /dev/null
+++ b/src/stdio/ferror.c
@@ -0,0 +1,14 @@
+#include "stdio_impl.h"
+
+#undef ferror
+
+int ferror(FILE *f)
+{
+	FLOCK(f);
+	int ret = !!(f->flags & F_ERR);
+	FUNLOCK(f);
+	return ret;
+}
+
+weak_alias(ferror, ferror_unlocked);
+weak_alias(ferror, _IO_ferror_unlocked);
diff --git a/src/stdio/fflush.c b/src/stdio/fflush.c
new file mode 100644
index 0000000..b009437
--- /dev/null
+++ b/src/stdio/fflush.c
@@ -0,0 +1,47 @@
+#include "stdio_impl.h"
+
+/* stdout.c will override this if linked */
+static FILE *volatile dummy = 0;
+weak_alias(dummy, __stdout_used);
+weak_alias(dummy, __stderr_used);
+
+int fflush(FILE *f)
+{
+	if (!f) {
+		int r = 0;
+		if (__stdout_used) r |= fflush(__stdout_used);
+		if (__stderr_used) r |= fflush(__stderr_used);
+
+		for (f=*__ofl_lock(); f; f=f->next) {
+			FLOCK(f);
+			if (f->wpos != f->wbase) r |= fflush(f);
+			FUNLOCK(f);
+		}
+		__ofl_unlock();
+
+		return r;
+	}
+
+	FLOCK(f);
+
+	/* If writing, flush output */
+	if (f->wpos != f->wbase) {
+		f->write(f, 0, 0);
+		if (!f->wpos) {
+			FUNLOCK(f);
+			return EOF;
+		}
+	}
+
+	/* If reading, sync position, per POSIX */
+	if (f->rpos != f->rend) f->seek(f, f->rpos-f->rend, SEEK_CUR);
+
+	/* Clear read and write modes */
+	f->wpos = f->wbase = f->wend = 0;
+	f->rpos = f->rend = 0;
+
+	FUNLOCK(f);
+	return 0;
+}
+
+weak_alias(fflush, fflush_unlocked);
diff --git a/src/stdio/fgetc.c b/src/stdio/fgetc.c
new file mode 100644
index 0000000..2578afc
--- /dev/null
+++ b/src/stdio/fgetc.c
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include "getc.h"
+
+int fgetc(FILE *f)
+{
+	return do_getc(f);
+}
diff --git a/src/stdio/fgetln.c_ b/src/stdio/fgetln.c_
new file mode 100644
index 0000000..5748435
--- /dev/null
+++ b/src/stdio/fgetln.c_
@@ -0,0 +1,21 @@
+#define _GNU_SOURCE
+#include "stdio_impl.h"
+#include <string.h>
+
+char *fgetln(FILE *f, size_t *plen)
+{
+	char *ret = 0, *z;
+	ssize_t l;
+	FLOCK(f);
+	ungetc(getc_unlocked(f), f);
+	if (f->rend && (z=memchr(f->rpos, '\n', f->rend - f->rpos))) {
+		ret = (char *)f->rpos;
+		*plen = ++z - ret;
+		f->rpos = (void *)z;
+	} else if ((l = getline(&f->getln_buf, (size_t[]){0}, f)) > 0) {
+		*plen = l;
+		ret = f->getln_buf;
+	}
+	FUNLOCK(f);
+	return ret;
+}
diff --git a/src/stdio/fgetpos.c b/src/stdio/fgetpos.c
new file mode 100644
index 0000000..50813d2
--- /dev/null
+++ b/src/stdio/fgetpos.c
@@ -0,0 +1,11 @@
+#include "stdio_impl.h"
+
+int fgetpos(FILE *restrict f, fpos_t *restrict pos)
+{
+	off_t off = __ftello(f);
+	if (off < 0) return -1;
+	*(long long *)pos = off;
+	return 0;
+}
+
+weak_alias(fgetpos, fgetpos64);
diff --git a/src/stdio/fgets.c b/src/stdio/fgets.c
new file mode 100644
index 0000000..6171f39
--- /dev/null
+++ b/src/stdio/fgets.c
@@ -0,0 +1,48 @@
+#include "stdio_impl.h"
+#include <string.h>
+
+#define MIN(a,b) ((a)<(b) ? (a) : (b))
+
+char *fgets(char *restrict s, int n, FILE *restrict f)
+{
+	char *p = s;
+	unsigned char *z;
+	size_t k;
+	int c;
+
+	FLOCK(f);
+
+	if (n--<=1) {
+		f->mode |= f->mode-1;
+		FUNLOCK(f);
+		if (n) return 0;
+		*s = 0;
+		return s;
+	}
+
+	while (n) {
+		if (f->rpos != f->rend) {
+			z = memchr(f->rpos, '\n', f->rend - f->rpos);
+			k = z ? z - f->rpos + 1 : f->rend - f->rpos;
+			k = MIN(k, n);
+			memcpy(p, f->rpos, k);
+			f->rpos += k;
+			p += k;
+			n -= k;
+			if (z || !n) break;
+		}
+		if ((c = getc_unlocked(f)) < 0) {
+			if (p==s || !feof(f)) s = 0;
+			break;
+		}
+		n--;
+		if ((*p++ = c) == '\n') break;
+	}
+	if (s) *p = 0;
+
+	FUNLOCK(f);
+
+	return s;
+}
+
+weak_alias(fgets, fgets_unlocked);
diff --git a/src/stdio/fgetwc.c_ b/src/stdio/fgetwc.c_
new file mode 100644
index 0000000..aa10b81
--- /dev/null
+++ b/src/stdio/fgetwc.c_
@@ -0,0 +1,68 @@
+#include "stdio_impl.h"
+#include "locale_impl.h"
+#include <wchar.h>
+#include <errno.h>
+
+static wint_t __fgetwc_unlocked_internal(FILE *f)
+{
+	wchar_t wc;
+	int c;
+	size_t l;
+
+	/* Convert character from buffer if possible */
+	if (f->rpos != f->rend) {
+		l = mbtowc(&wc, (void *)f->rpos, f->rend - f->rpos);
+		if (l+1 >= 1) {
+			f->rpos += l + !l; /* l==0 means 1 byte, null */
+			return wc;
+		}
+	}
+
+	/* Convert character byte-by-byte */
+	mbstate_t st = { 0 };
+	unsigned char b;
+	int first = 1;
+	do {
+		b = c = getc_unlocked(f);
+		if (c < 0) {
+			if (!first) {
+				f->flags |= F_ERR;
+				errno = EILSEQ;
+			}
+			return WEOF;
+		}
+		l = mbrtowc(&wc, (void *)&b, 1, &st);
+		if (l == -1) {
+			if (!first) {
+				f->flags |= F_ERR;
+				ungetc(b, f);
+			}
+			return WEOF;
+		}
+		first = 0;
+	} while (l == -2);
+
+	return wc;
+}
+
+wint_t __fgetwc_unlocked(FILE *f)
+{
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
+	if (f->mode <= 0) fwide(f, 1);
+	*ploc = f->locale;
+	wchar_t wc = __fgetwc_unlocked_internal(f);
+	*ploc = loc;
+	return wc;
+}
+
+wint_t fgetwc(FILE *f)
+{
+	wint_t c;
+	FLOCK(f);
+	c = __fgetwc_unlocked(f);
+	FUNLOCK(f);
+	return c;
+}
+
+weak_alias(__fgetwc_unlocked, fgetwc_unlocked);
+weak_alias(__fgetwc_unlocked, getwc_unlocked);
diff --git a/src/stdio/fgetws.c_ b/src/stdio/fgetws.c_
new file mode 100644
index 0000000..b08b304
--- /dev/null
+++ b/src/stdio/fgetws.c_
@@ -0,0 +1,33 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+#include <errno.h>
+
+wint_t __fgetwc_unlocked(FILE *);
+
+wchar_t *fgetws(wchar_t *restrict s, int n, FILE *restrict f)
+{
+	wchar_t *p = s;
+
+	if (!n--) return s;
+
+	FLOCK(f);
+
+	/* Setup a dummy errno so we can detect EILSEQ. This is
+	 * the only way to catch encoding errors in the form of a
+	 * partial character just before EOF. */
+	errno = EAGAIN;
+	for (; n; n--) {
+		wint_t c = __fgetwc_unlocked(f);
+		if (c == WEOF) break;
+		*p++ = c;
+		if (c == '\n') break;
+	}
+	*p = 0;
+	if (ferror(f) || errno==EILSEQ) p = s;
+
+	FUNLOCK(f);
+
+	return (p == s) ? NULL : s;
+}
+
+weak_alias(fgetws, fgetws_unlocked);
diff --git a/src/stdio/fileno.c_ b/src/stdio/fileno.c_
new file mode 100644
index 0000000..0bd0e98
--- /dev/null
+++ b/src/stdio/fileno.c_
@@ -0,0 +1,16 @@
+#include "stdio_impl.h"
+#include <errno.h>
+
+int fileno(FILE *f)
+{
+	FLOCK(f);
+	int fd = f->fd;
+	FUNLOCK(f);
+	if (fd < 0) {
+		errno = EBADF;
+		return -1;
+	}
+	return fd;
+}
+
+weak_alias(fileno, fileno_unlocked);
diff --git a/src/stdio/flockfile.c_ b/src/stdio/flockfile.c_
new file mode 100644
index 0000000..8e22065
--- /dev/null
+++ b/src/stdio/flockfile.c_
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+#include "pthread_impl.h"
+
+void flockfile(FILE *f)
+{
+	if (!ftrylockfile(f)) return;
+	__lockfile(f);
+	__register_locked_file(f, __pthread_self());
+}
diff --git a/src/stdio/fmemopen.c_ b/src/stdio/fmemopen.c_
new file mode 100644
index 0000000..343e3e3
--- /dev/null
+++ b/src/stdio/fmemopen.c_
@@ -0,0 +1,127 @@
+#include "stdio_impl.h"
+#include <errno.h>
+#include <string.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <inttypes.h>
+#include "libc.h"
+
+struct cookie {
+	size_t pos, len, size;
+	unsigned char *buf;
+	int mode;
+};
+
+struct mem_FILE {
+	FILE f;
+	struct cookie c;
+	unsigned char buf[UNGET+BUFSIZ], buf2[];
+};
+
+static off_t mseek(FILE *f, off_t off, int whence)
+{
+	ssize_t base;
+	struct cookie *c = f->cookie;
+	if (whence>2U) {
+fail:
+		errno = EINVAL;
+		return -1;
+	}
+	base = (size_t [3]){0, c->pos, c->len}[whence];
+	if (off < -base || off > (ssize_t)c->size-base) goto fail;
+	return c->pos = base+off;
+}
+
+static size_t mread(FILE *f, unsigned char *buf, size_t len)
+{
+	struct cookie *c = f->cookie;
+	size_t rem = c->len - c->pos;
+	if (c->pos > c->len) rem = 0;
+	if (len > rem) {
+		len = rem;
+		f->flags |= F_EOF;
+	}
+	memcpy(buf, c->buf+c->pos, len);
+	c->pos += len;
+	rem -= len;
+	if (rem > f->buf_size) rem = f->buf_size;
+	f->rpos = f->buf;
+	f->rend = f->buf + rem;
+	memcpy(f->rpos, c->buf+c->pos, rem);
+	c->pos += rem;
+	return len;
+}
+
+static size_t mwrite(FILE *f, const unsigned char *buf, size_t len)
+{
+	struct cookie *c = f->cookie;
+	size_t rem;
+	size_t len2 = f->wpos - f->wbase;
+	if (len2) {
+		f->wpos = f->wbase;
+		if (mwrite(f, f->wpos, len2) < len2) return 0;
+	}
+	if (c->mode == 'a') c->pos = c->len;
+	rem = c->size - c->pos;
+	if (len > rem) len = rem;
+	memcpy(c->buf+c->pos, buf, len);
+	c->pos += len;
+	if (c->pos > c->len) {
+		c->len = c->pos;
+		if (c->len < c->size) c->buf[c->len] = 0;
+		else if ((f->flags&F_NORD) && c->size) c->buf[c->size-1] = 0;
+	}
+	return len;
+}
+
+static int mclose(FILE *m)
+{
+	return 0;
+}
+
+FILE *fmemopen(void *restrict buf, size_t size, const char *restrict mode)
+{
+	struct mem_FILE *f;
+	int plus = !!strchr(mode, '+');
+	
+	if (!strchr("rwa", *mode)) {
+		errno = EINVAL;
+		return 0;
+	}
+
+	if (!buf && size > PTRDIFF_MAX) {
+		errno = ENOMEM;
+		return 0;
+	}
+
+	f = malloc(sizeof *f + (buf?0:size));
+	if (!f) return 0;
+	memset(f, 0, offsetof(struct mem_FILE, buf));
+	f->f.cookie = &f->c;
+	f->f.fd = -1;
+	f->f.lbf = EOF;
+	f->f.buf = f->buf + UNGET;
+	f->f.buf_size = sizeof f->buf - UNGET;
+	if (!buf) {
+		buf = f->buf2;
+		memset(buf, 0, size);
+	}
+
+	f->c.buf = buf;
+	f->c.size = size;
+	f->c.mode = *mode;
+	
+	if (!plus) f->f.flags = (*mode == 'r') ? F_NOWR : F_NORD;
+	if (*mode == 'r') f->c.len = size;
+	else if (*mode == 'a') f->c.len = f->c.pos = strnlen(buf, size);
+	else if (plus) *f->c.buf = 0;
+
+	f->f.read = mread;
+	f->f.write = mwrite;
+	f->f.seek = mseek;
+	f->f.close = mclose;
+
+	if (!libc.threaded) f->f.lock = -1;
+
+	return __ofl_add(&f->f);
+}
diff --git a/src/stdio/fopen.c b/src/stdio/fopen.c
new file mode 100644
index 0000000..a6154a8
--- /dev/null
+++ b/src/stdio/fopen.c
@@ -0,0 +1,30 @@
+#include "stdio_impl.h"
+#include <string.h>
+#include <errno.h>
+
+FILE *fopen(const char *restrict filename, const char *restrict mode)
+{
+	FILE *f;
+	int fd;
+	int flags;
+
+	/* Check for valid initial mode character */
+	if (!strchr("rwa", *mode)) {
+		errno = EINVAL;
+		return 0;
+	}
+
+	/* Compute the flags to pass to open() */
+	flags = __fmodeflags(mode);
+
+	fd = NX_FileOpen(filename, flags, 0666);
+	if (fd < 0) return 0;
+
+	f = __fdopen(fd, mode);
+	if (f) return f;
+
+    NX_FileClose(fd);
+	return 0;
+}
+
+weak_alias(fopen, fopen64);
diff --git a/src/stdio/fopencookie.c_ b/src/stdio/fopencookie.c_
new file mode 100644
index 0000000..da042fe
--- /dev/null
+++ b/src/stdio/fopencookie.c_
@@ -0,0 +1,135 @@
+#define _GNU_SOURCE
+#include "stdio_impl.h"
+#include <stdlib.h>
+#include <sys/ioctl.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <string.h>
+
+struct fcookie {
+	void *cookie;
+	cookie_io_functions_t iofuncs;
+};
+
+struct cookie_FILE {
+	FILE f;
+	struct fcookie fc;
+	unsigned char buf[UNGET+BUFSIZ];
+};
+
+static size_t cookieread(FILE *f, unsigned char *buf, size_t len)
+{
+	struct fcookie *fc = f->cookie;
+	ssize_t ret = -1;
+	size_t remain = len, readlen = 0;
+	size_t len2 = len - !!f->buf_size;
+
+	if (!fc->iofuncs.read) goto bail;
+
+	if (len2) {
+		ret = fc->iofuncs.read(fc->cookie, (char *) buf, len2);
+		if (ret <= 0) goto bail;
+
+		readlen += ret;
+		remain -= ret;
+	}
+
+	if (!f->buf_size || remain > !!f->buf_size) return readlen;
+
+	f->rpos = f->buf;
+	ret = fc->iofuncs.read(fc->cookie, (char *) f->rpos, f->buf_size);
+	if (ret <= 0) goto bail;
+	f->rend = f->rpos + ret;
+
+	buf[readlen++] = *f->rpos++;
+
+	return readlen;
+
+bail:
+	f->flags |= ret == 0 ? F_EOF : F_ERR;
+	f->rpos = f->rend = f->buf;
+	return readlen;
+}
+
+static size_t cookiewrite(FILE *f, const unsigned char *buf, size_t len)
+{
+	struct fcookie *fc = f->cookie;
+	ssize_t ret;
+	size_t len2 = f->wpos - f->wbase;
+	if (!fc->iofuncs.write) return len;
+	if (len2) {
+		f->wpos = f->wbase;
+		if (cookiewrite(f, f->wpos, len2) < len2) return 0;
+	}
+	ret = fc->iofuncs.write(fc->cookie, (const char *) buf, len);
+	if (ret < 0) {
+		f->wpos = f->wbase = f->wend = 0;
+		f->flags |= F_ERR;
+		return 0;
+	}
+	return ret;
+}
+
+static off_t cookieseek(FILE *f, off_t off, int whence)
+{
+	struct fcookie *fc = f->cookie;
+	int res;
+	if (whence > 2U) {
+		errno = EINVAL;
+		return -1;
+	}
+	if (!fc->iofuncs.seek) {
+		errno = ENOTSUP;
+		return -1;
+	}
+	res = fc->iofuncs.seek(fc->cookie, &off, whence);
+	if (res < 0)
+		return res;
+	return off;
+}
+
+static int cookieclose(FILE *f)
+{
+	struct fcookie *fc = f->cookie;
+	if (fc->iofuncs.close) return fc->iofuncs.close(fc->cookie);
+	return 0;
+}
+
+FILE *fopencookie(void *cookie, const char *mode, cookie_io_functions_t iofuncs)
+{
+	struct cookie_FILE *f;
+
+	/* Check for valid initial mode character */
+	if (!strchr("rwa", *mode)) {
+		errno = EINVAL;
+		return 0;
+	}
+
+	/* Allocate FILE+fcookie+buffer or fail */
+	if (!(f=malloc(sizeof *f))) return 0;
+
+	/* Zero-fill only the struct, not the buffer */
+	memset(&f->f, 0, sizeof f->f);
+
+	/* Impose mode restrictions */
+	if (!strchr(mode, '+')) f->f.flags = (*mode == 'r') ? F_NOWR : F_NORD;
+
+	/* Set up our fcookie */
+	f->fc.cookie = cookie;
+	f->fc.iofuncs = iofuncs;
+
+	f->f.fd = -1;
+	f->f.cookie = &f->fc;
+	f->f.buf = f->buf + UNGET;
+	f->f.buf_size = sizeof f->buf - UNGET;
+	f->f.lbf = EOF;
+
+	/* Initialize op ptrs. No problem if some are unneeded. */
+	f->f.read = cookieread;
+	f->f.write = cookiewrite;
+	f->f.seek = cookieseek;
+	f->f.close = cookieclose;
+
+	/* Add new FILE to open file list */
+	return __ofl_add(&f->f);
+}
diff --git a/src/stdio/fprintf.c b/src/stdio/fprintf.c
new file mode 100644
index 0000000..948743f
--- /dev/null
+++ b/src/stdio/fprintf.c
@@ -0,0 +1,12 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int fprintf(FILE *restrict f, const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vfprintf(f, fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/fputc.c b/src/stdio/fputc.c
new file mode 100644
index 0000000..f364ed3
--- /dev/null
+++ b/src/stdio/fputc.c
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include "putc.h"
+
+int fputc(int c, FILE *f)
+{
+	return do_putc(c, f);
+}
diff --git a/src/stdio/fputs.c b/src/stdio/fputs.c
new file mode 100644
index 0000000..1cf344f
--- /dev/null
+++ b/src/stdio/fputs.c
@@ -0,0 +1,10 @@
+#include "stdio_impl.h"
+#include <string.h>
+
+int fputs(const char *restrict s, FILE *restrict f)
+{
+	size_t l = strlen(s);
+	return (fwrite(s, 1, l, f)==l) - 1;
+}
+
+weak_alias(fputs, fputs_unlocked);
diff --git a/src/stdio/fputwc.c_ b/src/stdio/fputwc.c_
new file mode 100644
index 0000000..789fe9c
--- /dev/null
+++ b/src/stdio/fputwc.c_
@@ -0,0 +1,40 @@
+#include "stdio_impl.h"
+#include "locale_impl.h"
+#include <wchar.h>
+#include <limits.h>
+#include <ctype.h>
+
+wint_t __fputwc_unlocked(wchar_t c, FILE *f)
+{
+	char mbc[MB_LEN_MAX];
+	int l;
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
+
+	if (f->mode <= 0) fwide(f, 1);
+	*ploc = f->locale;
+
+	if (isascii(c)) {
+		c = putc_unlocked(c, f);
+	} else if (f->wpos + MB_LEN_MAX < f->wend) {
+		l = wctomb((void *)f->wpos, c);
+		if (l < 0) c = WEOF;
+		else f->wpos += l;
+	} else {
+		l = wctomb(mbc, c);
+		if (l < 0 || __fwritex((void *)mbc, l, f) < l) c = WEOF;
+	}
+	if (c==WEOF) f->flags |= F_ERR;
+	*ploc = loc;
+	return c;
+}
+
+wint_t fputwc(wchar_t c, FILE *f)
+{
+	FLOCK(f);
+	c = __fputwc_unlocked(c, f);
+	FUNLOCK(f);
+	return c;
+}
+
+weak_alias(__fputwc_unlocked, fputwc_unlocked);
+weak_alias(__fputwc_unlocked, putwc_unlocked);
diff --git a/src/stdio/fputws.c_ b/src/stdio/fputws.c_
new file mode 100644
index 0000000..0ed02f1
--- /dev/null
+++ b/src/stdio/fputws.c_
@@ -0,0 +1,29 @@
+#include "stdio_impl.h"
+#include "locale_impl.h"
+#include <wchar.h>
+
+int fputws(const wchar_t *restrict ws, FILE *restrict f)
+{
+	unsigned char buf[BUFSIZ];
+	size_t l=0;
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
+
+	FLOCK(f);
+
+	fwide(f, 1);
+	*ploc = f->locale;
+
+	while (ws && (l = wcsrtombs((void *)buf, (void*)&ws, sizeof buf, 0))+1 > 1)
+		if (__fwritex(buf, l, f) < l) {
+			FUNLOCK(f);
+			*ploc = loc;
+			return -1;
+		}
+
+	FUNLOCK(f);
+
+	*ploc = loc;
+	return l; /* 0 or -1 */
+}
+
+weak_alias(fputws, fputws_unlocked);
diff --git a/src/stdio/fread.c b/src/stdio/fread.c
new file mode 100644
index 0000000..a2116da
--- /dev/null
+++ b/src/stdio/fread.c
@@ -0,0 +1,38 @@
+#include "stdio_impl.h"
+#include <string.h>
+
+#define MIN(a,b) ((a)<(b) ? (a) : (b))
+
+size_t fread(void *restrict destv, size_t size, size_t nmemb, FILE *restrict f)
+{
+	unsigned char *dest = destv;
+	size_t len = size*nmemb, l = len, k;
+	if (!size) nmemb = 0;
+
+	FLOCK(f);
+
+	f->mode |= f->mode-1;
+
+	if (f->rpos != f->rend) {
+		/* First exhaust the buffer. */
+		k = MIN(f->rend - f->rpos, l);
+		memcpy(dest, f->rpos, k);
+		f->rpos += k;
+		dest += k;
+		l -= k;
+	}
+	
+	/* Read the remainder directly */
+	for (; l; l-=k, dest+=k) {
+		k = __toread(f) ? 0 : f->read(f, dest, l);
+		if (!k) {
+			FUNLOCK(f);
+			return (len-l)/size;
+		}
+	}
+
+	FUNLOCK(f);
+	return nmemb;
+}
+
+weak_alias(fread, fread_unlocked);
diff --git a/src/stdio/freopen.c b/src/stdio/freopen.c
new file mode 100644
index 0000000..11b8faf
--- /dev/null
+++ b/src/stdio/freopen.c
@@ -0,0 +1,48 @@
+#include "stdio_impl.h"
+#include <nxbase.h>
+
+/* The basic idea of this implementation is to open a new FILE,
+ * hack the necessary parts of the new FILE into the old one, then
+ * close the new FILE. */
+
+/* Locking IS necessary because another thread may provably hold the
+ * lock, via flockfile or otherwise, when freopen is called, and in that
+ * case, freopen cannot act until the lock is released. */
+
+FILE *freopen(const char *restrict filename, const char *restrict mode, FILE *restrict f)
+{
+	int fl = __fmodeflags(mode);
+	FILE *f2;
+
+	FLOCK(f);
+
+	fflush(f);
+
+	if (!filename) {
+		fl &= ~(NX_VFS_O_CREAT|NX_VFS_O_EXCL);
+	} else {
+		f2 = fopen(filename, mode);
+		if (!f2) goto fail;
+		if (f2->fd == f->fd) f2->fd = -1; /* avoid closing in fclose */
+		else goto fail2;
+
+		f->flags = (f->flags & F_PERM) | f2->flags;
+		f->read = f2->read;
+		f->write = f2->write;
+		f->seek = f2->seek;
+		f->close = f2->close;
+
+		fclose(f2);
+	}
+
+	FUNLOCK(f);
+	return f;
+
+fail2:
+	fclose(f2);
+fail:
+	fclose(f);
+	return NULL;
+}
+
+weak_alias(freopen, freopen64);
diff --git a/src/stdio/fscanf.c b/src/stdio/fscanf.c
new file mode 100644
index 0000000..f639e11
--- /dev/null
+++ b/src/stdio/fscanf.c
@@ -0,0 +1,14 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int fscanf(FILE *restrict f, const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vfscanf(f, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+weak_alias(fscanf, __isoc99_fscanf);
diff --git a/src/stdio/fseek.c b/src/stdio/fseek.c
new file mode 100644
index 0000000..439308f
--- /dev/null
+++ b/src/stdio/fseek.c
@@ -0,0 +1,43 @@
+#include "stdio_impl.h"
+
+int __fseeko_unlocked(FILE *f, off_t off, int whence)
+{
+	/* Adjust relative offset for unread data in buffer, if any. */
+	if (whence == SEEK_CUR && f->rend) off -= f->rend - f->rpos;
+
+	/* Flush write buffer, and report error on failure. */
+	if (f->wpos != f->wbase) {
+		f->write(f, 0, 0);
+		if (!f->wpos) return -1;
+	}
+
+	/* Leave writing mode */
+	f->wpos = f->wbase = f->wend = 0;
+
+	/* Perform the underlying seek. */
+	if (f->seek(f, off, whence) < 0) return -1;
+
+	/* If seek succeeded, file is seekable and we discard read buffer. */
+	f->rpos = f->rend = 0;
+	f->flags &= ~F_EOF;
+	
+	return 0;
+}
+
+int __fseeko(FILE *f, off_t off, int whence)
+{
+	int result;
+	FLOCK(f);
+	result = __fseeko_unlocked(f, off, whence);
+	FUNLOCK(f);
+	return result;
+}
+
+int fseek(FILE *f, long off, int whence)
+{
+	return __fseeko(f, off, whence);
+}
+
+weak_alias(__fseeko, fseeko);
+
+weak_alias(fseeko, fseeko64);
diff --git a/src/stdio/fsetpos.c b/src/stdio/fsetpos.c
new file mode 100644
index 0000000..77ab8d8
--- /dev/null
+++ b/src/stdio/fsetpos.c
@@ -0,0 +1,8 @@
+#include "stdio_impl.h"
+
+int fsetpos(FILE *f, const fpos_t *pos)
+{
+	return __fseeko(f, *(const long long *)pos, SEEK_SET);
+}
+
+weak_alias(fsetpos, fsetpos64);
diff --git a/src/stdio/ftell.c b/src/stdio/ftell.c
new file mode 100644
index 0000000..1a2afbb
--- /dev/null
+++ b/src/stdio/ftell.c
@@ -0,0 +1,41 @@
+#include "stdio_impl.h"
+#include <limits.h>
+#include <errno.h>
+
+off_t __ftello_unlocked(FILE *f)
+{
+	off_t pos = f->seek(f, 0,
+		(f->flags & F_APP) && f->wpos != f->wbase
+		? SEEK_END : SEEK_CUR);
+	if (pos < 0) return pos;
+
+	/* Adjust for data in buffer. */
+	if (f->rend)
+		pos += f->rpos - f->rend;
+	else if (f->wbase)
+		pos += f->wpos - f->wbase;
+	return pos;
+}
+
+off_t __ftello(FILE *f)
+{
+	off_t pos;
+	FLOCK(f);
+	pos = __ftello_unlocked(f);
+	FUNLOCK(f);
+	return pos;
+}
+
+long ftell(FILE *f)
+{
+	off_t pos = __ftello(f);
+	if (pos > LONG_MAX) {
+		errno = EOVERFLOW;
+		return -1;
+	}
+	return pos;
+}
+
+weak_alias(__ftello, ftello);
+
+weak_alias(ftello, ftello64);
diff --git a/src/stdio/ftrylockfile.c_ b/src/stdio/ftrylockfile.c_
new file mode 100644
index 0000000..5065058
--- /dev/null
+++ b/src/stdio/ftrylockfile.c_
@@ -0,0 +1,46 @@
+#include "stdio_impl.h"
+#include "pthread_impl.h"
+#include <limits.h>
+
+void __do_orphaned_stdio_locks()
+{
+	FILE *f;
+	for (f=__pthread_self()->stdio_locks; f; f=f->next_locked)
+		a_store(&f->lock, 0x40000000);
+}
+
+void __unlist_locked_file(FILE *f)
+{
+	if (f->lockcount) {
+		if (f->next_locked) f->next_locked->prev_locked = f->prev_locked;
+		if (f->prev_locked) f->prev_locked->next_locked = f->next_locked;
+		else __pthread_self()->stdio_locks = f->next_locked;
+	}
+}
+
+void __register_locked_file(FILE *f, pthread_t self)
+{
+	f->lockcount = 1;
+	f->prev_locked = 0;
+	f->next_locked = self->stdio_locks;
+	if (f->next_locked) f->next_locked->prev_locked = f;
+	self->stdio_locks = f;
+}
+
+int ftrylockfile(FILE *f)
+{
+	pthread_t self = __pthread_self();
+	int tid = self->tid;
+	int owner = f->lock;
+	if ((owner & ~MAYBE_WAITERS) == tid) {
+		if (f->lockcount == LONG_MAX)
+			return -1;
+		f->lockcount++;
+		return 0;
+	}
+	if (owner < 0) f->lock = owner = 0;
+	if (owner || a_cas(&f->lock, 0, tid))
+		return -1;
+	__register_locked_file(f, self);
+	return 0;
+}
diff --git a/src/stdio/funlockfile.c_ b/src/stdio/funlockfile.c_
new file mode 100644
index 0000000..44d8b0d
--- /dev/null
+++ b/src/stdio/funlockfile.c_
@@ -0,0 +1,13 @@
+#include "stdio_impl.h"
+#include "pthread_impl.h"
+
+void funlockfile(FILE *f)
+{
+	if (f->lockcount == 1) {
+		__unlist_locked_file(f);
+		f->lockcount = 0;
+		__unlockfile(f);
+	} else {
+		f->lockcount--;
+	}
+}
diff --git a/src/stdio/fwide.c_ b/src/stdio/fwide.c_
new file mode 100644
index 0000000..8bab634
--- /dev/null
+++ b/src/stdio/fwide.c_
@@ -0,0 +1,16 @@
+#include <wchar.h>
+#include "stdio_impl.h"
+#include "locale_impl.h"
+
+int fwide(FILE *f, int mode)
+{
+	FLOCK(f);
+	if (mode) {
+		if (!f->locale) f->locale = MB_CUR_MAX==1
+			? C_LOCALE : UTF8_LOCALE;
+		if (!f->mode) f->mode = mode>0 ? 1 : -1;
+	}
+	mode = f->mode;
+	FUNLOCK(f);
+	return mode;
+}
diff --git a/src/stdio/fwprintf.c_ b/src/stdio/fwprintf.c_
new file mode 100644
index 0000000..9ce4f01
--- /dev/null
+++ b/src/stdio/fwprintf.c_
@@ -0,0 +1,13 @@
+#include <stdio.h>
+#include <stdarg.h>
+#include <wchar.h>
+
+int fwprintf(FILE *restrict f, const wchar_t *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vfwprintf(f, fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/fwrite.c b/src/stdio/fwrite.c
new file mode 100644
index 0000000..7a567b2
--- /dev/null
+++ b/src/stdio/fwrite.c
@@ -0,0 +1,38 @@
+#include "stdio_impl.h"
+#include <string.h>
+
+size_t __fwritex(const unsigned char *restrict s, size_t l, FILE *restrict f)
+{
+	size_t i=0;
+
+	if (!f->wend && __towrite(f)) return 0;
+
+	if (l > f->wend - f->wpos) return f->write(f, s, l);
+
+	if (f->lbf >= 0) {
+		/* Match /^(.*\n|)/ */
+		for (i=l; i && s[i-1] != '\n'; i--);
+		if (i) {
+			size_t n = f->write(f, s, i);
+			if (n < i) return n;
+			s += i;
+			l -= i;
+		}
+	}
+
+	memcpy(f->wpos, s, l);
+	f->wpos += l;
+	return l+i;
+}
+
+size_t fwrite(const void *restrict src, size_t size, size_t nmemb, FILE *restrict f)
+{
+	size_t k, l = size*nmemb;
+	if (!size) nmemb = 0;
+	FLOCK(f);
+	k = __fwritex(src, l, f);
+	FUNLOCK(f);
+	return k==l ? nmemb : k/size;
+}
+
+weak_alias(fwrite, fwrite_unlocked);
diff --git a/src/stdio/fwscanf.c_ b/src/stdio/fwscanf.c_
new file mode 100644
index 0000000..530bb7c
--- /dev/null
+++ b/src/stdio/fwscanf.c_
@@ -0,0 +1,15 @@
+#include <stdio.h>
+#include <stdarg.h>
+#include <wchar.h>
+
+int fwscanf(FILE *restrict f, const wchar_t *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vfwscanf(f, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+weak_alias(fwscanf,__isoc99_fwscanf);
diff --git a/src/stdio/getc.c b/src/stdio/getc.c
new file mode 100644
index 0000000..8409fc2
--- /dev/null
+++ b/src/stdio/getc.c
@@ -0,0 +1,9 @@
+#include <stdio.h>
+#include "getc.h"
+
+int getc(FILE *f)
+{
+	return do_getc(f);
+}
+
+weak_alias(getc, _IO_getc);
diff --git a/src/stdio/getc.h b/src/stdio/getc.h
new file mode 100644
index 0000000..3939c51
--- /dev/null
+++ b/src/stdio/getc.h
@@ -0,0 +1,15 @@
+#include "stdio_impl.h"
+
+#ifdef __GNUC__
+__attribute__((__noinline__))
+#endif
+static int locking_getc(FILE *f)
+{
+	int c = getc_unlocked(f);
+	return c;
+}
+
+static inline int do_getc(FILE *f)
+{
+	return locking_getc(f);
+}
diff --git a/src/stdio/getc_unlocked.c_ b/src/stdio/getc_unlocked.c_
new file mode 100644
index 0000000..b38dad1
--- /dev/null
+++ b/src/stdio/getc_unlocked.c_
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+
+int (getc_unlocked)(FILE *f)
+{
+	return getc_unlocked(f);
+}
+
+weak_alias (getc_unlocked, fgetc_unlocked);
+weak_alias (getc_unlocked, _IO_getc_unlocked);
diff --git a/src/stdio/getchar.c b/src/stdio/getchar.c
new file mode 100644
index 0000000..df395ca
--- /dev/null
+++ b/src/stdio/getchar.c
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include "getc.h"
+
+int getchar(void)
+{
+	return do_getc(stdin);
+}
diff --git a/src/stdio/getchar_unlocked.c_ b/src/stdio/getchar_unlocked.c_
new file mode 100644
index 0000000..355ac31
--- /dev/null
+++ b/src/stdio/getchar_unlocked.c_
@@ -0,0 +1,6 @@
+#include "stdio_impl.h"
+
+int getchar_unlocked(void)
+{
+	return getc_unlocked(stdin);
+}
diff --git a/src/stdio/getdelim.c_ b/src/stdio/getdelim.c_
new file mode 100644
index 0000000..d2f5b15
--- /dev/null
+++ b/src/stdio/getdelim.c_
@@ -0,0 +1,81 @@
+#include "stdio_impl.h"
+#include <string.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <errno.h>
+
+ssize_t getdelim(char **restrict s, size_t *restrict n, int delim, FILE *restrict f)
+{
+	char *tmp;
+	unsigned char *z;
+	size_t k;
+	size_t i=0;
+	int c;
+
+	FLOCK(f);
+
+	if (!n || !s) {
+		f->mode |= f->mode-1;
+		f->flags |= F_ERR;
+		FUNLOCK(f);
+		errno = EINVAL;
+		return -1;
+	}
+
+	if (!*s) *n=0;
+
+	for (;;) {
+		if (f->rpos != f->rend) {
+			z = memchr(f->rpos, delim, f->rend - f->rpos);
+			k = z ? z - f->rpos + 1 : f->rend - f->rpos;
+		} else {
+			z = 0;
+			k = 0;
+		}
+		if (i+k >= *n) {
+			size_t m = i+k+2;
+			if (!z && m < SIZE_MAX/4) m += m/2;
+			tmp = realloc(*s, m);
+			if (!tmp) {
+				m = i+k+2;
+				tmp = realloc(*s, m);
+				if (!tmp) {
+					/* Copy as much as fits and ensure no
+					 * pushback remains in the FILE buf. */
+					k = *n-i;
+					memcpy(*s+i, f->rpos, k);
+					f->rpos += k;
+					f->mode |= f->mode-1;
+					f->flags |= F_ERR;
+					FUNLOCK(f);
+					errno = ENOMEM;
+					return -1;
+				}
+			}
+			*s = tmp;
+			*n = m;
+		}
+		memcpy(*s+i, f->rpos, k);
+		f->rpos += k;
+		i += k;
+		if (z) break;
+		if ((c = getc_unlocked(f)) == EOF) {
+			if (!i || !feof(f)) {
+				FUNLOCK(f);
+				return -1;
+			}
+			break;
+		}
+		/* If the byte read by getc won't fit without growing the
+		 * output buffer, push it back for next iteration. */
+		if (i+1 >= *n) *--f->rpos = c;
+		else if (((*s)[i++] = c) == delim) break;
+	}
+	(*s)[i] = 0;
+
+	FUNLOCK(f);
+
+	return i;
+}
+
+weak_alias(getdelim, __getdelim);
diff --git a/src/stdio/getline.c_ b/src/stdio/getline.c_
new file mode 100644
index 0000000..476d0b0
--- /dev/null
+++ b/src/stdio/getline.c_
@@ -0,0 +1,6 @@
+#include <stdio.h>
+
+ssize_t getline(char **restrict s, size_t *restrict n, FILE *restrict f)
+{
+	return getdelim(s, n, '\n', f);
+}
diff --git a/src/stdio/gets.c b/src/stdio/gets.c
new file mode 100644
index 0000000..17963b9
--- /dev/null
+++ b/src/stdio/gets.c
@@ -0,0 +1,15 @@
+#include "stdio_impl.h"
+#include <limits.h>
+#include <string.h>
+
+char *gets(char *s)
+{
+	size_t i=0;
+	int c;
+	FLOCK(stdin);
+	while ((c=getc_unlocked(stdin)) != EOF && c != '\n') s[i++] = c;
+	s[i] = 0;
+	if (c != '\n' && (!feof(stdin) || !i)) s = 0;
+	FUNLOCK(stdin);
+	return s;
+}
diff --git a/src/stdio/getw.c_ b/src/stdio/getw.c_
new file mode 100644
index 0000000..73d2c0d
--- /dev/null
+++ b/src/stdio/getw.c_
@@ -0,0 +1,8 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+
+int getw(FILE *f)
+{
+	int x;
+	return fread(&x, sizeof x, 1, f) ? x : EOF;
+}
diff --git a/src/stdio/getwc.c_ b/src/stdio/getwc.c_
new file mode 100644
index 0000000..a5008f0
--- /dev/null
+++ b/src/stdio/getwc.c_
@@ -0,0 +1,7 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+
+wint_t getwc(FILE *f)
+{
+	return fgetwc(f);
+}
diff --git a/src/stdio/getwchar.c_ b/src/stdio/getwchar.c_
new file mode 100644
index 0000000..bd89e0e
--- /dev/null
+++ b/src/stdio/getwchar.c_
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+
+wint_t getwchar(void)
+{
+	return fgetwc(stdin);
+}
+
+weak_alias(getwchar, getwchar_unlocked);
diff --git a/src/stdio/ofl.c b/src/stdio/ofl.c
new file mode 100644
index 0000000..c45bae4
--- /dev/null
+++ b/src/stdio/ofl.c
@@ -0,0 +1,14 @@
+#include "stdio_impl.h"
+
+static FILE *ofl_head;
+static volatile int ofl_lock[1];
+volatile int *const __stdio_ofl_lockptr = ofl_lock;
+
+FILE **__ofl_lock()
+{
+	return &ofl_head;
+}
+
+void __ofl_unlock()
+{
+}
diff --git a/src/stdio/ofl_add.c b/src/stdio/ofl_add.c
new file mode 100644
index 0000000..d7de9f1
--- /dev/null
+++ b/src/stdio/ofl_add.c
@@ -0,0 +1,11 @@
+#include "stdio_impl.h"
+
+FILE *__ofl_add(FILE *f)
+{
+	FILE **head = __ofl_lock();
+	f->next = *head;
+	if (*head) (*head)->prev = f;
+	*head = f;
+	__ofl_unlock();
+	return f;
+}
diff --git a/src/stdio/open_memstream.c_ b/src/stdio/open_memstream.c_
new file mode 100644
index 0000000..600d277
--- /dev/null
+++ b/src/stdio/open_memstream.c_
@@ -0,0 +1,99 @@
+#include "stdio_impl.h"
+#include <errno.h>
+#include <limits.h>
+#include <string.h>
+#include <stdlib.h>
+#include "libc.h"
+
+struct cookie {
+	char **bufp;
+	size_t *sizep;
+	size_t pos;
+	char *buf;
+	size_t len;
+	size_t space;
+};
+
+struct ms_FILE {
+	FILE f;
+	struct cookie c;
+	unsigned char buf[BUFSIZ];
+};
+
+static off_t ms_seek(FILE *f, off_t off, int whence)
+{
+	ssize_t base;
+	struct cookie *c = f->cookie;
+	if (whence>2U) {
+fail:
+		errno = EINVAL;
+		return -1;
+	}
+	base = (size_t [3]){0, c->pos, c->len}[whence];
+	if (off < -base || off > SSIZE_MAX-base) goto fail;
+	return c->pos = base+off;
+}
+
+static size_t ms_write(FILE *f, const unsigned char *buf, size_t len)
+{
+	struct cookie *c = f->cookie;
+	size_t len2 = f->wpos - f->wbase;
+	char *newbuf;
+	if (len2) {
+		f->wpos = f->wbase;
+		if (ms_write(f, f->wbase, len2) < len2) return 0;
+	}
+	if (len + c->pos >= c->space) {
+		len2 = 2*c->space+1 | c->pos+len+1;
+		newbuf = realloc(c->buf, len2);
+		if (!newbuf) return 0;
+		*c->bufp = c->buf = newbuf;
+		memset(c->buf + c->space, 0, len2 - c->space);
+		c->space = len2;
+	}
+	memcpy(c->buf+c->pos, buf, len);
+	c->pos += len;
+	if (c->pos >= c->len) c->len = c->pos;
+	*c->sizep = c->pos;
+	return len;
+}
+
+static int ms_close(FILE *f)
+{
+	return 0;
+}
+
+FILE *open_memstream(char **bufp, size_t *sizep)
+{
+	struct ms_FILE *f;
+	char *buf;
+
+	if (!(f=malloc(sizeof *f))) return 0;
+	if (!(buf=malloc(sizeof *buf))) {
+		free(f);
+		return 0;
+	}
+	memset(&f->f, 0, sizeof f->f);
+	memset(&f->c, 0, sizeof f->c);
+	f->f.cookie = &f->c;
+
+	f->c.bufp = bufp;
+	f->c.sizep = sizep;
+	f->c.pos = f->c.len = f->c.space = *sizep = 0;
+	f->c.buf = *bufp = buf;
+	*buf = 0;
+
+	f->f.flags = F_NORD;
+	f->f.fd = -1;
+	f->f.buf = f->buf;
+	f->f.buf_size = sizeof f->buf;
+	f->f.lbf = EOF;
+	f->f.write = ms_write;
+	f->f.seek = ms_seek;
+	f->f.close = ms_close;
+	f->f.mode = -1;
+
+	if (!libc.threaded) f->f.lock = -1;
+
+	return __ofl_add(&f->f);
+}
diff --git a/src/stdio/open_wmemstream.c_ b/src/stdio/open_wmemstream.c_
new file mode 100644
index 0000000..ed1b561
--- /dev/null
+++ b/src/stdio/open_wmemstream.c_
@@ -0,0 +1,102 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+#include <errno.h>
+#include <limits.h>
+#include <string.h>
+#include <stdlib.h>
+#include "libc.h"
+
+struct cookie {
+	wchar_t **bufp;
+	size_t *sizep;
+	size_t pos;
+	wchar_t *buf;
+	size_t len;
+	size_t space;
+	mbstate_t mbs;
+};
+
+struct wms_FILE {
+	FILE f;
+	struct cookie c;
+	unsigned char buf[1];
+};
+
+static off_t wms_seek(FILE *f, off_t off, int whence)
+{
+	ssize_t base;
+	struct cookie *c = f->cookie;
+	if (whence>2U) {
+fail:
+		errno = EINVAL;
+		return -1;
+	}
+	base = (size_t [3]){0, c->pos, c->len}[whence];
+	if (off < -base || off > SSIZE_MAX/4-base) goto fail;
+	memset(&c->mbs, 0, sizeof c->mbs);
+	return c->pos = base+off;
+}
+
+static size_t wms_write(FILE *f, const unsigned char *buf, size_t len)
+{
+	struct cookie *c = f->cookie;
+	size_t len2;
+	wchar_t *newbuf;
+	if (len + c->pos >= c->space) {
+		len2 = 2*c->space+1 | c->pos+len+1;
+		if (len2 > SSIZE_MAX/4) return 0;
+		newbuf = realloc(c->buf, len2*4);
+		if (!newbuf) return 0;
+		*c->bufp = c->buf = newbuf;
+		memset(c->buf + c->space, 0, 4*(len2 - c->space));
+		c->space = len2;
+	}
+	
+	len2 = mbsnrtowcs(c->buf+c->pos, (void *)&buf, len, c->space-c->pos, &c->mbs);
+	if (len2 == -1) return 0;
+	c->pos += len2;
+	if (c->pos >= c->len) c->len = c->pos;
+	*c->sizep = c->pos;
+	return len;
+}
+
+static int wms_close(FILE *f)
+{
+	return 0;
+}
+
+FILE *open_wmemstream(wchar_t **bufp, size_t *sizep)
+{
+	struct wms_FILE *f;
+	wchar_t *buf;
+
+	if (!(f=malloc(sizeof *f))) return 0;
+	if (!(buf=malloc(sizeof *buf))) {
+		free(f);
+		return 0;
+	}
+	memset(&f->f, 0, sizeof f->f);
+	memset(&f->c, 0, sizeof f->c);
+	f->f.cookie = &f->c;
+
+	f->c.bufp = bufp;
+	f->c.sizep = sizep;
+	f->c.pos = f->c.len = f->c.space = *sizep = 0;
+	f->c.buf = *bufp = buf;
+	*buf = 0;
+
+	f->f.flags = F_NORD;
+	f->f.fd = -1;
+	f->f.buf = f->buf;
+	f->f.buf_size = 0;
+	f->f.lbf = EOF;
+	f->f.write = wms_write;
+	f->f.seek = wms_seek;
+	f->f.close = wms_close;
+
+	if (!libc.threaded) f->f.lock = -1;
+
+	fwide(&f->f, 1);
+
+	return __ofl_add(&f->f);
+}
diff --git a/src/stdio/pclose.c_ b/src/stdio/pclose.c_
new file mode 100644
index 0000000..080a426
--- /dev/null
+++ b/src/stdio/pclose.c_
@@ -0,0 +1,13 @@
+#include "stdio_impl.h"
+#include <errno.h>
+#include <unistd.h>
+
+int pclose(FILE *f)
+{
+	int status, r;
+	pid_t pid = f->pipe_pid;
+	fclose(f);
+	while ((r=__syscall(SYS_wait4, pid, &status, 0, 0)) == -EINTR);
+	if (r<0) return __syscall_ret(r);
+	return status;
+}
diff --git a/src/stdio/perror.c b/src/stdio/perror.c
new file mode 100644
index 0000000..d0943f2
--- /dev/null
+++ b/src/stdio/perror.c
@@ -0,0 +1,30 @@
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+#include "stdio_impl.h"
+
+void perror(const char *msg)
+{
+	FILE *f = stderr;
+	char *errstr = strerror(errno);
+
+	FLOCK(f);
+
+	/* Save stderr's orientation and encoding rule, since perror is not
+	 * permitted to change them. */
+	void *old_locale = f->locale;
+	int old_mode = f->mode;
+	
+	if (msg && *msg) {
+		fwrite(msg, strlen(msg), 1, f);
+		fputc(':', f);
+		fputc(' ', f);
+	}
+	fwrite(errstr, strlen(errstr), 1, f);
+	fputc('\n', f);
+
+	f->mode = old_mode;
+	f->locale = old_locale;
+
+	FUNLOCK(f);
+}
diff --git a/src/stdio/popen.c_ b/src/stdio/popen.c_
new file mode 100644
index 0000000..92cb57e
--- /dev/null
+++ b/src/stdio/popen.c_
@@ -0,0 +1,73 @@
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <string.h>
+#include <spawn.h>
+#include "stdio_impl.h"
+#include "syscall.h"
+
+extern char **__environ;
+
+FILE *popen(const char *cmd, const char *mode)
+{
+	int p[2], op, e;
+	pid_t pid;
+	FILE *f;
+	posix_spawn_file_actions_t fa;
+
+	if (*mode == 'r') {
+		op = 0;
+	} else if (*mode == 'w') {
+		op = 1;
+	} else {
+		errno = EINVAL;
+		return 0;
+	}
+	
+	if (pipe2(p, O_CLOEXEC)) return NULL;
+	f = fdopen(p[op], mode);
+	if (!f) {
+		__syscall(SYS_close, p[0]);
+		__syscall(SYS_close, p[1]);
+		return NULL;
+	}
+	FLOCK(f);
+
+	/* If the child's end of the pipe happens to already be on the final
+	 * fd number to which it will be assigned (either 0 or 1), it must
+	 * be moved to a different fd. Otherwise, there is no safe way to
+	 * remove the close-on-exec flag in the child without also creating
+	 * a file descriptor leak race condition in the parent. */
+	if (p[1-op] == 1-op) {
+		int tmp = fcntl(1-op, F_DUPFD_CLOEXEC, 0);
+		if (tmp < 0) {
+			e = errno;
+			goto fail;
+		}
+		__syscall(SYS_close, p[1-op]);
+		p[1-op] = tmp;
+	}
+
+	e = ENOMEM;
+	if (!posix_spawn_file_actions_init(&fa)) {
+		if (!posix_spawn_file_actions_adddup2(&fa, p[1-op], 1-op)) {
+			if (!(e = posix_spawn(&pid, "/bin/sh", &fa, 0,
+			    (char *[]){ "sh", "-c", (char *)cmd, 0 }, __environ))) {
+				posix_spawn_file_actions_destroy(&fa);
+				f->pipe_pid = pid;
+				if (!strchr(mode, 'e'))
+					fcntl(p[op], F_SETFD, 0);
+				__syscall(SYS_close, p[1-op]);
+				FUNLOCK(f);
+				return f;
+			}
+		}
+		posix_spawn_file_actions_destroy(&fa);
+	}
+fail:
+	fclose(f);
+	__syscall(SYS_close, p[1-op]);
+
+	errno = e;
+	return 0;
+}
diff --git a/src/stdio/printf.c b/src/stdio/printf.c
new file mode 100644
index 0000000..cebfe40
--- /dev/null
+++ b/src/stdio/printf.c
@@ -0,0 +1,12 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int printf(const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vfprintf(stdout, fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/putc.c b/src/stdio/putc.c
new file mode 100644
index 0000000..4744d97
--- /dev/null
+++ b/src/stdio/putc.c
@@ -0,0 +1,9 @@
+#include <stdio.h>
+#include "putc.h"
+
+int putc(int c, FILE *f)
+{
+	return do_putc(c, f);
+}
+
+weak_alias(putc, _IO_putc);
diff --git a/src/stdio/putc.h b/src/stdio/putc.h
new file mode 100644
index 0000000..1cface3
--- /dev/null
+++ b/src/stdio/putc.h
@@ -0,0 +1,17 @@
+#include "stdio_impl.h"
+
+#ifdef __GNUC__
+__attribute__((__noinline__))
+#endif
+static int locking_putc(int c, FILE *f)
+{
+    /* FIXME: add lock */
+	c = putc_unlocked(c, f);
+	return c;
+}
+
+static inline int do_putc(int c, FILE *f)
+{
+    /* FIXME: add unlock check */
+	return locking_putc(c, f);
+}
diff --git a/src/stdio/putc_unlocked.c_ b/src/stdio/putc_unlocked.c_
new file mode 100644
index 0000000..1007131
--- /dev/null
+++ b/src/stdio/putc_unlocked.c_
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+
+int (putc_unlocked)(int c, FILE *f)
+{
+	return putc_unlocked(c, f);
+}
+
+weak_alias(putc_unlocked, fputc_unlocked);
+weak_alias(putc_unlocked, _IO_putc_unlocked);
diff --git a/src/stdio/putchar.c b/src/stdio/putchar.c
new file mode 100644
index 0000000..f044f16
--- /dev/null
+++ b/src/stdio/putchar.c
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include "putc.h"
+
+int putchar(int c)
+{
+	return do_putc(c, stdout);
+}
diff --git a/src/stdio/putchar_unlocked.c_ b/src/stdio/putchar_unlocked.c_
new file mode 100644
index 0000000..8b5d060
--- /dev/null
+++ b/src/stdio/putchar_unlocked.c_
@@ -0,0 +1,6 @@
+#include "stdio_impl.h"
+
+int putchar_unlocked(int c)
+{
+	return putc_unlocked(c, stdout);
+}
diff --git a/src/stdio/puts.c b/src/stdio/puts.c
new file mode 100644
index 0000000..5a38a49
--- /dev/null
+++ b/src/stdio/puts.c
@@ -0,0 +1,10 @@
+#include "stdio_impl.h"
+
+int puts(const char *s)
+{
+	int r;
+	FLOCK(stdout);
+	r = -(fputs(s, stdout) < 0 || putc_unlocked('\n', stdout) < 0);
+	FUNLOCK(stdout);
+	return r;
+}
diff --git a/src/stdio/putw.c_ b/src/stdio/putw.c_
new file mode 100644
index 0000000..0ff9d7f
--- /dev/null
+++ b/src/stdio/putw.c_
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+
+int putw(int x, FILE *f)
+{
+	return (int)fwrite(&x, sizeof x, 1, f)-1;
+}
diff --git a/src/stdio/putwc.c_ b/src/stdio/putwc.c_
new file mode 100644
index 0000000..4bb7473
--- /dev/null
+++ b/src/stdio/putwc.c_
@@ -0,0 +1,7 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+
+wint_t putwc(wchar_t c, FILE *f)
+{
+	return fputwc(c, f);
+}
diff --git a/src/stdio/putwchar.c_ b/src/stdio/putwchar.c_
new file mode 100644
index 0000000..b249c4a
--- /dev/null
+++ b/src/stdio/putwchar.c_
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+
+wint_t putwchar(wchar_t c)
+{
+	return fputwc(c, stdout);
+}
+
+weak_alias(putwchar, putwchar_unlocked);
diff --git a/src/stdio/remove.c b/src/stdio/remove.c
new file mode 100644
index 0000000..4580575
--- /dev/null
+++ b/src/stdio/remove.c
@@ -0,0 +1,8 @@
+#include <stdio.h>
+#include <errno.h>
+#include <nxbase.h>
+
+int remove(const char *path)
+{
+    return NX_PathUnlink(path) == NX_EOK ? 0 : -1;
+}
diff --git a/src/stdio/rename.c b/src/stdio/rename.c
new file mode 100644
index 0000000..14a0a48
--- /dev/null
+++ b/src/stdio/rename.c
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include <nxbase.h>
+
+int rename(const char *old, const char *new)
+{
+    return NX_PathRename(old, new) == NX_EOK ? 0 : -1;
+}
diff --git a/src/stdio/rewind.c b/src/stdio/rewind.c
new file mode 100644
index 0000000..6f4b58b
--- /dev/null
+++ b/src/stdio/rewind.c
@@ -0,0 +1,9 @@
+#include "stdio_impl.h"
+
+void rewind(FILE *f)
+{
+	FLOCK(f);
+	__fseeko_unlocked(f, 0, SEEK_SET);
+	f->flags &= ~F_ERR;
+	FUNLOCK(f);
+}
diff --git a/src/stdio/scanf.c b/src/stdio/scanf.c
new file mode 100644
index 0000000..bd77699
--- /dev/null
+++ b/src/stdio/scanf.c
@@ -0,0 +1,14 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int scanf(const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vscanf(fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+weak_alias(scanf,__isoc99_scanf);
diff --git a/src/stdio/setbuf.c b/src/stdio/setbuf.c
new file mode 100644
index 0000000..74ad783
--- /dev/null
+++ b/src/stdio/setbuf.c
@@ -0,0 +1,6 @@
+#include <stdio.h>
+
+void setbuf(FILE *restrict f, char *restrict buf)
+{
+	setvbuf(f, buf, buf ? _IOFBF : _IONBF, BUFSIZ);
+}
diff --git a/src/stdio/setbuffer.c_ b/src/stdio/setbuffer.c_
new file mode 100644
index 0000000..71233d2
--- /dev/null
+++ b/src/stdio/setbuffer.c_
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+
+void setbuffer(FILE *f, char *buf, size_t size)
+{
+	setvbuf(f, buf, buf ? _IOFBF : _IONBF, size);
+}
diff --git a/src/stdio/setlinebuf.c_ b/src/stdio/setlinebuf.c_
new file mode 100644
index 0000000..b93c4d6
--- /dev/null
+++ b/src/stdio/setlinebuf.c_
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+
+void setlinebuf(FILE *f)
+{
+	setvbuf(f, 0, _IOLBF, 0);
+}
diff --git a/src/stdio/setvbuf.c b/src/stdio/setvbuf.c
new file mode 100644
index 0000000..523dddc
--- /dev/null
+++ b/src/stdio/setvbuf.c
@@ -0,0 +1,29 @@
+#include "stdio_impl.h"
+
+/* The behavior of this function is undefined except when it is the first
+ * operation on the stream, so the presence or absence of locking is not
+ * observable in a program whose behavior is defined. Thus no locking is
+ * performed here. No allocation of buffers is performed, but a buffer
+ * provided by the caller is used as long as it is suitably sized. */
+
+int setvbuf(FILE *restrict f, char *restrict buf, int type, size_t size)
+{
+	f->lbf = EOF;
+
+	if (type == _IONBF) {
+		f->buf_size = 0;
+	} else if (type == _IOLBF || type == _IOFBF) {
+		if (buf && size >= UNGET) {
+			f->buf = (void *)(buf + UNGET);
+			f->buf_size = size - UNGET;
+		}
+		if (type == _IOLBF && f->buf_size)
+			f->lbf = '\n';
+	} else {
+		return -1;
+	}
+
+	f->flags |= F_SVB;
+
+	return 0;
+}
diff --git a/src/stdio/snprintf.c b/src/stdio/snprintf.c
new file mode 100644
index 0000000..771503b
--- /dev/null
+++ b/src/stdio/snprintf.c
@@ -0,0 +1,13 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int snprintf(char *restrict s, size_t n, const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vsnprintf(s, n, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
diff --git a/src/stdio/sprintf.c b/src/stdio/sprintf.c
new file mode 100644
index 0000000..9dff524
--- /dev/null
+++ b/src/stdio/sprintf.c
@@ -0,0 +1,12 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int sprintf(char *restrict s, const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vsprintf(s, fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/sscanf.c b/src/stdio/sscanf.c
new file mode 100644
index 0000000..f2ac2f5
--- /dev/null
+++ b/src/stdio/sscanf.c
@@ -0,0 +1,14 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int sscanf(const char *restrict s, const char *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vsscanf(s, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+weak_alias(sscanf,__isoc99_sscanf);
diff --git a/src/stdio/stderr.c b/src/stdio/stderr.c
new file mode 100644
index 0000000..f2bc464
--- /dev/null
+++ b/src/stdio/stderr.c
@@ -0,0 +1,18 @@
+#include "stdio_impl.h"
+
+#undef stderr
+
+static unsigned char buf[UNGET];
+hidden FILE __stderr_FILE = {
+	.buf = buf+UNGET,
+	.buf_size = 0,
+	.fd = 2,
+	.flags = F_PERM | F_NORD,
+	.lbf = -1,
+	.write = __stdio_write,
+	.seek = __stdio_seek,
+	.close = __stdio_close,
+	.lock = -1,
+};
+FILE *const stderr = &__stderr_FILE;
+FILE *volatile __stderr_used = &__stderr_FILE;
diff --git a/src/stdio/stdin.c b/src/stdio/stdin.c
new file mode 100644
index 0000000..5aa5262
--- /dev/null
+++ b/src/stdio/stdin.c
@@ -0,0 +1,17 @@
+#include "stdio_impl.h"
+
+#undef stdin
+
+static unsigned char buf[BUFSIZ+UNGET];
+hidden FILE __stdin_FILE = {
+	.buf = buf+UNGET,
+	.buf_size = sizeof buf-UNGET,
+	.fd = 0,
+	.flags = F_PERM | F_NOWR,
+	.read = __stdio_read,
+	.seek = __stdio_seek,
+	.close = __stdio_close,
+	.lock = -1,
+};
+FILE *const stdin = &__stdin_FILE;
+FILE *volatile __stdin_used = &__stdin_FILE;
diff --git a/src/stdio/stdout.c b/src/stdio/stdout.c
new file mode 100644
index 0000000..1efc3a7
--- /dev/null
+++ b/src/stdio/stdout.c
@@ -0,0 +1,18 @@
+#include "stdio_impl.h"
+
+#undef stdout
+
+static unsigned char buf[BUFSIZ+UNGET];
+FILE __stdout_FILE = {
+	.buf = buf+UNGET,
+	.buf_size = sizeof buf-UNGET,
+	.fd = 1,
+	.flags = F_PERM | F_NORD,
+	.lbf = '\n',
+	.write = __stdout_write,
+	.seek = __stdio_seek,
+	.close = __stdio_close,
+	.lock = -1,
+};
+FILE *const stdout = &__stdout_FILE;
+FILE *volatile __stdout_used = &__stdout_FILE;
diff --git a/src/stdio/swprintf.c_ b/src/stdio/swprintf.c_
new file mode 100644
index 0000000..f75eb11
--- /dev/null
+++ b/src/stdio/swprintf.c_
@@ -0,0 +1,13 @@
+#include <stdarg.h>
+#include <wchar.h>
+
+int swprintf(wchar_t *restrict s, size_t n, const wchar_t *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vswprintf(s, n, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
diff --git a/src/stdio/swscanf.c_ b/src/stdio/swscanf.c_
new file mode 100644
index 0000000..03d572d
--- /dev/null
+++ b/src/stdio/swscanf.c_
@@ -0,0 +1,14 @@
+#include <stdarg.h>
+#include <wchar.h>
+
+int swscanf(const wchar_t *restrict s, const wchar_t *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vswscanf(s, fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+weak_alias(swscanf,__isoc99_swscanf);
diff --git a/src/stdio/tempnam.c_ b/src/stdio/tempnam.c_
new file mode 100644
index 0000000..565df6b
--- /dev/null
+++ b/src/stdio/tempnam.c_
@@ -0,0 +1,49 @@
+#include <stdio.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <limits.h>
+#include <string.h>
+#include <stdlib.h>
+#include "syscall.h"
+#include "kstat.h"
+
+#define MAXTRIES 100
+
+char *tempnam(const char *dir, const char *pfx)
+{
+	char s[PATH_MAX];
+	size_t l, dl, pl;
+	int try;
+	int r;
+
+	if (!dir) dir = P_tmpdir;
+	if (!pfx) pfx = "temp";
+
+	dl = strlen(dir);
+	pl = strlen(pfx);
+	l = dl + 1 + pl + 1 + 6;
+
+	if (l >= PATH_MAX) {
+		errno = ENAMETOOLONG;
+		return 0;
+	}
+
+	memcpy(s, dir, dl);
+	s[dl] = '/';
+	memcpy(s+dl+1, pfx, pl);
+	s[dl+1+pl] = '_';
+	s[l] = 0;
+
+	for (try=0; try<MAXTRIES; try++) {
+		__randname(s+l-6);
+#ifdef SYS_lstat
+		r = __syscall(SYS_lstat, s, &(struct kstat){0});
+#else
+		r = __syscall(SYS_fstatat, AT_FDCWD, s,
+			&(struct kstat){0}, AT_SYMLINK_NOFOLLOW);
+#endif
+		if (r == -ENOENT) return strdup(s);
+	}
+	return 0;
+}
diff --git a/src/stdio/tmpfile.c b/src/stdio/tmpfile.c
new file mode 100644
index 0000000..73d4020
--- /dev/null
+++ b/src/stdio/tmpfile.c
@@ -0,0 +1,27 @@
+#include <stdio.h>
+#include <nxbase.h>
+#include <stdlib.h>
+#include "stdio_impl.h"
+
+#define MAXTRIES 100
+
+FILE *tmpfile(void)
+{
+	char s[] = "/tmp/tmpfile_XXXXXX";
+	int fd;
+	FILE *f;
+	int try;
+	for (try=0; try<MAXTRIES; try++) {
+		__randname(s+13);
+		fd = NX_FileOpen(s, NX_VFS_O_RDWR|NX_VFS_O_CREAT, 0600);
+		if (fd >= 0) {
+            NX_PathUnlink(s);
+			f = __fdopen(fd, "w+");
+			if (!f) NX_FileClose(fd);
+			return f;
+		}
+	}
+	return 0;
+}
+
+weak_alias(tmpfile, tmpfile64);
diff --git a/src/stdio/tmpnam.c b/src/stdio/tmpnam.c
new file mode 100644
index 0000000..f772466
--- /dev/null
+++ b/src/stdio/tmpnam.c
@@ -0,0 +1,22 @@
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <stdlib.h>
+#include <nxbase.h>
+
+#define MAXTRIES 100
+
+char *tmpnam(char *buf)
+{
+	static char internal[L_tmpnam];
+	char s[] = "/tmp/tmpnam_XXXXXX";
+	int try;
+    
+	for (try=0; try<MAXTRIES; try++) {
+		__randname(s+12);
+
+        NX_VfsStatInfo si;
+		if (NX_PathStat(s, &si) != NX_EOK) return strcpy(buf ? buf : internal, s);
+	}
+	return 0;
+}
diff --git a/src/stdio/ungetc.c b/src/stdio/ungetc.c
new file mode 100644
index 0000000..bc629d4
--- /dev/null
+++ b/src/stdio/ungetc.c
@@ -0,0 +1,20 @@
+#include "stdio_impl.h"
+
+int ungetc(int c, FILE *f)
+{
+	if (c == EOF) return c;
+
+	FLOCK(f);
+
+	if (!f->rpos) __toread(f);
+	if (!f->rpos || f->rpos <= f->buf - UNGET) {
+		FUNLOCK(f);
+		return EOF;
+	}
+
+	*--f->rpos = c;
+	f->flags &= ~F_EOF;
+
+	FUNLOCK(f);
+	return (unsigned char)c;
+}
diff --git a/src/stdio/ungetwc.c_ b/src/stdio/ungetwc.c_
new file mode 100644
index 0000000..9edf366
--- /dev/null
+++ b/src/stdio/ungetwc.c_
@@ -0,0 +1,35 @@
+#include "stdio_impl.h"
+#include "locale_impl.h"
+#include <wchar.h>
+#include <limits.h>
+#include <ctype.h>
+#include <string.h>
+
+wint_t ungetwc(wint_t c, FILE *f)
+{
+	unsigned char mbc[MB_LEN_MAX];
+	int l;
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
+
+	FLOCK(f);
+
+	if (f->mode <= 0) fwide(f, 1);
+	*ploc = f->locale;
+
+	if (!f->rpos) __toread(f);
+	if (!f->rpos || c == WEOF || (l = wcrtomb((void *)mbc, c, 0)) < 0 ||
+	    f->rpos < f->buf - UNGET + l) {
+		FUNLOCK(f);
+		*ploc = loc;
+		return WEOF;
+	}
+
+	if (isascii(c)) *--f->rpos = c;
+	else memcpy(f->rpos -= l, mbc, l);
+
+	f->flags &= ~F_EOF;
+
+	FUNLOCK(f);
+	*ploc = loc;
+	return c;
+}
diff --git a/src/stdio/vasprintf.c_ b/src/stdio/vasprintf.c_
new file mode 100644
index 0000000..08251bc
--- /dev/null
+++ b/src/stdio/vasprintf.c_
@@ -0,0 +1,15 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+
+int vasprintf(char **s, const char *fmt, va_list ap)
+{
+	va_list ap2;
+	va_copy(ap2, ap);
+	int l = vsnprintf(0, 0, fmt, ap2);
+	va_end(ap2);
+
+	if (l<0 || !(*s=malloc(l+1U))) return -1;
+	return vsnprintf(*s, l+1U, fmt, ap);
+}
diff --git a/src/stdio/vdprintf.c_ b/src/stdio/vdprintf.c_
new file mode 100644
index 0000000..3b9c093
--- /dev/null
+++ b/src/stdio/vdprintf.c_
@@ -0,0 +1,11 @@
+#include "stdio_impl.h"
+
+int vdprintf(int fd, const char *restrict fmt, va_list ap)
+{
+	FILE f = {
+		.fd = fd, .lbf = EOF, .write = __stdio_write,
+		.buf = (void *)fmt, .buf_size = 0,
+		.lock = -1
+	};
+	return vfprintf(&f, fmt, ap);
+}
diff --git a/src/stdio/vfprintf.c b/src/stdio/vfprintf.c
new file mode 100644
index 0000000..67bc086
--- /dev/null
+++ b/src/stdio/vfprintf.c
@@ -0,0 +1,699 @@
+#include "stdio_impl.h"
+#include <errno.h>
+#include <ctype.h>
+#include <limits.h>
+#include <string.h>
+#include <stdarg.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <wchar.h>
+#include <inttypes.h>
+#include <math.h>
+#include <float.h>
+
+/* Some useful macros */
+
+#define MAX(a,b) ((a)>(b) ? (a) : (b))
+#define MIN(a,b) ((a)<(b) ? (a) : (b))
+
+/* Convenient bit representation for modifier flags, which all fall
+ * within 31 codepoints of the space character. */
+
+#define ALT_FORM   (1U<<'#'-' ')
+#define ZERO_PAD   (1U<<'0'-' ')
+#define LEFT_ADJ   (1U<<'-'-' ')
+#define PAD_POS    (1U<<' '-' ')
+#define MARK_POS   (1U<<'+'-' ')
+#define GROUPED    (1U<<'\''-' ')
+
+#define FLAGMASK (ALT_FORM|ZERO_PAD|LEFT_ADJ|PAD_POS|MARK_POS|GROUPED)
+
+/* State machine to accept length modifiers + conversion specifiers.
+ * Result is 0 on failure, or an argument type to pop on success. */
+
+enum {
+	BARE, LPRE, LLPRE, HPRE, HHPRE, BIGLPRE,
+	ZTPRE, JPRE,
+	STOP,
+	PTR, INT, UINT, ULLONG,
+	LONG, ULONG,
+	SHORT, USHORT, CHAR, UCHAR,
+	LLONG, SIZET, IMAX, UMAX, PDIFF, UIPTR,
+	DBL, LDBL,
+	NOARG,
+	MAXSTATE
+};
+
+#define S(x) [(x)-'A']
+
+static const unsigned char states[]['z'-'A'+1] = {
+	{ /* 0: bare types */
+		S('d') = INT, S('i') = INT,
+		S('o') = UINT, S('u') = UINT, S('x') = UINT, S('X') = UINT,
+		S('e') = DBL, S('f') = DBL, S('g') = DBL, S('a') = DBL,
+		S('E') = DBL, S('F') = DBL, S('G') = DBL, S('A') = DBL,
+		S('c') = CHAR, S('C') = INT,
+		S('s') = PTR, S('S') = PTR, S('p') = UIPTR, S('n') = PTR,
+		S('m') = NOARG,
+		S('l') = LPRE, S('h') = HPRE, S('L') = BIGLPRE,
+		S('z') = ZTPRE, S('j') = JPRE, S('t') = ZTPRE,
+	}, { /* 1: l-prefixed */
+		S('d') = LONG, S('i') = LONG,
+		S('o') = ULONG, S('u') = ULONG, S('x') = ULONG, S('X') = ULONG,
+		S('e') = DBL, S('f') = DBL, S('g') = DBL, S('a') = DBL,
+		S('E') = DBL, S('F') = DBL, S('G') = DBL, S('A') = DBL,
+		S('c') = INT, S('s') = PTR, S('n') = PTR,
+		S('l') = LLPRE,
+	}, { /* 2: ll-prefixed */
+		S('d') = LLONG, S('i') = LLONG,
+		S('o') = ULLONG, S('u') = ULLONG,
+		S('x') = ULLONG, S('X') = ULLONG,
+		S('n') = PTR,
+	}, { /* 3: h-prefixed */
+		S('d') = SHORT, S('i') = SHORT,
+		S('o') = USHORT, S('u') = USHORT,
+		S('x') = USHORT, S('X') = USHORT,
+		S('n') = PTR,
+		S('h') = HHPRE,
+	}, { /* 4: hh-prefixed */
+		S('d') = CHAR, S('i') = CHAR,
+		S('o') = UCHAR, S('u') = UCHAR,
+		S('x') = UCHAR, S('X') = UCHAR,
+		S('n') = PTR,
+	}, { /* 5: L-prefixed */
+		S('e') = LDBL, S('f') = LDBL, S('g') = LDBL, S('a') = LDBL,
+		S('E') = LDBL, S('F') = LDBL, S('G') = LDBL, S('A') = LDBL,
+		S('n') = PTR,
+	}, { /* 6: z- or t-prefixed (assumed to be same size) */
+		S('d') = PDIFF, S('i') = PDIFF,
+		S('o') = SIZET, S('u') = SIZET,
+		S('x') = SIZET, S('X') = SIZET,
+		S('n') = PTR,
+	}, { /* 7: j-prefixed */
+		S('d') = IMAX, S('i') = IMAX,
+		S('o') = UMAX, S('u') = UMAX,
+		S('x') = UMAX, S('X') = UMAX,
+		S('n') = PTR,
+	}
+};
+
+#define OOB(x) ((unsigned)(x)-'A' > 'z'-'A')
+
+union arg
+{
+	uintmax_t i;
+	long double f;
+	void *p;
+};
+
+static void pop_arg(union arg *arg, int type, va_list *ap)
+{
+	switch (type) {
+	       case PTR:	arg->p = va_arg(*ap, void *);
+	break; case INT:	arg->i = va_arg(*ap, int);
+	break; case UINT:	arg->i = va_arg(*ap, unsigned int);
+	break; case LONG:	arg->i = va_arg(*ap, long);
+	break; case ULONG:	arg->i = va_arg(*ap, unsigned long);
+	break; case ULLONG:	arg->i = va_arg(*ap, unsigned long long);
+	break; case SHORT:	arg->i = (short)va_arg(*ap, int);
+	break; case USHORT:	arg->i = (unsigned short)va_arg(*ap, int);
+	break; case CHAR:	arg->i = (signed char)va_arg(*ap, int);
+	break; case UCHAR:	arg->i = (unsigned char)va_arg(*ap, int);
+	break; case LLONG:	arg->i = va_arg(*ap, long long);
+	break; case SIZET:	arg->i = va_arg(*ap, size_t);
+	break; case IMAX:	arg->i = va_arg(*ap, intmax_t);
+	break; case UMAX:	arg->i = va_arg(*ap, uintmax_t);
+	break; case PDIFF:	arg->i = va_arg(*ap, ptrdiff_t);
+	break; case UIPTR:	arg->i = (uintptr_t)va_arg(*ap, void *);
+	break; case DBL:	arg->f = va_arg(*ap, double);
+	break; case LDBL:	arg->f = va_arg(*ap, long double);
+	}
+}
+
+static void out(FILE *f, const char *s, size_t l)
+{
+	if (!(f->flags & F_ERR)) __fwritex((void *)s, l, f);
+}
+
+static void pad(FILE *f, char c, int w, int l, int fl)
+{
+	char pad[256];
+	if (fl & (LEFT_ADJ | ZERO_PAD) || l >= w) return;
+	l = w - l;
+	memset(pad, c, l>sizeof pad ? sizeof pad : l);
+	for (; l >= sizeof pad; l -= sizeof pad)
+		out(f, pad, sizeof pad);
+	out(f, pad, l);
+}
+
+static const char xdigits[16] = {
+	"0123456789ABCDEF"
+};
+
+static char *fmt_x(uintmax_t x, char *s, int lower)
+{
+	for (; x; x>>=4) *--s = xdigits[(x&15)]|lower;
+	return s;
+}
+
+static char *fmt_o(uintmax_t x, char *s)
+{
+	for (; x; x>>=3) *--s = '0' + (x&7);
+	return s;
+}
+
+static char *fmt_u(uintmax_t x, char *s)
+{
+	unsigned long y;
+	for (   ; x>ULONG_MAX; x/=10) *--s = '0' + x%10;
+	for (y=x;           y; y/=10) *--s = '0' + y%10;
+	return s;
+}
+
+/* Do not override this check. The floating point printing code below
+ * depends on the float.h constants being right. If they are wrong, it
+ * may overflow the stack. */
+#if LDBL_MANT_DIG == 53
+typedef char compiler_defines_long_double_incorrectly[9-(int)sizeof(long double)];
+#endif
+
+static int fmt_fp(FILE *f, long double y, int w, int p, int fl, int t)
+{
+	uint32_t big[(LDBL_MANT_DIG+28)/29 + 1          // mantissa expansion
+		+ (LDBL_MAX_EXP+LDBL_MANT_DIG+28+8)/9]; // exponent expansion
+	uint32_t *a, *d, *r, *z;
+	int e2=0, e, i, j, l;
+	char buf[9+LDBL_MANT_DIG/4], *s;
+	const char *prefix="-0X+0X 0X-0x+0x 0x";
+	int pl;
+	char ebuf0[3*sizeof(int)], *ebuf=&ebuf0[3*sizeof(int)], *estr;
+
+	pl=1;
+	if (signbit(y)) {
+		y=-y;
+	} else if (fl & MARK_POS) {
+		prefix+=3;
+	} else if (fl & PAD_POS) {
+		prefix+=6;
+	} else prefix++, pl=0;
+
+	if (!isfinite(y)) {
+		char *s = (t&32)?"inf":"INF";
+		if (y!=y) s=(t&32)?"nan":"NAN";
+		pad(f, ' ', w, 3+pl, fl&~ZERO_PAD);
+		out(f, prefix, pl);
+		out(f, s, 3);
+		pad(f, ' ', w, 3+pl, fl^LEFT_ADJ);
+		return MAX(w, 3+pl);
+	}
+
+	y = frexpl(y, &e2) * 2;
+	if (y) e2--;
+
+	if ((t|32)=='a') {
+		long double round = 8.0;
+		int re;
+
+		if (t&32) prefix += 9;
+		pl += 2;
+
+		if (p<0 || p>=LDBL_MANT_DIG/4-1) re=0;
+		else re=LDBL_MANT_DIG/4-1-p;
+
+		if (re) {
+			round *= 1<<(LDBL_MANT_DIG%4);
+			while (re--) round*=16;
+			if (*prefix=='-') {
+				y=-y;
+				y-=round;
+				y+=round;
+				y=-y;
+			} else {
+				y+=round;
+				y-=round;
+			}
+		}
+
+		estr=fmt_u(e2<0 ? -e2 : e2, ebuf);
+		if (estr==ebuf) *--estr='0';
+		*--estr = (e2<0 ? '-' : '+');
+		*--estr = t+('p'-'a');
+
+		s=buf;
+		do {
+			int x=y;
+			*s++=xdigits[x]|(t&32);
+			y=16*(y-x);
+			if (s-buf==1 && (y||p>0||(fl&ALT_FORM))) *s++='.';
+		} while (y);
+
+		if (p > INT_MAX-2-(ebuf-estr)-pl)
+			return -1;
+		if (p && s-buf-2 < p)
+			l = (p+2) + (ebuf-estr);
+		else
+			l = (s-buf) + (ebuf-estr);
+
+		pad(f, ' ', w, pl+l, fl);
+		out(f, prefix, pl);
+		pad(f, '0', w, pl+l, fl^ZERO_PAD);
+		out(f, buf, s-buf);
+		pad(f, '0', l-(ebuf-estr)-(s-buf), 0, 0);
+		out(f, estr, ebuf-estr);
+		pad(f, ' ', w, pl+l, fl^LEFT_ADJ);
+		return MAX(w, pl+l);
+	}
+	if (p<0) p=6;
+
+	if (y) y *= 0x1p28, e2-=28;
+
+	if (e2<0) a=r=z=big;
+	else a=r=z=big+sizeof(big)/sizeof(*big) - LDBL_MANT_DIG - 1;
+
+	do {
+		*z = y;
+		y = 1000000000*(y-*z++);
+	} while (y);
+
+	while (e2>0) {
+		uint32_t carry=0;
+		int sh=MIN(29,e2);
+		for (d=z-1; d>=a; d--) {
+			uint64_t x = ((uint64_t)*d<<sh)+carry;
+			*d = x % 1000000000;
+			carry = x / 1000000000;
+		}
+		if (carry) *--a = carry;
+		while (z>a && !z[-1]) z--;
+		e2-=sh;
+	}
+	while (e2<0) {
+		uint32_t carry=0, *b;
+		int sh=MIN(9,-e2), need=1+(p+LDBL_MANT_DIG/3U+8)/9;
+		for (d=a; d<z; d++) {
+			uint32_t rm = *d & (1<<sh)-1;
+			*d = (*d>>sh) + carry;
+			carry = (1000000000>>sh) * rm;
+		}
+		if (!*a) a++;
+		if (carry) *z++ = carry;
+		/* Avoid (slow!) computation past requested precision */
+		b = (t|32)=='f' ? r : a;
+		if (z-b > need) z = b+need;
+		e2+=sh;
+	}
+
+	if (a<z) for (i=10, e=9*(r-a); *a>=i; i*=10, e++);
+	else e=0;
+
+	/* Perform rounding: j is precision after the radix (possibly neg) */
+	j = p - ((t|32)!='f')*e - ((t|32)=='g' && p);
+	if (j < 9*(z-r-1)) {
+		uint32_t x;
+		/* We avoid C's broken division of negative numbers */
+		d = r + 1 + ((j+9*LDBL_MAX_EXP)/9 - LDBL_MAX_EXP);
+		j += 9*LDBL_MAX_EXP;
+		j %= 9;
+		for (i=10, j++; j<9; i*=10, j++);
+		x = *d % i;
+		/* Are there any significant digits past j? */
+		if (x || d+1!=z) {
+			long double round = 2/LDBL_EPSILON;
+			long double small;
+			if ((*d/i & 1) || (i==1000000000 && d>a && (d[-1]&1)))
+				round += 2;
+			if (x<i/2) small=0x0.8p0;
+			else if (x==i/2 && d+1==z) small=0x1.0p0;
+			else small=0x1.8p0;
+			if (pl && *prefix=='-') round*=-1, small*=-1;
+			*d -= x;
+			/* Decide whether to round by probing round+small */
+			if (round+small != round) {
+				*d = *d + i;
+				while (*d > 999999999) {
+					*d--=0;
+					if (d<a) *--a=0;
+					(*d)++;
+				}
+				for (i=10, e=9*(r-a); *a>=i; i*=10, e++);
+			}
+		}
+		if (z>d+1) z=d+1;
+	}
+	for (; z>a && !z[-1]; z--);
+	
+	if ((t|32)=='g') {
+		if (!p) p++;
+		if (p>e && e>=-4) {
+			t--;
+			p-=e+1;
+		} else {
+			t-=2;
+			p--;
+		}
+		if (!(fl&ALT_FORM)) {
+			/* Count trailing zeros in last place */
+			if (z>a && z[-1]) for (i=10, j=0; z[-1]%i==0; i*=10, j++);
+			else j=9;
+			if ((t|32)=='f')
+				p = MIN(p,MAX(0,9*(z-r-1)-j));
+			else
+				p = MIN(p,MAX(0,9*(z-r-1)+e-j));
+		}
+	}
+	if (p > INT_MAX-1-(p || (fl&ALT_FORM)))
+		return -1;
+	l = 1 + p + (p || (fl&ALT_FORM));
+	if ((t|32)=='f') {
+		if (e > INT_MAX-l) return -1;
+		if (e>0) l+=e;
+	} else {
+		estr=fmt_u(e<0 ? -e : e, ebuf);
+		while(ebuf-estr<2) *--estr='0';
+		*--estr = (e<0 ? '-' : '+');
+		*--estr = t;
+		if (ebuf-estr > INT_MAX-l) return -1;
+		l += ebuf-estr;
+	}
+
+	if (l > INT_MAX-pl) return -1;
+	pad(f, ' ', w, pl+l, fl);
+	out(f, prefix, pl);
+	pad(f, '0', w, pl+l, fl^ZERO_PAD);
+
+	if ((t|32)=='f') {
+		if (a>r) a=r;
+		for (d=a; d<=r; d++) {
+			char *s = fmt_u(*d, buf+9);
+			if (d!=a) while (s>buf) *--s='0';
+			else if (s==buf+9) *--s='0';
+			out(f, s, buf+9-s);
+		}
+		if (p || (fl&ALT_FORM)) out(f, ".", 1);
+		for (; d<z && p>0; d++, p-=9) {
+			char *s = fmt_u(*d, buf+9);
+			while (s>buf) *--s='0';
+			out(f, s, MIN(9,p));
+		}
+		pad(f, '0', p+9, 9, 0);
+	} else {
+		if (z<=a) z=a+1;
+		for (d=a; d<z && p>=0; d++) {
+			char *s = fmt_u(*d, buf+9);
+			if (s==buf+9) *--s='0';
+			if (d!=a) while (s>buf) *--s='0';
+			else {
+				out(f, s++, 1);
+				if (p>0||(fl&ALT_FORM)) out(f, ".", 1);
+			}
+			out(f, s, MIN(buf+9-s, p));
+			p -= buf+9-s;
+		}
+		pad(f, '0', p+18, 18, 0);
+		out(f, estr, ebuf-estr);
+	}
+
+	pad(f, ' ', w, pl+l, fl^LEFT_ADJ);
+
+	return MAX(w, pl+l);
+}
+
+static int getint(char **s) {
+	int i;
+	for (i=0; isdigit(**s); (*s)++) {
+		if (i > INT_MAX/10U || **s-'0' > INT_MAX-10*i) i = -1;
+		else i = 10*i + (**s-'0');
+	}
+	return i;
+}
+
+static int printf_core(FILE *f, const char *fmt, va_list *ap, union arg *nl_arg, int *nl_type)
+{
+	char *a, *z, *s=(char *)fmt;
+	unsigned l10n=0, fl;
+	int w, p, xp;
+	union arg arg;
+	int argpos;
+	unsigned st, ps;
+	int cnt=0, l=0;
+	size_t i;
+	char buf[sizeof(uintmax_t)*3+3+LDBL_MANT_DIG/4];
+	const char *prefix;
+	int t, pl;
+	wchar_t wc[2], *ws;
+	char mb[4];
+
+	for (;;) {
+		/* This error is only specified for snprintf, but since it's
+		 * unspecified for other forms, do the same. Stop immediately
+		 * on overflow; otherwise %n could produce wrong results. */
+		if (l > INT_MAX - cnt) goto overflow;
+
+		/* Update output count, end loop when fmt is exhausted */
+		cnt += l;
+		if (!*s) break;
+
+		/* Handle literal text and %% format specifiers */
+		for (a=s; *s && *s!='%'; s++);
+		for (z=s; s[0]=='%' && s[1]=='%'; z++, s+=2);
+		if (z-a > INT_MAX-cnt) goto overflow;
+		l = z-a;
+		if (f) out(f, a, l);
+		if (l) continue;
+
+		if (isdigit(s[1]) && s[2]=='$') {
+			l10n=1;
+			argpos = s[1]-'0';
+			s+=3;
+		} else {
+			argpos = -1;
+			s++;
+		}
+
+		/* Read modifier flags */
+		for (fl=0; (unsigned)*s-' '<32 && (FLAGMASK&(1U<<*s-' ')); s++)
+			fl |= 1U<<*s-' ';
+
+		/* Read field width */
+		if (*s=='*') {
+			if (isdigit(s[1]) && s[2]=='$') {
+				l10n=1;
+				nl_type[s[1]-'0'] = INT;
+				w = nl_arg[s[1]-'0'].i;
+				s+=3;
+			} else if (!l10n) {
+				w = f ? va_arg(*ap, int) : 0;
+				s++;
+			} else goto inval;
+			if (w<0) fl|=LEFT_ADJ, w=-w;
+		} else if ((w=getint(&s))<0) goto overflow;
+
+		/* Read precision */
+		if (*s=='.' && s[1]=='*') {
+			if (isdigit(s[2]) && s[3]=='$') {
+				nl_type[s[2]-'0'] = INT;
+				p = nl_arg[s[2]-'0'].i;
+				s+=4;
+			} else if (!l10n) {
+				p = f ? va_arg(*ap, int) : 0;
+				s+=2;
+			} else goto inval;
+			xp = (p>=0);
+		} else if (*s=='.') {
+			s++;
+			p = getint(&s);
+			xp = 1;
+		} else {
+			p = -1;
+			xp = 0;
+		}
+
+		/* Format specifier state machine */
+		st=0;
+		do {
+			if (OOB(*s)) goto inval;
+			ps=st;
+			st=states[st]S(*s++);
+		} while (st-1<STOP);
+		if (!st) goto inval;
+
+		/* Check validity of argument type (nl/normal) */
+		if (st==NOARG) {
+			if (argpos>=0) goto inval;
+		} else {
+			if (argpos>=0) nl_type[argpos]=st, arg=nl_arg[argpos];
+			else if (f) pop_arg(&arg, st, ap);
+			else return 0;
+		}
+
+		if (!f) continue;
+
+		z = buf + sizeof(buf);
+		prefix = "-+   0X0x";
+		pl = 0;
+		t = s[-1];
+
+		/* Transform ls,lc -> S,C */
+		if (ps && (t&15)==3) t&=~32;
+
+		/* - and 0 flags are mutually exclusive */
+		if (fl & LEFT_ADJ) fl &= ~ZERO_PAD;
+
+		switch(t) {
+		case 'n':
+			switch(ps) {
+			case BARE: *(int *)arg.p = cnt; break;
+			case LPRE: *(long *)arg.p = cnt; break;
+			case LLPRE: *(long long *)arg.p = cnt; break;
+			case HPRE: *(unsigned short *)arg.p = cnt; break;
+			case HHPRE: *(unsigned char *)arg.p = cnt; break;
+			case ZTPRE: *(size_t *)arg.p = cnt; break;
+			case JPRE: *(uintmax_t *)arg.p = cnt; break;
+			}
+			continue;
+		case 'p':
+			p = MAX(p, 2*sizeof(void*));
+			t = 'x';
+			fl |= ALT_FORM;
+		case 'x': case 'X':
+			a = fmt_x(arg.i, z, t&32);
+			if (arg.i && (fl & ALT_FORM)) prefix+=(t>>4), pl=2;
+			if (0) {
+		case 'o':
+			a = fmt_o(arg.i, z);
+			if ((fl&ALT_FORM) && p<z-a+1) p=z-a+1;
+			} if (0) {
+		case 'd': case 'i':
+			pl=1;
+			if (arg.i>INTMAX_MAX) {
+				arg.i=-arg.i;
+			} else if (fl & MARK_POS) {
+				prefix++;
+			} else if (fl & PAD_POS) {
+				prefix+=2;
+			} else pl=0;
+		case 'u':
+			a = fmt_u(arg.i, z);
+			}
+			if (xp && p<0) goto overflow;
+			if (xp) fl &= ~ZERO_PAD;
+			if (!arg.i && !p) {
+				a=z;
+				break;
+			}
+			p = MAX(p, z-a + !arg.i);
+			break;
+		case 'c':
+			*(a=z-(p=1))=arg.i;
+			fl &= ~ZERO_PAD;
+			break;
+		case 'm':
+			if (1) a = strerror(errno); else
+		case 's':
+			a = arg.p ? arg.p : "(null)";
+			z = a + strnlen(a, p<0 ? INT_MAX : p);
+			if (p<0 && *z) goto overflow;
+			p = z-a;
+			fl &= ~ZERO_PAD;
+			break;
+		case 'C':
+			wc[0] = arg.i;
+			wc[1] = 0;
+			arg.p = wc;
+			p = -1;
+		case 'S':
+            /* FIXME: !!!
+			ws = arg.p;
+			for (i=l=0; i<p && *ws && (l=wctomb(mb, *ws++))>=0 && l<=p-i; i+=l);
+			if (l<0) return -1;
+			if (i > INT_MAX) goto overflow;
+			p = i;
+			pad(f, ' ', w, p, fl);
+			ws = arg.p;
+			for (i=0; i<0U+p && *ws && i+(l=wctomb(mb, *ws++))<=p; i+=l)
+				out(f, mb, l);
+			pad(f, ' ', w, p, fl^LEFT_ADJ);
+			l = w>p ? w : p;*/
+			continue;
+		case 'e': case 'f': case 'g': case 'a':
+		case 'E': case 'F': case 'G': case 'A':
+			if (xp && p<0) goto overflow;
+			l = fmt_fp(f, arg.f, w, p, fl, t);
+			if (l<0) goto overflow;
+			continue;
+		}
+
+		if (p < z-a) p = z-a;
+		if (p > INT_MAX-pl) goto overflow;
+		if (w < pl+p) w = pl+p;
+		if (w > INT_MAX-cnt) goto overflow;
+
+		pad(f, ' ', w, pl+p, fl);
+		out(f, prefix, pl);
+		pad(f, '0', w, pl+p, fl^ZERO_PAD);
+		pad(f, '0', p, z-a, 0);
+		out(f, a, z-a);
+		pad(f, ' ', w, pl+p, fl^LEFT_ADJ);
+
+		l = w;
+	}
+
+	if (f) return cnt;
+	if (!l10n) return 0;
+
+	for (i=1; i<=NL_ARGMAX && nl_type[i]; i++)
+		pop_arg(nl_arg+i, nl_type[i], ap);
+	for (; i<=NL_ARGMAX && !nl_type[i]; i++);
+	if (i<=NL_ARGMAX) goto inval;
+	return 1;
+
+inval:
+	errno = EINVAL;
+	return -1;
+overflow:
+	errno = EOVERFLOW;
+	return -1;
+}
+
+int vfprintf(FILE *restrict f, const char *restrict fmt, va_list ap)
+{
+	va_list ap2;
+	int nl_type[NL_ARGMAX+1] = {0};
+	union arg nl_arg[NL_ARGMAX+1];
+	unsigned char internal_buf[80], *saved_buf = 0;
+	int olderr;
+	int ret;
+
+	/* the copy allows passing va_list* even if va_list is an array */
+	va_copy(ap2, ap);
+	if (printf_core(0, fmt, &ap2, nl_arg, nl_type) < 0) {
+		va_end(ap2);
+		return -1;
+	}
+
+	FLOCK(f);
+	olderr = f->flags & F_ERR;
+	if (f->mode < 1) f->flags &= ~F_ERR;
+	if (!f->buf_size) {
+		saved_buf = f->buf;
+		f->buf = internal_buf;
+		f->buf_size = sizeof internal_buf;
+		f->wpos = f->wbase = f->wend = 0;
+	}
+	if (!f->wend && __towrite(f)) ret = -1;
+	else ret = printf_core(f, fmt, &ap2, nl_arg, nl_type);
+    
+	if (saved_buf) {
+		f->write(f, 0, 0);
+		if (!f->wpos) ret = -1;
+		f->buf = saved_buf;
+		f->buf_size = 0;
+		f->wpos = f->wbase = f->wend = 0;
+	}
+    
+	if (f->flags & F_ERR) ret = -1;
+	f->flags |= olderr;
+	FUNLOCK(f);
+	va_end(ap2);
+	return ret;
+}
diff --git a/src/stdio/vfscanf.c b/src/stdio/vfscanf.c
new file mode 100644
index 0000000..87bda14
--- /dev/null
+++ b/src/stdio/vfscanf.c
@@ -0,0 +1,340 @@
+#include <stdlib.h>
+#include <stdarg.h>
+#include <ctype.h>
+#include <wchar.h>
+#include <wctype.h>
+#include <limits.h>
+#include <string.h>
+#include <stdint.h>
+
+#include "stdio_impl.h"
+#include "shgetc.h"
+#include "intscan.h"
+#include "floatscan.h"
+
+#define SIZE_hh -2
+#define SIZE_h  -1
+#define SIZE_def 0
+#define SIZE_l   1
+#define SIZE_L   2
+#define SIZE_ll  3
+
+static void store_int(void *dest, int size, unsigned long long i)
+{
+	if (!dest) return;
+	switch (size) {
+	case SIZE_hh:
+		*(char *)dest = i;
+		break;
+	case SIZE_h:
+		*(short *)dest = i;
+		break;
+	case SIZE_def:
+		*(int *)dest = i;
+		break;
+	case SIZE_l:
+		*(long *)dest = i;
+		break;
+	case SIZE_ll:
+		*(long long *)dest = i;
+		break;
+	}
+}
+
+static void *arg_n(va_list ap, unsigned int n)
+{
+	void *p;
+	unsigned int i;
+	va_list ap2;
+	va_copy(ap2, ap);
+	for (i=n; i>1; i--) va_arg(ap2, void *);
+	p = va_arg(ap2, void *);
+	va_end(ap2);
+	return p;
+}
+
+int vfscanf(FILE *restrict f, const char *restrict fmt, va_list ap)
+{
+	int width;
+	int size;
+	int alloc = 0;
+	int base;
+	const unsigned char *p;
+	int c, t;
+	char *s;
+	wchar_t *wcs;
+	mbstate_t st;
+	void *dest=NULL;
+	int invert;
+	int matches=0;
+	unsigned long long x;
+	long double y;
+	off_t pos = 0;
+	unsigned char scanset[257];
+	size_t i, k;
+	wchar_t wc;
+
+	FLOCK(f);
+
+	if (!f->rpos) __toread(f);
+	if (!f->rpos) goto input_fail;
+
+	for (p=(const unsigned char *)fmt; *p; p++) {
+
+		alloc = 0;
+
+		if (isspace(*p)) {
+			while (isspace(p[1])) p++;
+			shlim(f, 0);
+			while (isspace(shgetc(f)));
+			shunget(f);
+			pos += shcnt(f);
+			continue;
+		}
+		if (*p != '%' || p[1] == '%') {
+			shlim(f, 0);
+			if (*p == '%') {
+				p++;
+				while (isspace((c=shgetc(f))));
+			} else {
+				c = shgetc(f);
+			}
+			if (c!=*p) {
+				shunget(f);
+				if (c<0) goto input_fail;
+				goto match_fail;
+			}
+			pos += shcnt(f);
+			continue;
+		}
+
+		p++;
+		if (*p=='*') {
+			dest = 0; p++;
+		} else if (isdigit(*p) && p[1]=='$') {
+			dest = arg_n(ap, *p-'0'); p+=2;
+		} else {
+			dest = va_arg(ap, void *);
+		}
+
+		for (width=0; isdigit(*p); p++) {
+			width = 10*width + *p - '0';
+		}
+
+		if (*p=='m') {
+			wcs = 0;
+			s = 0;
+			alloc = !!dest;
+			p++;
+		} else {
+			alloc = 0;
+		}
+
+		size = SIZE_def;
+		switch (*p++) {
+		case 'h':
+			if (*p == 'h') p++, size = SIZE_hh;
+			else size = SIZE_h;
+			break;
+		case 'l':
+			if (*p == 'l') p++, size = SIZE_ll;
+			else size = SIZE_l;
+			break;
+		case 'j':
+			size = SIZE_ll;
+			break;
+		case 'z':
+		case 't':
+			size = SIZE_l;
+			break;
+		case 'L':
+			size = SIZE_L;
+			break;
+		case 'd': case 'i': case 'o': case 'u': case 'x':
+		case 'a': case 'e': case 'f': case 'g':
+		case 'A': case 'E': case 'F': case 'G': case 'X':
+		case 's': case 'c': case '[':
+		case 'S': case 'C':
+		case 'p': case 'n':
+			p--;
+			break;
+		default:
+			goto fmt_fail;
+		}
+
+		t = *p;
+
+		/* C or S */
+		if ((t&0x2f) == 3) {
+			t |= 32;
+			size = SIZE_l;
+		}
+
+		switch (t) {
+		case 'c':
+			if (width < 1) width = 1;
+		case '[':
+			break;
+		case 'n':
+			store_int(dest, size, pos);
+			/* do not increment match count, etc! */
+			continue;
+		default:
+			shlim(f, 0);
+			while (isspace(shgetc(f)));
+			shunget(f);
+			pos += shcnt(f);
+		}
+
+		shlim(f, width);
+		if (shgetc(f) < 0) goto input_fail;
+		shunget(f);
+
+		switch (t) {
+		case 's':
+		case 'c':
+		case '[':
+			if (t == 'c' || t == 's') {
+				memset(scanset, -1, sizeof scanset);
+				scanset[0] = 0;
+				if (t == 's') {
+					scanset[1+'\t'] = 0;
+					scanset[1+'\n'] = 0;
+					scanset[1+'\v'] = 0;
+					scanset[1+'\f'] = 0;
+					scanset[1+'\r'] = 0;
+					scanset[1+' '] = 0;
+				}
+			} else {
+				if (*++p == '^') p++, invert = 1;
+				else invert = 0;
+				memset(scanset, invert, sizeof scanset);
+				scanset[0] = 0;
+				if (*p == '-') p++, scanset[1+'-'] = 1-invert;
+				else if (*p == ']') p++, scanset[1+']'] = 1-invert;
+				for (; *p != ']'; p++) {
+					if (!*p) goto fmt_fail;
+					if (*p=='-' && p[1] && p[1] != ']')
+						for (c=p++[-1]; c<*p; c++)
+							scanset[1+c] = 1-invert;
+					scanset[1+*p] = 1-invert;
+				}
+			}
+			wcs = 0;
+			s = 0;
+			i = 0;
+			k = t=='c' ? width+1U : 31;
+			if (size == SIZE_l) {
+				if (alloc) {
+					wcs = malloc(k*sizeof(wchar_t));
+					if (!wcs) goto alloc_fail;
+				} else {
+					wcs = dest;
+				}
+				st = (mbstate_t){0};
+				while (scanset[(c=shgetc(f))+1]) {
+                    /*FIXME: !!!
+					switch (mbrtowc(&wc, &(char){c}, 1, &st)) {
+					case -1:
+						goto input_fail;
+					case -2:
+						continue;
+					}*/
+					if (wcs) wcs[i++] = wc;
+					if (alloc && i==k) {
+						k+=k+1;
+						wchar_t *tmp = realloc(wcs, k*sizeof(wchar_t));
+						if (!tmp) goto alloc_fail;
+						wcs = tmp;
+					}
+				}
+				/* FIXME:!!! if (!mbsinit(&st)) goto input_fail;*/
+			} else if (alloc) {
+				s = malloc(k);
+				if (!s) goto alloc_fail;
+				while (scanset[(c=shgetc(f))+1]) {
+					s[i++] = c;
+					if (i==k) {
+						k+=k+1;
+						char *tmp = realloc(s, k);
+						if (!tmp) goto alloc_fail;
+						s = tmp;
+					}
+				}
+			} else if ((s = dest)) {
+				while (scanset[(c=shgetc(f))+1])
+					s[i++] = c;
+			} else {
+				while (scanset[(c=shgetc(f))+1]);
+			}
+			shunget(f);
+			if (!shcnt(f)) goto match_fail;
+			if (t == 'c' && shcnt(f) != width) goto match_fail;
+			if (alloc) {
+				if (size == SIZE_l) *(wchar_t **)dest = wcs;
+				else *(char **)dest = s;
+			}
+			if (t != 'c') {
+				if (wcs) wcs[i] = 0;
+				if (s) s[i] = 0;
+			}
+			break;
+		case 'p':
+		case 'X':
+		case 'x':
+			base = 16;
+			goto int_common;
+		case 'o':
+			base = 8;
+			goto int_common;
+		case 'd':
+		case 'u':
+			base = 10;
+			goto int_common;
+		case 'i':
+			base = 0;
+		int_common:
+			x = __intscan(f, base, 0, ULLONG_MAX);
+			if (!shcnt(f)) goto match_fail;
+			if (t=='p' && dest) *(void **)dest = (void *)(uintptr_t)x;
+			else store_int(dest, size, x);
+			break;
+		case 'a': case 'A':
+		case 'e': case 'E':
+		case 'f': case 'F':
+		case 'g': case 'G':
+			y = __floatscan(f, size, 0);
+			if (!shcnt(f)) goto match_fail;
+			if (dest) switch (size) {
+			case SIZE_def:
+				*(float *)dest = y;
+				break;
+			case SIZE_l:
+				*(double *)dest = y;
+				break;
+			case SIZE_L:
+				*(long double *)dest = y;
+				break;
+			}
+			break;
+		}
+
+		pos += shcnt(f);
+		if (dest) matches++;
+	}
+	if (0) {
+fmt_fail:
+alloc_fail:
+input_fail:
+		if (!matches) matches--;
+match_fail:
+		if (alloc) {
+			free(s);
+			free(wcs);
+		}
+	}
+	FUNLOCK(f);
+	return matches;
+}
+
+weak_alias(vfscanf,__isoc99_vfscanf);
diff --git a/src/stdio/vfwprintf.c_ b/src/stdio/vfwprintf.c_
new file mode 100644
index 0000000..85b036c
--- /dev/null
+++ b/src/stdio/vfwprintf.c_
@@ -0,0 +1,365 @@
+#include "stdio_impl.h"
+#include <errno.h>
+#include <ctype.h>
+#include <limits.h>
+#include <string.h>
+#include <stdarg.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <wchar.h>
+#include <inttypes.h>
+
+/* Convenient bit representation for modifier flags, which all fall
+ * within 31 codepoints of the space character. */
+
+#define ALT_FORM   (1U<<'#'-' ')
+#define ZERO_PAD   (1U<<'0'-' ')
+#define LEFT_ADJ   (1U<<'-'-' ')
+#define PAD_POS    (1U<<' '-' ')
+#define MARK_POS   (1U<<'+'-' ')
+#define GROUPED    (1U<<'\''-' ')
+
+#define FLAGMASK (ALT_FORM|ZERO_PAD|LEFT_ADJ|PAD_POS|MARK_POS|GROUPED)
+
+/* State machine to accept length modifiers + conversion specifiers.
+ * Result is 0 on failure, or an argument type to pop on success. */
+
+enum {
+	BARE, LPRE, LLPRE, HPRE, HHPRE, BIGLPRE,
+	ZTPRE, JPRE,
+	STOP,
+	PTR, INT, UINT, ULLONG,
+	LONG, ULONG,
+	SHORT, USHORT, CHAR, UCHAR,
+	LLONG, SIZET, IMAX, UMAX, PDIFF, UIPTR,
+	DBL, LDBL,
+	NOARG,
+	MAXSTATE
+};
+
+#define S(x) [(x)-'A']
+
+static const unsigned char states[]['z'-'A'+1] = {
+	{ /* 0: bare types */
+		S('d') = INT, S('i') = INT,
+		S('o') = UINT, S('u') = UINT, S('x') = UINT, S('X') = UINT,
+		S('e') = DBL, S('f') = DBL, S('g') = DBL, S('a') = DBL,
+		S('E') = DBL, S('F') = DBL, S('G') = DBL, S('A') = DBL,
+		S('c') = CHAR, S('C') = INT,
+		S('s') = PTR, S('S') = PTR, S('p') = UIPTR, S('n') = PTR,
+		S('m') = NOARG,
+		S('l') = LPRE, S('h') = HPRE, S('L') = BIGLPRE,
+		S('z') = ZTPRE, S('j') = JPRE, S('t') = ZTPRE,
+	}, { /* 1: l-prefixed */
+		S('d') = LONG, S('i') = LONG,
+		S('o') = ULONG, S('u') = ULONG, S('x') = ULONG, S('X') = ULONG,
+		S('e') = DBL, S('f') = DBL, S('g') = DBL, S('a') = DBL,
+		S('E') = DBL, S('F') = DBL, S('G') = DBL, S('A') = DBL,
+		S('c') = INT, S('s') = PTR, S('n') = PTR,
+		S('l') = LLPRE,
+	}, { /* 2: ll-prefixed */
+		S('d') = LLONG, S('i') = LLONG,
+		S('o') = ULLONG, S('u') = ULLONG,
+		S('x') = ULLONG, S('X') = ULLONG,
+		S('n') = PTR,
+	}, { /* 3: h-prefixed */
+		S('d') = SHORT, S('i') = SHORT,
+		S('o') = USHORT, S('u') = USHORT,
+		S('x') = USHORT, S('X') = USHORT,
+		S('n') = PTR,
+		S('h') = HHPRE,
+	}, { /* 4: hh-prefixed */
+		S('d') = CHAR, S('i') = CHAR,
+		S('o') = UCHAR, S('u') = UCHAR,
+		S('x') = UCHAR, S('X') = UCHAR,
+		S('n') = PTR,
+	}, { /* 5: L-prefixed */
+		S('e') = LDBL, S('f') = LDBL, S('g') = LDBL, S('a') = LDBL,
+		S('E') = LDBL, S('F') = LDBL, S('G') = LDBL, S('A') = LDBL,
+		S('n') = PTR,
+	}, { /* 6: z- or t-prefixed (assumed to be same size) */
+		S('d') = PDIFF, S('i') = PDIFF,
+		S('o') = SIZET, S('u') = SIZET,
+		S('x') = SIZET, S('X') = SIZET,
+		S('n') = PTR,
+	}, { /* 7: j-prefixed */
+		S('d') = IMAX, S('i') = IMAX,
+		S('o') = UMAX, S('u') = UMAX,
+		S('x') = UMAX, S('X') = UMAX,
+		S('n') = PTR,
+	}
+};
+
+#define OOB(x) ((unsigned)(x)-'A' > 'z'-'A')
+
+union arg
+{
+	uintmax_t i;
+	long double f;
+	void *p;
+};
+
+static void pop_arg(union arg *arg, int type, va_list *ap)
+{
+	switch (type) {
+	       case PTR:	arg->p = va_arg(*ap, void *);
+	break; case INT:	arg->i = va_arg(*ap, int);
+	break; case UINT:	arg->i = va_arg(*ap, unsigned int);
+	break; case LONG:	arg->i = va_arg(*ap, long);
+	break; case ULONG:	arg->i = va_arg(*ap, unsigned long);
+	break; case ULLONG:	arg->i = va_arg(*ap, unsigned long long);
+	break; case SHORT:	arg->i = (short)va_arg(*ap, int);
+	break; case USHORT:	arg->i = (unsigned short)va_arg(*ap, int);
+	break; case CHAR:	arg->i = (signed char)va_arg(*ap, int);
+	break; case UCHAR:	arg->i = (unsigned char)va_arg(*ap, int);
+	break; case LLONG:	arg->i = va_arg(*ap, long long);
+	break; case SIZET:	arg->i = va_arg(*ap, size_t);
+	break; case IMAX:	arg->i = va_arg(*ap, intmax_t);
+	break; case UMAX:	arg->i = va_arg(*ap, uintmax_t);
+	break; case PDIFF:	arg->i = va_arg(*ap, ptrdiff_t);
+	break; case UIPTR:	arg->i = (uintptr_t)va_arg(*ap, void *);
+	break; case DBL:	arg->f = va_arg(*ap, double);
+	break; case LDBL:	arg->f = va_arg(*ap, long double);
+	}
+}
+
+static void out(FILE *f, const wchar_t *s, size_t l)
+{
+	while (l-- && !(f->flags & F_ERR)) fputwc(*s++, f);
+}
+
+static int getint(wchar_t **s) {
+	int i;
+	for (i=0; iswdigit(**s); (*s)++) {
+		if (i > INT_MAX/10U || **s-'0' > INT_MAX-10*i) i = -1;
+		else i = 10*i + (**s-'0');
+	}
+	return i;
+}
+
+static const char sizeprefix['y'-'a'] = {
+['a'-'a']='L', ['e'-'a']='L', ['f'-'a']='L', ['g'-'a']='L',
+['d'-'a']='j', ['i'-'a']='j', ['o'-'a']='j', ['u'-'a']='j', ['x'-'a']='j',
+['p'-'a']='j'
+};
+
+static int wprintf_core(FILE *f, const wchar_t *fmt, va_list *ap, union arg *nl_arg, int *nl_type)
+{
+	wchar_t *a, *z, *s=(wchar_t *)fmt;
+	unsigned l10n=0, fl;
+	int w, p, xp;
+	union arg arg;
+	int argpos;
+	unsigned st, ps;
+	int cnt=0, l=0;
+	int i;
+	int t;
+	char *bs;
+	char charfmt[16];
+	wchar_t wc;
+
+	for (;;) {
+		/* This error is only specified for snprintf, but since it's
+		 * unspecified for other forms, do the same. Stop immediately
+		 * on overflow; otherwise %n could produce wrong results. */
+		if (l > INT_MAX - cnt) goto overflow;
+
+		/* Update output count, end loop when fmt is exhausted */
+		cnt += l;
+		if (!*s) break;
+
+		/* Handle literal text and %% format specifiers */
+		for (a=s; *s && *s!='%'; s++);
+		for (z=s; s[0]=='%' && s[1]=='%'; z++, s+=2);
+		if (z-a > INT_MAX-cnt) goto overflow;
+		l = z-a;
+		if (f) out(f, a, l);
+		if (l) continue;
+
+		if (iswdigit(s[1]) && s[2]=='$') {
+			l10n=1;
+			argpos = s[1]-'0';
+			s+=3;
+		} else {
+			argpos = -1;
+			s++;
+		}
+
+		/* Read modifier flags */
+		for (fl=0; (unsigned)*s-' '<32 && (FLAGMASK&(1U<<*s-' ')); s++)
+			fl |= 1U<<*s-' ';
+
+		/* Read field width */
+		if (*s=='*') {
+			if (iswdigit(s[1]) && s[2]=='$') {
+				l10n=1;
+				nl_type[s[1]-'0'] = INT;
+				w = nl_arg[s[1]-'0'].i;
+				s+=3;
+			} else if (!l10n) {
+				w = f ? va_arg(*ap, int) : 0;
+				s++;
+			} else goto inval;
+			if (w<0) fl|=LEFT_ADJ, w=-w;
+		} else if ((w=getint(&s))<0) goto overflow;
+
+		/* Read precision */
+		if (*s=='.' && s[1]=='*') {
+			if (isdigit(s[2]) && s[3]=='$') {
+				nl_type[s[2]-'0'] = INT;
+				p = nl_arg[s[2]-'0'].i;
+				s+=4;
+			} else if (!l10n) {
+				p = f ? va_arg(*ap, int) : 0;
+				s+=2;
+			} else goto inval;
+			xp = (p>=0);
+		} else if (*s=='.') {
+			s++;
+			p = getint(&s);
+			xp = 1;
+		} else {
+			p = -1;
+			xp = 0;
+		}
+
+		/* Format specifier state machine */
+		st=0;
+		do {
+			if (OOB(*s)) goto inval;
+			ps=st;
+			st=states[st]S(*s++);
+		} while (st-1<STOP);
+		if (!st) goto inval;
+
+		/* Check validity of argument type (nl/normal) */
+		if (st==NOARG) {
+			if (argpos>=0) goto inval;
+		} else {
+			if (argpos>=0) nl_type[argpos]=st, arg=nl_arg[argpos];
+			else if (f) pop_arg(&arg, st, ap);
+			else return 0;
+		}
+
+		if (!f) continue;
+		t = s[-1];
+		if (ps && (t&15)==3) t&=~32;
+
+		switch (t) {
+		case 'n':
+			switch(ps) {
+			case BARE: *(int *)arg.p = cnt; break;
+			case LPRE: *(long *)arg.p = cnt; break;
+			case LLPRE: *(long long *)arg.p = cnt; break;
+			case HPRE: *(unsigned short *)arg.p = cnt; break;
+			case HHPRE: *(unsigned char *)arg.p = cnt; break;
+			case ZTPRE: *(size_t *)arg.p = cnt; break;
+			case JPRE: *(uintmax_t *)arg.p = cnt; break;
+			}
+			continue;
+		case 'c':
+			if (w<1) w=1;
+			if (w>1 && !(fl&LEFT_ADJ)) fprintf(f, "%*s", w-1, "");
+			fputwc(btowc(arg.i), f);
+			if (w>1 && (fl&LEFT_ADJ)) fprintf(f, "%*s", w-1, "");
+			l = w;
+			continue;
+		case 'C':
+			fputwc(arg.i, f);
+			l = 1;
+			continue;
+		case 'S':
+			a = arg.p;
+			z = a + wcsnlen(a, p<0 ? INT_MAX : p);
+			if (p<0 && *z) goto overflow;
+			p = z-a;
+			if (w<p) w=p;
+			if (!(fl&LEFT_ADJ)) fprintf(f, "%*s", w-p, "");
+			out(f, a, p);
+			if ((fl&LEFT_ADJ)) fprintf(f, "%*s", w-p, "");
+			l=w;
+			continue;
+		case 'm':
+			arg.p = strerror(errno);
+		case 's':
+			if (!arg.p) arg.p = "(null)";
+			bs = arg.p;
+			for (i=l=0; l<(p<0?INT_MAX:p) && (i=mbtowc(&wc, bs, MB_LEN_MAX))>0; bs+=i, l++);
+			if (i<0) return -1;
+			if (p<0 && *bs) goto overflow;
+			p=l;
+			if (w<p) w=p;
+			if (!(fl&LEFT_ADJ)) fprintf(f, "%*s", w-p, "");
+			bs = arg.p;
+			while (l--) {
+				i=mbtowc(&wc, bs, MB_LEN_MAX);
+				bs+=i;
+				fputwc(wc, f);
+			}
+			if ((fl&LEFT_ADJ)) fprintf(f, "%*s", w-p, "");
+			l=w;
+			continue;
+		}
+
+		if (xp && p<0) goto overflow;
+		snprintf(charfmt, sizeof charfmt, "%%%s%s%s%s%s*.*%c%c",
+			"#"+!(fl & ALT_FORM),
+			"+"+!(fl & MARK_POS),
+			"-"+!(fl & LEFT_ADJ),
+			" "+!(fl & PAD_POS),
+			"0"+!(fl & ZERO_PAD),
+			sizeprefix[(t|32)-'a'], t);
+
+		switch (t|32) {
+		case 'a': case 'e': case 'f': case 'g':
+			l = fprintf(f, charfmt, w, p, arg.f);
+			break;
+		case 'd': case 'i': case 'o': case 'u': case 'x': case 'p':
+			l = fprintf(f, charfmt, w, p, arg.i);
+			break;
+		}
+	}
+
+	if (f) return cnt;
+	if (!l10n) return 0;
+
+	for (i=1; i<=NL_ARGMAX && nl_type[i]; i++)
+		pop_arg(nl_arg+i, nl_type[i], ap);
+	for (; i<=NL_ARGMAX && !nl_type[i]; i++);
+	if (i<=NL_ARGMAX) return -1;
+	return 1;
+
+inval:
+	errno = EINVAL;
+	return -1;
+overflow:
+	errno = EOVERFLOW;
+	return -1;
+}
+
+int vfwprintf(FILE *restrict f, const wchar_t *restrict fmt, va_list ap)
+{
+	va_list ap2;
+	int nl_type[NL_ARGMAX] = {0};
+	union arg nl_arg[NL_ARGMAX];
+	int olderr;
+	int ret;
+
+	/* the copy allows passing va_list* even if va_list is an array */
+	va_copy(ap2, ap);
+	if (wprintf_core(0, fmt, &ap2, nl_arg, nl_type) < 0) {
+		va_end(ap2);
+		return -1;
+	}
+
+	FLOCK(f);
+	fwide(f, 1);
+	olderr = f->flags & F_ERR;
+	f->flags &= ~F_ERR;
+	ret = wprintf_core(f, fmt, &ap2, nl_arg, nl_type);
+	if (f->flags & F_ERR) ret = -1;
+	f->flags |= olderr;
+	FUNLOCK(f);
+	va_end(ap2);
+	return ret;
+}
diff --git a/src/stdio/vfwscanf.c_ b/src/stdio/vfwscanf.c_
new file mode 100644
index 0000000..82f4860
--- /dev/null
+++ b/src/stdio/vfwscanf.c_
@@ -0,0 +1,332 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <ctype.h>
+#include <wchar.h>
+#include <wctype.h>
+#include <limits.h>
+#include <string.h>
+
+#include "stdio_impl.h"
+#include "shgetc.h"
+#include "intscan.h"
+#include "floatscan.h"
+
+#define SIZE_hh -2
+#define SIZE_h  -1
+#define SIZE_def 0
+#define SIZE_l   1
+#define SIZE_L   2
+#define SIZE_ll  3
+
+static void store_int(void *dest, int size, unsigned long long i)
+{
+	if (!dest) return;
+	switch (size) {
+	case SIZE_hh:
+		*(char *)dest = i;
+		break;
+	case SIZE_h:
+		*(short *)dest = i;
+		break;
+	case SIZE_def:
+		*(int *)dest = i;
+		break;
+	case SIZE_l:
+		*(long *)dest = i;
+		break;
+	case SIZE_ll:
+		*(long long *)dest = i;
+		break;
+	}
+}
+
+static void *arg_n(va_list ap, unsigned int n)
+{
+	void *p;
+	unsigned int i;
+	va_list ap2;
+	va_copy(ap2, ap);
+	for (i=n; i>1; i--) va_arg(ap2, void *);
+	p = va_arg(ap2, void *);
+	va_end(ap2);
+	return p;
+}
+
+static int in_set(const wchar_t *set, int c)
+{
+	int j;
+	const wchar_t *p = set;
+	if (*p == '-') {
+		if (c=='-') return 1;
+		p++;
+	} else if (*p == ']') {
+		if (c==']') return 1;
+		p++;
+	}
+	for (; *p && *p != ']'; p++) {
+		if (*p=='-' && p[1] && p[1] != ']')
+			for (j=p++[-1]; j<*p; j++)
+				if (c==j) return 1;
+		if (c==*p) return 1;
+	}
+	return 0;
+}
+
+#if 1
+#undef getwc
+#define getwc(f) \
+	((f)->rpos != (f)->rend && *(f)->rpos < 128 ? *(f)->rpos++ : (getwc)(f))
+
+#undef ungetwc
+#define ungetwc(c,f) \
+	((f)->rend && (c)<128U ? *--(f)->rpos : ungetwc((c),(f)))
+#endif
+
+int vfwscanf(FILE *restrict f, const wchar_t *restrict fmt, va_list ap)
+{
+	int width;
+	int size;
+	int alloc;
+	const wchar_t *p;
+	int c, t;
+	char *s;
+	wchar_t *wcs;
+	void *dest=NULL;
+	int invert;
+	int matches=0;
+	off_t pos = 0, cnt;
+	static const char size_pfx[][3] = { "hh", "h", "", "l", "L", "ll" };
+	char tmp[3*sizeof(int)+10];
+	const wchar_t *set;
+	size_t i, k;
+
+	FLOCK(f);
+
+	fwide(f, 1);
+
+	for (p=fmt; *p; p++) {
+
+		alloc = 0;
+
+		if (iswspace(*p)) {
+			while (iswspace(p[1])) p++;
+			while (iswspace((c=getwc(f)))) pos++;
+			ungetwc(c, f);
+			continue;
+		}
+		if (*p != '%' || p[1] == '%') {
+			if (*p == '%') {
+				p++;
+				while (iswspace((c=getwc(f)))) pos++;
+			} else {
+				c = getwc(f);
+			}
+			if (c!=*p) {
+				ungetwc(c, f);
+				if (c<0) goto input_fail;
+				goto match_fail;
+			}
+			pos++;
+			continue;
+		}
+
+		p++;
+		if (*p=='*') {
+			dest = 0; p++;
+		} else if (iswdigit(*p) && p[1]=='$') {
+			dest = arg_n(ap, *p-'0'); p+=2;
+		} else {
+			dest = va_arg(ap, void *);
+		}
+
+		for (width=0; iswdigit(*p); p++) {
+			width = 10*width + *p - '0';
+		}
+
+		if (*p=='m') {
+			wcs = 0;
+			s = 0;
+			alloc = !!dest;
+			p++;
+		} else {
+			alloc = 0;
+		}
+
+		size = SIZE_def;
+		switch (*p++) {
+		case 'h':
+			if (*p == 'h') p++, size = SIZE_hh;
+			else size = SIZE_h;
+			break;
+		case 'l':
+			if (*p == 'l') p++, size = SIZE_ll;
+			else size = SIZE_l;
+			break;
+		case 'j':
+			size = SIZE_ll;
+			break;
+		case 'z':
+		case 't':
+			size = SIZE_l;
+			break;
+		case 'L':
+			size = SIZE_L;
+			break;
+		case 'd': case 'i': case 'o': case 'u': case 'x':
+		case 'a': case 'e': case 'f': case 'g':
+		case 'A': case 'E': case 'F': case 'G': case 'X':
+		case 's': case 'c': case '[':
+		case 'S': case 'C':
+		case 'p': case 'n':
+			p--;
+			break;
+		default:
+			goto fmt_fail;
+		}
+
+		t = *p;
+
+		/* Transform S,C -> ls,lc */
+		if ((t&0x2f)==3) {
+			size = SIZE_l;
+			t |= 32;
+		}
+
+		if (t != 'n') {
+			if (t != '[' && (t|32) != 'c')
+				while (iswspace((c=getwc(f)))) pos++;
+			else
+				c=getwc(f);
+			if (c < 0) goto input_fail;
+			ungetwc(c, f);
+		}
+
+		switch (t) {
+		case 'n':
+			store_int(dest, size, pos);
+			/* do not increment match count, etc! */
+			continue;
+
+		case 's':
+		case 'c':
+		case '[':
+			if (t == 'c') {
+				if (width<1) width = 1;
+				invert = 1;
+				set = L"";
+			} else if (t == 's') {
+				invert = 1;
+				static const wchar_t spaces[] = {
+					' ', '\t', '\n', '\r', 11, 12,  0x0085,
+					0x2000, 0x2001, 0x2002, 0x2003, 0x2004, 0x2005,
+					0x2006, 0x2008, 0x2009, 0x200a,
+					0x2028, 0x2029, 0x205f, 0x3000, 0 };
+				set = spaces;
+			} else {
+				if (*++p == '^') p++, invert = 1;
+				else invert = 0;
+				set = p;
+				if (*p==']') p++;
+				while (*p!=']') {
+					if (!*p) goto fmt_fail;
+					p++;
+				}
+			}
+
+			s = (size == SIZE_def) ? dest : 0;
+			wcs = (size == SIZE_l) ? dest : 0;
+
+			int gotmatch = 0;
+
+			if (width < 1) width = -1;
+
+			i = 0;
+			if (alloc) {
+				k = t=='c' ? width+1U : 31;
+				if (size == SIZE_l) {
+					wcs = malloc(k*sizeof(wchar_t));
+					if (!wcs) goto alloc_fail;
+				} else {
+					s = malloc(k);
+					if (!s) goto alloc_fail;
+				}
+			}
+			while (width) {
+				if ((c=getwc(f))<0) break;
+				if (in_set(set, c) == invert)
+					break;
+				if (wcs) {
+					wcs[i++] = c;
+					if (alloc && i==k) {
+						k += k+1;
+						wchar_t *tmp = realloc(wcs, k*sizeof(wchar_t));
+						if (!tmp) goto alloc_fail;
+						wcs = tmp;
+					}
+				} else if (size != SIZE_l) {
+					int l = wctomb(s?s+i:tmp, c);
+					if (l<0) goto input_fail;
+					i += l;
+					if (alloc && i > k-4) {
+						k += k+1;
+						char *tmp = realloc(s, k);
+						if (!tmp) goto alloc_fail;
+						s = tmp;
+					}
+				}
+				pos++;
+				width-=(width>0);
+				gotmatch=1;
+			}
+			if (width) {
+				ungetwc(c, f);
+				if (t == 'c' || !gotmatch) goto match_fail;
+			}
+
+			if (alloc) {
+				if (size == SIZE_l) *(wchar_t **)dest = wcs;
+				else *(char **)dest = s;
+			}
+			if (t != 'c') {
+				if (wcs) wcs[i] = 0;
+				if (s) s[i] = 0;
+			}
+			break;
+
+		case 'd': case 'i': case 'o': case 'u': case 'x':
+		case 'a': case 'e': case 'f': case 'g':
+		case 'A': case 'E': case 'F': case 'G': case 'X':
+		case 'p':
+			if (width < 1) width = 0;
+			snprintf(tmp, sizeof tmp, "%.*s%.0d%s%c%%lln",
+				1+!dest, "%*", width, size_pfx[size+2], t);
+			cnt = 0;
+			if (fscanf(f, tmp, dest?dest:&cnt, &cnt) == -1)
+				goto input_fail;
+			else if (!cnt)
+				goto match_fail;
+			pos += cnt;
+			break;
+		default:
+			goto fmt_fail;
+		}
+
+		if (dest) matches++;
+	}
+	if (0) {
+fmt_fail:
+alloc_fail:
+input_fail:
+		if (!matches) matches--;
+match_fail:
+		if (alloc) {
+			free(s);
+			free(wcs);
+		}
+	}
+	FUNLOCK(f);
+	return matches;
+}
+
+weak_alias(vfwscanf,__isoc99_vfwscanf);
diff --git a/src/stdio/vprintf.c b/src/stdio/vprintf.c
new file mode 100644
index 0000000..30d2bff
--- /dev/null
+++ b/src/stdio/vprintf.c
@@ -0,0 +1,6 @@
+#include <stdio.h>
+
+int vprintf(const char *restrict fmt, va_list ap)
+{
+	return vfprintf(stdout, fmt, ap);
+}
diff --git a/src/stdio/vscanf.c b/src/stdio/vscanf.c
new file mode 100644
index 0000000..9d46ab0
--- /dev/null
+++ b/src/stdio/vscanf.c
@@ -0,0 +1,9 @@
+#include <stdio.h>
+#include <stdarg.h>
+
+int vscanf(const char *restrict fmt, va_list ap)
+{
+	return vfscanf(stdin, fmt, ap);
+}
+
+weak_alias(vscanf,__isoc99_vscanf);
diff --git a/src/stdio/vsnprintf.c b/src/stdio/vsnprintf.c
new file mode 100644
index 0000000..b3510a6
--- /dev/null
+++ b/src/stdio/vsnprintf.c
@@ -0,0 +1,55 @@
+#include "stdio_impl.h"
+#include <limits.h>
+#include <string.h>
+#include <errno.h>
+#include <stdint.h>
+
+struct cookie {
+	char *s;
+	size_t n;
+};
+
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+
+static size_t sn_write(FILE *f, const unsigned char *s, size_t l)
+{
+	struct cookie *c = f->cookie;
+	size_t k = MIN(c->n, f->wpos - f->wbase);
+	if (k) {
+		memcpy(c->s, f->wbase, k);
+		c->s += k;
+		c->n -= k;
+	}
+	k = MIN(c->n, l);
+	if (k) {
+		memcpy(c->s, s, k);
+		c->s += k;
+		c->n -= k;
+	}
+	*c->s = 0;
+	f->wpos = f->wbase = f->buf;
+	/* pretend to succeed, even if we discarded extra data */
+	return l;
+}
+
+int vsnprintf(char *restrict s, size_t n, const char *restrict fmt, va_list ap)
+{
+	unsigned char buf[1];
+	char dummy[1];
+	struct cookie c = { .s = n ? s : dummy, .n = n ? n-1 : 0 };
+	FILE f = {
+		.lbf = EOF,
+		.write = sn_write,
+		.lock = -1,
+		.buf = buf,
+		.cookie = &c,
+	};
+
+	if (n > INT_MAX) {
+		errno = EOVERFLOW;
+		return -1;
+	}
+
+	*c.s = 0;
+	return vfprintf(&f, fmt, ap);
+}
diff --git a/src/stdio/vsprintf.c b/src/stdio/vsprintf.c
new file mode 100644
index 0000000..c57349d
--- /dev/null
+++ b/src/stdio/vsprintf.c
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include <limits.h>
+
+int vsprintf(char *restrict s, const char *restrict fmt, va_list ap)
+{
+	return vsnprintf(s, INT_MAX, fmt, ap);
+}
diff --git a/src/stdio/vsscanf.c b/src/stdio/vsscanf.c
new file mode 100644
index 0000000..4d6d259
--- /dev/null
+++ b/src/stdio/vsscanf.c
@@ -0,0 +1,27 @@
+#include "stdio_impl.h"
+#include <string.h>
+
+static size_t string_read(FILE *f, unsigned char *buf, size_t len)
+{
+	char *src = f->cookie;
+	size_t k = len+256;
+	char *end = memchr(src, 0, k);
+	if (end) k = end-src;
+	if (k < len) len = k;
+	memcpy(buf, src, len);
+	f->rpos = (void *)(src+len);
+	f->rend = (void *)(src+k);
+	f->cookie = src+k;
+	return len;
+}
+
+int vsscanf(const char *restrict s, const char *restrict fmt, va_list ap)
+{
+	FILE f = {
+		.buf = (void *)s, .cookie = (void *)s,
+		.read = string_read, .lock = -1
+	};
+	return vfscanf(&f, fmt, ap);
+}
+
+weak_alias(vsscanf,__isoc99_vsscanf);
diff --git a/src/stdio/vswprintf.c_ b/src/stdio/vswprintf.c_
new file mode 100644
index 0000000..7f98c5c
--- /dev/null
+++ b/src/stdio/vswprintf.c_
@@ -0,0 +1,60 @@
+#include "stdio_impl.h"
+#include <limits.h>
+#include <errno.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <wchar.h>
+
+struct cookie {
+	wchar_t *ws;
+	size_t l;
+};
+
+static size_t sw_write(FILE *f, const unsigned char *s, size_t l)
+{
+	size_t l0 = l;
+	int i = 0;
+	struct cookie *c = f->cookie;
+	if (s!=f->wbase && sw_write(f, f->wbase, f->wpos-f->wbase)==-1)
+		return -1;
+	while (c->l && l && (i=mbtowc(c->ws, (void *)s, l))>=0) {
+		s+=i;
+		l-=i;
+		c->l--;
+		c->ws++;
+	}
+	*c->ws = 0;
+	if (i < 0) {
+		f->wpos = f->wbase = f->wend = 0;
+		f->flags |= F_ERR;
+		return i;
+	}
+	f->wend = f->buf + f->buf_size;
+	f->wpos = f->wbase = f->buf;
+	return l0;
+}
+
+int vswprintf(wchar_t *restrict s, size_t n, const wchar_t *restrict fmt, va_list ap)
+{
+	int r;
+	unsigned char buf[256];
+	struct cookie c = { s, n-1 };
+	FILE f = {
+		.lbf = EOF,
+		.write = sw_write,
+		.lock = -1,
+		.buf = buf,
+		.buf_size = sizeof buf,
+		.cookie = &c,
+	};
+
+	if (!n) {
+		return -1;
+	} else if (n > INT_MAX) {
+		errno = EOVERFLOW;
+		return -1;
+	}
+	r = vfwprintf(&f, fmt, ap);
+	sw_write(&f, 0, 0);
+	return r>=n ? -1 : r;
+}
diff --git a/src/stdio/vswscanf.c_ b/src/stdio/vswscanf.c_
new file mode 100644
index 0000000..00b614b
--- /dev/null
+++ b/src/stdio/vswscanf.c_
@@ -0,0 +1,38 @@
+#include "stdio_impl.h"
+#include <wchar.h>
+
+static size_t wstring_read(FILE *f, unsigned char *buf, size_t len)
+{
+	const wchar_t *src = f->cookie;
+	size_t k;
+
+	if (!src) return 0;
+
+	k = wcsrtombs((void *)f->buf, &src, f->buf_size, 0);
+	if (k==(size_t)-1) {
+		f->rpos = f->rend = 0;
+		return 0;
+	}
+
+	f->rpos = f->buf;
+	f->rend = f->buf + k;
+	f->cookie = (void *)src;
+
+	if (!len || !k) return 0;
+
+	*buf = *f->rpos++;
+	return 1;
+}
+
+int vswscanf(const wchar_t *restrict s, const wchar_t *restrict fmt, va_list ap)
+{
+	unsigned char buf[256];
+	FILE f = {
+		.buf = buf, .buf_size = sizeof buf,
+		.cookie = (void *)s,
+		.read = wstring_read, .lock = -1
+	};
+	return vfwscanf(&f, fmt, ap);
+}
+
+weak_alias(vswscanf,__isoc99_vswscanf);
diff --git a/src/stdio/vwprintf.c_ b/src/stdio/vwprintf.c_
new file mode 100644
index 0000000..eeeecdc
--- /dev/null
+++ b/src/stdio/vwprintf.c_
@@ -0,0 +1,7 @@
+#include <stdio.h>
+#include <wchar.h>
+
+int vwprintf(const wchar_t *restrict fmt, va_list ap)
+{
+	return vfwprintf(stdout, fmt, ap);
+}
diff --git a/src/stdio/vwscanf.c_ b/src/stdio/vwscanf.c_
new file mode 100644
index 0000000..5a3931e
--- /dev/null
+++ b/src/stdio/vwscanf.c_
@@ -0,0 +1,10 @@
+#include <stdio.h>
+#include <stdarg.h>
+#include <wchar.h>
+
+int vwscanf(const wchar_t *restrict fmt, va_list ap)
+{
+	return vfwscanf(stdin, fmt, ap);
+}
+
+weak_alias(vwscanf,__isoc99_vwscanf);
diff --git a/src/stdio/wprintf.c_ b/src/stdio/wprintf.c_
new file mode 100644
index 0000000..342cd97
--- /dev/null
+++ b/src/stdio/wprintf.c_
@@ -0,0 +1,13 @@
+#include <stdio.h>
+#include <stdarg.h>
+#include <wchar.h>
+
+int wprintf(const wchar_t *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vwprintf(fmt, ap);
+	va_end(ap);
+	return ret;
+}
diff --git a/src/stdio/wscanf.c_ b/src/stdio/wscanf.c_
new file mode 100644
index 0000000..4dfec25
--- /dev/null
+++ b/src/stdio/wscanf.c_
@@ -0,0 +1,15 @@
+#include <stdio.h>
+#include <stdarg.h>
+#include <wchar.h>
+
+int wscanf(const wchar_t *restrict fmt, ...)
+{
+	int ret;
+	va_list ap;
+	va_start(ap, fmt);
+	ret = vwscanf(fmt, ap);
+	va_end(ap);
+	return ret;
+}
+
+weak_alias(wscanf,__isoc99_wscanf);
diff --git a/src/string/strnlen.c b/src/string/strnlen.c
new file mode 100644
index 0000000..6442eb7
--- /dev/null
+++ b/src/string/strnlen.c
@@ -0,0 +1,7 @@
+#include <string.h>
+
+size_t strnlen(const char *s, size_t n)
+{
+	const char *p = memchr(s, 0, n);
+	return p ? p-s : n;
+}
diff --git a/src/string/wcschr.c b/src/string/wcschr.c
new file mode 100644
index 0000000..8dfc2f3
--- /dev/null
+++ b/src/string/wcschr.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wcschr(const wchar_t *s, wchar_t c)
+{
+	if (!c) return (wchar_t *)s + wcslen(s);
+	for (; *s && *s != c; s++);
+	return *s ? (wchar_t *)s : 0;
+}
diff --git a/src/string/wcslen.c b/src/string/wcslen.c
new file mode 100644
index 0000000..1b7b665
--- /dev/null
+++ b/src/string/wcslen.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+size_t wcslen(const wchar_t *s)
+{
+	const wchar_t *a;
+	for (a=s; *s; s++);
+	return s-a;
+}
diff --git a/src/string/wcsrchr.c b/src/string/wcsrchr.c
new file mode 100644
index 0000000..8961b9e
--- /dev/null
+++ b/src/string/wcsrchr.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wcsrchr(const wchar_t *s, wchar_t c)
+{
+	const wchar_t *p;
+	for (p=s+wcslen(s); p>=s && *p!=c; p--);
+	return p>=s ? (wchar_t *)p : 0;
+}
diff --git a/src/temp/__randname.c b/src/temp/__randname.c
new file mode 100644
index 0000000..a719fec
--- /dev/null
+++ b/src/temp/__randname.c
@@ -0,0 +1,20 @@
+#include <stdint.h>
+#include <nxbase.h>
+
+/* This assumes that a check for the
+   template size has already been made */
+char *__randname(char *template)
+{
+	int i;
+    NX_TimeVal tv;
+	unsigned long r;
+
+    tv = NX_ClockGetMillisecond();
+
+	r = ((tv*65537) ^ ((uintptr_t)&tv)) / 16 + (uintptr_t)template;
+
+	for (i=0; i<6; i++, r>>=5)
+		template[i] = 'A'+(r&15)+(r&16)*2;
+
+	return template;
+}
-- 
Gitee


From a4a3833c85aa4de4b8a60f6b7471b62e12f3e9b0 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Fri, 2 Sep 2022 00:22:59 +0800
Subject: [PATCH 02/13] feat: add complex and fenv

---
 src/Makefile                          |   4 +
 src/arch/riscv64/include/bits/fenv.h  |  17 ++
 src/arch/riscv64/include/bits/float.h |  16 ++
 src/arch/x86/include/bits/fenv.h      |  33 ++++
 src/arch/x86/include/bits/float.h     |  20 ++
 src/complex/__cexp.c                  |  87 +++++++++
 src/complex/__cexpf.c                 |  68 +++++++
 src/complex/cabs.c                    |   6 +
 src/complex/cabsf.c                   |   6 +
 src/complex/cabsl.c                   |  13 ++
 src/complex/cacos.c                   |  11 ++
 src/complex/cacosf.c                  |   9 +
 src/complex/cacosh.c                  |  12 ++
 src/complex/cacoshf.c                 |  10 +
 src/complex/cacoshl.c                 |  17 ++
 src/complex/cacosl.c                  |  16 ++
 src/complex/carg.c                    |   6 +
 src/complex/cargf.c                   |   6 +
 src/complex/cargl.c                   |  13 ++
 src/complex/casin.c                   |  17 ++
 src/complex/casinf.c                  |  15 ++
 src/complex/casinh.c                  |   9 +
 src/complex/casinhf.c                 |   7 +
 src/complex/casinhl.c                 |  14 ++
 src/complex/casinl.c                  |  21 +++
 src/complex/catan.c                   | 107 +++++++++++
 src/complex/catanf.c                  | 103 ++++++++++
 src/complex/catanh.c                  |   9 +
 src/complex/catanhf.c                 |   7 +
 src/complex/catanhl.c                 |  14 ++
 src/complex/catanl.c                  | 114 ++++++++++++
 src/complex/ccos.c                    |   8 +
 src/complex/ccosf.c                   |   6 +
 src/complex/ccosh.c                   | 140 ++++++++++++++
 src/complex/ccoshf.c                  |  90 +++++++++
 src/complex/ccoshl.c                  |   7 +
 src/complex/ccosl.c                   |  13 ++
 src/complex/cexp.c                    |  83 +++++++++
 src/complex/cexpf.c                   |  83 +++++++++
 src/complex/cexpl.c                   |   7 +
 src/complex/cimag.c                   |   6 +
 src/complex/cimagf.c                  |   6 +
 src/complex/cimagl.c                  |   6 +
 src/complex/clog.c                    |  14 ++
 src/complex/clogf.c                   |  12 ++
 src/complex/clogl.c                   |  18 ++
 src/complex/conj.c                    |   6 +
 src/complex/conjf.c                   |   6 +
 src/complex/conjl.c                   |   6 +
 src/complex/cpow.c                    |   8 +
 src/complex/cpowf.c                   |   6 +
 src/complex/cpowl.c                   |  13 ++
 src/complex/cproj.c                   |   8 +
 src/complex/cprojf.c                  |   8 +
 src/complex/cprojl.c                  |  15 ++
 src/complex/creal.c                   |   6 +
 src/complex/crealf.c                  |   6 +
 src/complex/creall.c                  |   6 +
 src/complex/csin.c                    |   9 +
 src/complex/csinf.c                   |   7 +
 src/complex/csinh.c                   | 141 ++++++++++++++
 src/complex/csinhf.c                  |  90 +++++++++
 src/complex/csinhl.c                  |   7 +
 src/complex/csinl.c                   |  14 ++
 src/complex/csqrt.c                   | 100 ++++++++++
 src/complex/csqrtf.c                  |  81 ++++++++
 src/complex/csqrtl.c                  |   7 +
 src/complex/ctan.c                    |   9 +
 src/complex/ctanf.c                   |   7 +
 src/complex/ctanh.c                   | 129 +++++++++++++
 src/complex/ctanhf.c                  |  66 +++++++
 src/complex/ctanhl.c                  |   7 +
 src/complex/ctanl.c                   |  14 ++
 src/fenv/__flt_rounds.c               |  19 ++
 src/fenv/fegetexceptflag.c            |   7 +
 src/fenv/feholdexcept.c               |   8 +
 src/fenv/fenv.c                       |  38 ++++
 src/fenv/fesetexceptflag.c            |   8 +
 src/fenv/fesetround.c                 |  23 +++
 src/fenv/feupdateenv.c                |   9 +
 src/fenv/riscv64/fenv-sf.c            |   3 +
 src/fenv/riscv64/fenv.S               |  56 ++++++
 src/fenv/x86/fenv.S                   | 164 ++++++++++++++++
 src/include/complex.h                 | 120 ++++++++++++
 src/include/fenv.h                    |  28 +++
 src/include/float.h                   |  52 ++++++
 src/include/math.h                    |  10 +
 src/internal/complex_impl.h           |  22 +++
 src/internal/libm.h                   |  13 ++
 src/math/__math_invalidl.c            |   9 +
 src/math/hypotl.c                     |  66 +++++++
 src/math/sqrt_data.c                  |  19 ++
 src/math/sqrt_data.h                  |  13 ++
 src/math/sqrtl.c                      | 259 ++++++++++++++++++++++++++
 94 files changed, 3034 insertions(+)
 create mode 100644 src/arch/riscv64/include/bits/fenv.h
 create mode 100644 src/arch/riscv64/include/bits/float.h
 create mode 100644 src/arch/x86/include/bits/fenv.h
 create mode 100644 src/arch/x86/include/bits/float.h
 create mode 100644 src/complex/__cexp.c
 create mode 100644 src/complex/__cexpf.c
 create mode 100644 src/complex/cabs.c
 create mode 100644 src/complex/cabsf.c
 create mode 100644 src/complex/cabsl.c
 create mode 100644 src/complex/cacos.c
 create mode 100644 src/complex/cacosf.c
 create mode 100644 src/complex/cacosh.c
 create mode 100644 src/complex/cacoshf.c
 create mode 100644 src/complex/cacoshl.c
 create mode 100644 src/complex/cacosl.c
 create mode 100644 src/complex/carg.c
 create mode 100644 src/complex/cargf.c
 create mode 100644 src/complex/cargl.c
 create mode 100644 src/complex/casin.c
 create mode 100644 src/complex/casinf.c
 create mode 100644 src/complex/casinh.c
 create mode 100644 src/complex/casinhf.c
 create mode 100644 src/complex/casinhl.c
 create mode 100644 src/complex/casinl.c
 create mode 100644 src/complex/catan.c
 create mode 100644 src/complex/catanf.c
 create mode 100644 src/complex/catanh.c
 create mode 100644 src/complex/catanhf.c
 create mode 100644 src/complex/catanhl.c
 create mode 100644 src/complex/catanl.c
 create mode 100644 src/complex/ccos.c
 create mode 100644 src/complex/ccosf.c
 create mode 100644 src/complex/ccosh.c
 create mode 100644 src/complex/ccoshf.c
 create mode 100644 src/complex/ccoshl.c
 create mode 100644 src/complex/ccosl.c
 create mode 100644 src/complex/cexp.c
 create mode 100644 src/complex/cexpf.c
 create mode 100644 src/complex/cexpl.c
 create mode 100644 src/complex/cimag.c
 create mode 100644 src/complex/cimagf.c
 create mode 100644 src/complex/cimagl.c
 create mode 100644 src/complex/clog.c
 create mode 100644 src/complex/clogf.c
 create mode 100644 src/complex/clogl.c
 create mode 100644 src/complex/conj.c
 create mode 100644 src/complex/conjf.c
 create mode 100644 src/complex/conjl.c
 create mode 100644 src/complex/cpow.c
 create mode 100644 src/complex/cpowf.c
 create mode 100644 src/complex/cpowl.c
 create mode 100644 src/complex/cproj.c
 create mode 100644 src/complex/cprojf.c
 create mode 100644 src/complex/cprojl.c
 create mode 100644 src/complex/creal.c
 create mode 100644 src/complex/crealf.c
 create mode 100644 src/complex/creall.c
 create mode 100644 src/complex/csin.c
 create mode 100644 src/complex/csinf.c
 create mode 100644 src/complex/csinh.c
 create mode 100644 src/complex/csinhf.c
 create mode 100644 src/complex/csinhl.c
 create mode 100644 src/complex/csinl.c
 create mode 100644 src/complex/csqrt.c
 create mode 100644 src/complex/csqrtf.c
 create mode 100644 src/complex/csqrtl.c
 create mode 100644 src/complex/ctan.c
 create mode 100644 src/complex/ctanf.c
 create mode 100644 src/complex/ctanh.c
 create mode 100644 src/complex/ctanhf.c
 create mode 100644 src/complex/ctanhl.c
 create mode 100644 src/complex/ctanl.c
 create mode 100644 src/fenv/__flt_rounds.c
 create mode 100644 src/fenv/fegetexceptflag.c
 create mode 100644 src/fenv/feholdexcept.c
 create mode 100644 src/fenv/fenv.c
 create mode 100644 src/fenv/fesetexceptflag.c
 create mode 100644 src/fenv/fesetround.c
 create mode 100644 src/fenv/feupdateenv.c
 create mode 100644 src/fenv/riscv64/fenv-sf.c
 create mode 100644 src/fenv/riscv64/fenv.S
 create mode 100644 src/fenv/x86/fenv.S
 create mode 100644 src/include/complex.h
 create mode 100644 src/include/fenv.h
 create mode 100644 src/include/float.h
 create mode 100644 src/internal/complex_impl.h
 create mode 100644 src/math/__math_invalidl.c
 create mode 100644 src/math/hypotl.c
 create mode 100644 src/math/sqrt_data.c
 create mode 100644 src/math/sqrt_data.h
 create mode 100644 src/math/sqrtl.c

diff --git a/src/Makefile b/src/Makefile
index ce4073e..5f6399f 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -43,6 +43,10 @@ X_LIBDIRS	:= $(LIBS_DIR)
 # we must link nxbase lib.
 X_LIBS		+= libnxbase.a
 
+SRC			+= complex/
+SRC			+= fenv/
+SRC			+= fenv/$(ARCH)/
+SRC			+= fenv/$(ARCH)/*.S
 SRC			+= stdio/
 SRC			+= locale/
 SRC			+= temp/
diff --git a/src/arch/riscv64/include/bits/fenv.h b/src/arch/riscv64/include/bits/fenv.h
new file mode 100644
index 0000000..806ec40
--- /dev/null
+++ b/src/arch/riscv64/include/bits/fenv.h
@@ -0,0 +1,17 @@
+#define FE_INVALID      16
+#define FE_DIVBYZERO    8
+#define FE_OVERFLOW     4
+#define FE_UNDERFLOW    2
+#define FE_INEXACT      1
+
+#define FE_ALL_EXCEPT   31
+
+#define FE_TONEAREST    0
+#define FE_DOWNWARD     2
+#define FE_UPWARD       3
+#define FE_TOWARDZERO   1
+
+typedef unsigned int fexcept_t;
+typedef unsigned int fenv_t;
+
+#define FE_DFL_ENV      ((const fenv_t *) -1)
diff --git a/src/arch/riscv64/include/bits/float.h b/src/arch/riscv64/include/bits/float.h
new file mode 100644
index 0000000..719c790
--- /dev/null
+++ b/src/arch/riscv64/include/bits/float.h
@@ -0,0 +1,16 @@
+#define FLT_EVAL_METHOD 0
+
+#define LDBL_TRUE_MIN 6.47517511943802511092443895822764655e-4966L
+#define LDBL_MIN 3.36210314311209350626267781732175260e-4932L
+#define LDBL_MAX 1.18973149535723176508575932662800702e+4932L
+#define LDBL_EPSILON 1.92592994438723585305597794258492732e-34L
+
+#define LDBL_MANT_DIG 113
+#define LDBL_MIN_EXP (-16381)
+#define LDBL_MAX_EXP 16384
+
+#define LDBL_DIG 33
+#define LDBL_MIN_10_EXP (-4931)
+#define LDBL_MAX_10_EXP 4932
+
+#define DECIMAL_DIG 36
diff --git a/src/arch/x86/include/bits/fenv.h b/src/arch/x86/include/bits/fenv.h
new file mode 100644
index 0000000..4430009
--- /dev/null
+++ b/src/arch/x86/include/bits/fenv.h
@@ -0,0 +1,33 @@
+#define FE_INVALID    1
+#define __FE_DENORM   2
+#define FE_DIVBYZERO  4
+#define FE_OVERFLOW   8
+#define FE_UNDERFLOW  16
+#define FE_INEXACT    32
+
+#define FE_ALL_EXCEPT 63
+
+#define FE_TONEAREST  0
+#define FE_DOWNWARD   0x400
+#define FE_UPWARD     0x800
+#define FE_TOWARDZERO 0xc00
+
+typedef unsigned short fexcept_t;
+
+typedef struct {
+	unsigned short __control_word;
+	unsigned short __unused1;
+	unsigned short __status_word;
+	unsigned short __unused2;
+	unsigned short __tags;
+	unsigned short __unused3;
+	unsigned int __eip;
+	unsigned short __cs_selector;
+	unsigned int __opcode:11;
+	unsigned int __unused4:5;
+	unsigned int __data_offset;
+	unsigned short __data_selector;
+	unsigned short __unused5;
+} fenv_t;
+
+#define FE_DFL_ENV      ((const fenv_t *) -1)
diff --git a/src/arch/x86/include/bits/float.h b/src/arch/x86/include/bits/float.h
new file mode 100644
index 0000000..dd6e402
--- /dev/null
+++ b/src/arch/x86/include/bits/float.h
@@ -0,0 +1,20 @@
+#ifdef __FLT_EVAL_METHOD__
+#define FLT_EVAL_METHOD __FLT_EVAL_METHOD__
+#else
+#define FLT_EVAL_METHOD 2
+#endif
+
+#define LDBL_TRUE_MIN 3.6451995318824746025e-4951L
+#define LDBL_MIN     3.3621031431120935063e-4932L
+#define LDBL_MAX     1.1897314953572317650e+4932L
+#define LDBL_EPSILON 1.0842021724855044340e-19L
+
+#define LDBL_MANT_DIG 64
+#define LDBL_MIN_EXP (-16381)
+#define LDBL_MAX_EXP 16384
+
+#define LDBL_DIG 18
+#define LDBL_MIN_10_EXP (-4931)
+#define LDBL_MAX_10_EXP 4932
+
+#define DECIMAL_DIG 21
diff --git a/src/complex/__cexp.c b/src/complex/__cexp.c
new file mode 100644
index 0000000..003d20a
--- /dev/null
+++ b/src/complex/__cexp.c
@@ -0,0 +1,87 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/k_exp.c */
+/*-
+ * Copyright (c) 2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "complex_impl.h"
+
+static const uint32_t k = 1799; /* constant for reduction */
+static const double kln2 = 1246.97177782734161156; /* k * ln2 */
+
+/*
+ * Compute exp(x), scaled to avoid spurious overflow.  An exponent is
+ * returned separately in 'expt'.
+ *
+ * Input:  ln(DBL_MAX) <= x < ln(2 * DBL_MAX / DBL_MIN_DENORM) ~= 1454.91
+ * Output: 2**1023 <= y < 2**1024
+ */
+static double __frexp_exp(double x, int *expt)
+{
+	double exp_x;
+	uint32_t hx;
+
+	/*
+	 * We use exp(x) = exp(x - kln2) * 2**k, carefully chosen to
+	 * minimize |exp(kln2) - 2**k|.  We also scale the exponent of
+	 * exp_x to MAX_EXP so that the result can be multiplied by
+	 * a tiny number without losing accuracy due to denormalization.
+	 */
+	exp_x = exp(x - kln2);
+	GET_HIGH_WORD(hx, exp_x);
+	*expt = (hx >> 20) - (0x3ff + 1023) + k;
+	SET_HIGH_WORD(exp_x, (hx & 0xfffff) | ((0x3ff + 1023) << 20));
+	return exp_x;
+}
+
+/*
+ * __ldexp_cexp(x, expt) compute exp(x) * 2**expt.
+ * It is intended for large arguments (real part >= ln(DBL_MAX))
+ * where care is needed to avoid overflow.
+ *
+ * The present implementation is narrowly tailored for our hyperbolic and
+ * exponential functions.  We assume expt is small (0 or -1), and the caller
+ * has filtered out very large x, for which overflow would be inevitable.
+ */
+double complex __ldexp_cexp(double complex z, int expt)
+{
+	double x, y, exp_x, scale1, scale2;
+	int ex_expt, half_expt;
+
+	x = creal(z);
+	y = cimag(z);
+	exp_x = __frexp_exp(x, &ex_expt);
+	expt += ex_expt;
+
+	/*
+	 * Arrange so that scale1 * scale2 == 2**expt.  We use this to
+	 * compensate for scalbn being horrendously slow.
+	 */
+	half_expt = expt / 2;
+	INSERT_WORDS(scale1, (0x3ff + half_expt) << 20, 0);
+	half_expt = expt - half_expt;
+	INSERT_WORDS(scale2, (0x3ff + half_expt) << 20, 0);
+
+	return CMPLX(cos(y) * exp_x * scale1 * scale2, sin(y) * exp_x * scale1 * scale2);
+}
diff --git a/src/complex/__cexpf.c b/src/complex/__cexpf.c
new file mode 100644
index 0000000..ee5ff2b
--- /dev/null
+++ b/src/complex/__cexpf.c
@@ -0,0 +1,68 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/k_expf.c */
+/*-
+ * Copyright (c) 2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "complex_impl.h"
+
+static const uint32_t k = 235; /* constant for reduction */
+static const float kln2 = 162.88958740F; /* k * ln2 */
+
+/*
+ * See __cexp.c for details.
+ *
+ * Input:  ln(FLT_MAX) <= x < ln(2 * FLT_MAX / FLT_MIN_DENORM) ~= 192.7
+ * Output: 2**127 <= y < 2**128
+ */
+static float __frexp_expf(float x, int *expt)
+{
+	float exp_x;
+	uint32_t hx;
+
+	exp_x = expf(x - kln2);
+	GET_FLOAT_WORD(hx, exp_x);
+	*expt = (hx >> 23) - (0x7f + 127) + k;
+	SET_FLOAT_WORD(exp_x, (hx & 0x7fffff) | ((0x7f + 127) << 23));
+	return exp_x;
+}
+
+float complex __ldexp_cexpf(float complex z, int expt)
+{
+	float x, y, exp_x, scale1, scale2;
+	int ex_expt, half_expt;
+
+	x = crealf(z);
+	y = cimagf(z);
+	exp_x = __frexp_expf(x, &ex_expt);
+	expt += ex_expt;
+
+	half_expt = expt / 2;
+	SET_FLOAT_WORD(scale1, (0x7f + half_expt) << 23);
+	half_expt = expt - half_expt;
+	SET_FLOAT_WORD(scale2, (0x7f + half_expt) << 23);
+
+	return CMPLXF(cosf(y) * exp_x * scale1 * scale2,
+	  sinf(y) * exp_x * scale1 * scale2);
+}
diff --git a/src/complex/cabs.c b/src/complex/cabs.c
new file mode 100644
index 0000000..c5ad58a
--- /dev/null
+++ b/src/complex/cabs.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+double cabs(double complex z)
+{
+	return hypot(creal(z), cimag(z));
+}
diff --git a/src/complex/cabsf.c b/src/complex/cabsf.c
new file mode 100644
index 0000000..619f28d
--- /dev/null
+++ b/src/complex/cabsf.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+float cabsf(float complex z)
+{
+	return hypotf(crealf(z), cimagf(z));
+}
diff --git a/src/complex/cabsl.c b/src/complex/cabsl.c
new file mode 100644
index 0000000..d37e3f2
--- /dev/null
+++ b/src/complex/cabsl.c
@@ -0,0 +1,13 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double cabsl(long double complex z)
+{
+	return cabs(z);
+}
+#else
+long double cabsl(long double complex z)
+{
+	return hypotl(creall(z), cimagl(z));
+}
+#endif
diff --git a/src/complex/cacos.c b/src/complex/cacos.c
new file mode 100644
index 0000000..c39d257
--- /dev/null
+++ b/src/complex/cacos.c
@@ -0,0 +1,11 @@
+#include "complex_impl.h"
+
+// FIXME: Hull et al. "Implementing the complex arcsine and arccosine functions using exception handling" 1997
+
+/* acos(z) = pi/2 - asin(z) */
+
+double complex cacos(double complex z)
+{
+	z = casin(z);
+	return CMPLX(M_PI_2 - creal(z), -cimag(z));
+}
diff --git a/src/complex/cacosf.c b/src/complex/cacosf.c
new file mode 100644
index 0000000..2e04854
--- /dev/null
+++ b/src/complex/cacosf.c
@@ -0,0 +1,9 @@
+#include "complex_impl.h"
+
+// FIXME
+
+float complex cacosf(float complex z)
+{
+	z = casinf(z);
+	return CMPLXF((float)M_PI_2 - crealf(z), -cimagf(z));
+}
diff --git a/src/complex/cacosh.c b/src/complex/cacosh.c
new file mode 100644
index 0000000..76127f7
--- /dev/null
+++ b/src/complex/cacosh.c
@@ -0,0 +1,12 @@
+#include "complex_impl.h"
+
+/* acosh(z) = i acos(z) */
+
+double complex cacosh(double complex z)
+{
+	int zineg = signbit(cimag(z));
+
+	z = cacos(z);
+	if (zineg) return CMPLX(cimag(z), -creal(z));
+	else       return CMPLX(-cimag(z), creal(z));
+}
diff --git a/src/complex/cacoshf.c b/src/complex/cacoshf.c
new file mode 100644
index 0000000..8bd8058
--- /dev/null
+++ b/src/complex/cacoshf.c
@@ -0,0 +1,10 @@
+#include "complex_impl.h"
+
+float complex cacoshf(float complex z)
+{
+	int zineg = signbit(cimagf(z));
+
+	z = cacosf(z);
+	if (zineg) return CMPLXF(cimagf(z), -crealf(z));
+	else       return CMPLXF(-cimagf(z), crealf(z));
+}
diff --git a/src/complex/cacoshl.c b/src/complex/cacoshl.c
new file mode 100644
index 0000000..3a284be
--- /dev/null
+++ b/src/complex/cacoshl.c
@@ -0,0 +1,17 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex cacoshl(long double complex z)
+{
+	return cacosh(z);
+}
+#else
+long double complex cacoshl(long double complex z)
+{
+	int zineg = signbit(cimagl(z));
+
+	z = cacosl(z);
+	if (zineg) return CMPLXL(cimagl(z), -creall(z));
+	else       return CMPLXL(-cimagl(z), creall(z));
+}
+#endif
diff --git a/src/complex/cacosl.c b/src/complex/cacosl.c
new file mode 100644
index 0000000..cc20dcd
--- /dev/null
+++ b/src/complex/cacosl.c
@@ -0,0 +1,16 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex cacosl(long double complex z)
+{
+	return cacos(z);
+}
+#else
+// FIXME
+#define PI_2 1.57079632679489661923132169163975144L
+long double complex cacosl(long double complex z)
+{
+	z = casinl(z);
+	return CMPLXL(PI_2 - creall(z), -cimagl(z));
+}
+#endif
diff --git a/src/complex/carg.c b/src/complex/carg.c
new file mode 100644
index 0000000..dfe9b97
--- /dev/null
+++ b/src/complex/carg.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+double carg(double complex z)
+{
+	return atan2(cimag(z), creal(z));
+}
diff --git a/src/complex/cargf.c b/src/complex/cargf.c
new file mode 100644
index 0000000..9a6c19b
--- /dev/null
+++ b/src/complex/cargf.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+float cargf(float complex z)
+{
+	return atan2f(cimagf(z), crealf(z));
+}
diff --git a/src/complex/cargl.c b/src/complex/cargl.c
new file mode 100644
index 0000000..88f95f9
--- /dev/null
+++ b/src/complex/cargl.c
@@ -0,0 +1,13 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double cargl(long double complex z)
+{
+	return carg(z);
+}
+#else
+long double cargl(long double complex z)
+{
+	return atan2l(cimagl(z), creall(z));
+}
+#endif
diff --git a/src/complex/casin.c b/src/complex/casin.c
new file mode 100644
index 0000000..3244beb
--- /dev/null
+++ b/src/complex/casin.c
@@ -0,0 +1,17 @@
+#include "complex_impl.h"
+
+// FIXME
+
+/* asin(z) = -i log(i z + sqrt(1 - z*z)) */
+
+double complex casin(double complex z)
+{
+	double complex w;
+	double x, y;
+
+	x = creal(z);
+	y = cimag(z);
+	w = CMPLX(1.0 - (x - y)*(x + y), -2.0*x*y);
+	double complex r = clog(CMPLX(-y, x) + csqrt(w));
+	return CMPLX(cimag(r), -creal(r));
+}
diff --git a/src/complex/casinf.c b/src/complex/casinf.c
new file mode 100644
index 0000000..2cda2f0
--- /dev/null
+++ b/src/complex/casinf.c
@@ -0,0 +1,15 @@
+#include "complex_impl.h"
+
+// FIXME
+
+float complex casinf(float complex z)
+{
+	float complex w;
+	float x, y;
+
+	x = crealf(z);
+	y = cimagf(z);
+	w = CMPLXF(1.0 - (x - y)*(x + y), -2.0*x*y);
+	float complex r = clogf(CMPLXF(-y, x) + csqrtf(w));
+	return CMPLXF(cimagf(r), -crealf(r));
+}
diff --git a/src/complex/casinh.c b/src/complex/casinh.c
new file mode 100644
index 0000000..50bf27c
--- /dev/null
+++ b/src/complex/casinh.c
@@ -0,0 +1,9 @@
+#include "complex_impl.h"
+
+/* asinh(z) = -i asin(i z) */
+
+double complex casinh(double complex z)
+{
+	z = casin(CMPLX(-cimag(z), creal(z)));
+	return CMPLX(cimag(z), -creal(z));
+}
diff --git a/src/complex/casinhf.c b/src/complex/casinhf.c
new file mode 100644
index 0000000..93d82e5
--- /dev/null
+++ b/src/complex/casinhf.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+float complex casinhf(float complex z)
+{
+	z = casinf(CMPLXF(-cimagf(z), crealf(z)));
+	return CMPLXF(cimagf(z), -crealf(z));
+}
diff --git a/src/complex/casinhl.c b/src/complex/casinhl.c
new file mode 100644
index 0000000..68ba3dd
--- /dev/null
+++ b/src/complex/casinhl.c
@@ -0,0 +1,14 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex casinhl(long double complex z)
+{
+	return casinh(z);
+}
+#else
+long double complex casinhl(long double complex z)
+{
+	z = casinl(CMPLXL(-cimagl(z), creall(z)));
+	return CMPLXL(cimagl(z), -creall(z));
+}
+#endif
diff --git a/src/complex/casinl.c b/src/complex/casinl.c
new file mode 100644
index 0000000..072adc4
--- /dev/null
+++ b/src/complex/casinl.c
@@ -0,0 +1,21 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex casinl(long double complex z)
+{
+	return casin(z);
+}
+#else
+// FIXME
+long double complex casinl(long double complex z)
+{
+	long double complex w;
+	long double x, y;
+
+	x = creall(z);
+	y = cimagl(z);
+	w = CMPLXL(1.0 - (x - y)*(x + y), -2.0*x*y);
+	long double complex r = clogl(CMPLXL(-y, x) + csqrtl(w));
+	return CMPLXL(cimagl(r), -creall(r));
+}
+#endif
diff --git a/src/complex/catan.c b/src/complex/catan.c
new file mode 100644
index 0000000..ccc2fb5
--- /dev/null
+++ b/src/complex/catan.c
@@ -0,0 +1,107 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/s_catan.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Complex circular arc tangent
+ *
+ *
+ * SYNOPSIS:
+ *
+ * double complex catan();
+ * double complex z, w;
+ *
+ * w = catan (z);
+ *
+ *
+ * DESCRIPTION:
+ *
+ * If
+ *     z = x + iy,
+ *
+ * then
+ *          1       (    2x     )
+ * Re w  =  - arctan(-----------)  +  k PI
+ *          2       (     2    2)
+ *                  (1 - x  - y )
+ *
+ *               ( 2         2)
+ *          1    (x  +  (y+1) )
+ * Im w  =  - log(------------)
+ *          4    ( 2         2)
+ *               (x  +  (y-1) )
+ *
+ * Where k is an arbitrary integer.
+ *
+ * catan(z) = -i catanh(iz).
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    DEC       -10,+10      5900       1.3e-16     7.8e-18
+ *    IEEE      -10,+10     30000       2.3e-15     8.5e-17
+ * The check catan( ctan(z) )  =  z, with |x| and |y| < PI/2,
+ * had peak relative error 1.5e-16, rms relative error
+ * 2.9e-17.  See also clog().
+ */
+
+#include "complex_impl.h"
+
+#define MAXNUM 1.0e308
+
+static const double DP1 = 3.14159265160560607910E0;
+static const double DP2 = 1.98418714791870343106E-9;
+static const double DP3 = 1.14423774522196636802E-17;
+
+static double _redupi(double x)
+{
+	double t;
+	long i;
+
+	t = x/M_PI;
+	if (t >= 0.0)
+		t += 0.5;
+	else
+		t -= 0.5;
+
+	i = t;  /* the multiple */
+	t = i;
+	t = ((x - t * DP1) - t * DP2) - t * DP3;
+	return t;
+}
+
+double complex catan(double complex z)
+{
+	double complex w;
+	double a, t, x, x2, y;
+
+	x = creal(z);
+	y = cimag(z);
+
+	x2 = x * x;
+	a = 1.0 - x2 - (y * y);
+
+	t = 0.5 * atan2(2.0 * x, a);
+	w = _redupi(t);
+
+	t = y - 1.0;
+	a = x2 + (t * t);
+
+	t = y + 1.0;
+	a = (x2 + t * t)/a;
+	w = CMPLX(w, 0.25 * log(a));
+	return w;
+}
diff --git a/src/complex/catanf.c b/src/complex/catanf.c
new file mode 100644
index 0000000..ef3907a
--- /dev/null
+++ b/src/complex/catanf.c
@@ -0,0 +1,103 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/s_catanf.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Complex circular arc tangent
+ *
+ *
+ * SYNOPSIS:
+ *
+ * float complex catanf();
+ * float complex z, w;
+ *
+ * w = catanf( z );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * If
+ *     z = x + iy,
+ *
+ * then
+ *          1       (    2x     )
+ * Re w  =  - arctan(-----------)  +  k PI
+ *          2       (     2    2)
+ *                  (1 - x  - y )
+ *
+ *               ( 2         2)
+ *          1    (x  +  (y+1) )
+ * Im w  =  - log(------------)
+ *          4    ( 2         2)
+ *               (x  +  (y-1) )
+ *
+ * Where k is an arbitrary integer.
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE      -10,+10     30000        2.3e-6      5.2e-8
+ */
+
+#include "complex_impl.h"
+
+#define MAXNUMF 1.0e38F
+
+static const double DP1 = 3.140625;
+static const double DP2 = 9.67502593994140625E-4;
+static const double DP3 = 1.509957990978376432E-7;
+
+static float _redupif(float xx)
+{
+	float x, t;
+	long i;
+
+	x = xx;
+	t = x/(float)M_PI;
+	if (t >= 0.0f)
+		t += 0.5f;
+	else
+		t -= 0.5f;
+
+	i = t;  /* the multiple */
+	t = i;
+	t = ((x - t * DP1) - t * DP2) - t * DP3;
+	return t;
+}
+
+float complex catanf(float complex z)
+{
+	float complex w;
+	float a, t, x, x2, y;
+
+	x = crealf(z);
+	y = cimagf(z);
+
+	x2 = x * x;
+	a = 1.0f - x2 - (y * y);
+
+	t = 0.5f * atan2f(2.0f * x, a);
+	w = _redupif(t);
+
+	t = y - 1.0f;
+	a = x2 + (t * t);
+
+	t = y + 1.0f;
+	a = (x2 + (t * t))/a;
+	w = CMPLXF(w, 0.25f * logf(a));
+	return w;
+}
diff --git a/src/complex/catanh.c b/src/complex/catanh.c
new file mode 100644
index 0000000..c324c7f
--- /dev/null
+++ b/src/complex/catanh.c
@@ -0,0 +1,9 @@
+#include "complex_impl.h"
+
+/* atanh = -i atan(i z) */
+
+double complex catanh(double complex z)
+{
+	z = catan(CMPLX(-cimag(z), creal(z)));
+	return CMPLX(cimag(z), -creal(z));
+}
diff --git a/src/complex/catanhf.c b/src/complex/catanhf.c
new file mode 100644
index 0000000..b0505f6
--- /dev/null
+++ b/src/complex/catanhf.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+float complex catanhf(float complex z)
+{
+	z = catanf(CMPLXF(-cimagf(z), crealf(z)));
+	return CMPLXF(cimagf(z), -crealf(z));
+}
diff --git a/src/complex/catanhl.c b/src/complex/catanhl.c
new file mode 100644
index 0000000..6025c41
--- /dev/null
+++ b/src/complex/catanhl.c
@@ -0,0 +1,14 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex catanhl(long double complex z)
+{
+	return catanh(z);
+}
+#else
+long double complex catanhl(long double complex z)
+{
+	z = catanl(CMPLXL(-cimagl(z), creall(z)));
+	return CMPLXL(cimagl(z), -creall(z));
+}
+#endif
diff --git a/src/complex/catanl.c b/src/complex/catanl.c
new file mode 100644
index 0000000..e62526c
--- /dev/null
+++ b/src/complex/catanl.c
@@ -0,0 +1,114 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/s_catanl.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Complex circular arc tangent
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double complex catanl();
+ * long double complex z, w;
+ *
+ * w = catanl( z );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * If
+ *     z = x + iy,
+ *
+ * then
+ *          1       (    2x     )
+ * Re w  =  - arctan(-----------)  +  k PI
+ *          2       (     2    2)
+ *                  (1 - x  - y )
+ *
+ *               ( 2         2)
+ *          1    (x  +  (y+1) )
+ * Im w  =  - log(------------)
+ *          4    ( 2         2)
+ *               (x  +  (y-1) )
+ *
+ * Where k is an arbitrary integer.
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    DEC       -10,+10      5900       1.3e-16     7.8e-18
+ *    IEEE      -10,+10     30000       2.3e-15     8.5e-17
+ * The check catan( ctan(z) )  =  z, with |x| and |y| < PI/2,
+ * had peak relative error 1.5e-16, rms relative error
+ * 2.9e-17.  See also clog().
+ */
+
+#include <complex.h>
+#include <float.h>
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex catanl(long double complex z)
+{
+	return catan(z);
+}
+#else
+static const long double PIL = 3.141592653589793238462643383279502884197169L;
+static const long double DP1 = 3.14159265358979323829596852490908531763125L;
+static const long double DP2 = 1.6667485837041756656403424829301998703007e-19L;
+static const long double DP3 = 1.8830410776607851167459095484560349402753e-39L;
+
+static long double redupil(long double x)
+{
+	long double t;
+	long i;
+
+	t = x / PIL;
+	if (t >= 0.0L)
+		t += 0.5L;
+	else
+		t -= 0.5L;
+
+	i = t;  /* the multiple */
+	t = i;
+	t = ((x - t * DP1) - t * DP2) - t * DP3;
+	return t;
+}
+
+long double complex catanl(long double complex z)
+{
+	long double complex w;
+	long double a, t, x, x2, y;
+
+	x = creall(z);
+	y = cimagl(z);
+
+	x2 = x * x;
+	a = 1.0L - x2 - (y * y);
+
+	t = atan2l(2.0L * x, a) * 0.5L;
+	w = redupil(t);
+
+	t = y - 1.0L;
+	a = x2 + (t * t);
+
+	t = y + 1.0L;
+	a = (x2 + (t * t)) / a;
+	w = CMPLXF(w, 0.25L * logl(a));
+	return w;
+}
+#endif
diff --git a/src/complex/ccos.c b/src/complex/ccos.c
new file mode 100644
index 0000000..f32e1fa
--- /dev/null
+++ b/src/complex/ccos.c
@@ -0,0 +1,8 @@
+#include "complex_impl.h"
+
+/* cos(z) = cosh(i z) */
+
+double complex ccos(double complex z)
+{
+	return ccosh(CMPLX(-cimag(z), creal(z)));
+}
diff --git a/src/complex/ccosf.c b/src/complex/ccosf.c
new file mode 100644
index 0000000..490be9b
--- /dev/null
+++ b/src/complex/ccosf.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+float complex ccosf(float complex z)
+{
+	return ccoshf(CMPLXF(-cimagf(z), crealf(z)));
+}
diff --git a/src/complex/ccosh.c b/src/complex/ccosh.c
new file mode 100644
index 0000000..c995da7
--- /dev/null
+++ b/src/complex/ccosh.c
@@ -0,0 +1,140 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_ccosh.c */
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/*
+ * Hyperbolic cosine of a complex argument z = x + i y.
+ *
+ * cosh(z) = cosh(x+iy)
+ *         = cosh(x) cos(y) + i sinh(x) sin(y).
+ *
+ * Exceptional values are noted in the comments within the source code.
+ * These values and the return value were taken from n1124.pdf.
+ */
+
+#include "complex_impl.h"
+
+static const double huge = 0x1p1023;
+
+double complex ccosh(double complex z)
+{
+	double x, y, h;
+	int32_t hx, hy, ix, iy, lx, ly;
+
+	x = creal(z);
+	y = cimag(z);
+
+	EXTRACT_WORDS(hx, lx, x);
+	EXTRACT_WORDS(hy, ly, y);
+
+	ix = 0x7fffffff & hx;
+	iy = 0x7fffffff & hy;
+
+	/* Handle the nearly-non-exceptional cases where x and y are finite. */
+	if (ix < 0x7ff00000 && iy < 0x7ff00000) {
+		if ((iy | ly) == 0)
+			return CMPLX(cosh(x), x * y);
+		if (ix < 0x40360000)    /* small x: normal case */
+			return CMPLX(cosh(x) * cos(y), sinh(x) * sin(y));
+
+		/* |x| >= 22, so cosh(x) ~= exp(|x|) */
+		if (ix < 0x40862e42) {
+			/* x < 710: exp(|x|) won't overflow */
+			h = exp(fabs(x)) * 0.5;
+			return CMPLX(h * cos(y), copysign(h, x) * sin(y));
+		} else if (ix < 0x4096bbaa) {
+			/* x < 1455: scale to avoid overflow */
+			z = __ldexp_cexp(CMPLX(fabs(x), y), -1);
+			return CMPLX(creal(z), cimag(z) * copysign(1, x));
+		} else {
+			/* x >= 1455: the result always overflows */
+			h = huge * x;
+			return CMPLX(h * h * cos(y), h * sin(y));
+		}
+	}
+
+	/*
+	 * cosh(+-0 +- I Inf) = dNaN + I sign(d(+-0, dNaN))0.
+	 * The sign of 0 in the result is unspecified.  Choice = normally
+	 * the same as dNaN.  Raise the invalid floating-point exception.
+	 *
+	 * cosh(+-0 +- I NaN) = d(NaN) + I sign(d(+-0, NaN))0.
+	 * The sign of 0 in the result is unspecified.  Choice = normally
+	 * the same as d(NaN).
+	 */
+	if ((ix | lx) == 0 && iy >= 0x7ff00000)
+		return CMPLX(y - y, copysign(0, x * (y - y)));
+
+	/*
+	 * cosh(+-Inf +- I 0) = +Inf + I (+-)(+-)0.
+	 *
+	 * cosh(NaN +- I 0)   = d(NaN) + I sign(d(NaN, +-0))0.
+	 * The sign of 0 in the result is unspecified.
+	 */
+	if ((iy | ly) == 0 && ix >= 0x7ff00000) {
+		if (((hx & 0xfffff) | lx) == 0)
+			return CMPLX(x * x, copysign(0, x) * y);
+		return CMPLX(x * x, copysign(0, (x + x) * y));
+	}
+
+	/*
+	 * cosh(x +- I Inf) = dNaN + I dNaN.
+	 * Raise the invalid floating-point exception for finite nonzero x.
+	 *
+	 * cosh(x + I NaN) = d(NaN) + I d(NaN).
+	 * Optionally raises the invalid floating-point exception for finite
+	 * nonzero x.  Choice = don't raise (except for signaling NaNs).
+	 */
+	if (ix < 0x7ff00000 && iy >= 0x7ff00000)
+		return CMPLX(y - y, x * (y - y));
+
+	/*
+	 * cosh(+-Inf + I NaN)  = +Inf + I d(NaN).
+	 *
+	 * cosh(+-Inf +- I Inf) = +Inf + I dNaN.
+	 * The sign of Inf in the result is unspecified.  Choice = always +.
+	 * Raise the invalid floating-point exception.
+	 *
+	 * cosh(+-Inf + I y)   = +Inf cos(y) +- I Inf sin(y)
+	 */
+	if (ix >= 0x7ff00000 && ((hx & 0xfffff) | lx) == 0) {
+		if (iy >= 0x7ff00000)
+			return CMPLX(x * x, x * (y - y));
+		return CMPLX((x * x) * cos(y), x * sin(y));
+	}
+
+	/*
+	 * cosh(NaN + I NaN)  = d(NaN) + I d(NaN).
+	 *
+	 * cosh(NaN +- I Inf) = d(NaN) + I d(NaN).
+	 * Optionally raises the invalid floating-point exception.
+	 * Choice = raise.
+	 *
+	 * cosh(NaN + I y)    = d(NaN) + I d(NaN).
+	 * Optionally raises the invalid floating-point exception for finite
+	 * nonzero y.  Choice = don't raise (except for signaling NaNs).
+	 */
+	return CMPLX((x * x) * (y - y), (x + x) * (y - y));
+}
diff --git a/src/complex/ccoshf.c b/src/complex/ccoshf.c
new file mode 100644
index 0000000..189ce94
--- /dev/null
+++ b/src/complex/ccoshf.c
@@ -0,0 +1,90 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_ccoshf.c */
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/*
+ * Hyperbolic cosine of a complex argument.  See s_ccosh.c for details.
+ */
+
+#include "complex_impl.h"
+
+static const float huge = 0x1p127;
+
+float complex ccoshf(float complex z)
+{
+	float x, y, h;
+	int32_t hx, hy, ix, iy;
+
+	x = crealf(z);
+	y = cimagf(z);
+
+	GET_FLOAT_WORD(hx, x);
+	GET_FLOAT_WORD(hy, y);
+
+	ix = 0x7fffffff & hx;
+	iy = 0x7fffffff & hy;
+
+	if (ix < 0x7f800000 && iy < 0x7f800000) {
+		if (iy == 0)
+			return CMPLXF(coshf(x), x * y);
+		if (ix < 0x41100000)    /* small x: normal case */
+			return CMPLXF(coshf(x) * cosf(y), sinhf(x) * sinf(y));
+
+		/* |x| >= 9, so cosh(x) ~= exp(|x|) */
+		if (ix < 0x42b17218) {
+			/* x < 88.7: expf(|x|) won't overflow */
+			h = expf(fabsf(x)) * 0.5f;
+			return CMPLXF(h * cosf(y), copysignf(h, x) * sinf(y));
+		} else if (ix < 0x4340b1e7) {
+			/* x < 192.7: scale to avoid overflow */
+			z = __ldexp_cexpf(CMPLXF(fabsf(x), y), -1);
+			return CMPLXF(crealf(z), cimagf(z) * copysignf(1, x));
+		} else {
+			/* x >= 192.7: the result always overflows */
+			h = huge * x;
+			return CMPLXF(h * h * cosf(y), h * sinf(y));
+		}
+	}
+
+	if (ix == 0 && iy >= 0x7f800000)
+		return CMPLXF(y - y, copysignf(0, x * (y - y)));
+
+	if (iy == 0 && ix >= 0x7f800000) {
+		if ((hx & 0x7fffff) == 0)
+			return CMPLXF(x * x, copysignf(0, x) * y);
+		return CMPLXF(x * x, copysignf(0, (x + x) * y));
+	}
+
+	if (ix < 0x7f800000 && iy >= 0x7f800000)
+		return CMPLXF(y - y, x * (y - y));
+
+	if (ix >= 0x7f800000 && (hx & 0x7fffff) == 0) {
+		if (iy >= 0x7f800000)
+			return CMPLXF(x * x, x * (y - y));
+		return CMPLXF((x * x) * cosf(y), x * sinf(y));
+	}
+
+	return CMPLXF((x * x) * (y - y), (x + x) * (y - y));
+}
diff --git a/src/complex/ccoshl.c b/src/complex/ccoshl.c
new file mode 100644
index 0000000..ffb4d8a
--- /dev/null
+++ b/src/complex/ccoshl.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+//FIXME
+long double complex ccoshl(long double complex z)
+{
+	return ccosh(z);
+}
diff --git a/src/complex/ccosl.c b/src/complex/ccosl.c
new file mode 100644
index 0000000..2530006
--- /dev/null
+++ b/src/complex/ccosl.c
@@ -0,0 +1,13 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex ccosl(long double complex z)
+{
+	return ccos(z);
+}
+#else
+long double complex ccosl(long double complex z)
+{
+	return ccoshl(CMPLXL(-cimagl(z), creall(z)));
+}
+#endif
diff --git a/src/complex/cexp.c b/src/complex/cexp.c
new file mode 100644
index 0000000..7fb489b
--- /dev/null
+++ b/src/complex/cexp.c
@@ -0,0 +1,83 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_cexp.c */
+/*-
+ * Copyright (c) 2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "complex_impl.h"
+
+static const uint32_t
+exp_ovfl  = 0x40862e42,  /* high bits of MAX_EXP * ln2 ~= 710 */
+cexp_ovfl = 0x4096b8e4;  /* (MAX_EXP - MIN_DENORM_EXP) * ln2 */
+
+double complex cexp(double complex z)
+{
+	double x, y, exp_x;
+	uint32_t hx, hy, lx, ly;
+
+	x = creal(z);
+	y = cimag(z);
+
+	EXTRACT_WORDS(hy, ly, y);
+	hy &= 0x7fffffff;
+
+	/* cexp(x + I 0) = exp(x) + I 0 */
+	if ((hy | ly) == 0)
+		return CMPLX(exp(x), y);
+	EXTRACT_WORDS(hx, lx, x);
+	/* cexp(0 + I y) = cos(y) + I sin(y) */
+	if (((hx & 0x7fffffff) | lx) == 0)
+		return CMPLX(cos(y), sin(y));
+
+	if (hy >= 0x7ff00000) {
+		if (lx != 0 || (hx & 0x7fffffff) != 0x7ff00000) {
+			/* cexp(finite|NaN +- I Inf|NaN) = NaN + I NaN */
+			return CMPLX(y - y, y - y);
+		} else if (hx & 0x80000000) {
+			/* cexp(-Inf +- I Inf|NaN) = 0 + I 0 */
+			return CMPLX(0.0, 0.0);
+		} else {
+			/* cexp(+Inf +- I Inf|NaN) = Inf + I NaN */
+			return CMPLX(x, y - y);
+		}
+	}
+
+	if (hx >= exp_ovfl && hx <= cexp_ovfl) {
+		/*
+		 * x is between 709.7 and 1454.3, so we must scale to avoid
+		 * overflow in exp(x).
+		 */
+		return __ldexp_cexp(z, 0);
+	} else {
+		/*
+		 * Cases covered here:
+		 *  -  x < exp_ovfl and exp(x) won't overflow (common case)
+		 *  -  x > cexp_ovfl, so exp(x) * s overflows for all s > 0
+		 *  -  x = +-Inf (generated by exp())
+		 *  -  x = NaN (spurious inexact exception from y)
+		 */
+		exp_x = exp(x);
+		return CMPLX(exp_x * cos(y), exp_x * sin(y));
+	}
+}
diff --git a/src/complex/cexpf.c b/src/complex/cexpf.c
new file mode 100644
index 0000000..00d258f
--- /dev/null
+++ b/src/complex/cexpf.c
@@ -0,0 +1,83 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_cexpf.c */
+/*-
+ * Copyright (c) 2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "complex_impl.h"
+
+static const uint32_t
+exp_ovfl  = 0x42b17218,  /* MAX_EXP * ln2 ~= 88.722839355 */
+cexp_ovfl = 0x43400074;  /* (MAX_EXP - MIN_DENORM_EXP) * ln2 */
+
+float complex cexpf(float complex z)
+{
+	float x, y, exp_x;
+	uint32_t hx, hy;
+
+	x = crealf(z);
+	y = cimagf(z);
+
+	GET_FLOAT_WORD(hy, y);
+	hy &= 0x7fffffff;
+
+	/* cexp(x + I 0) = exp(x) + I 0 */
+	if (hy == 0)
+		return CMPLXF(expf(x), y);
+	GET_FLOAT_WORD(hx, x);
+	/* cexp(0 + I y) = cos(y) + I sin(y) */
+	if ((hx & 0x7fffffff) == 0)
+		return CMPLXF(cosf(y), sinf(y));
+
+	if (hy >= 0x7f800000) {
+		if ((hx & 0x7fffffff) != 0x7f800000) {
+			/* cexp(finite|NaN +- I Inf|NaN) = NaN + I NaN */
+			return CMPLXF(y - y, y - y);
+		} else if (hx & 0x80000000) {
+			/* cexp(-Inf +- I Inf|NaN) = 0 + I 0 */
+			return CMPLXF(0.0, 0.0);
+		} else {
+			/* cexp(+Inf +- I Inf|NaN) = Inf + I NaN */
+			return CMPLXF(x, y - y);
+		}
+	}
+
+	if (hx >= exp_ovfl && hx <= cexp_ovfl) {
+		/*
+		 * x is between 88.7 and 192, so we must scale to avoid
+		 * overflow in expf(x).
+		 */
+		return __ldexp_cexpf(z, 0);
+	} else {
+		/*
+		 * Cases covered here:
+		 *  -  x < exp_ovfl and exp(x) won't overflow (common case)
+		 *  -  x > cexp_ovfl, so exp(x) * s overflows for all s > 0
+		 *  -  x = +-Inf (generated by exp())
+		 *  -  x = NaN (spurious inexact exception from y)
+		 */
+		exp_x = expf(x);
+		return CMPLXF(exp_x * cosf(y), exp_x * sinf(y));
+	}
+}
diff --git a/src/complex/cexpl.c b/src/complex/cexpl.c
new file mode 100644
index 0000000..d4df950
--- /dev/null
+++ b/src/complex/cexpl.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+//FIXME
+long double complex cexpl(long double complex z)
+{
+	return cexp(z);
+}
diff --git a/src/complex/cimag.c b/src/complex/cimag.c
new file mode 100644
index 0000000..d6b0e68
--- /dev/null
+++ b/src/complex/cimag.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+double (cimag)(double complex z)
+{
+	return cimag(z);
+}
diff --git a/src/complex/cimagf.c b/src/complex/cimagf.c
new file mode 100644
index 0000000..b7166dc
--- /dev/null
+++ b/src/complex/cimagf.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+float (cimagf)(float complex z)
+{
+	return cimagf(z);
+}
diff --git a/src/complex/cimagl.c b/src/complex/cimagl.c
new file mode 100644
index 0000000..4db77f2
--- /dev/null
+++ b/src/complex/cimagl.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+long double (cimagl)(long double complex z)
+{
+	return cimagl(z);
+}
diff --git a/src/complex/clog.c b/src/complex/clog.c
new file mode 100644
index 0000000..b587c29
--- /dev/null
+++ b/src/complex/clog.c
@@ -0,0 +1,14 @@
+#include "complex_impl.h"
+
+// FIXME
+
+/* log(z) = log(|z|) + i arg(z) */
+
+double complex clog(double complex z)
+{
+	double r, phi;
+
+	r = cabs(z);
+	phi = carg(z);
+	return CMPLX(log(r), phi);
+}
diff --git a/src/complex/clogf.c b/src/complex/clogf.c
new file mode 100644
index 0000000..0389d47
--- /dev/null
+++ b/src/complex/clogf.c
@@ -0,0 +1,12 @@
+#include "complex_impl.h"
+
+// FIXME
+
+float complex clogf(float complex z)
+{
+	float r, phi;
+
+	r = cabsf(z);
+	phi = cargf(z);
+	return CMPLXF(logf(r), phi);
+}
diff --git a/src/complex/clogl.c b/src/complex/clogl.c
new file mode 100644
index 0000000..88e83e8
--- /dev/null
+++ b/src/complex/clogl.c
@@ -0,0 +1,18 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex clogl(long double complex z)
+{
+	return clog(z);
+}
+#else
+// FIXME
+long double complex clogl(long double complex z)
+{
+	long double r, phi;
+
+	r = cabsl(z);
+	phi = cargl(z);
+	return CMPLXL(logl(r), phi);
+}
+#endif
diff --git a/src/complex/conj.c b/src/complex/conj.c
new file mode 100644
index 0000000..a3b19a4
--- /dev/null
+++ b/src/complex/conj.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+double complex conj(double complex z)
+{
+	return CMPLX(creal(z), -cimag(z));
+}
diff --git a/src/complex/conjf.c b/src/complex/conjf.c
new file mode 100644
index 0000000..b2195c8
--- /dev/null
+++ b/src/complex/conjf.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+float complex conjf(float complex z)
+{
+	return CMPLXF(crealf(z), -cimagf(z));
+}
diff --git a/src/complex/conjl.c b/src/complex/conjl.c
new file mode 100644
index 0000000..87a4ebe
--- /dev/null
+++ b/src/complex/conjl.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+long double complex conjl(long double complex z)
+{
+	return CMPLXL(creall(z), -cimagl(z));
+}
diff --git a/src/complex/cpow.c b/src/complex/cpow.c
new file mode 100644
index 0000000..1137d39
--- /dev/null
+++ b/src/complex/cpow.c
@@ -0,0 +1,8 @@
+#include "complex_impl.h"
+
+/* pow(z, c) = exp(c log(z)), See C99 G.6.4.1 */
+
+double complex cpow(double complex z, double complex c)
+{
+	return cexp(c * clog(z));
+}
diff --git a/src/complex/cpowf.c b/src/complex/cpowf.c
new file mode 100644
index 0000000..f3fd4b7
--- /dev/null
+++ b/src/complex/cpowf.c
@@ -0,0 +1,6 @@
+#include "complex_impl.h"
+
+float complex cpowf(float complex z, float complex c)
+{
+	return cexpf(c * clogf(z));
+}
diff --git a/src/complex/cpowl.c b/src/complex/cpowl.c
new file mode 100644
index 0000000..be36f04
--- /dev/null
+++ b/src/complex/cpowl.c
@@ -0,0 +1,13 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex cpowl(long double complex z, long double complex c)
+{
+	return cpow(z, c);
+}
+#else
+long double complex cpowl(long double complex z, long double complex c)
+{
+	return cexpl(c * clogl(z));
+}
+#endif
diff --git a/src/complex/cproj.c b/src/complex/cproj.c
new file mode 100644
index 0000000..9ae1e17
--- /dev/null
+++ b/src/complex/cproj.c
@@ -0,0 +1,8 @@
+#include "complex_impl.h"
+
+double complex cproj(double complex z)
+{
+	if (isinf(creal(z)) || isinf(cimag(z)))
+		return CMPLX(INFINITY, copysign(0.0, creal(z)));
+	return z;
+}
diff --git a/src/complex/cprojf.c b/src/complex/cprojf.c
new file mode 100644
index 0000000..03fab33
--- /dev/null
+++ b/src/complex/cprojf.c
@@ -0,0 +1,8 @@
+#include "complex_impl.h"
+
+float complex cprojf(float complex z)
+{
+	if (isinf(crealf(z)) || isinf(cimagf(z)))
+		return CMPLXF(INFINITY, copysignf(0.0, crealf(z)));
+	return z;
+}
diff --git a/src/complex/cprojl.c b/src/complex/cprojl.c
new file mode 100644
index 0000000..38a494c
--- /dev/null
+++ b/src/complex/cprojl.c
@@ -0,0 +1,15 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex cprojl(long double complex z)
+{
+	return cproj(z);
+}
+#else
+long double complex cprojl(long double complex z)
+{
+	if (isinf(creall(z)) || isinf(cimagl(z)))
+		return CMPLXL(INFINITY, copysignl(0.0, creall(z)));
+	return z;
+}
+#endif
diff --git a/src/complex/creal.c b/src/complex/creal.c
new file mode 100644
index 0000000..f670304
--- /dev/null
+++ b/src/complex/creal.c
@@ -0,0 +1,6 @@
+#include <complex.h>
+
+double (creal)(double complex z)
+{
+	return creal(z);
+}
diff --git a/src/complex/crealf.c b/src/complex/crealf.c
new file mode 100644
index 0000000..5dc3ff1
--- /dev/null
+++ b/src/complex/crealf.c
@@ -0,0 +1,6 @@
+#include <complex.h>
+
+float (crealf)(float complex z)
+{
+	return crealf(z);
+}
diff --git a/src/complex/creall.c b/src/complex/creall.c
new file mode 100644
index 0000000..fd9dc34
--- /dev/null
+++ b/src/complex/creall.c
@@ -0,0 +1,6 @@
+#include <complex.h>
+
+long double (creall)(long double complex z)
+{
+	return creall(z);
+}
diff --git a/src/complex/csin.c b/src/complex/csin.c
new file mode 100644
index 0000000..535c4bf
--- /dev/null
+++ b/src/complex/csin.c
@@ -0,0 +1,9 @@
+#include "complex_impl.h"
+
+/* sin(z) = -i sinh(i z) */
+
+double complex csin(double complex z)
+{
+	z = csinh(CMPLX(-cimag(z), creal(z)));
+	return CMPLX(cimag(z), -creal(z));
+}
diff --git a/src/complex/csinf.c b/src/complex/csinf.c
new file mode 100644
index 0000000..69f5164
--- /dev/null
+++ b/src/complex/csinf.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+float complex csinf(float complex z)
+{
+	z = csinhf(CMPLXF(-cimagf(z), crealf(z)));
+	return CMPLXF(cimagf(z), -crealf(z));
+}
diff --git a/src/complex/csinh.c b/src/complex/csinh.c
new file mode 100644
index 0000000..eda0ab5
--- /dev/null
+++ b/src/complex/csinh.c
@@ -0,0 +1,141 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_csinh.c */
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/*
+ * Hyperbolic sine of a complex argument z = x + i y.
+ *
+ * sinh(z) = sinh(x+iy)
+ *         = sinh(x) cos(y) + i cosh(x) sin(y).
+ *
+ * Exceptional values are noted in the comments within the source code.
+ * These values and the return value were taken from n1124.pdf.
+ */
+
+#include "complex_impl.h"
+
+static const double huge = 0x1p1023;
+
+double complex csinh(double complex z)
+{
+	double x, y, h;
+	int32_t hx, hy, ix, iy, lx, ly;
+
+	x = creal(z);
+	y = cimag(z);
+
+	EXTRACT_WORDS(hx, lx, x);
+	EXTRACT_WORDS(hy, ly, y);
+
+	ix = 0x7fffffff & hx;
+	iy = 0x7fffffff & hy;
+
+	/* Handle the nearly-non-exceptional cases where x and y are finite. */
+	if (ix < 0x7ff00000 && iy < 0x7ff00000) {
+		if ((iy | ly) == 0)
+			return CMPLX(sinh(x), y);
+		if (ix < 0x40360000)    /* small x: normal case */
+			return CMPLX(sinh(x) * cos(y), cosh(x) * sin(y));
+
+		/* |x| >= 22, so cosh(x) ~= exp(|x|) */
+		if (ix < 0x40862e42) {
+			/* x < 710: exp(|x|) won't overflow */
+			h = exp(fabs(x)) * 0.5;
+			return CMPLX(copysign(h, x) * cos(y), h * sin(y));
+		} else if (ix < 0x4096bbaa) {
+			/* x < 1455: scale to avoid overflow */
+			z = __ldexp_cexp(CMPLX(fabs(x), y), -1);
+			return CMPLX(creal(z) * copysign(1, x), cimag(z));
+		} else {
+			/* x >= 1455: the result always overflows */
+			h = huge * x;
+			return CMPLX(h * cos(y), h * h * sin(y));
+		}
+	}
+
+	/*
+	 * sinh(+-0 +- I Inf) = sign(d(+-0, dNaN))0 + I dNaN.
+	 * The sign of 0 in the result is unspecified.  Choice = normally
+	 * the same as dNaN.  Raise the invalid floating-point exception.
+	 *
+	 * sinh(+-0 +- I NaN) = sign(d(+-0, NaN))0 + I d(NaN).
+	 * The sign of 0 in the result is unspecified.  Choice = normally
+	 * the same as d(NaN).
+	 */
+	if ((ix | lx) == 0 && iy >= 0x7ff00000)
+		return CMPLX(copysign(0, x * (y - y)), y - y);
+
+	/*
+	 * sinh(+-Inf +- I 0) = +-Inf + I +-0.
+	 *
+	 * sinh(NaN +- I 0)   = d(NaN) + I +-0.
+	 */
+	if ((iy | ly) == 0 && ix >= 0x7ff00000) {
+		if (((hx & 0xfffff) | lx) == 0)
+			return CMPLX(x, y);
+		return CMPLX(x, copysign(0, y));
+	}
+
+	/*
+	 * sinh(x +- I Inf) = dNaN + I dNaN.
+	 * Raise the invalid floating-point exception for finite nonzero x.
+	 *
+	 * sinh(x + I NaN) = d(NaN) + I d(NaN).
+	 * Optionally raises the invalid floating-point exception for finite
+	 * nonzero x.  Choice = don't raise (except for signaling NaNs).
+	 */
+	if (ix < 0x7ff00000 && iy >= 0x7ff00000)
+		return CMPLX(y - y, x * (y - y));
+
+	/*
+	 * sinh(+-Inf + I NaN)  = +-Inf + I d(NaN).
+	 * The sign of Inf in the result is unspecified.  Choice = normally
+	 * the same as d(NaN).
+	 *
+	 * sinh(+-Inf +- I Inf) = +Inf + I dNaN.
+	 * The sign of Inf in the result is unspecified.  Choice = always +.
+	 * Raise the invalid floating-point exception.
+	 *
+	 * sinh(+-Inf + I y)   = +-Inf cos(y) + I Inf sin(y)
+	 */
+	if (ix >= 0x7ff00000 && ((hx & 0xfffff) | lx) == 0) {
+		if (iy >= 0x7ff00000)
+			return CMPLX(x * x, x * (y - y));
+		return CMPLX(x * cos(y), INFINITY * sin(y));
+	}
+
+	/*
+	 * sinh(NaN + I NaN)  = d(NaN) + I d(NaN).
+	 *
+	 * sinh(NaN +- I Inf) = d(NaN) + I d(NaN).
+	 * Optionally raises the invalid floating-point exception.
+	 * Choice = raise.
+	 *
+	 * sinh(NaN + I y)    = d(NaN) + I d(NaN).
+	 * Optionally raises the invalid floating-point exception for finite
+	 * nonzero y.  Choice = don't raise (except for signaling NaNs).
+	 */
+	return CMPLX((x * x) * (y - y), (x + x) * (y - y));
+}
diff --git a/src/complex/csinhf.c b/src/complex/csinhf.c
new file mode 100644
index 0000000..eb1d98c
--- /dev/null
+++ b/src/complex/csinhf.c
@@ -0,0 +1,90 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_csinhf.c */
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/*
+ * Hyperbolic sine of a complex argument z.  See s_csinh.c for details.
+ */
+
+#include "complex_impl.h"
+
+static const float huge = 0x1p127;
+
+float complex csinhf(float complex z)
+{
+	float x, y, h;
+	int32_t hx, hy, ix, iy;
+
+	x = crealf(z);
+	y = cimagf(z);
+
+	GET_FLOAT_WORD(hx, x);
+	GET_FLOAT_WORD(hy, y);
+
+	ix = 0x7fffffff & hx;
+	iy = 0x7fffffff & hy;
+
+	if (ix < 0x7f800000 && iy < 0x7f800000) {
+		if (iy == 0)
+			return CMPLXF(sinhf(x), y);
+		if (ix < 0x41100000)    /* small x: normal case */
+			return CMPLXF(sinhf(x) * cosf(y), coshf(x) * sinf(y));
+
+		/* |x| >= 9, so cosh(x) ~= exp(|x|) */
+		if (ix < 0x42b17218) {
+			/* x < 88.7: expf(|x|) won't overflow */
+			h = expf(fabsf(x)) * 0.5f;
+			return CMPLXF(copysignf(h, x) * cosf(y), h * sinf(y));
+		} else if (ix < 0x4340b1e7) {
+			/* x < 192.7: scale to avoid overflow */
+			z = __ldexp_cexpf(CMPLXF(fabsf(x), y), -1);
+			return CMPLXF(crealf(z) * copysignf(1, x), cimagf(z));
+		} else {
+			/* x >= 192.7: the result always overflows */
+			h = huge * x;
+			return CMPLXF(h * cosf(y), h * h * sinf(y));
+		}
+	}
+
+	if (ix == 0 && iy >= 0x7f800000)
+		return CMPLXF(copysignf(0, x * (y - y)), y - y);
+
+	if (iy == 0 && ix >= 0x7f800000) {
+		if ((hx & 0x7fffff) == 0)
+			return CMPLXF(x, y);
+		return CMPLXF(x, copysignf(0, y));
+	}
+
+	if (ix < 0x7f800000 && iy >= 0x7f800000)
+		return CMPLXF(y - y, x * (y - y));
+
+	if (ix >= 0x7f800000 && (hx & 0x7fffff) == 0) {
+		if (iy >= 0x7f800000)
+			return CMPLXF(x * x, x * (y - y));
+		return CMPLXF(x * cosf(y), INFINITY * sinf(y));
+	}
+
+	return CMPLXF((x * x) * (y - y), (x + x) * (y - y));
+}
diff --git a/src/complex/csinhl.c b/src/complex/csinhl.c
new file mode 100644
index 0000000..09fd18f
--- /dev/null
+++ b/src/complex/csinhl.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+//FIXME
+long double complex csinhl(long double complex z)
+{
+	return csinh(z);
+}
diff --git a/src/complex/csinl.c b/src/complex/csinl.c
new file mode 100644
index 0000000..90a4eb3
--- /dev/null
+++ b/src/complex/csinl.c
@@ -0,0 +1,14 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex csinl(long double complex z)
+{
+	return csin(z);
+}
+#else
+long double complex csinl(long double complex z)
+{
+	z = csinhl(CMPLXL(-cimagl(z), creall(z)));
+	return CMPLXL(cimagl(z), -creall(z));
+}
+#endif
diff --git a/src/complex/csqrt.c b/src/complex/csqrt.c
new file mode 100644
index 0000000..c36de00
--- /dev/null
+++ b/src/complex/csqrt.c
@@ -0,0 +1,100 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_csqrt.c */
+/*-
+ * Copyright (c) 2007 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "complex_impl.h"
+
+/*
+ * gcc doesn't implement complex multiplication or division correctly,
+ * so we need to handle infinities specially. We turn on this pragma to
+ * notify conforming c99 compilers that the fast-but-incorrect code that
+ * gcc generates is acceptable, since the special cases have already been
+ * handled.
+ */
+#pragma STDC CX_LIMITED_RANGE ON
+
+/* We risk spurious overflow for components >= DBL_MAX / (1 + sqrt(2)). */
+#define THRESH  0x1.a827999fcef32p+1022
+
+double complex csqrt(double complex z)
+{
+	double complex result;
+	double a, b;
+	double t;
+	int scale;
+
+	a = creal(z);
+	b = cimag(z);
+
+	/* Handle special cases. */
+	if (z == 0)
+		return CMPLX(0, b);
+	if (isinf(b))
+		return CMPLX(INFINITY, b);
+	if (isnan(a)) {
+		t = (b - b) / (b - b);  /* raise invalid if b is not a NaN */
+		return CMPLX(a, t);   /* return NaN + NaN i */
+	}
+	if (isinf(a)) {
+		/*
+		 * csqrt(inf + NaN i)  = inf +  NaN i
+		 * csqrt(inf + y i)    = inf +  0 i
+		 * csqrt(-inf + NaN i) = NaN +- inf i
+		 * csqrt(-inf + y i)   = 0   +  inf i
+		 */
+		if (signbit(a))
+			return CMPLX(fabs(b - b), copysign(a, b));
+		else
+			return CMPLX(a, copysign(b - b, b));
+	}
+	/*
+	 * The remaining special case (b is NaN) is handled just fine by
+	 * the normal code path below.
+	 */
+
+	/* Scale to avoid overflow. */
+	if (fabs(a) >= THRESH || fabs(b) >= THRESH) {
+		a *= 0.25;
+		b *= 0.25;
+		scale = 1;
+	} else {
+		scale = 0;
+	}
+
+	/* Algorithm 312, CACM vol 10, Oct 1967. */
+	if (a >= 0) {
+		t = sqrt((a + hypot(a, b)) * 0.5);
+		result = CMPLX(t, b / (2 * t));
+	} else {
+		t = sqrt((-a + hypot(a, b)) * 0.5);
+		result = CMPLX(fabs(b) / (2 * t), copysign(t, b));
+	}
+
+	/* Rescale. */
+	if (scale)
+		result *= 2;
+	return result;
+}
diff --git a/src/complex/csqrtf.c b/src/complex/csqrtf.c
new file mode 100644
index 0000000..46b3e89
--- /dev/null
+++ b/src/complex/csqrtf.c
@@ -0,0 +1,81 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_csqrtf.c */
+/*-
+ * Copyright (c) 2007 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "complex_impl.h"
+
+/*
+ * gcc doesn't implement complex multiplication or division correctly,
+ * so we need to handle infinities specially. We turn on this pragma to
+ * notify conforming c99 compilers that the fast-but-incorrect code that
+ * gcc generates is acceptable, since the special cases have already been
+ * handled.
+ */
+
+float complex csqrtf(float complex z)
+{
+	float a = crealf(z), b = cimagf(z);
+	double t;
+
+	/* Handle special cases. */
+	if (z == 0)
+		return CMPLXF(0, b);
+	if (isinf(b))
+		return CMPLXF(INFINITY, b);
+	if (isnan(a)) {
+		t = (b - b) / (b - b);  /* raise invalid if b is not a NaN */
+		return CMPLXF(a, t);  /* return NaN + NaN i */
+	}
+	if (isinf(a)) {
+		/*
+		 * csqrtf(inf + NaN i)  = inf +  NaN i
+		 * csqrtf(inf + y i)    = inf +  0 i
+		 * csqrtf(-inf + NaN i) = NaN +- inf i
+		 * csqrtf(-inf + y i)   = 0   +  inf i
+		 */
+		if (signbit(a))
+			return CMPLXF(fabsf(b - b), copysignf(a, b));
+		else
+			return CMPLXF(a, copysignf(b - b, b));
+	}
+	/*
+	 * The remaining special case (b is NaN) is handled just fine by
+	 * the normal code path below.
+	 */
+
+	/*
+	 * We compute t in double precision to avoid overflow and to
+	 * provide correct rounding in nearly all cases.
+	 * This is Algorithm 312, CACM vol 10, Oct 1967.
+	 */
+	if (a >= 0) {
+		t = sqrt((a + hypot(a, b)) * 0.5);
+		return CMPLXF(t, b / (2.0 * t));
+	} else {
+		t = sqrt((-a + hypot(a, b)) * 0.5);
+		return CMPLXF(fabsf(b) / (2.0 * t), copysignf(t, b));
+	}
+}
diff --git a/src/complex/csqrtl.c b/src/complex/csqrtl.c
new file mode 100644
index 0000000..2253937
--- /dev/null
+++ b/src/complex/csqrtl.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+//FIXME
+long double complex csqrtl(long double complex z)
+{
+	return csqrt(z);
+}
diff --git a/src/complex/ctan.c b/src/complex/ctan.c
new file mode 100644
index 0000000..918717b
--- /dev/null
+++ b/src/complex/ctan.c
@@ -0,0 +1,9 @@
+#include "complex_impl.h"
+
+/* tan(z) = -i tanh(i z) */
+
+double complex ctan(double complex z)
+{
+	z = ctanh(CMPLX(-cimag(z), creal(z)));
+	return CMPLX(cimag(z), -creal(z));
+}
diff --git a/src/complex/ctanf.c b/src/complex/ctanf.c
new file mode 100644
index 0000000..04c3ff1
--- /dev/null
+++ b/src/complex/ctanf.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+float complex ctanf(float complex z)
+{
+	z = ctanhf(CMPLXF(-cimagf(z), crealf(z)));
+	return CMPLXF(cimagf(z), -crealf(z));
+}
diff --git a/src/complex/ctanh.c b/src/complex/ctanh.c
new file mode 100644
index 0000000..54004cd
--- /dev/null
+++ b/src/complex/ctanh.c
@@ -0,0 +1,129 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_ctanh.c */
+/*-
+ * Copyright (c) 2011 David Schultz
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/*
+ * Hyperbolic tangent of a complex argument z = x + i y.
+ *
+ * The algorithm is from:
+ *
+ *   W. Kahan.  Branch Cuts for Complex Elementary Functions or Much
+ *   Ado About Nothing's Sign Bit.  In The State of the Art in
+ *   Numerical Analysis, pp. 165 ff.  Iserles and Powell, eds., 1987.
+ *
+ * Method:
+ *
+ *   Let t    = tan(x)
+ *       beta = 1/cos^2(y)
+ *       s    = sinh(x)
+ *       rho  = cosh(x)
+ *
+ *   We have:
+ *
+ *   tanh(z) = sinh(z) / cosh(z)
+ *
+ *             sinh(x) cos(y) + i cosh(x) sin(y)
+ *           = ---------------------------------
+ *             cosh(x) cos(y) + i sinh(x) sin(y)
+ *
+ *             cosh(x) sinh(x) / cos^2(y) + i tan(y)
+ *           = -------------------------------------
+ *                    1 + sinh^2(x) / cos^2(y)
+ *
+ *             beta rho s + i t
+ *           = ----------------
+ *               1 + beta s^2
+ *
+ * Modifications:
+ *
+ *   I omitted the original algorithm's handling of overflow in tan(x) after
+ *   verifying with nearpi.c that this can't happen in IEEE single or double
+ *   precision.  I also handle large x differently.
+ */
+
+#include "complex_impl.h"
+
+double complex ctanh(double complex z)
+{
+	double x, y;
+	double t, beta, s, rho, denom;
+	uint32_t hx, ix, lx;
+
+	x = creal(z);
+	y = cimag(z);
+
+	EXTRACT_WORDS(hx, lx, x);
+	ix = hx & 0x7fffffff;
+
+	/*
+	 * ctanh(NaN + i 0) = NaN + i 0
+	 *
+	 * ctanh(NaN + i y) = NaN + i NaN               for y != 0
+	 *
+	 * The imaginary part has the sign of x*sin(2*y), but there's no
+	 * special effort to get this right.
+	 *
+	 * ctanh(+-Inf +- i Inf) = +-1 +- 0
+	 *
+	 * ctanh(+-Inf + i y) = +-1 + 0 sin(2y)         for y finite
+	 *
+	 * The imaginary part of the sign is unspecified.  This special
+	 * case is only needed to avoid a spurious invalid exception when
+	 * y is infinite.
+	 */
+	if (ix >= 0x7ff00000) {
+		if ((ix & 0xfffff) | lx)        /* x is NaN */
+			return CMPLX(x, (y == 0 ? y : x * y));
+		SET_HIGH_WORD(x, hx - 0x40000000);      /* x = copysign(1, x) */
+		return CMPLX(x, copysign(0, isinf(y) ? y : sin(y) * cos(y)));
+	}
+
+	/*
+	 * ctanh(+-0 + i NAN) = +-0 + i NaN
+	 * ctanh(+-0 +- i Inf) = +-0 + i NaN
+	 * ctanh(x + i NAN) = NaN + i NaN
+	 * ctanh(x +- i Inf) = NaN + i NaN
+	 */
+	if (!isfinite(y))
+		return CMPLX(x ? y - y : x, y - y);
+
+	/*
+	 * ctanh(+-huge + i +-y) ~= +-1 +- i 2sin(2y)/exp(2x), using the
+	 * approximation sinh^2(huge) ~= exp(2*huge) / 4.
+	 * We use a modified formula to avoid spurious overflow.
+	 */
+	if (ix >= 0x40360000) { /* x >= 22 */
+		double exp_mx = exp(-fabs(x));
+		return CMPLX(copysign(1, x), 4 * sin(y) * cos(y) * exp_mx * exp_mx);
+	}
+
+	/* Kahan's algorithm */
+	t = tan(y);
+	beta = 1.0 + t * t;     /* = 1 / cos^2(y) */
+	s = sinh(x);
+	rho = sqrt(1 + s * s);  /* = cosh(x) */
+	denom = 1 + beta * s * s;
+	return CMPLX((beta * rho * s) / denom, t / denom);
+}
diff --git a/src/complex/ctanhf.c b/src/complex/ctanhf.c
new file mode 100644
index 0000000..7f422ba
--- /dev/null
+++ b/src/complex/ctanhf.c
@@ -0,0 +1,66 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_ctanhf.c */
+/*-
+ * Copyright (c) 2011 David Schultz
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/*
+ * Hyperbolic tangent of a complex argument z.  See s_ctanh.c for details.
+ */
+
+#include "complex_impl.h"
+
+float complex ctanhf(float complex z)
+{
+	float x, y;
+	float t, beta, s, rho, denom;
+	uint32_t hx, ix;
+
+	x = crealf(z);
+	y = cimagf(z);
+
+	GET_FLOAT_WORD(hx, x);
+	ix = hx & 0x7fffffff;
+
+	if (ix >= 0x7f800000) {
+		if (ix & 0x7fffff)
+			return CMPLXF(x, (y == 0 ? y : x * y));
+		SET_FLOAT_WORD(x, hx - 0x40000000);
+		return CMPLXF(x, copysignf(0, isinf(y) ? y : sinf(y) * cosf(y)));
+	}
+
+	if (!isfinite(y))
+		return CMPLXF(ix ? y - y : x, y - y);
+
+	if (ix >= 0x41300000) { /* x >= 11 */
+		float exp_mx = expf(-fabsf(x));
+		return CMPLXF(copysignf(1, x), 4 * sinf(y) * cosf(y) * exp_mx * exp_mx);
+	}
+
+	t = tanf(y);
+	beta = 1.0 + t * t;
+	s = sinhf(x);
+	rho = sqrtf(1 + s * s);
+	denom = 1 + beta * s * s;
+	return CMPLXF((beta * rho * s) / denom, t / denom);
+}
diff --git a/src/complex/ctanhl.c b/src/complex/ctanhl.c
new file mode 100644
index 0000000..45d5862
--- /dev/null
+++ b/src/complex/ctanhl.c
@@ -0,0 +1,7 @@
+#include "complex_impl.h"
+
+//FIXME
+long double complex ctanhl(long double complex z)
+{
+	return ctanh(z);
+}
diff --git a/src/complex/ctanl.c b/src/complex/ctanl.c
new file mode 100644
index 0000000..4b87420
--- /dev/null
+++ b/src/complex/ctanl.c
@@ -0,0 +1,14 @@
+#include "complex_impl.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double complex ctanl(long double complex z)
+{
+	return ctan(z);
+}
+#else
+long double complex ctanl(long double complex z)
+{
+	z = ctanhl(CMPLXL(-cimagl(z), creall(z)));
+	return CMPLXL(cimagl(z), -creall(z));
+}
+#endif
diff --git a/src/fenv/__flt_rounds.c b/src/fenv/__flt_rounds.c
new file mode 100644
index 0000000..ec0b368
--- /dev/null
+++ b/src/fenv/__flt_rounds.c
@@ -0,0 +1,19 @@
+#include <float.h>
+#include <fenv.h>
+
+int __flt_rounds()
+{
+	switch (fegetround()) {
+#ifdef FE_TOWARDZERO
+	case FE_TOWARDZERO: return 0;
+#endif
+	case FE_TONEAREST: return 1;
+#ifdef FE_UPWARD
+	case FE_UPWARD: return 2;
+#endif
+#ifdef FE_DOWNWARD
+	case FE_DOWNWARD: return 3;
+#endif
+	}
+	return -1;
+}
diff --git a/src/fenv/fegetexceptflag.c b/src/fenv/fegetexceptflag.c
new file mode 100644
index 0000000..bab0b44
--- /dev/null
+++ b/src/fenv/fegetexceptflag.c
@@ -0,0 +1,7 @@
+#include <fenv.h>
+
+int fegetexceptflag(fexcept_t *fp, int mask)
+{
+	*fp = fetestexcept(mask);
+	return 0;
+}
diff --git a/src/fenv/feholdexcept.c b/src/fenv/feholdexcept.c
new file mode 100644
index 0000000..73ff1fa
--- /dev/null
+++ b/src/fenv/feholdexcept.c
@@ -0,0 +1,8 @@
+#include <fenv.h>
+
+int feholdexcept(fenv_t *envp)
+{
+	fegetenv(envp);
+	feclearexcept(FE_ALL_EXCEPT);
+	return 0;
+}
diff --git a/src/fenv/fenv.c b/src/fenv/fenv.c
new file mode 100644
index 0000000..5588dad
--- /dev/null
+++ b/src/fenv/fenv.c
@@ -0,0 +1,38 @@
+#include <fenv.h>
+
+/* Dummy functions for archs lacking fenv implementation */
+
+int feclearexcept(int mask)
+{
+	return 0;
+}
+
+int feraiseexcept(int mask)
+{
+	return 0;
+}
+
+int fetestexcept(int mask)
+{
+	return 0;
+}
+
+int fegetround(void)
+{
+	return FE_TONEAREST;
+}
+
+int __fesetround(int r)
+{
+	return 0;
+}
+
+int fegetenv(fenv_t *envp)
+{
+	return 0;
+}
+
+int fesetenv(const fenv_t *envp)
+{
+	return 0;
+}
diff --git a/src/fenv/fesetexceptflag.c b/src/fenv/fesetexceptflag.c
new file mode 100644
index 0000000..af5f102
--- /dev/null
+++ b/src/fenv/fesetexceptflag.c
@@ -0,0 +1,8 @@
+#include <fenv.h>
+
+int fesetexceptflag(const fexcept_t *fp, int mask)
+{
+	feclearexcept(~*fp & mask);
+	feraiseexcept(*fp & mask);
+	return 0;
+}
diff --git a/src/fenv/fesetround.c b/src/fenv/fesetround.c
new file mode 100644
index 0000000..4e2f164
--- /dev/null
+++ b/src/fenv/fesetround.c
@@ -0,0 +1,23 @@
+#include <fenv.h>
+#include <features.h>
+
+/* __fesetround wrapper for arch independent argument check */
+
+hidden int __fesetround(int);
+
+int fesetround(int r)
+{
+	if (r != FE_TONEAREST
+#ifdef FE_DOWNWARD
+		&& r != FE_DOWNWARD
+#endif
+#ifdef FE_UPWARD
+		&& r != FE_UPWARD
+#endif
+#ifdef FE_TOWARDZERO
+		&& r != FE_TOWARDZERO
+#endif
+	)
+		return -1;
+	return __fesetround(r);
+}
diff --git a/src/fenv/feupdateenv.c b/src/fenv/feupdateenv.c
new file mode 100644
index 0000000..50cef8e
--- /dev/null
+++ b/src/fenv/feupdateenv.c
@@ -0,0 +1,9 @@
+#include <fenv.h>
+
+int feupdateenv(const fenv_t *envp)
+{
+	int ex = fetestexcept(FE_ALL_EXCEPT);
+	fesetenv(envp);
+	feraiseexcept(ex);
+	return 0;
+}
diff --git a/src/fenv/riscv64/fenv-sf.c b/src/fenv/riscv64/fenv-sf.c
new file mode 100644
index 0000000..ecd3cb5
--- /dev/null
+++ b/src/fenv/riscv64/fenv-sf.c
@@ -0,0 +1,3 @@
+#ifndef __riscv_flen
+#include "../fenv.c"
+#endif
diff --git a/src/fenv/riscv64/fenv.S b/src/fenv/riscv64/fenv.S
new file mode 100644
index 0000000..0ea78bf
--- /dev/null
+++ b/src/fenv/riscv64/fenv.S
@@ -0,0 +1,56 @@
+#ifdef __riscv_flen
+
+.global feclearexcept
+.type feclearexcept, %function
+feclearexcept:
+	csrc fflags, a0
+	li a0, 0
+	ret
+
+.global feraiseexcept
+.type feraiseexcept, %function
+feraiseexcept:
+	csrs fflags, a0
+	li a0, 0
+	ret
+
+.global fetestexcept
+.type fetestexcept, %function
+fetestexcept:
+	frflags t0
+	and a0, t0, a0
+	ret
+
+.global fegetround
+.type fegetround, %function
+fegetround:
+	frrm a0
+	ret
+
+.global __fesetround
+.type __fesetround, %function
+__fesetround:
+	fsrm t0, a0
+	li a0, 0
+	ret
+
+.global fegetenv
+.type fegetenv, %function
+fegetenv:
+	frcsr t0
+	sw t0, 0(a0)
+	li a0, 0
+	ret
+
+.global fesetenv
+.type fesetenv, %function
+fesetenv:
+	li t2, -1
+	li t1, 0
+	beq a0, t2, 1f
+	lw t1, 0(a0)
+1:	fscsr t1
+	li a0, 0
+	ret
+
+#endif
diff --git a/src/fenv/x86/fenv.S b/src/fenv/x86/fenv.S
new file mode 100644
index 0000000..e7f7932
--- /dev/null
+++ b/src/fenv/x86/fenv.S
@@ -0,0 +1,164 @@
+.hidden __hwcap
+
+.global feclearexcept
+.type feclearexcept,@function
+feclearexcept:	
+	mov 4(%esp),%ecx
+	and $0x3f,%ecx
+	fnstsw %ax
+		# consider sse fenv as well if the cpu has XMM capability
+	call 1f
+1:	addl $__hwcap-1b,(%esp)
+	pop %edx
+	testl $0x02000000,(%edx)
+	jz 2f
+		# maintain exceptions in the sse mxcsr, clear x87 exceptions
+	test %eax,%ecx
+	jz 1f
+	fnclex
+1:	push %edx
+	stmxcsr (%esp)
+	pop %edx
+	and $0x3f,%eax
+	or %eax,%edx
+	test %edx,%ecx
+	jz 1f
+	not %ecx
+	and %ecx,%edx
+	push %edx
+	ldmxcsr (%esp)
+	pop %edx
+1:	xor %eax,%eax
+	ret
+		# only do the expensive x87 fenv load/store when needed
+2:	test %eax,%ecx
+	jz 1b
+	not %ecx
+	and %ecx,%eax
+	test $0x3f,%eax
+	jz 1f
+	fnclex
+	jmp 1b
+1:	sub $32,%esp
+	fnstenv (%esp)
+	mov %al,4(%esp)
+	fldenv (%esp)
+	add $32,%esp
+	xor %eax,%eax
+	ret
+
+.global feraiseexcept
+.type feraiseexcept,@function
+feraiseexcept:	
+	mov 4(%esp),%eax
+	and $0x3f,%eax
+	sub $32,%esp
+	fnstenv (%esp)
+	or %al,4(%esp)
+	fldenv (%esp)
+	add $32,%esp
+	xor %eax,%eax
+	ret
+
+.global __fesetround
+.hidden __fesetround
+.type __fesetround,@function
+__fesetround:
+	mov 4(%esp),%ecx
+	push %eax
+	xor %eax,%eax
+	fnstcw (%esp)
+	andb $0xf3,1(%esp)
+	or %ch,1(%esp)
+	fldcw (%esp)
+		# consider sse fenv as well if the cpu has XMM capability
+	call 1f
+1:	addl $__hwcap-1b,(%esp)
+	pop %edx
+	testl $0x02000000,(%edx)
+	jz 1f
+	stmxcsr (%esp)
+	shl $3,%ch
+	andb $0x9f,1(%esp)
+	or %ch,1(%esp)
+	ldmxcsr (%esp)
+1:	pop %ecx
+	ret
+
+.global fegetround
+.type fegetround,@function
+fegetround:
+	push %eax
+	fnstcw (%esp)
+	pop %eax
+	and $0xc00,%eax
+	ret
+
+.global fegetenv
+.type fegetenv,@function
+fegetenv:
+	mov 4(%esp),%ecx
+	xor %eax,%eax
+	fnstenv (%ecx)
+		# consider sse fenv as well if the cpu has XMM capability
+	call 1f
+1:	addl $__hwcap-1b,(%esp)
+	pop %edx
+	testl $0x02000000,(%edx)
+	jz 1f
+	push %eax
+	stmxcsr (%esp)
+	pop %edx
+	and $0x3f,%edx
+	or %edx,4(%ecx)
+1:	ret
+
+.global fesetenv
+.type fesetenv,@function
+fesetenv:
+	mov 4(%esp),%ecx
+	xor %eax,%eax
+	inc %ecx
+	jz 1f
+	fldenv -1(%ecx)
+	movl -1(%ecx),%ecx
+	jmp 2f
+1:	push %eax
+	push %eax
+	push %eax
+	push %eax
+	pushl $0xffff
+	push %eax
+	pushl $0x37f
+	fldenv (%esp)
+	add $28,%esp
+		# consider sse fenv as well if the cpu has XMM capability
+2:	call 1f
+1:	addl $__hwcap-1b,(%esp)
+	pop %edx
+	testl $0x02000000,(%edx)
+	jz 1f
+		# mxcsr := same rounding mode, cleared exceptions, default mask
+	and $0xc00,%ecx
+	shl $3,%ecx
+	or $0x1f80,%ecx
+	mov %ecx,4(%esp)
+	ldmxcsr 4(%esp)
+1:	ret
+
+.global fetestexcept
+.type fetestexcept,@function
+fetestexcept:
+	mov 4(%esp),%ecx
+	and $0x3f,%ecx
+	fnstsw %ax
+		# consider sse fenv as well if the cpu has XMM capability
+	call 1f
+1:	addl $__hwcap-1b,(%esp)
+	pop %edx
+	testl $0x02000000,(%edx)
+	jz 1f
+	stmxcsr 4(%esp)
+	or 4(%esp),%eax
+1:	and %ecx,%eax
+	ret
diff --git a/src/include/complex.h b/src/include/complex.h
new file mode 100644
index 0000000..1a0dcb0
--- /dev/null
+++ b/src/include/complex.h
@@ -0,0 +1,120 @@
+#ifndef _COMPLEX_H
+#define _COMPLEX_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define complex _Complex
+#ifdef __GNUC__
+#define _Complex_I (__extension__ (0.0f+1.0fi))
+#else
+#define _Complex_I (0.0f+1.0fi)
+#endif
+#define I _Complex_I
+
+double complex cacos(double complex);
+float complex cacosf(float complex);
+long double complex cacosl(long double complex);
+
+double complex casin(double complex);
+float complex casinf(float complex);
+long double complex casinl(long double complex);
+
+double complex catan(double complex);
+float complex catanf(float complex);
+long double complex catanl(long double complex);
+
+double complex ccos(double complex);
+float complex ccosf(float complex);
+long double complex ccosl(long double complex);
+
+double complex csin(double complex);
+float complex csinf(float complex);
+long double complex csinl(long double complex);
+
+double complex ctan(double complex);
+float complex ctanf(float complex);
+long double complex ctanl(long double complex);
+
+double complex cacosh(double complex);
+float complex cacoshf(float complex);
+long double complex cacoshl(long double complex);
+
+double complex casinh(double complex);
+float complex casinhf(float complex);
+long double complex casinhl(long double complex);
+
+double complex catanh(double complex);
+float complex catanhf(float complex);
+long double complex catanhl(long double complex);
+
+double complex ccosh(double complex);
+float complex ccoshf(float complex);
+long double complex ccoshl(long double complex);
+
+double complex csinh(double complex);
+float complex csinhf(float complex);
+long double complex csinhl(long double complex);
+
+double complex ctanh(double complex);
+float complex ctanhf(float complex);
+long double complex ctanhl(long double complex);
+
+double complex cexp(double complex);
+float complex cexpf(float complex);
+long double complex cexpl(long double complex);
+
+double complex clog(double complex);
+float complex clogf(float complex);
+long double complex clogl(long double complex);
+
+double cabs(double complex);
+float cabsf(float complex);
+long double cabsl(long double complex);
+
+double complex cpow(double complex, double complex);
+float complex cpowf(float complex, float complex);
+long double complex cpowl(long double complex, long double complex);
+
+double complex csqrt(double complex);
+float complex csqrtf(float complex);
+long double complex csqrtl(long double complex);
+
+double carg(double complex);
+float cargf(float complex);
+long double cargl(long double complex);
+
+double cimag(double complex);
+float cimagf(float complex);
+long double cimagl(long double complex);
+
+double complex conj(double complex);
+float complex conjf(float complex);
+long double complex conjl(long double complex);
+
+double complex cproj(double complex);
+float complex cprojf(float complex);
+long double complex cprojl(long double complex);
+
+double creal(double complex);
+float crealf(float complex);
+long double creall(long double complex);
+
+#ifndef __cplusplus
+#define __CIMAG(x, t) \
+	(+(union { _Complex t __z; t __xy[2]; }){(_Complex t)(x)}.__xy[1])
+
+#define creal(x) ((double)(x))
+#define crealf(x) ((float)(x))
+#define creall(x) ((long double)(x))
+
+#define cimag(x) __CIMAG(x, double)
+#define cimagf(x) __CIMAG(x, float)
+#define cimagl(x) __CIMAG(x, long double)
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff --git a/src/include/fenv.h b/src/include/fenv.h
new file mode 100644
index 0000000..05de990
--- /dev/null
+++ b/src/include/fenv.h
@@ -0,0 +1,28 @@
+#ifndef _FENV_H
+#define _FENV_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <bits/fenv.h>
+
+int feclearexcept(int);
+int fegetexceptflag(fexcept_t *, int);
+int feraiseexcept(int);
+int fesetexceptflag(const fexcept_t *, int);
+int fetestexcept(int);
+
+int fegetround(void);
+int fesetround(int);
+
+int fegetenv(fenv_t *);
+int feholdexcept(fenv_t *);
+int fesetenv(const fenv_t *);
+int feupdateenv(const fenv_t *);
+
+#ifdef __cplusplus
+}
+#endif
+#endif
+
diff --git a/src/include/float.h b/src/include/float.h
new file mode 100644
index 0000000..713aadb
--- /dev/null
+++ b/src/include/float.h
@@ -0,0 +1,52 @@
+#ifndef _FLOAT_H
+#define _FLOAT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+int __flt_rounds(void);
+#define FLT_ROUNDS (__flt_rounds())
+
+#define FLT_RADIX 2
+
+#define FLT_TRUE_MIN 1.40129846432481707092e-45F
+#define FLT_MIN 1.17549435082228750797e-38F
+#define FLT_MAX 3.40282346638528859812e+38F
+#define FLT_EPSILON 1.1920928955078125e-07F
+
+#define FLT_MANT_DIG 24
+#define FLT_MIN_EXP (-125)
+#define FLT_MAX_EXP 128
+#define FLT_HAS_SUBNORM 1
+
+#define FLT_DIG 6
+#define FLT_DECIMAL_DIG 9
+#define FLT_MIN_10_EXP (-37)
+#define FLT_MAX_10_EXP 38
+
+#define DBL_TRUE_MIN 4.94065645841246544177e-324
+#define DBL_MIN 2.22507385850720138309e-308
+#define DBL_MAX 1.79769313486231570815e+308
+#define DBL_EPSILON 2.22044604925031308085e-16
+
+#define DBL_MANT_DIG 53
+#define DBL_MIN_EXP (-1021)
+#define DBL_MAX_EXP 1024
+#define DBL_HAS_SUBNORM 1
+
+#define DBL_DIG 15
+#define DBL_DECIMAL_DIG 17
+#define DBL_MIN_10_EXP (-307)
+#define DBL_MAX_10_EXP 308
+
+#define LDBL_HAS_SUBNORM 1
+#define LDBL_DECIMAL_DIG DECIMAL_DIG
+
+#include <bits/float.h>
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/include/math.h b/src/include/math.h
index a21cd8e..95f384e 100644
--- a/src/include/math.h
+++ b/src/include/math.h
@@ -187,8 +187,12 @@ double	atan(double);
 float	atanf(float);
 double	atan2(double, double);
 float	atan2f(float, float);
+long double atan2l(long double, long double);
+
 double	atanh(double);
 float	atanhf(float);
+long double atanhl(long double);
+
 double	cbrt(double);
 float	cbrtf(float);
 double	ceil(double);
@@ -224,10 +228,14 @@ long double frexpl(long double, int *);
 
 double	hypot(double, double);
 float	hypotf(float, float);
+long double hypotl(long double, long double);
+
 double	ldexp(double, int);
 float	ldexpf(float, int);
 double	log(double);
 float	logf(float);
+long double logl(long double);
+
 double	log10(double);
 float	log10f(float);
 double	log1p(double);
@@ -260,6 +268,8 @@ double	sinh(double);
 float	sinhf(float);
 double	sqrt(double);
 float	sqrtf(float);
+long double sqrtl(long double);
+
 double	tan(double);
 float	tanf(float);
 double	tanh(double);
diff --git a/src/internal/complex_impl.h b/src/internal/complex_impl.h
new file mode 100644
index 0000000..51fb298
--- /dev/null
+++ b/src/internal/complex_impl.h
@@ -0,0 +1,22 @@
+#ifndef _COMPLEX_IMPL_H
+#define _COMPLEX_IMPL_H
+
+#include <complex.h>
+#include "libm.h"
+
+#undef __CMPLX
+#undef CMPLX
+#undef CMPLXF
+#undef CMPLXL
+
+#define __CMPLX(x, y, t) \
+	((union { _Complex t __z; t __xy[2]; }){.__xy = {(x),(y)}}.__z)
+
+#define CMPLX(x, y) __CMPLX(x, y, double)
+#define CMPLXF(x, y) __CMPLX(x, y, float)
+#define CMPLXL(x, y) __CMPLX(x, y, long double)
+
+hidden double complex __ldexp_cexp(double complex,int);
+hidden float complex __ldexp_cexpf(float complex,int);
+
+#endif
diff --git a/src/internal/libm.h b/src/internal/libm.h
index 1c44964..240332e 100644
--- a/src/internal/libm.h
+++ b/src/internal/libm.h
@@ -58,4 +58,17 @@ union ldshape {
 #error Unsupported long double representation
 #endif
 
+/* Helps static branch prediction so hot path can be better optimized.  */
+#ifdef __GNUC__
+#define predict_true(x) __builtin_expect(!!(x), 1)
+#define predict_false(x) __builtin_expect(x, 0)
+#else
+#define predict_true(x) (x)
+#define predict_false(x) (x)
+#endif
+
+#if LDBL_MANT_DIG != DBL_MANT_DIG
+hidden long double __math_invalidl(long double);
+#endif
+
 #endif
diff --git a/src/math/__math_invalidl.c b/src/math/__math_invalidl.c
new file mode 100644
index 0000000..1fca99d
--- /dev/null
+++ b/src/math/__math_invalidl.c
@@ -0,0 +1,9 @@
+#include <float.h>
+#include "libm.h"
+
+#if LDBL_MANT_DIG != DBL_MANT_DIG
+long double __math_invalidl(long double x)
+{
+	return (x - x) / (x - x);
+}
+#endif
diff --git a/src/math/hypotl.c b/src/math/hypotl.c
new file mode 100644
index 0000000..479aa92
--- /dev/null
+++ b/src/math/hypotl.c
@@ -0,0 +1,66 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double hypotl(long double x, long double y)
+{
+	return hypot(x, y);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#if LDBL_MANT_DIG == 64
+#define SPLIT (0x1p32L+1)
+#elif LDBL_MANT_DIG == 113
+#define SPLIT (0x1p57L+1)
+#endif
+
+static void sq(long double *hi, long double *lo, long double x)
+{
+	long double xh, xl, xc;
+	xc = x*SPLIT;
+	xh = x - xc + xc;
+	xl = x - xh;
+	*hi = x*x;
+	*lo = xh*xh - *hi + 2*xh*xl + xl*xl;
+}
+
+long double hypotl(long double x, long double y)
+{
+	union ldshape ux = {x}, uy = {y};
+	int ex, ey;
+	long double hx, lx, hy, ly, z;
+
+	ux.i.se &= 0x7fff;
+	uy.i.se &= 0x7fff;
+	if (ux.i.se < uy.i.se) {
+		ex = uy.i.se;
+		ey = ux.i.se;
+		x = uy.f;
+		y = ux.f;
+	} else {
+		ex = ux.i.se;
+		ey = uy.i.se;
+		x = ux.f;
+		y = uy.f;
+	}
+
+	if (ex == 0x7fff && isinf(y))
+		return y;
+	if (ex == 0x7fff || y == 0)
+		return x;
+	if (ex - ey > LDBL_MANT_DIG)
+		return x + y;
+
+	z = 1;
+	if (ex > 0x3fff+8000) {
+		z = 0x1p10000L;
+		x *= 0x1p-10000L;
+		y *= 0x1p-10000L;
+	} else if (ey < 0x3fff-8000) {
+		z = 0x1p-10000L;
+		x *= 0x1p10000L;
+		y *= 0x1p10000L;
+	}
+	sq(&hx, &lx, x);
+	sq(&hy, &ly, y);
+	return z*sqrtl(ly+lx+hy+hx);
+}
+#endif
diff --git a/src/math/sqrt_data.c b/src/math/sqrt_data.c
new file mode 100644
index 0000000..61bc22f
--- /dev/null
+++ b/src/math/sqrt_data.c
@@ -0,0 +1,19 @@
+#include "sqrt_data.h"
+const uint16_t __rsqrt_tab[128] = {
+0xb451,0xb2f0,0xb196,0xb044,0xaef9,0xadb6,0xac79,0xab43,
+0xaa14,0xa8eb,0xa7c8,0xa6aa,0xa592,0xa480,0xa373,0xa26b,
+0xa168,0xa06a,0x9f70,0x9e7b,0x9d8a,0x9c9d,0x9bb5,0x9ad1,
+0x99f0,0x9913,0x983a,0x9765,0x9693,0x95c4,0x94f8,0x9430,
+0x936b,0x92a9,0x91ea,0x912e,0x9075,0x8fbe,0x8f0a,0x8e59,
+0x8daa,0x8cfe,0x8c54,0x8bac,0x8b07,0x8a64,0x89c4,0x8925,
+0x8889,0x87ee,0x8756,0x86c0,0x862b,0x8599,0x8508,0x8479,
+0x83ec,0x8361,0x82d8,0x8250,0x81c9,0x8145,0x80c2,0x8040,
+0xff02,0xfd0e,0xfb25,0xf947,0xf773,0xf5aa,0xf3ea,0xf234,
+0xf087,0xeee3,0xed47,0xebb3,0xea27,0xe8a3,0xe727,0xe5b2,
+0xe443,0xe2dc,0xe17a,0xe020,0xdecb,0xdd7d,0xdc34,0xdaf1,
+0xd9b3,0xd87b,0xd748,0xd61a,0xd4f1,0xd3cd,0xd2ad,0xd192,
+0xd07b,0xcf69,0xce5b,0xcd51,0xcc4a,0xcb48,0xca4a,0xc94f,
+0xc858,0xc764,0xc674,0xc587,0xc49d,0xc3b7,0xc2d4,0xc1f4,
+0xc116,0xc03c,0xbf65,0xbe90,0xbdbe,0xbcef,0xbc23,0xbb59,
+0xba91,0xb9cc,0xb90a,0xb84a,0xb78c,0xb6d0,0xb617,0xb560,
+};
diff --git a/src/math/sqrt_data.h b/src/math/sqrt_data.h
new file mode 100644
index 0000000..260c7f9
--- /dev/null
+++ b/src/math/sqrt_data.h
@@ -0,0 +1,13 @@
+#ifndef _SQRT_DATA_H
+#define _SQRT_DATA_H
+
+#include <features.h>
+#include <stdint.h>
+
+/* if x in [1,2): i = (int)(64*x);
+   if x in [2,4): i = (int)(32*x-64);
+   __rsqrt_tab[i]*2^-16 is estimating 1/sqrt(x) with small relative error:
+   |__rsqrt_tab[i]*0x1p-16*sqrt(x) - 1| < -0x1.fdp-9 < 2^-8 */
+extern hidden const uint16_t __rsqrt_tab[128];
+
+#endif
diff --git a/src/math/sqrtl.c b/src/math/sqrtl.c
new file mode 100644
index 0000000..1b9f19c
--- /dev/null
+++ b/src/math/sqrtl.c
@@ -0,0 +1,259 @@
+#include <stdint.h>
+#include <math.h>
+#include <float.h>
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double sqrtl(long double x)
+{
+	return sqrt(x);
+}
+#elif (LDBL_MANT_DIG == 113 || LDBL_MANT_DIG == 64) && LDBL_MAX_EXP == 16384
+#include "sqrt_data.h"
+
+#define FENV_SUPPORT 1
+
+typedef struct {
+	uint64_t hi;
+	uint64_t lo;
+} u128;
+
+/* top: 16 bit sign+exponent, x: significand.  */
+static inline long double mkldbl(uint64_t top, u128 x)
+{
+	union ldshape u;
+#if LDBL_MANT_DIG == 113
+	u.i2.hi = x.hi;
+	u.i2.lo = x.lo;
+	u.i2.hi &= 0x0000ffffffffffff;
+	u.i2.hi |= top << 48;
+#elif LDBL_MANT_DIG == 64
+	u.i.se = top;
+	u.i.m = x.lo;
+	/* force the top bit on non-zero (and non-subnormal) results.  */
+	if (top & 0x7fff)
+		u.i.m |= 0x8000000000000000;
+#endif
+	return u.f;
+}
+
+/* return: top 16 bit is sign+exp and following bits are the significand.  */
+static inline u128 asu128(long double x)
+{
+	union ldshape u = {.f=x};
+	u128 r;
+#if LDBL_MANT_DIG == 113
+	r.hi = u.i2.hi;
+	r.lo = u.i2.lo;
+#elif LDBL_MANT_DIG == 64
+	r.lo = u.i.m<<49;
+	/* ignore the top bit: pseudo numbers are not handled. */
+	r.hi = u.i.m>>15;
+	r.hi &= 0x0000ffffffffffff;
+	r.hi |= (uint64_t)u.i.se << 48;
+#endif
+	return r;
+}
+
+/* returns a*b*2^-32 - e, with error 0 <= e < 1.  */
+static inline uint32_t mul32(uint32_t a, uint32_t b)
+{
+	return (uint64_t)a*b >> 32;
+}
+
+/* returns a*b*2^-64 - e, with error 0 <= e < 3.  */
+static inline uint64_t mul64(uint64_t a, uint64_t b)
+{
+	uint64_t ahi = a>>32;
+	uint64_t alo = a&0xffffffff;
+	uint64_t bhi = b>>32;
+	uint64_t blo = b&0xffffffff;
+	return ahi*bhi + (ahi*blo >> 32) + (alo*bhi >> 32);
+}
+
+static inline u128 add64(u128 a, uint64_t b)
+{
+	u128 r;
+	r.lo = a.lo + b;
+	r.hi = a.hi;
+	if (r.lo < a.lo)
+		r.hi++;
+	return r;
+}
+
+static inline u128 add128(u128 a, u128 b)
+{
+	u128 r;
+	r.lo = a.lo + b.lo;
+	r.hi = a.hi + b.hi;
+	if (r.lo < a.lo)
+		r.hi++;
+	return r;
+}
+
+static inline u128 sub64(u128 a, uint64_t b)
+{
+	u128 r;
+	r.lo = a.lo - b;
+	r.hi = a.hi;
+	if (a.lo < b)
+		r.hi--;
+	return r;
+}
+
+static inline u128 sub128(u128 a, u128 b)
+{
+	u128 r;
+	r.lo = a.lo - b.lo;
+	r.hi = a.hi - b.hi;
+	if (a.lo < b.lo)
+		r.hi--;
+	return r;
+}
+
+/* a<<n, 0 <= n <= 127 */
+static inline u128 lsh(u128 a, int n)
+{
+	if (n == 0)
+		return a;
+	if (n >= 64) {
+		a.hi = a.lo<<(n-64);
+		a.lo = 0;
+	} else {
+		a.hi = (a.hi<<n) | (a.lo>>(64-n));
+		a.lo = a.lo<<n;
+	}
+	return a;
+}
+
+/* a>>n, 0 <= n <= 127 */
+static inline u128 rsh(u128 a, int n)
+{
+	if (n == 0)
+		return a;
+	if (n >= 64) {
+		a.lo = a.hi>>(n-64);
+		a.hi = 0;
+	} else {
+		a.lo = (a.lo>>n) | (a.hi<<(64-n));
+		a.hi = a.hi>>n;
+	}
+	return a;
+}
+
+/* returns a*b exactly.  */
+static inline u128 mul64_128(uint64_t a, uint64_t b)
+{
+	u128 r;
+	uint64_t ahi = a>>32;
+	uint64_t alo = a&0xffffffff;
+	uint64_t bhi = b>>32;
+	uint64_t blo = b&0xffffffff;
+	uint64_t lo1 = ((ahi*blo)&0xffffffff) + ((alo*bhi)&0xffffffff) + (alo*blo>>32);
+	uint64_t lo2 = (alo*blo)&0xffffffff;
+	r.hi = ahi*bhi + (ahi*blo>>32) + (alo*bhi>>32) + (lo1>>32);
+	r.lo = (lo1<<32) + lo2;
+	return r;
+}
+
+/* returns a*b*2^-128 - e, with error 0 <= e < 7.  */
+static inline u128 mul128(u128 a, u128 b)
+{
+	u128 hi = mul64_128(a.hi, b.hi);
+	uint64_t m1 = mul64(a.hi, b.lo);
+	uint64_t m2 = mul64(a.lo, b.hi);
+	return add64(add64(hi, m1), m2);
+}
+
+/* returns a*b % 2^128.  */
+static inline u128 mul128_tail(u128 a, u128 b)
+{
+	u128 lo = mul64_128(a.lo, b.lo);
+	lo.hi += a.hi*b.lo + a.lo*b.hi;
+	return lo;
+}
+
+
+/* see sqrt.c for detailed comments.  */
+
+long double sqrtl(long double x)
+{
+	u128 ix, ml;
+	uint64_t top;
+
+	ix = asu128(x);
+	top = ix.hi >> 48;
+	if (predict_false(top - 0x0001 >= 0x7fff - 0x0001)) {
+		/* x < 0x1p-16382 or inf or nan.  */
+		if (2*ix.hi == 0 && ix.lo == 0)
+			return x;
+		if (ix.hi == 0x7fff000000000000 && ix.lo == 0)
+			return x;
+		if (top >= 0x7fff)
+			return __math_invalidl(x);
+		/* x is subnormal, normalize it.  */
+		ix = asu128(x * 0x1p112);
+		top = ix.hi >> 48;
+		top -= 112;
+	}
+
+	/* x = 4^e m; with int e and m in [1, 4) */
+	int even = top & 1;
+	ml = lsh(ix, 15);
+	ml.hi |= 0x8000000000000000;
+	if (even) ml = rsh(ml, 1);
+	top = (top + 0x3fff) >> 1;
+
+	/* r ~ 1/sqrt(m) */
+	static const uint64_t three = 0xc0000000;
+	uint64_t r, s, d, u, i;
+	i = (ix.hi >> 42) % 128;
+	r = (uint32_t)__rsqrt_tab[i] << 16;
+	/* |r sqrt(m) - 1| < 0x1p-8 */
+	s = mul32(ml.hi>>32, r);
+	d = mul32(s, r);
+	u = three - d;
+	r = mul32(u, r) << 1;
+	/* |r sqrt(m) - 1| < 0x1.7bp-16, switch to 64bit */
+	r = r<<32;
+	s = mul64(ml.hi, r);
+	d = mul64(s, r);
+	u = (three<<32) - d;
+	r = mul64(u, r) << 1;
+	/* |r sqrt(m) - 1| < 0x1.a5p-31 */
+	s = mul64(u, s) << 1;
+	d = mul64(s, r);
+	u = (three<<32) - d;
+	r = mul64(u, r) << 1;
+	/* |r sqrt(m) - 1| < 0x1.c001p-59, switch to 128bit */
+
+	static const u128 threel = {.hi=three<<32, .lo=0};
+	u128 rl, sl, dl, ul;
+	rl.hi = r;
+	rl.lo = 0;
+	sl = mul128(ml, rl);
+	dl = mul128(sl, rl);
+	ul = sub128(threel, dl);
+	sl = mul128(ul, sl); /* repr: 3.125 */
+	/* -0x1p-116 < s - sqrt(m) < 0x3.8001p-125 */
+	sl = rsh(sub64(sl, 4), 125-(LDBL_MANT_DIG-1));
+	/* s < sqrt(m) < s + 1 ULP + tiny */
+
+	long double y;
+	u128 d2, d1, d0;
+	d0 = sub128(lsh(ml, 2*(LDBL_MANT_DIG-1)-126), mul128_tail(sl,sl));
+	d1 = sub128(sl, d0);
+	d2 = add128(add64(sl, 1), d1);
+	sl = add64(sl, d1.hi >> 63);
+	y = mkldbl(top, sl);
+	if (FENV_SUPPORT) {
+		/* handle rounding modes and inexact exception.  */
+		top = predict_false((d2.hi|d2.lo)==0) ? 0 : 1;
+		top |= ((d1.hi^d2.hi)&0x8000000000000000) >> 48;
+		y += mkldbl(top, (u128){0});
+	}
+	return y;
+}
+#else
+#error unsupported long double format
+#endif
-- 
Gitee


From 2c7a52790135405d01d890d6f07660559826cf6d Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Sat, 3 Sep 2022 15:25:38 +0800
Subject: [PATCH 03/13] feat: update stdlib, exit

---
 doc/c99.md                               |   9 +-
 src/Makefile                             |   2 +
 src/arch/riscv64/include/atomic_arch.h   |  38 +++
 src/arch/riscv64/include/bits/alltypes.h |   2 +
 src/arch/riscv64/include/endian.h        |  33 ---
 src/arch/riscv64/include/float.h         |  65 -----
 src/arch/riscv64/include/limits.h        |  66 -----
 src/arch/x86/include/atomic_arch.h       | 108 ++++++++
 src/arch/x86/include/bits/alltypes.h     |   2 +
 src/arch/x86/include/endian.h            |  33 ---
 src/arch/x86/include/float.h             |  65 -----
 src/arch/x86/include/limits.h            |  66 -----
 src/complex/csqrt.c                      |   2 +-
 src/env/__libc_start_main.c              |  18 +-
 src/env/getenv.c                         |   8 +
 src/exit/_Exit.c                         |   7 +
 src/exit/abort.c                         |   3 +-
 src/exit/atexit.c                        |  73 +++++
 src/exit/exit.c                          |  30 +-
 src/include/endian.h                     |  62 +++++
 src/include/features.h                   |   1 +
 src/include/iso646.h                     |  20 ++
 src/include/limits.h                     |  43 +++
 src/include/stdint.h                     |   4 +
 src/include/stdlib.h                     |  47 ++--
 src/internal/atomic.h                    | 333 +++++++++++++++++++++++
 src/internal/libc.h                      |   3 +-
 src/internal/libm.h                      |   8 +-
 src/internal/lock.h                      |   7 +
 src/prng/rand.c                          |  15 +
 src/process/system.c                     |  42 +++
 src/stdlib/bsearch.c                     |  20 ++
 src/stdlib/malloc.c                      |  22 ++
 src/stdlib/qsort.c                       | 218 +++++++++++++++
 34 files changed, 1104 insertions(+), 371 deletions(-)
 create mode 100644 src/arch/riscv64/include/atomic_arch.h
 create mode 100644 src/arch/riscv64/include/bits/alltypes.h
 delete mode 100644 src/arch/riscv64/include/endian.h
 delete mode 100644 src/arch/riscv64/include/float.h
 delete mode 100644 src/arch/riscv64/include/limits.h
 create mode 100644 src/arch/x86/include/atomic_arch.h
 create mode 100644 src/arch/x86/include/bits/alltypes.h
 delete mode 100644 src/arch/x86/include/endian.h
 delete mode 100644 src/arch/x86/include/float.h
 delete mode 100644 src/arch/x86/include/limits.h
 create mode 100644 src/env/getenv.c
 create mode 100644 src/exit/_Exit.c
 create mode 100644 src/exit/atexit.c
 create mode 100644 src/include/endian.h
 create mode 100644 src/include/iso646.h
 create mode 100644 src/include/limits.h
 create mode 100644 src/internal/atomic.h
 create mode 100644 src/internal/lock.h
 create mode 100644 src/prng/rand.c
 create mode 100644 src/process/system.c
 create mode 100644 src/stdlib/bsearch.c
 create mode 100644 src/stdlib/malloc.c
 create mode 100644 src/stdlib/qsort.c

diff --git a/doc/c99.md b/doc/c99.md
index a2637b8..0ee71ad 100644
--- a/doc/c99.md
+++ b/doc/c99.md
@@ -31,4 +31,11 @@
 | wchar.h |   | 首先在1995年第一次修订时引进,用于支持多字节和宽字节函数 |
 | wctype.h |   | 首先在1995年第一次修订时引进,用于支持多字节和宽字节分类函数 |
 
-## assert.h
+## 比较复杂的头文件
+
+* stdlib.h
+* locale.h
+* math.h
+* time.h
+* wchar.h
+* wctype.h
\ No newline at end of file
diff --git a/src/Makefile b/src/Makefile
index 5f6399f..5bb5262 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -43,6 +43,8 @@ X_LIBDIRS	:= $(LIBS_DIR)
 # we must link nxbase lib.
 X_LIBS		+= libnxbase.a
 
+SRC			+= process/
+SRC			+= prng/
 SRC			+= complex/
 SRC			+= fenv/
 SRC			+= fenv/$(ARCH)/
diff --git a/src/arch/riscv64/include/atomic_arch.h b/src/arch/riscv64/include/atomic_arch.h
new file mode 100644
index 0000000..0c38258
--- /dev/null
+++ b/src/arch/riscv64/include/atomic_arch.h
@@ -0,0 +1,38 @@
+#define a_barrier a_barrier
+static inline void a_barrier()
+{
+	__asm__ __volatile__ ("fence rw,rw" : : : "memory");
+}
+
+#define a_cas a_cas
+static inline int a_cas(volatile int *p, int t, int s)
+{
+	int old, tmp;
+	__asm__ __volatile__ (
+		"\n1:	lr.w.aqrl %0, (%2)\n"
+		"	bne %0, %3, 1f\n"
+		"	sc.w.aqrl %1, %4, (%2)\n"
+		"	bnez %1, 1b\n"
+		"1:"
+		: "=&r"(old), "=&r"(tmp)
+		: "r"(p), "r"((long)t), "r"((long)s)
+		: "memory");
+	return old;
+}
+
+#define a_cas_p a_cas_p
+static inline void *a_cas_p(volatile void *p, void *t, void *s)
+{
+	void *old;
+	int tmp;
+	__asm__ __volatile__ (
+		"\n1:	lr.d.aqrl %0, (%2)\n"
+		"	bne %0, %3, 1f\n"
+		"	sc.d.aqrl %1, %4, (%2)\n"
+		"	bnez %1, 1b\n"
+		"1:"
+		: "=&r"(old), "=&r"(tmp)
+		: "r"(p), "r"(t), "r"(s)
+		: "memory");
+	return old;
+}
diff --git a/src/arch/riscv64/include/bits/alltypes.h b/src/arch/riscv64/include/bits/alltypes.h
new file mode 100644
index 0000000..39b22aa
--- /dev/null
+++ b/src/arch/riscv64/include/bits/alltypes.h
@@ -0,0 +1,2 @@
+#define __BYTE_ORDER 1234
+#define __LONG_MAX 0x7fffffffffffffffL
diff --git a/src/arch/riscv64/include/endian.h b/src/arch/riscv64/include/endian.h
deleted file mode 100644
index 4f9cd6e..0000000
--- a/src/arch/riscv64/include/endian.h
+++ /dev/null
@@ -1,33 +0,0 @@
-#ifndef _RISCV64_ENDIAN_H
-#define _RISCV64_ENDIAN_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#ifndef LITTLE_ENDIAN
-#define LITTLE_ENDIAN	(1234)
-#endif
-#ifndef BIG_ENDIAN
-#define BIG_ENDIAN		(4321)
-#endif
-
-#if ( !defined(__LITTLE_ENDIAN) && !defined(__BIG_ENDIAN) )
-#define __LITTLE_ENDIAN
-#endif
-
-#if defined(__LITTLE_ENDIAN)
-#define __BYTE_ORDER		LITTLE_ENDIAN
-#elif defined(__BIG_ENDIAN)
-#define __BYTE_ORDER		BIG_ENDIAN
-#else
-#error "Unknown byte order!"
-#endif
-
-#define BYTE_ORDER __BYTE_ORDER
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _RISCV64_ENDIAN_H */
diff --git a/src/arch/riscv64/include/float.h b/src/arch/riscv64/include/float.h
deleted file mode 100644
index 6b405f7..0000000
--- a/src/arch/riscv64/include/float.h
+++ /dev/null
@@ -1,65 +0,0 @@
-#ifndef _RISCV64_FLOAT_H
-#define _RISCV64_FLOAT_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define FLT_RADIX			(2)
-
-#define FLT_TRUE_MIN		(1.40129846432481707092e-45F)
-#define FLT_MIN				(1.17549435082228750797e-38F)
-#define FLT_MAX				(3.40282346638528859812e+38F)
-#define FLT_EPSILON			(1.1920928955078125e-07F)
-
-#define FLT_MANT_DIG		(24)
-#define FLT_MIN_EXP			(-125)
-#define FLT_MAX_EXP			(128)
-#define FLT_HAS_SUBNORM		(1)
-
-#define FLT_DIG				(6)
-#define FLT_DECIMAL_DIG		(9)
-#define FLT_MIN_10_EXP		(-37)
-#define FLT_MAX_10_EXP		(38)
-
-#define DBL_TRUE_MIN		(4.94065645841246544177e-324)
-#define DBL_MIN				(2.22507385850720138309e-308)
-#define DBL_MAX				(1.79769313486231570815e+308)
-#define DBL_EPSILON			(2.22044604925031308085e-16)
-
-#define DBL_MANT_DIG		(53)
-#define DBL_MIN_EXP			(-1021)
-#define DBL_MAX_EXP			(1024)
-#define DBL_HAS_SUBNORM 	(1)
-
-#define DBL_DIG				(15)
-#define DBL_DECIMAL_DIG		(17)
-#define DBL_MIN_10_EXP		(-307)
-#define DBL_MAX_10_EXP		(308)
-
-#define LDBL_HAS_SUBNORM	(1)
-#define LDBL_DECIMAL_DIG	(DECIMAL_DIG)
-
-/* RISCV64 */
-#define FLT_EVAL_METHOD		(0)
-
-#define LDBL_TRUE_MIN		(6.47517511943802511092443895822764655e-4966L)
-#define LDBL_MIN			(3.36210314311209350626267781732175260e-4932L)
-#define LDBL_MAX			(1.18973149535723176508575932662800702e+4932L)
-#define LDBL_EPSILON		(1.92592994438723585305597794258492732e-34L)
-
-#define LDBL_MANT_DIG		(113)
-#define LDBL_MIN_EXP		(-16381)
-#define LDBL_MAX_EXP		(16384)
-
-#define LDBL_DIG			(33)
-#define LDBL_MIN_10_EXP		(-4931)
-#define LDBL_MAX_10_EXP		(4932)
-
-#define DECIMAL_DIG			(36)
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _RISCV64_FLOAT_H */
diff --git a/src/arch/riscv64/include/limits.h b/src/arch/riscv64/include/limits.h
deleted file mode 100644
index 4908d22..0000000
--- a/src/arch/riscv64/include/limits.h
+++ /dev/null
@@ -1,66 +0,0 @@
-#ifndef _RISCV64_LIMITS_H
-#define _RISCV64_LIMITS_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/* Number of bits in a 'char' */
-#define CHAR_BIT		(8)
-
-/* Minimum and maximum values a 'signed char' can hold */
-#define SCHAR_MIN		(-128)
-#define SCHAR_MAX		127
-
-/* Minimum and maximum values a 'char' can hold */
-#define CHAR_MIN		(-128)
-#define CHAR_MAX		127
-
-/* Maximum value an 'unsigned char' can hold (Minimum is 0) */
-#define UCHAR_MAX		255
-
-/* Minimum and maximum values a 'signed short int' can hold */
-#define SHRT_MIN		(-1 - 0x7fff)
-#define SHRT_MAX		0x7fff
-
-/* Maximum value an 'unsigned short int' can hold (Minimum is 0) */
-#define USHRT_MAX		0xffff
-
-/* Minimum and maximum values a 'signed int' can hold */
-#define INT_MIN			(-1 - 0x7fffffff)
-#define INT_MAX			0x7fffffff
-
-/* Maximum value an 'unsigned int' can hold (Minimum is 0) */
-#define UINT_MAX		0xffffffffU
-
-/* Minimum and maximum values a 'signed long int' can hold */
-#define LONG_MIN		(-LONG_MAX - 1)
-#define LONG_MAX		0x7fffffffffffffffL
-
-/* Maximum value an 'unsigned long int' can hold (Minimum is 0) */
-#define ULONG_MAX		(2UL * LONG_MAX + 1)
-
-/* Minimum and maximum values a 'signed long long int' can hold */
-#define LLONG_MIN		(-LLONG_MAX - 1)
-#define LLONG_MAX		0x7fffffffffffffffLL
-
-/* Maximum value an 'unsigned long long int' can hold (Minimum is 0) */
-#define ULLONG_MAX		(2ULL * LLONG_MAX + 1)
-
-/* Minimum and maximum values a 'max int' can hold */
-#define INTMAX_MIN		LLONG_MIN
-#define INTMAX_MAX		LLONG_MAX
-
-/* Maximum value an 'max uint' can hold (Minimum is 0) */
-#define UINTMAX_MAX		ULLONG_MAX
-
-#define NL_ARGMAX 9
-#define NL_MSGMAX 32767
-#define NL_SETMAX 255
-#define NL_TEXTMAX 2048
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _RISCV64_LIMITS_H */
diff --git a/src/arch/x86/include/atomic_arch.h b/src/arch/x86/include/atomic_arch.h
new file mode 100644
index 0000000..047fb68
--- /dev/null
+++ b/src/arch/x86/include/atomic_arch.h
@@ -0,0 +1,108 @@
+#define a_cas a_cas
+static inline int a_cas(volatile int *p, int t, int s)
+{
+	__asm__ __volatile__ (
+		"lock ; cmpxchg %3, %1"
+		: "=a"(t), "=m"(*p) : "a"(t), "r"(s) : "memory" );
+	return t;
+}
+
+#define a_swap a_swap
+static inline int a_swap(volatile int *p, int v)
+{
+	__asm__ __volatile__(
+		"xchg %0, %1"
+		: "=r"(v), "=m"(*p) : "0"(v) : "memory" );
+	return v;
+}
+
+#define a_fetch_add a_fetch_add
+static inline int a_fetch_add(volatile int *p, int v)
+{
+	__asm__ __volatile__(
+		"lock ; xadd %0, %1"
+		: "=r"(v), "=m"(*p) : "0"(v) : "memory" );
+	return v;
+}
+
+#define a_and a_and
+static inline void a_and(volatile int *p, int v)
+{
+	__asm__ __volatile__(
+		"lock ; and %1, %0"
+		: "=m"(*p) : "r"(v) : "memory" );
+}
+
+#define a_or a_or
+static inline void a_or(volatile int *p, int v)
+{
+	__asm__ __volatile__(
+		"lock ; or %1, %0"
+		: "=m"(*p) : "r"(v) : "memory" );
+}
+
+#define a_inc a_inc
+static inline void a_inc(volatile int *p)
+{
+	__asm__ __volatile__(
+		"lock ; incl %0"
+		: "=m"(*p) : "m"(*p) : "memory" );
+}
+
+#define a_dec a_dec
+static inline void a_dec(volatile int *p)
+{
+	__asm__ __volatile__(
+		"lock ; decl %0"
+		: "=m"(*p) : "m"(*p) : "memory" );
+}
+
+#define a_store a_store
+static inline void a_store(volatile int *p, int x)
+{
+	__asm__ __volatile__(
+		"mov %1, %0 ; lock ; orl $0,(%%esp)"
+		: "=m"(*p) : "r"(x) : "memory" );
+}
+
+#define a_barrier a_barrier
+static inline void a_barrier()
+{
+	__asm__ __volatile__( "" : : : "memory" );
+}
+
+#define a_spin a_spin
+static inline void a_spin()
+{
+	__asm__ __volatile__( "pause" : : : "memory" );
+}
+
+#define a_crash a_crash
+static inline void a_crash()
+{
+	__asm__ __volatile__( "hlt" : : : "memory" );
+}
+
+#define a_ctz_64 a_ctz_64
+static inline int a_ctz_64(uint64_t x)
+{
+	int r;
+	__asm__( "bsf %1,%0 ; jnz 1f ; bsf %2,%0 ; add $32,%0\n1:"
+		: "=&r"(r) : "r"((unsigned)x), "r"((unsigned)(x>>32)) );
+	return r;
+}
+
+#define a_ctz_32 a_ctz_32
+static inline int a_ctz_32(uint32_t x)
+{
+	int r;
+	__asm__( "bsf %1,%0" : "=r"(r) : "r"(x) );
+	return r;
+}
+
+#define a_clz_32 a_clz_32
+static inline int a_clz_32(uint32_t x)
+{
+	__asm__( "bsr %1,%0 ; xor $31,%0" : "=r"(x) : "r"(x) );
+	return x;
+}
diff --git a/src/arch/x86/include/bits/alltypes.h b/src/arch/x86/include/bits/alltypes.h
new file mode 100644
index 0000000..abf403b
--- /dev/null
+++ b/src/arch/x86/include/bits/alltypes.h
@@ -0,0 +1,2 @@
+#define __BYTE_ORDER 1234
+#define __LONG_MAX 0x7fffffffL
diff --git a/src/arch/x86/include/endian.h b/src/arch/x86/include/endian.h
deleted file mode 100644
index 1c36e3b..0000000
--- a/src/arch/x86/include/endian.h
+++ /dev/null
@@ -1,33 +0,0 @@
-#ifndef _X86_ENDIAN_H
-#define _X86_ENDIAN_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#ifndef LITTLE_ENDIAN
-#define LITTLE_ENDIAN	(1234)
-#endif
-#ifndef BIG_ENDIAN
-#define BIG_ENDIAN		(4321)
-#endif
-
-#if ( !defined(__LITTLE_ENDIAN) && !defined(__BIG_ENDIAN) )
-#define __LITTLE_ENDIAN
-#endif
-
-#if defined(__LITTLE_ENDIAN)
-#define __BYTE_ORDER		LITTLE_ENDIAN
-#elif defined(__BIG_ENDIAN)
-#define __BYTE_ORDER		BIG_ENDIAN
-#else
-#error "Unknown byte order!"
-#endif
-
-#define BYTE_ORDER __BYTE_ORDER
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _X86_ENDIAN_H */
diff --git a/src/arch/x86/include/float.h b/src/arch/x86/include/float.h
deleted file mode 100644
index b29fd0a..0000000
--- a/src/arch/x86/include/float.h
+++ /dev/null
@@ -1,65 +0,0 @@
-#ifndef _X86_FLOAT_H
-#define _X86_FLOAT_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define FLT_RADIX			(2)
-
-#define FLT_TRUE_MIN		(1.40129846432481707092e-45F)
-#define FLT_MIN				(1.17549435082228750797e-38F)
-#define FLT_MAX				(3.40282346638528859812e+38F)
-#define FLT_EPSILON			(1.1920928955078125e-07F)
-
-#define FLT_MANT_DIG		(24)
-#define FLT_MIN_EXP			(-125)
-#define FLT_MAX_EXP			(128)
-#define FLT_HAS_SUBNORM		(1)
-
-#define FLT_DIG				(6)
-#define FLT_DECIMAL_DIG		(9)
-#define FLT_MIN_10_EXP		(-37)
-#define FLT_MAX_10_EXP		(38)
-
-#define DBL_TRUE_MIN		(4.94065645841246544177e-324)
-#define DBL_MIN				(2.22507385850720138309e-308)
-#define DBL_MAX				(1.79769313486231570815e+308)
-#define DBL_EPSILON			(2.22044604925031308085e-16)
-
-#define DBL_MANT_DIG		(53)
-#define DBL_MIN_EXP			(-1021)
-#define DBL_MAX_EXP			(1024)
-#define DBL_HAS_SUBNORM 	(1)
-
-#define DBL_DIG				(15)
-#define DBL_DECIMAL_DIG		(17)
-#define DBL_MIN_10_EXP		(-307)
-#define DBL_MAX_10_EXP		(308)
-
-#define LDBL_HAS_SUBNORM	(1)
-#define LDBL_DECIMAL_DIG	(DECIMAL_DIG)
-
-/* x86 */
-#define FLT_EVAL_METHOD 2
-
-#define LDBL_TRUE_MIN 3.6451995318824746025e-4951L
-#define LDBL_MIN     3.3621031431120935063e-4932L
-#define LDBL_MAX     1.1897314953572317650e+4932L
-#define LDBL_EPSILON 1.0842021724855044340e-19L
-
-#define LDBL_MANT_DIG 64
-#define LDBL_MIN_EXP (-16381)
-#define LDBL_MAX_EXP 16384
-
-#define LDBL_DIG 18
-#define LDBL_MIN_10_EXP (-4931)
-#define LDBL_MAX_10_EXP 4932
-
-#define DECIMAL_DIG 21
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _X86_FLOAT_H */
diff --git a/src/arch/x86/include/limits.h b/src/arch/x86/include/limits.h
deleted file mode 100644
index 637d049..0000000
--- a/src/arch/x86/include/limits.h
+++ /dev/null
@@ -1,66 +0,0 @@
-#ifndef _X86_LIMITS_H
-#define _X86_LIMITS_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/* Number of bits in a 'char' */
-#define CHAR_BIT		(8)
-
-/* Minimum and maximum values a 'signed char' can hold */
-#define SCHAR_MIN		(-128)
-#define SCHAR_MAX		127
-
-/* Minimum and maximum values a 'char' can hold */
-#define CHAR_MIN		(-128)
-#define CHAR_MAX		127
-
-/* Maximum value an 'unsigned char' can hold (Minimum is 0) */
-#define UCHAR_MAX		255
-
-/* Minimum and maximum values a 'signed short int' can hold */
-#define SHRT_MIN		(-1 - 0x7fff)
-#define SHRT_MAX		0x7fff
-
-/* Maximum value an 'unsigned short int' can hold (Minimum is 0) */
-#define USHRT_MAX		0xffff
-
-/* Minimum and maximum values a 'signed int' can hold */
-#define INT_MIN			(-1 - 0x7fffffff)
-#define INT_MAX			0x7fffffff
-
-/* Maximum value an 'unsigned int' can hold (Minimum is 0) */
-#define UINT_MAX		0xffffffffU
-
-/* Minimum and maximum values a 'signed long int' can hold */
-#define LONG_MIN		(-LONG_MAX - 1)
-#define LONG_MAX		0x7fffffffL
-
-/* Maximum value an 'unsigned long int' can hold (Minimum is 0) */
-#define ULONG_MAX		(2UL * LONG_MAX + 1)
-
-/* Minimum and maximum values a 'signed long long int' can hold */
-#define LLONG_MIN		(-LLONG_MAX - 1)
-#define LLONG_MAX		0x7fffffffffffffffLL
-
-/* Maximum value an 'unsigned long long int' can hold (Minimum is 0) */
-#define ULLONG_MAX		(2ULL * LLONG_MAX + 1)
-
-/* Minimum and maximum values a 'max int' can hold */
-#define INTMAX_MIN		LLONG_MIN
-#define INTMAX_MAX		LLONG_MAX
-
-/* Maximum value an 'max uint' can hold (Minimum is 0) */
-#define UINTMAX_MAX		ULLONG_MAX
-
-#define NL_ARGMAX 9
-#define NL_MSGMAX 32767
-#define NL_SETMAX 255
-#define NL_TEXTMAX 2048
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _X86_LIMITS_H */
diff --git a/src/complex/csqrt.c b/src/complex/csqrt.c
index c36de00..e61745d 100644
--- a/src/complex/csqrt.c
+++ b/src/complex/csqrt.c
@@ -34,7 +34,7 @@
  * gcc generates is acceptable, since the special cases have already been
  * handled.
  */
-#pragma STDC CX_LIMITED_RANGE ON
+// #pragma STDC CX_LIMITED_RANGE ON
 
 /* We risk spurious overflow for components >= DBL_MAX / (1 + sqrt(2)). */
 #define THRESH  0x1.a827999fcef32p+1022
diff --git a/src/env/__libc_start_main.c b/src/env/__libc_start_main.c
index 3109e61..e583225 100644
--- a/src/env/__libc_start_main.c
+++ b/src/env/__libc_start_main.c
@@ -8,10 +8,10 @@
 
 extern int main(int argc, char *argv[]);
 
-void __init_libc(char **envp, char *pn)
+void __init_libc(char * envline)
 {
     libc.page_size = 4096;
-
+    libc.envline = envline;
     __init_stdio();
 }
 
@@ -20,18 +20,10 @@ NX_Error NX_WEAK_SYM NX_Main(char * cmdline, char * envline)
     /* build cmdline and envline */
     char *argv[MAX_ARGS + 1];
     int argc;
-    int ret;
 
     argc = NX_CmdToArray(cmdline, argv, MAX_ARGS);
 
-    __init_libc(NX_NULL, NX_NULL);
-    ret = main(argc, argv);
-    if (ret == EXIT_SUCCESS)
-    {
-        return NX_EOK;
-    }
-    else
-    {
-        return NX_ERROR;
-    }
+    __init_libc(envline);
+    exit(main(argc, argv));
+    return NX_EOK;
 }
diff --git a/src/env/getenv.c b/src/env/getenv.c
new file mode 100644
index 0000000..13bd824
--- /dev/null
+++ b/src/env/getenv.c
@@ -0,0 +1,8 @@
+#include <stdlib.h>
+#include <string.h>
+
+char *getenv(const char *name)
+{
+    /* TODO: add env name */
+	return 0;
+}
diff --git a/src/exit/_Exit.c b/src/exit/_Exit.c
new file mode 100644
index 0000000..a01c356
--- /dev/null
+++ b/src/exit/_Exit.c
@@ -0,0 +1,7 @@
+#include <stdlib.h>
+#include <nxbase.h>
+
+_Noreturn void _Exit(int ec)
+{
+	for (;;) NX_ProcessExit(ec);
+}
diff --git a/src/exit/abort.c b/src/exit/abort.c
index a7e65ca..1ad3566 100644
--- a/src/exit/abort.c
+++ b/src/exit/abort.c
@@ -16,5 +16,6 @@ _Noreturn void abort(void)
 
     NX_SignalSend(pid, NX_SIGNAL_ABORT, 0);
     NX_SignalSend(pid, NX_SIGNAL_KILL, 0);
-    while (1);
+    
+	_Exit(127);
 }
diff --git a/src/exit/atexit.c b/src/exit/atexit.c
new file mode 100644
index 0000000..5324c16
--- /dev/null
+++ b/src/exit/atexit.c
@@ -0,0 +1,73 @@
+#include <stdlib.h>
+#include <stdint.h>
+#include "libc.h"
+#include "lock.h"
+
+/* Ensure that at least 32 atexit handlers can be registered without malloc */
+#define COUNT 32
+
+static struct fl
+{
+	struct fl *next;
+	void (*f[COUNT])(void *);
+	void *a[COUNT];
+} builtin, *head;
+
+static int slot;
+static volatile int lock[1];
+volatile int *const __atexit_lockptr = lock;
+
+void __funcs_on_exit()
+{
+	void (*func)(void *), *arg;
+	LOCK(lock);
+	for (; head; head=head->next, slot=COUNT) while(slot-->0) {
+		func = head->f[slot];
+		arg = head->a[slot];
+		UNLOCK(lock);
+		func(arg);
+		LOCK(lock);
+	}
+}
+
+void __cxa_finalize(void *dso)
+{
+}
+
+int __cxa_atexit(void (*func)(void *), void *arg, void *dso)
+{
+	LOCK(lock);
+
+	/* Defer initialization of head so it can be in BSS */
+	if (!head) head = &builtin;
+
+	/* If the current function list is full, add a new one */
+	if (slot==COUNT) {
+		struct fl *new_fl = calloc(sizeof(struct fl), 1);
+		if (!new_fl) {
+			UNLOCK(lock);
+			return -1;
+		}
+		new_fl->next = head;
+		head = new_fl;
+		slot = 0;
+	}
+
+	/* Append function to the list. */
+	head->f[slot] = func;
+	head->a[slot] = arg;
+	slot++;
+
+	UNLOCK(lock);
+	return 0;
+}
+
+static void call(void *p)
+{
+	((void (*)(void))(uintptr_t)p)();
+}
+
+int atexit(void (*func)(void))
+{
+	return __cxa_atexit(call, (void *)(uintptr_t)func, 0);
+}
diff --git a/src/exit/exit.c b/src/exit/exit.c
index 50872f2..04a9308 100644
--- a/src/exit/exit.c
+++ b/src/exit/exit.c
@@ -1,8 +1,32 @@
 #include <stdlib.h>
-#include <nxbase.h>
+#include "libc.h"
+
+static void dummy()
+{
+}
+
+/* atexit.c and __stdio_exit.c override these. the latter is linked
+ * as a consequence of linking either __toread.c or __towrite.c. */
+weak_alias(dummy, __funcs_on_exit);
+weak_alias(dummy, __stdio_exit);
+weak_alias(dummy, _fini);
+
+extern weak_sym hidden void (*const __fini_array_start)(void), (*const __fini_array_end)(void);
+
+static void libc_exit_fini(void)
+{
+	uintptr_t a = (uintptr_t)&__fini_array_end;
+	for (; a>(uintptr_t)&__fini_array_start; a-=sizeof(void(*)()))
+		(*(void (**)())(a-sizeof(void(*)())))();
+	_fini();
+}
+
+weak_alias(libc_exit_fini, __libc_exit_fini);
 
 _Noreturn void exit(int code)
 {
-    NX_ProcessExit(code);
-    while (1);
+    __funcs_on_exit();
+	__libc_exit_fini();
+	__stdio_exit();
+	_Exit(code);
 }
diff --git a/src/include/endian.h b/src/include/endian.h
new file mode 100644
index 0000000..6dc4ac1
--- /dev/null
+++ b/src/include/endian.h
@@ -0,0 +1,62 @@
+#ifndef _ENDIAN_H
+#define _ENDIAN_H
+
+#include <features.h>
+#include <stdint.h>
+
+#include <bits/alltypes.h>
+
+#define __PDP_ENDIAN 3412
+
+#define __LITTLE_ENDIAN 1234
+#define __BIG_ENDIAN 4321
+
+#define BIG_ENDIAN __BIG_ENDIAN
+#define LITTLE_ENDIAN __LITTLE_ENDIAN
+#define PDP_ENDIAN __PDP_ENDIAN
+#define BYTE_ORDER __BYTE_ORDER
+
+static __inline uint16_t __bswap16(uint16_t __x)
+{
+	return (__x<<8) | (__x>>8);
+}
+
+static __inline uint32_t __bswap32(uint32_t __x)
+{
+	return (__x>>24) | (__x>>8&0xff00) | (__x<<8&0xff0000) | (__x<<24);
+}
+
+static __inline uint64_t __bswap64(uint64_t __x)
+{
+	return (__bswap32(__x)+(0ULL<<32)) | __bswap32(__x>>32);
+}
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define htobe16(x) __bswap16(x)
+#define be16toh(x) __bswap16(x)
+#define htobe32(x) __bswap32(x)
+#define be32toh(x) __bswap32(x)
+#define htobe64(x) __bswap64(x)
+#define be64toh(x) __bswap64(x)
+#define htole16(x) (uint16_t)(x)
+#define le16toh(x) (uint16_t)(x)
+#define htole32(x) (uint32_t)(x)
+#define le32toh(x) (uint32_t)(x)
+#define htole64(x) (uint64_t)(x)
+#define le64toh(x) (uint64_t)(x)
+#else
+#define htobe16(x) (uint16_t)(x)
+#define be16toh(x) (uint16_t)(x)
+#define htobe32(x) (uint32_t)(x)
+#define be32toh(x) (uint32_t)(x)
+#define htobe64(x) (uint64_t)(x)
+#define be64toh(x) (uint64_t)(x)
+#define htole16(x) __bswap16(x)
+#define le16toh(x) __bswap16(x)
+#define htole32(x) __bswap32(x)
+#define le32toh(x) __bswap32(x)
+#define htole64(x) __bswap64(x)
+#define le64toh(x) __bswap64(x)
+#endif
+
+#endif
diff --git a/src/include/features.h b/src/include/features.h
index 46f7f81..13f55d9 100644
--- a/src/include/features.h
+++ b/src/include/features.h
@@ -22,6 +22,7 @@
 
 #define __REDIR(x,y) __typeof__(x) x __asm__(#y)
 
+#define weak_sym __attribute__((__weak__))
 #define hidden __attribute__((__visibility__("hidden")))
 #define weak_alias(old, new) \
 	extern __typeof(old) new __attribute__((__weak__, __alias__(#old)))
diff --git a/src/include/iso646.h b/src/include/iso646.h
new file mode 100644
index 0000000..88ff53d
--- /dev/null
+++ b/src/include/iso646.h
@@ -0,0 +1,20 @@
+#ifndef _ISO646_H
+#define _ISO646_H
+
+#ifndef __cplusplus
+
+#define and    &&
+#define and_eq &=
+#define bitand &
+#define bitor  |
+#define compl  ~
+#define not    !
+#define not_eq !=
+#define or     ||
+#define or_eq  |=
+#define xor    ^
+#define xor_eq ^=
+
+#endif
+
+#endif
diff --git a/src/include/limits.h b/src/include/limits.h
new file mode 100644
index 0000000..981833f
--- /dev/null
+++ b/src/include/limits.h
@@ -0,0 +1,43 @@
+#ifndef _LIMITS_H
+#define _LIMITS_H
+
+#include <features.h>
+
+#include <bits/alltypes.h> /* __LONG_MAX */
+
+/* Support signed or unsigned plain-char */
+
+#if '\xff' > 0
+#define CHAR_MIN 0
+#define CHAR_MAX 255
+#else
+#define CHAR_MIN (-128)
+#define CHAR_MAX 127
+#endif
+
+#define CHAR_BIT 8
+#define SCHAR_MIN (-128)
+#define SCHAR_MAX 127
+#define UCHAR_MAX 255
+#define SHRT_MIN  (-1-0x7fff)
+#define SHRT_MAX  0x7fff
+#define USHRT_MAX 0xffff
+#define INT_MIN  (-1-0x7fffffff)
+#define INT_MAX  0x7fffffff
+#define UINT_MAX 0xffffffffU
+#define LONG_MIN (-LONG_MAX-1)
+#define LONG_MAX __LONG_MAX
+#define ULONG_MAX (2UL*LONG_MAX+1)
+#define LLONG_MIN (-LLONG_MAX-1)
+#define LLONG_MAX  0x7fffffffffffffffLL
+#define ULLONG_MAX (2ULL*LLONG_MAX+1)
+
+#define MB_LEN_MAX 4
+
+
+#define NL_ARGMAX 9
+#define NL_MSGMAX 32767
+#define NL_SETMAX 255
+#define NL_TEXTMAX 2048
+
+#endif
diff --git a/src/include/stdint.h b/src/include/stdint.h
index 49740c1..7b71c13 100644
--- a/src/include/stdint.h
+++ b/src/include/stdint.h
@@ -76,6 +76,10 @@ typedef uint64_t uint_least64_t;
 #define UINT_LEAST32_MAX UINT32_MAX
 #define UINT_LEAST64_MAX UINT64_MAX
 
+#define INTMAX_MIN  INT64_MIN
+#define INTMAX_MAX  INT64_MAX
+#define UINTMAX_MAX UINT64_MAX
+
 #define WINT_MIN 0U
 #define WINT_MAX UINT32_MAX
 
diff --git a/src/include/stdlib.h b/src/include/stdlib.h
index e70f7e7..c6ca66e 100644
--- a/src/include/stdlib.h
+++ b/src/include/stdlib.h
@@ -1,7 +1,6 @@
 #ifndef _STDLIB_H
 #define _STDLIB_H
 
-#include <nxbase.h>
 #include <features.h>
 
 #include <bits/types.h>
@@ -10,31 +9,41 @@
 extern "C" {
 #endif
 
+#ifndef NULL
+#ifdef __cplusplus
+#define NULL 0L
+#else
+#define NULL ((void*)0)
+#endif
+#endif
+
+/* program support utilities */
 #define EXIT_FAILURE 1
 #define EXIT_SUCCESS 0
 
-_Noreturn void exit (int);
-_Noreturn void abort(void);
+#define RAND_MAX (0x7fffffff)
 
-static inline void* malloc( size_t size )
-{
-    return (void *)NX_MemAlloc(size);
-}
+int rand (void);
+void srand (unsigned);
 
-static inline void *realloc( void *ptr, size_t new_size )
-{
-    return (void *)NX_MemReAlloc(ptr, new_size);
-}
+/* memory management */
+void* malloc( size_t size );
+void *realloc( void *ptr, size_t new_size );
+void* calloc( size_t num, size_t size );
+void free( void* ptr );
 
-static inline void* calloc( size_t num, size_t size )
-{
-    return (void *)NX_MemAlloc(num * size);
-}
+_Noreturn void exit (int);
+int atexit (void (*) (void));
+_Noreturn void abort(void);
+_Noreturn void _Exit (int);
 
-static inline void free( void* ptr )
-{
-    NX_MemFree(ptr);
-}
+/* Communicating with the environment */
+char *getenv (const char *);
+int system (const char *);
+
+/* Algorithms */
+void *bsearch (const void *, const void *, size_t, size_t, int (*)(const void *, const void *));
+void qsort (void *, size_t, size_t, int (*)(const void *, const void *));
 
 int mblen (const char *, size_t);
 int mbtowc (wchar_t *__restrict, const char *__restrict, size_t);
diff --git a/src/internal/atomic.h b/src/internal/atomic.h
new file mode 100644
index 0000000..96c1552
--- /dev/null
+++ b/src/internal/atomic.h
@@ -0,0 +1,333 @@
+#ifndef _ATOMIC_H
+#define _ATOMIC_H
+
+#include <stdint.h>
+
+#include "atomic_arch.h"
+
+#ifdef a_ll
+
+#ifndef a_pre_llsc
+#define a_pre_llsc()
+#endif
+
+#ifndef a_post_llsc
+#define a_post_llsc()
+#endif
+
+#ifndef a_cas
+#define a_cas a_cas
+static inline int a_cas(volatile int *p, int t, int s)
+{
+	int old;
+	a_pre_llsc();
+	do old = a_ll(p);
+	while (old==t && !a_sc(p, s));
+	a_post_llsc();
+	return old;
+}
+#endif
+
+#ifndef a_swap
+#define a_swap a_swap
+static inline int a_swap(volatile int *p, int v)
+{
+	int old;
+	a_pre_llsc();
+	do old = a_ll(p);
+	while (!a_sc(p, v));
+	a_post_llsc();
+	return old;
+}
+#endif
+
+#ifndef a_fetch_add
+#define a_fetch_add a_fetch_add
+static inline int a_fetch_add(volatile int *p, int v)
+{
+	int old;
+	a_pre_llsc();
+	do old = a_ll(p);
+	while (!a_sc(p, (unsigned)old + v));
+	a_post_llsc();
+	return old;
+}
+#endif
+
+#ifndef a_fetch_and
+#define a_fetch_and a_fetch_and
+static inline int a_fetch_and(volatile int *p, int v)
+{
+	int old;
+	a_pre_llsc();
+	do old = a_ll(p);
+	while (!a_sc(p, old & v));
+	a_post_llsc();
+	return old;
+}
+#endif
+
+#ifndef a_fetch_or
+#define a_fetch_or a_fetch_or
+static inline int a_fetch_or(volatile int *p, int v)
+{
+	int old;
+	a_pre_llsc();
+	do old = a_ll(p);
+	while (!a_sc(p, old | v));
+	a_post_llsc();
+	return old;
+}
+#endif
+
+#endif
+
+#ifdef a_ll_p
+
+#ifndef a_cas_p
+#define a_cas_p a_cas_p
+static inline void *a_cas_p(volatile void *p, void *t, void *s)
+{
+	void *old;
+	a_pre_llsc();
+	do old = a_ll_p(p);
+	while (old==t && !a_sc_p(p, s));
+	a_post_llsc();
+	return old;
+}
+#endif
+
+#endif
+
+#ifndef a_cas
+#error missing definition of a_cas
+#endif
+
+#ifndef a_swap
+#define a_swap a_swap
+static inline int a_swap(volatile int *p, int v)
+{
+	int old;
+	do old = *p;
+	while (a_cas(p, old, v) != old);
+	return old;
+}
+#endif
+
+#ifndef a_fetch_add
+#define a_fetch_add a_fetch_add
+static inline int a_fetch_add(volatile int *p, int v)
+{
+	int old;
+	do old = *p;
+	while (a_cas(p, old, (unsigned)old+v) != old);
+	return old;
+}
+#endif
+
+#ifndef a_fetch_and
+#define a_fetch_and a_fetch_and
+static inline int a_fetch_and(volatile int *p, int v)
+{
+	int old;
+	do old = *p;
+	while (a_cas(p, old, old&v) != old);
+	return old;
+}
+#endif
+#ifndef a_fetch_or
+#define a_fetch_or a_fetch_or
+static inline int a_fetch_or(volatile int *p, int v)
+{
+	int old;
+	do old = *p;
+	while (a_cas(p, old, old|v) != old);
+	return old;
+}
+#endif
+
+#ifndef a_and
+#define a_and a_and
+static inline void a_and(volatile int *p, int v)
+{
+	a_fetch_and(p, v);
+}
+#endif
+
+#ifndef a_or
+#define a_or a_or
+static inline void a_or(volatile int *p, int v)
+{
+	a_fetch_or(p, v);
+}
+#endif
+
+#ifndef a_inc
+#define a_inc a_inc
+static inline void a_inc(volatile int *p)
+{
+	a_fetch_add(p, 1);
+}
+#endif
+
+#ifndef a_dec
+#define a_dec a_dec
+static inline void a_dec(volatile int *p)
+{
+	a_fetch_add(p, -1);
+}
+#endif
+
+#ifndef a_store
+#define a_store a_store
+static inline void a_store(volatile int *p, int v)
+{
+#ifdef a_barrier
+	a_barrier();
+	*p = v;
+	a_barrier();
+#else
+	a_swap(p, v);
+#endif
+}
+#endif
+
+#ifndef a_barrier
+#define a_barrier a_barrier
+static void a_barrier()
+{
+	volatile int tmp = 0;
+	a_cas(&tmp, 0, 0);
+}
+#endif
+
+#ifndef a_spin
+#define a_spin a_barrier
+#endif
+
+#ifndef a_and_64
+#define a_and_64 a_and_64
+static inline void a_and_64(volatile uint64_t *p, uint64_t v)
+{
+	union { uint64_t v; uint32_t r[2]; } u = { v };
+	if (u.r[0]+1) a_and((int *)p, u.r[0]);
+	if (u.r[1]+1) a_and((int *)p+1, u.r[1]);
+}
+#endif
+
+#ifndef a_or_64
+#define a_or_64 a_or_64
+static inline void a_or_64(volatile uint64_t *p, uint64_t v)
+{
+	union { uint64_t v; uint32_t r[2]; } u = { v };
+	if (u.r[0]) a_or((int *)p, u.r[0]);
+	if (u.r[1]) a_or((int *)p+1, u.r[1]);
+}
+#endif
+
+#ifndef a_cas_p
+typedef char a_cas_p_undefined_but_pointer_not_32bit[-sizeof(char) == 0xffffffff ? 1 : -1];
+#define a_cas_p a_cas_p
+static inline void *a_cas_p(volatile void *p, void *t, void *s)
+{
+	return (void *)a_cas((volatile int *)p, (int)t, (int)s);
+}
+#endif
+
+#ifndef a_or_l
+#define a_or_l a_or_l
+static inline void a_or_l(volatile void *p, long v)
+{
+	if (sizeof(long) == sizeof(int)) a_or(p, v);
+	else a_or_64(p, v);
+}
+#endif
+
+#ifndef a_crash
+#define a_crash a_crash
+static inline void a_crash()
+{
+	*(volatile char *)0=0;
+}
+#endif
+
+#ifndef a_ctz_32
+#define a_ctz_32 a_ctz_32
+static inline int a_ctz_32(uint32_t x)
+{
+#ifdef a_clz_32
+	return 31-a_clz_32(x&-x);
+#else
+	static const char debruijn32[32] = {
+		0, 1, 23, 2, 29, 24, 19, 3, 30, 27, 25, 11, 20, 8, 4, 13,
+		31, 22, 28, 18, 26, 10, 7, 12, 21, 17, 9, 6, 16, 5, 15, 14
+	};
+	return debruijn32[(x&-x)*0x076be629 >> 27];
+#endif
+}
+#endif
+
+#ifndef a_ctz_64
+#define a_ctz_64 a_ctz_64
+static inline int a_ctz_64(uint64_t x)
+{
+	static const char debruijn64[64] = {
+		0, 1, 2, 53, 3, 7, 54, 27, 4, 38, 41, 8, 34, 55, 48, 28,
+		62, 5, 39, 46, 44, 42, 22, 9, 24, 35, 59, 56, 49, 18, 29, 11,
+		63, 52, 6, 26, 37, 40, 33, 47, 61, 45, 43, 21, 23, 58, 17, 10,
+		51, 25, 36, 32, 60, 20, 57, 16, 50, 31, 19, 15, 30, 14, 13, 12
+	};
+	if (sizeof(long) < 8) {
+		uint32_t y = x;
+		if (!y) {
+			y = x>>32;
+			return 32 + a_ctz_32(y);
+		}
+		return a_ctz_32(y);
+	}
+	return debruijn64[(x&-x)*0x022fdd63cc95386dull >> 58];
+}
+#endif
+
+static inline int a_ctz_l(unsigned long x)
+{
+	return (sizeof(long) < 8) ? a_ctz_32(x) : a_ctz_64(x);
+}
+
+#ifndef a_clz_64
+#define a_clz_64 a_clz_64
+static inline int a_clz_64(uint64_t x)
+{
+#ifdef a_clz_32
+	if (x>>32)
+		return a_clz_32(x>>32);
+	return a_clz_32(x) + 32;
+#else
+	uint32_t y;
+	int r;
+	if (x>>32) y=x>>32, r=0; else y=x, r=32;
+	if (y>>16) y>>=16; else r |= 16;
+	if (y>>8) y>>=8; else r |= 8;
+	if (y>>4) y>>=4; else r |= 4;
+	if (y>>2) y>>=2; else r |= 2;
+	return r | !(y>>1);
+#endif
+}
+#endif
+
+#ifndef a_clz_32
+#define a_clz_32 a_clz_32
+static inline int a_clz_32(uint32_t x)
+{
+	x >>= 1;
+	x |= x >> 1;
+	x |= x >> 2;
+	x |= x >> 4;
+	x |= x >> 8;
+	x |= x >> 16;
+	x++;
+	return 31-a_ctz_32(x);
+}
+#endif
+
+#endif
diff --git a/src/internal/libc.h b/src/internal/libc.h
index c43ef96..eb0be32 100644
--- a/src/internal/libc.h
+++ b/src/internal/libc.h
@@ -14,6 +14,7 @@ struct __locale_struct {
 struct __libc {
 	size_t page_size;
 	struct __locale_struct global_locale;
+    char * envline; /* nxos envline */
 };
 
 #ifndef PAGE_SIZE
@@ -23,7 +24,7 @@ struct __libc {
 extern hidden struct __libc __libc;
 #define libc __libc
 
-hidden void __init_libc(char **, char *);
+void __init_libc(char *);
 
 extern hidden const char __libc_version[];
 
diff --git a/src/internal/libm.h b/src/internal/libm.h
index 240332e..60847c4 100644
--- a/src/internal/libm.h
+++ b/src/internal/libm.h
@@ -7,7 +7,7 @@
 #include <endian.h>
 
 #if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
-#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384 && BYTE_ORDER == LITTLE_ENDIAN
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384 && __BYTE_ORDER == __LITTLE_ENDIAN
 union ldshape {
 	long double f;
 	struct {
@@ -15,7 +15,7 @@ union ldshape {
 		uint16_t se;
 	} i;
 };
-#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384 && BYTE_ORDER == __BIG_ENDIAN
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384 && __BYTE_ORDER == __BIG_ENDIAN
 /* This is the m68k variant of 80-bit long double, and this definition only works
  * on archs where the alignment requirement of uint64_t is <= 4. */
 union ldshape {
@@ -26,7 +26,7 @@ union ldshape {
 		uint64_t m;
 	} i;
 };
-#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384 && BYTE_ORDER == LITTLE_ENDIAN
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384 && __BYTE_ORDER == __LITTLE_ENDIAN
 union ldshape {
 	long double f;
 	struct {
@@ -40,7 +40,7 @@ union ldshape {
 		uint64_t hi;
 	} i2;
 };
-#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384 && BYTE_ORDER == __BIG_ENDIAN
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384 && __BYTE_ORDER == __BIG_ENDIAN
 union ldshape {
 	long double f;
 	struct {
diff --git a/src/internal/lock.h b/src/internal/lock.h
new file mode 100644
index 0000000..5596e6a
--- /dev/null
+++ b/src/internal/lock.h
@@ -0,0 +1,7 @@
+#ifndef LOCK_H
+#define LOCK_H
+
+#define LOCK(x) 
+#define UNLOCK(x) 
+
+#endif
diff --git a/src/prng/rand.c b/src/prng/rand.c
new file mode 100644
index 0000000..c000cd2
--- /dev/null
+++ b/src/prng/rand.c
@@ -0,0 +1,15 @@
+#include <stdlib.h>
+#include <stdint.h>
+
+static uint64_t seed;
+
+void srand(unsigned s)
+{
+	seed = s-1;
+}
+
+int rand(void)
+{
+	seed = 6364136223846793005ULL*seed + 1;
+	return seed>>33;
+}
diff --git a/src/process/system.c b/src/process/system.c
new file mode 100644
index 0000000..f2e39b0
--- /dev/null
+++ b/src/process/system.c
@@ -0,0 +1,42 @@
+#include <stdlib.h>
+#include <errno.h>
+#include <nxbase.h>
+#include "libc.h"
+
+int system(const char *cmd)
+{
+    NX_Error err;
+    char appPath[NX_VFS_MAX_NAME] = {0};
+    char *spacePos;
+    char * cmdline = (char *)cmd;
+    NX_U32 exitCode;
+
+    if (!cmdline)
+    {
+        return -1;
+    }
+
+    while (*cmdline == ' ') /* skip space in the head */
+    {
+        cmdline++;
+    }
+    
+    /* get app name */
+    spacePos = NX_StrChr(cmdline, ' ');
+    if (spacePos == NX_NULL) /* no space, only app name */
+    {
+        NX_StrCopy(appPath, cmdline);
+    }
+    else
+    {
+        NX_StrCopyN(appPath, cmdline, spacePos - cmdline + 1);
+    }
+
+	err = NX_ProcessLaunch(appPath, NX_PROC_FLAG_WAIT, &exitCode, cmdline, libc.envline);
+    if (err != NX_EOK)
+    {
+        return -1;
+    }
+
+	return 0;
+}
diff --git a/src/stdlib/bsearch.c b/src/stdlib/bsearch.c
new file mode 100644
index 0000000..fe050ea
--- /dev/null
+++ b/src/stdlib/bsearch.c
@@ -0,0 +1,20 @@
+#include <stdlib.h>
+
+void *bsearch(const void *key, const void *base, size_t nel, size_t width, int (*cmp)(const void *, const void *))
+{
+	void *try;
+	int sign;
+	while (nel > 0) {
+		try = (char *)base + width*(nel/2);
+		sign = cmp(key, try);
+		if (sign < 0) {
+			nel /= 2;
+		} else if (sign > 0) {
+			base = (char *)try + width;
+			nel -= nel/2+1;
+		} else {
+			return try;
+		}
+	}
+	return NULL;
+}
diff --git a/src/stdlib/malloc.c b/src/stdlib/malloc.c
new file mode 100644
index 0000000..fffb96b
--- /dev/null
+++ b/src/stdlib/malloc.c
@@ -0,0 +1,22 @@
+#include <stdlib.h>
+#include <nxbase.h>
+
+void* malloc( size_t size )
+{
+    return (void *)NX_MemAlloc(size);
+}
+
+void *realloc( void *ptr, size_t new_size )
+{
+    return (void *)NX_MemReAlloc(ptr, new_size);
+}
+
+void* calloc( size_t num, size_t size )
+{
+    return (void *)NX_MemAlloc(num * size);
+}
+
+void free( void* ptr )
+{
+    NX_MemFree(ptr);
+}
diff --git a/src/stdlib/qsort.c b/src/stdlib/qsort.c
new file mode 100644
index 0000000..da58fd3
--- /dev/null
+++ b/src/stdlib/qsort.c
@@ -0,0 +1,218 @@
+/* Copyright (C) 2011 by Valentin Ochs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/* Minor changes by Rich Felker for integration in musl, 2011-04-27. */
+
+/* Smoothsort, an adaptive variant of Heapsort.  Memory usage: O(1).
+   Run time: Worst case O(n log n), close to O(n) in the mostly-sorted case. */
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "atomic.h"
+#define ntz(x) a_ctz_l((x))
+
+typedef int (*cmpfun)(const void *, const void *);
+
+static inline int pntz(size_t p[2]) {
+	int r = ntz(p[0] - 1);
+	if(r != 0 || (r = 8*sizeof(size_t) + ntz(p[1])) != 8*sizeof(size_t)) {
+		return r;
+	}
+	return 0;
+}
+
+static void cycle(size_t width, unsigned char* ar[], int n)
+{
+	unsigned char tmp[256];
+	size_t l;
+	int i;
+
+	if(n < 2) {
+		return;
+	}
+
+	ar[n] = tmp;
+	while(width) {
+		l = sizeof(tmp) < width ? sizeof(tmp) : width;
+		memcpy(ar[n], ar[0], l);
+		for(i = 0; i < n; i++) {
+			memcpy(ar[i], ar[i + 1], l);
+			ar[i] += l;
+		}
+		width -= l;
+	}
+}
+
+/* shl() and shr() need n > 0 */
+static inline void shl(size_t p[2], int n)
+{
+	if(n >= 8 * sizeof(size_t)) {
+		n -= 8 * sizeof(size_t);
+		p[1] = p[0];
+		p[0] = 0;
+	}
+	p[1] <<= n;
+	p[1] |= p[0] >> (sizeof(size_t) * 8 - n);
+	p[0] <<= n;
+}
+
+static inline void shr(size_t p[2], int n)
+{
+	if(n >= 8 * sizeof(size_t)) {
+		n -= 8 * sizeof(size_t);
+		p[0] = p[1];
+		p[1] = 0;
+	}
+	p[0] >>= n;
+	p[0] |= p[1] << (sizeof(size_t) * 8 - n);
+	p[1] >>= n;
+}
+
+static void sift(unsigned char *head, size_t width, cmpfun cmp, int pshift, size_t lp[])
+{
+	unsigned char *rt, *lf;
+	unsigned char *ar[14 * sizeof(size_t) + 1];
+	int i = 1;
+
+	ar[0] = head;
+	while(pshift > 1) {
+		rt = head - width;
+		lf = head - width - lp[pshift - 2];
+
+		if((*cmp)(ar[0], lf) >= 0 && (*cmp)(ar[0], rt) >= 0) {
+			break;
+		}
+		if((*cmp)(lf, rt) >= 0) {
+			ar[i++] = lf;
+			head = lf;
+			pshift -= 1;
+		} else {
+			ar[i++] = rt;
+			head = rt;
+			pshift -= 2;
+		}
+	}
+	cycle(width, ar, i);
+}
+
+static void trinkle(unsigned char *head, size_t width, cmpfun cmp, size_t pp[2], int pshift, int trusty, size_t lp[])
+{
+	unsigned char *stepson,
+	              *rt, *lf;
+	size_t p[2];
+	unsigned char *ar[14 * sizeof(size_t) + 1];
+	int i = 1;
+	int trail;
+
+	p[0] = pp[0];
+	p[1] = pp[1];
+
+	ar[0] = head;
+	while(p[0] != 1 || p[1] != 0) {
+		stepson = head - lp[pshift];
+		if((*cmp)(stepson, ar[0]) <= 0) {
+			break;
+		}
+		if(!trusty && pshift > 1) {
+			rt = head - width;
+			lf = head - width - lp[pshift - 2];
+			if((*cmp)(rt, stepson) >= 0 || (*cmp)(lf, stepson) >= 0) {
+				break;
+			}
+		}
+
+		ar[i++] = stepson;
+		head = stepson;
+		trail = pntz(p);
+		shr(p, trail);
+		pshift += trail;
+		trusty = 0;
+	}
+	if(!trusty) {
+		cycle(width, ar, i);
+		sift(head, width, cmp, pshift, lp);
+	}
+}
+
+void qsort(void *base, size_t nel, size_t width, cmpfun cmp)
+{
+	size_t lp[12*sizeof(size_t)];
+	size_t i, size = width * nel;
+	unsigned char *head, *high;
+	size_t p[2] = {1, 0};
+	int pshift = 1;
+	int trail;
+
+	if (!size) return;
+
+	head = base;
+	high = head + size - width;
+
+	/* Precompute Leonardo numbers, scaled by element width */
+	for(lp[0]=lp[1]=width, i=2; (lp[i]=lp[i-2]+lp[i-1]+width) < size; i++);
+
+	while(head < high) {
+		if((p[0] & 3) == 3) {
+			sift(head, width, cmp, pshift, lp);
+			shr(p, 2);
+			pshift += 2;
+		} else {
+			if(lp[pshift - 1] >= high - head) {
+				trinkle(head, width, cmp, p, pshift, 0, lp);
+			} else {
+				sift(head, width, cmp, pshift, lp);
+			}
+			
+			if(pshift == 1) {
+				shl(p, 1);
+				pshift = 0;
+			} else {
+				shl(p, pshift - 1);
+				pshift = 1;
+			}
+		}
+		
+		p[0] |= 1;
+		head += width;
+	}
+
+	trinkle(head, width, cmp, p, pshift, 0, lp);
+
+	while(pshift != 1 || p[0] != 1 || p[1] != 0) {
+		if(pshift <= 1) {
+			trail = pntz(p);
+			shr(p, trail);
+			pshift += trail;
+		} else {
+			shl(p, 2);
+			pshift -= 2;
+			p[0] ^= 7;
+			shr(p, 1);
+			trinkle(head - lp[pshift] - width, width, cmp, p, pshift + 1, 1, lp);
+			shl(p, 1);
+			p[0] |= 1;
+			trinkle(head - width, width, cmp, p, pshift, 1, lp);
+		}
+		head -= width;
+	}
+}
-- 
Gitee


From 658c3926066e7108392b50b4743d06fee5775aa6 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Sat, 3 Sep 2022 17:55:44 +0800
Subject: [PATCH 04/13] feat: update stdlib

---
 src/include/stdlib.h       |  27 +++++++
 src/internal/string_impl.h |   9 +++
 src/stdlib/abs.c           |   6 ++
 src/stdlib/atof.c          |   6 ++
 src/stdlib/atoi.c          |  16 ++++
 src/stdlib/atol.c          |  17 ++++
 src/stdlib/atoll.c         |  17 ++++
 src/stdlib/div.c           |   6 ++
 src/stdlib/labs.c          |   6 ++
 src/stdlib/ldiv.c          |   6 ++
 src/stdlib/llabs.c         |   6 ++
 src/stdlib/lldiv.c         |   6 ++
 src/stdlib/strtod.c        |  37 +++++++++
 src/stdlib/strtol.c        |  56 ++++++++++++++
 src/string/stpncpy.c       |  32 ++++++++
 src/string/strchrnul.c     |  28 +++++++
 src/string/strcspn.c       |  18 +++++
 src/string/strncat.c       |  10 +++
 src/string/strncmp.c       |   9 +++
 src/string/strncpy.c       |   7 ++
 src/string/strpbrk.c       |   7 ++
 src/string/strspn.c        |  20 +++++
 src/string/strstr.c        | 154 +++++++++++++++++++++++++++++++++++++
 src/string/strtok.c        |  13 ++++
 24 files changed, 519 insertions(+)
 create mode 100644 src/internal/string_impl.h
 create mode 100644 src/stdlib/abs.c
 create mode 100644 src/stdlib/atof.c
 create mode 100644 src/stdlib/atoi.c
 create mode 100644 src/stdlib/atol.c
 create mode 100644 src/stdlib/atoll.c
 create mode 100644 src/stdlib/div.c
 create mode 100644 src/stdlib/labs.c
 create mode 100644 src/stdlib/ldiv.c
 create mode 100644 src/stdlib/llabs.c
 create mode 100644 src/stdlib/lldiv.c
 create mode 100644 src/stdlib/strtod.c
 create mode 100644 src/stdlib/strtol.c
 create mode 100644 src/string/stpncpy.c
 create mode 100644 src/string/strchrnul.c
 create mode 100644 src/string/strcspn.c
 create mode 100644 src/string/strncat.c
 create mode 100644 src/string/strncmp.c
 create mode 100644 src/string/strncpy.c
 create mode 100644 src/string/strpbrk.c
 create mode 100644 src/string/strspn.c
 create mode 100644 src/string/strstr.c
 create mode 100644 src/string/strtok.c

diff --git a/src/include/stdlib.h b/src/include/stdlib.h
index c6ca66e..eb938f1 100644
--- a/src/include/stdlib.h
+++ b/src/include/stdlib.h
@@ -45,6 +45,33 @@ int system (const char *);
 void *bsearch (const void *, const void *, size_t, size_t, int (*)(const void *, const void *));
 void qsort (void *, size_t, size_t, int (*)(const void *, const void *));
 
+/* Conversions to numeric formats */
+int atoi (const char *);
+long atol (const char *);
+long long atoll (const char *);
+double atof (const char *);
+
+float strtof (const char *__restrict, char **__restrict);
+double strtod (const char *__restrict, char **__restrict);
+long double strtold (const char *__restrict, char **__restrict);
+
+long strtol (const char *__restrict, char **__restrict, int);
+unsigned long strtoul (const char *__restrict, char **__restrict, int);
+long long strtoll (const char *__restrict, char **__restrict, int);
+unsigned long long strtoull (const char *__restrict, char **__restrict, int);
+
+int abs (int);
+long labs (long);
+long long llabs (long long);
+
+typedef struct { int quot, rem; } div_t;
+typedef struct { long quot, rem; } ldiv_t;
+typedef struct { long long quot, rem; } lldiv_t;
+
+div_t div (int, int);
+ldiv_t ldiv (long, long);
+lldiv_t lldiv (long long, long long);
+
 int mblen (const char *, size_t);
 int mbtowc (wchar_t *__restrict, const char *__restrict, size_t);
 int wctomb (char *, wchar_t);
diff --git a/src/internal/string_impl.h b/src/internal/string_impl.h
new file mode 100644
index 0000000..bd9021f
--- /dev/null
+++ b/src/internal/string_impl.h
@@ -0,0 +1,9 @@
+#ifndef STRING_H
+#define STRING_H
+
+hidden void *__memrchr(const void *, int, size_t);
+hidden char *__stpcpy(char *, const char *);
+hidden char *__stpncpy(char *, const char *, size_t);
+hidden char *__strchrnul(const char *, int);
+
+#endif
diff --git a/src/stdlib/abs.c b/src/stdlib/abs.c
new file mode 100644
index 0000000..e721fdc
--- /dev/null
+++ b/src/stdlib/abs.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+int abs(int a)
+{
+	return a>0 ? a : -a;
+}
diff --git a/src/stdlib/atof.c b/src/stdlib/atof.c
new file mode 100644
index 0000000..f7fcd82
--- /dev/null
+++ b/src/stdlib/atof.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+double atof(const char *s)
+{
+	return strtod(s, 0);
+}
diff --git a/src/stdlib/atoi.c b/src/stdlib/atoi.c
new file mode 100644
index 0000000..9baca7b
--- /dev/null
+++ b/src/stdlib/atoi.c
@@ -0,0 +1,16 @@
+#include <stdlib.h>
+#include <ctype.h>
+
+int atoi(const char *s)
+{
+	int n=0, neg=0;
+	while (isspace(*s)) s++;
+	switch (*s) {
+	case '-': neg=1;
+	case '+': s++;
+	}
+	/* Compute n as a negative number to avoid overflow on INT_MIN */
+	while (isdigit(*s))
+		n = 10*n - (*s++ - '0');
+	return neg ? n : -n;
+}
diff --git a/src/stdlib/atol.c b/src/stdlib/atol.c
new file mode 100644
index 0000000..140ea3e
--- /dev/null
+++ b/src/stdlib/atol.c
@@ -0,0 +1,17 @@
+#include <stdlib.h>
+#include <ctype.h>
+
+long atol(const char *s)
+{
+	long n=0;
+	int neg=0;
+	while (isspace(*s)) s++;
+	switch (*s) {
+	case '-': neg=1;
+	case '+': s++;
+	}
+	/* Compute n as a negative number to avoid overflow on LONG_MIN */
+	while (isdigit(*s))
+		n = 10*n - (*s++ - '0');
+	return neg ? n : -n;
+}
diff --git a/src/stdlib/atoll.c b/src/stdlib/atoll.c
new file mode 100644
index 0000000..b693048
--- /dev/null
+++ b/src/stdlib/atoll.c
@@ -0,0 +1,17 @@
+#include <stdlib.h>
+#include <ctype.h>
+
+long long atoll(const char *s)
+{
+	long long n=0;
+	int neg=0;
+	while (isspace(*s)) s++;
+	switch (*s) {
+	case '-': neg=1;
+	case '+': s++;
+	}
+	/* Compute n as a negative number to avoid overflow on LLONG_MIN */
+	while (isdigit(*s))
+		n = 10*n - (*s++ - '0');
+	return neg ? n : -n;
+}
diff --git a/src/stdlib/div.c b/src/stdlib/div.c
new file mode 100644
index 0000000..e42c1f1
--- /dev/null
+++ b/src/stdlib/div.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+div_t div(int num, int den)
+{
+	return (div_t){ num/den, num%den };
+}
diff --git a/src/stdlib/labs.c b/src/stdlib/labs.c
new file mode 100644
index 0000000..83ddb14
--- /dev/null
+++ b/src/stdlib/labs.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+long labs(long a)
+{
+	return a>0 ? a : -a;
+}
diff --git a/src/stdlib/ldiv.c b/src/stdlib/ldiv.c
new file mode 100644
index 0000000..36eb960
--- /dev/null
+++ b/src/stdlib/ldiv.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+ldiv_t ldiv(long num, long den)
+{
+	return (ldiv_t){ num/den, num%den };
+}
diff --git a/src/stdlib/llabs.c b/src/stdlib/llabs.c
new file mode 100644
index 0000000..9dfaf5c
--- /dev/null
+++ b/src/stdlib/llabs.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+long long llabs(long long a)
+{
+	return a>0 ? a : -a;
+}
diff --git a/src/stdlib/lldiv.c b/src/stdlib/lldiv.c
new file mode 100644
index 0000000..7aaf7a0
--- /dev/null
+++ b/src/stdlib/lldiv.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+lldiv_t lldiv(long long num, long long den)
+{
+	return (lldiv_t){ num/den, num%den };
+}
diff --git a/src/stdlib/strtod.c b/src/stdlib/strtod.c
new file mode 100644
index 0000000..a5d0118
--- /dev/null
+++ b/src/stdlib/strtod.c
@@ -0,0 +1,37 @@
+#include <stdlib.h>
+#include "shgetc.h"
+#include "floatscan.h"
+#include "stdio_impl.h"
+
+static long double strtox(const char *s, char **p, int prec)
+{
+	FILE f;
+	sh_fromstring(&f, s);
+	shlim(&f, 0);
+	long double y = __floatscan(&f, prec, 1);
+	off_t cnt = shcnt(&f);
+	if (p) *p = cnt ? (char *)s + cnt : (char *)s;
+	return y;
+}
+
+float strtof(const char *restrict s, char **restrict p)
+{
+	return strtox(s, p, 0);
+}
+
+double strtod(const char *restrict s, char **restrict p)
+{
+	return strtox(s, p, 1);
+}
+
+long double strtold(const char *restrict s, char **restrict p)
+{
+	return strtox(s, p, 2);
+}
+
+weak_alias(strtof, strtof_l);
+weak_alias(strtod, strtod_l);
+weak_alias(strtold, strtold_l);
+weak_alias(strtof, __strtof_l);
+weak_alias(strtod, __strtod_l);
+weak_alias(strtold, __strtold_l);
diff --git a/src/stdlib/strtol.c b/src/stdlib/strtol.c
new file mode 100644
index 0000000..bfefea6
--- /dev/null
+++ b/src/stdlib/strtol.c
@@ -0,0 +1,56 @@
+#include "stdio_impl.h"
+#include "intscan.h"
+#include "shgetc.h"
+#include <inttypes.h>
+#include <limits.h>
+#include <ctype.h>
+
+static unsigned long long strtox(const char *s, char **p, int base, unsigned long long lim)
+{
+	FILE f;
+	sh_fromstring(&f, s);
+	shlim(&f, 0);
+	unsigned long long y = __intscan(&f, base, 1, lim);
+	if (p) {
+		size_t cnt = shcnt(&f);
+		*p = (char *)s + cnt;
+	}
+	return y;
+}
+
+unsigned long long strtoull(const char *restrict s, char **restrict p, int base)
+{
+	return strtox(s, p, base, ULLONG_MAX);
+}
+
+long long strtoll(const char *restrict s, char **restrict p, int base)
+{
+	return strtox(s, p, base, LLONG_MIN);
+}
+
+unsigned long strtoul(const char *restrict s, char **restrict p, int base)
+{
+	return strtox(s, p, base, ULONG_MAX);
+}
+
+long strtol(const char *restrict s, char **restrict p, int base)
+{
+	return strtox(s, p, base, 0UL+LONG_MIN);
+}
+
+intmax_t strtoimax(const char *restrict s, char **restrict p, int base)
+{
+	return strtoll(s, p, base);
+}
+
+uintmax_t strtoumax(const char *restrict s, char **restrict p, int base)
+{
+	return strtoull(s, p, base);
+}
+
+weak_alias(strtol, __strtol_internal);
+weak_alias(strtoul, __strtoul_internal);
+weak_alias(strtoll, __strtoll_internal);
+weak_alias(strtoull, __strtoull_internal);
+weak_alias(strtoimax, __strtoimax_internal);
+weak_alias(strtoumax, __strtoumax_internal);
diff --git a/src/string/stpncpy.c b/src/string/stpncpy.c
new file mode 100644
index 0000000..f57fa6b
--- /dev/null
+++ b/src/string/stpncpy.c
@@ -0,0 +1,32 @@
+#include <string.h>
+#include <stdint.h>
+#include <limits.h>
+
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+char *__stpncpy(char *restrict d, const char *restrict s, size_t n)
+{
+#ifdef __GNUC__
+	typedef size_t __attribute__((__may_alias__)) word;
+	word *wd;
+	const word *ws;
+	if (((uintptr_t)s & ALIGN) == ((uintptr_t)d & ALIGN)) {
+		for (; ((uintptr_t)s & ALIGN) && n && (*d=*s); n--, s++, d++);
+		if (!n || !*s) goto tail;
+		wd=(void *)d; ws=(const void *)s;
+		for (; n>=sizeof(size_t) && !HASZERO(*ws);
+		       n-=sizeof(size_t), ws++, wd++) *wd = *ws;
+		d=(void *)wd; s=(const void *)ws;
+	}
+#endif
+	for (; n && (*d=*s); n--, s++, d++);
+tail:
+	memset(d, 0, n);
+	return d;
+}
+
+weak_alias(__stpncpy, stpncpy);
+
diff --git a/src/string/strchrnul.c b/src/string/strchrnul.c
new file mode 100644
index 0000000..39e2635
--- /dev/null
+++ b/src/string/strchrnul.c
@@ -0,0 +1,28 @@
+#include <string.h>
+#include <stdint.h>
+#include <limits.h>
+
+#define ALIGN (sizeof(size_t))
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+char *__strchrnul(const char *s, int c)
+{
+	c = (unsigned char)c;
+	if (!c) return (char *)s + strlen(s);
+
+#ifdef __GNUC__
+	typedef size_t __attribute__((__may_alias__)) word;
+	const word *w;
+	for (; (uintptr_t)s % ALIGN; s++)
+		if (!*s || *(unsigned char *)s == c) return (char *)s;
+	size_t k = ONES * c;
+	for (w = (void *)s; !HASZERO(*w) && !HASZERO(*w^k); w++);
+	s = (void *)w;
+#endif
+	for (; *s && *(unsigned char *)s != c; s++);
+	return (char *)s;
+}
+
+weak_alias(__strchrnul, strchrnul);
diff --git a/src/string/strcspn.c b/src/string/strcspn.c
new file mode 100644
index 0000000..1e7ba7e
--- /dev/null
+++ b/src/string/strcspn.c
@@ -0,0 +1,18 @@
+#include <string.h>
+#include "string_impl.h"
+
+#define BITOP(a,b,op) \
+ ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
+
+size_t strcspn(const char *s, const char *c)
+{
+	const char *a = s;
+	size_t byteset[32/sizeof(size_t)];
+
+	if (!c[0] || !c[1]) return __strchrnul(s, *c)-a;
+
+	memset(byteset, 0, sizeof byteset);
+	for (; *c && BITOP(byteset, *(unsigned char *)c, |=); c++);
+	for (; *s && !BITOP(byteset, *(unsigned char *)s, &); s++);
+	return s-a;
+}
diff --git a/src/string/strncat.c b/src/string/strncat.c
new file mode 100644
index 0000000..01ca2a2
--- /dev/null
+++ b/src/string/strncat.c
@@ -0,0 +1,10 @@
+#include <string.h>
+
+char *strncat(char *restrict d, const char *restrict s, size_t n)
+{
+	char *a = d;
+	d += strlen(d);
+	while (n && *s) n--, *d++ = *s++;
+	*d++ = 0;
+	return a;
+}
diff --git a/src/string/strncmp.c b/src/string/strncmp.c
new file mode 100644
index 0000000..e228843
--- /dev/null
+++ b/src/string/strncmp.c
@@ -0,0 +1,9 @@
+#include <string.h>
+
+int strncmp(const char *_l, const char *_r, size_t n)
+{
+	const unsigned char *l=(void *)_l, *r=(void *)_r;
+	if (!n--) return 0;
+	for (; *l && *r && n && *l == *r ; l++, r++, n--);
+	return *l - *r;
+}
diff --git a/src/string/strncpy.c b/src/string/strncpy.c
new file mode 100644
index 0000000..545892e
--- /dev/null
+++ b/src/string/strncpy.c
@@ -0,0 +1,7 @@
+#include <string.h>
+
+char *strncpy(char *restrict d, const char *restrict s, size_t n)
+{
+	__stpncpy(d, s, n);
+	return d;
+}
diff --git a/src/string/strpbrk.c b/src/string/strpbrk.c
new file mode 100644
index 0000000..55947c6
--- /dev/null
+++ b/src/string/strpbrk.c
@@ -0,0 +1,7 @@
+#include <string.h>
+
+char *strpbrk(const char *s, const char *b)
+{
+	s += strcspn(s, b);
+	return *s ? (char *)s : 0;
+}
diff --git a/src/string/strspn.c b/src/string/strspn.c
new file mode 100644
index 0000000..9543dad
--- /dev/null
+++ b/src/string/strspn.c
@@ -0,0 +1,20 @@
+#include <string.h>
+
+#define BITOP(a,b,op) \
+ ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
+
+size_t strspn(const char *s, const char *c)
+{
+	const char *a = s;
+	size_t byteset[32/sizeof(size_t)] = { 0 };
+
+	if (!c[0]) return 0;
+	if (!c[1]) {
+		for (; *s == *c; s++);
+		return s-a;
+	}
+
+	for (; *c && BITOP(byteset, *(unsigned char *)c, |=); c++);
+	for (; *s && BITOP(byteset, *(unsigned char *)s, &); s++);
+	return s-a;
+}
diff --git a/src/string/strstr.c b/src/string/strstr.c
new file mode 100644
index 0000000..96657bc
--- /dev/null
+++ b/src/string/strstr.c
@@ -0,0 +1,154 @@
+#include <string.h>
+#include <stdint.h>
+
+static char *twobyte_strstr(const unsigned char *h, const unsigned char *n)
+{
+	uint16_t nw = n[0]<<8 | n[1], hw = h[0]<<8 | h[1];
+	for (h++; *h && hw != nw; hw = hw<<8 | *++h);
+	return *h ? (char *)h-1 : 0;
+}
+
+static char *threebyte_strstr(const unsigned char *h, const unsigned char *n)
+{
+	uint32_t nw = (uint32_t)n[0]<<24 | n[1]<<16 | n[2]<<8;
+	uint32_t hw = (uint32_t)h[0]<<24 | h[1]<<16 | h[2]<<8;
+	for (h+=2; *h && hw != nw; hw = (hw|*++h)<<8);
+	return *h ? (char *)h-2 : 0;
+}
+
+static char *fourbyte_strstr(const unsigned char *h, const unsigned char *n)
+{
+	uint32_t nw = (uint32_t)n[0]<<24 | n[1]<<16 | n[2]<<8 | n[3];
+	uint32_t hw = (uint32_t)h[0]<<24 | h[1]<<16 | h[2]<<8 | h[3];
+	for (h+=3; *h && hw != nw; hw = hw<<8 | *++h);
+	return *h ? (char *)h-3 : 0;
+}
+
+#define MAX(a,b) ((a)>(b)?(a):(b))
+#define MIN(a,b) ((a)<(b)?(a):(b))
+
+#define BITOP(a,b,op) \
+ ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
+
+static char *twoway_strstr(const unsigned char *h, const unsigned char *n)
+{
+	const unsigned char *z;
+	size_t l, ip, jp, k, p, ms, p0, mem, mem0;
+	size_t byteset[32 / sizeof(size_t)] = { 0 };
+	size_t shift[256];
+
+	/* Computing length of needle and fill shift table */
+	for (l=0; n[l] && h[l]; l++)
+		BITOP(byteset, n[l], |=), shift[n[l]] = l+1;
+	if (n[l]) return 0; /* hit the end of h */
+
+	/* Compute maximal suffix */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] > n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	ms = ip;
+	p0 = p;
+
+	/* And with the opposite comparison */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] < n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	if (ip+1 > ms+1) ms = ip;
+	else p = p0;
+
+	/* Periodic needle? */
+	if (memcmp(n, n+p, ms+1)) {
+		mem0 = 0;
+		p = MAX(ms, l-ms-1) + 1;
+	} else mem0 = l-p;
+	mem = 0;
+
+	/* Initialize incremental end-of-haystack pointer */
+	z = h;
+
+	/* Search loop */
+	for (;;) {
+		/* Update incremental end-of-haystack pointer */
+		if (z-h < l) {
+			/* Fast estimate for MAX(l,63) */
+			size_t grow = l | 63;
+			const unsigned char *z2 = memchr(z, 0, grow);
+			if (z2) {
+				z = z2;
+				if (z-h < l) return 0;
+			} else z += grow;
+		}
+
+		/* Check last byte first; advance by shift on mismatch */
+		if (BITOP(byteset, h[l-1], &)) {
+			k = l-shift[h[l-1]];
+			if (k) {
+				if (k < mem) k = mem;
+				h += k;
+				mem = 0;
+				continue;
+			}
+		} else {
+			h += l;
+			mem = 0;
+			continue;
+		}
+
+		/* Compare right half */
+		for (k=MAX(ms+1,mem); n[k] && n[k] == h[k]; k++);
+		if (n[k]) {
+			h += k-ms;
+			mem = 0;
+			continue;
+		}
+		/* Compare left half */
+		for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
+		if (k <= mem) return (char *)h;
+		h += p;
+		mem = mem0;
+	}
+}
+
+char *strstr(const char *h, const char *n)
+{
+	/* Return immediately on empty needle */
+	if (!n[0]) return (char *)h;
+
+	/* Use faster algorithms for short needles */
+	h = strchr(h, *n);
+	if (!h || !n[1]) return (char *)h;
+	if (!h[1]) return 0;
+	if (!n[2]) return twobyte_strstr((void *)h, (void *)n);
+	if (!h[2]) return 0;
+	if (!n[3]) return threebyte_strstr((void *)h, (void *)n);
+	if (!h[3]) return 0;
+	if (!n[4]) return fourbyte_strstr((void *)h, (void *)n);
+
+	return twoway_strstr((void *)h, (void *)n);
+}
diff --git a/src/string/strtok.c b/src/string/strtok.c
new file mode 100644
index 0000000..3508790
--- /dev/null
+++ b/src/string/strtok.c
@@ -0,0 +1,13 @@
+#include <string.h>
+
+char *strtok(char *restrict s, const char *restrict sep)
+{
+	static char *p;
+	if (!s && !(s = p)) return NULL;
+	s += strspn(s, sep);
+	if (!*s) return p = 0;
+	p = s + strcspn(s, sep);
+	if (*p) *p++ = 0;
+	else p = 0;
+	return s;
+}
-- 
Gitee


From c431894d6ef996849ef5816c2a9a3a62c06e49cd Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Sat, 3 Sep 2022 21:19:24 +0800
Subject: [PATCH 05/13] feat: update math

---
 src/complex/cpow.c    |   4 +
 src/complex/cpowf.c   |   4 +
 src/complex/cpowl.c   |   8 ++
 src/include/tgmath.h  | 270 ++++++++++++++++++++++++++++++++++++++++++
 src/math/__invtrigl.c |  63 ++++++++++
 src/math/__invtrigl.h |   8 ++
 src/math/__polevll.c  |  93 +++++++++++++++
 src/math/atan2l.c     |  85 +++++++++++++
 src/math/atanl.c      | 184 ++++++++++++++++++++++++++++
 src/math/logl.c       | 175 +++++++++++++++++++++++++++
 10 files changed, 894 insertions(+)
 create mode 100644 src/include/tgmath.h
 create mode 100644 src/math/__invtrigl.c
 create mode 100644 src/math/__invtrigl.h
 create mode 100644 src/math/__polevll.c
 create mode 100644 src/math/atan2l.c
 create mode 100644 src/math/atanl.c
 create mode 100644 src/math/logl.c

diff --git a/src/complex/cpow.c b/src/complex/cpow.c
index 1137d39..328058e 100644
--- a/src/complex/cpow.c
+++ b/src/complex/cpow.c
@@ -4,5 +4,9 @@
 
 double complex cpow(double complex z, double complex c)
 {
+#if 0
 	return cexp(c * clog(z));
+#else
+    return 0;
+#endif
 }
diff --git a/src/complex/cpowf.c b/src/complex/cpowf.c
index f3fd4b7..7b3e7e8 100644
--- a/src/complex/cpowf.c
+++ b/src/complex/cpowf.c
@@ -2,5 +2,9 @@
 
 float complex cpowf(float complex z, float complex c)
 {
+#if 0
 	return cexpf(c * clogf(z));
+#else
+    return 0;
+#endif
 }
diff --git a/src/complex/cpowl.c b/src/complex/cpowl.c
index be36f04..4ef8efc 100644
--- a/src/complex/cpowl.c
+++ b/src/complex/cpowl.c
@@ -3,11 +3,19 @@
 #if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
 long double complex cpowl(long double complex z, long double complex c)
 {
+#if 0
 	return cpow(z, c);
+#else
+    return 0;
+#endif
 }
 #else
 long double complex cpowl(long double complex z, long double complex c)
 {
+#if 0
 	return cexpl(c * clogl(z));
+#else
+    return 0;
+#endif
 }
 #endif
diff --git a/src/include/tgmath.h b/src/include/tgmath.h
new file mode 100644
index 0000000..e41ccac
--- /dev/null
+++ b/src/include/tgmath.h
@@ -0,0 +1,270 @@
+#ifndef _TGMATH_H
+#define _TGMATH_H
+
+/*
+the return types are only correct with gcc (__GNUC__)
+otherwise they are long double or long double complex
+
+the long double version of a function is never chosen when
+sizeof(double) == sizeof(long double)
+(but the return type is set correctly with gcc)
+*/
+
+#include <math.h>
+#include <complex.h>
+
+#define __IS_FP(x) (sizeof((x)+1ULL) == sizeof((x)+1.0f))
+#define __IS_CX(x) (__IS_FP(x) && sizeof(x) == sizeof((x)+I))
+#define __IS_REAL(x) (__IS_FP(x) && 2*sizeof(x) == sizeof((x)+I))
+
+#define __FLT(x) (__IS_REAL(x) && sizeof(x) == sizeof(float))
+#define __LDBL(x) (__IS_REAL(x) && sizeof(x) == sizeof(long double) && sizeof(long double) != sizeof(double))
+
+#define __FLTCX(x) (__IS_CX(x) && sizeof(x) == sizeof(float complex))
+#define __DBLCX(x) (__IS_CX(x) && sizeof(x) == sizeof(double complex))
+#define __LDBLCX(x) (__IS_CX(x) && sizeof(x) == sizeof(long double complex) && sizeof(long double) != sizeof(double))
+
+/* return type */
+
+#ifdef __GNUC__
+/*
+the result must be casted to the right type
+(otherwise the result type is determined by the conversion
+rules applied to all the function return types so it is long
+double or long double complex except for integral functions)
+
+this cannot be done in c99, so the typeof gcc extension is
+used and that the type of ?: depends on wether an operand is
+a null pointer constant or not
+(in c11 _Generic can be used)
+
+the c arguments below must be integer constant expressions
+so they can be in null pointer constants
+(__IS_FP above was carefully chosen this way)
+*/
+/* if c then t else void */
+#define __type1(c,t) __typeof__(*(0?(t*)0:(void*)!(c)))
+/* if c then t1 else t2 */
+#define __type2(c,t1,t2) __typeof__(*(0?(__type1(c,t1)*)0:(__type1(!(c),t2)*)0))
+/* cast to double when x is integral, otherwise use typeof(x) */
+#define __RETCAST(x) ( \
+	__type2(__IS_FP(x), __typeof__(x), double))
+/* 2 args case, should work for complex types (cpow) */
+#define __RETCAST_2(x, y) ( \
+	__type2(__IS_FP(x) && __IS_FP(y), \
+		__typeof__((x)+(y)), \
+		__typeof__((x)+(y)+1.0)))
+/* 3 args case (fma only) */
+#define __RETCAST_3(x, y, z) ( \
+	__type2(__IS_FP(x) && __IS_FP(y) && __IS_FP(z), \
+		__typeof__((x)+(y)+(z)), \
+		__typeof__((x)+(y)+(z)+1.0)))
+/* drop complex from the type of x */
+/* TODO: wrong when sizeof(long double)==sizeof(double) */
+#define __RETCAST_REAL(x) (  \
+	__type2(__IS_FP(x) && sizeof((x)+I) == sizeof(float complex), float, \
+	__type2(sizeof((x)+1.0+I) == sizeof(double complex), double, \
+		long double)))
+/* add complex to the type of x */
+#define __RETCAST_CX(x) (__typeof__(__RETCAST(x)0+I))
+#else
+#define __RETCAST(x)
+#define __RETCAST_2(x, y)
+#define __RETCAST_3(x, y, z)
+#define __RETCAST_REAL(x)
+#define __RETCAST_CX(x)
+#endif
+
+/* function selection */
+
+#define __tg_real_nocast(fun, x) ( \
+	__FLT(x) ? fun ## f (x) : \
+	__LDBL(x) ? fun ## l (x) : \
+	fun(x) )
+
+#define __tg_real(fun, x) (__RETCAST(x)__tg_real_nocast(fun, x))
+
+#define __tg_real_2_1(fun, x, y) (__RETCAST(x)( \
+	__FLT(x) ? fun ## f (x, y) : \
+	__LDBL(x) ? fun ## l (x, y) : \
+	fun(x, y) ))
+
+#define __tg_real_2(fun, x, y) (__RETCAST_2(x, y)( \
+	__FLT(x) && __FLT(y) ? fun ## f (x, y) : \
+	__LDBL((x)+(y)) ? fun ## l (x, y) : \
+	fun(x, y) ))
+
+#define __tg_complex(fun, x) (__RETCAST_CX(x)( \
+	__FLTCX((x)+I) && __IS_FP(x) ? fun ## f (x) : \
+	__LDBLCX((x)+I) ? fun ## l (x) : \
+	fun(x) ))
+
+#define __tg_complex_retreal(fun, x) (__RETCAST_REAL(x)( \
+	__FLTCX((x)+I) && __IS_FP(x) ? fun ## f (x) : \
+	__LDBLCX((x)+I) ? fun ## l (x) : \
+	fun(x) ))
+
+#define __tg_real_complex(fun, x) (__RETCAST(x)( \
+	__FLTCX(x) ? c ## fun ## f (x) : \
+	__DBLCX(x) ? c ## fun (x) : \
+	__LDBLCX(x) ? c ## fun ## l (x) : \
+	__FLT(x) ? fun ## f (x) : \
+	__LDBL(x) ? fun ## l (x) : \
+	fun(x) ))
+
+/* special cases */
+
+#define __tg_real_remquo(x, y, z) (__RETCAST_2(x, y)( \
+	__FLT(x) && __FLT(y) ? remquof(x, y, z) : \
+	__LDBL((x)+(y)) ? remquol(x, y, z) : \
+	remquo(x, y, z) ))
+
+#define __tg_real_fma(x, y, z) (__RETCAST_3(x, y, z)( \
+	__FLT(x) && __FLT(y) && __FLT(z) ? fmaf(x, y, z) : \
+	__LDBL((x)+(y)+(z)) ? fmal(x, y, z) : \
+	fma(x, y, z) ))
+
+#define __tg_real_complex_pow(x, y) (__RETCAST_2(x, y)( \
+	__FLTCX((x)+(y)) && __IS_FP(x) && __IS_FP(y) ? cpowf(x, y) : \
+	__FLTCX((x)+(y)) ? cpow(x, y) : \
+	__DBLCX((x)+(y)) ? cpow(x, y) : \
+	__LDBLCX((x)+(y)) ? cpowl(x, y) : \
+	__FLT(x) && __FLT(y) ? powf(x, y) : \
+	__LDBL((x)+(y)) ? powl(x, y) : \
+	pow(x, y) ))
+
+#define __tg_real_complex_fabs(x) (__RETCAST_REAL(x)( \
+	__FLTCX(x) ? cabsf(x) : \
+	__DBLCX(x) ? cabs(x) : \
+	__LDBLCX(x) ? cabsl(x) : \
+	__FLT(x) ? fabsf(x) : \
+	__LDBL(x) ? fabsl(x) : \
+	fabs(x) ))
+
+/* suppress any macros in math.h or complex.h */
+
+#undef acos
+#undef acosh
+#undef asin
+#undef asinh
+#undef atan
+#undef atan2
+#undef atanh
+#undef carg
+#undef cbrt
+#undef ceil
+#undef cimag
+#undef conj
+#undef copysign
+#undef cos
+#undef cosh
+#undef cproj
+#undef creal
+#undef erf
+#undef erfc
+#undef exp
+#undef exp2
+#undef expm1
+#undef fabs
+#undef fdim
+#undef floor
+#undef fma
+#undef fmax
+#undef fmin
+#undef fmod
+#undef frexp
+#undef hypot
+#undef ilogb
+#undef ldexp
+#undef lgamma
+#undef llrint
+#undef llround
+#undef log
+#undef log10
+#undef log1p
+#undef log2
+#undef logb
+#undef lrint
+#undef lround
+#undef nearbyint
+#undef nextafter
+#undef nexttoward
+#undef pow
+#undef remainder
+#undef remquo
+#undef rint
+#undef round
+#undef scalbln
+#undef scalbn
+#undef sin
+#undef sinh
+#undef sqrt
+#undef tan
+#undef tanh
+#undef tgamma
+#undef trunc
+
+/* tg functions */
+
+#define acos(x)         __tg_real_complex(acos, (x))
+#define acosh(x)        __tg_real_complex(acosh, (x))
+#define asin(x)         __tg_real_complex(asin, (x))
+#define asinh(x)        __tg_real_complex(asinh, (x))
+#define atan(x)         __tg_real_complex(atan, (x))
+#define atan2(x,y)      __tg_real_2(atan2, (x), (y))
+#define atanh(x)        __tg_real_complex(atanh, (x))
+#define carg(x)         __tg_complex_retreal(carg, (x))
+#define cbrt(x)         __tg_real(cbrt, (x))
+#define ceil(x)         __tg_real(ceil, (x))
+#define cimag(x)        __tg_complex_retreal(cimag, (x))
+#define conj(x)         __tg_complex(conj, (x))
+#define copysign(x,y)   __tg_real_2(copysign, (x), (y))
+#define cos(x)          __tg_real_complex(cos, (x))
+#define cosh(x)         __tg_real_complex(cosh, (x))
+#define cproj(x)        __tg_complex(cproj, (x))
+#define creal(x)        __tg_complex_retreal(creal, (x))
+#define erf(x)          __tg_real(erf, (x))
+#define erfc(x)         __tg_real(erfc, (x))
+#define exp(x)          __tg_real_complex(exp, (x))
+#define exp2(x)         __tg_real(exp2, (x))
+#define expm1(x)        __tg_real(expm1, (x))
+#define fabs(x)         __tg_real_complex_fabs(x)
+#define fdim(x,y)       __tg_real_2(fdim, (x), (y))
+#define floor(x)        __tg_real(floor, (x))
+#define fma(x,y,z)      __tg_real_fma((x), (y), (z))
+#define fmax(x,y)       __tg_real_2(fmax, (x), (y))
+#define fmin(x,y)       __tg_real_2(fmin, (x), (y))
+#define fmod(x,y)       __tg_real_2(fmod, (x), (y))
+#define frexp(x,y)      __tg_real_2_1(frexp, (x), (y))
+#define hypot(x,y)      __tg_real_2(hypot, (x), (y))
+#define ilogb(x)        __tg_real_nocast(ilogb, (x))
+#define ldexp(x,y)      __tg_real_2_1(ldexp, (x), (y))
+#define lgamma(x)       __tg_real(lgamma, (x))
+#define llrint(x)       __tg_real_nocast(llrint, (x))
+#define llround(x)      __tg_real_nocast(llround, (x))
+#define log(x)          __tg_real_complex(log, (x))
+#define log10(x)        __tg_real(log10, (x))
+#define log1p(x)        __tg_real(log1p, (x))
+#define log2(x)         __tg_real(log2, (x))
+#define logb(x)         __tg_real(logb, (x))
+#define lrint(x)        __tg_real_nocast(lrint, (x))
+#define lround(x)       __tg_real_nocast(lround, (x))
+#define nearbyint(x)    __tg_real(nearbyint, (x))
+#define nextafter(x,y)  __tg_real_2(nextafter, (x), (y))
+#define nexttoward(x,y) __tg_real_2(nexttoward, (x), (y))
+#define pow(x,y)        __tg_real_complex_pow((x), (y))
+#define remainder(x,y)  __tg_real_2(remainder, (x), (y))
+#define remquo(x,y,z)   __tg_real_remquo((x), (y), (z))
+#define rint(x)         __tg_real(rint, (x))
+#define round(x)        __tg_real(round, (x))
+#define scalbln(x,y)    __tg_real_2_1(scalbln, (x), (y))
+#define scalbn(x,y)     __tg_real_2_1(scalbn, (x), (y))
+#define sin(x)          __tg_real_complex(sin, (x))
+#define sinh(x)         __tg_real_complex(sinh, (x))
+#define sqrt(x)         __tg_real_complex(sqrt, (x))
+#define tan(x)          __tg_real_complex(tan, (x))
+#define tanh(x)         __tg_real_complex(tanh, (x))
+#define tgamma(x)       __tg_real(tgamma, (x))
+#define trunc(x)        __tg_real(trunc, (x))
+
+#endif
diff --git a/src/math/__invtrigl.c b/src/math/__invtrigl.c
new file mode 100644
index 0000000..48f83aa
--- /dev/null
+++ b/src/math/__invtrigl.c
@@ -0,0 +1,63 @@
+#include <float.h>
+#include "__invtrigl.h"
+
+#if LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+static const long double
+pS0 =  1.66666666666666666631e-01L,
+pS1 = -4.16313987993683104320e-01L,
+pS2 =  3.69068046323246813704e-01L,
+pS3 = -1.36213932016738603108e-01L,
+pS4 =  1.78324189708471965733e-02L,
+pS5 = -2.19216428382605211588e-04L,
+pS6 = -7.10526623669075243183e-06L,
+qS1 = -2.94788392796209867269e+00L,
+qS2 =  3.27309890266528636716e+00L,
+qS3 = -1.68285799854822427013e+00L,
+qS4 =  3.90699412641738801874e-01L,
+qS5 = -3.14365703596053263322e-02L;
+
+const long double pio2_hi = 1.57079632679489661926L;
+const long double pio2_lo = -2.50827880633416601173e-20L;
+
+/* used in asinl() and acosl() */
+/* R(x^2) is a rational approximation of (asin(x)-x)/x^3 with Remez algorithm */
+long double __invtrigl_R(long double z)
+{
+	long double p, q;
+	p = z*(pS0+z*(pS1+z*(pS2+z*(pS3+z*(pS4+z*(pS5+z*pS6))))));
+	q = 1.0+z*(qS1+z*(qS2+z*(qS3+z*(qS4+z*qS5))));
+	return p/q;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+static const long double
+pS0 =  1.66666666666666666666666666666700314e-01L,
+pS1 = -7.32816946414566252574527475428622708e-01L,
+pS2 =  1.34215708714992334609030036562143589e+00L,
+pS3 = -1.32483151677116409805070261790752040e+00L,
+pS4 =  7.61206183613632558824485341162121989e-01L,
+pS5 = -2.56165783329023486777386833928147375e-01L,
+pS6 =  4.80718586374448793411019434585413855e-02L,
+pS7 = -4.42523267167024279410230886239774718e-03L,
+pS8 =  1.44551535183911458253205638280410064e-04L,
+pS9 = -2.10558957916600254061591040482706179e-07L,
+qS1 = -4.84690167848739751544716485245697428e+00L,
+qS2 =  9.96619113536172610135016921140206980e+00L,
+qS3 = -1.13177895428973036660836798461641458e+01L,
+qS4 =  7.74004374389488266169304117714658761e+00L,
+qS5 = -3.25871986053534084709023539900339905e+00L,
+qS6 =  8.27830318881232209752469022352928864e-01L,
+qS7 = -1.18768052702942805423330715206348004e-01L,
+qS8 =  8.32600764660522313269101537926539470e-03L,
+qS9 = -1.99407384882605586705979504567947007e-04L;
+
+const long double pio2_hi = 1.57079632679489661923132169163975140L;
+const long double pio2_lo = 4.33590506506189051239852201302167613e-35L;
+
+long double __invtrigl_R(long double z)
+{
+	long double p, q;
+	p = z*(pS0+z*(pS1+z*(pS2+z*(pS3+z*(pS4+z*(pS5+z*(pS6+z*(pS7+z*(pS8+z*pS9)))))))));
+	q = 1.0+z*(qS1+z*(qS2+z*(qS3+z*(qS4+z*(qS5+z*(qS6+z*(qS7+z*(qS8+z*qS9))))))));
+	return p/q;
+}
+#endif
diff --git a/src/math/__invtrigl.h b/src/math/__invtrigl.h
new file mode 100644
index 0000000..bee7931
--- /dev/null
+++ b/src/math/__invtrigl.h
@@ -0,0 +1,8 @@
+#include <features.h>
+
+/* shared by acosl, asinl and atan2l */
+#define pio2_hi __pio2_hi
+#define pio2_lo __pio2_lo
+hidden extern const long double pio2_hi, pio2_lo;
+
+hidden long double __invtrigl_R(long double z);
diff --git a/src/math/__polevll.c b/src/math/__polevll.c
new file mode 100644
index 0000000..ce1a840
--- /dev/null
+++ b/src/math/__polevll.c
@@ -0,0 +1,93 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/polevll.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Evaluate polynomial
+ *
+ *
+ * SYNOPSIS:
+ *
+ * int N;
+ * long double x, y, coef[N+1], polevl[];
+ *
+ * y = polevll( x, coef, N );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Evaluates polynomial of degree N:
+ *
+ *                     2          N
+ * y  =  C  + C x + C x  +...+ C x
+ *        0    1     2          N
+ *
+ * Coefficients are stored in reverse order:
+ *
+ * coef[0] = C  , ..., coef[N] = C  .
+ *            N                   0
+ *
+ *  The function p1evll() assumes that coef[N] = 1.0 and is
+ * omitted from the array.  Its calling arguments are
+ * otherwise the same as polevll().
+ *
+ *
+ * SPEED:
+ *
+ * In the interest of speed, there are no checks for out
+ * of bounds arithmetic.  This routine is used by most of
+ * the functions in the library.  Depending on available
+ * equipment features, the user may wish to rewrite the
+ * program in microcode or assembly language.
+ *
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+#else
+/*
+ * Polynomial evaluator:
+ *  P[0] x^n  +  P[1] x^(n-1)  +  ...  +  P[n]
+ */
+long double __polevll(long double x, const long double *P, int n)
+{
+	long double y;
+
+	y = *P++;
+	do {
+		y = y * x + *P++;
+	} while (--n);
+
+	return y;
+}
+
+/*
+ * Polynomial evaluator:
+ *  x^n  +  P[0] x^(n-1)  +  P[1] x^(n-2)  +  ...  +  P[n]
+ */
+long double __p1evll(long double x, const long double *P, int n)
+{
+	long double y;
+
+	n -= 1;
+	y = x + *P++;
+	do {
+		y = y * x + *P++;
+	} while (--n);
+
+	return y;
+}
+#endif
diff --git a/src/math/atan2l.c b/src/math/atan2l.c
new file mode 100644
index 0000000..f0937a9
--- /dev/null
+++ b/src/math/atan2l.c
@@ -0,0 +1,85 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_atan2l.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ *
+ */
+/*
+ * See comments in atan2.c.
+ * Converted to long double by David Schultz <das@FreeBSD.ORG>.
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double atan2l(long double y, long double x)
+{
+	return atan2(y, x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#include "__invtrigl.h"
+
+long double atan2l(long double y, long double x)
+{
+	union ldshape ux, uy;
+	long double z;
+	int m, ex, ey;
+
+	if (isnan(x) || isnan(y))
+		return x+y;
+	if (x == 1)
+		return atanl(y);
+	ux.f = x;
+	uy.f = y;
+	ex = ux.i.se & 0x7fff;
+	ey = uy.i.se & 0x7fff;
+	m = 2*(ux.i.se>>15) | uy.i.se>>15;
+	if (y == 0) {
+		switch(m) {
+		case 0:
+		case 1: return y;           /* atan(+-0,+anything)=+-0 */
+		case 2: return  2*pio2_hi;  /* atan(+0,-anything) = pi */
+		case 3: return -2*pio2_hi;  /* atan(-0,-anything) =-pi */
+		}
+	}
+	if (x == 0)
+		return m&1 ? -pio2_hi : pio2_hi;
+	if (ex == 0x7fff) {
+		if (ey == 0x7fff) {
+			switch(m) {
+			case 0: return  pio2_hi/2;   /* atan(+INF,+INF) */
+			case 1: return -pio2_hi/2;   /* atan(-INF,+INF) */
+			case 2: return  1.5*pio2_hi; /* atan(+INF,-INF) */
+			case 3: return -1.5*pio2_hi; /* atan(-INF,-INF) */
+			}
+		} else {
+			switch(m) {
+			case 0: return  0.0;        /* atan(+...,+INF) */
+			case 1: return -0.0;        /* atan(-...,+INF) */
+			case 2: return  2*pio2_hi;  /* atan(+...,-INF) */
+			case 3: return -2*pio2_hi;  /* atan(-...,-INF) */
+			}
+		}
+	}
+	if (ex+120 < ey || ey == 0x7fff)
+		return m&1 ? -pio2_hi : pio2_hi;
+	/* z = atan(|y/x|) without spurious underflow */
+	if ((m&2) && ey+120 < ex)  /* |y/x| < 0x1p-120, x<0 */
+		z = 0.0;
+	else
+		z = atanl(fabsl(y/x));
+	switch (m) {
+	case 0: return z;               /* atan(+,+) */
+	case 1: return -z;              /* atan(-,+) */
+	case 2: return 2*pio2_hi-(z-2*pio2_lo); /* atan(+,-) */
+	default: /* case 3 */
+		return (z-2*pio2_lo)-2*pio2_hi; /* atan(-,-) */
+	}
+}
+#endif
diff --git a/src/math/atanl.c b/src/math/atanl.c
new file mode 100644
index 0000000..c3b0c92
--- /dev/null
+++ b/src/math/atanl.c
@@ -0,0 +1,184 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_atanl.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+/*
+ * See comments in atan.c.
+ * Converted to long double by David Schultz <das@FreeBSD.ORG>.
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double atanl(long double x)
+{
+	return atan(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+
+#if LDBL_MANT_DIG == 64
+#define EXPMAN(u) ((u.i.se & 0x7fff)<<8 | (u.i.m>>55 & 0xff))
+
+static const long double atanhi[] = {
+	 4.63647609000806116202e-01L,
+	 7.85398163397448309628e-01L,
+	 9.82793723247329067960e-01L,
+	 1.57079632679489661926e+00L,
+};
+
+static const long double atanlo[] = {
+	 1.18469937025062860669e-20L,
+	-1.25413940316708300586e-20L,
+	 2.55232234165405176172e-20L,
+	-2.50827880633416601173e-20L,
+};
+
+static const long double aT[] = {
+	 3.33333333333333333017e-01L,
+	-1.99999999999999632011e-01L,
+	 1.42857142857046531280e-01L,
+	-1.11111111100562372733e-01L,
+	 9.09090902935647302252e-02L,
+	-7.69230552476207730353e-02L,
+	 6.66661718042406260546e-02L,
+	-5.88158892835030888692e-02L,
+	 5.25499891539726639379e-02L,
+	-4.70119845393155721494e-02L,
+	 4.03539201366454414072e-02L,
+	-2.91303858419364158725e-02L,
+	 1.24822046299269234080e-02L,
+};
+
+static long double T_even(long double x)
+{
+	return aT[0] + x * (aT[2] + x * (aT[4] + x * (aT[6] +
+		x * (aT[8] + x * (aT[10] + x * aT[12])))));
+}
+
+static long double T_odd(long double x)
+{
+	return aT[1] + x * (aT[3] + x * (aT[5] + x * (aT[7] +
+		x * (aT[9] + x * aT[11]))));
+}
+#elif LDBL_MANT_DIG == 113
+#define EXPMAN(u) ((u.i.se & 0x7fff)<<8 | u.i.top>>8)
+
+static const long double atanhi[] = {
+	 4.63647609000806116214256231461214397e-01L,
+	 7.85398163397448309615660845819875699e-01L,
+	 9.82793723247329067985710611014666038e-01L,
+	 1.57079632679489661923132169163975140e+00L,
+};
+
+static const long double atanlo[] = {
+	 4.89509642257333492668618435220297706e-36L,
+	 2.16795253253094525619926100651083806e-35L,
+	-2.31288434538183565909319952098066272e-35L,
+	 4.33590506506189051239852201302167613e-35L,
+};
+
+static const long double aT[] = {
+	 3.33333333333333333333333333333333125e-01L,
+	-1.99999999999999999999999999999180430e-01L,
+	 1.42857142857142857142857142125269827e-01L,
+	-1.11111111111111111111110834490810169e-01L,
+	 9.09090909090909090908522355708623681e-02L,
+	-7.69230769230769230696553844935357021e-02L,
+	 6.66666666666666660390096773046256096e-02L,
+	-5.88235294117646671706582985209643694e-02L,
+	 5.26315789473666478515847092020327506e-02L,
+	-4.76190476189855517021024424991436144e-02L,
+	 4.34782608678695085948531993458097026e-02L,
+	-3.99999999632663469330634215991142368e-02L,
+	 3.70370363987423702891250829918659723e-02L,
+	-3.44827496515048090726669907612335954e-02L,
+	 3.22579620681420149871973710852268528e-02L,
+	-3.03020767654269261041647570626778067e-02L,
+	 2.85641979882534783223403715930946138e-02L,
+	-2.69824879726738568189929461383741323e-02L,
+	 2.54194698498808542954187110873675769e-02L,
+	-2.35083879708189059926183138130183215e-02L,
+	 2.04832358998165364349957325067131428e-02L,
+	-1.54489555488544397858507248612362957e-02L,
+	 8.64492360989278761493037861575248038e-03L,
+	-2.58521121597609872727919154569765469e-03L,
+};
+
+static long double T_even(long double x)
+{
+	return (aT[0] + x * (aT[2] + x * (aT[4] + x * (aT[6] + x * (aT[8] +
+		x * (aT[10] + x * (aT[12] + x * (aT[14] + x * (aT[16] +
+		x * (aT[18] + x * (aT[20] + x * aT[22])))))))))));
+}
+
+static long double T_odd(long double x)
+{
+	return (aT[1] + x * (aT[3] + x * (aT[5] + x * (aT[7] + x * (aT[9] +
+		x * (aT[11] + x * (aT[13] + x * (aT[15] + x * (aT[17] +
+		x * (aT[19] + x * (aT[21] + x * aT[23])))))))))));
+}
+#endif
+
+long double atanl(long double x)
+{
+	union ldshape u = {x};
+	long double w, s1, s2, z;
+	int id;
+	unsigned e = u.i.se & 0x7fff;
+	unsigned sign = u.i.se >> 15;
+	unsigned expman;
+
+	if (e >= 0x3fff + LDBL_MANT_DIG + 1) { /* if |x| is large, atan(x)~=pi/2 */
+		if (isnan(x))
+			return x;
+		return sign ? -atanhi[3] : atanhi[3];
+	}
+	/* Extract the exponent and the first few bits of the mantissa. */
+	expman = EXPMAN(u);
+	if (expman < ((0x3fff - 2) << 8) + 0xc0) {  /* |x| < 0.4375 */
+		if (e < 0x3fff - (LDBL_MANT_DIG+1)/2) {   /* if |x| is small, atanl(x)~=x */
+			/* raise underflow if subnormal */
+			if (e == 0)
+				FORCE_EVAL((float)x);
+			return x;
+		}
+		id = -1;
+	} else {
+		x = fabsl(x);
+		if (expman < (0x3fff << 8) + 0x30) {  /* |x| < 1.1875 */
+			if (expman < ((0x3fff - 1) << 8) + 0x60) { /*  7/16 <= |x| < 11/16 */
+				id = 0;
+				x = (2.0*x-1.0)/(2.0+x);
+			} else {                                 /* 11/16 <= |x| < 19/16 */
+				id = 1;
+				x = (x-1.0)/(x+1.0);
+			}
+		} else {
+			if (expman < ((0x3fff + 1) << 8) + 0x38) { /* |x| < 2.4375 */
+				id = 2;
+				x = (x-1.5)/(1.0+1.5*x);
+			} else {                                 /* 2.4375 <= |x| */
+				id = 3;
+				x = -1.0/x;
+			}
+		}
+	}
+	/* end of argument reduction */
+	z = x*x;
+	w = z*z;
+	/* break sum aT[i]z**(i+1) into odd and even poly */
+	s1 = z*T_even(w);
+	s2 = w*T_odd(w);
+	if (id < 0)
+		return x - x*(s1+s2);
+	z = atanhi[id] - ((x*(s1+s2) - atanlo[id]) - x);
+	return sign ? -z : z;
+}
+#endif
diff --git a/src/math/logl.c b/src/math/logl.c
new file mode 100644
index 0000000..5d53659
--- /dev/null
+++ b/src/math/logl.c
@@ -0,0 +1,175 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_logl.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Natural logarithm, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, logl();
+ *
+ * y = logl( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns the base e (2.718...) logarithm of x.
+ *
+ * The argument is separated into its exponent and fractional
+ * parts.  If the exponent is between -1 and +1, the logarithm
+ * of the fraction is approximated by
+ *
+ *     log(1+x) = x - 0.5 x**2 + x**3 P(x)/Q(x).
+ *
+ * Otherwise, setting  z = 2(x-1)/(x+1),
+ *
+ *     log(x) = log(1+z/2) - log(1-z/2) = z + z**3 P(z)/Q(z).
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE      0.5, 2.0    150000      8.71e-20    2.75e-20
+ *    IEEE     exp(+-10000) 100000      5.39e-20    2.34e-20
+ *
+ * In the tests over the interval exp(+-10000), the logarithms
+ * of the random arguments were uniformly distributed over
+ * [-10000, +10000].
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double logl(long double x)
+{
+	return log(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/* Coefficients for log(1+x) = x - x**2/2 + x**3 P(x)/Q(x)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 2.32e-20
+ */
+static const long double P[] = {
+ 4.5270000862445199635215E-5L,
+ 4.9854102823193375972212E-1L,
+ 6.5787325942061044846969E0L,
+ 2.9911919328553073277375E1L,
+ 6.0949667980987787057556E1L,
+ 5.7112963590585538103336E1L,
+ 2.0039553499201281259648E1L,
+};
+static const long double Q[] = {
+/* 1.0000000000000000000000E0,*/
+ 1.5062909083469192043167E1L,
+ 8.3047565967967209469434E1L,
+ 2.2176239823732856465394E2L,
+ 3.0909872225312059774938E2L,
+ 2.1642788614495947685003E2L,
+ 6.0118660497603843919306E1L,
+};
+
+/* Coefficients for log(x) = z + z^3 P(z^2)/Q(z^2),
+ * where z = 2(x-1)/(x+1)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 6.16e-22
+ */
+static const long double R[4] = {
+ 1.9757429581415468984296E-3L,
+-7.1990767473014147232598E-1L,
+ 1.0777257190312272158094E1L,
+-3.5717684488096787370998E1L,
+};
+static const long double S[4] = {
+/* 1.00000000000000000000E0L,*/
+-2.6201045551331104417768E1L,
+ 1.9361891836232102174846E2L,
+-4.2861221385716144629696E2L,
+};
+static const long double C1 = 6.9314575195312500000000E-1L;
+static const long double C2 = 1.4286068203094172321215E-6L;
+
+#define SQRTH 0.70710678118654752440L
+
+long double logl(long double x)
+{
+	long double y, z;
+	int e;
+
+	if (isnan(x))
+		return x;
+	if (x == INFINITY)
+		return x;
+	if (x <= 0.0) {
+		if (x == 0.0)
+			return -1/(x*x); /* -inf with divbyzero */
+		return 0/0.0f; /* nan with invalid */
+	}
+
+	/* separate mantissa from exponent */
+	/* Note, frexp is used so that denormal numbers
+	 * will be handled properly.
+	 */
+	x = frexpl(x, &e);
+
+	/* logarithm using log(x) = z + z**3 P(z)/Q(z),
+	 * where z = 2(x-1)/(x+1)
+	 */
+	if (e > 2 || e < -2) {
+		if (x < SQRTH) {  /* 2(2x-1)/(2x+1) */
+			e -= 1;
+			z = x - 0.5;
+			y = 0.5 * z + 0.5;
+		} else {  /*  2 (x-1)/(x+1)   */
+			z = x - 0.5;
+			z -= 0.5;
+			y = 0.5 * x  + 0.5;
+		}
+		x = z / y;
+		z = x*x;
+		z = x * (z * __polevll(z, R, 3) / __p1evll(z, S, 3));
+		z = z + e * C2;
+		z = z + x;
+		z = z + e * C1;
+		return z;
+	}
+
+	/* logarithm using log(1+x) = x - .5x**2 + x**3 P(x)/Q(x) */
+	if (x < SQRTH) {
+		e -= 1;
+		x = 2.0*x - 1.0;
+	} else {
+		x = x - 1.0;
+	}
+	z = x*x;
+	y = x * (z * __polevll(x, P, 6) / __p1evll(x, Q, 6));
+	y = y + e * C2;
+	z = y - 0.5*z;
+	/* Note, the sum of above terms does not exceed x/4,
+	 * so it contributes at most about 1/4 lsb to the error.
+	 */
+	z = z + x;
+	z = z + e * C1; /* This sum has an error of 1/2 lsb. */
+	return z;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double logl(long double x)
+{
+	return log(x);
+}
+#endif
-- 
Gitee


From fd15ce416f019a5df4a3e72eae2a439648aa7f6f Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Mon, 5 Sep 2022 01:36:25 +0800
Subject: [PATCH 06/13] feat: update math and time

---
 src/Makefile             |   1 +
 src/gcc/x86/divmod.c     |   5 +
 src/include/math.h       |   3 +
 src/include/time.h       |  61 +++++
 src/internal/libm.h      |   3 +
 src/math/exp2f_data.c    |  35 +++
 src/math/exp2f_data.h    |  23 ++
 src/math/powf_data.c     |  34 +++
 src/math/powf_data.h     |  26 ++
 src/math/powl.c          | 522 +++++++++++++++++++++++++++++++++++++++
 src/stdio/__init_stdio.c |   1 +
 src/string/strcoll.c     |   7 +
 src/time/__secs_to_tm.c  |  95 +++++++
 src/time/__tm_to_secs.c  | 110 +++++++++
 src/time/asctime.c       |  23 ++
 src/time/clock.c         |   8 +
 src/time/ctime.c         |  10 +
 src/time/difftime.c      |   6 +
 src/time/gmtime.c        |  17 ++
 src/time/localtime.c     |  11 +
 src/time/mktime.c        |  11 +
 src/time/strftime.c      | 224 +++++++++++++++++
 src/time/time.c          |  29 +++
 src/time/time_impl.h     |   4 +
 src/time/timespec_get.c  |  18 ++
 25 files changed, 1287 insertions(+)
 create mode 100644 src/include/time.h
 create mode 100644 src/math/exp2f_data.c
 create mode 100644 src/math/exp2f_data.h
 create mode 100644 src/math/powf_data.c
 create mode 100644 src/math/powf_data.h
 create mode 100644 src/math/powl.c
 create mode 100644 src/string/strcoll.c
 create mode 100644 src/time/__secs_to_tm.c
 create mode 100644 src/time/__tm_to_secs.c
 create mode 100644 src/time/asctime.c
 create mode 100644 src/time/clock.c
 create mode 100644 src/time/ctime.c
 create mode 100644 src/time/difftime.c
 create mode 100644 src/time/gmtime.c
 create mode 100644 src/time/localtime.c
 create mode 100644 src/time/mktime.c
 create mode 100644 src/time/strftime.c
 create mode 100644 src/time/time.c
 create mode 100644 src/time/time_impl.h
 create mode 100644 src/time/timespec_get.c

diff --git a/src/Makefile b/src/Makefile
index 5bb5262..c9d82d9 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -43,6 +43,7 @@ X_LIBDIRS	:= $(LIBS_DIR)
 # we must link nxbase lib.
 X_LIBS		+= libnxbase.a
 
+SRC			+= time/
 SRC			+= process/
 SRC			+= prng/
 SRC			+= complex/
diff --git a/src/gcc/x86/divmod.c b/src/gcc/x86/divmod.c
index fc5ce08..e2dfa3a 100644
--- a/src/gcc/x86/divmod.c
+++ b/src/gcc/x86/divmod.c
@@ -7,3 +7,8 @@ unsigned long long __umoddi3(unsigned long long a, unsigned long long b)
 {
     return (unsigned long)a % (unsigned long)b;
 }
+
+long long __moddi3(long long a, long long b)
+{
+    return (long)a % (long)b;
+}
diff --git a/src/include/math.h b/src/include/math.h
index 95f384e..47b90b6 100644
--- a/src/include/math.h
+++ b/src/include/math.h
@@ -185,6 +185,8 @@ double	asinh(double);
 float	asinhf(float);
 double	atan(double);
 float	atanf(float);
+long double atanl(long double);
+
 double	atan2(double, double);
 float	atan2f(float, float);
 long double atan2l(long double, long double);
@@ -249,6 +251,7 @@ double	modf(double, double *);
 float	modff(float, float *);
 double	pow(double, double);
 float	powf(float, float);
+long double powl(long double x, long double y);
 double	rint(double);
 float	rintf(float);
 double	round(double);
diff --git a/src/include/time.h b/src/include/time.h
new file mode 100644
index 0000000..e1c1317
--- /dev/null
+++ b/src/include/time.h
@@ -0,0 +1,61 @@
+#ifndef	_TIME_H
+#define _TIME_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <features.h>
+
+#ifdef __cplusplus
+#define NULL 0L
+#else
+#define NULL ((void*)0)
+#endif
+
+typedef long    time_t;
+typedef long    clock_t;
+
+struct timespec {
+    time_t tv_sec; // seconds
+    long tv_nsec; // and nanoseconds
+};
+
+#include <bits/alltypes.h>
+#include <bits/types.h>
+
+struct tm {
+	int tm_sec;
+	int tm_min;
+	int tm_hour;
+	int tm_mday;
+	int tm_mon;
+	int tm_year;
+	int tm_wday;
+	int tm_yday;
+	int tm_isdst;
+	long __tm_gmtoff;
+	const char *__tm_zone;
+};
+
+clock_t clock (void);
+time_t time (time_t *);
+double difftime (time_t, time_t);
+time_t mktime (struct tm *);
+size_t strftime (char *__restrict, size_t, const char *__restrict, const struct tm *__restrict);
+struct tm *gmtime (const time_t *);
+struct tm *localtime (const time_t *);
+char *asctime (const struct tm *);
+char *ctime (const time_t *);
+int timespec_get(struct timespec *, int);
+
+#define CLOCKS_PER_SEC 1000L
+
+#define TIME_UTC 1
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif
diff --git a/src/internal/libm.h b/src/internal/libm.h
index 60847c4..cf56e50 100644
--- a/src/internal/libm.h
+++ b/src/internal/libm.h
@@ -71,4 +71,7 @@ union ldshape {
 hidden long double __math_invalidl(long double);
 #endif
 
+hidden long double __polevll(long double, const long double *, int);
+hidden long double __p1evll(long double, const long double *, int);
+
 #endif
diff --git a/src/math/exp2f_data.c b/src/math/exp2f_data.c
new file mode 100644
index 0000000..be32472
--- /dev/null
+++ b/src/math/exp2f_data.c
@@ -0,0 +1,35 @@
+/*
+ * Shared data between expf, exp2f and powf.
+ *
+ * Copyright (c) 2017-2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "exp2f_data.h"
+
+#define N (1 << EXP2F_TABLE_BITS)
+
+const struct exp2f_data __exp2f_data = {
+  /* tab[i] = uint(2^(i/N)) - (i << 52-BITS)
+     used for computing 2^(k/N) for an int |k| < 150 N as
+     double(tab[k%N] + (k << 52-BITS)) */
+  .tab = {
+0x3ff0000000000000, 0x3fefd9b0d3158574, 0x3fefb5586cf9890f, 0x3fef9301d0125b51,
+0x3fef72b83c7d517b, 0x3fef54873168b9aa, 0x3fef387a6e756238, 0x3fef1e9df51fdee1,
+0x3fef06fe0a31b715, 0x3feef1a7373aa9cb, 0x3feedea64c123422, 0x3feece086061892d,
+0x3feebfdad5362a27, 0x3feeb42b569d4f82, 0x3feeab07dd485429, 0x3feea47eb03a5585,
+0x3feea09e667f3bcd, 0x3fee9f75e8ec5f74, 0x3feea11473eb0187, 0x3feea589994cce13,
+0x3feeace5422aa0db, 0x3feeb737b0cdc5e5, 0x3feec49182a3f090, 0x3feed503b23e255d,
+0x3feee89f995ad3ad, 0x3feeff76f2fb5e47, 0x3fef199bdd85529c, 0x3fef3720dcef9069,
+0x3fef5818dcfba487, 0x3fef7c97337b9b5f, 0x3fefa4afa2a490da, 0x3fefd0765b6e4540,
+  },
+  .shift_scaled = 0x1.8p+52 / N,
+  .poly = {
+  0x1.c6af84b912394p-5, 0x1.ebfce50fac4f3p-3, 0x1.62e42ff0c52d6p-1,
+  },
+  .shift = 0x1.8p+52,
+  .invln2_scaled = 0x1.71547652b82fep+0 * N,
+  .poly_scaled = {
+  0x1.c6af84b912394p-5/N/N/N, 0x1.ebfce50fac4f3p-3/N/N, 0x1.62e42ff0c52d6p-1/N,
+  },
+};
diff --git a/src/math/exp2f_data.h b/src/math/exp2f_data.h
new file mode 100644
index 0000000..fe744f1
--- /dev/null
+++ b/src/math/exp2f_data.h
@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2017-2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _EXP2F_DATA_H
+#define _EXP2F_DATA_H
+
+#include <features.h>
+#include <stdint.h>
+
+/* Shared between expf, exp2f and powf.  */
+#define EXP2F_TABLE_BITS 5
+#define EXP2F_POLY_ORDER 3
+extern hidden const struct exp2f_data {
+	uint64_t tab[1 << EXP2F_TABLE_BITS];
+	double shift_scaled;
+	double poly[EXP2F_POLY_ORDER];
+	double shift;
+	double invln2_scaled;
+	double poly_scaled[EXP2F_POLY_ORDER];
+} __exp2f_data;
+
+#endif
diff --git a/src/math/powf_data.c b/src/math/powf_data.c
new file mode 100644
index 0000000..13e1d9a
--- /dev/null
+++ b/src/math/powf_data.c
@@ -0,0 +1,34 @@
+/*
+ * Data definition for powf.
+ *
+ * Copyright (c) 2017-2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "powf_data.h"
+
+const struct powf_log2_data __powf_log2_data = {
+  .tab = {
+  { 0x1.661ec79f8f3bep+0, -0x1.efec65b963019p-2 * POWF_SCALE },
+  { 0x1.571ed4aaf883dp+0, -0x1.b0b6832d4fca4p-2 * POWF_SCALE },
+  { 0x1.49539f0f010bp+0, -0x1.7418b0a1fb77bp-2 * POWF_SCALE },
+  { 0x1.3c995b0b80385p+0, -0x1.39de91a6dcf7bp-2 * POWF_SCALE },
+  { 0x1.30d190c8864a5p+0, -0x1.01d9bf3f2b631p-2 * POWF_SCALE },
+  { 0x1.25e227b0b8eap+0, -0x1.97c1d1b3b7afp-3 * POWF_SCALE },
+  { 0x1.1bb4a4a1a343fp+0, -0x1.2f9e393af3c9fp-3 * POWF_SCALE },
+  { 0x1.12358f08ae5bap+0, -0x1.960cbbf788d5cp-4 * POWF_SCALE },
+  { 0x1.0953f419900a7p+0, -0x1.a6f9db6475fcep-5 * POWF_SCALE },
+  { 0x1p+0, 0x0p+0 * POWF_SCALE },
+  { 0x1.e608cfd9a47acp-1, 0x1.338ca9f24f53dp-4 * POWF_SCALE },
+  { 0x1.ca4b31f026aap-1, 0x1.476a9543891bap-3 * POWF_SCALE },
+  { 0x1.b2036576afce6p-1, 0x1.e840b4ac4e4d2p-3 * POWF_SCALE },
+  { 0x1.9c2d163a1aa2dp-1, 0x1.40645f0c6651cp-2 * POWF_SCALE },
+  { 0x1.886e6037841edp-1, 0x1.88e9c2c1b9ff8p-2 * POWF_SCALE },
+  { 0x1.767dcf5534862p-1, 0x1.ce0a44eb17bccp-2 * POWF_SCALE },
+  },
+  .poly = {
+  0x1.27616c9496e0bp-2 * POWF_SCALE, -0x1.71969a075c67ap-2 * POWF_SCALE,
+  0x1.ec70a6ca7baddp-2 * POWF_SCALE, -0x1.7154748bef6c8p-1 * POWF_SCALE,
+  0x1.71547652ab82bp0 * POWF_SCALE,
+  }
+};
diff --git a/src/math/powf_data.h b/src/math/powf_data.h
new file mode 100644
index 0000000..5b136e2
--- /dev/null
+++ b/src/math/powf_data.h
@@ -0,0 +1,26 @@
+/*
+ * Copyright (c) 2017-2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _POWF_DATA_H
+#define _POWF_DATA_H
+
+#include "libm.h"
+#include "exp2f_data.h"
+
+#define POWF_LOG2_TABLE_BITS 4
+#define POWF_LOG2_POLY_ORDER 5
+#if TOINT_INTRINSICS
+#define POWF_SCALE_BITS EXP2F_TABLE_BITS
+#else
+#define POWF_SCALE_BITS 0
+#endif
+#define POWF_SCALE ((double)(1 << POWF_SCALE_BITS))
+extern hidden const struct powf_log2_data {
+	struct {
+		double invc, logc;
+	} tab[1 << POWF_LOG2_TABLE_BITS];
+	double poly[POWF_LOG2_POLY_ORDER];
+} __powf_log2_data;
+
+#endif
diff --git a/src/math/powl.c b/src/math/powl.c
new file mode 100644
index 0000000..5b6da07
--- /dev/null
+++ b/src/math/powl.c
@@ -0,0 +1,522 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_powl.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*                                                      powl.c
+ *
+ *      Power function, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, z, powl();
+ *
+ * z = powl( x, y );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Computes x raised to the yth power.  Analytically,
+ *
+ *      x**y  =  exp( y log(x) ).
+ *
+ * Following Cody and Waite, this program uses a lookup table
+ * of 2**-i/32 and pseudo extended precision arithmetic to
+ * obtain several extra bits of accuracy in both the logarithm
+ * and the exponential.
+ *
+ *
+ * ACCURACY:
+ *
+ * The relative error of pow(x,y) can be estimated
+ * by   y dl ln(2),   where dl is the absolute error of
+ * the internally computed base 2 logarithm.  At the ends
+ * of the approximation interval the logarithm equal 1/32
+ * and its relative error is about 1 lsb = 1.1e-19.  Hence
+ * the predicted relative error in the result is 2.3e-21 y .
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *
+ *    IEEE     +-1000       40000      2.8e-18      3.7e-19
+ * .001 < x < 1000, with log(x) uniformly distributed.
+ * -1000 < y < 1000, y uniformly distributed.
+ *
+ *    IEEE     0,8700       60000      6.5e-18      1.0e-18
+ * 0.99 < x < 1.01, 0 < y < 8700, uniformly distributed.
+ *
+ *
+ * ERROR MESSAGES:
+ *
+ *   message         condition      value returned
+ * pow overflow     x**y > MAXNUM      INFINITY
+ * pow underflow   x**y < 1/MAXNUM       0.0
+ * pow domain      x<0 and y noninteger  0.0
+ *
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double powl(long double x, long double y)
+{
+	return pow(x, y);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+
+/* Table size */
+#define NXT 32
+
+/* log(1+x) =  x - .5x^2 + x^3 *  P(z)/Q(z)
+ * on the domain  2^(-1/32) - 1  <=  x  <=  2^(1/32) - 1
+ */
+static const long double P[] = {
+ 8.3319510773868690346226E-4L,
+ 4.9000050881978028599627E-1L,
+ 1.7500123722550302671919E0L,
+ 1.4000100839971580279335E0L,
+};
+static const long double Q[] = {
+/* 1.0000000000000000000000E0L,*/
+ 5.2500282295834889175431E0L,
+ 8.4000598057587009834666E0L,
+ 4.2000302519914740834728E0L,
+};
+/* A[i] = 2^(-i/32), rounded to IEEE long double precision.
+ * If i is even, A[i] + B[i/2] gives additional accuracy.
+ */
+static const long double A[33] = {
+ 1.0000000000000000000000E0L,
+ 9.7857206208770013448287E-1L,
+ 9.5760328069857364691013E-1L,
+ 9.3708381705514995065011E-1L,
+ 9.1700404320467123175367E-1L,
+ 8.9735453750155359320742E-1L,
+ 8.7812608018664974155474E-1L,
+ 8.5930964906123895780165E-1L,
+ 8.4089641525371454301892E-1L,
+ 8.2287773907698242225554E-1L,
+ 8.0524516597462715409607E-1L,
+ 7.8799042255394324325455E-1L,
+ 7.7110541270397041179298E-1L,
+ 7.5458221379671136985669E-1L,
+ 7.3841307296974965571198E-1L,
+ 7.2259040348852331001267E-1L,
+ 7.0710678118654752438189E-1L,
+ 6.9195494098191597746178E-1L,
+ 6.7712777346844636413344E-1L,
+ 6.6261832157987064729696E-1L,
+ 6.4841977732550483296079E-1L,
+ 6.3452547859586661129850E-1L,
+ 6.2092890603674202431705E-1L,
+ 6.0762367999023443907803E-1L,
+ 5.9460355750136053334378E-1L,
+ 5.8186242938878875689693E-1L,
+ 5.6939431737834582684856E-1L,
+ 5.5719337129794626814472E-1L,
+ 5.4525386633262882960438E-1L,
+ 5.3357020033841180906486E-1L,
+ 5.2213689121370692017331E-1L,
+ 5.1094857432705833910408E-1L,
+ 5.0000000000000000000000E-1L,
+};
+static const long double B[17] = {
+ 0.0000000000000000000000E0L,
+ 2.6176170809902549338711E-20L,
+-1.0126791927256478897086E-20L,
+ 1.3438228172316276937655E-21L,
+ 1.2207982955417546912101E-20L,
+-6.3084814358060867200133E-21L,
+ 1.3164426894366316434230E-20L,
+-1.8527916071632873716786E-20L,
+ 1.8950325588932570796551E-20L,
+ 1.5564775779538780478155E-20L,
+ 6.0859793637556860974380E-21L,
+-2.0208749253662532228949E-20L,
+ 1.4966292219224761844552E-20L,
+ 3.3540909728056476875639E-21L,
+-8.6987564101742849540743E-22L,
+-1.2327176863327626135542E-20L,
+ 0.0000000000000000000000E0L,
+};
+
+/* 2^x = 1 + x P(x),
+ * on the interval -1/32 <= x <= 0
+ */
+static const long double R[] = {
+ 1.5089970579127659901157E-5L,
+ 1.5402715328927013076125E-4L,
+ 1.3333556028915671091390E-3L,
+ 9.6181291046036762031786E-3L,
+ 5.5504108664798463044015E-2L,
+ 2.4022650695910062854352E-1L,
+ 6.9314718055994530931447E-1L,
+};
+
+#define MEXP (NXT*16384.0L)
+/* The following if denormal numbers are supported, else -MEXP: */
+#define MNEXP (-NXT*(16384.0L+64.0L))
+/* log2(e) - 1 */
+#define LOG2EA 0.44269504088896340735992L
+
+#define F W
+#define Fa Wa
+#define Fb Wb
+#define G W
+#define Ga Wa
+#define Gb u
+#define H W
+#define Ha Wb
+#define Hb Wb
+
+static const long double MAXLOGL = 1.1356523406294143949492E4L;
+static const long double MINLOGL = -1.13994985314888605586758E4L;
+static const long double LOGE2L = 6.9314718055994530941723E-1L;
+static const long double huge = 0x1p10000L;
+/* XXX Prevent gcc from erroneously constant folding this. */
+static const volatile long double twom10000 = 0x1p-10000L;
+
+static long double reducl(long double);
+static long double powil(long double, int);
+
+long double powl(long double x, long double y)
+{
+	/* double F, Fa, Fb, G, Ga, Gb, H, Ha, Hb */
+	int i, nflg, iyflg, yoddint;
+	long e;
+	volatile long double z=0;
+	long double w=0, W=0, Wa=0, Wb=0, ya=0, yb=0, u=0;
+
+	/* make sure no invalid exception is raised by nan comparision */
+	if (isnan(x)) {
+		if (!isnan(y) && y == 0.0)
+			return 1.0;
+		return x;
+	}
+	if (isnan(y)) {
+		if (x == 1.0)
+			return 1.0;
+		return y;
+	}
+	if (x == 1.0)
+		return 1.0; /* 1**y = 1, even if y is nan */
+	if (x == -1.0 && !isfinite(y))
+		return 1.0; /* -1**inf = 1 */
+	if (y == 0.0)
+		return 1.0; /* x**0 = 1, even if x is nan */
+	if (y == 1.0)
+		return x;
+	if (y >= LDBL_MAX) {
+		if (x > 1.0 || x < -1.0)
+			return INFINITY;
+		if (x != 0.0)
+			return 0.0;
+	}
+	if (y <= -LDBL_MAX) {
+		if (x > 1.0 || x < -1.0)
+			return 0.0;
+		if (x != 0.0 || y == -INFINITY)
+			return INFINITY;
+	}
+	if (x >= LDBL_MAX) {
+		if (y > 0.0)
+			return INFINITY;
+		return 0.0;
+	}
+
+	w = floorl(y);
+
+	/* Set iyflg to 1 if y is an integer. */
+	iyflg = 0;
+	if (w == y)
+		iyflg = 1;
+
+	/* Test for odd integer y. */
+	yoddint = 0;
+	if (iyflg) {
+		ya = fabsl(y);
+		ya = floorl(0.5 * ya);
+		yb = 0.5 * fabsl(w);
+		if( ya != yb )
+			yoddint = 1;
+	}
+
+	if (x <= -LDBL_MAX) {
+		if (y > 0.0) {
+			if (yoddint)
+				return -INFINITY;
+			return INFINITY;
+		}
+		if (y < 0.0) {
+			if (yoddint)
+				return -0.0;
+			return 0.0;
+		}
+	}
+	nflg = 0; /* (x<0)**(odd int) */
+	if (x <= 0.0) {
+		if (x == 0.0) {
+			if (y < 0.0) {
+				if (signbit(x) && yoddint)
+					/* (-0.0)**(-odd int) = -inf, divbyzero */
+					return -1.0/0.0;
+				/* (+-0.0)**(negative) = inf, divbyzero */
+				return 1.0/0.0;
+			}
+			if (signbit(x) && yoddint)
+				return -0.0;
+			return 0.0;
+		}
+		if (iyflg == 0)
+			return (x - x) / (x - x); /* (x<0)**(non-int) is NaN */
+		/* (x<0)**(integer) */
+		if (yoddint)
+			nflg = 1; /* negate result */
+		x = -x;
+	}
+	/* (+integer)**(integer)  */
+	if (iyflg && floorl(x) == x && fabsl(y) < 32768.0) {
+		w = powil(x, (int)y);
+		return nflg ? -w : w;
+	}
+
+	/* separate significand from exponent */
+	x = frexpl(x, &i);
+	e = i;
+
+	/* find significand in antilog table A[] */
+	i = 1;
+	if (x <= A[17])
+		i = 17;
+	if (x <= A[i+8])
+		i += 8;
+	if (x <= A[i+4])
+		i += 4;
+	if (x <= A[i+2])
+		i += 2;
+	if (x >= A[1])
+		i = -1;
+	i += 1;
+
+	/* Find (x - A[i])/A[i]
+	 * in order to compute log(x/A[i]):
+	 *
+	 * log(x) = log( a x/a ) = log(a) + log(x/a)
+	 *
+	 * log(x/a) = log(1+v),  v = x/a - 1 = (x-a)/a
+	 */
+	x -= A[i];
+	x -= B[i/2];
+	x /= A[i];
+
+	/* rational approximation for log(1+v):
+	 *
+	 * log(1+v)  =  v  -  v**2/2  +  v**3 P(v) / Q(v)
+	 */
+	z = x*x;
+	w = x * (z * __polevll(x, P, 3) / __p1evll(x, Q, 3));
+	w = w - 0.5*z;
+
+	/* Convert to base 2 logarithm:
+	 * multiply by log2(e) = 1 + LOG2EA
+	 */
+	z = LOG2EA * w;
+	z += w;
+	z += LOG2EA * x;
+	z += x;
+
+	/* Compute exponent term of the base 2 logarithm. */
+	w = -i;
+	w /= NXT;
+	w += e;
+	/* Now base 2 log of x is w + z. */
+
+	/* Multiply base 2 log by y, in extended precision. */
+
+	/* separate y into large part ya
+	 * and small part yb less than 1/NXT
+	 */
+	ya = reducl(y);
+	yb = y - ya;
+
+	/* (w+z)(ya+yb)
+	 * = w*ya + w*yb + z*y
+	 */
+	F = z * y  +  w * yb;
+	Fa = reducl(F);
+	Fb = F - Fa;
+
+	G = Fa + w * ya;
+	Ga = reducl(G);
+	Gb = G - Ga;
+
+	H = Fb + Gb;
+	Ha = reducl(H);
+	w = (Ga + Ha) * NXT;
+
+	/* Test the power of 2 for overflow */
+	if (w > MEXP)
+		return huge * huge;  /* overflow */
+	if (w < MNEXP)
+		return twom10000 * twom10000;  /* underflow */
+
+	e = w;
+	Hb = H - Ha;
+
+	if (Hb > 0.0) {
+		e += 1;
+		Hb -= 1.0/NXT;  /*0.0625L;*/
+	}
+
+	/* Now the product y * log2(x)  =  Hb + e/NXT.
+	 *
+	 * Compute base 2 exponential of Hb,
+	 * where -0.0625 <= Hb <= 0.
+	 */
+	z = Hb * __polevll(Hb, R, 6);  /*  z = 2**Hb - 1  */
+
+	/* Express e/NXT as an integer plus a negative number of (1/NXT)ths.
+	 * Find lookup table entry for the fractional power of 2.
+	 */
+	if (e < 0)
+		i = 0;
+	else
+		i = 1;
+	i = e/NXT + i;
+	e = NXT*i - e;
+	w = A[e];
+	z = w * z;  /*  2**-e * ( 1 + (2**Hb-1) )  */
+	z = z + w;
+	z = scalbnl(z, i);  /* multiply by integer power of 2 */
+
+	if (nflg)
+		z = -z;
+	return z;
+}
+
+
+/* Find a multiple of 1/NXT that is within 1/NXT of x. */
+static long double reducl(long double x)
+{
+	long double t;
+
+	t = x * NXT;
+	t = floorl(t);
+	t = t / NXT;
+	return t;
+}
+
+/*
+ *      Positive real raised to integer power, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, powil();
+ * int n;
+ *
+ * y = powil( x, n );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns argument x>0 raised to the nth power.
+ * The routine efficiently decomposes n as a sum of powers of
+ * two. The desired power is a product of two-to-the-kth
+ * powers of x.  Thus to compute the 32767 power of x requires
+ * 28 multiplications instead of 32767 multiplications.
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   x domain   n domain  # trials      peak         rms
+ *    IEEE     .001,1000  -1022,1023  50000       4.3e-17     7.8e-18
+ *    IEEE        1,2     -1022,1023  20000       3.9e-17     7.6e-18
+ *    IEEE     .99,1.01     0,8700    10000       3.6e-16     7.2e-17
+ *
+ * Returns MAXNUM on overflow, zero on underflow.
+ */
+
+static long double powil(long double x, int nn)
+{
+	long double ww, y;
+	long double s;
+	int n, e, sign, lx;
+
+	if (nn == 0)
+		return 1.0;
+
+	if (nn < 0) {
+		sign = -1;
+		n = -nn;
+	} else {
+		sign = 1;
+		n = nn;
+	}
+
+	/* Overflow detection */
+
+	/* Calculate approximate logarithm of answer */
+	s = x;
+	s = frexpl( s, &lx);
+	e = (lx - 1)*n;
+	if ((e == 0) || (e > 64) || (e < -64)) {
+		s = (s - 7.0710678118654752e-1L) / (s +  7.0710678118654752e-1L);
+		s = (2.9142135623730950L * s - 0.5 + lx) * nn * LOGE2L;
+	} else {
+		s = LOGE2L * e;
+	}
+
+	if (s > MAXLOGL)
+		return huge * huge;  /* overflow */
+
+	if (s < MINLOGL)
+		return twom10000 * twom10000;  /* underflow */
+	/* Handle tiny denormal answer, but with less accuracy
+	 * since roundoff error in 1.0/x will be amplified.
+	 * The precise demarcation should be the gradual underflow threshold.
+	 */
+	if (s < -MAXLOGL+2.0) {
+		x = 1.0/x;
+		sign = -sign;
+	}
+
+	/* First bit of the power */
+	if (n & 1)
+		y = x;
+	else
+		y = 1.0;
+
+	ww = x;
+	n >>= 1;
+	while (n) {
+		ww = ww * ww;   /* arg to the 2-to-the-kth power */
+		if (n & 1)     /* if that bit is set, then include in product */
+			y *= ww;
+		n >>= 1;
+	}
+
+	if (sign < 0)
+		y = 1.0/y;
+	return y;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double powl(long double x, long double y)
+{
+	return pow(x, y);
+}
+#endif
diff --git a/src/stdio/__init_stdio.c b/src/stdio/__init_stdio.c
index 8634d09..18918bf 100644
--- a/src/stdio/__init_stdio.c
+++ b/src/stdio/__init_stdio.c
@@ -16,4 +16,5 @@ void __init_stdio(void)
     /* no buffer for stdin & stdout */
     setbuf(stdin, NULL);
     setbuf(stdout, NULL);
+    setbuf(stderr, NULL);
 }
diff --git a/src/string/strcoll.c b/src/string/strcoll.c
new file mode 100644
index 0000000..c96907b
--- /dev/null
+++ b/src/string/strcoll.c
@@ -0,0 +1,7 @@
+#include <string.h>
+#include <locale.h>
+
+int strcoll(const char *l, const char *r)
+{
+	return strcmp(l, r);
+}
diff --git a/src/time/__secs_to_tm.c b/src/time/__secs_to_tm.c
new file mode 100644
index 0000000..c32ba03
--- /dev/null
+++ b/src/time/__secs_to_tm.c
@@ -0,0 +1,95 @@
+/*
+ * libc/time/__secs_to_tm.c
+ */
+
+#include <limits.h>
+#include <time.h>
+#include <stdio.h>
+
+/* 2000-03-01 (mod 400 year, immediately after feb29 */
+#define LEAPOCH			(946684800LL + 86400 * (31 + 29))
+#define DAYS_PER_400Y	(365 * 400 + 97)
+#define DAYS_PER_100Y	(365 * 100 + 24)
+#define DAYS_PER_4Y		(365 * 4 + 1)
+
+int __secs_to_tm(time_t t, struct tm * tm)
+{
+	time_t days, secs, years;
+	int remdays, remsecs, remyears;
+	int qc_cycles, c_cycles, q_cycles;
+	int months;
+	int wday, yday, leap;
+	static const char days_in_month[] = { 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 31, 29 };
+
+	/* Reject time_t values whose year would overflow int */
+	if(t < INT_MIN * 31622400LL || t > INT_MAX * 31622400LL)
+		return -1;
+
+	secs = t - LEAPOCH;
+
+	days = secs / 86400;
+	remsecs = secs % 86400;
+	if(remsecs < 0)
+	{
+		remsecs += 86400;
+		days--;
+	}
+
+	wday = (3 + days) % 7;
+	if(wday < 0)
+		wday += 7;
+
+	qc_cycles = days / DAYS_PER_400Y;
+	remdays = days % DAYS_PER_400Y;
+	if(remdays < 0)
+	{
+		remdays += DAYS_PER_400Y;
+		qc_cycles--;
+	}
+
+	c_cycles = remdays / DAYS_PER_100Y;
+	if(c_cycles == 4)
+		c_cycles--;
+	remdays -= c_cycles * DAYS_PER_100Y;
+
+	q_cycles = remdays / DAYS_PER_4Y;
+	if(q_cycles == 25)
+		q_cycles--;
+	remdays -= q_cycles * DAYS_PER_4Y;
+
+	remyears = remdays / 365;
+	if(remyears == 4)
+		remyears--;
+	remdays -= remyears * 365;
+
+	leap = !remyears && (q_cycles || !c_cycles);
+	yday = remdays + 31 + 28 + leap;
+	if(yday >= 365 + leap)
+		yday -= 365 + leap;
+
+	years = remyears + 4 * q_cycles + 100 * c_cycles + 400LL * qc_cycles;
+
+	for(months = 0; days_in_month[months] <= remdays; months++)
+		remdays -= days_in_month[months];
+
+	if(months >= 10)
+	{
+		months -= 12;
+		years++;
+	}
+
+	if(years + 100 > INT_MAX || years + 100 < INT_MIN)
+		return -1;
+
+	tm->tm_year = years + 100;
+	tm->tm_mon = months + 2;
+	tm->tm_mday = remdays + 1;
+	tm->tm_wday = wday;
+	tm->tm_yday = yday;
+
+	tm->tm_hour = remsecs / 3600;
+	tm->tm_min = remsecs / 60 % 60;
+	tm->tm_sec = remsecs % 60;
+
+	return 0;
+}
diff --git a/src/time/__tm_to_secs.c b/src/time/__tm_to_secs.c
new file mode 100644
index 0000000..9ec2a06
--- /dev/null
+++ b/src/time/__tm_to_secs.c
@@ -0,0 +1,110 @@
+/*
+ * libc/time/__tm_to_secs.c
+ */
+
+#include <time.h>
+
+static time_t __year_to_secs(time_t year, int * is_leap)
+{
+	if(year - 2ULL <= 136)
+	{
+		int y = year;
+		int leaps = (y - 68) >> 2;
+		if(!((y - 68) & 3))
+		{
+			leaps--;
+			if(is_leap)
+				*is_leap = 1;
+		}
+		else if(is_leap)
+		{
+			*is_leap = 0;
+		}
+		return 31536000 * (y - 70) + 86400 * leaps;
+	}
+
+	int cycles, centuries, leaps, rem;
+
+	if(!is_leap)
+		is_leap = &(int ){ 0 };
+	cycles = (year - 100) / 400;
+	rem = (year - 100) % 400;
+	if(rem < 0)
+	{
+		cycles--;
+		rem += 400;
+	}
+	if(!rem)
+	{
+		*is_leap = 1;
+		centuries = 0;
+		leaps = 0;
+	}
+	else
+	{
+		if(rem >= 200)
+		{
+			if(rem >= 300)
+				centuries = 3, rem -= 300;
+			else
+				centuries = 2, rem -= 200;
+		}
+		else
+		{
+			if(rem >= 100)
+				centuries = 1, rem -= 100;
+			else
+				centuries = 0;
+		}
+		if(!rem)
+		{
+			*is_leap = 0;
+			leaps = 0;
+		}
+		else
+		{
+			leaps = rem / 4U;
+			rem %= 4U;
+			*is_leap = !rem;
+		}
+	}
+
+	leaps += 97 * cycles + 24 * centuries - *is_leap;
+
+	return (year - 100) * 31536000LL + leaps * 86400LL + 946684800 + 86400;
+}
+
+static int __month_to_secs(int month, int is_leap)
+{
+	static const int secs_through_month[] = { 0, 31 * 86400, 59 * 86400, 90 * 86400, 120 * 86400,
+		151 * 86400, 181 * 86400, 212 * 86400, 243 * 86400, 273 * 86400, 304 * 86400, 334 * 86400 };
+	int t = secs_through_month[month];
+	if(is_leap && month >= 2)
+		t += 86400;
+	return t;
+}
+
+time_t __tm_to_secs(const struct tm * tm)
+{
+	int is_leap;
+	time_t year = tm->tm_year;
+	int month = tm->tm_mon;
+	if(month >= 12 || month < 0)
+	{
+		int adj = month / 12;
+		month %= 12;
+		if(month < 0)
+		{
+			adj--;
+			month += 12;
+		}
+		year += adj;
+	}
+	time_t t = __year_to_secs(year, &is_leap);
+	t += __month_to_secs(month, is_leap);
+	t += 86400LL * (tm->tm_mday - 1);
+	t += 3600LL * tm->tm_hour;
+	t += 60LL * tm->tm_min;
+	t += tm->tm_sec;
+	return t;
+}
diff --git a/src/time/asctime.c b/src/time/asctime.c
new file mode 100644
index 0000000..d045366
--- /dev/null
+++ b/src/time/asctime.c
@@ -0,0 +1,23 @@
+/*
+ * libc/time/asctime.c
+ */
+
+#include <stddef.h>
+#include <stdio.h>
+#include <time.h>
+
+static const char * week_days[] = {"Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"};
+static const char * month_days[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"};
+
+char * asctime(const struct tm * tm)
+{
+	static char ascbuf[26];
+
+	if(!tm)
+		return NULL;
+
+	snprintf(ascbuf, 26, "%.3s %.3s%3d %.2d:%.2d:%.2d %d\n",
+			week_days[tm->tm_wday], month_days[tm->tm_mon], tm->tm_mday,
+			tm->tm_hour, tm->tm_min, tm->tm_sec, 1900 + tm->tm_year);
+	return ascbuf;
+}
diff --git a/src/time/clock.c b/src/time/clock.c
new file mode 100644
index 0000000..90c9cbf
--- /dev/null
+++ b/src/time/clock.c
@@ -0,0 +1,8 @@
+#include <time.h>
+#include <limits.h>
+#include <nxbase.h>
+
+clock_t clock(void)
+{
+    return NX_ClockGetMillisecond();
+}
diff --git a/src/time/ctime.c b/src/time/ctime.c
new file mode 100644
index 0000000..aed224e
--- /dev/null
+++ b/src/time/ctime.c
@@ -0,0 +1,10 @@
+/*
+ * libc/time/ctime.c
+ */
+
+#include <time.h>
+
+char * ctime(const time_t * t)
+{
+	return asctime((const struct tm *)localtime(t));
+}
diff --git a/src/time/difftime.c b/src/time/difftime.c
new file mode 100644
index 0000000..80a18cc
--- /dev/null
+++ b/src/time/difftime.c
@@ -0,0 +1,6 @@
+#include <time.h>
+
+double difftime(time_t t1, time_t t0)
+{
+	return t1-t0;
+}
diff --git a/src/time/gmtime.c b/src/time/gmtime.c
new file mode 100644
index 0000000..53ff15c
--- /dev/null
+++ b/src/time/gmtime.c
@@ -0,0 +1,17 @@
+/*
+ * libc/time/gmtime.c
+ */
+
+#include <time.h>
+#include "time_impl.h"
+
+struct tm * gmtime(const time_t * t)
+{
+	static struct tm tm;
+	if(__secs_to_tm(*t, &tm) < 0)
+		return NULL;
+	tm.tm_isdst = 0;
+	tm.__tm_gmtoff = 0;
+	tm.__tm_zone = "UTC";
+	return &tm;
+}
diff --git a/src/time/localtime.c b/src/time/localtime.c
new file mode 100644
index 0000000..31d6e7d
--- /dev/null
+++ b/src/time/localtime.c
@@ -0,0 +1,11 @@
+/*
+ * libc/time/localtime.c
+ */
+
+#include <time.h>
+#include "time_impl.h"
+
+struct tm * localtime(const time_t * t)
+{
+	return gmtime(t);
+}
diff --git a/src/time/mktime.c b/src/time/mktime.c
new file mode 100644
index 0000000..cb2fc62
--- /dev/null
+++ b/src/time/mktime.c
@@ -0,0 +1,11 @@
+/*
+ * libc/time/mktime.c
+ */
+
+#include <time.h>
+#include "time_impl.h"
+
+time_t mktime(struct tm * tm)
+{
+	return __tm_to_secs(tm);
+}
diff --git a/src/time/strftime.c b/src/time/strftime.c
new file mode 100644
index 0000000..a07dbf5
--- /dev/null
+++ b/src/time/strftime.c
@@ -0,0 +1,224 @@
+/*
+ * libc/time/strftime.c
+ */
+
+#include <stdarg.h>
+#include <time.h>
+
+static char * aday[] = {
+    "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"
+};
+
+static char * day[] = {
+    "Sunday", "Monday", "Tuesday", "Wednesday",
+    "Thursday", "Friday", "Saturday"
+};
+
+static char * amonth[] = {
+    "Jan", "Feb", "Mar", "Apr", "May", "Jun",
+    "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
+};
+
+static char * month[] = {
+    "January", "February", "March", "April", "May", "June",
+    "July", "August", "September", "October", "November", "December"
+};
+
+static char buf[26];
+
+static int powers[5] = { 1, 10, 100, 1000, 10000 };
+
+static void strfmt(char * str, const char * fmt, ...)
+{
+	int ival, ilen;
+	char *sval;
+	va_list vp;
+
+	va_start(vp, fmt);
+	while (*fmt)
+	{
+		if (*fmt++ == '%')
+		{
+			ilen = *fmt++ - '0';
+			if (ilen == 0)
+			{
+				sval = va_arg(vp, char *);
+				while (*sval)
+					*str++ = *sval++;
+			}
+			else
+			{
+				ival = va_arg(vp, int);
+
+				while (ilen)
+				{
+					ival %= powers[ilen--];
+					*str++ = (char) ('0' + ival / powers[ilen]);
+				}
+			}
+		}
+		else
+			*str++ = fmt[-1];
+	}
+	*str = '\0';
+	va_end(vp);
+}
+
+size_t strftime(char * s, size_t max, const char * fmt, const struct tm * t)
+{
+	int w, d;
+	char *p, *q, *r;
+
+	p = s;
+	q = s + max - 1;
+	while ((*fmt != '\0'))
+	{
+		if (*fmt++ == '%')
+		{
+			r = buf;
+			switch (*fmt++)
+			{
+			case '%':
+				r = "%";
+				break;
+
+			case 'a':
+				r = aday[t->tm_wday];
+				break;
+
+			case 'A':
+				r = day[t->tm_wday];
+				break;
+
+			case 'b':
+				r = amonth[t->tm_mon];
+				break;
+
+			case 'B':
+				r = month[t->tm_mon];
+				break;
+
+			case 'c':
+				strfmt(r, "%0 %0 %2 %2:%2:%2 %4", aday[t->tm_wday],
+						amonth[t->tm_mon], t->tm_mday, t->tm_hour, t->tm_min,
+						t->tm_sec, t->tm_year + 1900);
+				break;
+
+			case 'd':
+				strfmt(r, "%2", t->tm_mday);
+				break;
+
+			case 'H':
+				strfmt(r, "%2", t->tm_hour);
+				break;
+
+			case 'I':
+				strfmt(r, "%2", (t->tm_hour % 12) ? t->tm_hour % 12 : 12);
+				break;
+
+			case 'j':
+				strfmt(r, "%3", t->tm_yday + 1);
+				break;
+
+			case 'm':
+				strfmt(r, "%2", t->tm_mon + 1);
+				break;
+
+			case 'M':
+				strfmt(r, "%2", t->tm_min);
+				break;
+
+			case 'p':
+				r = (t->tm_hour > 11) ? "PM" : "AM";
+				break;
+
+			case 'S':
+				strfmt(r, "%2", t->tm_sec);
+				break;
+
+			case 'U':
+				w = t->tm_yday / 7;
+				if (t->tm_yday % 7 > t->tm_wday)
+					w++;
+				strfmt(r, "%2", w);
+				break;
+
+			case 'W':
+				w = t->tm_yday / 7;
+				if (t->tm_yday % 7 > (t->tm_wday + 6) % 7)
+					w++;
+				strfmt(r, "%2", w);
+				break;
+
+			case 'V':
+				w = (t->tm_yday + 7 - (t->tm_wday ? t->tm_wday - 1 : 6)) / 7;
+				d = (t->tm_yday + 7 - (t->tm_wday ? t->tm_wday - 1 : 6)) % 7;
+
+				if (d >= 4)
+				{
+					w++;
+				}
+				else if (w == 0)
+				{
+					w = 53;
+				}
+				strfmt(r, "%2", w);
+				break;
+
+			case 'w':
+				strfmt(r, "%1", t->tm_wday);
+				break;
+
+			case 'x':
+				strfmt(r, "%3s %3s %2 %4", aday[t->tm_wday], amonth[t->tm_mon],
+						t->tm_mday, t->tm_year + 1900);
+				break;
+
+			case 'X':
+				strfmt(r, "%2:%2:%2", t->tm_hour, t->tm_min, t->tm_sec);
+				break;
+
+			case 'y':
+				strfmt(r, "%2", t->tm_year % 100);
+				break;
+
+			case 'Y':
+				strfmt(r, "%4", t->tm_year + 1900);
+				break;
+
+			case 'Z':
+				r = t->tm_isdst ? "DST" : "GMT";
+				break;
+
+			default:
+				buf[0] = '%';
+				buf[1] = fmt[-1];
+				buf[2] = '\0';
+				if (buf[1] == 0)
+					fmt--;
+				break;
+			}
+			while (*r)
+			{
+				if (p == q)
+				{
+					*q = '\0';
+					return 0;
+				}
+				*p++ = *r++;
+			}
+		}
+		else
+		{
+			if (p == q)
+			{
+				*q = '\0';
+				return 0;
+			}
+			*p++ = fmt[-1];
+		}
+	}
+
+	*p = '\0';
+	return p - s;
+}
diff --git a/src/time/time.c b/src/time/time.c
new file mode 100644
index 0000000..98bd1e5
--- /dev/null
+++ b/src/time/time.c
@@ -0,0 +1,29 @@
+/*
+ * libc/time/time.c
+ */
+
+#include <time.h>
+#include <nxbase.h>
+#include "time_impl.h"
+
+time_t time(time_t * t)
+{
+    time_t secs;
+    struct tm tm;
+    NX_Time nxtm;
+    NX_TimeGet(&nxtm);
+
+    tm.tm_year = nxtm.year - 1900; /* start from */
+    tm.tm_mon = nxtm.month - 1;
+    tm.tm_mday = nxtm.day;
+    tm.tm_hour = nxtm.hour;
+    tm.tm_min = nxtm.minute;
+    tm.tm_sec = nxtm.second;
+    
+	secs = __tm_to_secs(&tm);
+
+    if (t)
+        *t = secs;
+
+    return secs;
+}
diff --git a/src/time/time_impl.h b/src/time/time_impl.h
new file mode 100644
index 0000000..693c66f
--- /dev/null
+++ b/src/time/time_impl.h
@@ -0,0 +1,4 @@
+#include <time.h>
+
+int __secs_to_tm(time_t t, struct tm * tm);
+time_t __tm_to_secs(const struct tm * tm);
diff --git a/src/time/timespec_get.c b/src/time/timespec_get.c
new file mode 100644
index 0000000..0476877
--- /dev/null
+++ b/src/time/timespec_get.c
@@ -0,0 +1,18 @@
+#include <time.h>
+#include <nxbase.h>
+
+/* There is no other implemented value than TIME_UTC; all other values
+ * are considered erroneous. */
+int timespec_get(struct timespec * ts, int base)
+{
+    time_t tm;
+	if (base != TIME_UTC) return 0;
+
+    tm = time(&tm);
+    
+    if (ts) {
+        ts->tv_sec = tm;
+        ts->tv_nsec = NX_ClockGetMillisecond() * 1000UL;
+    }
+	return base;
+}
-- 
Gitee


From dd5934a1cc2c431f605d20eefe346cb60a6ecd6c Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Tue, 6 Sep 2022 01:29:35 +0800
Subject: [PATCH 07/13] feat: add weak sym for fenv

---
 src/fenv/fenv.c | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/src/fenv/fenv.c b/src/fenv/fenv.c
index 5588dad..dc310eb 100644
--- a/src/fenv/fenv.c
+++ b/src/fenv/fenv.c
@@ -1,38 +1,39 @@
 #include <fenv.h>
+#include <features.h>
 
 /* Dummy functions for archs lacking fenv implementation */
 
-int feclearexcept(int mask)
+weak_sym int feclearexcept(int mask)
 {
 	return 0;
 }
 
-int feraiseexcept(int mask)
+weak_sym int feraiseexcept(int mask)
 {
 	return 0;
 }
 
-int fetestexcept(int mask)
+weak_sym int fetestexcept(int mask)
 {
 	return 0;
 }
 
-int fegetround(void)
+weak_sym int fegetround(void)
 {
 	return FE_TONEAREST;
 }
 
-int __fesetround(int r)
+weak_sym int __fesetround(int r)
 {
 	return 0;
 }
 
-int fegetenv(fenv_t *envp)
+weak_sym int fegetenv(fenv_t *envp)
 {
 	return 0;
 }
 
-int fesetenv(const fenv_t *envp)
+weak_sym int fesetenv(const fenv_t *envp)
 {
 	return 0;
 }
-- 
Gitee


From 9df25a81b3fc9fe55247ff59df29be31317e1c81 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Wed, 7 Sep 2022 23:44:52 +0800
Subject: [PATCH 08/13] feat: update math and string

---
 doc/c99.md                  |  12 +-
 src/Makefile                |   1 +
 src/ctype/wcswidth.c        |   1 +
 src/env/__libc_start_main.c |  21 +-
 src/env/getenv.c            |   8 +-
 src/include/math.h          | 541 +++++++++++++++++--------------
 src/internal/libc.h         |   6 +-
 src/internal/libm.h         | 203 +++++++++++-
 src/internal/locale_impl.h  |  49 +++
 src/internal/wchar_impl.h   |  14 +
 src/locale/__mo_lookup.c    |  42 +++
 src/locale/c_locale.c       |  15 +
 src/locale/locale_map.c     | 107 +++++++
 src/locale/setlocale.c      |  73 ++++-
 src/math/__cosl.c           |  96 ++++++
 src/math/__expo2.c          |   7 +-
 src/math/__expo2f.c         |   7 +-
 src/math/__fpclassifyl.c    |  42 +++
 src/math/__math_divzero.c   |   6 +
 src/math/__math_divzerof.c  |   6 +
 src/math/__math_invalid.c   |   6 +
 src/math/__math_invalidf.c  |   6 +
 src/math/__math_oflow.c     |   6 +
 src/math/__math_oflowf.c    |   6 +
 src/math/__math_uflow.c     |   6 +
 src/math/__math_uflowf.c    |   6 +
 src/math/__math_xflow.c     |   6 +
 src/math/__math_xflowf.c    |   6 +
 src/math/__rem_pio2l.c      | 155 +++++++++
 src/math/__sinl.c           |  78 +++++
 src/math/__tan.c            |  12 +-
 src/math/__tandf.c          |   2 +-
 src/math/__tanl.c           | 143 +++++++++
 src/math/acos.c             |  16 +-
 src/math/acosf.c            |   3 +-
 src/math/acosh.c            |   3 +-
 src/math/acoshf.c           |   3 +-
 src/math/acoshl.c           |  29 ++
 src/math/acosl.c            |  67 ++++
 src/math/asin.c             |  16 +-
 src/math/asinf.c            |   3 +-
 src/math/asinh.c            |   3 +-
 src/math/asinhf.c           |   3 +-
 src/math/asinhl.c           |  41 +++
 src/math/asinl.c            |  71 +++++
 src/math/atan.c             |  16 +-
 src/math/atan2.c            |  17 +-
 src/math/atan2f.c           |   3 +-
 src/math/atanf.c            |   3 +-
 src/math/atanh.c            |   3 +-
 src/math/atanhf.c           |   3 +-
 src/math/atanhl.c           |  35 ++
 src/math/cbrt.c             |   3 +-
 src/math/cbrtf.c            |   3 +-
 src/math/cbrtl.c            | 124 ++++++++
 src/math/ceil.c             |   3 +-
 src/math/ceilf.c            |   3 +-
 src/math/ceill.c            |  34 ++
 src/math/copysign.c         |  17 +-
 src/math/copysignf.c        |  12 +-
 src/math/cos.c              |  14 +-
 src/math/cosf.c             |   3 +-
 src/math/cosh.c             |   5 +-
 src/math/coshf.c            |   5 +-
 src/math/coshl.c            |  47 +++
 src/math/cosl.c             |  39 +++
 src/math/erf.c              | 273 ++++++++++++++++
 src/math/erff.c             | 183 +++++++++++
 src/math/erfl.c             | 353 ++++++++++++++++++++
 src/math/exp.c              | 231 +++++++-------
 src/math/exp10.c            |  24 ++
 src/math/exp10f.c           |  22 ++
 src/math/exp10l.c           |  32 ++
 src/math/exp2.c             |   3 +-
 src/math/exp2f.c            |   3 +-
 src/math/exp2l.c            | 619 ++++++++++++++++++++++++++++++++++++
 src/math/exp_data.c         | 182 +++++++++++
 src/math/exp_data.h         |  26 ++
 src/math/expf.c             |   3 +-
 src/math/expl.c             | 128 ++++++++
 src/math/expm1.c            |  14 +-
 src/math/expm1f.c           |   3 +-
 src/math/expm1l.c           | 123 +++++++
 src/math/fabs.c             |   3 +-
 src/math/fabsf.c            |   3 +-
 src/math/fdim.c             |   3 +-
 src/math/fdimf.c            |   3 +-
 src/math/fdiml.c            |   2 +-
 src/math/finite.c           |   7 +
 src/math/finitef.c          |   7 +
 src/math/floor.c            |   3 +-
 src/math/floorf.c           |   3 +-
 src/math/fma.c              | 183 +++++++++++
 src/math/fmaf.c             |  93 ++++++
 src/math/fmal.c             | 293 +++++++++++++++++
 src/math/fmax.c             |   3 +-
 src/math/fmaxf.c            |   3 +-
 src/math/fmaxl.c            |   2 +-
 src/math/fmin.c             |   3 +-
 src/math/fminf.c            |   3 +-
 src/math/fminl.c            |   2 +-
 src/math/fmod.c             |   3 +-
 src/math/fmodf.c            |   3 +-
 src/math/frexp.c            |   3 +-
 src/math/frexpf.c           |   3 +-
 src/math/hypot.c            |   3 +-
 src/math/hypotf.c           |   3 +-
 src/math/ilogb.c            |  26 ++
 src/math/ilogbf.c           |  26 ++
 src/math/ilogbl.c           |  55 ++++
 src/math/ldexp.c            |   3 +-
 src/math/ldexpf.c           |   3 +-
 src/math/ldexpl.c           |   6 +
 src/math/log10.c            |   3 +-
 src/math/log10f.c           |   3 +-
 src/math/log10l.c           | 191 +++++++++++
 src/math/log1p.c            |  14 +-
 src/math/log1pf.c           |   3 +-
 src/math/log1pl.c           | 177 +++++++++++
 src/math/log2.c             |   3 +-
 src/math/log2_data.c        | 201 ++++++++++++
 src/math/log2_data.h        |  28 ++
 src/math/log2f.c            |   3 +-
 src/math/log2l.c            | 182 +++++++++++
 src/math/log_data.c         | 328 +++++++++++++++++++
 src/math/log_data.h         |  28 ++
 src/math/logb.c             |  17 +
 src/math/logbf.c            |  10 +
 src/math/logbl.c            |  16 +
 src/math/logf.c             |   3 +-
 src/math/logf_data.c        |  33 ++
 src/math/logf_data.h        |  20 ++
 src/math/lrint.c            |  72 +++++
 src/math/lrintf.c           |   8 +
 src/math/lrintl.c           |  36 +++
 src/math/lround.c           |   2 +-
 src/math/lroundf.c          |   2 +-
 src/math/lroundl.c          |   2 +-
 src/math/modf.c             |   3 +-
 src/math/modff.c            |   3 +-
 src/math/modfl.c            |  53 +++
 src/math/nan.c              |   6 +
 src/math/nanf.c             |   6 +
 src/math/nanl.c             |   6 +
 src/math/nearbyint.c        |  20 ++
 src/math/nearbyintf.c       |  18 ++
 src/math/nearbyintl.c       |  26 ++
 src/math/nextafter.c        |  31 ++
 src/math/nextafterf.c       |  30 ++
 src/math/nextafterl.c       |  75 +++++
 src/math/nexttoward.c       |  42 +++
 src/math/nexttowardf.c      |  35 ++
 src/math/nexttowardl.c      |   6 +
 src/math/pow.c              | 612 ++++++++++++++++++-----------------
 src/math/pow_data.c         | 180 +++++++++++
 src/math/pow_data.h         |  22 ++
 src/math/powf.c             |   3 +-
 src/math/remainder.c        |   9 +
 src/math/remainderf.c       |   9 +
 src/math/remainderl.c       |  15 +
 src/math/remquo.c           |  82 +++++
 src/math/remquof.c          |  82 +++++
 src/math/remquol.c          | 124 ++++++++
 src/math/rint.c             |   3 +-
 src/math/rintf.c            |   3 +-
 src/math/rintl.c            |  29 ++
 src/math/round.c            |   3 +-
 src/math/roundf.c           |   3 +-
 src/math/scalb.c            |  35 ++
 src/math/scalbf.c           |  32 ++
 src/math/scalbln.c          |   3 +-
 src/math/scalblnf.c         |   3 +-
 src/math/scalbn.c           |   3 +-
 src/math/scalbnf.c          |   3 +-
 src/math/signgam.c          |   6 +
 src/math/significand.c      |   7 +
 src/math/significandf.c     |   7 +
 src/math/sin.c              |   3 +-
 src/math/sincos.c           |  69 ++++
 src/math/sincosf.c          | 117 +++++++
 src/math/sincosl.c          |  60 ++++
 src/math/sinf.c             |   3 +-
 src/math/sinh.c             |   5 +-
 src/math/sinhf.c            |   5 +-
 src/math/sinhl.c            |  43 +++
 src/math/sinl.c             |  41 +++
 src/math/sqrt.c             |   3 +-
 src/math/sqrtf.c            |   3 +-
 src/math/tan.c              |   3 +-
 src/math/tanf.c             |   3 +-
 src/math/tanh.c             |   3 +-
 src/math/tanhf.c            |   3 +-
 src/math/tanhl.c            |  48 +++
 src/math/tanl.c             |  29 ++
 src/math/tgamma.c           | 222 +++++++++++++
 src/math/tgammaf.c          |   6 +
 src/math/tgammal.c          | 281 ++++++++++++++++
 src/math/trunc.c            |   4 +-
 src/math/truncf.c           |   3 +-
 src/math/truncl.c           |  34 ++
 src/multibyte/btowc.c       |  10 +
 src/multibyte/c16rtomb.c    |  35 ++
 src/multibyte/c32rtomb.c    |   7 +
 src/multibyte/internal.c    |  26 ++
 src/multibyte/internal.h    |  24 ++
 src/multibyte/mblen.c       |   6 +
 src/multibyte/mbrlen.c      |   7 +
 src/multibyte/mbrtowc.c     |  51 +++
 src/multibyte/mbsinit.c     |   6 +
 src/multibyte/mbsnrtowcs.c  |  55 ++++
 src/multibyte/mbsrtowcs.c   | 120 +++++++
 src/multibyte/mbstowcs.c    |   7 +
 src/multibyte/mbtowc.c      |  47 +++
 src/multibyte/wcsnrtombs.c  |  35 ++
 src/multibyte/wcsrtombs.c   |  55 ++++
 src/multibyte/wcstombs.c    |   7 +
 src/multibyte/wctob.c       |  11 +
 src/multibyte/wctomb.c      |   8 +
 src/stdlib/wcstod.c         |  64 ++++
 src/stdlib/wcstol.c         |  81 +++++
 src/string/wcpcpy.c         |   6 +
 src/string/wcpncpy.c        |   7 +
 src/string/wcscasecmp.c     |   8 +
 src/string/wcscasecmp_l.c   |   7 +
 src/string/wcscat.c         |   7 +
 src/string/wcscmp.c         |   7 +
 src/string/wcscpy.c         |   8 +
 src/string/wcscspn.c        |  10 +
 src/string/wcsdup.c         |  10 +
 src/string/wcsncasecmp.c    |   9 +
 src/string/wcsncasecmp_l.c  |   7 +
 src/string/wcsncat.c        |  10 +
 src/string/wcsncmp.c        |   7 +
 src/string/wcsncpy.c        |   9 +
 src/string/wcsnlen.c        |   8 +
 src/string/wcspbrk.c        |   7 +
 src/string/wcsspn.c         |   8 +
 src/string/wcsstr.c         | 105 ++++++
 src/string/wcstok.c         |  12 +
 src/string/wcswcs.c         |   6 +
 src/string/wmemchr.c        |   7 +
 src/string/wmemcmp.c        |   7 +
 src/string/wmemcpy.c        |   8 +
 src/string/wmemmove.c       |  13 +
 src/string/wmemset.c        |   8 +
 src/time/wcsftime.c         |  11 +
 246 files changed, 9462 insertions(+), 849 deletions(-)
 create mode 100644 src/internal/locale_impl.h
 create mode 100644 src/internal/wchar_impl.h
 create mode 100644 src/locale/__mo_lookup.c
 create mode 100644 src/locale/c_locale.c
 create mode 100644 src/locale/locale_map.c
 create mode 100644 src/math/__cosl.c
 create mode 100644 src/math/__fpclassifyl.c
 create mode 100644 src/math/__math_divzero.c
 create mode 100644 src/math/__math_divzerof.c
 create mode 100644 src/math/__math_invalid.c
 create mode 100644 src/math/__math_invalidf.c
 create mode 100644 src/math/__math_oflow.c
 create mode 100644 src/math/__math_oflowf.c
 create mode 100644 src/math/__math_uflow.c
 create mode 100644 src/math/__math_uflowf.c
 create mode 100644 src/math/__math_xflow.c
 create mode 100644 src/math/__math_xflowf.c
 create mode 100644 src/math/__rem_pio2l.c
 create mode 100644 src/math/__sinl.c
 create mode 100644 src/math/__tanl.c
 create mode 100644 src/math/acoshl.c
 create mode 100644 src/math/acosl.c
 create mode 100644 src/math/asinhl.c
 create mode 100644 src/math/asinl.c
 create mode 100644 src/math/atanhl.c
 create mode 100644 src/math/cbrtl.c
 create mode 100644 src/math/ceill.c
 create mode 100644 src/math/coshl.c
 create mode 100644 src/math/cosl.c
 create mode 100644 src/math/erf.c
 create mode 100644 src/math/erff.c
 create mode 100644 src/math/erfl.c
 create mode 100644 src/math/exp10.c
 create mode 100644 src/math/exp10f.c
 create mode 100644 src/math/exp10l.c
 create mode 100644 src/math/exp2l.c
 create mode 100644 src/math/exp_data.c
 create mode 100644 src/math/exp_data.h
 create mode 100644 src/math/expl.c
 create mode 100644 src/math/expm1l.c
 create mode 100644 src/math/finite.c
 create mode 100644 src/math/finitef.c
 create mode 100644 src/math/fma.c
 create mode 100644 src/math/fmaf.c
 create mode 100644 src/math/fmal.c
 create mode 100644 src/math/ilogb.c
 create mode 100644 src/math/ilogbf.c
 create mode 100644 src/math/ilogbl.c
 create mode 100644 src/math/ldexpl.c
 create mode 100644 src/math/log10l.c
 create mode 100644 src/math/log1pl.c
 create mode 100644 src/math/log2_data.c
 create mode 100644 src/math/log2_data.h
 create mode 100644 src/math/log2l.c
 create mode 100644 src/math/log_data.c
 create mode 100644 src/math/log_data.h
 create mode 100644 src/math/logb.c
 create mode 100644 src/math/logbf.c
 create mode 100644 src/math/logbl.c
 create mode 100644 src/math/logf_data.c
 create mode 100644 src/math/logf_data.h
 create mode 100644 src/math/lrint.c
 create mode 100644 src/math/lrintf.c
 create mode 100644 src/math/lrintl.c
 create mode 100644 src/math/modfl.c
 create mode 100644 src/math/nan.c
 create mode 100644 src/math/nanf.c
 create mode 100644 src/math/nanl.c
 create mode 100644 src/math/nearbyint.c
 create mode 100644 src/math/nearbyintf.c
 create mode 100644 src/math/nearbyintl.c
 create mode 100644 src/math/nextafter.c
 create mode 100644 src/math/nextafterf.c
 create mode 100644 src/math/nextafterl.c
 create mode 100644 src/math/nexttoward.c
 create mode 100644 src/math/nexttowardf.c
 create mode 100644 src/math/nexttowardl.c
 create mode 100644 src/math/pow_data.c
 create mode 100644 src/math/pow_data.h
 create mode 100644 src/math/remainder.c
 create mode 100644 src/math/remainderf.c
 create mode 100644 src/math/remainderl.c
 create mode 100644 src/math/remquo.c
 create mode 100644 src/math/remquof.c
 create mode 100644 src/math/remquol.c
 create mode 100644 src/math/rintl.c
 create mode 100644 src/math/scalb.c
 create mode 100644 src/math/scalbf.c
 create mode 100644 src/math/signgam.c
 create mode 100644 src/math/significand.c
 create mode 100644 src/math/significandf.c
 create mode 100644 src/math/sincos.c
 create mode 100644 src/math/sincosf.c
 create mode 100644 src/math/sincosl.c
 create mode 100644 src/math/sinhl.c
 create mode 100644 src/math/sinl.c
 create mode 100644 src/math/tanhl.c
 create mode 100644 src/math/tanl.c
 create mode 100644 src/math/tgamma.c
 create mode 100644 src/math/tgammaf.c
 create mode 100644 src/math/tgammal.c
 create mode 100644 src/math/truncl.c
 create mode 100644 src/multibyte/btowc.c
 create mode 100644 src/multibyte/c16rtomb.c
 create mode 100644 src/multibyte/c32rtomb.c
 create mode 100644 src/multibyte/internal.c
 create mode 100644 src/multibyte/internal.h
 create mode 100644 src/multibyte/mblen.c
 create mode 100644 src/multibyte/mbrlen.c
 create mode 100644 src/multibyte/mbrtowc.c
 create mode 100644 src/multibyte/mbsinit.c
 create mode 100644 src/multibyte/mbsnrtowcs.c
 create mode 100644 src/multibyte/mbsrtowcs.c
 create mode 100644 src/multibyte/mbstowcs.c
 create mode 100644 src/multibyte/mbtowc.c
 create mode 100644 src/multibyte/wcsnrtombs.c
 create mode 100644 src/multibyte/wcsrtombs.c
 create mode 100644 src/multibyte/wcstombs.c
 create mode 100644 src/multibyte/wctob.c
 create mode 100644 src/multibyte/wctomb.c
 create mode 100644 src/stdlib/wcstod.c
 create mode 100644 src/stdlib/wcstol.c
 create mode 100644 src/string/wcpcpy.c
 create mode 100644 src/string/wcpncpy.c
 create mode 100644 src/string/wcscasecmp.c
 create mode 100644 src/string/wcscasecmp_l.c
 create mode 100644 src/string/wcscat.c
 create mode 100644 src/string/wcscmp.c
 create mode 100644 src/string/wcscpy.c
 create mode 100644 src/string/wcscspn.c
 create mode 100644 src/string/wcsdup.c
 create mode 100644 src/string/wcsncasecmp.c
 create mode 100644 src/string/wcsncasecmp_l.c
 create mode 100644 src/string/wcsncat.c
 create mode 100644 src/string/wcsncmp.c
 create mode 100644 src/string/wcsncpy.c
 create mode 100644 src/string/wcsnlen.c
 create mode 100644 src/string/wcspbrk.c
 create mode 100644 src/string/wcsspn.c
 create mode 100644 src/string/wcsstr.c
 create mode 100644 src/string/wcstok.c
 create mode 100644 src/string/wcswcs.c
 create mode 100644 src/string/wmemchr.c
 create mode 100644 src/string/wmemcmp.c
 create mode 100644 src/string/wmemcpy.c
 create mode 100644 src/string/wmemmove.c
 create mode 100644 src/string/wmemset.c
 create mode 100644 src/time/wcsftime.c

diff --git a/doc/c99.md b/doc/c99.md
index 0ee71ad..ac1d6ca 100644
--- a/doc/c99.md
+++ b/doc/c99.md
@@ -33,9 +33,9 @@
 
 ## 比较复杂的头文件
 
-* stdlib.h
-* locale.h
-* math.h
-* time.h
-* wchar.h
-* wctype.h
\ No newline at end of file
+* stdlib.h 
+* locale.h ok
+* math.h ok
+* time.h ok
+* wchar.h ok
+* wctype.h ok
\ No newline at end of file
diff --git a/src/Makefile b/src/Makefile
index c9d82d9..e1ed3ef 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -43,6 +43,7 @@ X_LIBDIRS	:= $(LIBS_DIR)
 # we must link nxbase lib.
 X_LIBS		+= libnxbase.a
 
+SRC			+= multibyte/
 SRC			+= time/
 SRC			+= process/
 SRC			+= prng/
diff --git a/src/ctype/wcswidth.c b/src/ctype/wcswidth.c
index 5c8a5a4..6570b5b 100644
--- a/src/ctype/wcswidth.c
+++ b/src/ctype/wcswidth.c
@@ -1,4 +1,5 @@
 #include <wchar.h>
+#include "wchar_impl.h"
 
 int wcswidth(const wchar_t *wcs, size_t n)
 {
diff --git a/src/env/__libc_start_main.c b/src/env/__libc_start_main.c
index e583225..0c9b4ed 100644
--- a/src/env/__libc_start_main.c
+++ b/src/env/__libc_start_main.c
@@ -6,24 +6,33 @@
 
 #define MAX_ARGS 1024
 
-extern int main(int argc, char *argv[]);
+extern int main(int argc, char *argv[], char *envp[]);
 
-void __init_libc(char * envline)
+void __init_libc(char * envline, char ** envp)
 {
     libc.page_size = 4096;
     libc.envline = envline;
+    libc.environ = envp;
+	libc.secure = 1;
+    NX_MemSet(&libc.global_locale, 0, sizeof(libc.global_locale));
     __init_stdio();
 }
 
 NX_Error NX_WEAK_SYM NX_Main(char * cmdline, char * envline)
 {
     /* build cmdline and envline */
-    char *argv[MAX_ARGS + 1];
+    static char *argv[MAX_ARGS + 1] = {0,};
+    static char *envp[MAX_ARGS + 1] = {0,};
+
+    static char _envline[1024] = {0,};
+
     int argc;
 
+    NX_StrCopyN(_envline, envline, 1024);
+    NX_EnvToArray(_envline, envp, MAX_ARGS);
     argc = NX_CmdToArray(cmdline, argv, MAX_ARGS);
-
-    __init_libc(envline);
-    exit(main(argc, argv));
+    
+    __init_libc(envline, envp);
+    exit(main(argc, argv, envp));
     return NX_EOK;
 }
diff --git a/src/env/getenv.c b/src/env/getenv.c
index 13bd824..dc96d7a 100644
--- a/src/env/getenv.c
+++ b/src/env/getenv.c
@@ -1,8 +1,14 @@
 #include <stdlib.h>
 #include <string.h>
+#include "libc.h"
+#include "string_impl.h"
 
 char *getenv(const char *name)
 {
-    /* TODO: add env name */
+	size_t l = __strchrnul(name, '=') - name;
+	if (l && !name[l] && __environ)
+		for (char **e = __environ; *e; e++)
+			if (!strncmp(name, *e, l) && l[*e] == '=')
+				return *e + l+1;
 	return 0;
 }
diff --git a/src/include/math.h b/src/include/math.h
index 47b90b6..3327fb0 100644
--- a/src/include/math.h
+++ b/src/include/math.h
@@ -12,115 +12,46 @@ extern "C" {
 typedef float	float_t;
 typedef double	double_t;
 
-#define FORCE_EVAL(x) do {									\
-	if (sizeof(x) == sizeof(float)) {						\
-		volatile float __x __attribute__((unused));			\
-		__x = (x);											\
-	} else if (sizeof(x) == sizeof(double)) {				\
-		volatile double __x __attribute__((unused));		\
-		__x = (x);											\
-	} else {												\
-		volatile long double __x __attribute__((unused));	\
-		__x = (x);											\
-	}														\
-} while(0)
-
-/* Get two 32 bit ints from a double */
-#define EXTRACT_WORDS(hi, lo, d)							\
-do {														\
-	union {double f; uint64_t i;} __u;						\
-	__u.f = (d);											\
-	(hi) = __u.i >> 32;										\
-	(lo) = (uint32_t)__u.i;									\
-} while (0)
-
-/* Get the more significant 32 bit int from a double */
-#define GET_HIGH_WORD(hi, d)								\
-do {														\
-	union {double f; uint64_t i;} __u;						\
-	__u.f = (d);											\
-	(hi) = __u.i >> 32;										\
-} while (0)
-
-/* Get the less significant 32 bit int from a double */
-#define GET_LOW_WORD(lo, d)									\
-do {														\
-	union {double f; uint64_t i;} __u;						\
-	__u.f = (d);											\
-	(lo) = (uint32_t)__u.i;									\
-} while (0)
-
-/* Set a double from two 32 bit ints */
-#define INSERT_WORDS(d, hi, lo)								\
-do {														\
-	union {double f; uint64_t i;} __u;						\
-	__u.i = ((uint64_t)(hi)<<32) | (uint32_t)(lo);			\
-	(d) = __u.f;											\
-} while (0)
-
-/* Set the more significant 32 bits of a double from an int */
-#define SET_HIGH_WORD(d, hi)								\
-do {														\
-	union {double f; uint64_t i;} __u;						\
-	__u.f = (d);											\
-	__u.i &= 0xffffffff;									\
-	__u.i |= (uint64_t)(hi) << 32;							\
-	(d) = __u.f;											\
-} while (0)
-
-/* Set the less significant 32 bits of a double from an int */
-#define SET_LOW_WORD(d, lo)									\
-do {														\
-	union {double f; uint64_t i;} __u;						\
-	__u.f = (d);											\
-	__u.i &= 0xffffffff00000000ull;							\
-	__u.i |= (uint32_t)(lo);								\
-	(d) = __u.f;											\
-} while (0)
-
-/* Get a 32 bit int from a float */
-#define GET_FLOAT_WORD(w, d)								\
-do {														\
-	union {float f; uint32_t i;} __u;						\
-	__u.f = (d);											\
-	(w) = __u.i;											\
-} while (0)
-
-/* Set a float from a 32 bit int */
-#define SET_FLOAT_WORD(d, w)								\
-do {														\
-	union {float f; uint32_t i;} __u;						\
-	__u.i = (w);											\
-	(d) = __u.f;											\
-} while (0)
+#if 100*__GNUC__+__GNUC_MINOR__ >= 303
+#define NAN       __builtin_nanf("")
+#define INFINITY  __builtin_inff()
+#else
+#define NAN       (0.0f/0.0f)
+#define INFINITY  1e5000f
+#endif
 
-int __signbit(double);
-int __signbitf(float);
-int __signbitl(long double);
+#define HUGE_VALF INFINITY
+#define HUGE_VAL  ((double)INFINITY)
+#define HUGE_VALL ((long double)INFINITY)
 
-#define signbit(x) ( \
-	sizeof(x) == sizeof(float) ? (int)(__FLOAT_BITS(x)>>31) : \
-	sizeof(x) == sizeof(double) ? (int)(__DOUBLE_BITS(x)>>63) : \
-	__signbitl(x) )
+#define MATH_ERRNO  1
+#define MATH_ERREXCEPT 2
+#define math_errhandling 2
+
+#define FP_ILOGBNAN (-1-0x7fffffff)
+#define FP_ILOGB0 FP_ILOGBNAN
 
-#define FP_NAN					0
-#define FP_INFINITE				1
-#define FP_ZERO					2
-#define FP_SUBNORMAL			3
-#define FP_NORMAL				4
-
-#define NAN						__builtin_nan("")
-#define INFINITY				__builtin_inf()
-#define	HUGE_VALF				__builtin_huge_valf()
-#define	HUGE_VAL				__builtin_huge_val()
-#define	HUGE_VALL				__builtin_huge_vall()
-
-#define	isgreater(x, y)			__builtin_isgreater((x), (y))
-#define	isgreaterequal(x, y)	__builtin_isgreaterequal((x), (y))
-#define	isless(x, y)			__builtin_isless((x), (y))
-#define	islessequal(x, y)		__builtin_islessequal((x), (y))
-#define	islessgreater(x, y)		__builtin_islessgreater((x), (y))
-#define	isunordered(x, y)		__builtin_isunordered((x), (y))
+#define FP_NAN       0
+#define FP_INFINITE  1
+#define FP_ZERO      2
+#define FP_SUBNORMAL 3
+#define FP_NORMAL    4
+
+#ifdef __FP_FAST_FMA
+#define FP_FAST_FMA 1
+#endif
+
+#ifdef __FP_FAST_FMAF
+#define FP_FAST_FMAF 1
+#endif
+
+#ifdef __FP_FAST_FMAL
+#define FP_FAST_FMAL 1
+#endif
+
+int __fpclassify(double);
+int __fpclassifyf(float);
+int __fpclassifyl(long double);
 
 static __inline unsigned __FLOAT_BITS(float __f)
 {
@@ -128,7 +59,6 @@ static __inline unsigned __FLOAT_BITS(float __f)
 	__u.__f = __f;
 	return __u.__i;
 }
-
 static __inline unsigned long long __DOUBLE_BITS(double __f)
 {
 	union {double __f; unsigned long long __i;} __u;
@@ -139,174 +69,311 @@ static __inline unsigned long long __DOUBLE_BITS(double __f)
 #define fpclassify(x) ( \
 	sizeof(x) == sizeof(float) ? __fpclassifyf(x) : \
 	sizeof(x) == sizeof(double) ? __fpclassify(x) : \
-	__fpclassify(x) )
+	__fpclassifyl(x) )
 
 #define isinf(x) ( \
 	sizeof(x) == sizeof(float) ? (__FLOAT_BITS(x) & 0x7fffffff) == 0x7f800000 : \
 	sizeof(x) == sizeof(double) ? (__DOUBLE_BITS(x) & -1ULL>>1) == 0x7ffULL<<52 : \
-	__fpclassify(x) == FP_INFINITE)
+	__fpclassifyl(x) == FP_INFINITE)
 
 #define isnan(x) ( \
 	sizeof(x) == sizeof(float) ? (__FLOAT_BITS(x) & 0x7fffffff) > 0x7f800000 : \
 	sizeof(x) == sizeof(double) ? (__DOUBLE_BITS(x) & -1ULL>>1) > 0x7ffULL<<52 : \
-	__fpclassify(x) == FP_NAN)
+	__fpclassifyl(x) == FP_NAN)
 
 #define isnormal(x) ( \
 	sizeof(x) == sizeof(float) ? ((__FLOAT_BITS(x)+0x00800000) & 0x7fffffff) >= 0x01000000 : \
 	sizeof(x) == sizeof(double) ? ((__DOUBLE_BITS(x)+(1ULL<<52)) & -1ULL>>1) >= 1ULL<<53 : \
-	__fpclassify(x) == FP_NORMAL)
+	__fpclassifyl(x) == FP_NORMAL)
 
 #define isfinite(x) ( \
 	sizeof(x) == sizeof(float) ? (__FLOAT_BITS(x) & 0x7fffffff) < 0x7f800000 : \
 	sizeof(x) == sizeof(double) ? (__DOUBLE_BITS(x) & -1ULL>>1) < 0x7ffULL<<52 : \
-	__fpclassify(x) > FP_INFINITE)
-
-#define M_E						2.7182818284590452354	/* e */
-#define M_LOG2E					1.4426950408889634074	/* log_2 e */
-#define M_LOG10E				0.43429448190325182765	/* log_10 e */
-#define M_LN2					0.69314718055994530942	/* log_e 2 */
-#define M_LN10					2.30258509299404568402	/* log_e 10 */
-#define M_PI					3.14159265358979323846	/* pi */
-#define M_PI_2					1.57079632679489661923	/* pi/2 */
-#define M_PI_4					0.78539816339744830962	/* pi/4 */
-#define M_1_PI					0.31830988618379067154	/* 1/pi */
-#define M_2_PI					0.63661977236758134308	/* 2/pi */
-#define M_2_SQRTPI				1.12837916709551257390	/* 2/sqrt(pi) */
-#define M_SQRT2					1.41421356237309504880	/* sqrt(2) */
-#define M_SQRT1_2				0.70710678118654752440	/* 1/sqrt(2) */
-
-double	acos(double);
-float	acosf(float);
-double	acosh(double);
-float	acoshf(float);
-double	asin(double);
-float	asinf(float);
-double	asinh(double);
-float	asinhf(float);
-double	atan(double);
-float	atanf(float);
+	__fpclassifyl(x) > FP_INFINITE)
+
+int __signbit(double);
+int __signbitf(float);
+int __signbitl(long double);
+
+#define signbit(x) ( \
+	sizeof(x) == sizeof(float) ? (int)(__FLOAT_BITS(x)>>31) : \
+	sizeof(x) == sizeof(double) ? (int)(__DOUBLE_BITS(x)>>63) : \
+	__signbitl(x) )
+
+#define isunordered(x,y) (isnan((x)) ? ((void)(y),1) : isnan((y)))
+
+#define __ISREL_DEF(rel, op, type) \
+static __inline int __is##rel(type __x, type __y) \
+{ return !isunordered(__x,__y) && __x op __y; }
+
+__ISREL_DEF(lessf, <, float_t)
+__ISREL_DEF(less, <, double_t)
+__ISREL_DEF(lessl, <, long double)
+__ISREL_DEF(lessequalf, <=, float_t)
+__ISREL_DEF(lessequal, <=, double_t)
+__ISREL_DEF(lessequall, <=, long double)
+__ISREL_DEF(lessgreaterf, !=, float_t)
+__ISREL_DEF(lessgreater, !=, double_t)
+__ISREL_DEF(lessgreaterl, !=, long double)
+__ISREL_DEF(greaterf, >, float_t)
+__ISREL_DEF(greater, >, double_t)
+__ISREL_DEF(greaterl, >, long double)
+__ISREL_DEF(greaterequalf, >=, float_t)
+__ISREL_DEF(greaterequal, >=, double_t)
+__ISREL_DEF(greaterequall, >=, long double)
+
+#define __tg_pred_2(x, y, p) ( \
+	sizeof((x)+(y)) == sizeof(float) ? p##f(x, y) : \
+	sizeof((x)+(y)) == sizeof(double) ? p(x, y) : \
+	p##l(x, y) )
+
+#define isless(x, y)            __tg_pred_2(x, y, __isless)
+#define islessequal(x, y)       __tg_pred_2(x, y, __islessequal)
+#define islessgreater(x, y)     __tg_pred_2(x, y, __islessgreater)
+#define isgreater(x, y)         __tg_pred_2(x, y, __isgreater)
+#define isgreaterequal(x, y)    __tg_pred_2(x, y, __isgreaterequal)
+
+double      acos(double);
+float       acosf(float);
+long double acosl(long double);
+
+double      acosh(double);
+float       acoshf(float);
+long double acoshl(long double);
+
+double      asin(double);
+float       asinf(float);
+long double asinl(long double);
+
+double      asinh(double);
+float       asinhf(float);
+long double asinhl(long double);
+
+double      atan(double);
+float       atanf(float);
 long double atanl(long double);
 
-double	atan2(double, double);
-float	atan2f(float, float);
+double      atan2(double, double);
+float       atan2f(float, float);
 long double atan2l(long double, long double);
 
-double	atanh(double);
-float	atanhf(float);
+double      atanh(double);
+float       atanhf(float);
 long double atanhl(long double);
 
-double	cbrt(double);
-float	cbrtf(float);
-double	ceil(double);
-float	ceilf(float);
-double	cos(double);
-float	cosf(float);
-double	cosh(double);
-float	coshf(float);
-double	exp(double);
-float	expf(float);
-double	exp2(double);
-float	exp2f(float);
-double	expm1(double);
-float	expm1f(float);
-double	fabs(double);
-float	fabsf(float);
+double      cbrt(double);
+float       cbrtf(float);
+long double cbrtl(long double);
+
+double      ceil(double);
+float       ceilf(float);
+long double ceill(long double);
+
+double      copysign(double, double);
+float       copysignf(float, float);
+long double copysignl(long double, long double);
+
+double      cos(double);
+float       cosf(float);
+long double cosl(long double);
+
+double      cosh(double);
+float       coshf(float);
+long double coshl(long double);
+
+double      erf(double);
+float       erff(float);
+long double erfl(long double);
+
+double      erfc(double);
+float       erfcf(float);
+long double erfcl(long double);
+
+double      exp(double);
+float       expf(float);
+long double expl(long double);
+
+double      exp2(double);
+float       exp2f(float);
+long double exp2l(long double);
+
+double      expm1(double);
+float       expm1f(float);
+long double expm1l(long double);
+
+double      fabs(double);
+float       fabsf(float);
 long double fabsl(long double);
-double	fdim(double, double);
-float	fdimf(float, float);
+
+double      fdim(double, double);
+float       fdimf(float, float);
 long double fdiml(long double, long double);
 
-double	floor(double);
-float	floorf(float);
+double      floor(double);
+float       floorf(float);
 long double floorl(long double);
 
-double	fmod(double, double);
-float	fmodf(float, float);
+double      fma(double, double, double);
+float       fmaf(float, float, float);
+long double fmal(long double, long double, long double);
+
+double      fmax(double, double);
+float       fmaxf(float, float);
+long double fmaxl(long double, long double);
+
+double      fmin(double, double);
+float       fminf(float, float);
+long double fminl(long double, long double);
+
+double      fmod(double, double);
+float       fmodf(float, float);
 long double fmodl(long double, long double);
 
-double	frexp(double, int *);
-float	frexpf(float, int *);
+double      frexp(double, int *);
+float       frexpf(float, int *);
 long double frexpl(long double, int *);
 
-double	hypot(double, double);
-float	hypotf(float, float);
+double      hypot(double, double);
+float       hypotf(float, float);
 long double hypotl(long double, long double);
 
-double	ldexp(double, int);
-float	ldexpf(float, int);
-double	log(double);
-float	logf(float);
+int         ilogb(double);
+int         ilogbf(float);
+int         ilogbl(long double);
+
+double      ldexp(double, int);
+float       ldexpf(float, int);
+long double ldexpl(long double, int);
+
+double      lgamma(double);
+float       lgammaf(float);
+long double lgammal(long double);
+
+long long   llrint(double);
+long long   llrintf(float);
+long long   llrintl(long double);
+
+long long   llround(double);
+long long   llroundf(float);
+long long   llroundl(long double);
+
+double      log(double);
+float       logf(float);
 long double logl(long double);
 
-double	log10(double);
-float	log10f(float);
-double	log1p(double);
-float	log1pf(float);
-double	log2(double);
-float	log2f(float);
-long    lround(double);
-long    lroundf(float);
-long    lroundl(long double);
-double	modf(double, double *);
-float	modff(float, float *);
-double	pow(double, double);
-float	powf(float, float);
-long double powl(long double x, long double y);
-double	rint(double);
-float	rintf(float);
-double	round(double);
-float	roundf(float);
-long double roundl(long double x);
-double	scalbn(double, int);
-float	scalbnf(float, int);
-long double scalblnl(long double x, long n);
-
-double	scalbln(double, long);
-float	scalblnf(float, long);
-long double scalbnl(long double x, int n);
-
-double	sin(double);
-float	sinf(float);
-double	sinh(double);
-float	sinhf(float);
-double	sqrt(double);
-float	sqrtf(float);
-long double sqrtl(long double);
+double      log10(double);
+float       log10f(float);
+long double log10l(long double);
 
-double	tan(double);
-float	tanf(float);
-double	tanh(double);
-float	tanhf(float);
-double	trunc(double);
-float	truncf(float);
-double fmax(double, double);
-float fmaxf(float, float);
-long double fmaxl(long double, long double);
+double      log1p(double);
+float       log1pf(float);
+long double log1pl(long double);
 
-double fmin(double, double);
-float fminf(float, float);
-long double fminl(long double, long double);
+double      log2(double);
+float       log2f(float);
+long double log2l(long double);
+
+double      logb(double);
+float       logbf(float);
+long double logbl(long double);
+
+long        lrint(double);
+long        lrintf(float);
+long        lrintl(long double);
+
+long        lround(double);
+long        lroundf(float);
+long        lroundl(long double);
+
+double      modf(double, double *);
+float       modff(float, float *);
+long double modfl(long double, long double *);
+
+double      nan(const char *);
+float       nanf(const char *);
+long double nanl(const char *);
+
+double      nearbyint(double);
+float       nearbyintf(float);
+long double nearbyintl(long double);
+
+double      nextafter(double, double);
+float       nextafterf(float, float);
+long double nextafterl(long double, long double);
+
+double      nexttoward(double, long double);
+float       nexttowardf(float, long double);
+long double nexttowardl(long double, long double);
+
+double      pow(double, double);
+float       powf(float, float);
+long double powl(long double, long double);
+
+double      remainder(double, double);
+float       remainderf(float, float);
+long double remainderl(long double, long double);
+
+double      remquo(double, double, int *);
+float       remquof(float, float, int *);
+long double remquol(long double, long double, int *);
+
+double      rint(double);
+float       rintf(float);
+long double rintl(long double);
+
+double      round(double);
+float       roundf(float);
+long double roundl(long double);
+
+double      scalbln(double, long);
+float       scalblnf(float, long);
+long double scalblnl(long double, long);
+
+double      scalbn(double, int);
+float       scalbnf(float, int);
+long double scalbnl(long double, int);
+
+double      sin(double);
+float       sinf(float);
+long double sinl(long double);
+
+double      sinh(double);
+float       sinhf(float);
+long double sinhl(long double);
+
+double      sqrt(double);
+float       sqrtf(float);
+long double sqrtl(long double);
 
-float copysignf(float x, float y);
-double copysign(double x, double y);
-long double copysignl(long double x, long double y);
-
-/*
- * libm kernel functions
- */
-double	__cos(double, double);
-float	__cosdf(double);
-double	__expo2(double);
-float	__expo2f(float);
-int		__fpclassify(double);
-int		__fpclassifyf(float);
-int		__rem_pio2_large(double *, double *, int, int, int);
-int		__rem_pio2(double, double *);
-int		__rem_pio2f(float, double *);
-double	__sin(double, double, int);
-float	__sindf(double);
-double	__tan(double, double, int);
-float	__tandf(double, int);
+double      tan(double);
+float       tanf(float);
+long double tanl(long double);
+
+double      tanh(double);
+float       tanhf(float);
+long double tanhl(long double);
+
+double      tgamma(double);
+float       tgammaf(float);
+long double tgammal(long double);
+
+double      trunc(double);
+float       truncf(float);
+long double truncl(long double);
+
+#define M_E             2.7182818284590452354   /* e */
+#define M_LOG2E         1.4426950408889634074   /* log_2 e */
+#define M_LOG10E        0.43429448190325182765  /* log_10 e */
+#define M_LN2           0.69314718055994530942  /* log_e 2 */
+#define M_LN10          2.30258509299404568402  /* log_e 10 */
+#define M_PI            3.14159265358979323846  /* pi */
+#define M_PI_2          1.57079632679489661923  /* pi/2 */
+#define M_PI_4          0.78539816339744830962  /* pi/4 */
+#define M_1_PI          0.31830988618379067154  /* 1/pi */
+#define M_2_PI          0.63661977236758134308  /* 2/pi */
+#define M_2_SQRTPI      1.12837916709551257390  /* 2/sqrt(pi) */
+#define M_SQRT2         1.41421356237309504880  /* sqrt(2) */
+#define M_SQRT1_2       0.70710678118654752440  /* 1/sqrt(2) */
 
 #ifdef __cplusplus
 }
diff --git a/src/internal/libc.h b/src/internal/libc.h
index eb0be32..21adc89 100644
--- a/src/internal/libc.h
+++ b/src/internal/libc.h
@@ -15,8 +15,12 @@ struct __libc {
 	size_t page_size;
 	struct __locale_struct global_locale;
     char * envline; /* nxos envline */
+    char **environ; /* unix environ */
+	char secure;
 };
 
+#define __environ libc.environ
+
 #ifndef PAGE_SIZE
 #define PAGE_SIZE libc.page_size
 #endif
@@ -24,7 +28,7 @@ struct __libc {
 extern hidden struct __libc __libc;
 #define libc __libc
 
-void __init_libc(char *);
+void __init_libc(char *, char **);
 
 extern hidden const char __libc_version[];
 
diff --git a/src/internal/libm.h b/src/internal/libm.h
index cf56e50..42eb7fe 100644
--- a/src/internal/libm.h
+++ b/src/internal/libm.h
@@ -58,6 +58,34 @@ union ldshape {
 #error Unsupported long double representation
 #endif
 
+/* Support non-nearest rounding mode.  */
+#define WANT_ROUNDING 1
+/* Support signaling NaNs.  */
+#define WANT_SNAN 0
+
+#if WANT_SNAN
+#error SNaN is unsupported
+#else
+#define issignalingf_inline(x) 0
+#define issignaling_inline(x) 0
+#endif
+
+#ifndef TOINT_INTRINSICS
+#define TOINT_INTRINSICS 0
+#endif
+
+#if TOINT_INTRINSICS
+/* Round x to nearest int in all rounding modes, ties have to be rounded
+   consistently with converttoint so the results match.  If the result
+   would be outside of [-2^31, 2^31-1] then the semantics is unspecified.  */
+static double_t roundtoint(double_t);
+
+/* Convert x to nearest int in all rounding modes, ties have to be rounded
+   consistently with roundtoint.  If the result is not representible in an
+   int32_t then the semantics is unspecified.  */
+static int32_t converttoint(double_t);
+#endif
+
 /* Helps static branch prediction so hot path can be better optimized.  */
 #ifdef __GNUC__
 #define predict_true(x) __builtin_expect(!!(x), 1)
@@ -67,11 +95,182 @@ union ldshape {
 #define predict_false(x) (x)
 #endif
 
-#if LDBL_MANT_DIG != DBL_MANT_DIG
-hidden long double __math_invalidl(long double);
+/* Evaluate an expression as the specified type. With standard excess
+   precision handling a type cast or assignment is enough (with
+   -ffloat-store an assignment is required, in old compilers argument
+   passing and return statement may not drop excess precision).  */
+
+static inline float eval_as_float(float x)
+{
+	float y = x;
+	return y;
+}
+
+static inline double eval_as_double(double x)
+{
+	double y = x;
+	return y;
+}
+
+/* fp_barrier returns its input, but limits code transformations
+   as if it had a side-effect (e.g. observable io) and returned
+   an arbitrary value.  */
+
+#ifndef fp_barrierf
+#define fp_barrierf fp_barrierf
+static inline float fp_barrierf(float x)
+{
+	volatile float y = x;
+	return y;
+}
 #endif
 
+#ifndef fp_barrier
+#define fp_barrier fp_barrier
+static inline double fp_barrier(double x)
+{
+	volatile double y = x;
+	return y;
+}
+#endif
+
+#ifndef fp_barrierl
+#define fp_barrierl fp_barrierl
+static inline long double fp_barrierl(long double x)
+{
+	volatile long double y = x;
+	return y;
+}
+#endif
+
+/* fp_force_eval ensures that the input value is computed when that's
+   otherwise unused.  To prevent the constant folding of the input
+   expression, an additional fp_barrier may be needed or a compilation
+   mode that does so (e.g. -frounding-math in gcc). Then it can be
+   used to evaluate an expression for its fenv side-effects only.   */
+
+#ifndef fp_force_evalf
+#define fp_force_evalf fp_force_evalf
+static inline void fp_force_evalf(float x)
+{
+	volatile float y;
+	y = x;
+    x = y;
+}
+#endif
+
+#ifndef fp_force_eval
+#define fp_force_eval fp_force_eval
+static inline void fp_force_eval(double x)
+{
+	volatile double y;
+	y = x;
+    x = y;
+}
+#endif
+
+#ifndef fp_force_evall
+#define fp_force_evall fp_force_evall
+static inline void fp_force_evall(long double x)
+{
+	volatile long double y;
+	y = x;
+    x = y;
+}
+#endif
+
+#define FORCE_EVAL(x) do {                        \
+	if (sizeof(x) == sizeof(float)) {         \
+		fp_force_evalf(x);                \
+	} else if (sizeof(x) == sizeof(double)) { \
+		fp_force_eval(x);                 \
+	} else {                                  \
+		fp_force_evall(x);                \
+	}                                         \
+} while(0)
+
+#define asuint(f) ((union{float _f; uint32_t _i;}){f})._i
+#define asfloat(i) ((union{uint32_t _i; float _f;}){i})._f
+#define asuint64(f) ((union{double _f; uint64_t _i;}){f})._i
+#define asdouble(i) ((union{uint64_t _i; double _f;}){i})._f
+
+#define EXTRACT_WORDS(hi,lo,d)                    \
+do {                                              \
+  uint64_t __u = asuint64(d);                     \
+  (hi) = __u >> 32;                               \
+  (lo) = (uint32_t)__u;                           \
+} while (0)
+
+#define GET_HIGH_WORD(hi,d)                       \
+do {                                              \
+  (hi) = asuint64(d) >> 32;                       \
+} while (0)
+
+#define GET_LOW_WORD(lo,d)                        \
+do {                                              \
+  (lo) = (uint32_t)asuint64(d);                   \
+} while (0)
+
+#define INSERT_WORDS(d,hi,lo)                     \
+do {                                              \
+  (d) = asdouble(((uint64_t)(hi)<<32) | (uint32_t)(lo)); \
+} while (0)
+
+#define SET_HIGH_WORD(d,hi)                       \
+  INSERT_WORDS(d, hi, (uint32_t)asuint64(d))
+
+#define SET_LOW_WORD(d,lo)                        \
+  INSERT_WORDS(d, asuint64(d)>>32, lo)
+
+#define GET_FLOAT_WORD(w,d)                       \
+do {                                              \
+  (w) = asuint(d);                                \
+} while (0)
+
+#define SET_FLOAT_WORD(d,w)                       \
+do {                                              \
+  (d) = asfloat(w);                               \
+} while (0)
+
+hidden int    __rem_pio2_large(double*,double*,int,int,int);
+
+hidden int    __rem_pio2(double,double*);
+hidden double __sin(double,double,int);
+hidden double __cos(double,double);
+hidden double __tan(double,double,int);
+hidden double __expo2(double,double);
+
+hidden int    __rem_pio2f(float,double*);
+hidden float  __sindf(double);
+hidden float  __cosdf(double);
+hidden float  __tandf(double,int);
+hidden float  __expo2f(float,float);
+
+hidden int __rem_pio2l(long double, long double *);
+hidden long double __sinl(long double, long double, int);
+hidden long double __cosl(long double, long double);
+hidden long double __tanl(long double, long double, int);
+
 hidden long double __polevll(long double, const long double *, int);
 hidden long double __p1evll(long double, const long double *, int);
 
+extern int __signgam;
+hidden double __lgamma_r(double, int *);
+hidden float __lgammaf_r(float, int *);
+
+/* error handling functions */
+hidden float __math_xflowf(uint32_t, float);
+hidden float __math_uflowf(uint32_t);
+hidden float __math_oflowf(uint32_t);
+hidden float __math_divzerof(uint32_t);
+hidden float __math_invalidf(float);
+hidden double __math_xflow(uint32_t, double);
+hidden double __math_uflow(uint32_t);
+hidden double __math_oflow(uint32_t);
+hidden double __math_divzero(uint32_t);
+hidden double __math_invalid(double);
+#if LDBL_MANT_DIG != DBL_MANT_DIG
+hidden long double __math_invalidl(long double);
+#endif
+
 #endif
diff --git a/src/internal/locale_impl.h b/src/internal/locale_impl.h
new file mode 100644
index 0000000..0b26439
--- /dev/null
+++ b/src/internal/locale_impl.h
@@ -0,0 +1,49 @@
+#ifndef _LOCALE_IMPL_H
+#define _LOCALE_IMPL_H
+
+#include <locale.h>
+#include <stdlib.h>
+#include "libc.h"
+#include <nxbase.h>
+
+#define LOCALE_NAME_MAX 23
+
+struct __locale_map {
+	const void *map;
+	size_t map_size;
+	char name[LOCALE_NAME_MAX+1];
+	const struct __locale_map *next;
+};
+
+typedef struct __locale_struct * locale_t;
+
+extern hidden volatile int __locale_lock[1];
+
+extern hidden const struct __locale_map __c_dot_utf8;
+extern hidden const struct __locale_struct __c_locale;
+extern hidden const struct __locale_struct __c_dot_utf8_locale;
+
+hidden const struct __locale_map *__get_locale(int, const char *);
+hidden const char *__mo_lookup(const void *, size_t, const char *);
+hidden const char *__lctrans(const char *, const struct __locale_map *);
+hidden const char *__lctrans_cur(const char *);
+hidden const char *__lctrans_impl(const char *, const struct __locale_map *);
+hidden int __loc_is_allocated(locale_t);
+hidden char *__gettextdomain(void);
+
+#define LOC_MAP_FAILED ((const struct __locale_map *)-1)
+
+#define LCTRANS(msg, lc, loc) __lctrans(msg, (loc)->cat[(lc)])
+#define LCTRANS_CUR(msg) __lctrans_cur(msg)
+
+#define C_LOCALE ((locale_t)&__c_locale)
+#define UTF8_LOCALE ((locale_t)&__c_dot_utf8_locale)
+
+#define CURRENT_LOCALE (&libc.global_locale)
+
+#define CURRENT_UTF8 (!!libc.global_locale.cat[LC_CTYPE])
+
+#undef MB_CUR_MAX
+#define MB_CUR_MAX (CURRENT_UTF8 ? 4 : 1)
+
+#endif
diff --git a/src/internal/wchar_impl.h b/src/internal/wchar_impl.h
new file mode 100644
index 0000000..39ae985
--- /dev/null
+++ b/src/internal/wchar_impl.h
@@ -0,0 +1,14 @@
+#ifndef _WCHAR_IMPL_H
+#define _WCHAR_IMPL_H
+
+#include "locale_impl.h"
+
+int wcscasecmp(const wchar_t *, const wchar_t *);
+int wcscasecmp_l(const wchar_t *, const wchar_t *, locale_t);
+int wcsncasecmp(const wchar_t *, const wchar_t *, size_t);
+int wcsncasecmp_l(const wchar_t *, const wchar_t *, size_t, locale_t);
+
+int wcwidth(wchar_t wc);
+size_t wcsnlen (const wchar_t *, size_t);
+
+#endif
diff --git a/src/locale/__mo_lookup.c b/src/locale/__mo_lookup.c
new file mode 100644
index 0000000..4bc38af
--- /dev/null
+++ b/src/locale/__mo_lookup.c
@@ -0,0 +1,42 @@
+#include <stdint.h>
+#include <string.h>
+
+static inline uint32_t swapc(uint32_t x, int c)
+{
+	return c ? ((x>>24) | ((x>>8)&0xff00) | ((x<<8)&0xff0000) | (x<<24)) : x;
+}
+
+const char *__mo_lookup(const void *p, size_t size, const char *s)
+{
+	const uint32_t *mo = p;
+	int sw = *mo - 0x950412de;
+	uint32_t b = 0, n = swapc(mo[2], sw);
+	uint32_t o = swapc(mo[3], sw);
+	uint32_t t = swapc(mo[4], sw);
+	if (n>=size/4 || o>=size-4*n || t>=size-4*n || ((o|t)%4))
+		return 0;
+	o/=4;
+	t/=4;
+	for (;;) {
+		uint32_t ol = swapc(mo[o+2*(b+n/2)], sw);
+		uint32_t os = swapc(mo[o+2*(b+n/2)+1], sw);
+		if (os >= size || ol >= size-os || ((char *)p)[os+ol])
+			return 0;
+		int sign = strcmp(s, (char *)p + os);
+		if (!sign) {
+			uint32_t tl = swapc(mo[t+2*(b+n/2)], sw);
+			uint32_t ts = swapc(mo[t+2*(b+n/2)+1], sw);
+			if (ts >= size || tl >= size-ts || ((char *)p)[ts+tl])
+				return 0;
+			return (char *)p + ts;
+		}
+		else if (n == 1) return 0;
+		else if (sign < 0)
+			n /= 2;
+		else {
+			b += n/2;
+			n -= n/2;
+		}
+	}
+	return 0;
+}
diff --git a/src/locale/c_locale.c b/src/locale/c_locale.c
new file mode 100644
index 0000000..77ccf58
--- /dev/null
+++ b/src/locale/c_locale.c
@@ -0,0 +1,15 @@
+#include "locale_impl.h"
+#include <stdint.h>
+
+static const uint32_t empty_mo[] = { 0x950412de, 0, -1, -1, -1 };
+
+const struct __locale_map __c_dot_utf8 = {
+	.map = empty_mo,
+	.map_size = sizeof empty_mo,
+	.name = "C.UTF-8"
+};
+
+const struct __locale_struct __c_locale = { 0 };
+const struct __locale_struct __c_dot_utf8_locale = {
+	.cat[LC_CTYPE] = &__c_dot_utf8
+};
diff --git a/src/locale/locale_map.c b/src/locale/locale_map.c
new file mode 100644
index 0000000..c411668
--- /dev/null
+++ b/src/locale/locale_map.c
@@ -0,0 +1,107 @@
+#include <locale.h>
+#include <string.h>
+#include <stdlib.h>
+#include "locale_impl.h"
+#include "libc.h"
+#include "lock.h"
+#include "string_impl.h"
+
+const char *__lctrans_impl(const char *msg, const struct __locale_map *lm)
+{
+	const char *trans = 0;
+	if (lm) trans = __mo_lookup(lm->map, lm->map_size, msg);
+	return trans ? trans : msg;
+}
+
+static const char envvars[][12] = {
+	"LC_CTYPE",
+	"LC_NUMERIC",
+	"LC_TIME",
+	"LC_COLLATE",
+	"LC_MONETARY",
+	"LC_MESSAGES",
+};
+
+volatile int __locale_lock[1];
+volatile int *const __locale_lockptr = __locale_lock;
+
+const struct __locale_map *__get_locale(int cat, const char *val)
+{
+	static void *volatile loc_head;
+	const struct __locale_map *p;
+	struct __locale_map *new = 0;
+	size_t n;
+
+	if (!*val) {
+		((val = getenv("LC_ALL")) && *val) ||
+		((val = getenv(envvars[cat])) && *val) ||
+		((val = getenv("LANG")) && *val) ||
+		(val = "C.UTF-8");
+	}
+
+	/* Limit name length and forbid leading dot or any slashes. */
+	for (n=0; n<LOCALE_NAME_MAX && val[n] && val[n]!='/'; n++);
+	if (val[0]=='.' || val[n]) val = "C.UTF-8";
+	int builtin = (val[0]=='C' && !val[1])
+		|| !strcmp(val, "C.UTF-8")
+		|| !strcmp(val, "POSIX");
+
+	if (builtin) {
+		if (cat == LC_CTYPE && val[1]=='.')
+			return (void *)&__c_dot_utf8;
+		return 0;
+	}
+
+	for (p=loc_head; p; p=p->next)
+		if (!strcmp(val, p->name)) return p;
+
+#if 0
+	if (!libc.secure) path = getenv("MUSL_LOCPATH");
+	/* FIXME: add a default path? */
+
+	if (path) for (; *path; path=z+!!*z) {
+		z = __strchrnul(path, ':');
+		l = z - path;
+		if (l >= sizeof buf - n - 2) continue;
+		memcpy(buf, path, l);
+		buf[l] = '/';
+		memcpy(buf+l+1, val, n);
+		buf[l+1+n] = 0;
+		size_t map_size;
+		const void *map = __map_file(buf, &map_size);
+		if (map) {
+			new = malloc(sizeof *new);
+			if (!new) {
+				__munmap((void *)map, map_size);
+				break;
+			}
+			new->map = map;
+			new->map_size = map_size;
+			memcpy(new->name, val, n);
+			new->name[n] = 0;
+			new->next = loc_head;
+			loc_head = new;
+			break;
+		}
+	}
+#endif
+
+	/* If no locale definition was found, make a locale map
+	 * object anyway to store the name, which is kept for the
+	 * sake of being able to do message translations at the
+	 * application level. */
+	if (!new && (new = malloc(sizeof *new))) {
+		new->map = __c_dot_utf8.map;
+		new->map_size = __c_dot_utf8.map_size;
+		memcpy(new->name, val, n);
+		new->name[n] = 0;
+		new->next = loc_head;
+		loc_head = new;
+	}
+
+	/* For LC_CTYPE, never return a null pointer unless the
+	 * requested name was "C" or "POSIX". */
+	if (!new && cat == LC_CTYPE) new = (void *)&__c_dot_utf8;
+
+	return new;
+}
diff --git a/src/locale/setlocale.c b/src/locale/setlocale.c
index cb31b01..fc2d77b 100644
--- a/src/locale/setlocale.c
+++ b/src/locale/setlocale.c
@@ -1,8 +1,79 @@
 #include <locale.h>
 #include <stdlib.h>
 #include <string.h>
+#include "locale_impl.h"
+#include "libc.h"
+#include "lock.h"
+#include "string_impl.h"
+
+static char buf[LC_ALL*(LOCALE_NAME_MAX+1)];
 
 char *setlocale(int cat, const char *name)
 {
-	return "C.UTF-8";
+	const struct __locale_map *lm;
+
+	if ((unsigned)cat > LC_ALL) return 0;
+
+	LOCK(__locale_lock);
+
+	/* For LC_ALL, setlocale is required to return a string which
+	 * encodes the current setting for all categories. The format of
+	 * this string is unspecified, and only the following code, which
+	 * performs both the serialization and deserialization, depends
+	 * on the format, so it can easily be changed if needed. */
+	if (cat == LC_ALL) {
+		int i;
+		if (name) {
+			struct __locale_struct tmp_locale;
+			char part[LOCALE_NAME_MAX+1] = "C.UTF-8";
+			const char *p = name;
+			for (i=0; i<LC_ALL; i++) {
+				const char *z = __strchrnul(p, ';');
+				if (z-p <= LOCALE_NAME_MAX) {
+					memcpy(part, p, z-p);
+					part[z-p] = 0;
+					if (*z) p = z+1;
+				}
+				lm = __get_locale(i, part);
+				if (lm == LOC_MAP_FAILED) {
+					UNLOCK(__locale_lock);
+					return 0;
+				}
+				tmp_locale.cat[i] = lm;
+			}
+			libc.global_locale = tmp_locale;
+		}
+		char *s = buf;
+		const char *part;
+		int same = 0;
+		for (i=0; i<LC_ALL; i++) {
+			const struct __locale_map *lm =
+				libc.global_locale.cat[i];
+			if (lm == libc.global_locale.cat[0]) same++;
+			part = lm ? lm->name : "C";
+			size_t l = strlen(part);
+			memcpy(s, part, l);
+			s[l] = ';';
+			s += l+1;
+		}
+		*--s = 0;
+		UNLOCK(__locale_lock);
+		return same==LC_ALL ? (char *)part : buf;
+	}
+
+	if (name) {
+		lm = __get_locale(cat, name);
+		if (lm == LOC_MAP_FAILED) {
+			UNLOCK(__locale_lock);
+			return 0;
+		}
+		libc.global_locale.cat[cat] = lm;
+	} else {
+		lm = libc.global_locale.cat[cat];
+	}
+	char *ret = lm ? (char *)lm->name : "C";
+
+	UNLOCK(__locale_lock);
+
+	return ret;
 }
diff --git a/src/math/__cosl.c b/src/math/__cosl.c
new file mode 100644
index 0000000..fa522dd
--- /dev/null
+++ b/src/math/__cosl.c
@@ -0,0 +1,96 @@
+/* origin: FreeBSD /usr/src/lib/msun/ld80/k_cosl.c */
+/* origin: FreeBSD /usr/src/lib/msun/ld128/k_cosl.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ * Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+
+#include "libm.h"
+
+#if (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#if LDBL_MANT_DIG == 64
+/*
+ * ld80 version of __cos.c.  See __cos.c for most comments.
+ */
+/*
+ * Domain [-0.7854, 0.7854], range ~[-2.43e-23, 2.425e-23]:
+ * |cos(x) - c(x)| < 2**-75.1
+ *
+ * The coefficients of c(x) were generated by a pari-gp script using
+ * a Remez algorithm that searches for the best higher coefficients
+ * after rounding leading coefficients to a specified precision.
+ *
+ * Simpler methods like Chebyshev or basic Remez barely suffice for
+ * cos() in 64-bit precision, because we want the coefficient of x^2
+ * to be precisely -0.5 so that multiplying by it is exact, and plain
+ * rounding of the coefficients of a good polynomial approximation only
+ * gives this up to about 64-bit precision.  Plain rounding also gives
+ * a mediocre approximation for the coefficient of x^4, but a rounding
+ * error of 0.5 ulps for this coefficient would only contribute ~0.01
+ * ulps to the final error, so this is unimportant.  Rounding errors in
+ * higher coefficients are even less important.
+ *
+ * In fact, coefficients above the x^4 one only need to have 53-bit
+ * precision, and this is more efficient.  We get this optimization
+ * almost for free from the complications needed to search for the best
+ * higher coefficients.
+ */
+static const long double
+C1 =  0.0416666666666666666136L;        /*  0xaaaaaaaaaaaaaa9b.0p-68 */
+static const double
+C2 = -0.0013888888888888874,            /* -0x16c16c16c16c10.0p-62 */
+C3 =  0.000024801587301571716,          /*  0x1a01a01a018e22.0p-68 */
+C4 = -0.00000027557319215507120,        /* -0x127e4fb7602f22.0p-74 */
+C5 =  0.0000000020876754400407278,      /*  0x11eed8caaeccf1.0p-81 */
+C6 = -1.1470297442401303e-11,           /* -0x19393412bd1529.0p-89 */
+C7 =  4.7383039476436467e-14;           /*  0x1aac9d9af5c43e.0p-97 */
+#define POLY(z) (z*(C1+z*(C2+z*(C3+z*(C4+z*(C5+z*(C6+z*C7)))))))
+#elif LDBL_MANT_DIG == 113
+/*
+ * ld128 version of __cos.c.  See __cos.c for most comments.
+ */
+/*
+ * Domain [-0.7854, 0.7854], range ~[-1.80e-37, 1.79e-37]:
+ * |cos(x) - c(x))| < 2**-122.0
+ *
+ * 113-bit precision requires more care than 64-bit precision, since
+ * simple methods give a minimax polynomial with coefficient for x^2
+ * that is 1 ulp below 0.5, but we want it to be precisely 0.5.  See
+ * above for more details.
+ */
+static const long double
+C1 =  0.04166666666666666666666666666666658424671L,
+C2 = -0.001388888888888888888888888888863490893732L,
+C3 =  0.00002480158730158730158730158600795304914210L,
+C4 = -0.2755731922398589065255474947078934284324e-6L,
+C5 =  0.2087675698786809897659225313136400793948e-8L,
+C6 = -0.1147074559772972315817149986812031204775e-10L,
+C7 =  0.4779477332386808976875457937252120293400e-13L;
+static const double
+C8 = -0.1561920696721507929516718307820958119868e-15,
+C9 =  0.4110317413744594971475941557607804508039e-18,
+C10 = -0.8896592467191938803288521958313920156409e-21,
+C11 =  0.1601061435794535138244346256065192782581e-23;
+#define POLY(z) (z*(C1+z*(C2+z*(C3+z*(C4+z*(C5+z*(C6+z*(C7+ \
+	z*(C8+z*(C9+z*(C10+z*C11)))))))))))
+#endif
+
+long double __cosl(long double x, long double y)
+{
+	long double hz,z,r,w;
+
+	z  = x*x;
+	r  = POLY(z);
+	hz = 0.5*z;
+	w  = 1.0-hz;
+	return w + (((1.0-w)-hz) + (z*r-x*y));
+}
+#endif
diff --git a/src/math/__expo2.c b/src/math/__expo2.c
index 0e18b2e..248f052 100644
--- a/src/math/__expo2.c
+++ b/src/math/__expo2.c
@@ -1,16 +1,17 @@
-#include <math.h>
+#include "libm.h"
 
 /* k is such that k*ln2 has minimal relative error and x - kln2 > log(DBL_MIN) */
 static const int k = 2043;
 static const double kln2 = 0x1.62066151add8bp+10;
 
 /* exp(x)/2 for x >= log(DBL_MAX), slightly better than 0.5*exp(x/2)*exp(x/2) */
-double __expo2(double x)
+double __expo2(double x, double sign)
 {
 	double scale;
 
 	/* note that k is odd and scale*scale overflows */
 	INSERT_WORDS(scale, (uint32_t)(0x3ff + k/2) << 20, 0);
 	/* exp(x - k ln2) * 2**(k-1) */
-	return exp(x - kln2) * scale * scale;
+	/* in directed rounding correct sign before rounding or overflow is important */
+	return exp(x - kln2) * (sign * scale) * scale;
 }
diff --git a/src/math/__expo2f.c b/src/math/__expo2f.c
index 0057941..538eb09 100644
--- a/src/math/__expo2f.c
+++ b/src/math/__expo2f.c
@@ -1,16 +1,17 @@
-#include <math.h>
+#include "libm.h"
 
 /* k is such that k*ln2 has minimal relative error and x - kln2 > log(FLT_MIN) */
 static const int k = 235;
 static const float kln2 = 0x1.45c778p+7f;
 
 /* expf(x)/2 for x >= log(FLT_MAX), slightly better than 0.5f*expf(x/2)*expf(x/2) */
-float __expo2f(float x)
+float __expo2f(float x, float sign)
 {
 	float scale;
 
 	/* note that k is odd and scale*scale overflows */
 	SET_FLOAT_WORD(scale, (uint32_t)(0x7f + k/2) << 23);
 	/* exp(x - k ln2) * 2**(k-1) */
-	return expf(x - kln2) * scale * scale;
+	/* in directed rounding correct sign before rounding or overflow is important */
+	return expf(x - kln2) * (sign * scale) * scale;
 }
diff --git a/src/math/__fpclassifyl.c b/src/math/__fpclassifyl.c
new file mode 100644
index 0000000..e41781b
--- /dev/null
+++ b/src/math/__fpclassifyl.c
@@ -0,0 +1,42 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+int __fpclassifyl(long double x)
+{
+	return __fpclassify(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+int __fpclassifyl(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	int msb = u.i.m>>63;
+	if (!e && !msb)
+		return u.i.m ? FP_SUBNORMAL : FP_ZERO;
+	if (e == 0x7fff) {
+		/* The x86 variant of 80-bit extended precision only admits
+		 * one representation of each infinity, with the mantissa msb
+		 * necessarily set. The version with it clear is invalid/nan.
+		 * The m68k variant, however, allows either, and tooling uses
+		 * the version with it clear. */
+		if (__BYTE_ORDER == __LITTLE_ENDIAN && !msb)
+			return FP_NAN;
+		return u.i.m << 1 ? FP_NAN : FP_INFINITE;
+	}
+	if (!msb)
+		return FP_NAN;
+	return FP_NORMAL;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+int __fpclassifyl(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	u.i.se = 0;
+	if (!e)
+		return u.i2.lo | u.i2.hi ? FP_SUBNORMAL : FP_ZERO;
+	if (e == 0x7fff)
+		return u.i2.lo | u.i2.hi ? FP_NAN : FP_INFINITE;
+	return FP_NORMAL;
+}
+#endif
diff --git a/src/math/__math_divzero.c b/src/math/__math_divzero.c
new file mode 100644
index 0000000..59d2135
--- /dev/null
+++ b/src/math/__math_divzero.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+double __math_divzero(uint32_t sign)
+{
+	return fp_barrier(sign ? -1.0 : 1.0) / 0.0;
+}
diff --git a/src/math/__math_divzerof.c b/src/math/__math_divzerof.c
new file mode 100644
index 0000000..ce046f3
--- /dev/null
+++ b/src/math/__math_divzerof.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+float __math_divzerof(uint32_t sign)
+{
+	return fp_barrierf(sign ? -1.0f : 1.0f) / 0.0f;
+}
diff --git a/src/math/__math_invalid.c b/src/math/__math_invalid.c
new file mode 100644
index 0000000..1774049
--- /dev/null
+++ b/src/math/__math_invalid.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+double __math_invalid(double x)
+{
+	return (x - x) / (x - x);
+}
diff --git a/src/math/__math_invalidf.c b/src/math/__math_invalidf.c
new file mode 100644
index 0000000..357d4b1
--- /dev/null
+++ b/src/math/__math_invalidf.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+float __math_invalidf(float x)
+{
+	return (x - x) / (x - x);
+}
diff --git a/src/math/__math_oflow.c b/src/math/__math_oflow.c
new file mode 100644
index 0000000..c85dbf9
--- /dev/null
+++ b/src/math/__math_oflow.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+double __math_oflow(uint32_t sign)
+{
+	return __math_xflow(sign, 0x1p769);
+}
diff --git a/src/math/__math_oflowf.c b/src/math/__math_oflowf.c
new file mode 100644
index 0000000..fa7d062
--- /dev/null
+++ b/src/math/__math_oflowf.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+float __math_oflowf(uint32_t sign)
+{
+	return __math_xflowf(sign, 0x1p97f);
+}
diff --git a/src/math/__math_uflow.c b/src/math/__math_uflow.c
new file mode 100644
index 0000000..b90594a
--- /dev/null
+++ b/src/math/__math_uflow.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+double __math_uflow(uint32_t sign)
+{
+	return __math_xflow(sign, 0x1p-767);
+}
diff --git a/src/math/__math_uflowf.c b/src/math/__math_uflowf.c
new file mode 100644
index 0000000..94d50f2
--- /dev/null
+++ b/src/math/__math_uflowf.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+float __math_uflowf(uint32_t sign)
+{
+	return __math_xflowf(sign, 0x1p-95f);
+}
diff --git a/src/math/__math_xflow.c b/src/math/__math_xflow.c
new file mode 100644
index 0000000..744203c
--- /dev/null
+++ b/src/math/__math_xflow.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+double __math_xflow(uint32_t sign, double y)
+{
+	return eval_as_double(fp_barrier(sign ? -y : y) * y);
+}
diff --git a/src/math/__math_xflowf.c b/src/math/__math_xflowf.c
new file mode 100644
index 0000000..f2c8478
--- /dev/null
+++ b/src/math/__math_xflowf.c
@@ -0,0 +1,6 @@
+#include "libm.h"
+
+float __math_xflowf(uint32_t sign, float y)
+{
+	return eval_as_float(fp_barrierf(sign ? -y : y) * y);
+}
diff --git a/src/math/__rem_pio2l.c b/src/math/__rem_pio2l.c
new file mode 100644
index 0000000..236b2de
--- /dev/null
+++ b/src/math/__rem_pio2l.c
@@ -0,0 +1,155 @@
+/* origin: FreeBSD /usr/src/lib/msun/ld80/e_rem_pio2.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ * Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ *
+ * Optimized by Bruce D. Evans.
+ */
+#include "libm.h"
+#if (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+/* ld80 and ld128 version of __rem_pio2(x,y)
+ *
+ * return the remainder of x rem pi/2 in y[0]+y[1]
+ * use __rem_pio2_large() for large x
+ */
+
+static const long double toint = 1.5/LDBL_EPSILON;
+
+#if LDBL_MANT_DIG == 64
+/* u ~< 0x1p25*pi/2 */
+#define SMALL(u) (((u.i.se & 0x7fffU)<<16 | u.i.m>>48) < ((0x3fff + 25)<<16 | 0x921f>>1 | 0x8000))
+#define QUOBITS(x) ((uint32_t)(int32_t)x & 0x7fffffff)
+#define ROUND1 22
+#define ROUND2 61
+#define NX 3
+#define NY 2
+/*
+ * invpio2:  64 bits of 2/pi
+ * pio2_1:   first  39 bits of pi/2
+ * pio2_1t:  pi/2 - pio2_1
+ * pio2_2:   second 39 bits of pi/2
+ * pio2_2t:  pi/2 - (pio2_1+pio2_2)
+ * pio2_3:   third  39 bits of pi/2
+ * pio2_3t:  pi/2 - (pio2_1+pio2_2+pio2_3)
+ */
+static const double
+pio2_1 =  1.57079632679597125389e+00, /* 0x3FF921FB, 0x54444000 */
+pio2_2 = -1.07463465549783099519e-12, /* -0x12e7b967674000.0p-92 */
+pio2_3 =  6.36831716351370313614e-25; /*  0x18a2e037074000.0p-133 */
+static const long double
+pio4    =  0x1.921fb54442d1846ap-1L,
+invpio2 =  6.36619772367581343076e-01L, /*  0xa2f9836e4e44152a.0p-64 */
+pio2_1t = -1.07463465549719416346e-12L, /* -0x973dcb3b399d747f.0p-103 */
+pio2_2t =  6.36831716351095013979e-25L, /*  0xc51701b839a25205.0p-144 */
+pio2_3t = -2.75299651904407171810e-37L; /* -0xbb5bf6c7ddd660ce.0p-185 */
+#elif LDBL_MANT_DIG == 113
+/* u ~< 0x1p45*pi/2 */
+#define SMALL(u) (((u.i.se & 0x7fffU)<<16 | u.i.top) < ((0x3fff + 45)<<16 | 0x921f))
+#define QUOBITS(x) ((uint32_t)(int64_t)x & 0x7fffffff)
+#define ROUND1 51
+#define ROUND2 119
+#define NX 5
+#define NY 3
+static const long double
+pio4    =  0x1.921fb54442d18469898cc51701b8p-1L,
+invpio2 =  6.3661977236758134307553505349005747e-01L,	/*  0x145f306dc9c882a53f84eafa3ea6a.0p-113 */
+pio2_1  =  1.5707963267948966192292994253909555e+00L,	/*  0x1921fb54442d18469800000000000.0p-112 */
+pio2_1t =  2.0222662487959507323996846200947577e-21L,	/*  0x13198a2e03707344a4093822299f3.0p-181 */
+pio2_2  =  2.0222662487959507323994779168837751e-21L,	/*  0x13198a2e03707344a400000000000.0p-181 */
+pio2_2t =  2.0670321098263988236496903051604844e-43L,	/*  0x127044533e63a0105df531d89cd91.0p-254 */
+pio2_3  =  2.0670321098263988236499468110329591e-43L,	/*  0x127044533e63a0105e00000000000.0p-254 */
+pio2_3t = -2.5650587247459238361625433492959285e-65L;	/* -0x159c4ec64ddaeb5f78671cbfb2210.0p-327 */
+#endif
+
+int __rem_pio2l(long double x, long double *y)
+{
+	union ldshape u,uz;
+	long double z,w,t,r,fn;
+	double tx[NX],ty[NY];
+	int ex,ey,n,i;
+
+	u.f = x;
+	ex = u.i.se & 0x7fff;
+	if (SMALL(u)) {
+		/* rint(x/(pi/2)) */
+		fn = x*invpio2 + toint - toint;
+		n = QUOBITS(fn);
+		r = x-fn*pio2_1;
+		w = fn*pio2_1t;  /* 1st round good to 102/180 bits (ld80/ld128) */
+		/* Matters with directed rounding. */
+		if (predict_false(r - w < -pio4)) {
+			n--;
+			fn--;
+			r = x - fn*pio2_1;
+			w = fn*pio2_1t;
+		} else if (predict_false(r - w > pio4)) {
+			n++;
+			fn++;
+			r = x - fn*pio2_1;
+			w = fn*pio2_1t;
+		}
+		y[0] = r-w;
+		u.f = y[0];
+		ey = u.i.se & 0x7fff;
+		if (ex - ey > ROUND1) {  /* 2nd iteration needed, good to 141/248 (ld80/ld128) */
+			t = r;
+			w = fn*pio2_2;
+			r = t-w;
+			w = fn*pio2_2t-((t-r)-w);
+			y[0] = r-w;
+			u.f = y[0];
+			ey = u.i.se & 0x7fff;
+			if (ex - ey > ROUND2) {  /* 3rd iteration, good to 180/316 bits */
+				t = r; /* will cover all possible cases (not verified for ld128) */
+				w = fn*pio2_3;
+				r = t-w;
+				w = fn*pio2_3t-((t-r)-w);
+				y[0] = r-w;
+			}
+		}
+		y[1] = (r - y[0]) - w;
+		return n;
+	}
+	/*
+	 * all other (large) arguments
+	 */
+	if (ex == 0x7fff) {                /* x is inf or NaN */
+		y[0] = y[1] = x - x;
+		return 0;
+	}
+	/* set z = scalbn(|x|,-ilogb(x)+23) */
+	uz.f = x;
+	uz.i.se = 0x3fff + 23;
+	z = uz.f;
+	for (i=0; i < NX - 1; i++) {
+		tx[i] = (double)(int32_t)z;
+		z     = (z-tx[i])*0x1p24;
+	}
+	tx[i] = z;
+	while (tx[i] == 0)
+		i--;
+	n = __rem_pio2_large(tx, ty, ex-0x3fff-23, i+1, NY);
+	w = ty[1];
+	if (NY == 3)
+		w += ty[2];
+	r = ty[0] + w;
+	/* TODO: for ld128 this does not follow the recommendation of the
+	comments of __rem_pio2_large which seem wrong if |ty[0]| > |ty[1]+ty[2]| */
+	w -= r - ty[0];
+	if (u.i.se >> 15) {
+		y[0] = -r;
+		y[1] = -w;
+		return -n;
+	}
+	y[0] = r;
+	y[1] = w;
+	return n;
+}
+#endif
diff --git a/src/math/__sinl.c b/src/math/__sinl.c
new file mode 100644
index 0000000..2525bbe
--- /dev/null
+++ b/src/math/__sinl.c
@@ -0,0 +1,78 @@
+/* origin: FreeBSD /usr/src/lib/msun/ld80/k_sinl.c */
+/* origin: FreeBSD /usr/src/lib/msun/ld128/k_sinl.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ * Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+#include "libm.h"
+
+#if (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#if LDBL_MANT_DIG == 64
+/*
+ * ld80 version of __sin.c.  See __sin.c for most comments.
+ */
+/*
+ * Domain [-0.7854, 0.7854], range ~[-1.89e-22, 1.915e-22]
+ * |sin(x)/x - s(x)| < 2**-72.1
+ *
+ * See __cosl.c for more details about the polynomial.
+ */
+static const long double
+S1 = -0.166666666666666666671L;   /* -0xaaaaaaaaaaaaaaab.0p-66 */
+static const double
+S2 =  0.0083333333333333332,      /*  0x11111111111111.0p-59 */
+S3 = -0.00019841269841269427,     /* -0x1a01a01a019f81.0p-65 */
+S4 =  0.0000027557319223597490,   /*  0x171de3a55560f7.0p-71 */
+S5 = -0.000000025052108218074604, /* -0x1ae64564f16cad.0p-78 */
+S6 =  1.6059006598854211e-10,     /*  0x161242b90243b5.0p-85 */
+S7 = -7.6429779983024564e-13,     /* -0x1ae42ebd1b2e00.0p-93 */
+S8 =  2.6174587166648325e-15;     /*  0x179372ea0b3f64.0p-101 */
+#define POLY(z) (S2+z*(S3+z*(S4+z*(S5+z*(S6+z*(S7+z*S8))))))
+#elif LDBL_MANT_DIG == 113
+/*
+ * ld128 version of __sin.c.  See __sin.c for most comments.
+ */
+/*
+ * Domain [-0.7854, 0.7854], range ~[-1.53e-37, 1.659e-37]
+ * |sin(x)/x - s(x)| < 2**-122.1
+ *
+ * See __cosl.c for more details about the polynomial.
+ */
+static const long double
+S1 = -0.16666666666666666666666666666666666606732416116558L,
+S2 =  0.0083333333333333333333333333333331135404851288270047L,
+S3 = -0.00019841269841269841269841269839935785325638310428717L,
+S4 =  0.27557319223985890652557316053039946268333231205686e-5L,
+S5 = -0.25052108385441718775048214826384312253862930064745e-7L,
+S6 =  0.16059043836821614596571832194524392581082444805729e-9L,
+S7 = -0.76471637318198151807063387954939213287488216303768e-12L,
+S8 =  0.28114572543451292625024967174638477283187397621303e-14L;
+static const double
+S9  = -0.82206352458348947812512122163446202498005154296863e-17,
+S10 =  0.19572940011906109418080609928334380560135358385256e-19,
+S11 = -0.38680813379701966970673724299207480965452616911420e-22,
+S12 =  0.64038150078671872796678569586315881020659912139412e-25;
+#define POLY(z) (S2+z*(S3+z*(S4+z*(S5+z*(S6+z*(S7+z*(S8+ \
+	z*(S9+z*(S10+z*(S11+z*S12))))))))))
+#endif
+
+long double __sinl(long double x, long double y, int iy)
+{
+	long double z,r,v;
+
+	z = x*x;
+	v = z*x;
+	r = POLY(z);
+	if (iy == 0)
+		return x+v*(S1+z*r);
+	return x-((z*(0.5*y-v*r)-y)-v*S1);
+}
+#endif
diff --git a/src/math/__tan.c b/src/math/__tan.c
index e8d10d0..8019844 100644
--- a/src/math/__tan.c
+++ b/src/math/__tan.c
@@ -1,3 +1,13 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/k_tan.c */
+/*
+ * ====================================================
+ * Copyright 2004 Sun Microsystems, Inc.  All Rights Reserved.
+ *
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* __tan( x, y, k )
  * kernel tan function on ~[-pi/4, pi/4] (except on -0), pi/4 ~ 0.7854
  * Input x is assumed to be bounded by ~pi/4 in magnitude.
@@ -33,7 +43,7 @@
  *                     = 1 - 2*(tan(y) - (tan(y)^2)/(1+tan(y)))
  */
 
-#include <math.h>
+#include "libm.h"
 
 static const double T[] = {
              3.33333333333334091986e-01, /* 3FD55555, 55555563 */
diff --git a/src/math/__tandf.c b/src/math/__tandf.c
index f4da1ca..9ff577f 100644
--- a/src/math/__tandf.c
+++ b/src/math/__tandf.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 
 /* |tan(x)/x - t(x)| < 2**-25.5 (~[-2e-08, 2e-08]). */
 static const double T[] = {
diff --git a/src/math/__tanl.c b/src/math/__tanl.c
new file mode 100644
index 0000000..54abc3d
--- /dev/null
+++ b/src/math/__tanl.c
@@ -0,0 +1,143 @@
+/* origin: FreeBSD /usr/src/lib/msun/ld80/k_tanl.c */
+/* origin: FreeBSD /usr/src/lib/msun/ld128/k_tanl.c */
+/*
+ * ====================================================
+ * Copyright 2004 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
+ *
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+#include "libm.h"
+
+#if (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#if LDBL_MANT_DIG == 64
+/*
+ * ld80 version of __tan.c.  See __tan.c for most comments.
+ */
+/*
+ * Domain [-0.67434, 0.67434], range ~[-2.25e-22, 1.921e-22]
+ * |tan(x)/x - t(x)| < 2**-71.9
+ *
+ * See __cosl.c for more details about the polynomial.
+ */
+static const long double
+T3 =  0.333333333333333333180L,         /*  0xaaaaaaaaaaaaaaa5.0p-65 */
+T5 =  0.133333333333333372290L,         /*  0x88888888888893c3.0p-66 */
+T7 =  0.0539682539682504975744L,        /*  0xdd0dd0dd0dc13ba2.0p-68 */
+pio4   =  0.785398163397448309628L,     /*  0xc90fdaa22168c235.0p-64 */
+pio4lo = -1.25413940316708300586e-20L;  /* -0xece675d1fc8f8cbb.0p-130 */
+static const double
+T9  =  0.021869488536312216,            /*  0x1664f4882cc1c2.0p-58 */
+T11 =  0.0088632355256619590,           /*  0x1226e355c17612.0p-59 */
+T13 =  0.0035921281113786528,           /*  0x1d6d3d185d7ff8.0p-61 */
+T15 =  0.0014558334756312418,           /*  0x17da354aa3f96b.0p-62 */
+T17 =  0.00059003538700862256,          /*  0x13559358685b83.0p-63 */
+T19 =  0.00023907843576635544,          /*  0x1f56242026b5be.0p-65 */
+T21 =  0.000097154625656538905,         /*  0x1977efc26806f4.0p-66 */
+T23 =  0.000038440165747303162,         /*  0x14275a09b3ceac.0p-67 */
+T25 =  0.000018082171885432524,         /*  0x12f5e563e5487e.0p-68 */
+T27 =  0.0000024196006108814377,        /*  0x144c0d80cc6896.0p-71 */
+T29 =  0.0000078293456938132840,        /*  0x106b59141a6cb3.0p-69 */
+T31 = -0.0000032609076735050182,        /* -0x1b5abef3ba4b59.0p-71 */
+T33 =  0.0000023261313142559411;        /*  0x13835436c0c87f.0p-71 */
+#define RPOLY(w) (T5 + w * (T9 + w * (T13 + w * (T17 + w * (T21 + \
+	w * (T25 + w * (T29 + w * T33)))))))
+#define VPOLY(w) (T7 + w * (T11 + w * (T15 + w * (T19 + w * (T23 + \
+	w * (T27 + w * T31))))))
+#elif LDBL_MANT_DIG == 113
+/*
+ * ld128 version of __tan.c.  See __tan.c for most comments.
+ */
+/*
+ * Domain [-0.67434, 0.67434], range ~[-3.37e-36, 1.982e-37]
+ * |tan(x)/x - t(x)| < 2**-117.8 (XXX should be ~1e-37)
+ *
+ * See __cosl.c for more details about the polynomial.
+ */
+static const long double
+T3 = 0x1.5555555555555555555555555553p-2L,
+T5 = 0x1.1111111111111111111111111eb5p-3L,
+T7 = 0x1.ba1ba1ba1ba1ba1ba1ba1b694cd6p-5L,
+T9 = 0x1.664f4882c10f9f32d6bbe09d8bcdp-6L,
+T11 = 0x1.226e355e6c23c8f5b4f5762322eep-7L,
+T13 = 0x1.d6d3d0e157ddfb5fed8e84e27b37p-9L,
+T15 = 0x1.7da36452b75e2b5fce9ee7c2c92ep-10L,
+T17 = 0x1.355824803674477dfcf726649efep-11L,
+T19 = 0x1.f57d7734d1656e0aceb716f614c2p-13L,
+T21 = 0x1.967e18afcb180ed942dfdc518d6cp-14L,
+T23 = 0x1.497d8eea21e95bc7e2aa79b9f2cdp-15L,
+T25 = 0x1.0b132d39f055c81be49eff7afd50p-16L,
+T27 = 0x1.b0f72d33eff7bfa2fbc1059d90b6p-18L,
+T29 = 0x1.5ef2daf21d1113df38d0fbc00267p-19L,
+T31 = 0x1.1c77d6eac0234988cdaa04c96626p-20L,
+T33 = 0x1.cd2a5a292b180e0bdd701057dfe3p-22L,
+T35 = 0x1.75c7357d0298c01a31d0a6f7d518p-23L,
+T37 = 0x1.2f3190f4718a9a520f98f50081fcp-24L,
+pio4 = 0x1.921fb54442d18469898cc51701b8p-1L,
+pio4lo = 0x1.cd129024e088a67cc74020bbea60p-116L;
+static const double
+T39 =  0.000000028443389121318352,	/*  0x1e8a7592977938.0p-78 */
+T41 =  0.000000011981013102001973,	/*  0x19baa1b1223219.0p-79 */
+T43 =  0.0000000038303578044958070,	/*  0x107385dfb24529.0p-80 */
+T45 =  0.0000000034664378216909893,	/*  0x1dc6c702a05262.0p-81 */
+T47 = -0.0000000015090641701997785,	/* -0x19ecef3569ebb6.0p-82 */
+T49 =  0.0000000029449552300483952,	/*  0x194c0668da786a.0p-81 */
+T51 = -0.0000000022006995706097711,	/* -0x12e763b8845268.0p-81 */
+T53 =  0.0000000015468200913196612,	/*  0x1a92fc98c29554.0p-82 */
+T55 = -0.00000000061311613386849674,	/* -0x151106cbc779a9.0p-83 */
+T57 =  1.4912469681508012e-10;		/*  0x147edbdba6f43a.0p-85 */
+#define RPOLY(w) (T5 + w * (T9 + w * (T13 + w * (T17 + w * (T21 + \
+	w * (T25 + w * (T29 + w * (T33 + w * (T37 + w * (T41 + \
+	w * (T45 + w * (T49 + w * (T53 + w * T57)))))))))))))
+#define VPOLY(w) (T7 + w * (T11 + w * (T15 + w * (T19 + w * (T23 + \
+	w * (T27 + w * (T31 + w * (T35 + w * (T39 + w * (T43 + \
+	w * (T47 + w * (T51 + w * T55))))))))))))
+#endif
+
+long double __tanl(long double x, long double y, int odd) {
+	long double z, r, v, w, s, a, t;
+	int big, sign;
+
+	big = fabsl(x) >= 0.67434;
+	if (big) {
+		sign = 0;
+		if (x < 0) {
+			sign = 1;
+			x = -x;
+			y = -y;
+		}
+		x = (pio4 - x) + (pio4lo - y);
+		y = 0.0;
+	}
+	z = x * x;
+	w = z * z;
+	r = RPOLY(w);
+	v = z * VPOLY(w);
+	s = z * x;
+	r = y + z * (s * (r + v) + y) + T3 * s;
+	w = x + r;
+	if (big) {
+		s = 1 - 2*odd;
+		v = s - 2.0 * (x + (r - w * w / (w + s)));
+		return sign ? -v : v;
+	}
+	if (!odd)
+		return w;
+	/*
+	 * if allow error up to 2 ulp, simply return
+	 * -1.0 / (x+r) here
+	 */
+	/* compute -1.0 / (x+r) accurately */
+	z = w;
+	z = z + 0x1p32 - 0x1p32;
+	v = r - (z - x);        /* z+v = r+x */
+	t = a = -1.0 / w;       /* a = -1.0/w */
+	t = t + 0x1p32 - 0x1p32;
+	s = 1.0 + t * z;
+	return t + a * (s + t * v);
+}
+#endif
diff --git a/src/math/acos.c b/src/math/acos.c
index fbc2579..ea9c87b 100644
--- a/src/math/acos.c
+++ b/src/math/acos.c
@@ -1,3 +1,14 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_acos.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* acos(x)
  * Method :
  *      acos(x)  = pi/2 - asin(x)
@@ -22,8 +33,7 @@
  * Function needed: sqrt
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 pio2_hi = 1.57079632679489655800e+00, /* 0x3FF921FB, 0x54442D18 */
@@ -59,7 +69,7 @@ double acos(double x)
 		uint32_t lx;
 
 		GET_LOW_WORD(lx,x);
-		if (((ix-0x3ff00000) | lx) == 0) {
+		if ((ix-0x3ff00000 | lx) == 0) {
 			/* acos(1)=0, acos(-1)=pi */
 			if (hx >> 31)
 				return 2*pio2_hi + 0x1p-120f;
diff --git a/src/math/acosf.c b/src/math/acosf.c
index 34cf814..9532947 100644
--- a/src/math/acosf.c
+++ b/src/math/acosf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 pio2_hi = 1.5707962513e+00, /* 0x3fc90fda */
diff --git a/src/math/acosh.c b/src/math/acosh.c
index fdca32a..4ce9b3d 100644
--- a/src/math/acosh.c
+++ b/src/math/acosh.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* acosh(x) = log(x + sqrt(x*x-1)) */
 double acosh(double x)
diff --git a/src/math/acoshf.c b/src/math/acoshf.c
index 2c0652c..16550f1 100644
--- a/src/math/acoshf.c
+++ b/src/math/acoshf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* acosh(x) = log(x + sqrt(x*x-1)) */
 float acoshf(float x)
diff --git a/src/math/acoshl.c b/src/math/acoshl.c
new file mode 100644
index 0000000..8d4b43f
--- /dev/null
+++ b/src/math/acoshl.c
@@ -0,0 +1,29 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double acoshl(long double x)
+{
+	return acosh(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/* acosh(x) = log(x + sqrt(x*x-1)) */
+long double acoshl(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+
+	if (e < 0x3fff + 1)
+		/* |x| < 2, invalid if x < 1 or nan */
+		return log1pl(x-1 + sqrtl((x-1)*(x-1)+2*(x-1)));
+	if (e < 0x3fff + 32)
+		/* |x| < 0x1p32 */
+		return logl(2*x - 1/(x+sqrtl(x*x-1)));
+	return logl(x) + 0.693147180559945309417232121458176568L;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double acoshl(long double x)
+{
+	return acosh(x);
+}
+#endif
diff --git a/src/math/acosl.c b/src/math/acosl.c
new file mode 100644
index 0000000..c03bdf0
--- /dev/null
+++ b/src/math/acosl.c
@@ -0,0 +1,67 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_acosl.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+/*
+ * See comments in acos.c.
+ * Converted to long double by David Schultz <das@FreeBSD.ORG>.
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double acosl(long double x)
+{
+	return acos(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#include "__invtrigl.h"
+#if LDBL_MANT_DIG == 64
+#define CLEARBOTTOM(u) (u.i.m &= -1ULL << 32)
+#elif LDBL_MANT_DIG == 113
+#define CLEARBOTTOM(u) (u.i.lo = 0)
+#endif
+
+long double acosl(long double x)
+{
+	union ldshape u = {x};
+	long double z, s, c, f;
+	uint16_t e = u.i.se & 0x7fff;
+
+	/* |x| >= 1 or nan */
+	if (e >= 0x3fff) {
+		if (x == 1)
+			return 0;
+		if (x == -1)
+			return 2*pio2_hi + 0x1p-120f;
+		return 0/(x-x);
+	}
+	/* |x| < 0.5 */
+	if (e < 0x3fff - 1) {
+		if (e < 0x3fff - LDBL_MANT_DIG - 1)
+			return pio2_hi + 0x1p-120f;
+		return pio2_hi - (__invtrigl_R(x*x)*x - pio2_lo + x);
+	}
+	/* x < -0.5 */
+	if (u.i.se >> 15) {
+		z = (1 + x)*0.5;
+		s = sqrtl(z);
+		return 2*(pio2_hi - (__invtrigl_R(z)*s - pio2_lo + s));
+	}
+	/* x > 0.5 */
+	z = (1 - x)*0.5;
+	s = sqrtl(z);
+	u.f = s;
+	CLEARBOTTOM(u);
+	f = u.f;
+	c = (z - f*f)/(s + f);
+	return 2*(__invtrigl_R(z)*s + c + f);
+}
+#endif
diff --git a/src/math/asin.c b/src/math/asin.c
index f8a4b5d..c926b18 100644
--- a/src/math/asin.c
+++ b/src/math/asin.c
@@ -1,3 +1,14 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_asin.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* asin(x)
  * Method :
  *      Since  asin(x) = x + x^3/6 + x^5*3/40 + x^7*15/336 + ...
@@ -28,8 +39,7 @@
  *
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 pio2_hi = 1.57079632679489655800e+00, /* 0x3FF921FB, 0x54442D18 */
@@ -65,7 +75,7 @@ double asin(double x)
 	if (ix >= 0x3ff00000) {
 		uint32_t lx;
 		GET_LOW_WORD(lx, x);
-		if (((ix-0x3ff00000) | lx) == 0)
+		if ((ix-0x3ff00000 | lx) == 0)
 			/* asin(1) = +-pi/2 with inexact */
 			return x*pio2_hi + 0x1p-120f;
 		return 0/(x-x);
diff --git a/src/math/asinf.c b/src/math/asinf.c
index ddf056a..c103d49 100644
--- a/src/math/asinf.c
+++ b/src/math/asinf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 pio2 = 1.570796326794896558e+00;
diff --git a/src/math/asinh.c b/src/math/asinh.c
index 80543ee..0829f22 100644
--- a/src/math/asinh.c
+++ b/src/math/asinh.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* asinh(x) = sign(x)*log(|x|+sqrt(x*x+1)) ~= x - x^3/6 + o(x^5) */
 double asinh(double x)
diff --git a/src/math/asinhf.c b/src/math/asinhf.c
index 0cd7289..fc9f091 100644
--- a/src/math/asinhf.c
+++ b/src/math/asinhf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* asinh(x) = sign(x)*log(|x|+sqrt(x*x+1)) ~= x - x^3/6 + o(x^5) */
 float asinhf(float x)
diff --git a/src/math/asinhl.c b/src/math/asinhl.c
new file mode 100644
index 0000000..8635f52
--- /dev/null
+++ b/src/math/asinhl.c
@@ -0,0 +1,41 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double asinhl(long double x)
+{
+	return asinh(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/* asinh(x) = sign(x)*log(|x|+sqrt(x*x+1)) ~= x - x^3/6 + o(x^5) */
+long double asinhl(long double x)
+{
+	union ldshape u = {x};
+	unsigned e = u.i.se & 0x7fff;
+	unsigned s = u.i.se >> 15;
+
+	/* |x| */
+	u.i.se = e;
+	x = u.f;
+
+	if (e >= 0x3fff + 32) {
+		/* |x| >= 0x1p32 or inf or nan */
+		x = logl(x) + 0.693147180559945309417232121458176568L;
+	} else if (e >= 0x3fff + 1) {
+		/* |x| >= 2 */
+		x = logl(2*x + 1/(sqrtl(x*x+1)+x));
+	} else if (e >= 0x3fff - 32) {
+		/* |x| >= 0x1p-32 */
+		x = log1pl(x + x*x/(sqrtl(x*x+1)+1));
+	} else {
+		/* |x| < 0x1p-32, raise inexact if x!=0 */
+		FORCE_EVAL(x + 0x1p120f);
+	}
+	return s ? -x : x;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double asinhl(long double x)
+{
+	return asinh(x);
+}
+#endif
diff --git a/src/math/asinl.c b/src/math/asinl.c
new file mode 100644
index 0000000..347c535
--- /dev/null
+++ b/src/math/asinl.c
@@ -0,0 +1,71 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_asinl.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+/*
+ * See comments in asin.c.
+ * Converted to long double by David Schultz <das@FreeBSD.ORG>.
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double asinl(long double x)
+{
+	return asin(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#include "__invtrigl.h"
+#if LDBL_MANT_DIG == 64
+#define CLOSETO1(u) (u.i.m>>56 >= 0xf7)
+#define CLEARBOTTOM(u) (u.i.m &= -1ULL << 32)
+#elif LDBL_MANT_DIG == 113
+#define CLOSETO1(u) (u.i.top >= 0xee00)
+#define CLEARBOTTOM(u) (u.i.lo = 0)
+#endif
+
+long double asinl(long double x)
+{
+	union ldshape u = {x};
+	long double z, r, s;
+	uint16_t e = u.i.se & 0x7fff;
+	int sign = u.i.se >> 15;
+
+	if (e >= 0x3fff) {   /* |x| >= 1 or nan */
+		/* asin(+-1)=+-pi/2 with inexact */
+		if (x == 1 || x == -1)
+			return x*pio2_hi + 0x1p-120f;
+		return 0/(x-x);
+	}
+	if (e < 0x3fff - 1) {  /* |x| < 0.5 */
+		if (e < 0x3fff - (LDBL_MANT_DIG+1)/2) {
+			/* return x with inexact if x!=0 */
+			FORCE_EVAL(x + 0x1p120f);
+			return x;
+		}
+		return x + x*__invtrigl_R(x*x);
+	}
+	/* 1 > |x| >= 0.5 */
+	z = (1.0 - fabsl(x))*0.5;
+	s = sqrtl(z);
+	r = __invtrigl_R(z);
+	if (CLOSETO1(u)) {
+		x = pio2_hi - (2*(s+s*r)-pio2_lo);
+	} else {
+		long double f, c;
+		u.f = s;
+		CLEARBOTTOM(u);
+		f = u.f;
+		c = (z - f*f)/(s + f);
+		x = 0.5*pio2_hi-(2*s*r - (pio2_lo-2*c) - (0.5*pio2_hi-2*f));
+	}
+	return sign ? -x : x;
+}
+#endif
diff --git a/src/math/atan.c b/src/math/atan.c
index 583a88a..63b0ab2 100644
--- a/src/math/atan.c
+++ b/src/math/atan.c
@@ -1,3 +1,14 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_atan.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* atan(x)
  * Method
  *   1. Reduce x to positive by atan(x) = -atan(-x).
@@ -19,8 +30,7 @@
  */
 
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double atanhi[] = {
   4.63647609000806093515e-01, /* atan(0.5)hi 0x3FDDAC67, 0x0561BB4F */
@@ -103,4 +113,4 @@ double atan(double x)
 		return x - x*(s1+s2);
 	z = atanhi[id] - (x*(s1+s2) - atanlo[id] - x);
 	return sign ? -z : z;
-}
\ No newline at end of file
+}
diff --git a/src/math/atan2.c b/src/math/atan2.c
index 3d15b8e..5a1903c 100644
--- a/src/math/atan2.c
+++ b/src/math/atan2.c
@@ -1,3 +1,15 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_atan2.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ *
+ */
 /* atan2(y,x)
  * Method :
  *      1. Reduce y to positive by atan2(y,x)=-atan2(-y,x).
@@ -25,8 +37,7 @@
  * to produce the hexadecimal values shown.
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 pi     = 3.1415926535897931160E+00, /* 0x400921FB, 0x54442D18 */
@@ -41,7 +52,7 @@ double atan2(double y, double x)
 		return x+y;
 	EXTRACT_WORDS(ix, lx, x);
 	EXTRACT_WORDS(iy, ly, y);
-	if (((ix-0x3ff00000) | lx) == 0)  /* x = 1.0 */
+	if ((ix-0x3ff00000 | lx) == 0)  /* x = 1.0 */
 		return atan(y);
 	m = ((iy>>31)&1) | ((ix>>30)&2);  /* 2*sign(x)+sign(y) */
 	ix = ix & 0x7fffffff;
diff --git a/src/math/atan2f.c b/src/math/atan2f.c
index 3bc78c3..d154eb4 100644
--- a/src/math/atan2f.c
+++ b/src/math/atan2f.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 pi     = 3.1415927410e+00, /* 0x40490fdb */
diff --git a/src/math/atanf.c b/src/math/atanf.c
index b73b2e7..5824b04 100644
--- a/src/math/atanf.c
+++ b/src/math/atanf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float atanhi[] = {
   4.6364760399e-01, /* atan(0.5)hi 0x3eed6338 */
diff --git a/src/math/atanh.c b/src/math/atanh.c
index cbd976c..397cc23 100644
--- a/src/math/atanh.c
+++ b/src/math/atanh.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* atanh(x) = log((1+x)/(1-x))/2 = log1p(2x/(1-x))/2 ~= x + x^3/3 + o(x^5) */
 double atanh(double x)
diff --git a/src/math/atanhf.c b/src/math/atanhf.c
index b454879..e531edd 100644
--- a/src/math/atanhf.c
+++ b/src/math/atanhf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* atanh(x) = log((1+x)/(1-x))/2 = log1p(2x/(1-x))/2 ~= x + x^3/3 + o(x^5) */
 float atanhf(float x)
diff --git a/src/math/atanhl.c b/src/math/atanhl.c
new file mode 100644
index 0000000..87cd1cd
--- /dev/null
+++ b/src/math/atanhl.c
@@ -0,0 +1,35 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double atanhl(long double x)
+{
+	return atanh(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+/* atanh(x) = log((1+x)/(1-x))/2 = log1p(2x/(1-x))/2 ~= x + x^3/3 + o(x^5) */
+long double atanhl(long double x)
+{
+	union ldshape u = {x};
+	unsigned e = u.i.se & 0x7fff;
+	unsigned s = u.i.se >> 15;
+
+	/* |x| */
+	u.i.se = e;
+	x = u.f;
+
+	if (e < 0x3ff - 1) {
+		if (e < 0x3ff - LDBL_MANT_DIG/2) {
+			/* handle underflow */
+			if (e == 0)
+				FORCE_EVAL((float)x);
+		} else {
+			/* |x| < 0.5, up to 1.7ulp error */
+			x = 0.5*log1pl(2*x + 2*x*x/(1-x));
+		}
+	} else {
+		/* avoid overflow */
+		x = 0.5*log1pl(2*(x/(1-x)));
+	}
+	return s ? -x : x;
+}
+#endif
diff --git a/src/math/cbrt.c b/src/math/cbrt.c
index c98a929..37ede17 100644
--- a/src/math/cbrt.c
+++ b/src/math/cbrt.c
@@ -2,8 +2,7 @@
  * Return cube root of x
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const uint32_t
 B1 = 715094163, /* B1 = (1023-1023/3-0.03306235651)*2**20 */
diff --git a/src/math/cbrtf.c b/src/math/cbrtf.c
index a758df9..b5ebc76 100644
--- a/src/math/cbrtf.c
+++ b/src/math/cbrtf.c
@@ -2,8 +2,7 @@
  * Return cube root of x
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const unsigned
 B1 = 709958130, /* B1 = (127-127.0/3-0.03306235651)*2**23 */
diff --git a/src/math/cbrtl.c b/src/math/cbrtl.c
new file mode 100644
index 0000000..ceff913
--- /dev/null
+++ b/src/math/cbrtl.c
@@ -0,0 +1,124 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_cbrtl.c */
+/*-
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ * Copyright (c) 2009-2011, Bruce D. Evans, Steven G. Kargl, David Schultz.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ *
+ * The argument reduction and testing for exceptional cases was
+ * written by Steven G. Kargl with input from Bruce D. Evans
+ * and David A. Schultz.
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double cbrtl(long double x)
+{
+	return cbrt(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+static const unsigned B1 = 709958130; /* B1 = (127-127.0/3-0.03306235651)*2**23 */
+
+long double cbrtl(long double x)
+{
+	union ldshape u = {x}, v;
+	union {float f; uint32_t i;} uft;
+	long double r, s, t, w;
+	double_t dr, dt, dx;
+	float_t ft;
+	int e = u.i.se & 0x7fff;
+	int sign = u.i.se & 0x8000;
+
+	/*
+	 * If x = +-Inf, then cbrt(x) = +-Inf.
+	 * If x = NaN, then cbrt(x) = NaN.
+	 */
+	if (e == 0x7fff)
+		return x + x;
+	if (e == 0) {
+		/* Adjust subnormal numbers. */
+		u.f *= 0x1p120;
+		e = u.i.se & 0x7fff;
+		/* If x = +-0, then cbrt(x) = +-0. */
+		if (e == 0)
+			return x;
+		e -= 120;
+	}
+	e -= 0x3fff;
+	u.i.se = 0x3fff;
+	x = u.f;
+	switch (e % 3) {
+	case 1:
+	case -2:
+		x *= 2;
+		e--;
+		break;
+	case 2:
+	case -1:
+		x *= 4;
+		e -= 2;
+		break;
+	}
+	v.f = 1.0;
+	v.i.se = sign | (0x3fff + e/3);
+
+	/*
+	 * The following is the guts of s_cbrtf, with the handling of
+	 * special values removed and extra care for accuracy not taken,
+	 * but with most of the extra accuracy not discarded.
+	 */
+
+	/* ~5-bit estimate: */
+	uft.f = x;
+	uft.i = (uft.i & 0x7fffffff)/3 + B1;
+	ft = uft.f;
+
+	/* ~16-bit estimate: */
+	dx = x;
+	dt = ft;
+	dr = dt * dt * dt;
+	dt = dt * (dx + dx + dr) / (dx + dr + dr);
+
+	/* ~47-bit estimate: */
+	dr = dt * dt * dt;
+	dt = dt * (dx + dx + dr) / (dx + dr + dr);
+
+#if LDBL_MANT_DIG == 64
+	/*
+	 * dt is cbrtl(x) to ~47 bits (after x has been reduced to 1 <= x < 8).
+	 * Round it away from zero to 32 bits (32 so that t*t is exact, and
+	 * away from zero for technical reasons).
+	 */
+	t = dt + (0x1.0p32L + 0x1.0p-31L) - 0x1.0p32;
+#elif LDBL_MANT_DIG == 113
+	/*
+	 * Round dt away from zero to 47 bits.  Since we don't trust the 47,
+	 * add 2 47-bit ulps instead of 1 to round up.  Rounding is slow and
+	 * might be avoidable in this case, since on most machines dt will
+	 * have been evaluated in 53-bit precision and the technical reasons
+	 * for rounding up might not apply to either case in cbrtl() since
+	 * dt is much more accurate than needed.
+	 */
+	t = dt + 0x2.0p-46 + 0x1.0p60L - 0x1.0p60;
+#endif
+
+	/*
+	 * Final step Newton iteration to 64 or 113 bits with
+	 * error < 0.667 ulps
+	 */
+	s = t*t;         /* t*t is exact */
+	r = x/s;         /* error <= 0.5 ulps; |r| < |t| */
+	w = t+t;         /* t+t is exact */
+	r = (r-t)/(w+r); /* r-t is exact; w+r ~= 3*t */
+	t = t+t*r;       /* error <= 0.5 + 0.5/3 + epsilon */
+
+	t *= v.f;
+	return t;
+}
+#endif
diff --git a/src/math/ceil.c b/src/math/ceil.c
index 06717a6..55468f3 100644
--- a/src/math/ceil.c
+++ b/src/math/ceil.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define EPS DBL_EPSILON
 static const double_t toint = 1/EPS;
diff --git a/src/math/ceilf.c b/src/math/ceilf.c
index c148bc6..0a0d14e 100644
--- a/src/math/ceilf.c
+++ b/src/math/ceilf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static float __ceilf(float x)
 {
diff --git a/src/math/ceill.c b/src/math/ceill.c
new file mode 100644
index 0000000..60a8302
--- /dev/null
+++ b/src/math/ceill.c
@@ -0,0 +1,34 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double ceill(long double x)
+{
+	return ceil(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+
+static const long double toint = 1/LDBL_EPSILON;
+
+long double ceill(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	long double y;
+
+	if (e >= 0x3fff+LDBL_MANT_DIG-1 || x == 0)
+		return x;
+	/* y = int(x) - x, where int(x) is an integer neighbor of x */
+	if (u.i.se >> 15)
+		y = x - toint + toint - x;
+	else
+		y = x + toint - toint - x;
+	/* special case because of non-nearest rounding modes */
+	if (e <= 0x3fff-1) {
+		FORCE_EVAL(y);
+		return u.i.se >> 15 ? -0.0 : 1;
+	}
+	if (y < 0)
+		return x + y + 1;
+	return x + y;
+}
+#endif
diff --git a/src/math/copysign.c b/src/math/copysign.c
index b6589c1..b09331b 100644
--- a/src/math/copysign.c
+++ b/src/math/copysign.c
@@ -1,13 +1,8 @@
-#include <math.h>
-#include <stdint.h>
+#include "libm.h"
 
-// #include "math_private.h"
-
-double copysign(double x, double y)
-{
-    uint32_t hx, hy;
-    GET_HIGH_WORD (hx, x);
-    GET_HIGH_WORD (hy, y);
-    SET_HIGH_WORD (x, (hx & 0x7fffffff) | (hy & 0x80000000));
-    return x;
+double copysign(double x, double y) {
+	union {double f; uint64_t i;} ux={x}, uy={y};
+	ux.i &= -1ULL/2;
+	ux.i |= uy.i & 1ULL<<63;
+	return ux.f;
 }
diff --git a/src/math/copysignf.c b/src/math/copysignf.c
index fc65379..0af6ae9 100644
--- a/src/math/copysignf.c
+++ b/src/math/copysignf.c
@@ -1,12 +1,10 @@
 #include <math.h>
 #include <stdint.h>
-// #include "math_private.h"
 
 float copysignf(float x, float y)
 {
-	uint32_t ix,iy;
-	GET_FLOAT_WORD(ix,x);
-	GET_FLOAT_WORD(iy,y);
-	SET_FLOAT_WORD(x,(ix&0x7fffffff)|(iy&0x80000000));
-    return x;
-}
\ No newline at end of file
+	union {float f; uint32_t i;} ux={x}, uy={y};
+	ux.i &= 0x7fffffff;
+	ux.i |= uy.i & 0x80000000;
+	return ux.f;
+}
diff --git a/src/math/cos.c b/src/math/cos.c
index c18a2c8..ee97f68 100644
--- a/src/math/cos.c
+++ b/src/math/cos.c
@@ -1,3 +1,14 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_cos.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* cos(x)
  * Return cosine function of x.
  *
@@ -29,8 +40,7 @@
  *      TRIG(x) returns trig(x) nearly rounded
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 double cos(double x)
 {
diff --git a/src/math/cosf.c b/src/math/cosf.c
index 32d149e..be6b25e 100644
--- a/src/math/cosf.c
+++ b/src/math/cosf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* Small multiples of pi/2 rounded to double precision. */
 static const double
diff --git a/src/math/cosh.c b/src/math/cosh.c
index 596dac6..490c15f 100644
--- a/src/math/cosh.c
+++ b/src/math/cosh.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* cosh(x) = (exp(x) + 1/exp(x))/2
  *         = 1 + 0.5*(exp(x)-1)*(exp(x)-1)/exp(x)
@@ -36,6 +35,6 @@ double cosh(double x)
 
 	/* |x| > log(DBL_MAX) or nan */
 	/* note: the result is stored to handle overflow */
-	t = __expo2(x);
+	t = __expo2(x, 1.0);
 	return t;
 }
diff --git a/src/math/coshf.c b/src/math/coshf.c
index 8b95247..e739cff 100644
--- a/src/math/coshf.c
+++ b/src/math/coshf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float coshf(float x)
 {
@@ -29,6 +28,6 @@ float coshf(float x)
 	}
 
 	/* |x| > log(FLT_MAX) or nan */
-	t = __expo2f(x);
+	t = __expo2f(x, 1.0f);
 	return t;
 }
diff --git a/src/math/coshl.c b/src/math/coshl.c
new file mode 100644
index 0000000..06a56fe
--- /dev/null
+++ b/src/math/coshl.c
@@ -0,0 +1,47 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double coshl(long double x)
+{
+	return cosh(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+long double coshl(long double x)
+{
+	union ldshape u = {x};
+	unsigned ex = u.i.se & 0x7fff;
+	uint32_t w;
+	long double t;
+
+	/* |x| */
+	u.i.se = ex;
+	x = u.f;
+	w = u.i.m >> 32;
+
+	/* |x| < log(2) */
+	if (ex < 0x3fff-1 || (ex == 0x3fff-1 && w < 0xb17217f7)) {
+		if (ex < 0x3fff-32) {
+			FORCE_EVAL(x + 0x1p120f);
+			return 1;
+		}
+		t = expm1l(x);
+		return 1 + t*t/(2*(1+t));
+	}
+
+	/* |x| < log(LDBL_MAX) */
+	if (ex < 0x3fff+13 || (ex == 0x3fff+13 && w < 0xb17217f7)) {
+		t = expl(x);
+		return 0.5*(t + 1/t);
+	}
+
+	/* |x| > log(LDBL_MAX) or nan */
+	t = expl(0.5*x);
+	return 0.5*t*t;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double coshl(long double x)
+{
+	return cosh(x);
+}
+#endif
diff --git a/src/math/cosl.c b/src/math/cosl.c
new file mode 100644
index 0000000..79c41c7
--- /dev/null
+++ b/src/math/cosl.c
@@ -0,0 +1,39 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double cosl(long double x) {
+	return cos(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double cosl(long double x)
+{
+	union ldshape u = {x};
+	unsigned n;
+	long double y[2], hi, lo;
+
+	u.i.se &= 0x7fff;
+	if (u.i.se == 0x7fff)
+		return x - x;
+	x = u.f;
+	if (x < M_PI_4) {
+		if (u.i.se < 0x3fff - LDBL_MANT_DIG)
+			/* raise inexact if x!=0 */
+			return 1.0 + x;
+		return __cosl(x, 0);
+	}
+	n = __rem_pio2l(x, y);
+	hi = y[0];
+	lo = y[1];
+	switch (n & 3) {
+	case 0:
+		return __cosl(hi, lo);
+	case 1:
+		return -__sinl(hi, lo, 1);
+	case 2:
+		return -__cosl(hi, lo);
+	case 3:
+	default:
+		return __sinl(hi, lo, 1);
+	}
+}
+#endif
diff --git a/src/math/erf.c b/src/math/erf.c
new file mode 100644
index 0000000..2f30a29
--- /dev/null
+++ b/src/math/erf.c
@@ -0,0 +1,273 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_erf.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+/* double erf(double x)
+ * double erfc(double x)
+ *                           x
+ *                    2      |\
+ *     erf(x)  =  ---------  | exp(-t*t)dt
+ *                 sqrt(pi) \|
+ *                           0
+ *
+ *     erfc(x) =  1-erf(x)
+ *  Note that
+ *              erf(-x) = -erf(x)
+ *              erfc(-x) = 2 - erfc(x)
+ *
+ * Method:
+ *      1. For |x| in [0, 0.84375]
+ *          erf(x)  = x + x*R(x^2)
+ *          erfc(x) = 1 - erf(x)           if x in [-.84375,0.25]
+ *                  = 0.5 + ((0.5-x)-x*R)  if x in [0.25,0.84375]
+ *         where R = P/Q where P is an odd poly of degree 8 and
+ *         Q is an odd poly of degree 10.
+ *                                               -57.90
+ *                      | R - (erf(x)-x)/x | <= 2
+ *
+ *
+ *         Remark. The formula is derived by noting
+ *          erf(x) = (2/sqrt(pi))*(x - x^3/3 + x^5/10 - x^7/42 + ....)
+ *         and that
+ *          2/sqrt(pi) = 1.128379167095512573896158903121545171688
+ *         is close to one. The interval is chosen because the fix
+ *         point of erf(x) is near 0.6174 (i.e., erf(x)=x when x is
+ *         near 0.6174), and by some experiment, 0.84375 is chosen to
+ *         guarantee the error is less than one ulp for erf.
+ *
+ *      2. For |x| in [0.84375,1.25], let s = |x| - 1, and
+ *         c = 0.84506291151 rounded to single (24 bits)
+ *              erf(x)  = sign(x) * (c  + P1(s)/Q1(s))
+ *              erfc(x) = (1-c)  - P1(s)/Q1(s) if x > 0
+ *                        1+(c+P1(s)/Q1(s))    if x < 0
+ *              |P1/Q1 - (erf(|x|)-c)| <= 2**-59.06
+ *         Remark: here we use the taylor series expansion at x=1.
+ *              erf(1+s) = erf(1) + s*Poly(s)
+ *                       = 0.845.. + P1(s)/Q1(s)
+ *         That is, we use rational approximation to approximate
+ *                      erf(1+s) - (c = (single)0.84506291151)
+ *         Note that |P1/Q1|< 0.078 for x in [0.84375,1.25]
+ *         where
+ *              P1(s) = degree 6 poly in s
+ *              Q1(s) = degree 6 poly in s
+ *
+ *      3. For x in [1.25,1/0.35(~2.857143)],
+ *              erfc(x) = (1/x)*exp(-x*x-0.5625+R1/S1)
+ *              erf(x)  = 1 - erfc(x)
+ *         where
+ *              R1(z) = degree 7 poly in z, (z=1/x^2)
+ *              S1(z) = degree 8 poly in z
+ *
+ *      4. For x in [1/0.35,28]
+ *              erfc(x) = (1/x)*exp(-x*x-0.5625+R2/S2) if x > 0
+ *                      = 2.0 - (1/x)*exp(-x*x-0.5625+R2/S2) if -6<x<0
+ *                      = 2.0 - tiny            (if x <= -6)
+ *              erf(x)  = sign(x)*(1.0 - erfc(x)) if x < 6, else
+ *              erf(x)  = sign(x)*(1.0 - tiny)
+ *         where
+ *              R2(z) = degree 6 poly in z, (z=1/x^2)
+ *              S2(z) = degree 7 poly in z
+ *
+ *      Note1:
+ *         To compute exp(-x*x-0.5625+R/S), let s be a single
+ *         precision number and s := x; then
+ *              -x*x = -s*s + (s-x)*(s+x)
+ *              exp(-x*x-0.5626+R/S) =
+ *                      exp(-s*s-0.5625)*exp((s-x)*(s+x)+R/S);
+ *      Note2:
+ *         Here 4 and 5 make use of the asymptotic series
+ *                        exp(-x*x)
+ *              erfc(x) ~ ---------- * ( 1 + Poly(1/x^2) )
+ *                        x*sqrt(pi)
+ *         We use rational approximation to approximate
+ *              g(s)=f(1/x^2) = log(erfc(x)*x) - x*x + 0.5625
+ *         Here is the error bound for R1/S1 and R2/S2
+ *              |R1/S1 - f(x)|  < 2**(-62.57)
+ *              |R2/S2 - f(x)|  < 2**(-61.52)
+ *
+ *      5. For inf > x >= 28
+ *              erf(x)  = sign(x) *(1 - tiny)  (raise inexact)
+ *              erfc(x) = tiny*tiny (raise underflow) if x > 0
+ *                      = 2 - tiny if x<0
+ *
+ *      7. Special case:
+ *              erf(0)  = 0, erf(inf)  = 1, erf(-inf) = -1,
+ *              erfc(0) = 1, erfc(inf) = 0, erfc(-inf) = 2,
+ *              erfc/erf(NaN) is NaN
+ */
+
+#include "libm.h"
+
+static const double
+erx  = 8.45062911510467529297e-01, /* 0x3FEB0AC1, 0x60000000 */
+/*
+ * Coefficients for approximation to  erf on [0,0.84375]
+ */
+efx8 =  1.02703333676410069053e+00, /* 0x3FF06EBA, 0x8214DB69 */
+pp0  =  1.28379167095512558561e-01, /* 0x3FC06EBA, 0x8214DB68 */
+pp1  = -3.25042107247001499370e-01, /* 0xBFD4CD7D, 0x691CB913 */
+pp2  = -2.84817495755985104766e-02, /* 0xBF9D2A51, 0xDBD7194F */
+pp3  = -5.77027029648944159157e-03, /* 0xBF77A291, 0x236668E4 */
+pp4  = -2.37630166566501626084e-05, /* 0xBEF8EAD6, 0x120016AC */
+qq1  =  3.97917223959155352819e-01, /* 0x3FD97779, 0xCDDADC09 */
+qq2  =  6.50222499887672944485e-02, /* 0x3FB0A54C, 0x5536CEBA */
+qq3  =  5.08130628187576562776e-03, /* 0x3F74D022, 0xC4D36B0F */
+qq4  =  1.32494738004321644526e-04, /* 0x3F215DC9, 0x221C1A10 */
+qq5  = -3.96022827877536812320e-06, /* 0xBED09C43, 0x42A26120 */
+/*
+ * Coefficients for approximation to  erf  in [0.84375,1.25]
+ */
+pa0  = -2.36211856075265944077e-03, /* 0xBF6359B8, 0xBEF77538 */
+pa1  =  4.14856118683748331666e-01, /* 0x3FDA8D00, 0xAD92B34D */
+pa2  = -3.72207876035701323847e-01, /* 0xBFD7D240, 0xFBB8C3F1 */
+pa3  =  3.18346619901161753674e-01, /* 0x3FD45FCA, 0x805120E4 */
+pa4  = -1.10894694282396677476e-01, /* 0xBFBC6398, 0x3D3E28EC */
+pa5  =  3.54783043256182359371e-02, /* 0x3FA22A36, 0x599795EB */
+pa6  = -2.16637559486879084300e-03, /* 0xBF61BF38, 0x0A96073F */
+qa1  =  1.06420880400844228286e-01, /* 0x3FBB3E66, 0x18EEE323 */
+qa2  =  5.40397917702171048937e-01, /* 0x3FE14AF0, 0x92EB6F33 */
+qa3  =  7.18286544141962662868e-02, /* 0x3FB2635C, 0xD99FE9A7 */
+qa4  =  1.26171219808761642112e-01, /* 0x3FC02660, 0xE763351F */
+qa5  =  1.36370839120290507362e-02, /* 0x3F8BEDC2, 0x6B51DD1C */
+qa6  =  1.19844998467991074170e-02, /* 0x3F888B54, 0x5735151D */
+/*
+ * Coefficients for approximation to  erfc in [1.25,1/0.35]
+ */
+ra0  = -9.86494403484714822705e-03, /* 0xBF843412, 0x600D6435 */
+ra1  = -6.93858572707181764372e-01, /* 0xBFE63416, 0xE4BA7360 */
+ra2  = -1.05586262253232909814e+01, /* 0xC0251E04, 0x41B0E726 */
+ra3  = -6.23753324503260060396e+01, /* 0xC04F300A, 0xE4CBA38D */
+ra4  = -1.62396669462573470355e+02, /* 0xC0644CB1, 0x84282266 */
+ra5  = -1.84605092906711035994e+02, /* 0xC067135C, 0xEBCCABB2 */
+ra6  = -8.12874355063065934246e+01, /* 0xC0545265, 0x57E4D2F2 */
+ra7  = -9.81432934416914548592e+00, /* 0xC023A0EF, 0xC69AC25C */
+sa1  =  1.96512716674392571292e+01, /* 0x4033A6B9, 0xBD707687 */
+sa2  =  1.37657754143519042600e+02, /* 0x4061350C, 0x526AE721 */
+sa3  =  4.34565877475229228821e+02, /* 0x407B290D, 0xD58A1A71 */
+sa4  =  6.45387271733267880336e+02, /* 0x40842B19, 0x21EC2868 */
+sa5  =  4.29008140027567833386e+02, /* 0x407AD021, 0x57700314 */
+sa6  =  1.08635005541779435134e+02, /* 0x405B28A3, 0xEE48AE2C */
+sa7  =  6.57024977031928170135e+00, /* 0x401A47EF, 0x8E484A93 */
+sa8  = -6.04244152148580987438e-02, /* 0xBFAEEFF2, 0xEE749A62 */
+/*
+ * Coefficients for approximation to  erfc in [1/.35,28]
+ */
+rb0  = -9.86494292470009928597e-03, /* 0xBF843412, 0x39E86F4A */
+rb1  = -7.99283237680523006574e-01, /* 0xBFE993BA, 0x70C285DE */
+rb2  = -1.77579549177547519889e+01, /* 0xC031C209, 0x555F995A */
+rb3  = -1.60636384855821916062e+02, /* 0xC064145D, 0x43C5ED98 */
+rb4  = -6.37566443368389627722e+02, /* 0xC083EC88, 0x1375F228 */
+rb5  = -1.02509513161107724954e+03, /* 0xC0900461, 0x6A2E5992 */
+rb6  = -4.83519191608651397019e+02, /* 0xC07E384E, 0x9BDC383F */
+sb1  =  3.03380607434824582924e+01, /* 0x403E568B, 0x261D5190 */
+sb2  =  3.25792512996573918826e+02, /* 0x40745CAE, 0x221B9F0A */
+sb3  =  1.53672958608443695994e+03, /* 0x409802EB, 0x189D5118 */
+sb4  =  3.19985821950859553908e+03, /* 0x40A8FFB7, 0x688C246A */
+sb5  =  2.55305040643316442583e+03, /* 0x40A3F219, 0xCEDF3BE6 */
+sb6  =  4.74528541206955367215e+02, /* 0x407DA874, 0xE79FE763 */
+sb7  = -2.24409524465858183362e+01; /* 0xC03670E2, 0x42712D62 */
+
+static double erfc1(double x)
+{
+	double_t s,P,Q;
+
+	s = fabs(x) - 1;
+	P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
+	Q = 1+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
+	return 1 - erx - P/Q;
+}
+
+static double erfc2(uint32_t ix, double x)
+{
+	double_t s,R,S;
+	double z;
+
+	if (ix < 0x3ff40000)  /* |x| < 1.25 */
+		return erfc1(x);
+
+	x = fabs(x);
+	s = 1/(x*x);
+	if (ix < 0x4006db6d) {  /* |x| < 1/.35 ~ 2.85714 */
+		R = ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(
+		     ra5+s*(ra6+s*ra7))))));
+		S = 1.0+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(
+		     sa5+s*(sa6+s*(sa7+s*sa8)))))));
+	} else {                /* |x| > 1/.35 */
+		R = rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(
+		     rb5+s*rb6)))));
+		S = 1.0+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(
+		     sb5+s*(sb6+s*sb7))))));
+	}
+	z = x;
+	SET_LOW_WORD(z,0);
+	return exp(-z*z-0.5625)*exp((z-x)*(z+x)+R/S)/x;
+}
+
+double erf(double x)
+{
+	double r,s,z,y;
+	uint32_t ix;
+	int sign;
+
+	GET_HIGH_WORD(ix, x);
+	sign = ix>>31;
+	ix &= 0x7fffffff;
+	if (ix >= 0x7ff00000) {
+		/* erf(nan)=nan, erf(+-inf)=+-1 */
+		return 1-2*sign + 1/x;
+	}
+	if (ix < 0x3feb0000) {  /* |x| < 0.84375 */
+		if (ix < 0x3e300000) {  /* |x| < 2**-28 */
+			/* avoid underflow */
+			return 0.125*(8*x + efx8*x);
+		}
+		z = x*x;
+		r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
+		s = 1.0+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
+		y = r/s;
+		return x + x*y;
+	}
+	if (ix < 0x40180000)  /* 0.84375 <= |x| < 6 */
+		y = 1 - erfc2(ix,x);
+	else
+		y = 1 - 0x1p-1022;
+	return sign ? -y : y;
+}
+
+double erfc(double x)
+{
+	double r,s,z,y;
+	uint32_t ix;
+	int sign;
+
+	GET_HIGH_WORD(ix, x);
+	sign = ix>>31;
+	ix &= 0x7fffffff;
+	if (ix >= 0x7ff00000) {
+		/* erfc(nan)=nan, erfc(+-inf)=0,2 */
+		return 2*sign + 1/x;
+	}
+	if (ix < 0x3feb0000) {  /* |x| < 0.84375 */
+		if (ix < 0x3c700000)  /* |x| < 2**-56 */
+			return 1.0 - x;
+		z = x*x;
+		r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
+		s = 1.0+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
+		y = r/s;
+		if (sign || ix < 0x3fd00000) {  /* x < 1/4 */
+			return 1.0 - (x+x*y);
+		}
+		return 0.5 - (x - 0.5 + x*y);
+	}
+	if (ix < 0x403c0000) {  /* 0.84375 <= |x| < 28 */
+		return sign ? 2 - erfc2(ix,x) : erfc2(ix,x);
+	}
+	return sign ? 2 - 0x1p-1022 : 0x1p-1022*0x1p-1022;
+}
diff --git a/src/math/erff.c b/src/math/erff.c
new file mode 100644
index 0000000..ed5f397
--- /dev/null
+++ b/src/math/erff.c
@@ -0,0 +1,183 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_erff.c */
+/*
+ * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
+ */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+#include "libm.h"
+
+static const float
+erx  =  8.4506291151e-01, /* 0x3f58560b */
+/*
+ * Coefficients for approximation to  erf on [0,0.84375]
+ */
+efx8 =  1.0270333290e+00, /* 0x3f8375d4 */
+pp0  =  1.2837916613e-01, /* 0x3e0375d4 */
+pp1  = -3.2504209876e-01, /* 0xbea66beb */
+pp2  = -2.8481749818e-02, /* 0xbce9528f */
+pp3  = -5.7702702470e-03, /* 0xbbbd1489 */
+pp4  = -2.3763017452e-05, /* 0xb7c756b1 */
+qq1  =  3.9791721106e-01, /* 0x3ecbbbce */
+qq2  =  6.5022252500e-02, /* 0x3d852a63 */
+qq3  =  5.0813062117e-03, /* 0x3ba68116 */
+qq4  =  1.3249473704e-04, /* 0x390aee49 */
+qq5  = -3.9602282413e-06, /* 0xb684e21a */
+/*
+ * Coefficients for approximation to  erf  in [0.84375,1.25]
+ */
+pa0  = -2.3621185683e-03, /* 0xbb1acdc6 */
+pa1  =  4.1485610604e-01, /* 0x3ed46805 */
+pa2  = -3.7220788002e-01, /* 0xbebe9208 */
+pa3  =  3.1834661961e-01, /* 0x3ea2fe54 */
+pa4  = -1.1089469492e-01, /* 0xbde31cc2 */
+pa5  =  3.5478305072e-02, /* 0x3d1151b3 */
+pa6  = -2.1663755178e-03, /* 0xbb0df9c0 */
+qa1  =  1.0642088205e-01, /* 0x3dd9f331 */
+qa2  =  5.4039794207e-01, /* 0x3f0a5785 */
+qa3  =  7.1828655899e-02, /* 0x3d931ae7 */
+qa4  =  1.2617121637e-01, /* 0x3e013307 */
+qa5  =  1.3637083583e-02, /* 0x3c5f6e13 */
+qa6  =  1.1984500103e-02, /* 0x3c445aa3 */
+/*
+ * Coefficients for approximation to  erfc in [1.25,1/0.35]
+ */
+ra0  = -9.8649440333e-03, /* 0xbc21a093 */
+ra1  = -6.9385856390e-01, /* 0xbf31a0b7 */
+ra2  = -1.0558626175e+01, /* 0xc128f022 */
+ra3  = -6.2375331879e+01, /* 0xc2798057 */
+ra4  = -1.6239666748e+02, /* 0xc322658c */
+ra5  = -1.8460508728e+02, /* 0xc3389ae7 */
+ra6  = -8.1287437439e+01, /* 0xc2a2932b */
+ra7  = -9.8143291473e+00, /* 0xc11d077e */
+sa1  =  1.9651271820e+01, /* 0x419d35ce */
+sa2  =  1.3765776062e+02, /* 0x4309a863 */
+sa3  =  4.3456588745e+02, /* 0x43d9486f */
+sa4  =  6.4538726807e+02, /* 0x442158c9 */
+sa5  =  4.2900814819e+02, /* 0x43d6810b */
+sa6  =  1.0863500214e+02, /* 0x42d9451f */
+sa7  =  6.5702495575e+00, /* 0x40d23f7c */
+sa8  = -6.0424413532e-02, /* 0xbd777f97 */
+/*
+ * Coefficients for approximation to  erfc in [1/.35,28]
+ */
+rb0  = -9.8649431020e-03, /* 0xbc21a092 */
+rb1  = -7.9928326607e-01, /* 0xbf4c9dd4 */
+rb2  = -1.7757955551e+01, /* 0xc18e104b */
+rb3  = -1.6063638306e+02, /* 0xc320a2ea */
+rb4  = -6.3756646729e+02, /* 0xc41f6441 */
+rb5  = -1.0250950928e+03, /* 0xc480230b */
+rb6  = -4.8351919556e+02, /* 0xc3f1c275 */
+sb1  =  3.0338060379e+01, /* 0x41f2b459 */
+sb2  =  3.2579251099e+02, /* 0x43a2e571 */
+sb3  =  1.5367296143e+03, /* 0x44c01759 */
+sb4  =  3.1998581543e+03, /* 0x4547fdbb */
+sb5  =  2.5530502930e+03, /* 0x451f90ce */
+sb6  =  4.7452853394e+02, /* 0x43ed43a7 */
+sb7  = -2.2440952301e+01; /* 0xc1b38712 */
+
+static float erfc1(float x)
+{
+	float_t s,P,Q;
+
+	s = fabsf(x) - 1;
+	P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
+	Q = 1+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
+	return 1 - erx - P/Q;
+}
+
+static float erfc2(uint32_t ix, float x)
+{
+	float_t s,R,S;
+	float z;
+
+	if (ix < 0x3fa00000)  /* |x| < 1.25 */
+		return erfc1(x);
+
+	x = fabsf(x);
+	s = 1/(x*x);
+	if (ix < 0x4036db6d) {   /* |x| < 1/0.35 */
+		R = ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(
+		     ra5+s*(ra6+s*ra7))))));
+		S = 1.0f+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(
+		     sa5+s*(sa6+s*(sa7+s*sa8)))))));
+	} else {                 /* |x| >= 1/0.35 */
+		R = rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(
+		     rb5+s*rb6)))));
+		S = 1.0f+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(
+		     sb5+s*(sb6+s*sb7))))));
+	}
+	GET_FLOAT_WORD(ix, x);
+	SET_FLOAT_WORD(z, ix&0xffffe000);
+	return expf(-z*z - 0.5625f) * expf((z-x)*(z+x) + R/S)/x;
+}
+
+float erff(float x)
+{
+	float r,s,z,y;
+	uint32_t ix;
+	int sign;
+
+	GET_FLOAT_WORD(ix, x);
+	sign = ix>>31;
+	ix &= 0x7fffffff;
+	if (ix >= 0x7f800000) {
+		/* erf(nan)=nan, erf(+-inf)=+-1 */
+		return 1-2*sign + 1/x;
+	}
+	if (ix < 0x3f580000) {  /* |x| < 0.84375 */
+		if (ix < 0x31800000) {  /* |x| < 2**-28 */
+			/*avoid underflow */
+			return 0.125f*(8*x + efx8*x);
+		}
+		z = x*x;
+		r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
+		s = 1+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
+		y = r/s;
+		return x + x*y;
+	}
+	if (ix < 0x40c00000)  /* |x| < 6 */
+		y = 1 - erfc2(ix,x);
+	else
+		y = 1 - 0x1p-120f;
+	return sign ? -y : y;
+}
+
+float erfcf(float x)
+{
+	float r,s,z,y;
+	uint32_t ix;
+	int sign;
+
+	GET_FLOAT_WORD(ix, x);
+	sign = ix>>31;
+	ix &= 0x7fffffff;
+	if (ix >= 0x7f800000) {
+		/* erfc(nan)=nan, erfc(+-inf)=0,2 */
+		return 2*sign + 1/x;
+	}
+
+	if (ix < 0x3f580000) {  /* |x| < 0.84375 */
+		if (ix < 0x23800000)  /* |x| < 2**-56 */
+			return 1.0f - x;
+		z = x*x;
+		r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
+		s = 1.0f+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
+		y = r/s;
+		if (sign || ix < 0x3e800000)  /* x < 1/4 */
+			return 1.0f - (x+x*y);
+		return 0.5f - (x - 0.5f + x*y);
+	}
+	if (ix < 0x41e00000) {  /* |x| < 28 */
+		return sign ? 2 - erfc2(ix,x) : erfc2(ix,x);
+	}
+	return sign ? 2 - 0x1p-120f : 0x1p-120f*0x1p-120f;
+}
diff --git a/src/math/erfl.c b/src/math/erfl.c
new file mode 100644
index 0000000..e267c23
--- /dev/null
+++ b/src/math/erfl.c
@@ -0,0 +1,353 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_erfl.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/* double erf(double x)
+ * double erfc(double x)
+ *                           x
+ *                    2      |\
+ *     erf(x)  =  ---------  | exp(-t*t)dt
+ *                 sqrt(pi) \|
+ *                           0
+ *
+ *     erfc(x) =  1-erf(x)
+ *  Note that
+ *              erf(-x) = -erf(x)
+ *              erfc(-x) = 2 - erfc(x)
+ *
+ * Method:
+ *      1. For |x| in [0, 0.84375]
+ *          erf(x)  = x + x*R(x^2)
+ *          erfc(x) = 1 - erf(x)           if x in [-.84375,0.25]
+ *                  = 0.5 + ((0.5-x)-x*R)  if x in [0.25,0.84375]
+ *         Remark. The formula is derived by noting
+ *          erf(x) = (2/sqrt(pi))*(x - x^3/3 + x^5/10 - x^7/42 + ....)
+ *         and that
+ *          2/sqrt(pi) = 1.128379167095512573896158903121545171688
+ *         is close to one. The interval is chosen because the fix
+ *         point of erf(x) is near 0.6174 (i.e., erf(x)=x when x is
+ *         near 0.6174), and by some experiment, 0.84375 is chosen to
+ *         guarantee the error is less than one ulp for erf.
+ *
+ *      2. For |x| in [0.84375,1.25], let s = |x| - 1, and
+ *         c = 0.84506291151 rounded to single (24 bits)
+ *      erf(x)  = sign(x) * (c  + P1(s)/Q1(s))
+ *      erfc(x) = (1-c)  - P1(s)/Q1(s) if x > 0
+ *                        1+(c+P1(s)/Q1(s))    if x < 0
+ *         Remark: here we use the taylor series expansion at x=1.
+ *              erf(1+s) = erf(1) + s*Poly(s)
+ *                       = 0.845.. + P1(s)/Q1(s)
+ *         Note that |P1/Q1|< 0.078 for x in [0.84375,1.25]
+ *
+ *      3. For x in [1.25,1/0.35(~2.857143)],
+ *      erfc(x) = (1/x)*exp(-x*x-0.5625+R1(z)/S1(z))
+ *              z=1/x^2
+ *      erf(x)  = 1 - erfc(x)
+ *
+ *      4. For x in [1/0.35,107]
+ *      erfc(x) = (1/x)*exp(-x*x-0.5625+R2/S2) if x > 0
+ *                      = 2.0 - (1/x)*exp(-x*x-0.5625+R2(z)/S2(z))
+ *                             if -6.666<x<0
+ *                      = 2.0 - tiny            (if x <= -6.666)
+ *              z=1/x^2
+ *      erf(x)  = sign(x)*(1.0 - erfc(x)) if x < 6.666, else
+ *      erf(x)  = sign(x)*(1.0 - tiny)
+ *      Note1:
+ *         To compute exp(-x*x-0.5625+R/S), let s be a single
+ *         precision number and s := x; then
+ *              -x*x = -s*s + (s-x)*(s+x)
+ *              exp(-x*x-0.5626+R/S) =
+ *                      exp(-s*s-0.5625)*exp((s-x)*(s+x)+R/S);
+ *      Note2:
+ *         Here 4 and 5 make use of the asymptotic series
+ *                        exp(-x*x)
+ *              erfc(x) ~ ---------- * ( 1 + Poly(1/x^2) )
+ *                        x*sqrt(pi)
+ *
+ *      5. For inf > x >= 107
+ *      erf(x)  = sign(x) *(1 - tiny)  (raise inexact)
+ *      erfc(x) = tiny*tiny (raise underflow) if x > 0
+ *                      = 2 - tiny if x<0
+ *
+ *      7. Special case:
+ *      erf(0)  = 0, erf(inf)  = 1, erf(-inf) = -1,
+ *      erfc(0) = 1, erfc(inf) = 0, erfc(-inf) = 2,
+ *              erfc/erf(NaN) is NaN
+ */
+
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double erfl(long double x)
+{
+	return erf(x);
+}
+long double erfcl(long double x)
+{
+	return erfc(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+static const long double
+erx = 0.845062911510467529296875L,
+
+/*
+ * Coefficients for approximation to  erf on [0,0.84375]
+ */
+/* 8 * (2/sqrt(pi) - 1) */
+efx8 = 1.0270333367641005911692712249723613735048E0L,
+pp[6] = {
+	1.122751350964552113068262337278335028553E6L,
+	-2.808533301997696164408397079650699163276E6L,
+	-3.314325479115357458197119660818768924100E5L,
+	-6.848684465326256109712135497895525446398E4L,
+	-2.657817695110739185591505062971929859314E3L,
+	-1.655310302737837556654146291646499062882E2L,
+},
+qq[6] = {
+	8.745588372054466262548908189000448124232E6L,
+	3.746038264792471129367533128637019611485E6L,
+	7.066358783162407559861156173539693900031E5L,
+	7.448928604824620999413120955705448117056E4L,
+	4.511583986730994111992253980546131408924E3L,
+	1.368902937933296323345610240009071254014E2L,
+	/* 1.000000000000000000000000000000000000000E0 */
+},
+
+/*
+ * Coefficients for approximation to  erf  in [0.84375,1.25]
+ */
+/* erf(x+1) = 0.845062911510467529296875 + pa(x)/qa(x)
+   -0.15625 <= x <= +.25
+   Peak relative error 8.5e-22  */
+pa[8] = {
+	-1.076952146179812072156734957705102256059E0L,
+	 1.884814957770385593365179835059971587220E2L,
+	-5.339153975012804282890066622962070115606E1L,
+	 4.435910679869176625928504532109635632618E1L,
+	 1.683219516032328828278557309642929135179E1L,
+	-2.360236618396952560064259585299045804293E0L,
+	 1.852230047861891953244413872297940938041E0L,
+	 9.394994446747752308256773044667843200719E-2L,
+},
+qa[7] =  {
+	4.559263722294508998149925774781887811255E2L,
+	3.289248982200800575749795055149780689738E2L,
+	2.846070965875643009598627918383314457912E2L,
+	1.398715859064535039433275722017479994465E2L,
+	6.060190733759793706299079050985358190726E1L,
+	2.078695677795422351040502569964299664233E1L,
+	4.641271134150895940966798357442234498546E0L,
+	/* 1.000000000000000000000000000000000000000E0 */
+},
+
+/*
+ * Coefficients for approximation to  erfc in [1.25,1/0.35]
+ */
+/* erfc(1/x) = x exp (-1/x^2 - 0.5625 + ra(x^2)/sa(x^2))
+   1/2.85711669921875 < 1/x < 1/1.25
+   Peak relative error 3.1e-21  */
+ra[] = {
+	1.363566591833846324191000679620738857234E-1L,
+	1.018203167219873573808450274314658434507E1L,
+	1.862359362334248675526472871224778045594E2L,
+	1.411622588180721285284945138667933330348E3L,
+	5.088538459741511988784440103218342840478E3L,
+	8.928251553922176506858267311750789273656E3L,
+	7.264436000148052545243018622742770549982E3L,
+	2.387492459664548651671894725748959751119E3L,
+	2.220916652813908085449221282808458466556E2L,
+},
+sa[] = {
+	-1.382234625202480685182526402169222331847E1L,
+	-3.315638835627950255832519203687435946482E2L,
+	-2.949124863912936259747237164260785326692E3L,
+	-1.246622099070875940506391433635999693661E4L,
+	-2.673079795851665428695842853070996219632E4L,
+	-2.880269786660559337358397106518918220991E4L,
+	-1.450600228493968044773354186390390823713E4L,
+	-2.874539731125893533960680525192064277816E3L,
+	-1.402241261419067750237395034116942296027E2L,
+	/* 1.000000000000000000000000000000000000000E0 */
+},
+
+/*
+ * Coefficients for approximation to  erfc in [1/.35,107]
+ */
+/* erfc(1/x) = x exp (-1/x^2 - 0.5625 + rb(x^2)/sb(x^2))
+   1/6.6666259765625 < 1/x < 1/2.85711669921875
+   Peak relative error 4.2e-22  */
+rb[] = {
+	-4.869587348270494309550558460786501252369E-5L,
+	-4.030199390527997378549161722412466959403E-3L,
+	-9.434425866377037610206443566288917589122E-2L,
+	-9.319032754357658601200655161585539404155E-1L,
+	-4.273788174307459947350256581445442062291E0L,
+	-8.842289940696150508373541814064198259278E0L,
+	-7.069215249419887403187988144752613025255E0L,
+	-1.401228723639514787920274427443330704764E0L,
+},
+sb[] = {
+	4.936254964107175160157544545879293019085E-3L,
+	1.583457624037795744377163924895349412015E-1L,
+	1.850647991850328356622940552450636420484E0L,
+	9.927611557279019463768050710008450625415E0L,
+	2.531667257649436709617165336779212114570E1L,
+	2.869752886406743386458304052862814690045E1L,
+	1.182059497870819562441683560749192539345E1L,
+	/* 1.000000000000000000000000000000000000000E0 */
+},
+/* erfc(1/x) = x exp (-1/x^2 - 0.5625 + rc(x^2)/sc(x^2))
+   1/107 <= 1/x <= 1/6.6666259765625
+   Peak relative error 1.1e-21  */
+rc[] = {
+	-8.299617545269701963973537248996670806850E-5L,
+	-6.243845685115818513578933902532056244108E-3L,
+	-1.141667210620380223113693474478394397230E-1L,
+	-7.521343797212024245375240432734425789409E-1L,
+	-1.765321928311155824664963633786967602934E0L,
+	-1.029403473103215800456761180695263439188E0L,
+},
+sc[] = {
+	8.413244363014929493035952542677768808601E-3L,
+	2.065114333816877479753334599639158060979E-1L,
+	1.639064941530797583766364412782135680148E0L,
+	4.936788463787115555582319302981666347450E0L,
+	5.005177727208955487404729933261347679090E0L,
+	/* 1.000000000000000000000000000000000000000E0 */
+};
+
+static long double erfc1(long double x)
+{
+	long double s,P,Q;
+
+	s = fabsl(x) - 1;
+	P = pa[0] + s * (pa[1] + s * (pa[2] +
+	     s * (pa[3] + s * (pa[4] + s * (pa[5] + s * (pa[6] + s * pa[7]))))));
+	Q = qa[0] + s * (qa[1] + s * (qa[2] +
+	     s * (qa[3] + s * (qa[4] + s * (qa[5] + s * (qa[6] + s))))));
+	return 1 - erx - P / Q;
+}
+
+static long double erfc2(uint32_t ix, long double x)
+{
+	union ldshape u;
+	long double s,z,R,S;
+
+	if (ix < 0x3fffa000)  /* 0.84375 <= |x| < 1.25 */
+		return erfc1(x);
+
+	x = fabsl(x);
+	s = 1 / (x * x);
+	if (ix < 0x4000b6db) {  /* 1.25 <= |x| < 2.857 ~ 1/.35 */
+		R = ra[0] + s * (ra[1] + s * (ra[2] + s * (ra[3] + s * (ra[4] +
+		     s * (ra[5] + s * (ra[6] + s * (ra[7] + s * ra[8])))))));
+		S = sa[0] + s * (sa[1] + s * (sa[2] + s * (sa[3] + s * (sa[4] +
+		     s * (sa[5] + s * (sa[6] + s * (sa[7] + s * (sa[8] + s))))))));
+	} else if (ix < 0x4001d555) {  /* 2.857 <= |x| < 6.6666259765625 */
+		R = rb[0] + s * (rb[1] + s * (rb[2] + s * (rb[3] + s * (rb[4] +
+		     s * (rb[5] + s * (rb[6] + s * rb[7]))))));
+		S = sb[0] + s * (sb[1] + s * (sb[2] + s * (sb[3] + s * (sb[4] +
+		     s * (sb[5] + s * (sb[6] + s))))));
+	} else { /* 6.666 <= |x| < 107 (erfc only) */
+		R = rc[0] + s * (rc[1] + s * (rc[2] + s * (rc[3] +
+		     s * (rc[4] + s * rc[5]))));
+		S = sc[0] + s * (sc[1] + s * (sc[2] + s * (sc[3] +
+		     s * (sc[4] + s))));
+	}
+	u.f = x;
+	u.i.m &= -1ULL << 40;
+	z = u.f;
+	return expl(-z*z - 0.5625) * expl((z - x) * (z + x) + R / S) / x;
+}
+
+long double erfl(long double x)
+{
+	long double r, s, z, y;
+	union ldshape u = {x};
+	uint32_t ix = (u.i.se & 0x7fffU)<<16 | u.i.m>>48;
+	int sign = u.i.se >> 15;
+
+	if (ix >= 0x7fff0000)
+		/* erf(nan)=nan, erf(+-inf)=+-1 */
+		return 1 - 2*sign + 1/x;
+	if (ix < 0x3ffed800) {  /* |x| < 0.84375 */
+		if (ix < 0x3fde8000) {  /* |x| < 2**-33 */
+			return 0.125 * (8 * x + efx8 * x);  /* avoid underflow */
+		}
+		z = x * x;
+		r = pp[0] + z * (pp[1] +
+		     z * (pp[2] + z * (pp[3] + z * (pp[4] + z * pp[5]))));
+		s = qq[0] + z * (qq[1] +
+		     z * (qq[2] + z * (qq[3] + z * (qq[4] + z * (qq[5] + z)))));
+		y = r / s;
+		return x + x * y;
+	}
+	if (ix < 0x4001d555)  /* |x| < 6.6666259765625 */
+		y = 1 - erfc2(ix,x);
+	else
+		y = 1 - 0x1p-16382L;
+	return sign ? -y : y;
+}
+
+long double erfcl(long double x)
+{
+	long double r, s, z, y;
+	union ldshape u = {x};
+	uint32_t ix = (u.i.se & 0x7fffU)<<16 | u.i.m>>48;
+	int sign = u.i.se >> 15;
+
+	if (ix >= 0x7fff0000)
+		/* erfc(nan) = nan, erfc(+-inf) = 0,2 */
+		return 2*sign + 1/x;
+	if (ix < 0x3ffed800) {  /* |x| < 0.84375 */
+		if (ix < 0x3fbe0000)  /* |x| < 2**-65 */
+			return 1.0 - x;
+		z = x * x;
+		r = pp[0] + z * (pp[1] +
+		     z * (pp[2] + z * (pp[3] + z * (pp[4] + z * pp[5]))));
+		s = qq[0] + z * (qq[1] +
+		     z * (qq[2] + z * (qq[3] + z * (qq[4] + z * (qq[5] + z)))));
+		y = r / s;
+		if (ix < 0x3ffd8000) /* x < 1/4 */
+			return 1.0 - (x + x * y);
+		return 0.5 - (x - 0.5 + x * y);
+	}
+	if (ix < 0x4005d600)  /* |x| < 107 */
+		return sign ? 2 - erfc2(ix,x) : erfc2(ix,x);
+	y = 0x1p-16382L;
+	return sign ? 2 - y : y*y;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double erfl(long double x)
+{
+	return erf(x);
+}
+long double erfcl(long double x)
+{
+	return erfc(x);
+}
+#endif
diff --git a/src/math/exp.c b/src/math/exp.c
index e415396..b764d73 100644
--- a/src/math/exp.c
+++ b/src/math/exp.c
@@ -1,125 +1,134 @@
-/* exp(x)
- * Returns the exponential of x.
+/*
+ * Double-precision e^x function.
  *
- * Method
- *   1. Argument reduction:
- *      Reduce x to an r so that |r| <= 0.5*ln2 ~ 0.34658.
- *      Given x, find r and integer k such that
- *
- *               x = k*ln2 + r,  |r| <= 0.5*ln2.
- *
- *      Here r will be represented as r = hi-lo for better
- *      accuracy.
- *
- *   2. Approximation of exp(r) by a special rational function on
- *      the interval [0,0.34658]:
- *      Write
- *          R(r**2) = r*(exp(r)+1)/(exp(r)-1) = 2 + r*r/6 - r**4/360 + ...
- *      We use a special Remez algorithm on [0,0.34658] to generate
- *      a polynomial of degree 5 to approximate R. The maximum error
- *      of this polynomial approximation is bounded by 2**-59. In
- *      other words,
- *          R(z) ~ 2.0 + P1*z + P2*z**2 + P3*z**3 + P4*z**4 + P5*z**5
- *      (where z=r*r, and the values of P1 to P5 are listed below)
- *      and
- *          |                  5          |     -59
- *          | 2.0+P1*z+...+P5*z   -  R(z) | <= 2
- *          |                             |
- *      The computation of exp(r) thus becomes
- *                              2*r
- *              exp(r) = 1 + ----------
- *                            R(r) - r
- *                                 r*c(r)
- *                     = 1 + r + ----------- (for better accuracy)
- *                                2 - c(r)
- *      where
- *                              2       4             10
- *              c(r) = r - (P1*r  + P2*r  + ... + P5*r   ).
- *
- *   3. Scale back to obtain exp(x):
- *      From step 1, we have
- *         exp(x) = 2^k * exp(r)
- *
- * Special cases:
- *      exp(INF) is INF, exp(NaN) is NaN;
- *      exp(-INF) is 0, and
- *      for finite argument, only exp(0)=1 is exact.
- *
- * Accuracy:
- *      according to an error analysis, the error is always less than
- *      1 ulp (unit in the last place).
- *
- * Misc. info.
- *      For IEEE double
- *          if x >  709.782712893383973096 then exp(x) overflows
- *          if x < -745.133219101941108420 then exp(x) underflows
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
  */
 
 #include <math.h>
+#include <stdint.h>
+#include "libm.h"
+#include "exp_data.h"
 
+#define N (1 << EXP_TABLE_BITS)
+#define InvLn2N __exp_data.invln2N
+#define NegLn2hiN __exp_data.negln2hiN
+#define NegLn2loN __exp_data.negln2loN
+#define Shift __exp_data.shift
+#define T __exp_data.tab
+#define C2 __exp_data.poly[5 - EXP_POLY_ORDER]
+#define C3 __exp_data.poly[6 - EXP_POLY_ORDER]
+#define C4 __exp_data.poly[7 - EXP_POLY_ORDER]
+#define C5 __exp_data.poly[8 - EXP_POLY_ORDER]
 
-static const double
-half[2] = {0.5,-0.5},
-ln2hi = 6.93147180369123816490e-01, /* 0x3fe62e42, 0xfee00000 */
-ln2lo = 1.90821492927058770002e-10, /* 0x3dea39ef, 0x35793c76 */
-invln2 = 1.44269504088896338700e+00, /* 0x3ff71547, 0x652b82fe */
-P1   =  1.66666666666666019037e-01, /* 0x3FC55555, 0x5555553E */
-P2   = -2.77777777770155933842e-03, /* 0xBF66C16C, 0x16BEBD93 */
-P3   =  6.61375632143793436117e-05, /* 0x3F11566A, 0xAF25DE2C */
-P4   = -1.65339022054652515390e-06, /* 0xBEBBBD41, 0xC5D26BF1 */
-P5   =  4.13813679705723846039e-08; /* 0x3E663769, 0x72BEA4D0 */
+/* Handle cases that may overflow or underflow when computing the result that
+   is scale*(1+TMP) without intermediate rounding.  The bit representation of
+   scale is in SBITS, however it has a computed exponent that may have
+   overflown into the sign bit so that needs to be adjusted before using it as
+   a double.  (int32_t)KI is the k used in the argument reduction and exponent
+   adjustment of scale, positive k here means the result may overflow and
+   negative k means the result may underflow.  */
+static inline double specialcase(double_t tmp, uint64_t sbits, uint64_t ki)
+{
+	double_t scale, y;
 
-double exp(double x)
+	if ((ki & 0x80000000) == 0) {
+		/* k > 0, the exponent of scale might have overflowed by <= 460.  */
+		sbits -= 1009ull << 52;
+		scale = asdouble(sbits);
+		y = 0x1p1009 * (scale + scale * tmp);
+		return eval_as_double(y);
+	}
+	/* k < 0, need special care in the subnormal range.  */
+	sbits += 1022ull << 52;
+	scale = asdouble(sbits);
+	y = scale + scale * tmp;
+	if (y < 1.0) {
+		/* Round y to the right precision before scaling it into the subnormal
+		 range to avoid double rounding that can cause 0.5+E/2 ulp error where
+		 E is the worst-case ulp error outside the subnormal range.  So this
+		 is only useful if the goal is better than 1 ulp worst-case error.  */
+		double_t hi, lo;
+		lo = scale - y + scale * tmp;
+		hi = 1.0 + y;
+		lo = 1.0 - hi + y + lo;
+		y = eval_as_double(hi + lo) - 1.0;
+		/* Avoid -0.0 with downward rounding.  */
+		if (WANT_ROUNDING && y == 0.0)
+			y = 0.0;
+		/* The underflow exception needs to be signaled explicitly.  */
+		fp_force_eval(fp_barrier(0x1p-1022) * 0x1p-1022);
+	}
+	y = 0x1p-1022 * y;
+	return eval_as_double(y);
+}
+
+/* Top 12 bits of a double (sign and exponent bits).  */
+static inline uint32_t top12(double x)
 {
-	double_t hi, lo, c, xx, y;
-	int k, sign;
-	uint32_t hx;
+	return asuint64(x) >> 52;
+}
 
-	GET_HIGH_WORD(hx, x);
-	sign = hx>>31;
-	hx &= 0x7fffffff;  /* high word of |x| */
+double exp(double x)
+{
+	uint32_t abstop;
+	uint64_t ki, idx, top, sbits;
+	double_t kd, z, r, r2, scale, tail, tmp;
 
-	/* special cases */
-	if (hx >= 0x4086232b) {  /* if |x| >= 708.39... */
-		if (isnan(x))
-			return x;
-		if (x > 709.782712893383973096) {
-			/* overflow if x!=inf */
-			x *= 0x1p1023;
-			return x;
-		}
-		if (x < -708.39641853226410622) {
-			/* underflow if x!=-inf */
-			FORCE_EVAL((float)(-0x1p-149/x));
-			if (x < -745.13321910194110842)
-				return 0;
+	abstop = top12(x) & 0x7ff;
+	if (predict_false(abstop - top12(0x1p-54) >= top12(512.0) - top12(0x1p-54))) {
+		if (abstop - top12(0x1p-54) >= 0x80000000)
+			/* Avoid spurious underflow for tiny x.  */
+			/* Note: 0 is common input.  */
+			return WANT_ROUNDING ? 1.0 + x : 1.0;
+		if (abstop >= top12(1024.0)) {
+			if (asuint64(x) == asuint64(-INFINITY))
+				return 0.0;
+			if (abstop >= top12(INFINITY))
+				return 1.0 + x;
+			if (asuint64(x) >> 63)
+				return __math_uflow(0);
+			else
+				return __math_oflow(0);
 		}
+		/* Large x is special cased below.  */
+		abstop = 0;
 	}
 
-	/* argument reduction */
-	if (hx > 0x3fd62e42) {  /* if |x| > 0.5 ln2 */
-		if (hx >= 0x3ff0a2b2)  /* if |x| >= 1.5 ln2 */
-			k = (int)(invln2*x + half[sign]);
-		else
-			k = 1 - sign - sign;
-		hi = x - k*ln2hi;  /* k*ln2hi is exact here */
-		lo = k*ln2lo;
-		x = hi - lo;
-	} else if (hx > 0x3e300000)  {  /* if |x| > 2**-28 */
-		k = 0;
-		hi = x;
-		lo = 0;
-	} else {
-		/* inexact if x!=0 */
-		FORCE_EVAL(0x1p1023 + x);
-		return 1 + x;
-	}
-
-	/* x is now in primary range */
-	xx = x*x;
-	c = x - xx*(P1+xx*(P2+xx*(P3+xx*(P4+xx*P5))));
-	y = 1 + (x*c/(2-c) - lo + hi);
-	if (k == 0)
-		return y;
-	return scalbn(y, k);
+	/* exp(x) = 2^(k/N) * exp(r), with exp(r) in [2^(-1/2N),2^(1/2N)].  */
+	/* x = ln2/N*k + r, with int k and r in [-ln2/2N, ln2/2N].  */
+	z = InvLn2N * x;
+#if TOINT_INTRINSICS
+	kd = roundtoint(z);
+	ki = converttoint(z);
+#elif EXP_USE_TOINT_NARROW
+	/* z - kd is in [-0.5-2^-16, 0.5] in all rounding modes.  */
+	kd = eval_as_double(z + Shift);
+	ki = asuint64(kd) >> 16;
+	kd = (double_t)(int32_t)ki;
+#else
+	/* z - kd is in [-1, 1] in non-nearest rounding modes.  */
+	kd = eval_as_double(z + Shift);
+	ki = asuint64(kd);
+	kd -= Shift;
+#endif
+	r = x + kd * NegLn2hiN + kd * NegLn2loN;
+	/* 2^(k/N) ~= scale * (1 + tail).  */
+	idx = 2 * (ki % N);
+	top = ki << (52 - EXP_TABLE_BITS);
+	tail = asdouble(T[idx]);
+	/* This is only a valid scale when -1023*N < k < 1024*N.  */
+	sbits = T[idx + 1] + top;
+	/* exp(x) = 2^(k/N) * exp(r) ~= scale + scale * (tail + exp(r) - 1).  */
+	/* Evaluation is optimized assuming superscalar pipelined execution.  */
+	r2 = r * r;
+	/* Without fma the worst case error is 0.25/N ulp larger.  */
+	/* Worst case error is less than 0.5+1.11/N+(abs poly error * 2^53) ulp.  */
+	tmp = tail + r + r2 * (C2 + r * C3) + r2 * r2 * (C4 + r * C5);
+	if (predict_false(abstop == 0))
+		return specialcase(tmp, sbits, ki);
+	scale = asdouble(sbits);
+	/* Note: tmp == 0 or |tmp| > 2^-200 and scale > 2^-739, so there
+	   is no spurious underflow here even without fma.  */
+	return eval_as_double(scale + scale * tmp);
 }
diff --git a/src/math/exp10.c b/src/math/exp10.c
new file mode 100644
index 0000000..26899eb
--- /dev/null
+++ b/src/math/exp10.c
@@ -0,0 +1,24 @@
+#define _GNU_SOURCE
+#include <math.h>
+#include <stdint.h>
+
+double exp10(double x)
+{
+	static const double p10[] = {
+		1e-15, 1e-14, 1e-13, 1e-12, 1e-11, 1e-10,
+		1e-9, 1e-8, 1e-7, 1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1,
+		1, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
+		1e10, 1e11, 1e12, 1e13, 1e14, 1e15
+	};
+	double n, y = modf(x, &n);
+	union {double f; uint64_t i;} u = {n};
+	/* fabs(n) < 16 without raising invalid on nan */
+	if ((u.i>>52 & 0x7ff) < 0x3ff+4) {
+		if (!y) return p10[(int)n+15];
+		y = exp2(3.32192809488736234787031942948939 * y);
+		return y * p10[(int)n+15];
+	}
+	return pow(10.0, x);
+}
+
+weak_alias(exp10, pow10);
diff --git a/src/math/exp10f.c b/src/math/exp10f.c
new file mode 100644
index 0000000..d009f0a
--- /dev/null
+++ b/src/math/exp10f.c
@@ -0,0 +1,22 @@
+#define _GNU_SOURCE
+#include <math.h>
+#include <stdint.h>
+
+float exp10f(float x)
+{
+	static const float p10[] = {
+		1e-7f, 1e-6f, 1e-5f, 1e-4f, 1e-3f, 1e-2f, 1e-1f,
+		1, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7
+	};
+	float n, y = modff(x, &n);
+	union {float f; uint32_t i;} u = {n};
+	/* fabsf(n) < 8 without raising invalid on nan */
+	if ((u.i>>23 & 0xff) < 0x7f+3) {
+		if (!y) return p10[(int)n+7];
+		y = exp2f(3.32192809488736234787031942948939f * y);
+		return y * p10[(int)n+7];
+	}
+	return exp2(3.32192809488736234787031942948939 * x);
+}
+
+weak_alias(exp10f, pow10f);
diff --git a/src/math/exp10l.c b/src/math/exp10l.c
new file mode 100644
index 0000000..f3da1a0
--- /dev/null
+++ b/src/math/exp10l.c
@@ -0,0 +1,32 @@
+#define _GNU_SOURCE
+#include <float.h>
+#include <math.h>
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double exp10l(long double x)
+{
+	return exp10(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double exp10l(long double x)
+{
+	static const long double p10[] = {
+		1e-15L, 1e-14L, 1e-13L, 1e-12L, 1e-11L, 1e-10L,
+		1e-9L, 1e-8L, 1e-7L, 1e-6L, 1e-5L, 1e-4L, 1e-3L, 1e-2L, 1e-1L,
+		1, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
+		1e10, 1e11, 1e12, 1e13, 1e14, 1e15
+	};
+	long double n, y = modfl(x, &n);
+	union ldshape u = {n};
+	/* fabsl(n) < 16 without raising invalid on nan */
+	if ((u.i.se & 0x7fff) < 0x3fff+4) {
+		if (!y) return p10[(int)n+15];
+		y = exp2l(3.32192809488736234787031942948939L * y);
+		return y * p10[(int)n+15];
+	}
+	return powl(10.0, x);
+}
+#endif
+
+weak_alias(exp10l, pow10l);
diff --git a/src/math/exp2.c b/src/math/exp2.c
index 2f75530..4c2560d 100644
--- a/src/math/exp2.c
+++ b/src/math/exp2.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define TBLSIZE 256
 
diff --git a/src/math/exp2f.c b/src/math/exp2f.c
index 6cf4272..9ce1e10 100644
--- a/src/math/exp2f.c
+++ b/src/math/exp2f.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define TBLSIZE 16
 
diff --git a/src/math/exp2l.c b/src/math/exp2l.c
new file mode 100644
index 0000000..3565c1e
--- /dev/null
+++ b/src/math/exp2l.c
@@ -0,0 +1,619 @@
+/* origin: FreeBSD /usr/src/lib/msun/ld80/s_exp2l.c and /usr/src/lib/msun/ld128/s_exp2l.c */
+/*-
+ * Copyright (c) 2005-2008 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double exp2l(long double x)
+{
+	return exp2(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+#define TBLBITS 7
+#define TBLSIZE (1 << TBLBITS)
+
+static const double
+redux = 0x1.8p63 / TBLSIZE,
+P1    = 0x1.62e42fefa39efp-1,
+P2    = 0x1.ebfbdff82c58fp-3,
+P3    = 0x1.c6b08d7049fap-5,
+P4    = 0x1.3b2ab6fba4da5p-7,
+P5    = 0x1.5d8804780a736p-10,
+P6    = 0x1.430918835e33dp-13;
+
+static const double tbl[TBLSIZE * 2] = {
+	0x1.6a09e667f3bcdp-1,   -0x1.bdd3413b2648p-55,
+	0x1.6c012750bdabfp-1,   -0x1.2895667ff0cp-57,
+	0x1.6dfb23c651a2fp-1,   -0x1.bbe3a683c88p-58,
+	0x1.6ff7df9519484p-1,   -0x1.83c0f25860fp-56,
+	0x1.71f75e8ec5f74p-1,   -0x1.16e4786887bp-56,
+	0x1.73f9a48a58174p-1,   -0x1.0a8d96c65d5p-55,
+	0x1.75feb564267c9p-1,   -0x1.0245957316ep-55,
+	0x1.780694fde5d3fp-1,    0x1.866b80a0216p-55,
+	0x1.7a11473eb0187p-1,   -0x1.41577ee0499p-56,
+	0x1.7c1ed0130c132p-1,    0x1.f124cd1164ep-55,
+	0x1.7e2f336cf4e62p-1,    0x1.05d02ba157ap-57,
+	0x1.80427543e1a12p-1,   -0x1.27c86626d97p-55,
+	0x1.82589994cce13p-1,   -0x1.d4c1dd41533p-55,
+	0x1.8471a4623c7adp-1,   -0x1.8d684a341cep-56,
+	0x1.868d99b4492edp-1,   -0x1.fc6f89bd4f68p-55,
+	0x1.88ac7d98a6699p-1,    0x1.994c2f37cb5p-55,
+	0x1.8ace5422aa0dbp-1,    0x1.6e9f156864bp-55,
+	0x1.8cf3216b5448cp-1,   -0x1.0d55e32e9e4p-57,
+	0x1.8f1ae99157736p-1,    0x1.5cc13a2e397p-56,
+	0x1.9145b0b91ffc6p-1,   -0x1.dd6792e5825p-55,
+	0x1.93737b0cdc5e5p-1,   -0x1.75fc781b58p-58,
+	0x1.95a44cbc8520fp-1,   -0x1.64b7c96a5fp-57,
+	0x1.97d829fde4e5p-1,    -0x1.d185b7c1b86p-55,
+	0x1.9a0f170ca07bap-1,   -0x1.173bd91cee6p-55,
+	0x1.9c49182a3f09p-1,     0x1.c7c46b071f2p-57,
+	0x1.9e86319e32323p-1,    0x1.824ca78e64cp-57,
+	0x1.a0c667b5de565p-1,   -0x1.359495d1cd5p-55,
+	0x1.a309bec4a2d33p-1,    0x1.6305c7ddc368p-55,
+	0x1.a5503b23e255dp-1,   -0x1.d2f6edb8d42p-55,
+	0x1.a799e1330b358p-1,    0x1.bcb7ecac564p-55,
+	0x1.a9e6b5579fdbfp-1,    0x1.0fac90ef7fdp-55,
+	0x1.ac36bbfd3f37ap-1,   -0x1.f9234cae76dp-56,
+	0x1.ae89f995ad3adp-1,    0x1.7a1cd345dcc8p-55,
+	0x1.b0e07298db666p-1,   -0x1.bdef54c80e4p-55,
+	0x1.b33a2b84f15fbp-1,   -0x1.2805e3084d8p-58,
+	0x1.b59728de5593ap-1,   -0x1.c71dfbbba6ep-55,
+	0x1.b7f76f2fb5e47p-1,   -0x1.5584f7e54acp-57,
+	0x1.ba5b030a1064ap-1,   -0x1.efcd30e5429p-55,
+	0x1.bcc1e904bc1d2p-1,    0x1.23dd07a2d9fp-56,
+	0x1.bf2c25bd71e09p-1,   -0x1.efdca3f6b9c8p-55,
+	0x1.c199bdd85529cp-1,    0x1.11065895049p-56,
+	0x1.c40ab5fffd07ap-1,    0x1.b4537e083c6p-55,
+	0x1.c67f12e57d14bp-1,    0x1.2884dff483c8p-55,
+	0x1.c8f6d9406e7b5p-1,    0x1.1acbc48805cp-57,
+	0x1.cb720dcef9069p-1,    0x1.503cbd1e94ap-57,
+	0x1.cdf0b555dc3fap-1,   -0x1.dd83b53829dp-56,
+	0x1.d072d4a07897cp-1,   -0x1.cbc3743797a8p-55,
+	0x1.d2f87080d89f2p-1,   -0x1.d487b719d858p-55,
+	0x1.d5818dcfba487p-1,    0x1.2ed02d75b37p-56,
+	0x1.d80e316c98398p-1,   -0x1.11ec18bedep-55,
+	0x1.da9e603db3285p-1,    0x1.c2300696db5p-55,
+	0x1.dd321f301b46p-1,     0x1.2da5778f019p-55,
+	0x1.dfc97337b9b5fp-1,   -0x1.1a5cd4f184b8p-55,
+	0x1.e264614f5a129p-1,   -0x1.7b627817a148p-55,
+	0x1.e502ee78b3ff6p-1,    0x1.39e8980a9cdp-56,
+	0x1.e7a51fbc74c83p-1,    0x1.2d522ca0c8ep-55,
+	0x1.ea4afa2a490dap-1,   -0x1.e9c23179c288p-55,
+	0x1.ecf482d8e67f1p-1,   -0x1.c93f3b411ad8p-55,
+	0x1.efa1bee615a27p-1,    0x1.dc7f486a4b68p-55,
+	0x1.f252b376bba97p-1,    0x1.3a1a5bf0d8e8p-55,
+	0x1.f50765b6e454p-1,     0x1.9d3e12dd8a18p-55,
+	0x1.f7bfdad9cbe14p-1,   -0x1.dbb12d00635p-55,
+	0x1.fa7c1819e90d8p-1,    0x1.74853f3a593p-56,
+	0x1.fd3c22b8f71f1p-1,    0x1.2eb74966578p-58,
+	0x1p+0,                  0x0p+0,
+	0x1.0163da9fb3335p+0,    0x1.b61299ab8cd8p-54,
+	0x1.02c9a3e778061p+0,   -0x1.19083535b08p-56,
+	0x1.04315e86e7f85p+0,   -0x1.0a31c1977c98p-54,
+	0x1.059b0d3158574p+0,    0x1.d73e2a475b4p-55,
+	0x1.0706b29ddf6dep+0,   -0x1.c91dfe2b13cp-55,
+	0x1.0874518759bc8p+0,    0x1.186be4bb284p-57,
+	0x1.09e3ecac6f383p+0,    0x1.14878183161p-54,
+	0x1.0b5586cf9890fp+0,    0x1.8a62e4adc61p-54,
+	0x1.0cc922b7247f7p+0,    0x1.01edc16e24f8p-54,
+	0x1.0e3ec32d3d1a2p+0,    0x1.03a1727c58p-59,
+	0x1.0fb66affed31bp+0,   -0x1.b9bedc44ebcp-57,
+	0x1.11301d0125b51p+0,   -0x1.6c51039449bp-54,
+	0x1.12abdc06c31ccp+0,   -0x1.1b514b36ca8p-58,
+	0x1.1429aaea92dep+0,    -0x1.32fbf9af1368p-54,
+	0x1.15a98c8a58e51p+0,    0x1.2406ab9eeabp-55,
+	0x1.172b83c7d517bp+0,   -0x1.19041b9d78ap-55,
+	0x1.18af9388c8deap+0,   -0x1.11023d1970f8p-54,
+	0x1.1a35beb6fcb75p+0,    0x1.e5b4c7b4969p-55,
+	0x1.1bbe084045cd4p+0,   -0x1.95386352ef6p-54,
+	0x1.1d4873168b9aap+0,    0x1.e016e00a264p-54,
+	0x1.1ed5022fcd91dp+0,   -0x1.1df98027bb78p-54,
+	0x1.2063b88628cd6p+0,    0x1.dc775814a85p-55,
+	0x1.21f49917ddc96p+0,    0x1.2a97e9494a6p-55,
+	0x1.2387a6e756238p+0,    0x1.9b07eb6c7058p-54,
+	0x1.251ce4fb2a63fp+0,    0x1.ac155bef4f5p-55,
+	0x1.26b4565e27cddp+0,    0x1.2bd339940eap-55,
+	0x1.284dfe1f56381p+0,   -0x1.a4c3a8c3f0d8p-54,
+	0x1.29e9df51fdee1p+0,    0x1.612e8afad12p-55,
+	0x1.2b87fd0dad99p+0,    -0x1.10adcd6382p-59,
+	0x1.2d285a6e4030bp+0,    0x1.0024754db42p-54,
+	0x1.2ecafa93e2f56p+0,    0x1.1ca0f45d524p-56,
+	0x1.306fe0a31b715p+0,    0x1.6f46ad23183p-55,
+	0x1.32170fc4cd831p+0,    0x1.a9ce78e1804p-55,
+	0x1.33c08b26416ffp+0,    0x1.327218436598p-54,
+	0x1.356c55f929ff1p+0,   -0x1.b5cee5c4e46p-55,
+	0x1.371a7373aa9cbp+0,   -0x1.63aeabf42ebp-54,
+	0x1.38cae6d05d866p+0,   -0x1.e958d3c99048p-54,
+	0x1.3a7db34e59ff7p+0,   -0x1.5e436d661f6p-56,
+	0x1.3c32dc313a8e5p+0,   -0x1.efff8375d2ap-54,
+	0x1.3dea64c123422p+0,    0x1.ada0911f09fp-55,
+	0x1.3fa4504ac801cp+0,   -0x1.7d023f956fap-54,
+	0x1.4160a21f72e2ap+0,   -0x1.ef3691c309p-58,
+	0x1.431f5d950a897p+0,   -0x1.1c7dde35f7ap-55,
+	0x1.44e086061892dp+0,    0x1.89b7a04ef8p-59,
+	0x1.46a41ed1d0057p+0,    0x1.c944bd1648a8p-54,
+	0x1.486a2b5c13cdp+0,     0x1.3c1a3b69062p-56,
+	0x1.4a32af0d7d3dep+0,    0x1.9cb62f3d1be8p-54,
+	0x1.4bfdad5362a27p+0,    0x1.d4397afec42p-56,
+	0x1.4dcb299fddd0dp+0,    0x1.8ecdbbc6a78p-54,
+	0x1.4f9b2769d2ca7p+0,   -0x1.4b309d25958p-54,
+	0x1.516daa2cf6642p+0,   -0x1.f768569bd94p-55,
+	0x1.5342b569d4f82p+0,   -0x1.07abe1db13dp-55,
+	0x1.551a4ca5d920fp+0,   -0x1.d689cefede6p-55,
+	0x1.56f4736b527dap+0,    0x1.9bb2c011d938p-54,
+	0x1.58d12d497c7fdp+0,    0x1.295e15b9a1ep-55,
+	0x1.5ab07dd485429p+0,    0x1.6324c0546478p-54,
+	0x1.5c9268a5946b7p+0,    0x1.c4b1b81698p-60,
+	0x1.5e76f15ad2148p+0,    0x1.ba6f93080e68p-54,
+	0x1.605e1b976dc09p+0,   -0x1.3e2429b56de8p-54,
+	0x1.6247eb03a5585p+0,   -0x1.383c17e40b48p-54,
+	0x1.6434634ccc32p+0,    -0x1.c483c759d89p-55,
+	0x1.6623882552225p+0,   -0x1.bb60987591cp-54,
+	0x1.68155d44ca973p+0,    0x1.038ae44f74p-57,
+};
+
+/*
+ * exp2l(x): compute the base 2 exponential of x
+ *
+ * Accuracy: Peak error < 0.511 ulp.
+ *
+ * Method: (equally-spaced tables)
+ *
+ *   Reduce x:
+ *     x = 2**k + y, for integer k and |y| <= 1/2.
+ *     Thus we have exp2l(x) = 2**k * exp2(y).
+ *
+ *   Reduce y:
+ *     y = i/TBLSIZE + z for integer i near y * TBLSIZE.
+ *     Thus we have exp2(y) = exp2(i/TBLSIZE) * exp2(z),
+ *     with |z| <= 2**-(TBLBITS+1).
+ *
+ *   We compute exp2(i/TBLSIZE) via table lookup and exp2(z) via a
+ *   degree-6 minimax polynomial with maximum error under 2**-69.
+ *   The table entries each have 104 bits of accuracy, encoded as
+ *   a pair of double precision values.
+ */
+long double exp2l(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	long double r, z;
+	uint32_t i0;
+	union {uint32_t u; int32_t i;} k;
+
+	/* Filter out exceptional cases. */
+	if (e >= 0x3fff + 13) {  /* |x| >= 8192 or x is NaN */
+		if (u.i.se >= 0x3fff + 14 && u.i.se >> 15 == 0)
+			/* overflow */
+			return x * 0x1p16383L;
+		if (e == 0x7fff)  /* -inf or -nan */
+			return -1/x;
+		if (x < -16382) {
+			if (x <= -16446 || x - 0x1p63 + 0x1p63 != x)
+				/* underflow */
+				FORCE_EVAL((float)(-0x1p-149/x));
+			if (x <= -16446)
+				return 0;
+		}
+	} else if (e < 0x3fff - 64) {
+		return 1 + x;
+	}
+
+	/*
+	 * Reduce x, computing z, i0, and k. The low bits of x + redux
+	 * contain the 16-bit integer part of the exponent (k) followed by
+	 * TBLBITS fractional bits (i0). We use bit tricks to extract these
+	 * as integers, then set z to the remainder.
+	 *
+	 * Example: Suppose x is 0xabc.123456p0 and TBLBITS is 8.
+	 * Then the low-order word of x + redux is 0x000abc12,
+	 * We split this into k = 0xabc and i0 = 0x12 (adjusted to
+	 * index into the table), then we compute z = 0x0.003456p0.
+	 */
+	u.f = x + redux;
+	i0 = u.i.m + TBLSIZE / 2;
+	k.u = i0 / TBLSIZE * TBLSIZE;
+	k.i /= TBLSIZE;
+	i0 %= TBLSIZE;
+	u.f -= redux;
+	z = x - u.f;
+
+	/* Compute r = exp2l(y) = exp2lt[i0] * p(z). */
+	long double t_hi = tbl[2*i0];
+	long double t_lo = tbl[2*i0 + 1];
+	/* XXX This gives > 1 ulp errors outside of FE_TONEAREST mode */
+	r = t_lo + (t_hi + t_lo) * z * (P1 + z * (P2 + z * (P3 + z * (P4
+	     + z * (P5 + z * P6))))) + t_hi;
+
+	return scalbnl(r, k.i);
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+#define TBLBITS 7
+#define TBLSIZE (1 << TBLBITS)
+
+static const long double
+    P1        = 0x1.62e42fefa39ef35793c7673007e6p-1L,
+    P2        = 0x1.ebfbdff82c58ea86f16b06ec9736p-3L,
+    P3        = 0x1.c6b08d704a0bf8b33a762bad3459p-5L,
+    P4        = 0x1.3b2ab6fba4e7729ccbbe0b4f3fc2p-7L,
+    P5        = 0x1.5d87fe78a67311071dee13fd11d9p-10L,
+    P6        = 0x1.430912f86c7876f4b663b23c5fe5p-13L;
+
+static const double
+    P7        = 0x1.ffcbfc588b041p-17,
+    P8        = 0x1.62c0223a5c7c7p-20,
+    P9        = 0x1.b52541ff59713p-24,
+    P10       = 0x1.e4cf56a391e22p-28,
+    redux     = 0x1.8p112 / TBLSIZE;
+
+static const long double tbl[TBLSIZE] = {
+	0x1.6a09e667f3bcc908b2fb1366dfeap-1L,
+	0x1.6c012750bdabeed76a99800f4edep-1L,
+	0x1.6dfb23c651a2ef220e2cbe1bc0d4p-1L,
+	0x1.6ff7df9519483cf87e1b4f3e1e98p-1L,
+	0x1.71f75e8ec5f73dd2370f2ef0b148p-1L,
+	0x1.73f9a48a58173bd5c9a4e68ab074p-1L,
+	0x1.75feb564267c8bf6e9aa33a489a8p-1L,
+	0x1.780694fde5d3f619ae02808592a4p-1L,
+	0x1.7a11473eb0186d7d51023f6ccb1ap-1L,
+	0x1.7c1ed0130c1327c49334459378dep-1L,
+	0x1.7e2f336cf4e62105d02ba1579756p-1L,
+	0x1.80427543e1a11b60de67649a3842p-1L,
+	0x1.82589994cce128acf88afab34928p-1L,
+	0x1.8471a4623c7acce52f6b97c6444cp-1L,
+	0x1.868d99b4492ec80e41d90ac2556ap-1L,
+	0x1.88ac7d98a669966530bcdf2d4cc0p-1L,
+	0x1.8ace5422aa0db5ba7c55a192c648p-1L,
+	0x1.8cf3216b5448bef2aa1cd161c57ap-1L,
+	0x1.8f1ae991577362b982745c72eddap-1L,
+	0x1.9145b0b91ffc588a61b469f6b6a0p-1L,
+	0x1.93737b0cdc5e4f4501c3f2540ae8p-1L,
+	0x1.95a44cbc8520ee9b483695a0e7fep-1L,
+	0x1.97d829fde4e4f8b9e920f91e8eb6p-1L,
+	0x1.9a0f170ca07b9ba3109b8c467844p-1L,
+	0x1.9c49182a3f0901c7c46b071f28dep-1L,
+	0x1.9e86319e323231824ca78e64c462p-1L,
+	0x1.a0c667b5de564b29ada8b8cabbacp-1L,
+	0x1.a309bec4a2d3358c171f770db1f4p-1L,
+	0x1.a5503b23e255c8b424491caf88ccp-1L,
+	0x1.a799e1330b3586f2dfb2b158f31ep-1L,
+	0x1.a9e6b5579fdbf43eb243bdff53a2p-1L,
+	0x1.ac36bbfd3f379c0db966a3126988p-1L,
+	0x1.ae89f995ad3ad5e8734d17731c80p-1L,
+	0x1.b0e07298db66590842acdfc6fb4ep-1L,
+	0x1.b33a2b84f15faf6bfd0e7bd941b0p-1L,
+	0x1.b59728de559398e3881111648738p-1L,
+	0x1.b7f76f2fb5e46eaa7b081ab53ff6p-1L,
+	0x1.ba5b030a10649840cb3c6af5b74cp-1L,
+	0x1.bcc1e904bc1d2247ba0f45b3d06cp-1L,
+	0x1.bf2c25bd71e088408d7025190cd0p-1L,
+	0x1.c199bdd85529c2220cb12a0916bap-1L,
+	0x1.c40ab5fffd07a6d14df820f17deap-1L,
+	0x1.c67f12e57d14b4a2137fd20f2a26p-1L,
+	0x1.c8f6d9406e7b511acbc48805c3f6p-1L,
+	0x1.cb720dcef90691503cbd1e949d0ap-1L,
+	0x1.cdf0b555dc3f9c44f8958fac4f12p-1L,
+	0x1.d072d4a07897b8d0f22f21a13792p-1L,
+	0x1.d2f87080d89f18ade123989ea50ep-1L,
+	0x1.d5818dcfba48725da05aeb66dff8p-1L,
+	0x1.d80e316c98397bb84f9d048807a0p-1L,
+	0x1.da9e603db3285708c01a5b6d480cp-1L,
+	0x1.dd321f301b4604b695de3c0630c0p-1L,
+	0x1.dfc97337b9b5eb968cac39ed284cp-1L,
+	0x1.e264614f5a128a12761fa17adc74p-1L,
+	0x1.e502ee78b3ff6273d130153992d0p-1L,
+	0x1.e7a51fbc74c834b548b2832378a4p-1L,
+	0x1.ea4afa2a490d9858f73a18f5dab4p-1L,
+	0x1.ecf482d8e67f08db0312fb949d50p-1L,
+	0x1.efa1bee615a27771fd21a92dabb6p-1L,
+	0x1.f252b376bba974e8696fc3638f24p-1L,
+	0x1.f50765b6e4540674f84b762861a6p-1L,
+	0x1.f7bfdad9cbe138913b4bfe72bd78p-1L,
+	0x1.fa7c1819e90d82e90a7e74b26360p-1L,
+	0x1.fd3c22b8f71f10975ba4b32bd006p-1L,
+	0x1.0000000000000000000000000000p+0L,
+	0x1.0163da9fb33356d84a66ae336e98p+0L,
+	0x1.02c9a3e778060ee6f7caca4f7a18p+0L,
+	0x1.04315e86e7f84bd738f9a20da442p+0L,
+	0x1.059b0d31585743ae7c548eb68c6ap+0L,
+	0x1.0706b29ddf6ddc6dc403a9d87b1ep+0L,
+	0x1.0874518759bc808c35f25d942856p+0L,
+	0x1.09e3ecac6f3834521e060c584d5cp+0L,
+	0x1.0b5586cf9890f6298b92b7184200p+0L,
+	0x1.0cc922b7247f7407b705b893dbdep+0L,
+	0x1.0e3ec32d3d1a2020742e4f8af794p+0L,
+	0x1.0fb66affed31af232091dd8a169ep+0L,
+	0x1.11301d0125b50a4ebbf1aed9321cp+0L,
+	0x1.12abdc06c31cbfb92bad324d6f84p+0L,
+	0x1.1429aaea92ddfb34101943b2588ep+0L,
+	0x1.15a98c8a58e512480d573dd562aep+0L,
+	0x1.172b83c7d517adcdf7c8c50eb162p+0L,
+	0x1.18af9388c8de9bbbf70b9a3c269cp+0L,
+	0x1.1a35beb6fcb753cb698f692d2038p+0L,
+	0x1.1bbe084045cd39ab1e72b442810ep+0L,
+	0x1.1d4873168b9aa7805b8028990be8p+0L,
+	0x1.1ed5022fcd91cb8819ff61121fbep+0L,
+	0x1.2063b88628cd63b8eeb0295093f6p+0L,
+	0x1.21f49917ddc962552fd29294bc20p+0L,
+	0x1.2387a6e75623866c1fadb1c159c0p+0L,
+	0x1.251ce4fb2a63f3582ab7de9e9562p+0L,
+	0x1.26b4565e27cdd257a673281d3068p+0L,
+	0x1.284dfe1f5638096cf15cf03c9fa0p+0L,
+	0x1.29e9df51fdee12c25d15f5a25022p+0L,
+	0x1.2b87fd0dad98ffddea46538fca24p+0L,
+	0x1.2d285a6e4030b40091d536d0733ep+0L,
+	0x1.2ecafa93e2f5611ca0f45d5239a4p+0L,
+	0x1.306fe0a31b7152de8d5a463063bep+0L,
+	0x1.32170fc4cd8313539cf1c3009330p+0L,
+	0x1.33c08b26416ff4c9c8610d96680ep+0L,
+	0x1.356c55f929ff0c94623476373be4p+0L,
+	0x1.371a7373aa9caa7145502f45452ap+0L,
+	0x1.38cae6d05d86585a9cb0d9bed530p+0L,
+	0x1.3a7db34e59ff6ea1bc9299e0a1fep+0L,
+	0x1.3c32dc313a8e484001f228b58cf0p+0L,
+	0x1.3dea64c12342235b41223e13d7eep+0L,
+	0x1.3fa4504ac801ba0bf701aa417b9cp+0L,
+	0x1.4160a21f72e29f84325b8f3dbacap+0L,
+	0x1.431f5d950a896dc704439410b628p+0L,
+	0x1.44e086061892d03136f409df0724p+0L,
+	0x1.46a41ed1d005772512f459229f0ap+0L,
+	0x1.486a2b5c13cd013c1a3b69062f26p+0L,
+	0x1.4a32af0d7d3de672d8bcf46f99b4p+0L,
+	0x1.4bfdad5362a271d4397afec42e36p+0L,
+	0x1.4dcb299fddd0d63b36ef1a9e19dep+0L,
+	0x1.4f9b2769d2ca6ad33d8b69aa0b8cp+0L,
+	0x1.516daa2cf6641c112f52c84d6066p+0L,
+	0x1.5342b569d4f81df0a83c49d86bf4p+0L,
+	0x1.551a4ca5d920ec52ec620243540cp+0L,
+	0x1.56f4736b527da66ecb004764e61ep+0L,
+	0x1.58d12d497c7fd252bc2b7343d554p+0L,
+	0x1.5ab07dd48542958c93015191e9a8p+0L,
+	0x1.5c9268a5946b701c4b1b81697ed4p+0L,
+	0x1.5e76f15ad21486e9be4c20399d12p+0L,
+	0x1.605e1b976dc08b076f592a487066p+0L,
+	0x1.6247eb03a5584b1f0fa06fd2d9eap+0L,
+	0x1.6434634ccc31fc76f8714c4ee122p+0L,
+	0x1.66238825522249127d9e29b92ea2p+0L,
+	0x1.68155d44ca973081c57227b9f69ep+0L,
+};
+
+static const float eps[TBLSIZE] = {
+	-0x1.5c50p-101,
+	-0x1.5d00p-106,
+	 0x1.8e90p-102,
+	-0x1.5340p-103,
+	 0x1.1bd0p-102,
+	-0x1.4600p-105,
+	-0x1.7a40p-104,
+	 0x1.d590p-102,
+	-0x1.d590p-101,
+	 0x1.b100p-103,
+	-0x1.0d80p-105,
+	 0x1.6b00p-103,
+	-0x1.9f00p-105,
+	 0x1.c400p-103,
+	 0x1.e120p-103,
+	-0x1.c100p-104,
+	-0x1.9d20p-103,
+	 0x1.a800p-108,
+	 0x1.4c00p-106,
+	-0x1.9500p-106,
+	 0x1.6900p-105,
+	-0x1.29d0p-100,
+	 0x1.4c60p-103,
+	 0x1.13a0p-102,
+	-0x1.5b60p-103,
+	-0x1.1c40p-103,
+	 0x1.db80p-102,
+	 0x1.91a0p-102,
+	 0x1.dc00p-105,
+	 0x1.44c0p-104,
+	 0x1.9710p-102,
+	 0x1.8760p-103,
+	-0x1.a720p-103,
+	 0x1.ed20p-103,
+	-0x1.49c0p-102,
+	-0x1.e000p-111,
+	 0x1.86a0p-103,
+	 0x1.2b40p-103,
+	-0x1.b400p-108,
+	 0x1.1280p-99,
+	-0x1.02d8p-102,
+	-0x1.e3d0p-103,
+	-0x1.b080p-105,
+	-0x1.f100p-107,
+	-0x1.16c0p-105,
+	-0x1.1190p-103,
+	-0x1.a7d2p-100,
+	 0x1.3450p-103,
+	-0x1.67c0p-105,
+	 0x1.4b80p-104,
+	-0x1.c4e0p-103,
+	 0x1.6000p-108,
+	-0x1.3f60p-105,
+	 0x1.93f0p-104,
+	 0x1.5fe0p-105,
+	 0x1.6f80p-107,
+	-0x1.7600p-106,
+	 0x1.21e0p-106,
+	-0x1.3a40p-106,
+	-0x1.40c0p-104,
+	-0x1.9860p-105,
+	-0x1.5d40p-108,
+	-0x1.1d70p-106,
+	 0x1.2760p-105,
+	 0x0.0000p+0,
+	 0x1.21e2p-104,
+	-0x1.9520p-108,
+	-0x1.5720p-106,
+	-0x1.4810p-106,
+	-0x1.be00p-109,
+	 0x1.0080p-105,
+	-0x1.5780p-108,
+	-0x1.d460p-105,
+	-0x1.6140p-105,
+	 0x1.4630p-104,
+	 0x1.ad50p-103,
+	 0x1.82e0p-105,
+	 0x1.1d3cp-101,
+	 0x1.6100p-107,
+	 0x1.ec30p-104,
+	 0x1.f200p-108,
+	 0x1.0b40p-103,
+	 0x1.3660p-102,
+	 0x1.d9d0p-103,
+	-0x1.02d0p-102,
+	 0x1.b070p-103,
+	 0x1.b9c0p-104,
+	-0x1.01c0p-103,
+	-0x1.dfe0p-103,
+	 0x1.1b60p-104,
+	-0x1.ae94p-101,
+	-0x1.3340p-104,
+	 0x1.b3d8p-102,
+	-0x1.6e40p-105,
+	-0x1.3670p-103,
+	 0x1.c140p-104,
+	 0x1.1840p-101,
+	 0x1.1ab0p-102,
+	-0x1.a400p-104,
+	 0x1.1f00p-104,
+	-0x1.7180p-103,
+	 0x1.4ce0p-102,
+	 0x1.9200p-107,
+	-0x1.54c0p-103,
+	 0x1.1b80p-105,
+	-0x1.1828p-101,
+	 0x1.5720p-102,
+	-0x1.a060p-100,
+	 0x1.9160p-102,
+	 0x1.a280p-104,
+	 0x1.3400p-107,
+	 0x1.2b20p-102,
+	 0x1.7800p-108,
+	 0x1.cfd0p-101,
+	 0x1.2ef0p-102,
+	-0x1.2760p-99,
+	 0x1.b380p-104,
+	 0x1.0048p-101,
+	-0x1.60b0p-102,
+	 0x1.a1ccp-100,
+	-0x1.a640p-104,
+	-0x1.08a0p-101,
+	 0x1.7e60p-102,
+	 0x1.22c0p-103,
+	-0x1.7200p-106,
+	 0x1.f0f0p-102,
+	 0x1.eb4ep-99,
+	 0x1.c6e0p-103,
+};
+
+/*
+ * exp2l(x): compute the base 2 exponential of x
+ *
+ * Accuracy: Peak error < 0.502 ulp.
+ *
+ * Method: (accurate tables)
+ *
+ *   Reduce x:
+ *     x = 2**k + y, for integer k and |y| <= 1/2.
+ *     Thus we have exp2(x) = 2**k * exp2(y).
+ *
+ *   Reduce y:
+ *     y = i/TBLSIZE + z - eps[i] for integer i near y * TBLSIZE.
+ *     Thus we have exp2(y) = exp2(i/TBLSIZE) * exp2(z - eps[i]),
+ *     with |z - eps[i]| <= 2**-8 + 2**-98 for the table used.
+ *
+ *   We compute exp2(i/TBLSIZE) via table lookup and exp2(z - eps[i]) via
+ *   a degree-10 minimax polynomial with maximum error under 2**-120.
+ *   The values in exp2t[] and eps[] are chosen such that
+ *   exp2t[i] = exp2(i/TBLSIZE + eps[i]), and eps[i] is a small offset such
+ *   that exp2t[i] is accurate to 2**-122.
+ *
+ *   Note that the range of i is +-TBLSIZE/2, so we actually index the tables
+ *   by i0 = i + TBLSIZE/2.
+ *
+ *   This method is due to Gal, with many details due to Gal and Bachelis:
+ *
+ *	Gal, S. and Bachelis, B.  An Accurate Elementary Mathematical Library
+ *	for the IEEE Floating Point Standard.  TOMS 17(1), 26-46 (1991).
+ */
+long double
+exp2l(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	long double r, z, t;
+	uint32_t i0;
+	union {uint32_t u; int32_t i;} k;
+
+	/* Filter out exceptional cases. */
+	if (e >= 0x3fff + 14) {  /* |x| >= 16384 or x is NaN */
+		if (u.i.se >= 0x3fff + 15 && u.i.se >> 15 == 0)
+			/* overflow */
+			return x * 0x1p16383L;
+		if (e == 0x7fff)  /* -inf or -nan */
+			return -1/x;
+		if (x < -16382) {
+			if (x <= -16495 || x - 0x1p112 + 0x1p112 != x)
+				/* underflow */
+				FORCE_EVAL((float)(-0x1p-149/x));
+			if (x <= -16446)
+				return 0;
+		}
+	} else if (e < 0x3fff - 114) {
+		return 1 + x;
+	}
+
+	/*
+	 * Reduce x, computing z, i0, and k. The low bits of x + redux
+	 * contain the 16-bit integer part of the exponent (k) followed by
+	 * TBLBITS fractional bits (i0). We use bit tricks to extract these
+	 * as integers, then set z to the remainder.
+	 *
+	 * Example: Suppose x is 0xabc.123456p0 and TBLBITS is 8.
+	 * Then the low-order word of x + redux is 0x000abc12,
+	 * We split this into k = 0xabc and i0 = 0x12 (adjusted to
+	 * index into the table), then we compute z = 0x0.003456p0.
+	 */
+	u.f = x + redux;
+	i0 = u.i2.lo + TBLSIZE / 2;
+	k.u = i0 / TBLSIZE * TBLSIZE;
+	k.i /= TBLSIZE;
+	i0 %= TBLSIZE;
+	u.f -= redux;
+	z = x - u.f;
+
+	/* Compute r = exp2(y) = exp2t[i0] * p(z - eps[i]). */
+	t = tbl[i0];
+	z -= eps[i0];
+	r = t + t * z * (P1 + z * (P2 + z * (P3 + z * (P4 + z * (P5 + z * (P6
+	    + z * (P7 + z * (P8 + z * (P9 + z * P10)))))))));
+
+	return scalbnl(r, k.i);
+}
+#endif
diff --git a/src/math/exp_data.c b/src/math/exp_data.c
new file mode 100644
index 0000000..21be014
--- /dev/null
+++ b/src/math/exp_data.c
@@ -0,0 +1,182 @@
+/*
+ * Shared data between exp, exp2 and pow.
+ *
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "exp_data.h"
+
+#define N (1 << EXP_TABLE_BITS)
+
+const struct exp_data __exp_data = {
+// N/ln2
+.invln2N = 0x1.71547652b82fep0 * N,
+// -ln2/N
+.negln2hiN = -0x1.62e42fefa0000p-8,
+.negln2loN = -0x1.cf79abc9e3b3ap-47,
+// Used for rounding when !TOINT_INTRINSICS
+#if EXP_USE_TOINT_NARROW
+.shift = 0x1800000000.8p0,
+#else
+.shift = 0x1.8p52,
+#endif
+// exp polynomial coefficients.
+.poly = {
+// abs error: 1.555*2^-66
+// ulp error: 0.509 (0.511 without fma)
+// if |x| < ln2/256+eps
+// abs error if |x| < ln2/256+0x1p-15: 1.09*2^-65
+// abs error if |x| < ln2/128: 1.7145*2^-56
+0x1.ffffffffffdbdp-2,
+0x1.555555555543cp-3,
+0x1.55555cf172b91p-5,
+0x1.1111167a4d017p-7,
+},
+.exp2_shift = 0x1.8p52 / N,
+// exp2 polynomial coefficients.
+.exp2_poly = {
+// abs error: 1.2195*2^-65
+// ulp error: 0.507 (0.511 without fma)
+// if |x| < 1/256
+// abs error if |x| < 1/128: 1.9941*2^-56
+0x1.62e42fefa39efp-1,
+0x1.ebfbdff82c424p-3,
+0x1.c6b08d70cf4b5p-5,
+0x1.3b2abd24650ccp-7,
+0x1.5d7e09b4e3a84p-10,
+},
+// 2^(k/N) ~= H[k]*(1 + T[k]) for int k in [0,N)
+// tab[2*k] = asuint64(T[k])
+// tab[2*k+1] = asuint64(H[k]) - (k << 52)/N
+.tab = {
+0x0, 0x3ff0000000000000,
+0x3c9b3b4f1a88bf6e, 0x3feff63da9fb3335,
+0xbc7160139cd8dc5d, 0x3fefec9a3e778061,
+0xbc905e7a108766d1, 0x3fefe315e86e7f85,
+0x3c8cd2523567f613, 0x3fefd9b0d3158574,
+0xbc8bce8023f98efa, 0x3fefd06b29ddf6de,
+0x3c60f74e61e6c861, 0x3fefc74518759bc8,
+0x3c90a3e45b33d399, 0x3fefbe3ecac6f383,
+0x3c979aa65d837b6d, 0x3fefb5586cf9890f,
+0x3c8eb51a92fdeffc, 0x3fefac922b7247f7,
+0x3c3ebe3d702f9cd1, 0x3fefa3ec32d3d1a2,
+0xbc6a033489906e0b, 0x3fef9b66affed31b,
+0xbc9556522a2fbd0e, 0x3fef9301d0125b51,
+0xbc5080ef8c4eea55, 0x3fef8abdc06c31cc,
+0xbc91c923b9d5f416, 0x3fef829aaea92de0,
+0x3c80d3e3e95c55af, 0x3fef7a98c8a58e51,
+0xbc801b15eaa59348, 0x3fef72b83c7d517b,
+0xbc8f1ff055de323d, 0x3fef6af9388c8dea,
+0x3c8b898c3f1353bf, 0x3fef635beb6fcb75,
+0xbc96d99c7611eb26, 0x3fef5be084045cd4,
+0x3c9aecf73e3a2f60, 0x3fef54873168b9aa,
+0xbc8fe782cb86389d, 0x3fef4d5022fcd91d,
+0x3c8a6f4144a6c38d, 0x3fef463b88628cd6,
+0x3c807a05b0e4047d, 0x3fef3f49917ddc96,
+0x3c968efde3a8a894, 0x3fef387a6e756238,
+0x3c875e18f274487d, 0x3fef31ce4fb2a63f,
+0x3c80472b981fe7f2, 0x3fef2b4565e27cdd,
+0xbc96b87b3f71085e, 0x3fef24dfe1f56381,
+0x3c82f7e16d09ab31, 0x3fef1e9df51fdee1,
+0xbc3d219b1a6fbffa, 0x3fef187fd0dad990,
+0x3c8b3782720c0ab4, 0x3fef1285a6e4030b,
+0x3c6e149289cecb8f, 0x3fef0cafa93e2f56,
+0x3c834d754db0abb6, 0x3fef06fe0a31b715,
+0x3c864201e2ac744c, 0x3fef0170fc4cd831,
+0x3c8fdd395dd3f84a, 0x3feefc08b26416ff,
+0xbc86a3803b8e5b04, 0x3feef6c55f929ff1,
+0xbc924aedcc4b5068, 0x3feef1a7373aa9cb,
+0xbc9907f81b512d8e, 0x3feeecae6d05d866,
+0xbc71d1e83e9436d2, 0x3feee7db34e59ff7,
+0xbc991919b3ce1b15, 0x3feee32dc313a8e5,
+0x3c859f48a72a4c6d, 0x3feedea64c123422,
+0xbc9312607a28698a, 0x3feeda4504ac801c,
+0xbc58a78f4817895b, 0x3feed60a21f72e2a,
+0xbc7c2c9b67499a1b, 0x3feed1f5d950a897,
+0x3c4363ed60c2ac11, 0x3feece086061892d,
+0x3c9666093b0664ef, 0x3feeca41ed1d0057,
+0x3c6ecce1daa10379, 0x3feec6a2b5c13cd0,
+0x3c93ff8e3f0f1230, 0x3feec32af0d7d3de,
+0x3c7690cebb7aafb0, 0x3feebfdad5362a27,
+0x3c931dbdeb54e077, 0x3feebcb299fddd0d,
+0xbc8f94340071a38e, 0x3feeb9b2769d2ca7,
+0xbc87deccdc93a349, 0x3feeb6daa2cf6642,
+0xbc78dec6bd0f385f, 0x3feeb42b569d4f82,
+0xbc861246ec7b5cf6, 0x3feeb1a4ca5d920f,
+0x3c93350518fdd78e, 0x3feeaf4736b527da,
+0x3c7b98b72f8a9b05, 0x3feead12d497c7fd,
+0x3c9063e1e21c5409, 0x3feeab07dd485429,
+0x3c34c7855019c6ea, 0x3feea9268a5946b7,
+0x3c9432e62b64c035, 0x3feea76f15ad2148,
+0xbc8ce44a6199769f, 0x3feea5e1b976dc09,
+0xbc8c33c53bef4da8, 0x3feea47eb03a5585,
+0xbc845378892be9ae, 0x3feea34634ccc320,
+0xbc93cedd78565858, 0x3feea23882552225,
+0x3c5710aa807e1964, 0x3feea155d44ca973,
+0xbc93b3efbf5e2228, 0x3feea09e667f3bcd,
+0xbc6a12ad8734b982, 0x3feea012750bdabf,
+0xbc6367efb86da9ee, 0x3fee9fb23c651a2f,
+0xbc80dc3d54e08851, 0x3fee9f7df9519484,
+0xbc781f647e5a3ecf, 0x3fee9f75e8ec5f74,
+0xbc86ee4ac08b7db0, 0x3fee9f9a48a58174,
+0xbc8619321e55e68a, 0x3fee9feb564267c9,
+0x3c909ccb5e09d4d3, 0x3feea0694fde5d3f,
+0xbc7b32dcb94da51d, 0x3feea11473eb0187,
+0x3c94ecfd5467c06b, 0x3feea1ed0130c132,
+0x3c65ebe1abd66c55, 0x3feea2f336cf4e62,
+0xbc88a1c52fb3cf42, 0x3feea427543e1a12,
+0xbc9369b6f13b3734, 0x3feea589994cce13,
+0xbc805e843a19ff1e, 0x3feea71a4623c7ad,
+0xbc94d450d872576e, 0x3feea8d99b4492ed,
+0x3c90ad675b0e8a00, 0x3feeaac7d98a6699,
+0x3c8db72fc1f0eab4, 0x3feeace5422aa0db,
+0xbc65b6609cc5e7ff, 0x3feeaf3216b5448c,
+0x3c7bf68359f35f44, 0x3feeb1ae99157736,
+0xbc93091fa71e3d83, 0x3feeb45b0b91ffc6,
+0xbc5da9b88b6c1e29, 0x3feeb737b0cdc5e5,
+0xbc6c23f97c90b959, 0x3feeba44cbc8520f,
+0xbc92434322f4f9aa, 0x3feebd829fde4e50,
+0xbc85ca6cd7668e4b, 0x3feec0f170ca07ba,
+0x3c71affc2b91ce27, 0x3feec49182a3f090,
+0x3c6dd235e10a73bb, 0x3feec86319e32323,
+0xbc87c50422622263, 0x3feecc667b5de565,
+0x3c8b1c86e3e231d5, 0x3feed09bec4a2d33,
+0xbc91bbd1d3bcbb15, 0x3feed503b23e255d,
+0x3c90cc319cee31d2, 0x3feed99e1330b358,
+0x3c8469846e735ab3, 0x3feede6b5579fdbf,
+0xbc82dfcd978e9db4, 0x3feee36bbfd3f37a,
+0x3c8c1a7792cb3387, 0x3feee89f995ad3ad,
+0xbc907b8f4ad1d9fa, 0x3feeee07298db666,
+0xbc55c3d956dcaeba, 0x3feef3a2b84f15fb,
+0xbc90a40e3da6f640, 0x3feef9728de5593a,
+0xbc68d6f438ad9334, 0x3feeff76f2fb5e47,
+0xbc91eee26b588a35, 0x3fef05b030a1064a,
+0x3c74ffd70a5fddcd, 0x3fef0c1e904bc1d2,
+0xbc91bdfbfa9298ac, 0x3fef12c25bd71e09,
+0x3c736eae30af0cb3, 0x3fef199bdd85529c,
+0x3c8ee3325c9ffd94, 0x3fef20ab5fffd07a,
+0x3c84e08fd10959ac, 0x3fef27f12e57d14b,
+0x3c63cdaf384e1a67, 0x3fef2f6d9406e7b5,
+0x3c676b2c6c921968, 0x3fef3720dcef9069,
+0xbc808a1883ccb5d2, 0x3fef3f0b555dc3fa,
+0xbc8fad5d3ffffa6f, 0x3fef472d4a07897c,
+0xbc900dae3875a949, 0x3fef4f87080d89f2,
+0x3c74a385a63d07a7, 0x3fef5818dcfba487,
+0xbc82919e2040220f, 0x3fef60e316c98398,
+0x3c8e5a50d5c192ac, 0x3fef69e603db3285,
+0x3c843a59ac016b4b, 0x3fef7321f301b460,
+0xbc82d52107b43e1f, 0x3fef7c97337b9b5f,
+0xbc892ab93b470dc9, 0x3fef864614f5a129,
+0x3c74b604603a88d3, 0x3fef902ee78b3ff6,
+0x3c83c5ec519d7271, 0x3fef9a51fbc74c83,
+0xbc8ff7128fd391f0, 0x3fefa4afa2a490da,
+0xbc8dae98e223747d, 0x3fefaf482d8e67f1,
+0x3c8ec3bc41aa2008, 0x3fefba1bee615a27,
+0x3c842b94c3a9eb32, 0x3fefc52b376bba97,
+0x3c8a64a931d185ee, 0x3fefd0765b6e4540,
+0xbc8e37bae43be3ed, 0x3fefdbfdad9cbe14,
+0x3c77893b4d91cd9d, 0x3fefe7c1819e90d8,
+0x3c5305c14160cc89, 0x3feff3c22b8f71f1,
+},
+};
diff --git a/src/math/exp_data.h b/src/math/exp_data.h
new file mode 100644
index 0000000..3e24bac
--- /dev/null
+++ b/src/math/exp_data.h
@@ -0,0 +1,26 @@
+/*
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _EXP_DATA_H
+#define _EXP_DATA_H
+
+#include <features.h>
+#include <stdint.h>
+
+#define EXP_TABLE_BITS 7
+#define EXP_POLY_ORDER 5
+#define EXP_USE_TOINT_NARROW 0
+#define EXP2_POLY_ORDER 5
+extern hidden const struct exp_data {
+	double invln2N;
+	double shift;
+	double negln2hiN;
+	double negln2loN;
+	double poly[4]; /* Last four coefficients.  */
+	double exp2_shift;
+	double exp2_poly[EXP2_POLY_ORDER];
+	uint64_t tab[2*(1 << EXP_TABLE_BITS)];
+} __exp_data;
+
+#endif
diff --git a/src/math/expf.c b/src/math/expf.c
index 34a41f4..e2c0237 100644
--- a/src/math/expf.c
+++ b/src/math/expf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 half[2] = {0.5,-0.5},
diff --git a/src/math/expl.c b/src/math/expl.c
new file mode 100644
index 0000000..0a7f44f
--- /dev/null
+++ b/src/math/expl.c
@@ -0,0 +1,128 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_expl.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Exponential function, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, expl();
+ *
+ * y = expl( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns e (2.71828...) raised to the x power.
+ *
+ * Range reduction is accomplished by separating the argument
+ * into an integer k and fraction f such that
+ *
+ *     x    k  f
+ *    e  = 2  e.
+ *
+ * A Pade' form of degree 5/6 is used to approximate exp(f) - 1
+ * in the basic range [-0.5 ln 2, 0.5 ln 2].
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE      +-10000     50000       1.12e-19    2.81e-20
+ *
+ *
+ * Error amplification in the exponential function can be
+ * a serious matter.  The error propagation involves
+ * exp( X(1+delta) ) = exp(X) ( 1 + X*delta + ... ),
+ * which shows that a 1 lsb error in representing X produces
+ * a relative error of X times 1 lsb in the function.
+ * While the routine gives an accurate result for arguments
+ * that are exactly represented by a long double precision
+ * computer number, the result contains amplified roundoff
+ * error for large arguments not exactly represented.
+ *
+ *
+ * ERROR MESSAGES:
+ *
+ *   message         condition      value returned
+ * exp underflow    x < MINLOG         0.0
+ * exp overflow     x > MAXLOG         MAXNUM
+ *
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double expl(long double x)
+{
+	return exp(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+
+static const long double P[3] = {
+ 1.2617719307481059087798E-4L,
+ 3.0299440770744196129956E-2L,
+ 9.9999999999999999991025E-1L,
+};
+static const long double Q[4] = {
+ 3.0019850513866445504159E-6L,
+ 2.5244834034968410419224E-3L,
+ 2.2726554820815502876593E-1L,
+ 2.0000000000000000000897E0L,
+};
+static const long double
+LN2HI = 6.9314575195312500000000E-1L,
+LN2LO = 1.4286068203094172321215E-6L,
+LOG2E = 1.4426950408889634073599E0L;
+
+long double expl(long double x)
+{
+	long double px, xx;
+	int k;
+
+	if (isnan(x))
+		return x;
+	if (x > 11356.5234062941439488L) /* x > ln(2^16384 - 0.5) */
+		return x * 0x1p16383L;
+	if (x < -11399.4985314888605581L) /* x < ln(2^-16446) */
+		return -0x1p-16445L/x;
+
+	/* Express e**x = e**f 2**k
+	 *   = e**(f + k ln(2))
+	 */
+	px = floorl(LOG2E * x + 0.5);
+	k = px;
+	x -= px * LN2HI;
+	x -= px * LN2LO;
+
+	/* rational approximation of the fractional part:
+	 * e**x =  1 + 2x P(x**2)/(Q(x**2) - x P(x**2))
+	 */
+	xx = x * x;
+	px = x * __polevll(xx, P, 2);
+	x = px/(__polevll(xx, Q, 3) - px);
+	x = 1.0 + 2.0 * x;
+	return scalbnl(x, k);
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double expl(long double x)
+{
+	return exp(x);
+}
+#endif
diff --git a/src/math/expm1.c b/src/math/expm1.c
index 8bd78d7..ac1e61e 100644
--- a/src/math/expm1.c
+++ b/src/math/expm1.c
@@ -1,3 +1,14 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_expm1.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* expm1(x)
  * Returns exp(x)-1, the exponential of x minus 1.
  *
@@ -93,8 +104,7 @@
  * to produce the hexadecimal values shown.
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 o_threshold = 7.09782712893383973096e+02, /* 0x40862E42, 0xFEFA39EF */
diff --git a/src/math/expm1f.c b/src/math/expm1f.c
index e0d7701..012c7c5 100644
--- a/src/math/expm1f.c
+++ b/src/math/expm1f.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 o_threshold = 8.8721679688e+01, /* 0x42b17180 */
diff --git a/src/math/expm1l.c b/src/math/expm1l.c
new file mode 100644
index 0000000..d171507
--- /dev/null
+++ b/src/math/expm1l.c
@@ -0,0 +1,123 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_expm1l.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Exponential function, minus 1
+ *      Long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, expm1l();
+ *
+ * y = expm1l( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns e (2.71828...) raised to the x power, minus 1.
+ *
+ * Range reduction is accomplished by separating the argument
+ * into an integer k and fraction f such that
+ *
+ *     x    k  f
+ *    e  = 2  e.
+ *
+ * An expansion x + .5 x^2 + x^3 R(x) approximates exp(f) - 1
+ * in the basic range [-0.5 ln 2, 0.5 ln 2].
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE    -45,+maxarg   200,000     1.2e-19     2.5e-20
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double expm1l(long double x)
+{
+	return expm1(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+
+/* exp(x) - 1 = x + 0.5 x^2 + x^3 P(x)/Q(x)
+   -.5 ln 2  <  x  <  .5 ln 2
+   Theoretical peak relative error = 3.4e-22  */
+static const long double
+P0 = -1.586135578666346600772998894928250240826E4L,
+P1 =  2.642771505685952966904660652518429479531E3L,
+P2 = -3.423199068835684263987132888286791620673E2L,
+P3 =  1.800826371455042224581246202420972737840E1L,
+P4 = -5.238523121205561042771939008061958820811E-1L,
+Q0 = -9.516813471998079611319047060563358064497E4L,
+Q1 =  3.964866271411091674556850458227710004570E4L,
+Q2 = -7.207678383830091850230366618190187434796E3L,
+Q3 =  7.206038318724600171970199625081491823079E2L,
+Q4 = -4.002027679107076077238836622982900945173E1L,
+/* Q5 = 1.000000000000000000000000000000000000000E0 */
+/* C1 + C2 = ln 2 */
+C1 = 6.93145751953125E-1L,
+C2 = 1.428606820309417232121458176568075500134E-6L,
+/* ln 2^-65 */
+minarg = -4.5054566736396445112120088E1L,
+/* ln 2^16384 */
+maxarg = 1.1356523406294143949492E4L;
+
+long double expm1l(long double x)
+{
+	long double px, qx, xx;
+	int k;
+
+	if (isnan(x))
+		return x;
+	if (x > maxarg)
+		return x*0x1p16383L; /* overflow, unless x==inf */
+	if (x == 0.0)
+		return x;
+	if (x < minarg)
+		return -1.0;
+
+	xx = C1 + C2;
+	/* Express x = ln 2 (k + remainder), remainder not exceeding 1/2. */
+	px = floorl(0.5 + x / xx);
+	k = px;
+	/* remainder times ln 2 */
+	x -= px * C1;
+	x -= px * C2;
+
+	/* Approximate exp(remainder ln 2).*/
+	px = (((( P4 * x + P3) * x + P2) * x + P1) * x + P0) * x;
+	qx = (((( x + Q4) * x + Q3) * x + Q2) * x + Q1) * x + Q0;
+	xx = x * x;
+	qx = x + (0.5 * xx + xx * px / qx);
+
+	/* exp(x) = exp(k ln 2) exp(remainder ln 2) = 2^k exp(remainder ln 2).
+	 We have qx = exp(remainder ln 2) - 1, so
+	 exp(x) - 1  =  2^k (qx + 1) - 1  =  2^k qx + 2^k - 1.  */
+	px = scalbnl(1.0, k);
+	x = px * qx + (px - 1.0);
+	return x;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double expm1l(long double x)
+{
+	return expm1(x);
+}
+#endif
diff --git a/src/math/fabs.c b/src/math/fabs.c
index 24fe550..60c8d03 100644
--- a/src/math/fabs.c
+++ b/src/math/fabs.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static double __fabs(double x)
 {
diff --git a/src/math/fabsf.c b/src/math/fabsf.c
index ad6520c..7921266 100644
--- a/src/math/fabsf.c
+++ b/src/math/fabsf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static float __fabsf(float x)
 {
diff --git a/src/math/fdim.c b/src/math/fdim.c
index 61bfbd4..fb25521 100644
--- a/src/math/fdim.c
+++ b/src/math/fdim.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double fdim(double x, double y)
 {
diff --git a/src/math/fdimf.c b/src/math/fdimf.c
index 3bb5839..5cfeac6 100644
--- a/src/math/fdimf.c
+++ b/src/math/fdimf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float fdimf(float x, float y)
 {
diff --git a/src/math/fdiml.c b/src/math/fdiml.c
index 62e29b7..041a1fc 100644
--- a/src/math/fdiml.c
+++ b/src/math/fdiml.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 #include <float.h>
 
 #if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
diff --git a/src/math/finite.c b/src/math/finite.c
new file mode 100644
index 0000000..25a0575
--- /dev/null
+++ b/src/math/finite.c
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <math.h>
+
+int finite(double x)
+{
+	return isfinite(x);
+}
diff --git a/src/math/finitef.c b/src/math/finitef.c
new file mode 100644
index 0000000..2c4c771
--- /dev/null
+++ b/src/math/finitef.c
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <math.h>
+
+int finitef(float x)
+{
+	return isfinite(x);
+}
diff --git a/src/math/floor.c b/src/math/floor.c
index f44460d..410185e 100644
--- a/src/math/floor.c
+++ b/src/math/floor.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define EPS DBL_EPSILON
 static const double_t toint = 1/EPS;
diff --git a/src/math/floorf.c b/src/math/floorf.c
index ffff0d2..f1ed9a2 100644
--- a/src/math/floorf.c
+++ b/src/math/floorf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static float __floorf(float x)
 {
diff --git a/src/math/fma.c b/src/math/fma.c
new file mode 100644
index 0000000..0c6f90c
--- /dev/null
+++ b/src/math/fma.c
@@ -0,0 +1,183 @@
+#include <stdint.h>
+#include <float.h>
+#include <math.h>
+#include "atomic.h"
+
+#define ASUINT64(x) ((union {double f; uint64_t i;}){x}).i
+#define ZEROINFNAN (0x7ff-0x3ff-52-1)
+
+struct num { uint64_t m; int e; int sign; };
+
+static struct num normalize(double x)
+{
+	uint64_t ix = ASUINT64(x);
+	int e = ix>>52;
+	int sign = e & 0x800;
+	e &= 0x7ff;
+	if (!e) {
+		ix = ASUINT64(x*0x1p63);
+		e = ix>>52 & 0x7ff;
+		e = e ? e-63 : 0x800;
+	}
+	ix &= (1ull<<52)-1;
+	ix |= 1ull<<52;
+	ix <<= 1;
+	e -= 0x3ff + 52 + 1;
+	return (struct num){ix,e,sign};
+}
+
+static void mul(uint64_t *hi, uint64_t *lo, uint64_t x, uint64_t y)
+{
+	uint64_t t1,t2,t3;
+	uint64_t xlo = (uint32_t)x, xhi = x>>32;
+	uint64_t ylo = (uint32_t)y, yhi = y>>32;
+
+	t1 = xlo*ylo;
+	t2 = xlo*yhi + xhi*ylo;
+	t3 = xhi*yhi;
+	*lo = t1 + (t2<<32);
+	*hi = t3 + (t2>>32) + (t1 > *lo);
+}
+
+double fma(double x, double y, double z)
+{
+	#pragma STDC FENV_ACCESS ON
+
+	/* normalize so top 10bits and last bit are 0 */
+	struct num nx, ny, nz;
+	nx = normalize(x);
+	ny = normalize(y);
+	nz = normalize(z);
+
+	if (nx.e >= ZEROINFNAN || ny.e >= ZEROINFNAN)
+		return x*y + z;
+	if (nz.e >= ZEROINFNAN) {
+		if (nz.e > ZEROINFNAN) /* z==0 */
+			return x*y + z;
+		return z;
+	}
+
+	/* mul: r = x*y */
+	uint64_t rhi, rlo, zhi, zlo;
+	mul(&rhi, &rlo, nx.m, ny.m);
+	/* either top 20 or 21 bits of rhi and last 2 bits of rlo are 0 */
+
+	/* align exponents */
+	int e = nx.e + ny.e;
+	int d = nz.e - e;
+	/* shift bits z<<=kz, r>>=kr, so kz+kr == d, set e = e+kr (== ez-kz) */
+	if (d > 0) {
+		if (d < 64) {
+			zlo = nz.m<<d;
+			zhi = nz.m>>64-d;
+		} else {
+			zlo = 0;
+			zhi = nz.m;
+			e = nz.e - 64;
+			d -= 64;
+			if (d == 0) {
+			} else if (d < 64) {
+				rlo = rhi<<64-d | rlo>>d | !!(rlo<<64-d);
+				rhi = rhi>>d;
+			} else {
+				rlo = 1;
+				rhi = 0;
+			}
+		}
+	} else {
+		zhi = 0;
+		d = -d;
+		if (d == 0) {
+			zlo = nz.m;
+		} else if (d < 64) {
+			zlo = nz.m>>d | !!(nz.m<<64-d);
+		} else {
+			zlo = 1;
+		}
+	}
+
+	/* add */
+	int sign = nx.sign^ny.sign;
+	int samesign = !(sign^nz.sign);
+	int nonzero = 1;
+	if (samesign) {
+		/* r += z */
+		rlo += zlo;
+		rhi += zhi + (rlo < zlo);
+	} else {
+		/* r -= z */
+		uint64_t t = rlo;
+		rlo -= zlo;
+		rhi = rhi - zhi - (t < rlo);
+		if (rhi>>63) {
+			rlo = -rlo;
+			rhi = -rhi-!!rlo;
+			sign = !sign;
+		}
+		nonzero = !!rhi;
+	}
+
+	/* set rhi to top 63bit of the result (last bit is sticky) */
+	if (nonzero) {
+		e += 64;
+		d = a_clz_64(rhi)-1;
+		/* note: d > 0 */
+		rhi = rhi<<d | rlo>>64-d | !!(rlo<<d);
+	} else if (rlo) {
+		d = a_clz_64(rlo)-1;
+		if (d < 0)
+			rhi = rlo>>1 | (rlo&1);
+		else
+			rhi = rlo<<d;
+	} else {
+		/* exact +-0 */
+		return x*y + z;
+	}
+	e -= d;
+
+	/* convert to double */
+	int64_t i = rhi; /* i is in [1<<62,(1<<63)-1] */
+	if (sign)
+		i = -i;
+	double r = i; /* |r| is in [0x1p62,0x1p63] */
+
+	if (e < -1022-62) {
+		/* result is subnormal before rounding */
+		if (e == -1022-63) {
+			double c = 0x1p63;
+			if (sign)
+				c = -c;
+			if (r == c) {
+				/* min normal after rounding, underflow depends
+				   on arch behaviour which can be imitated by
+				   a double to float conversion */
+				float fltmin = 0x0.ffffff8p-63*FLT_MIN * r;
+				return DBL_MIN/FLT_MIN * fltmin;
+			}
+			/* one bit is lost when scaled, add another top bit to
+			   only round once at conversion if it is inexact */
+			if (rhi << 53) {
+				i = rhi>>1 | (rhi&1) | 1ull<<62;
+				if (sign)
+					i = -i;
+				r = i;
+				r = 2*r - c; /* remove top bit */
+
+				/* raise underflow portably, such that it
+				   cannot be optimized away */
+				{
+					double_t tiny = DBL_MIN/FLT_MIN * r;
+					r += (double)(tiny*tiny) * (r-r);
+				}
+			}
+		} else {
+			/* only round once when scaled */
+			d = 10;
+			i = ( rhi>>d | !!(rhi<<64-d) ) << d;
+			if (sign)
+				i = -i;
+			r = i;
+		}
+	}
+	return scalbn(r, e);
+}
diff --git a/src/math/fmaf.c b/src/math/fmaf.c
new file mode 100644
index 0000000..80f5cd8
--- /dev/null
+++ b/src/math/fmaf.c
@@ -0,0 +1,93 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_fmaf.c */
+/*-
+ * Copyright (c) 2005-2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdint.h>
+
+/*
+ * Fused multiply-add: Compute x * y + z with a single rounding error.
+ *
+ * A double has more than twice as much precision than a float, so
+ * direct double-precision arithmetic suffices, except where double
+ * rounding occurs.
+ */
+float fmaf(float x, float y, float z)
+{
+	#pragma STDC FENV_ACCESS ON
+	double xy, result;
+	union {double f; uint64_t i;} u;
+	int e;
+
+	xy = (double)x * y;
+	result = xy + z;
+	u.f = result;
+	e = u.i>>52 & 0x7ff;
+	/* Common case: The double precision result is fine. */
+	if ((u.i & 0x1fffffff) != 0x10000000 || /* not a halfway case */
+		e == 0x7ff ||                   /* NaN */
+		(result - xy == z && result - z == xy) || /* exact */
+		fegetround() != FE_TONEAREST)       /* not round-to-nearest */
+	{
+		/*
+		underflow may not be raised correctly, example:
+		fmaf(0x1p-120f, 0x1p-120f, 0x1p-149f)
+		*/
+#if defined(FE_INEXACT) && defined(FE_UNDERFLOW)
+		if (e < 0x3ff-126 && e >= 0x3ff-149 && fetestexcept(FE_INEXACT)) {
+			feclearexcept(FE_INEXACT);
+			/* TODO: gcc and clang bug workaround */
+			volatile float vz = z;
+			result = xy + vz;
+			if (fetestexcept(FE_INEXACT))
+				feraiseexcept(FE_UNDERFLOW);
+			else
+				feraiseexcept(FE_INEXACT);
+		}
+#endif
+		z = result;
+		return z;
+	}
+
+	/*
+	 * If result is inexact, and exactly halfway between two float values,
+	 * we need to adjust the low-order bit in the direction of the error.
+	 */
+#ifdef FE_TOWARDZERO
+	fesetround(FE_TOWARDZERO);
+#endif
+	volatile double vxy = xy;  /* XXX work around gcc CSE bug */
+	double adjusted_result = vxy + z;
+	fesetround(FE_TONEAREST);
+	if (result == adjusted_result) {
+		u.f = adjusted_result;
+		u.i++;
+		adjusted_result = u.f;
+	}
+	z = adjusted_result;
+	return z;
+}
diff --git a/src/math/fmal.c b/src/math/fmal.c
new file mode 100644
index 0000000..4506aac
--- /dev/null
+++ b/src/math/fmal.c
@@ -0,0 +1,293 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_fmal.c */
+/*-
+ * Copyright (c) 2005-2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+
+#include "libm.h"
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double fmal(long double x, long double y, long double z)
+{
+	return fma(x, y, z);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+#include <fenv.h>
+#if LDBL_MANT_DIG == 64
+#define LASTBIT(u) (u.i.m & 1)
+#define SPLIT (0x1p32L + 1)
+#elif LDBL_MANT_DIG == 113
+#define LASTBIT(u) (u.i.lo & 1)
+#define SPLIT (0x1p57L + 1)
+#endif
+
+/*
+ * A struct dd represents a floating-point number with twice the precision
+ * of a long double.  We maintain the invariant that "hi" stores the high-order
+ * bits of the result.
+ */
+struct dd {
+	long double hi;
+	long double lo;
+};
+
+/*
+ * Compute a+b exactly, returning the exact result in a struct dd.  We assume
+ * that both a and b are finite, but make no assumptions about their relative
+ * magnitudes.
+ */
+static inline struct dd dd_add(long double a, long double b)
+{
+	struct dd ret;
+	long double s;
+
+	ret.hi = a + b;
+	s = ret.hi - a;
+	ret.lo = (a - (ret.hi - s)) + (b - s);
+	return (ret);
+}
+
+/*
+ * Compute a+b, with a small tweak:  The least significant bit of the
+ * result is adjusted into a sticky bit summarizing all the bits that
+ * were lost to rounding.  This adjustment negates the effects of double
+ * rounding when the result is added to another number with a higher
+ * exponent.  For an explanation of round and sticky bits, see any reference
+ * on FPU design, e.g.,
+ *
+ *     J. Coonen.  An Implementation Guide to a Proposed Standard for
+ *     Floating-Point Arithmetic.  Computer, vol. 13, no. 1, Jan 1980.
+ */
+static inline long double add_adjusted(long double a, long double b)
+{
+	struct dd sum;
+	union ldshape u;
+
+	sum = dd_add(a, b);
+	if (sum.lo != 0) {
+		u.f = sum.hi;
+		if (!LASTBIT(u))
+			sum.hi = nextafterl(sum.hi, INFINITY * sum.lo);
+	}
+	return (sum.hi);
+}
+
+/*
+ * Compute ldexp(a+b, scale) with a single rounding error. It is assumed
+ * that the result will be subnormal, and care is taken to ensure that
+ * double rounding does not occur.
+ */
+static inline long double add_and_denormalize(long double a, long double b, int scale)
+{
+	struct dd sum;
+	int bits_lost;
+	union ldshape u;
+
+	sum = dd_add(a, b);
+
+	/*
+	 * If we are losing at least two bits of accuracy to denormalization,
+	 * then the first lost bit becomes a round bit, and we adjust the
+	 * lowest bit of sum.hi to make it a sticky bit summarizing all the
+	 * bits in sum.lo. With the sticky bit adjusted, the hardware will
+	 * break any ties in the correct direction.
+	 *
+	 * If we are losing only one bit to denormalization, however, we must
+	 * break the ties manually.
+	 */
+	if (sum.lo != 0) {
+		u.f = sum.hi;
+		bits_lost = -u.i.se - scale + 1;
+		if ((bits_lost != 1) ^ LASTBIT(u))
+			sum.hi = nextafterl(sum.hi, INFINITY * sum.lo);
+	}
+	return scalbnl(sum.hi, scale);
+}
+
+/*
+ * Compute a*b exactly, returning the exact result in a struct dd.  We assume
+ * that both a and b are normalized, so no underflow or overflow will occur.
+ * The current rounding mode must be round-to-nearest.
+ */
+static inline struct dd dd_mul(long double a, long double b)
+{
+	struct dd ret;
+	long double ha, hb, la, lb, p, q;
+
+	p = a * SPLIT;
+	ha = a - p;
+	ha += p;
+	la = a - ha;
+
+	p = b * SPLIT;
+	hb = b - p;
+	hb += p;
+	lb = b - hb;
+
+	p = ha * hb;
+	q = ha * lb + la * hb;
+
+	ret.hi = p + q;
+	ret.lo = p - ret.hi + q + la * lb;
+	return (ret);
+}
+
+/*
+ * Fused multiply-add: Compute x * y + z with a single rounding error.
+ *
+ * We use scaling to avoid overflow/underflow, along with the
+ * canonical precision-doubling technique adapted from:
+ *
+ *      Dekker, T.  A Floating-Point Technique for Extending the
+ *      Available Precision.  Numer. Math. 18, 224-242 (1971).
+ */
+long double fmal(long double x, long double y, long double z)
+{
+	#pragma STDC FENV_ACCESS ON
+	long double xs, ys, zs, adj;
+	struct dd xy, r;
+	int oround;
+	int ex, ey, ez;
+	int spread;
+
+	/*
+	 * Handle special cases. The order of operations and the particular
+	 * return values here are crucial in handling special cases involving
+	 * infinities, NaNs, overflows, and signed zeroes correctly.
+	 */
+	if (!isfinite(x) || !isfinite(y))
+		return (x * y + z);
+	if (!isfinite(z))
+		return (z);
+	if (x == 0.0 || y == 0.0)
+		return (x * y + z);
+	if (z == 0.0)
+		return (x * y);
+
+	xs = frexpl(x, &ex);
+	ys = frexpl(y, &ey);
+	zs = frexpl(z, &ez);
+	oround = fegetround();
+	spread = ex + ey - ez;
+
+	/*
+	 * If x * y and z are many orders of magnitude apart, the scaling
+	 * will overflow, so we handle these cases specially.  Rounding
+	 * modes other than FE_TONEAREST are painful.
+	 */
+	if (spread < -LDBL_MANT_DIG) {
+#ifdef FE_INEXACT
+		feraiseexcept(FE_INEXACT);
+#endif
+#ifdef FE_UNDERFLOW
+		if (!isnormal(z))
+			feraiseexcept(FE_UNDERFLOW);
+#endif
+		switch (oround) {
+		default: /* FE_TONEAREST */
+			return (z);
+#ifdef FE_TOWARDZERO
+		case FE_TOWARDZERO:
+			if (x > 0.0 ^ y < 0.0 ^ z < 0.0)
+				return (z);
+			else
+				return (nextafterl(z, 0));
+#endif
+#ifdef FE_DOWNWARD
+		case FE_DOWNWARD:
+			if (x > 0.0 ^ y < 0.0)
+				return (z);
+			else
+				return (nextafterl(z, -INFINITY));
+#endif
+#ifdef FE_UPWARD
+		case FE_UPWARD:
+			if (x > 0.0 ^ y < 0.0)
+				return (nextafterl(z, INFINITY));
+			else
+				return (z);
+#endif
+		}
+	}
+	if (spread <= LDBL_MANT_DIG * 2)
+		zs = scalbnl(zs, -spread);
+	else
+		zs = copysignl(LDBL_MIN, zs);
+
+	fesetround(FE_TONEAREST);
+
+	/*
+	 * Basic approach for round-to-nearest:
+	 *
+	 *     (xy.hi, xy.lo) = x * y           (exact)
+	 *     (r.hi, r.lo)   = xy.hi + z       (exact)
+	 *     adj = xy.lo + r.lo               (inexact; low bit is sticky)
+	 *     result = r.hi + adj              (correctly rounded)
+	 */
+	xy = dd_mul(xs, ys);
+	r = dd_add(xy.hi, zs);
+
+	spread = ex + ey;
+
+	if (r.hi == 0.0) {
+		/*
+		 * When the addends cancel to 0, ensure that the result has
+		 * the correct sign.
+		 */
+		fesetround(oround);
+		volatile long double vzs = zs; /* XXX gcc CSE bug workaround */
+		return xy.hi + vzs + scalbnl(xy.lo, spread);
+	}
+
+	if (oround != FE_TONEAREST) {
+		/*
+		 * There is no need to worry about double rounding in directed
+		 * rounding modes.
+		 * But underflow may not be raised correctly, example in downward rounding:
+		 * fmal(0x1.0000000001p-16000L, 0x1.0000000001p-400L, -0x1p-16440L)
+		 */
+		long double ret;
+#if defined(FE_INEXACT) && defined(FE_UNDERFLOW)
+		int e = fetestexcept(FE_INEXACT);
+		feclearexcept(FE_INEXACT);
+#endif
+		fesetround(oround);
+		adj = r.lo + xy.lo;
+		ret = scalbnl(r.hi + adj, spread);
+#if defined(FE_INEXACT) && defined(FE_UNDERFLOW)
+		if (ilogbl(ret) < -16382 && fetestexcept(FE_INEXACT))
+			feraiseexcept(FE_UNDERFLOW);
+		else if (e)
+			feraiseexcept(FE_INEXACT);
+#endif
+		return ret;
+	}
+
+	adj = add_adjusted(r.lo, xy.lo);
+	if (spread + ilogbl(r.hi) > -16383)
+		return scalbnl(r.hi + adj, spread);
+	else
+		return add_and_denormalize(r.hi, adj, spread);
+}
+#endif
diff --git a/src/math/fmax.c b/src/math/fmax.c
index 537464a..c4354f2 100644
--- a/src/math/fmax.c
+++ b/src/math/fmax.c
@@ -1,5 +1,4 @@
-
-#include <math.h>
+#include "libm.h"
 
 double fmax(double x, double y) {
     return (double) (x > y ? x : y);
diff --git a/src/math/fmaxf.c b/src/math/fmaxf.c
index bb06b8f..8c80bf8 100644
--- a/src/math/fmaxf.c
+++ b/src/math/fmaxf.c
@@ -1,5 +1,4 @@
-
-#include <math.h>
+#include "libm.h"
 
 float fmaxf(float x, float y) {
     return (float) (x > y ? x : y);
diff --git a/src/math/fmaxl.c b/src/math/fmaxl.c
index 4b03158..d8b0446 100644
--- a/src/math/fmaxl.c
+++ b/src/math/fmaxl.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 #include <float.h>
 
 #if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
diff --git a/src/math/fmin.c b/src/math/fmin.c
index 7c9ea09..3c13968 100644
--- a/src/math/fmin.c
+++ b/src/math/fmin.c
@@ -1,5 +1,4 @@
-
-#include <math.h>
+#include "libm.h"
 
 double fmin(double x, double y) {
     return (double) (x < y ? x : y);
diff --git a/src/math/fminf.c b/src/math/fminf.c
index b88dc77..6539ac3 100644
--- a/src/math/fminf.c
+++ b/src/math/fminf.c
@@ -1,5 +1,4 @@
-
-#include <math.h>
+#include "libm.h"
 
 float fminf(float x, float y) {
     return (float) (x < y ? x : y);
diff --git a/src/math/fminl.c b/src/math/fminl.c
index 69bc24a..efc4b5d 100644
--- a/src/math/fminl.c
+++ b/src/math/fminl.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 #include <float.h>
 
 #if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
diff --git a/src/math/fmod.c b/src/math/fmod.c
index 0f9f786..1def2f1 100644
--- a/src/math/fmod.c
+++ b/src/math/fmod.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double fmod(double x, double y)
 {
diff --git a/src/math/fmodf.c b/src/math/fmodf.c
index 9cdda9b..d73f591 100644
--- a/src/math/fmodf.c
+++ b/src/math/fmodf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float fmodf(float x, float y)
 {
diff --git a/src/math/frexp.c b/src/math/frexp.c
index f224e3a..584c7a6 100644
--- a/src/math/frexp.c
+++ b/src/math/frexp.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double frexp(double x, int *e)
 {
diff --git a/src/math/frexpf.c b/src/math/frexpf.c
index 0abbdb7..114beda 100644
--- a/src/math/frexpf.c
+++ b/src/math/frexpf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float frexpf(float x, int *e)
 {
diff --git a/src/math/hypot.c b/src/math/hypot.c
index 85395dd..fe8d55f 100644
--- a/src/math/hypot.c
+++ b/src/math/hypot.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define SPLIT (0x1p27 + 1)
 
diff --git a/src/math/hypotf.c b/src/math/hypotf.c
index 3ff3f20..f168b50 100644
--- a/src/math/hypotf.c
+++ b/src/math/hypotf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float hypotf(float x, float y)
 {
diff --git a/src/math/ilogb.c b/src/math/ilogb.c
new file mode 100644
index 0000000..64d4015
--- /dev/null
+++ b/src/math/ilogb.c
@@ -0,0 +1,26 @@
+#include <limits.h>
+#include "libm.h"
+
+int ilogb(double x)
+{
+	#pragma STDC FENV_ACCESS ON
+	union {double f; uint64_t i;} u = {x};
+	uint64_t i = u.i;
+	int e = i>>52 & 0x7ff;
+
+	if (!e) {
+		i <<= 12;
+		if (i == 0) {
+			FORCE_EVAL(0/0.0f);
+			return FP_ILOGB0;
+		}
+		/* subnormal x */
+		for (e = -0x3ff; i>>63 == 0; e--, i<<=1);
+		return e;
+	}
+	if (e == 0x7ff) {
+		FORCE_EVAL(0/0.0f);
+		return i<<12 ? FP_ILOGBNAN : INT_MAX;
+	}
+	return e - 0x3ff;
+}
diff --git a/src/math/ilogbf.c b/src/math/ilogbf.c
new file mode 100644
index 0000000..e23ba20
--- /dev/null
+++ b/src/math/ilogbf.c
@@ -0,0 +1,26 @@
+#include <limits.h>
+#include "libm.h"
+
+int ilogbf(float x)
+{
+	#pragma STDC FENV_ACCESS ON
+	union {float f; uint32_t i;} u = {x};
+	uint32_t i = u.i;
+	int e = i>>23 & 0xff;
+
+	if (!e) {
+		i <<= 9;
+		if (i == 0) {
+			FORCE_EVAL(0/0.0f);
+			return FP_ILOGB0;
+		}
+		/* subnormal x */
+		for (e = -0x7f; i>>31 == 0; e--, i<<=1);
+		return e;
+	}
+	if (e == 0xff) {
+		FORCE_EVAL(0/0.0f);
+		return i<<9 ? FP_ILOGBNAN : INT_MAX;
+	}
+	return e - 0x7f;
+}
diff --git a/src/math/ilogbl.c b/src/math/ilogbl.c
new file mode 100644
index 0000000..7b1a9cf
--- /dev/null
+++ b/src/math/ilogbl.c
@@ -0,0 +1,55 @@
+#include <limits.h>
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+int ilogbl(long double x)
+{
+	return ilogb(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+int ilogbl(long double x)
+{
+	#pragma STDC FENV_ACCESS ON
+	union ldshape u = {x};
+	uint64_t m = u.i.m;
+	int e = u.i.se & 0x7fff;
+
+	if (!e) {
+		if (m == 0) {
+			FORCE_EVAL(0/0.0f);
+			return FP_ILOGB0;
+		}
+		/* subnormal x */
+		for (e = -0x3fff+1; m>>63 == 0; e--, m<<=1);
+		return e;
+	}
+	if (e == 0x7fff) {
+		FORCE_EVAL(0/0.0f);
+		return m<<1 ? FP_ILOGBNAN : INT_MAX;
+	}
+	return e - 0x3fff;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+int ilogbl(long double x)
+{
+	#pragma STDC FENV_ACCESS ON
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+
+	if (!e) {
+		if (x == 0) {
+			FORCE_EVAL(0/0.0f);
+			return FP_ILOGB0;
+		}
+		/* subnormal x */
+		x *= 0x1p120;
+		return ilogbl(x) - 120;
+	}
+	if (e == 0x7fff) {
+		FORCE_EVAL(0/0.0f);
+		u.i.se = 0;
+		return u.f ? FP_ILOGBNAN : INT_MAX;
+	}
+	return e - 0x3fff;
+}
+#endif
diff --git a/src/math/ldexp.c b/src/math/ldexp.c
index 6fb2c24..36835db 100644
--- a/src/math/ldexp.c
+++ b/src/math/ldexp.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double ldexp(double x, int n)
 {
diff --git a/src/math/ldexpf.c b/src/math/ldexpf.c
index d25f97b..f0981ae 100644
--- a/src/math/ldexpf.c
+++ b/src/math/ldexpf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float ldexpf(float x, int n)
 {
diff --git a/src/math/ldexpl.c b/src/math/ldexpl.c
new file mode 100644
index 0000000..fd145cc
--- /dev/null
+++ b/src/math/ldexpl.c
@@ -0,0 +1,6 @@
+#include <math.h>
+
+long double ldexpl(long double x, int n)
+{
+	return scalbnl(x, n);
+}
diff --git a/src/math/log10.c b/src/math/log10.c
index 0abf2f3..09f49ce 100644
--- a/src/math/log10.c
+++ b/src/math/log10.c
@@ -6,8 +6,7 @@
  *    log10(x) = (f - f*f/2 + r)/log(10) + k*log10(2)
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 ivln10hi  = 4.34294481878168880939e-01, /* 0x3fdbcb7b, 0x15200000 */
diff --git a/src/math/log10f.c b/src/math/log10f.c
index 1fe120c..48da8b8 100644
--- a/src/math/log10f.c
+++ b/src/math/log10f.c
@@ -2,8 +2,7 @@
  * See comments in log10.c.
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 ivln10hi  =  4.3432617188e-01, /* 0x3ede6000 */
diff --git a/src/math/log10l.c b/src/math/log10l.c
new file mode 100644
index 0000000..63dcc28
--- /dev/null
+++ b/src/math/log10l.c
@@ -0,0 +1,191 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_log10l.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Common logarithm, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, log10l();
+ *
+ * y = log10l( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns the base 10 logarithm of x.
+ *
+ * The argument is separated into its exponent and fractional
+ * parts.  If the exponent is between -1 and +1, the logarithm
+ * of the fraction is approximated by
+ *
+ *     log(1+x) = x - 0.5 x**2 + x**3 P(x)/Q(x).
+ *
+ * Otherwise, setting  z = 2(x-1)/x+1),
+ *
+ *     log(x) = z + z**3 P(z)/Q(z).
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE      0.5, 2.0     30000      9.0e-20     2.6e-20
+ *    IEEE     exp(+-10000)  30000      6.0e-20     2.3e-20
+ *
+ * In the tests over the interval exp(+-10000), the logarithms
+ * of the random arguments were uniformly distributed over
+ * [-10000, +10000].
+ *
+ * ERROR MESSAGES:
+ *
+ * log singularity:  x = 0; returns MINLOG
+ * log domain:       x < 0; returns MINLOG
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double log10l(long double x)
+{
+	return log10(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/* Coefficients for log(1+x) = x - x**2/2 + x**3 P(x)/Q(x)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 6.2e-22
+ */
+static const long double P[] = {
+ 4.9962495940332550844739E-1L,
+ 1.0767376367209449010438E1L,
+ 7.7671073698359539859595E1L,
+ 2.5620629828144409632571E2L,
+ 4.2401812743503691187826E2L,
+ 3.4258224542413922935104E2L,
+ 1.0747524399916215149070E2L,
+};
+static const long double Q[] = {
+/* 1.0000000000000000000000E0,*/
+ 2.3479774160285863271658E1L,
+ 1.9444210022760132894510E2L,
+ 7.7952888181207260646090E2L,
+ 1.6911722418503949084863E3L,
+ 2.0307734695595183428202E3L,
+ 1.2695660352705325274404E3L,
+ 3.2242573199748645407652E2L,
+};
+
+/* Coefficients for log(x) = z + z^3 P(z^2)/Q(z^2),
+ * where z = 2(x-1)/(x+1)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 6.16e-22
+ */
+static const long double R[4] = {
+ 1.9757429581415468984296E-3L,
+-7.1990767473014147232598E-1L,
+ 1.0777257190312272158094E1L,
+-3.5717684488096787370998E1L,
+};
+static const long double S[4] = {
+/* 1.00000000000000000000E0L,*/
+-2.6201045551331104417768E1L,
+ 1.9361891836232102174846E2L,
+-4.2861221385716144629696E2L,
+};
+/* log10(2) */
+#define L102A 0.3125L
+#define L102B -1.1470004336018804786261e-2L
+/* log10(e) */
+#define L10EA 0.5L
+#define L10EB -6.5705518096748172348871e-2L
+
+#define SQRTH 0.70710678118654752440L
+
+long double log10l(long double x)
+{
+	long double y, z;
+	int e;
+
+	if (isnan(x))
+		return x;
+	if(x <= 0.0) {
+		if(x == 0.0)
+			return -1.0 / (x*x);
+		return (x - x) / 0.0;
+	}
+	if (x == INFINITY)
+		return INFINITY;
+	/* separate mantissa from exponent */
+	/* Note, frexp is used so that denormal numbers
+	 * will be handled properly.
+	 */
+	x = frexpl(x, &e);
+
+	/* logarithm using log(x) = z + z**3 P(z)/Q(z),
+	 * where z = 2(x-1)/x+1)
+	 */
+	if (e > 2 || e < -2) {
+		if (x < SQRTH) {  /* 2(2x-1)/(2x+1) */
+			e -= 1;
+			z = x - 0.5;
+			y = 0.5 * z + 0.5;
+		} else {  /*  2 (x-1)/(x+1)   */
+			z = x - 0.5;
+			z -= 0.5;
+			y = 0.5 * x  + 0.5;
+		}
+		x = z / y;
+		z = x*x;
+		y = x * (z * __polevll(z, R, 3) / __p1evll(z, S, 3));
+		goto done;
+	}
+
+	/* logarithm using log(1+x) = x - .5x**2 + x**3 P(x)/Q(x) */
+	if (x < SQRTH) {
+		e -= 1;
+		x = 2.0*x - 1.0;
+	} else {
+		x = x - 1.0;
+	}
+	z = x*x;
+	y = x * (z * __polevll(x, P, 6) / __p1evll(x, Q, 7));
+	y = y - 0.5*z;
+
+done:
+	/* Multiply log of fraction by log10(e)
+	 * and base 2 exponent by log10(2).
+	 *
+	 * ***CAUTION***
+	 *
+	 * This sequence of operations is critical and it may
+	 * be horribly defeated by some compiler optimizers.
+	 */
+	z = y * (L10EB);
+	z += x * (L10EB);
+	z += e * (L102B);
+	z += y * (L10EA);
+	z += x * (L10EA);
+	z += e * (L102A);
+	return z;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double log10l(long double x)
+{
+	return log10(x);
+}
+#endif
diff --git a/src/math/log1p.c b/src/math/log1p.c
index 24d891f..0097134 100644
--- a/src/math/log1p.c
+++ b/src/math/log1p.c
@@ -1,3 +1,14 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_log1p.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
 /* double log1p(double x)
  * Return the natural logarithm of 1+x.
  *
@@ -42,8 +53,7 @@
  *       See HP-15C Advanced Functions Handbook, p.193.
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 ln2_hi = 6.93147180369123816490e-01,  /* 3fe62e42 fee00000 */
diff --git a/src/math/log1pf.c b/src/math/log1pf.c
index fda8775..0a3aa98 100644
--- a/src/math/log1pf.c
+++ b/src/math/log1pf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 ln2_hi = 6.9313812256e-01, /* 0x3f317180 */
diff --git a/src/math/log1pl.c b/src/math/log1pl.c
new file mode 100644
index 0000000..141b5f0
--- /dev/null
+++ b/src/math/log1pl.c
@@ -0,0 +1,177 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/s_log1pl.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Relative error logarithm
+ *      Natural logarithm of 1+x, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, log1pl();
+ *
+ * y = log1pl( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns the base e (2.718...) logarithm of 1+x.
+ *
+ * The argument 1+x is separated into its exponent and fractional
+ * parts.  If the exponent is between -1 and +1, the logarithm
+ * of the fraction is approximated by
+ *
+ *     log(1+x) = x - 0.5 x^2 + x^3 P(x)/Q(x).
+ *
+ * Otherwise, setting  z = 2(x-1)/x+1),
+ *
+ *     log(x) = z + z^3 P(z)/Q(z).
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE     -1.0, 9.0    100000      8.2e-20    2.5e-20
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double log1pl(long double x)
+{
+	return log1p(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/* Coefficients for log(1+x) = x - x^2 / 2 + x^3 P(x)/Q(x)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 2.32e-20
+ */
+static const long double P[] = {
+ 4.5270000862445199635215E-5L,
+ 4.9854102823193375972212E-1L,
+ 6.5787325942061044846969E0L,
+ 2.9911919328553073277375E1L,
+ 6.0949667980987787057556E1L,
+ 5.7112963590585538103336E1L,
+ 2.0039553499201281259648E1L,
+};
+static const long double Q[] = {
+/* 1.0000000000000000000000E0,*/
+ 1.5062909083469192043167E1L,
+ 8.3047565967967209469434E1L,
+ 2.2176239823732856465394E2L,
+ 3.0909872225312059774938E2L,
+ 2.1642788614495947685003E2L,
+ 6.0118660497603843919306E1L,
+};
+
+/* Coefficients for log(x) = z + z^3 P(z^2)/Q(z^2),
+ * where z = 2(x-1)/(x+1)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 6.16e-22
+ */
+static const long double R[4] = {
+ 1.9757429581415468984296E-3L,
+-7.1990767473014147232598E-1L,
+ 1.0777257190312272158094E1L,
+-3.5717684488096787370998E1L,
+};
+static const long double S[4] = {
+/* 1.00000000000000000000E0L,*/
+-2.6201045551331104417768E1L,
+ 1.9361891836232102174846E2L,
+-4.2861221385716144629696E2L,
+};
+static const long double C1 = 6.9314575195312500000000E-1L;
+static const long double C2 = 1.4286068203094172321215E-6L;
+
+#define SQRTH 0.70710678118654752440L
+
+long double log1pl(long double xm1)
+{
+	long double x, y, z;
+	int e;
+
+	if (isnan(xm1))
+		return xm1;
+	if (xm1 == INFINITY)
+		return xm1;
+	if (xm1 == 0.0)
+		return xm1;
+
+	x = xm1 + 1.0;
+
+	/* Test for domain errors.  */
+	if (x <= 0.0) {
+		if (x == 0.0)
+			return -1/(x*x); /* -inf with divbyzero */
+		return 0/0.0f; /* nan with invalid */
+	}
+
+	/* Separate mantissa from exponent.
+	   Use frexp so that denormal numbers will be handled properly.  */
+	x = frexpl(x, &e);
+
+	/* logarithm using log(x) = z + z^3 P(z)/Q(z),
+	   where z = 2(x-1)/x+1)  */
+	if (e > 2 || e < -2) {
+		if (x < SQRTH) { /* 2(2x-1)/(2x+1) */
+			e -= 1;
+			z = x - 0.5;
+			y = 0.5 * z + 0.5;
+		} else { /*  2 (x-1)/(x+1)   */
+			z = x - 0.5;
+			z -= 0.5;
+			y = 0.5 * x  + 0.5;
+		}
+		x = z / y;
+		z = x*x;
+		z = x * (z * __polevll(z, R, 3) / __p1evll(z, S, 3));
+		z = z + e * C2;
+		z = z + x;
+		z = z + e * C1;
+		return z;
+	}
+
+	/* logarithm using log(1+x) = x - .5x**2 + x**3 P(x)/Q(x) */
+	if (x < SQRTH) {
+		e -= 1;
+		if (e != 0)
+			x = 2.0 * x - 1.0;
+		else
+			x = xm1;
+	} else {
+		if (e != 0)
+			x = x - 1.0;
+		else
+			x = xm1;
+	}
+	z = x*x;
+	y = x * (z * __polevll(x, P, 6) / __p1evll(x, Q, 6));
+	y = y + e * C2;
+	z = y - 0.5 * z;
+	z = z + x;
+	z = z + e * C1;
+	return z;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double log1pl(long double x)
+{
+	return log1p(x);
+}
+#endif
diff --git a/src/math/log2.c b/src/math/log2.c
index 414180e..35f51e8 100644
--- a/src/math/log2.c
+++ b/src/math/log2.c
@@ -6,8 +6,7 @@
  *    log2(x) = (f - f*f/2 + r)/log(2) + k
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double
 ivln2hi = 1.44269504072144627571e+00, /* 0x3ff71547, 0x65200000 */
diff --git a/src/math/log2_data.c b/src/math/log2_data.c
new file mode 100644
index 0000000..3dd1ca5
--- /dev/null
+++ b/src/math/log2_data.c
@@ -0,0 +1,201 @@
+/*
+ * Data for log2.
+ *
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "log2_data.h"
+
+#define N (1 << LOG2_TABLE_BITS)
+
+const struct log2_data __log2_data = {
+// First coefficient: 0x1.71547652b82fe1777d0ffda0d24p0
+.invln2hi = 0x1.7154765200000p+0,
+.invln2lo = 0x1.705fc2eefa200p-33,
+.poly1 = {
+// relative error: 0x1.2fad8188p-63
+// in -0x1.5b51p-5 0x1.6ab2p-5
+-0x1.71547652b82fep-1,
+0x1.ec709dc3a03f7p-2,
+-0x1.71547652b7c3fp-2,
+0x1.2776c50f05be4p-2,
+-0x1.ec709dd768fe5p-3,
+0x1.a61761ec4e736p-3,
+-0x1.7153fbc64a79bp-3,
+0x1.484d154f01b4ap-3,
+-0x1.289e4a72c383cp-3,
+0x1.0b32f285aee66p-3,
+},
+.poly = {
+// relative error: 0x1.a72c2bf8p-58
+// abs error: 0x1.67a552c8p-66
+// in -0x1.f45p-8 0x1.f45p-8
+-0x1.71547652b8339p-1,
+0x1.ec709dc3a04bep-2,
+-0x1.7154764702ffbp-2,
+0x1.2776c50034c48p-2,
+-0x1.ec7b328ea92bcp-3,
+0x1.a6225e117f92ep-3,
+},
+/* Algorithm:
+
+	x = 2^k z
+	log2(x) = k + log2(c) + log2(z/c)
+	log2(z/c) = poly(z/c - 1)
+
+where z is in [1.6p-1; 1.6p0] which is split into N subintervals and z falls
+into the ith one, then table entries are computed as
+
+	tab[i].invc = 1/c
+	tab[i].logc = (double)log2(c)
+	tab2[i].chi = (double)c
+	tab2[i].clo = (double)(c - (double)c)
+
+where c is near the center of the subinterval and is chosen by trying +-2^29
+floating point invc candidates around 1/center and selecting one for which
+
+	1) the rounding error in 0x1.8p10 + logc is 0,
+	2) the rounding error in z - chi - clo is < 0x1p-64 and
+	3) the rounding error in (double)log2(c) is minimized (< 0x1p-68).
+
+Note: 1) ensures that k + logc can be computed without rounding error, 2)
+ensures that z/c - 1 can be computed as (z - chi - clo)*invc with close to a
+single rounding error when there is no fast fma for z*invc - 1, 3) ensures
+that logc + poly(z/c - 1) has small error, however near x == 1 when
+|log2(x)| < 0x1p-4, this is not enough so that is special cased.  */
+.tab = {
+{0x1.724286bb1acf8p+0, -0x1.1095feecdb000p-1},
+{0x1.6e1f766d2cca1p+0, -0x1.08494bd76d000p-1},
+{0x1.6a13d0e30d48ap+0, -0x1.00143aee8f800p-1},
+{0x1.661ec32d06c85p+0, -0x1.efec5360b4000p-2},
+{0x1.623fa951198f8p+0, -0x1.dfdd91ab7e000p-2},
+{0x1.5e75ba4cf026cp+0, -0x1.cffae0cc79000p-2},
+{0x1.5ac055a214fb8p+0, -0x1.c043811fda000p-2},
+{0x1.571ed0f166e1ep+0, -0x1.b0b67323ae000p-2},
+{0x1.53909590bf835p+0, -0x1.a152f5a2db000p-2},
+{0x1.5014fed61adddp+0, -0x1.9217f5af86000p-2},
+{0x1.4cab88e487bd0p+0, -0x1.8304db0719000p-2},
+{0x1.49539b4334feep+0, -0x1.74189f9a9e000p-2},
+{0x1.460cbdfafd569p+0, -0x1.6552bb5199000p-2},
+{0x1.42d664ee4b953p+0, -0x1.56b23a29b1000p-2},
+{0x1.3fb01111dd8a6p+0, -0x1.483650f5fa000p-2},
+{0x1.3c995b70c5836p+0, -0x1.39de937f6a000p-2},
+{0x1.3991c4ab6fd4ap+0, -0x1.2baa1538d6000p-2},
+{0x1.3698e0ce099b5p+0, -0x1.1d98340ca4000p-2},
+{0x1.33ae48213e7b2p+0, -0x1.0fa853a40e000p-2},
+{0x1.30d191985bdb1p+0, -0x1.01d9c32e73000p-2},
+{0x1.2e025cab271d7p+0, -0x1.e857da2fa6000p-3},
+{0x1.2b404cf13cd82p+0, -0x1.cd3c8633d8000p-3},
+{0x1.288b02c7ccb50p+0, -0x1.b26034c14a000p-3},
+{0x1.25e2263944de5p+0, -0x1.97c1c2f4fe000p-3},
+{0x1.234563d8615b1p+0, -0x1.7d6023f800000p-3},
+{0x1.20b46e33eaf38p+0, -0x1.633a71a05e000p-3},
+{0x1.1e2eefdcda3ddp+0, -0x1.494f5e9570000p-3},
+{0x1.1bb4a580b3930p+0, -0x1.2f9e424e0a000p-3},
+{0x1.19453847f2200p+0, -0x1.162595afdc000p-3},
+{0x1.16e06c0d5d73cp+0, -0x1.f9c9a75bd8000p-4},
+{0x1.1485f47b7e4c2p+0, -0x1.c7b575bf9c000p-4},
+{0x1.12358ad0085d1p+0, -0x1.960c60ff48000p-4},
+{0x1.0fef00f532227p+0, -0x1.64ce247b60000p-4},
+{0x1.0db2077d03a8fp+0, -0x1.33f78b2014000p-4},
+{0x1.0b7e6d65980d9p+0, -0x1.0387d1a42c000p-4},
+{0x1.0953efe7b408dp+0, -0x1.a6f9208b50000p-5},
+{0x1.07325cac53b83p+0, -0x1.47a954f770000p-5},
+{0x1.05197e40d1b5cp+0, -0x1.d23a8c50c0000p-6},
+{0x1.03091c1208ea2p+0, -0x1.16a2629780000p-6},
+{0x1.0101025b37e21p+0, -0x1.720f8d8e80000p-8},
+{0x1.fc07ef9caa76bp-1, 0x1.6fe53b1500000p-7},
+{0x1.f4465d3f6f184p-1, 0x1.11ccce10f8000p-5},
+{0x1.ecc079f84107fp-1, 0x1.c4dfc8c8b8000p-5},
+{0x1.e573a99975ae8p-1, 0x1.3aa321e574000p-4},
+{0x1.de5d6f0bd3de6p-1, 0x1.918a0d08b8000p-4},
+{0x1.d77b681ff38b3p-1, 0x1.e72e9da044000p-4},
+{0x1.d0cb5724de943p-1, 0x1.1dcd2507f6000p-3},
+{0x1.ca4b2dc0e7563p-1, 0x1.476ab03dea000p-3},
+{0x1.c3f8ee8d6cb51p-1, 0x1.7074377e22000p-3},
+{0x1.bdd2b4f020c4cp-1, 0x1.98ede8ba94000p-3},
+{0x1.b7d6c006015cap-1, 0x1.c0db86ad2e000p-3},
+{0x1.b20366e2e338fp-1, 0x1.e840aafcee000p-3},
+{0x1.ac57026295039p-1, 0x1.0790ab4678000p-2},
+{0x1.a6d01bc2731ddp-1, 0x1.1ac056801c000p-2},
+{0x1.a16d3bc3ff18bp-1, 0x1.2db11d4fee000p-2},
+{0x1.9c2d14967feadp-1, 0x1.406464ec58000p-2},
+{0x1.970e4f47c9902p-1, 0x1.52dbe093af000p-2},
+{0x1.920fb3982bcf2p-1, 0x1.651902050d000p-2},
+{0x1.8d30187f759f1p-1, 0x1.771d2cdeaf000p-2},
+{0x1.886e5ebb9f66dp-1, 0x1.88e9c857d9000p-2},
+{0x1.83c97b658b994p-1, 0x1.9a80155e16000p-2},
+{0x1.7f405ffc61022p-1, 0x1.abe186ed3d000p-2},
+{0x1.7ad22181415cap-1, 0x1.bd0f2aea0e000p-2},
+{0x1.767dcf99eff8cp-1, 0x1.ce0a43dbf4000p-2},
+},
+#if !__FP_FAST_FMA
+.tab2 = {
+{0x1.6200012b90a8ep-1, 0x1.904ab0644b605p-55},
+{0x1.66000045734a6p-1, 0x1.1ff9bea62f7a9p-57},
+{0x1.69fffc325f2c5p-1, 0x1.27ecfcb3c90bap-55},
+{0x1.6e00038b95a04p-1, 0x1.8ff8856739326p-55},
+{0x1.71fffe09994e3p-1, 0x1.afd40275f82b1p-55},
+{0x1.7600015590e1p-1, -0x1.2fd75b4238341p-56},
+{0x1.7a00012655bd5p-1, 0x1.808e67c242b76p-56},
+{0x1.7e0003259e9a6p-1, -0x1.208e426f622b7p-57},
+{0x1.81fffedb4b2d2p-1, -0x1.402461ea5c92fp-55},
+{0x1.860002dfafcc3p-1, 0x1.df7f4a2f29a1fp-57},
+{0x1.89ffff78c6b5p-1, -0x1.e0453094995fdp-55},
+{0x1.8e00039671566p-1, -0x1.a04f3bec77b45p-55},
+{0x1.91fffe2bf1745p-1, -0x1.7fa34400e203cp-56},
+{0x1.95fffcc5c9fd1p-1, -0x1.6ff8005a0695dp-56},
+{0x1.9a0003bba4767p-1, 0x1.0f8c4c4ec7e03p-56},
+{0x1.9dfffe7b92da5p-1, 0x1.e7fd9478c4602p-55},
+{0x1.a1fffd72efdafp-1, -0x1.a0c554dcdae7ep-57},
+{0x1.a5fffde04ff95p-1, 0x1.67da98ce9b26bp-55},
+{0x1.a9fffca5e8d2bp-1, -0x1.284c9b54c13dep-55},
+{0x1.adfffddad03eap-1, 0x1.812c8ea602e3cp-58},
+{0x1.b1ffff10d3d4dp-1, -0x1.efaddad27789cp-55},
+{0x1.b5fffce21165ap-1, 0x1.3cb1719c61237p-58},
+{0x1.b9fffd950e674p-1, 0x1.3f7d94194cep-56},
+{0x1.be000139ca8afp-1, 0x1.50ac4215d9bcp-56},
+{0x1.c20005b46df99p-1, 0x1.beea653e9c1c9p-57},
+{0x1.c600040b9f7aep-1, -0x1.c079f274a70d6p-56},
+{0x1.ca0006255fd8ap-1, -0x1.a0b4076e84c1fp-56},
+{0x1.cdfffd94c095dp-1, 0x1.8f933f99ab5d7p-55},
+{0x1.d1ffff975d6cfp-1, -0x1.82c08665fe1bep-58},
+{0x1.d5fffa2561c93p-1, -0x1.b04289bd295f3p-56},
+{0x1.d9fff9d228b0cp-1, 0x1.70251340fa236p-55},
+{0x1.de00065bc7e16p-1, -0x1.5011e16a4d80cp-56},
+{0x1.e200002f64791p-1, 0x1.9802f09ef62ep-55},
+{0x1.e600057d7a6d8p-1, -0x1.e0b75580cf7fap-56},
+{0x1.ea00027edc00cp-1, -0x1.c848309459811p-55},
+{0x1.ee0006cf5cb7cp-1, -0x1.f8027951576f4p-55},
+{0x1.f2000782b7dccp-1, -0x1.f81d97274538fp-55},
+{0x1.f6000260c450ap-1, -0x1.071002727ffdcp-59},
+{0x1.f9fffe88cd533p-1, -0x1.81bdce1fda8bp-58},
+{0x1.fdfffd50f8689p-1, 0x1.7f91acb918e6ep-55},
+{0x1.0200004292367p+0, 0x1.b7ff365324681p-54},
+{0x1.05fffe3e3d668p+0, 0x1.6fa08ddae957bp-55},
+{0x1.0a0000a85a757p+0, -0x1.7e2de80d3fb91p-58},
+{0x1.0e0001a5f3fccp+0, -0x1.1823305c5f014p-54},
+{0x1.11ffff8afbaf5p+0, -0x1.bfabb6680bac2p-55},
+{0x1.15fffe54d91adp+0, -0x1.d7f121737e7efp-54},
+{0x1.1a00011ac36e1p+0, 0x1.c000a0516f5ffp-54},
+{0x1.1e00019c84248p+0, -0x1.082fbe4da5dap-54},
+{0x1.220000ffe5e6ep+0, -0x1.8fdd04c9cfb43p-55},
+{0x1.26000269fd891p+0, 0x1.cfe2a7994d182p-55},
+{0x1.2a00029a6e6dap+0, -0x1.00273715e8bc5p-56},
+{0x1.2dfffe0293e39p+0, 0x1.b7c39dab2a6f9p-54},
+{0x1.31ffff7dcf082p+0, 0x1.df1336edc5254p-56},
+{0x1.35ffff05a8b6p+0, -0x1.e03564ccd31ebp-54},
+{0x1.3a0002e0eaeccp+0, 0x1.5f0e74bd3a477p-56},
+{0x1.3e000043bb236p+0, 0x1.c7dcb149d8833p-54},
+{0x1.4200002d187ffp+0, 0x1.e08afcf2d3d28p-56},
+{0x1.460000d387cb1p+0, 0x1.20837856599a6p-55},
+{0x1.4a00004569f89p+0, -0x1.9fa5c904fbcd2p-55},
+{0x1.4e000043543f3p+0, -0x1.81125ed175329p-56},
+{0x1.51fffcc027f0fp+0, 0x1.883d8847754dcp-54},
+{0x1.55ffffd87b36fp+0, -0x1.709e731d02807p-55},
+{0x1.59ffff21df7bap+0, 0x1.7f79f68727b02p-55},
+{0x1.5dfffebfc3481p+0, -0x1.180902e30e93ep-54},
+},
+#endif
+};
diff --git a/src/math/log2_data.h b/src/math/log2_data.h
new file mode 100644
index 0000000..276a786
--- /dev/null
+++ b/src/math/log2_data.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _LOG2_DATA_H
+#define _LOG2_DATA_H
+
+#include <features.h>
+
+#define LOG2_TABLE_BITS 6
+#define LOG2_POLY_ORDER 7
+#define LOG2_POLY1_ORDER 11
+extern hidden const struct log2_data {
+	double invln2hi;
+	double invln2lo;
+	double poly[LOG2_POLY_ORDER - 1];
+	double poly1[LOG2_POLY1_ORDER - 1];
+	struct {
+		double invc, logc;
+	} tab[1 << LOG2_TABLE_BITS];
+#if !__FP_FAST_FMA
+	struct {
+		double chi, clo;
+	} tab2[1 << LOG2_TABLE_BITS];
+#endif
+} __log2_data;
+
+#endif
diff --git a/src/math/log2f.c b/src/math/log2f.c
index 8b6ba37..dbaf8d5 100644
--- a/src/math/log2f.c
+++ b/src/math/log2f.c
@@ -2,8 +2,7 @@
  * See comments in log2.c.
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 ivln2hi =  1.4428710938e+00, /* 0x3fb8b000 */
diff --git a/src/math/log2l.c b/src/math/log2l.c
new file mode 100644
index 0000000..722b451
--- /dev/null
+++ b/src/math/log2l.c
@@ -0,0 +1,182 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_log2l.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Base 2 logarithm, long double precision
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, log2l();
+ *
+ * y = log2l( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns the base 2 logarithm of x.
+ *
+ * The argument is separated into its exponent and fractional
+ * parts.  If the exponent is between -1 and +1, the (natural)
+ * logarithm of the fraction is approximated by
+ *
+ *     log(1+x) = x - 0.5 x**2 + x**3 P(x)/Q(x).
+ *
+ * Otherwise, setting  z = 2(x-1)/x+1),
+ *
+ *     log(x) = z + z**3 P(z)/Q(z).
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE      0.5, 2.0     30000      9.8e-20     2.7e-20
+ *    IEEE     exp(+-10000)  70000      5.4e-20     2.3e-20
+ *
+ * In the tests over the interval exp(+-10000), the logarithms
+ * of the random arguments were uniformly distributed over
+ * [-10000, +10000].
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double log2l(long double x)
+{
+	return log2(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/* Coefficients for ln(1+x) = x - x**2/2 + x**3 P(x)/Q(x)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 6.2e-22
+ */
+static const long double P[] = {
+ 4.9962495940332550844739E-1L,
+ 1.0767376367209449010438E1L,
+ 7.7671073698359539859595E1L,
+ 2.5620629828144409632571E2L,
+ 4.2401812743503691187826E2L,
+ 3.4258224542413922935104E2L,
+ 1.0747524399916215149070E2L,
+};
+static const long double Q[] = {
+/* 1.0000000000000000000000E0,*/
+ 2.3479774160285863271658E1L,
+ 1.9444210022760132894510E2L,
+ 7.7952888181207260646090E2L,
+ 1.6911722418503949084863E3L,
+ 2.0307734695595183428202E3L,
+ 1.2695660352705325274404E3L,
+ 3.2242573199748645407652E2L,
+};
+
+/* Coefficients for log(x) = z + z^3 P(z^2)/Q(z^2),
+ * where z = 2(x-1)/(x+1)
+ * 1/sqrt(2) <= x < sqrt(2)
+ * Theoretical peak relative error = 6.16e-22
+ */
+static const long double R[4] = {
+ 1.9757429581415468984296E-3L,
+-7.1990767473014147232598E-1L,
+ 1.0777257190312272158094E1L,
+-3.5717684488096787370998E1L,
+};
+static const long double S[4] = {
+/* 1.00000000000000000000E0L,*/
+-2.6201045551331104417768E1L,
+ 1.9361891836232102174846E2L,
+-4.2861221385716144629696E2L,
+};
+/* log2(e) - 1 */
+#define LOG2EA 4.4269504088896340735992e-1L
+
+#define SQRTH 0.70710678118654752440L
+
+long double log2l(long double x)
+{
+	long double y, z;
+	int e;
+
+	if (isnan(x))
+		return x;
+	if (x == INFINITY)
+		return x;
+	if (x <= 0.0) {
+		if (x == 0.0)
+			return -1/(x*x); /* -inf with divbyzero */
+		return 0/0.0f; /* nan with invalid */
+	}
+
+	/* separate mantissa from exponent */
+	/* Note, frexp is used so that denormal numbers
+	 * will be handled properly.
+	 */
+	x = frexpl(x, &e);
+
+	/* logarithm using log(x) = z + z**3 P(z)/Q(z),
+	 * where z = 2(x-1)/x+1)
+	 */
+	if (e > 2 || e < -2) {
+		if (x < SQRTH) {  /* 2(2x-1)/(2x+1) */
+			e -= 1;
+			z = x - 0.5;
+			y = 0.5 * z + 0.5;
+		} else {  /*  2 (x-1)/(x+1)   */
+			z = x - 0.5;
+			z -= 0.5;
+			y = 0.5 * x + 0.5;
+		}
+		x = z / y;
+		z = x*x;
+		y = x * (z * __polevll(z, R, 3) / __p1evll(z, S, 3));
+		goto done;
+	}
+
+	/* logarithm using log(1+x) = x - .5x**2 + x**3 P(x)/Q(x) */
+	if (x < SQRTH) {
+		e -= 1;
+		x = 2.0*x - 1.0;
+	} else {
+		x = x - 1.0;
+	}
+	z = x*x;
+	y = x * (z * __polevll(x, P, 6) / __p1evll(x, Q, 7));
+	y = y - 0.5*z;
+
+done:
+	/* Multiply log of fraction by log2(e)
+	 * and base 2 exponent by 1
+	 *
+	 * ***CAUTION***
+	 *
+	 * This sequence of operations is critical and it may
+	 * be horribly defeated by some compiler optimizers.
+	 */
+	z = y * LOG2EA;
+	z += x * LOG2EA;
+	z += y;
+	z += x;
+	z += e;
+	return z;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double log2l(long double x)
+{
+	return log2(x);
+}
+#endif
diff --git a/src/math/log_data.c b/src/math/log_data.c
new file mode 100644
index 0000000..1a6ec71
--- /dev/null
+++ b/src/math/log_data.c
@@ -0,0 +1,328 @@
+/*
+ * Data for log.
+ *
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "log_data.h"
+
+#define N (1 << LOG_TABLE_BITS)
+
+const struct log_data __log_data = {
+.ln2hi = 0x1.62e42fefa3800p-1,
+.ln2lo = 0x1.ef35793c76730p-45,
+.poly1 = {
+// relative error: 0x1.c04d76cp-63
+// in -0x1p-4 0x1.09p-4 (|log(1+x)| > 0x1p-4 outside the interval)
+-0x1p-1,
+0x1.5555555555577p-2,
+-0x1.ffffffffffdcbp-3,
+0x1.999999995dd0cp-3,
+-0x1.55555556745a7p-3,
+0x1.24924a344de3p-3,
+-0x1.fffffa4423d65p-4,
+0x1.c7184282ad6cap-4,
+-0x1.999eb43b068ffp-4,
+0x1.78182f7afd085p-4,
+-0x1.5521375d145cdp-4,
+},
+.poly = {
+// relative error: 0x1.926199e8p-56
+// abs error: 0x1.882ff33p-65
+// in -0x1.fp-9 0x1.fp-9
+-0x1.0000000000001p-1,
+0x1.555555551305bp-2,
+-0x1.fffffffeb459p-3,
+0x1.999b324f10111p-3,
+-0x1.55575e506c89fp-3,
+},
+/* Algorithm:
+
+	x = 2^k z
+	log(x) = k ln2 + log(c) + log(z/c)
+	log(z/c) = poly(z/c - 1)
+
+where z is in [1.6p-1; 1.6p0] which is split into N subintervals and z falls
+into the ith one, then table entries are computed as
+
+	tab[i].invc = 1/c
+	tab[i].logc = (double)log(c)
+	tab2[i].chi = (double)c
+	tab2[i].clo = (double)(c - (double)c)
+
+where c is near the center of the subinterval and is chosen by trying +-2^29
+floating point invc candidates around 1/center and selecting one for which
+
+	1) the rounding error in 0x1.8p9 + logc is 0,
+	2) the rounding error in z - chi - clo is < 0x1p-66 and
+	3) the rounding error in (double)log(c) is minimized (< 0x1p-66).
+
+Note: 1) ensures that k*ln2hi + logc can be computed without rounding error,
+2) ensures that z/c - 1 can be computed as (z - chi - clo)*invc with close to
+a single rounding error when there is no fast fma for z*invc - 1, 3) ensures
+that logc + poly(z/c - 1) has small error, however near x == 1 when
+|log(x)| < 0x1p-4, this is not enough so that is special cased.  */
+.tab = {
+{0x1.734f0c3e0de9fp+0, -0x1.7cc7f79e69000p-2},
+{0x1.713786a2ce91fp+0, -0x1.76feec20d0000p-2},
+{0x1.6f26008fab5a0p+0, -0x1.713e31351e000p-2},
+{0x1.6d1a61f138c7dp+0, -0x1.6b85b38287800p-2},
+{0x1.6b1490bc5b4d1p+0, -0x1.65d5590807800p-2},
+{0x1.69147332f0cbap+0, -0x1.602d076180000p-2},
+{0x1.6719f18224223p+0, -0x1.5a8ca86909000p-2},
+{0x1.6524f99a51ed9p+0, -0x1.54f4356035000p-2},
+{0x1.63356aa8f24c4p+0, -0x1.4f637c36b4000p-2},
+{0x1.614b36b9ddc14p+0, -0x1.49da7fda85000p-2},
+{0x1.5f66452c65c4cp+0, -0x1.445923989a800p-2},
+{0x1.5d867b5912c4fp+0, -0x1.3edf439b0b800p-2},
+{0x1.5babccb5b90dep+0, -0x1.396ce448f7000p-2},
+{0x1.59d61f2d91a78p+0, -0x1.3401e17bda000p-2},
+{0x1.5805612465687p+0, -0x1.2e9e2ef468000p-2},
+{0x1.56397cee76bd3p+0, -0x1.2941b3830e000p-2},
+{0x1.54725e2a77f93p+0, -0x1.23ec58cda8800p-2},
+{0x1.52aff42064583p+0, -0x1.1e9e129279000p-2},
+{0x1.50f22dbb2bddfp+0, -0x1.1956d2b48f800p-2},
+{0x1.4f38f4734ded7p+0, -0x1.141679ab9f800p-2},
+{0x1.4d843cfde2840p+0, -0x1.0edd094ef9800p-2},
+{0x1.4bd3ec078a3c8p+0, -0x1.09aa518db1000p-2},
+{0x1.4a27fc3e0258ap+0, -0x1.047e65263b800p-2},
+{0x1.4880524d48434p+0, -0x1.feb224586f000p-3},
+{0x1.46dce1b192d0bp+0, -0x1.f474a7517b000p-3},
+{0x1.453d9d3391854p+0, -0x1.ea4443d103000p-3},
+{0x1.43a2744b4845ap+0, -0x1.e020d44e9b000p-3},
+{0x1.420b54115f8fbp+0, -0x1.d60a22977f000p-3},
+{0x1.40782da3ef4b1p+0, -0x1.cc00104959000p-3},
+{0x1.3ee8f5d57fe8fp+0, -0x1.c202956891000p-3},
+{0x1.3d5d9a00b4ce9p+0, -0x1.b81178d811000p-3},
+{0x1.3bd60c010c12bp+0, -0x1.ae2c9ccd3d000p-3},
+{0x1.3a5242b75dab8p+0, -0x1.a45402e129000p-3},
+{0x1.38d22cd9fd002p+0, -0x1.9a877681df000p-3},
+{0x1.3755bc5847a1cp+0, -0x1.90c6d69483000p-3},
+{0x1.35dce49ad36e2p+0, -0x1.87120a645c000p-3},
+{0x1.34679984dd440p+0, -0x1.7d68fb4143000p-3},
+{0x1.32f5cceffcb24p+0, -0x1.73cb83c627000p-3},
+{0x1.3187775a10d49p+0, -0x1.6a39a9b376000p-3},
+{0x1.301c8373e3990p+0, -0x1.60b3154b7a000p-3},
+{0x1.2eb4ebb95f841p+0, -0x1.5737d76243000p-3},
+{0x1.2d50a0219a9d1p+0, -0x1.4dc7b8fc23000p-3},
+{0x1.2bef9a8b7fd2ap+0, -0x1.4462c51d20000p-3},
+{0x1.2a91c7a0c1babp+0, -0x1.3b08abc830000p-3},
+{0x1.293726014b530p+0, -0x1.31b996b490000p-3},
+{0x1.27dfa5757a1f5p+0, -0x1.2875490a44000p-3},
+{0x1.268b39b1d3bbfp+0, -0x1.1f3b9f879a000p-3},
+{0x1.2539d838ff5bdp+0, -0x1.160c8252ca000p-3},
+{0x1.23eb7aac9083bp+0, -0x1.0ce7f57f72000p-3},
+{0x1.22a012ba940b6p+0, -0x1.03cdc49fea000p-3},
+{0x1.2157996cc4132p+0, -0x1.f57bdbc4b8000p-4},
+{0x1.201201dd2fc9bp+0, -0x1.e370896404000p-4},
+{0x1.1ecf4494d480bp+0, -0x1.d17983ef94000p-4},
+{0x1.1d8f5528f6569p+0, -0x1.bf9674ed8a000p-4},
+{0x1.1c52311577e7cp+0, -0x1.adc79202f6000p-4},
+{0x1.1b17c74cb26e9p+0, -0x1.9c0c3e7288000p-4},
+{0x1.19e010c2c1ab6p+0, -0x1.8a646b372c000p-4},
+{0x1.18ab07bb670bdp+0, -0x1.78d01b3ac0000p-4},
+{0x1.1778a25efbcb6p+0, -0x1.674f145380000p-4},
+{0x1.1648d354c31dap+0, -0x1.55e0e6d878000p-4},
+{0x1.151b990275fddp+0, -0x1.4485cdea1e000p-4},
+{0x1.13f0ea432d24cp+0, -0x1.333d94d6aa000p-4},
+{0x1.12c8b7210f9dap+0, -0x1.22079f8c56000p-4},
+{0x1.11a3028ecb531p+0, -0x1.10e4698622000p-4},
+{0x1.107fbda8434afp+0, -0x1.ffa6c6ad20000p-5},
+{0x1.0f5ee0f4e6bb3p+0, -0x1.dda8d4a774000p-5},
+{0x1.0e4065d2a9fcep+0, -0x1.bbcece4850000p-5},
+{0x1.0d244632ca521p+0, -0x1.9a1894012c000p-5},
+{0x1.0c0a77ce2981ap+0, -0x1.788583302c000p-5},
+{0x1.0af2f83c636d1p+0, -0x1.5715e67d68000p-5},
+{0x1.09ddb98a01339p+0, -0x1.35c8a49658000p-5},
+{0x1.08cabaf52e7dfp+0, -0x1.149e364154000p-5},
+{0x1.07b9f2f4e28fbp+0, -0x1.e72c082eb8000p-6},
+{0x1.06ab58c358f19p+0, -0x1.a55f152528000p-6},
+{0x1.059eea5ecf92cp+0, -0x1.63d62cf818000p-6},
+{0x1.04949cdd12c90p+0, -0x1.228fb8caa0000p-6},
+{0x1.038c6c6f0ada9p+0, -0x1.c317b20f90000p-7},
+{0x1.02865137932a9p+0, -0x1.419355daa0000p-7},
+{0x1.0182427ea7348p+0, -0x1.81203c2ec0000p-8},
+{0x1.008040614b195p+0, -0x1.0040979240000p-9},
+{0x1.fe01ff726fa1ap-1, 0x1.feff384900000p-9},
+{0x1.fa11cc261ea74p-1, 0x1.7dc41353d0000p-7},
+{0x1.f6310b081992ep-1, 0x1.3cea3c4c28000p-6},
+{0x1.f25f63ceeadcdp-1, 0x1.b9fc114890000p-6},
+{0x1.ee9c8039113e7p-1, 0x1.1b0d8ce110000p-5},
+{0x1.eae8078cbb1abp-1, 0x1.58a5bd001c000p-5},
+{0x1.e741aa29d0c9bp-1, 0x1.95c8340d88000p-5},
+{0x1.e3a91830a99b5p-1, 0x1.d276aef578000p-5},
+{0x1.e01e009609a56p-1, 0x1.07598e598c000p-4},
+{0x1.dca01e577bb98p-1, 0x1.253f5e30d2000p-4},
+{0x1.d92f20b7c9103p-1, 0x1.42edd8b380000p-4},
+{0x1.d5cac66fb5ccep-1, 0x1.606598757c000p-4},
+{0x1.d272caa5ede9dp-1, 0x1.7da76356a0000p-4},
+{0x1.cf26e3e6b2ccdp-1, 0x1.9ab434e1c6000p-4},
+{0x1.cbe6da2a77902p-1, 0x1.b78c7bb0d6000p-4},
+{0x1.c8b266d37086dp-1, 0x1.d431332e72000p-4},
+{0x1.c5894bd5d5804p-1, 0x1.f0a3171de6000p-4},
+{0x1.c26b533bb9f8cp-1, 0x1.067152b914000p-3},
+{0x1.bf583eeece73fp-1, 0x1.147858292b000p-3},
+{0x1.bc4fd75db96c1p-1, 0x1.2266ecdca3000p-3},
+{0x1.b951e0c864a28p-1, 0x1.303d7a6c55000p-3},
+{0x1.b65e2c5ef3e2cp-1, 0x1.3dfc33c331000p-3},
+{0x1.b374867c9888bp-1, 0x1.4ba366b7a8000p-3},
+{0x1.b094b211d304ap-1, 0x1.5933928d1f000p-3},
+{0x1.adbe885f2ef7ep-1, 0x1.66acd2418f000p-3},
+{0x1.aaf1d31603da2p-1, 0x1.740f8ec669000p-3},
+{0x1.a82e63fd358a7p-1, 0x1.815c0f51af000p-3},
+{0x1.a5740ef09738bp-1, 0x1.8e92954f68000p-3},
+{0x1.a2c2a90ab4b27p-1, 0x1.9bb3602f84000p-3},
+{0x1.a01a01393f2d1p-1, 0x1.a8bed1c2c0000p-3},
+{0x1.9d79f24db3c1bp-1, 0x1.b5b515c01d000p-3},
+{0x1.9ae2505c7b190p-1, 0x1.c2967ccbcc000p-3},
+{0x1.9852ef297ce2fp-1, 0x1.cf635d5486000p-3},
+{0x1.95cbaeea44b75p-1, 0x1.dc1bd3446c000p-3},
+{0x1.934c69de74838p-1, 0x1.e8c01b8cfe000p-3},
+{0x1.90d4f2f6752e6p-1, 0x1.f5509c0179000p-3},
+{0x1.8e6528effd79dp-1, 0x1.00e6c121fb800p-2},
+{0x1.8bfce9fcc007cp-1, 0x1.071b80e93d000p-2},
+{0x1.899c0dabec30ep-1, 0x1.0d46b9e867000p-2},
+{0x1.87427aa2317fbp-1, 0x1.13687334bd000p-2},
+{0x1.84f00acb39a08p-1, 0x1.1980d67234800p-2},
+{0x1.82a49e8653e55p-1, 0x1.1f8ffe0cc8000p-2},
+{0x1.8060195f40260p-1, 0x1.2595fd7636800p-2},
+{0x1.7e22563e0a329p-1, 0x1.2b9300914a800p-2},
+{0x1.7beb377dcb5adp-1, 0x1.3187210436000p-2},
+{0x1.79baa679725c2p-1, 0x1.377266dec1800p-2},
+{0x1.77907f2170657p-1, 0x1.3d54ffbaf3000p-2},
+{0x1.756cadbd6130cp-1, 0x1.432eee32fe000p-2},
+},
+#if !__FP_FAST_FMA
+.tab2 = {
+{0x1.61000014fb66bp-1, 0x1.e026c91425b3cp-56},
+{0x1.63000034db495p-1, 0x1.dbfea48005d41p-55},
+{0x1.650000d94d478p-1, 0x1.e7fa786d6a5b7p-55},
+{0x1.67000074e6fadp-1, 0x1.1fcea6b54254cp-57},
+{0x1.68ffffedf0faep-1, -0x1.c7e274c590efdp-56},
+{0x1.6b0000763c5bcp-1, -0x1.ac16848dcda01p-55},
+{0x1.6d0001e5cc1f6p-1, 0x1.33f1c9d499311p-55},
+{0x1.6efffeb05f63ep-1, -0x1.e80041ae22d53p-56},
+{0x1.710000e86978p-1, 0x1.bff6671097952p-56},
+{0x1.72ffffc67e912p-1, 0x1.c00e226bd8724p-55},
+{0x1.74fffdf81116ap-1, -0x1.e02916ef101d2p-57},
+{0x1.770000f679c9p-1, -0x1.7fc71cd549c74p-57},
+{0x1.78ffffa7ec835p-1, 0x1.1bec19ef50483p-55},
+{0x1.7affffe20c2e6p-1, -0x1.07e1729cc6465p-56},
+{0x1.7cfffed3fc9p-1, -0x1.08072087b8b1cp-55},
+{0x1.7efffe9261a76p-1, 0x1.dc0286d9df9aep-55},
+{0x1.81000049ca3e8p-1, 0x1.97fd251e54c33p-55},
+{0x1.8300017932c8fp-1, -0x1.afee9b630f381p-55},
+{0x1.850000633739cp-1, 0x1.9bfbf6b6535bcp-55},
+{0x1.87000204289c6p-1, -0x1.bbf65f3117b75p-55},
+{0x1.88fffebf57904p-1, -0x1.9006ea23dcb57p-55},
+{0x1.8b00022bc04dfp-1, -0x1.d00df38e04b0ap-56},
+{0x1.8cfffe50c1b8ap-1, -0x1.8007146ff9f05p-55},
+{0x1.8effffc918e43p-1, 0x1.3817bd07a7038p-55},
+{0x1.910001efa5fc7p-1, 0x1.93e9176dfb403p-55},
+{0x1.9300013467bb9p-1, 0x1.f804e4b980276p-56},
+{0x1.94fffe6ee076fp-1, -0x1.f7ef0d9ff622ep-55},
+{0x1.96fffde3c12d1p-1, -0x1.082aa962638bap-56},
+{0x1.98ffff4458a0dp-1, -0x1.7801b9164a8efp-55},
+{0x1.9afffdd982e3ep-1, -0x1.740e08a5a9337p-55},
+{0x1.9cfffed49fb66p-1, 0x1.fce08c19bep-60},
+{0x1.9f00020f19c51p-1, -0x1.a3faa27885b0ap-55},
+{0x1.a10001145b006p-1, 0x1.4ff489958da56p-56},
+{0x1.a300007bbf6fap-1, 0x1.cbeab8a2b6d18p-55},
+{0x1.a500010971d79p-1, 0x1.8fecadd78793p-55},
+{0x1.a70001df52e48p-1, -0x1.f41763dd8abdbp-55},
+{0x1.a90001c593352p-1, -0x1.ebf0284c27612p-55},
+{0x1.ab0002a4f3e4bp-1, -0x1.9fd043cff3f5fp-57},
+{0x1.acfffd7ae1ed1p-1, -0x1.23ee7129070b4p-55},
+{0x1.aefffee510478p-1, 0x1.a063ee00edea3p-57},
+{0x1.b0fffdb650d5bp-1, 0x1.a06c8381f0ab9p-58},
+{0x1.b2ffffeaaca57p-1, -0x1.9011e74233c1dp-56},
+{0x1.b4fffd995badcp-1, -0x1.9ff1068862a9fp-56},
+{0x1.b7000249e659cp-1, 0x1.aff45d0864f3ep-55},
+{0x1.b8ffff987164p-1, 0x1.cfe7796c2c3f9p-56},
+{0x1.bafffd204cb4fp-1, -0x1.3ff27eef22bc4p-57},
+{0x1.bcfffd2415c45p-1, -0x1.cffb7ee3bea21p-57},
+{0x1.beffff86309dfp-1, -0x1.14103972e0b5cp-55},
+{0x1.c0fffe1b57653p-1, 0x1.bc16494b76a19p-55},
+{0x1.c2ffff1fa57e3p-1, -0x1.4feef8d30c6edp-57},
+{0x1.c4fffdcbfe424p-1, -0x1.43f68bcec4775p-55},
+{0x1.c6fffed54b9f7p-1, 0x1.47ea3f053e0ecp-55},
+{0x1.c8fffeb998fd5p-1, 0x1.383068df992f1p-56},
+{0x1.cb0002125219ap-1, -0x1.8fd8e64180e04p-57},
+{0x1.ccfffdd94469cp-1, 0x1.e7ebe1cc7ea72p-55},
+{0x1.cefffeafdc476p-1, 0x1.ebe39ad9f88fep-55},
+{0x1.d1000169af82bp-1, 0x1.57d91a8b95a71p-56},
+{0x1.d30000d0ff71dp-1, 0x1.9c1906970c7dap-55},
+{0x1.d4fffea790fc4p-1, -0x1.80e37c558fe0cp-58},
+{0x1.d70002edc87e5p-1, -0x1.f80d64dc10f44p-56},
+{0x1.d900021dc82aap-1, -0x1.47c8f94fd5c5cp-56},
+{0x1.dafffd86b0283p-1, 0x1.c7f1dc521617ep-55},
+{0x1.dd000296c4739p-1, 0x1.8019eb2ffb153p-55},
+{0x1.defffe54490f5p-1, 0x1.e00d2c652cc89p-57},
+{0x1.e0fffcdabf694p-1, -0x1.f8340202d69d2p-56},
+{0x1.e2fffdb52c8ddp-1, 0x1.b00c1ca1b0864p-56},
+{0x1.e4ffff24216efp-1, 0x1.2ffa8b094ab51p-56},
+{0x1.e6fffe88a5e11p-1, -0x1.7f673b1efbe59p-58},
+{0x1.e9000119eff0dp-1, -0x1.4808d5e0bc801p-55},
+{0x1.eafffdfa51744p-1, 0x1.80006d54320b5p-56},
+{0x1.ed0001a127fa1p-1, -0x1.002f860565c92p-58},
+{0x1.ef00007babcc4p-1, -0x1.540445d35e611p-55},
+{0x1.f0ffff57a8d02p-1, -0x1.ffb3139ef9105p-59},
+{0x1.f30001ee58ac7p-1, 0x1.a81acf2731155p-55},
+{0x1.f4ffff5823494p-1, 0x1.a3f41d4d7c743p-55},
+{0x1.f6ffffca94c6bp-1, -0x1.202f41c987875p-57},
+{0x1.f8fffe1f9c441p-1, 0x1.77dd1f477e74bp-56},
+{0x1.fafffd2e0e37ep-1, -0x1.f01199a7ca331p-57},
+{0x1.fd0001c77e49ep-1, 0x1.181ee4bceacb1p-56},
+{0x1.feffff7e0c331p-1, -0x1.e05370170875ap-57},
+{0x1.00ffff465606ep+0, -0x1.a7ead491c0adap-55},
+{0x1.02ffff3867a58p+0, -0x1.77f69c3fcb2ep-54},
+{0x1.04ffffdfc0d17p+0, 0x1.7bffe34cb945bp-54},
+{0x1.0700003cd4d82p+0, 0x1.20083c0e456cbp-55},
+{0x1.08ffff9f2cbe8p+0, -0x1.dffdfbe37751ap-57},
+{0x1.0b000010cda65p+0, -0x1.13f7faee626ebp-54},
+{0x1.0d00001a4d338p+0, 0x1.07dfa79489ff7p-55},
+{0x1.0effffadafdfdp+0, -0x1.7040570d66bcp-56},
+{0x1.110000bbafd96p+0, 0x1.e80d4846d0b62p-55},
+{0x1.12ffffae5f45dp+0, 0x1.dbffa64fd36efp-54},
+{0x1.150000dd59ad9p+0, 0x1.a0077701250aep-54},
+{0x1.170000f21559ap+0, 0x1.dfdf9e2e3deeep-55},
+{0x1.18ffffc275426p+0, 0x1.10030dc3b7273p-54},
+{0x1.1b000123d3c59p+0, 0x1.97f7980030188p-54},
+{0x1.1cffff8299eb7p+0, -0x1.5f932ab9f8c67p-57},
+{0x1.1effff48ad4p+0, 0x1.37fbf9da75bebp-54},
+{0x1.210000c8b86a4p+0, 0x1.f806b91fd5b22p-54},
+{0x1.2300003854303p+0, 0x1.3ffc2eb9fbf33p-54},
+{0x1.24fffffbcf684p+0, 0x1.601e77e2e2e72p-56},
+{0x1.26ffff52921d9p+0, 0x1.ffcbb767f0c61p-56},
+{0x1.2900014933a3cp+0, -0x1.202ca3c02412bp-56},
+{0x1.2b00014556313p+0, -0x1.2808233f21f02p-54},
+{0x1.2cfffebfe523bp+0, -0x1.8ff7e384fdcf2p-55},
+{0x1.2f0000bb8ad96p+0, -0x1.5ff51503041c5p-55},
+{0x1.30ffffb7ae2afp+0, -0x1.10071885e289dp-55},
+{0x1.32ffffeac5f7fp+0, -0x1.1ff5d3fb7b715p-54},
+{0x1.350000ca66756p+0, 0x1.57f82228b82bdp-54},
+{0x1.3700011fbf721p+0, 0x1.000bac40dd5ccp-55},
+{0x1.38ffff9592fb9p+0, -0x1.43f9d2db2a751p-54},
+{0x1.3b00004ddd242p+0, 0x1.57f6b707638e1p-55},
+{0x1.3cffff5b2c957p+0, 0x1.a023a10bf1231p-56},
+{0x1.3efffeab0b418p+0, 0x1.87f6d66b152bp-54},
+{0x1.410001532aff4p+0, 0x1.7f8375f198524p-57},
+{0x1.4300017478b29p+0, 0x1.301e672dc5143p-55},
+{0x1.44fffe795b463p+0, 0x1.9ff69b8b2895ap-55},
+{0x1.46fffe80475ep+0, -0x1.5c0b19bc2f254p-54},
+{0x1.48fffef6fc1e7p+0, 0x1.b4009f23a2a72p-54},
+{0x1.4afffe5bea704p+0, -0x1.4ffb7bf0d7d45p-54},
+{0x1.4d000171027dep+0, -0x1.9c06471dc6a3dp-54},
+{0x1.4f0000ff03ee2p+0, 0x1.77f890b85531cp-54},
+{0x1.5100012dc4bd1p+0, 0x1.004657166a436p-57},
+{0x1.530001605277ap+0, -0x1.6bfcece233209p-54},
+{0x1.54fffecdb704cp+0, -0x1.902720505a1d7p-55},
+{0x1.56fffef5f54a9p+0, 0x1.bbfe60ec96412p-54},
+{0x1.5900017e61012p+0, 0x1.87ec581afef9p-55},
+{0x1.5b00003c93e92p+0, -0x1.f41080abf0ccp-54},
+{0x1.5d0001d4919bcp+0, -0x1.8812afb254729p-54},
+{0x1.5efffe7b87a89p+0, -0x1.47eb780ed6904p-54},
+},
+#endif
+};
diff --git a/src/math/log_data.h b/src/math/log_data.h
new file mode 100644
index 0000000..1be22ab
--- /dev/null
+++ b/src/math/log_data.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _LOG_DATA_H
+#define _LOG_DATA_H
+
+#include <features.h>
+
+#define LOG_TABLE_BITS 7
+#define LOG_POLY_ORDER 6
+#define LOG_POLY1_ORDER 12
+extern hidden const struct log_data {
+	double ln2hi;
+	double ln2lo;
+	double poly[LOG_POLY_ORDER - 1]; /* First coefficient is 1.  */
+	double poly1[LOG_POLY1_ORDER - 1];
+	struct {
+		double invc, logc;
+	} tab[1 << LOG_TABLE_BITS];
+#if !__FP_FAST_FMA
+	struct {
+		double chi, clo;
+	} tab2[1 << LOG_TABLE_BITS];
+#endif
+} __log_data;
+
+#endif
diff --git a/src/math/logb.c b/src/math/logb.c
new file mode 100644
index 0000000..7f8bdfa
--- /dev/null
+++ b/src/math/logb.c
@@ -0,0 +1,17 @@
+#include <math.h>
+
+/*
+special cases:
+	logb(+-0) = -inf, and raise divbyzero
+	logb(+-inf) = +inf
+	logb(nan) = nan
+*/
+
+double logb(double x)
+{
+	if (!isfinite(x))
+		return x * x;
+	if (x == 0)
+		return -1/(x*x);
+	return ilogb(x);
+}
diff --git a/src/math/logbf.c b/src/math/logbf.c
new file mode 100644
index 0000000..a0a0b5e
--- /dev/null
+++ b/src/math/logbf.c
@@ -0,0 +1,10 @@
+#include <math.h>
+
+float logbf(float x)
+{
+	if (!isfinite(x))
+		return x * x;
+	if (x == 0)
+		return -1/(x*x);
+	return ilogbf(x);
+}
diff --git a/src/math/logbl.c b/src/math/logbl.c
new file mode 100644
index 0000000..962973a
--- /dev/null
+++ b/src/math/logbl.c
@@ -0,0 +1,16 @@
+#include <math.h>
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double logbl(long double x)
+{
+	return logb(x);
+}
+#else
+long double logbl(long double x)
+{
+	if (!isfinite(x))
+		return x * x;
+	if (x == 0)
+		return -1/(x*x);
+	return ilogbl(x);
+}
+#endif
diff --git a/src/math/logf.c b/src/math/logf.c
index bbce38f..98853f0 100644
--- a/src/math/logf.c
+++ b/src/math/logf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 ln2_hi = 6.9313812256e-01, /* 0x3f317180 */
diff --git a/src/math/logf_data.c b/src/math/logf_data.c
new file mode 100644
index 0000000..857221f
--- /dev/null
+++ b/src/math/logf_data.c
@@ -0,0 +1,33 @@
+/*
+ * Data definition for logf.
+ *
+ * Copyright (c) 2017-2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "logf_data.h"
+
+const struct logf_data __logf_data = {
+  .tab = {
+  { 0x1.661ec79f8f3bep+0, -0x1.57bf7808caadep-2 },
+  { 0x1.571ed4aaf883dp+0, -0x1.2bef0a7c06ddbp-2 },
+  { 0x1.49539f0f010bp+0, -0x1.01eae7f513a67p-2 },
+  { 0x1.3c995b0b80385p+0, -0x1.b31d8a68224e9p-3 },
+  { 0x1.30d190c8864a5p+0, -0x1.6574f0ac07758p-3 },
+  { 0x1.25e227b0b8eap+0, -0x1.1aa2bc79c81p-3 },
+  { 0x1.1bb4a4a1a343fp+0, -0x1.a4e76ce8c0e5ep-4 },
+  { 0x1.12358f08ae5bap+0, -0x1.1973c5a611cccp-4 },
+  { 0x1.0953f419900a7p+0, -0x1.252f438e10c1ep-5 },
+  { 0x1p+0, 0x0p+0 },
+  { 0x1.e608cfd9a47acp-1, 0x1.aa5aa5df25984p-5 },
+  { 0x1.ca4b31f026aap-1, 0x1.c5e53aa362eb4p-4 },
+  { 0x1.b2036576afce6p-1, 0x1.526e57720db08p-3 },
+  { 0x1.9c2d163a1aa2dp-1, 0x1.bc2860d22477p-3 },
+  { 0x1.886e6037841edp-1, 0x1.1058bc8a07ee1p-2 },
+  { 0x1.767dcf5534862p-1, 0x1.4043057b6ee09p-2 },
+  },
+  .ln2 = 0x1.62e42fefa39efp-1,
+  .poly = {
+  -0x1.00ea348b88334p-2, 0x1.5575b0be00b6ap-2, -0x1.ffffef20a4123p-2,
+  }
+};
diff --git a/src/math/logf_data.h b/src/math/logf_data.h
new file mode 100644
index 0000000..00cff6f
--- /dev/null
+++ b/src/math/logf_data.h
@@ -0,0 +1,20 @@
+/*
+ * Copyright (c) 2017-2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _LOGF_DATA_H
+#define _LOGF_DATA_H
+
+#include <features.h>
+
+#define LOGF_TABLE_BITS 4
+#define LOGF_POLY_ORDER 4
+extern hidden const struct logf_data {
+	struct {
+		double invc, logc;
+	} tab[1 << LOGF_TABLE_BITS];
+	double ln2;
+	double poly[LOGF_POLY_ORDER - 1]; /* First order coefficient is 1.  */
+} __logf_data;
+
+#endif
diff --git a/src/math/lrint.c b/src/math/lrint.c
new file mode 100644
index 0000000..ddee7a0
--- /dev/null
+++ b/src/math/lrint.c
@@ -0,0 +1,72 @@
+#include <limits.h>
+#include <fenv.h>
+#include <math.h>
+#include "libm.h"
+
+/*
+If the result cannot be represented (overflow, nan), then
+lrint raises the invalid exception.
+
+Otherwise if the input was not an integer then the inexact
+exception is raised.
+
+C99 is a bit vague about whether inexact exception is
+allowed to be raised when invalid is raised.
+(F.9 explicitly allows spurious inexact exceptions, F.9.6.5
+does not make it clear if that rule applies to lrint, but
+IEEE 754r 7.8 seems to forbid spurious inexact exception in
+the ineger conversion functions)
+
+So we try to make sure that no spurious inexact exception is
+raised in case of an overflow.
+
+If the bit size of long > precision of double, then there
+cannot be inexact rounding in case the result overflows,
+otherwise LONG_MAX and LONG_MIN can be represented exactly
+as a double.
+*/
+
+#if LONG_MAX < 1U<<53 && defined(FE_INEXACT)
+#include <float.h>
+#include <stdint.h>
+#if FLT_EVAL_METHOD==0 || FLT_EVAL_METHOD==1
+#define EPS DBL_EPSILON
+#elif FLT_EVAL_METHOD==2
+#define EPS LDBL_EPSILON
+#endif
+#ifdef __GNUC__
+/* avoid stack frame in lrint */
+__attribute__((noinline))
+#endif
+static long lrint_slow(double x)
+{
+	#pragma STDC FENV_ACCESS ON
+	int e;
+
+	e = fetestexcept(FE_INEXACT);
+	x = rint(x);
+	if (!e && (x > LONG_MAX || x < LONG_MIN))
+		feclearexcept(FE_INEXACT);
+	/* conversion */
+	return x;
+}
+
+long lrint(double x)
+{
+	uint32_t abstop = asuint64(x)>>32 & 0x7fffffff;
+	uint64_t sign = asuint64(x) & (1ULL << 63);
+
+	if (abstop < 0x41dfffff) {
+		/* |x| < 0x7ffffc00, no overflow */
+		double_t toint = asdouble(asuint64(1/EPS) | sign);
+		double_t y = x + toint - toint;
+		return (long)y;
+	}
+	return lrint_slow(x);
+}
+#else
+long lrint(double x)
+{
+	return rint(x);
+}
+#endif
diff --git a/src/math/lrintf.c b/src/math/lrintf.c
new file mode 100644
index 0000000..ca0b6a4
--- /dev/null
+++ b/src/math/lrintf.c
@@ -0,0 +1,8 @@
+#include <math.h>
+
+/* uses LONG_MAX > 2^24, see comments in lrint.c */
+
+long lrintf(float x)
+{
+	return rintf(x);
+}
diff --git a/src/math/lrintl.c b/src/math/lrintl.c
new file mode 100644
index 0000000..b2a8106
--- /dev/null
+++ b/src/math/lrintl.c
@@ -0,0 +1,36 @@
+#include <limits.h>
+#include <fenv.h>
+#include "libm.h"
+
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long lrintl(long double x)
+{
+	return lrint(x);
+}
+#elif defined(FE_INEXACT)
+/*
+see comments in lrint.c
+
+Note that if LONG_MAX == 0x7fffffffffffffff && LDBL_MANT_DIG == 64
+then x == 2**63 - 0.5 is the only input that overflows and
+raises inexact (with tonearest or upward rounding mode)
+*/
+long lrintl(long double x)
+{
+	#pragma STDC FENV_ACCESS ON
+	int e;
+
+	e = fetestexcept(FE_INEXACT);
+	x = rintl(x);
+	if (!e && (x > LONG_MAX || x < LONG_MIN))
+		feclearexcept(FE_INEXACT);
+	/* conversion */
+	return x;
+}
+#else
+long lrintl(long double x)
+{
+	return rintl(x);
+}
+#endif
diff --git a/src/math/lround.c b/src/math/lround.c
index b8b7954..ca705b6 100644
--- a/src/math/lround.c
+++ b/src/math/lround.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 
 long lround(double x)
 {
diff --git a/src/math/lroundf.c b/src/math/lroundf.c
index c4707e7..7e75ba6 100644
--- a/src/math/lroundf.c
+++ b/src/math/lroundf.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 
 long lroundf(float x)
 {
diff --git a/src/math/lroundl.c b/src/math/lroundl.c
index 094fdf6..6c86bcf 100644
--- a/src/math/lroundl.c
+++ b/src/math/lroundl.c
@@ -1,4 +1,4 @@
-#include <math.h>
+#include "libm.h"
 
 long lroundl(long double x)
 {
diff --git a/src/math/modf.c b/src/math/modf.c
index ef64a4a..1c8a1db 100644
--- a/src/math/modf.c
+++ b/src/math/modf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double modf(double x, double *iptr)
 {
diff --git a/src/math/modff.c b/src/math/modff.c
index 0e495ac..639514e 100644
--- a/src/math/modff.c
+++ b/src/math/modff.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float modff(float x, float *iptr)
 {
diff --git a/src/math/modfl.c b/src/math/modfl.c
new file mode 100644
index 0000000..a47b192
--- /dev/null
+++ b/src/math/modfl.c
@@ -0,0 +1,53 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double modfl(long double x, long double *iptr)
+{
+	double d;
+	long double r;
+
+	r = modf(x, &d);
+	*iptr = d;
+	return r;
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+
+static const long double toint = 1/LDBL_EPSILON;
+
+long double modfl(long double x, long double *iptr)
+{
+	union ldshape u = {x};
+	int e = (u.i.se & 0x7fff) - 0x3fff;
+	int s = u.i.se >> 15;
+	long double absx;
+	long double y;
+
+	/* no fractional part */
+	if (e >= LDBL_MANT_DIG-1) {
+		*iptr = x;
+		if (isnan(x))
+			return x;
+		return s ? -0.0 : 0.0;
+	}
+
+	/* no integral part*/
+	if (e < 0) {
+		*iptr = s ? -0.0 : 0.0;
+		return x;
+	}
+
+	/* raises spurious inexact */
+	absx = s ? -x : x;
+	y = absx + toint - toint - absx;
+	if (y == 0) {
+		*iptr = x;
+		return s ? -0.0 : 0.0;
+	}
+	if (y > 0)
+		y -= 1;
+	if (s)
+		y = -y;
+	*iptr = x + y;
+	return -y;
+}
+#endif
diff --git a/src/math/nan.c b/src/math/nan.c
new file mode 100644
index 0000000..9e0826c
--- /dev/null
+++ b/src/math/nan.c
@@ -0,0 +1,6 @@
+#include <math.h>
+
+double nan(const char *s)
+{
+	return NAN;
+}
diff --git a/src/math/nanf.c b/src/math/nanf.c
new file mode 100644
index 0000000..752ce54
--- /dev/null
+++ b/src/math/nanf.c
@@ -0,0 +1,6 @@
+#include <math.h>
+
+float nanf(const char *s)
+{
+	return NAN;
+}
diff --git a/src/math/nanl.c b/src/math/nanl.c
new file mode 100644
index 0000000..969af56
--- /dev/null
+++ b/src/math/nanl.c
@@ -0,0 +1,6 @@
+#include <math.h>
+
+long double nanl(const char *s)
+{
+	return NAN;
+}
diff --git a/src/math/nearbyint.c b/src/math/nearbyint.c
new file mode 100644
index 0000000..f4e8aac
--- /dev/null
+++ b/src/math/nearbyint.c
@@ -0,0 +1,20 @@
+#include <fenv.h>
+#include <math.h>
+
+/* nearbyint is the same as rint, but it must not raise the inexact exception */
+
+double nearbyint(double x)
+{
+#ifdef FE_INEXACT
+	#pragma STDC FENV_ACCESS ON
+	int e;
+
+	e = fetestexcept(FE_INEXACT);
+#endif
+	x = rint(x);
+#ifdef FE_INEXACT
+	if (!e)
+		feclearexcept(FE_INEXACT);
+#endif
+	return x;
+}
diff --git a/src/math/nearbyintf.c b/src/math/nearbyintf.c
new file mode 100644
index 0000000..092e9ff
--- /dev/null
+++ b/src/math/nearbyintf.c
@@ -0,0 +1,18 @@
+#include <fenv.h>
+#include <math.h>
+
+float nearbyintf(float x)
+{
+#ifdef FE_INEXACT
+	#pragma STDC FENV_ACCESS ON
+	int e;
+
+	e = fetestexcept(FE_INEXACT);
+#endif
+	x = rintf(x);
+#ifdef FE_INEXACT
+	if (!e)
+		feclearexcept(FE_INEXACT);
+#endif
+	return x;
+}
diff --git a/src/math/nearbyintl.c b/src/math/nearbyintl.c
new file mode 100644
index 0000000..8285249
--- /dev/null
+++ b/src/math/nearbyintl.c
@@ -0,0 +1,26 @@
+#include <math.h>
+#include <float.h>
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double nearbyintl(long double x)
+{
+	return nearbyint(x);
+}
+#else
+#include <fenv.h>
+long double nearbyintl(long double x)
+{
+#ifdef FE_INEXACT
+	#pragma STDC FENV_ACCESS ON
+	int e;
+
+	e = fetestexcept(FE_INEXACT);
+#endif
+	x = rintl(x);
+#ifdef FE_INEXACT
+	if (!e)
+		feclearexcept(FE_INEXACT);
+#endif
+	return x;
+}
+#endif
diff --git a/src/math/nextafter.c b/src/math/nextafter.c
new file mode 100644
index 0000000..ab5795a
--- /dev/null
+++ b/src/math/nextafter.c
@@ -0,0 +1,31 @@
+#include "libm.h"
+
+double nextafter(double x, double y)
+{
+	union {double f; uint64_t i;} ux={x}, uy={y};
+	uint64_t ax, ay;
+	int e;
+
+	if (isnan(x) || isnan(y))
+		return x + y;
+	if (ux.i == uy.i)
+		return y;
+	ax = ux.i & -1ULL/2;
+	ay = uy.i & -1ULL/2;
+	if (ax == 0) {
+		if (ay == 0)
+			return y;
+		ux.i = (uy.i & 1ULL<<63) | 1;
+	} else if (ax > ay || ((ux.i ^ uy.i) & 1ULL<<63))
+		ux.i--;
+	else
+		ux.i++;
+	e = ux.i >> 52 & 0x7ff;
+	/* raise overflow if ux.f is infinite and x is finite */
+	if (e == 0x7ff)
+		FORCE_EVAL(x+x);
+	/* raise underflow if ux.f is subnormal or zero */
+	if (e == 0)
+		FORCE_EVAL(x*x + ux.f*ux.f);
+	return ux.f;
+}
diff --git a/src/math/nextafterf.c b/src/math/nextafterf.c
new file mode 100644
index 0000000..75a09f7
--- /dev/null
+++ b/src/math/nextafterf.c
@@ -0,0 +1,30 @@
+#include "libm.h"
+
+float nextafterf(float x, float y)
+{
+	union {float f; uint32_t i;} ux={x}, uy={y};
+	uint32_t ax, ay, e;
+
+	if (isnan(x) || isnan(y))
+		return x + y;
+	if (ux.i == uy.i)
+		return y;
+	ax = ux.i & 0x7fffffff;
+	ay = uy.i & 0x7fffffff;
+	if (ax == 0) {
+		if (ay == 0)
+			return y;
+		ux.i = (uy.i & 0x80000000) | 1;
+	} else if (ax > ay || ((ux.i ^ uy.i) & 0x80000000))
+		ux.i--;
+	else
+		ux.i++;
+	e = ux.i & 0x7f800000;
+	/* raise overflow if ux.f is infinite and x is finite */
+	if (e == 0x7f800000)
+		FORCE_EVAL(x+x);
+	/* raise underflow if ux.f is subnormal or zero */
+	if (e == 0)
+		FORCE_EVAL(x*x + ux.f*ux.f);
+	return ux.f;
+}
diff --git a/src/math/nextafterl.c b/src/math/nextafterl.c
new file mode 100644
index 0000000..37e858f
--- /dev/null
+++ b/src/math/nextafterl.c
@@ -0,0 +1,75 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double nextafterl(long double x, long double y)
+{
+	return nextafter(x, y);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+long double nextafterl(long double x, long double y)
+{
+	union ldshape ux, uy;
+
+	if (isnan(x) || isnan(y))
+		return x + y;
+	if (x == y)
+		return y;
+	ux.f = x;
+	if (x == 0) {
+		uy.f = y;
+		ux.i.m = 1;
+		ux.i.se = uy.i.se & 0x8000;
+	} else if ((x < y) == !(ux.i.se & 0x8000)) {
+		ux.i.m++;
+		if (ux.i.m << 1 == 0) {
+			ux.i.m = 1ULL << 63;
+			ux.i.se++;
+		}
+	} else {
+		if (ux.i.m << 1 == 0) {
+			ux.i.se--;
+			if (ux.i.se)
+				ux.i.m = 0;
+		}
+		ux.i.m--;
+	}
+	/* raise overflow if ux is infinite and x is finite */
+	if ((ux.i.se & 0x7fff) == 0x7fff)
+		return x + x;
+	/* raise underflow if ux is subnormal or zero */
+	if ((ux.i.se & 0x7fff) == 0)
+		FORCE_EVAL(x*x + ux.f*ux.f);
+	return ux.f;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+long double nextafterl(long double x, long double y)
+{
+	union ldshape ux, uy;
+
+	if (isnan(x) || isnan(y))
+		return x + y;
+	if (x == y)
+		return y;
+	ux.f = x;
+	if (x == 0) {
+		uy.f = y;
+		ux.i.lo = 1;
+		ux.i.se = uy.i.se & 0x8000;
+	} else if ((x < y) == !(ux.i.se & 0x8000)) {
+		ux.i2.lo++;
+		if (ux.i2.lo == 0)
+			ux.i2.hi++;
+	} else {
+		if (ux.i2.lo == 0)
+			ux.i2.hi--;
+		ux.i2.lo--;
+	}
+	/* raise overflow if ux is infinite and x is finite */
+	if ((ux.i.se & 0x7fff) == 0x7fff)
+		return x + x;
+	/* raise underflow if ux is subnormal or zero */
+	if ((ux.i.se & 0x7fff) == 0)
+		FORCE_EVAL(x*x + ux.f*ux.f);
+	return ux.f;
+}
+#endif
diff --git a/src/math/nexttoward.c b/src/math/nexttoward.c
new file mode 100644
index 0000000..827ee5c
--- /dev/null
+++ b/src/math/nexttoward.c
@@ -0,0 +1,42 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+double nexttoward(double x, long double y)
+{
+	return nextafter(x, y);
+}
+#else
+double nexttoward(double x, long double y)
+{
+	union {double f; uint64_t i;} ux = {x};
+	int e;
+
+	if (isnan(x) || isnan(y))
+		return x + y;
+	if (x == y)
+		return y;
+	if (x == 0) {
+		ux.i = 1;
+		if (signbit(y))
+			ux.i |= 1ULL<<63;
+	} else if (x < y) {
+		if (signbit(x))
+			ux.i--;
+		else
+			ux.i++;
+	} else {
+		if (signbit(x))
+			ux.i++;
+		else
+			ux.i--;
+	}
+	e = ux.i>>52 & 0x7ff;
+	/* raise overflow if ux.f is infinite and x is finite */
+	if (e == 0x7ff)
+		FORCE_EVAL(x+x);
+	/* raise underflow if ux.f is subnormal or zero */
+	if (e == 0)
+		FORCE_EVAL(x*x + ux.f*ux.f);
+	return ux.f;
+}
+#endif
diff --git a/src/math/nexttowardf.c b/src/math/nexttowardf.c
new file mode 100644
index 0000000..bbf172f
--- /dev/null
+++ b/src/math/nexttowardf.c
@@ -0,0 +1,35 @@
+#include "libm.h"
+
+float nexttowardf(float x, long double y)
+{
+	union {float f; uint32_t i;} ux = {x};
+	uint32_t e;
+
+	if (isnan(x) || isnan(y))
+		return x + y;
+	if (x == y)
+		return y;
+	if (x == 0) {
+		ux.i = 1;
+		if (signbit(y))
+			ux.i |= 0x80000000;
+	} else if (x < y) {
+		if (signbit(x))
+			ux.i--;
+		else
+			ux.i++;
+	} else {
+		if (signbit(x))
+			ux.i++;
+		else
+			ux.i--;
+	}
+	e = ux.i & 0x7f800000;
+	/* raise overflow if ux.f is infinite and x is finite */
+	if (e == 0x7f800000)
+		FORCE_EVAL(x+x);
+	/* raise underflow if ux.f is subnormal or zero */
+	if (e == 0)
+		FORCE_EVAL(x*x + ux.f*ux.f);
+	return ux.f;
+}
diff --git a/src/math/nexttowardl.c b/src/math/nexttowardl.c
new file mode 100644
index 0000000..67a6340
--- /dev/null
+++ b/src/math/nexttowardl.c
@@ -0,0 +1,6 @@
+#include <math.h>
+
+long double nexttowardl(long double x, long double y)
+{
+	return nextafterl(x, y);
+}
diff --git a/src/math/pow.c b/src/math/pow.c
index 42ab839..694c2ef 100644
--- a/src/math/pow.c
+++ b/src/math/pow.c
@@ -1,319 +1,343 @@
-/* pow(x,y) return x**y
+/*
+ * Double-precision x^y function.
  *
- *                    n
- * Method:  Let x =  2   * (1+f)
- *      1. Compute and return log2(x) in two pieces:
- *              log2(x) = w1 + w2,
- *         where w1 has 53-24 = 29 bit trailing zeros.
- *      2. Perform y*log2(x) = n+y' by simulating muti-precision
- *         arithmetic, where |y'|<=0.5.
- *      3. Return x**y = 2**n*exp(y'*log2)
- *
- * Special cases:
- *      1.  (anything) ** 0  is 1
- *      2.  1 ** (anything)  is 1
- *      3.  (anything except 1) ** NAN is NAN
- *      4.  NAN ** (anything except 0) is NAN
- *      5.  +-(|x| > 1) **  +INF is +INF
- *      6.  +-(|x| > 1) **  -INF is +0
- *      7.  +-(|x| < 1) **  +INF is +0
- *      8.  +-(|x| < 1) **  -INF is +INF
- *      9.  -1          ** +-INF is 1
- *      10. +0 ** (+anything except 0, NAN)               is +0
- *      11. -0 ** (+anything except 0, NAN, odd integer)  is +0
- *      12. +0 ** (-anything except 0, NAN)               is +INF, raise divbyzero
- *      13. -0 ** (-anything except 0, NAN, odd integer)  is +INF, raise divbyzero
- *      14. -0 ** (+odd integer) is -0
- *      15. -0 ** (-odd integer) is -INF, raise divbyzero
- *      16. +INF ** (+anything except 0,NAN) is +INF
- *      17. +INF ** (-anything except 0,NAN) is +0
- *      18. -INF ** (+odd integer) is -INF
- *      19. -INF ** (anything) = -0 ** (-anything), (anything except odd integer)
- *      20. (anything) ** 1 is (anything)
- *      21. (anything) ** -1 is 1/(anything)
- *      22. (-anything) ** (integer) is (-1)**(integer)*(+anything**integer)
- *      23. (-anything except 0 and inf) ** (non-integer) is NAN
- *
- * Accuracy:
- *      pow(x,y) returns x**y nearly rounded. In particular
- *                      pow(integer,integer)
- *      always returns the correct integer provided it is
- *      representable.
- *
- * Constants :
- * The hexadecimal values are the intended ones for the following
- * constants. The decimal values may be used, provided that the
- * compiler will convert from decimal to binary accurately enough
- * to produce the hexadecimal values shown.
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
  */
 
 #include <math.h>
+#include <stdint.h>
+#include "libm.h"
+#include "exp_data.h"
+#include "pow_data.h"
 
+/*
+Worst-case error: 0.54 ULP (~= ulperr_exp + 1024*Ln2*relerr_log*2^53)
+relerr_log: 1.3 * 2^-68 (Relative error of log, 1.5 * 2^-68 without fma)
+ulperr_exp: 0.509 ULP (ULP error of exp, 0.511 ULP without fma)
+*/
 
-static const double
-bp[]   = {1.0, 1.5,},
-dp_h[] = { 0.0, 5.84962487220764160156e-01,}, /* 0x3FE2B803, 0x40000000 */
-dp_l[] = { 0.0, 1.35003920212974897128e-08,}, /* 0x3E4CFDEB, 0x43CFD006 */
-two53  =  9007199254740992.0, /* 0x43400000, 0x00000000 */
-huge   =  1.0e300,
-tiny   =  1.0e-300,
-/* poly coefs for (3/2)*(log(x)-2s-2/3*s**3 */
-L1 =  5.99999999999994648725e-01, /* 0x3FE33333, 0x33333303 */
-L2 =  4.28571428578550184252e-01, /* 0x3FDB6DB6, 0xDB6FABFF */
-L3 =  3.33333329818377432918e-01, /* 0x3FD55555, 0x518F264D */
-L4 =  2.72728123808534006489e-01, /* 0x3FD17460, 0xA91D4101 */
-L5 =  2.30660745775561754067e-01, /* 0x3FCD864A, 0x93C9DB65 */
-L6 =  2.06975017800338417784e-01, /* 0x3FCA7E28, 0x4A454EEF */
-P1 =  1.66666666666666019037e-01, /* 0x3FC55555, 0x5555553E */
-P2 = -2.77777777770155933842e-03, /* 0xBF66C16C, 0x16BEBD93 */
-P3 =  6.61375632143793436117e-05, /* 0x3F11566A, 0xAF25DE2C */
-P4 = -1.65339022054652515390e-06, /* 0xBEBBBD41, 0xC5D26BF1 */
-P5 =  4.13813679705723846039e-08, /* 0x3E663769, 0x72BEA4D0 */
-lg2     =  6.93147180559945286227e-01, /* 0x3FE62E42, 0xFEFA39EF */
-lg2_h   =  6.93147182464599609375e-01, /* 0x3FE62E43, 0x00000000 */
-lg2_l   = -1.90465429995776804525e-09, /* 0xBE205C61, 0x0CA86C39 */
-ovt     =  8.0085662595372944372e-017, /* -(1024-log2(ovfl+.5ulp)) */
-cp      =  9.61796693925975554329e-01, /* 0x3FEEC709, 0xDC3A03FD =2/(3ln2) */
-cp_h    =  9.61796700954437255859e-01, /* 0x3FEEC709, 0xE0000000 =(float)cp */
-cp_l    = -7.02846165095275826516e-09, /* 0xBE3E2FE0, 0x145B01F5 =tail of cp_h*/
-ivln2   =  1.44269504088896338700e+00, /* 0x3FF71547, 0x652B82FE =1/ln2 */
-ivln2_h =  1.44269502162933349609e+00, /* 0x3FF71547, 0x60000000 =24b 1/ln2*/
-ivln2_l =  1.92596299112661746887e-08; /* 0x3E54AE0B, 0xF85DDF44 =1/ln2 tail*/
+#define T __pow_log_data.tab
+#define A __pow_log_data.poly
+#define Ln2hi __pow_log_data.ln2hi
+#define Ln2lo __pow_log_data.ln2lo
+#define N (1 << POW_LOG_TABLE_BITS)
+#define OFF 0x3fe6955500000000
 
-double pow(double x, double y)
+/* Top 12 bits of a double (sign and exponent bits).  */
+static inline uint32_t top12(double x)
 {
-	double z,ax,z_h,z_l,p_h,p_l;
-	double y1,t1,t2,r,s,t,u,v,w;
-	int32_t i,j,k,yisint,n;
-	int32_t hx,hy,ix,iy;
-	uint32_t lx,ly;
+	return asuint64(x) >> 52;
+}
 
-	EXTRACT_WORDS(hx, lx, x);
-	EXTRACT_WORDS(hy, ly, y);
-	ix = hx & 0x7fffffff;
-	iy = hy & 0x7fffffff;
+/* Compute y+TAIL = log(x) where the rounded result is y and TAIL has about
+   additional 15 bits precision.  IX is the bit representation of x, but
+   normalized in the subnormal range using the sign bit for the exponent.  */
+static inline double_t log_inline(uint64_t ix, double_t *tail)
+{
+	/* double_t for better performance on targets with FLT_EVAL_METHOD==2.  */
+	double_t z, r, y, invc, logc, logctail, kd, hi, t1, t2, lo, lo1, lo2, p;
+	uint64_t iz, tmp;
+	int k, i;
 
-	/* x**0 = 1, even if x is NaN */
-	if ((iy|ly) == 0)
-		return 1.0;
-	/* 1**y = 1, even if y is NaN */
-	if (hx == 0x3ff00000 && lx == 0)
-		return 1.0;
-	/* NaN if either arg is NaN */
-	if (ix > 0x7ff00000 || (ix == 0x7ff00000 && lx != 0) ||
-	    iy > 0x7ff00000 || (iy == 0x7ff00000 && ly != 0))
-		return x + y;
+	/* x = 2^k z; where z is in range [OFF,2*OFF) and exact.
+	   The range is split into N subintervals.
+	   The ith subinterval contains z and c is near its center.  */
+	tmp = ix - OFF;
+	i = (tmp >> (52 - POW_LOG_TABLE_BITS)) % N;
+	k = (int64_t)tmp >> 52; /* arithmetic shift */
+	iz = ix - (tmp & 0xfffULL << 52);
+	z = asdouble(iz);
+	kd = (double_t)k;
 
-	/* determine if y is an odd int when x < 0
-	 * yisint = 0       ... y is not an integer
-	 * yisint = 1       ... y is an odd int
-	 * yisint = 2       ... y is an even int
-	 */
-	yisint = 0;
-	if (hx < 0) {
-		if (iy >= 0x43400000)
-			yisint = 2; /* even integer y */
-		else if (iy >= 0x3ff00000) {
-			k = (iy>>20) - 0x3ff;  /* exponent */
-			if (k > 20) {
-				uint32_t j = ly>>(52-k);
-				if ((j<<(52-k)) == ly)
-					yisint = 2 - (j&1);
-			} else if (ly == 0) {
-				uint32_t j = iy>>(20-k);
-				if ((j<<(20-k)) == iy)
-					yisint = 2 - (j&1);
-			}
-		}
-	}
+	/* log(x) = k*Ln2 + log(c) + log1p(z/c-1).  */
+	invc = T[i].invc;
+	logc = T[i].logc;
+	logctail = T[i].logctail;
 
-	/* special value of y */
-	if (ly == 0) {
-		if (iy == 0x7ff00000) {  /* y is +-inf */
-			if (((ix-0x3ff00000)|lx) == 0)  /* (-1)**+-inf is 1 */
-				return 1.0;
-			else if (ix >= 0x3ff00000) /* (|x|>1)**+-inf = inf,0 */
-				return hy >= 0 ? y : 0.0;
-			else                       /* (|x|<1)**+-inf = 0,inf */
-				return hy >= 0 ? 0.0 : -y;
-		}
-		if (iy == 0x3ff00000) {    /* y is +-1 */
-			if (hy >= 0)
-				return x;
-			y = 1/x;
-#if FLT_EVAL_METHOD!=0
-			{
-				union {double f; uint64_t i;} u = {y};
-				uint64_t i = u.i & -1ULL/2;
-				if (i>>52 == 0 && (i&(i-1)))
-					FORCE_EVAL((float)y);
-			}
+	/* Note: 1/c is j/N or j/N/2 where j is an integer in [N,2N) and
+     |z/c - 1| < 1/N, so r = z/c - 1 is exactly representible.  */
+#if __FP_FAST_FMA
+	r = __builtin_fma(z, invc, -1.0);
+#else
+	/* Split z such that rhi, rlo and rhi*rhi are exact and |rlo| <= |r|.  */
+	double_t zhi = asdouble((iz + (1ULL << 31)) & (-1ULL << 32));
+	double_t zlo = z - zhi;
+	double_t rhi = zhi * invc - 1.0;
+	double_t rlo = zlo * invc;
+	r = rhi + rlo;
 #endif
-			return y;
-		}
-		if (hy == 0x40000000)    /* y is 2 */
-			return x*x;
-		if (hy == 0x3fe00000) {  /* y is 0.5 */
-			if (hx >= 0)     /* x >= +0 */
-				return sqrt(x);
-		}
+
+	/* k*Ln2 + log(c) + r.  */
+	t1 = kd * Ln2hi + logc;
+	t2 = t1 + r;
+	lo1 = kd * Ln2lo + logctail;
+	lo2 = t1 - t2 + r;
+
+	/* Evaluation is optimized assuming superscalar pipelined execution.  */
+	double_t ar, ar2, ar3, lo3, lo4;
+	ar = A[0] * r; /* A[0] = -0.5.  */
+	ar2 = r * ar;
+	ar3 = r * ar2;
+	/* k*Ln2 + log(c) + r + A[0]*r*r.  */
+#if __FP_FAST_FMA
+	hi = t2 + ar2;
+	lo3 = __builtin_fma(ar, r, -ar2);
+	lo4 = t2 - hi + ar2;
+#else
+	double_t arhi = A[0] * rhi;
+	double_t arhi2 = rhi * arhi;
+	hi = t2 + arhi2;
+	lo3 = rlo * (ar + arhi);
+	lo4 = t2 - hi + arhi2;
+#endif
+	/* p = log1p(r) - r - A[0]*r*r.  */
+	p = (ar3 * (A[1] + r * A[2] +
+		    ar2 * (A[3] + r * A[4] + ar2 * (A[5] + r * A[6]))));
+	lo = lo1 + lo2 + lo3 + lo4 + p;
+	y = hi + lo;
+	*tail = hi - y + lo;
+	return y;
+}
+
+#undef N
+#undef T
+#define N (1 << EXP_TABLE_BITS)
+#define InvLn2N __exp_data.invln2N
+#define NegLn2hiN __exp_data.negln2hiN
+#define NegLn2loN __exp_data.negln2loN
+#define Shift __exp_data.shift
+#define T __exp_data.tab
+#define C2 __exp_data.poly[5 - EXP_POLY_ORDER]
+#define C3 __exp_data.poly[6 - EXP_POLY_ORDER]
+#define C4 __exp_data.poly[7 - EXP_POLY_ORDER]
+#define C5 __exp_data.poly[8 - EXP_POLY_ORDER]
+#define C6 __exp_data.poly[9 - EXP_POLY_ORDER]
+
+/* Handle cases that may overflow or underflow when computing the result that
+   is scale*(1+TMP) without intermediate rounding.  The bit representation of
+   scale is in SBITS, however it has a computed exponent that may have
+   overflown into the sign bit so that needs to be adjusted before using it as
+   a double.  (int32_t)KI is the k used in the argument reduction and exponent
+   adjustment of scale, positive k here means the result may overflow and
+   negative k means the result may underflow.  */
+static inline double specialcase(double_t tmp, uint64_t sbits, uint64_t ki)
+{
+	double_t scale, y;
+
+	if ((ki & 0x80000000) == 0) {
+		/* k > 0, the exponent of scale might have overflowed by <= 460.  */
+		sbits -= 1009ull << 52;
+		scale = asdouble(sbits);
+		y = 0x1p1009 * (scale + scale * tmp);
+		return eval_as_double(y);
 	}
+	/* k < 0, need special care in the subnormal range.  */
+	sbits += 1022ull << 52;
+	/* Note: sbits is signed scale.  */
+	scale = asdouble(sbits);
+	y = scale + scale * tmp;
+	if (fabs(y) < 1.0) {
+		/* Round y to the right precision before scaling it into the subnormal
+		   range to avoid double rounding that can cause 0.5+E/2 ulp error where
+		   E is the worst-case ulp error outside the subnormal range.  So this
+		   is only useful if the goal is better than 1 ulp worst-case error.  */
+		double_t hi, lo, one = 1.0;
+		if (y < 0.0)
+			one = -1.0;
+		lo = scale - y + scale * tmp;
+		hi = one + y;
+		lo = one - hi + y + lo;
+		y = eval_as_double(hi + lo) - one;
+		/* Fix the sign of 0.  */
+		if (y == 0.0)
+			y = asdouble(sbits & 0x8000000000000000);
+		/* The underflow exception needs to be signaled explicitly.  */
+		fp_force_eval(fp_barrier(0x1p-1022) * 0x1p-1022);
+	}
+	y = 0x1p-1022 * y;
+	return eval_as_double(y);
+}
 
-	ax = fabs(x);
-	/* special value of x */
-	if (lx == 0) {
-		if (ix == 0x7ff00000 || ix == 0 || ix == 0x3ff00000) { /* x is +-0,+-inf,+-1 */
-			z = ax;
-			if (hy < 0)   /* z = (1/|x|) */
-				z = 1.0/z;
-			if (hx < 0) {
-				if (((ix-0x3ff00000)|yisint) == 0) {
-					z = (z-z)/(z-z); /* (-1)**non-int is NaN */
-				} else if (yisint == 1)
-					z = -z;          /* (x<0)**odd = -(|x|**odd) */
-			}
-			return z;
+#define SIGN_BIAS (0x800 << EXP_TABLE_BITS)
+
+/* Computes sign*exp(x+xtail) where |xtail| < 2^-8/N and |xtail| <= |x|.
+   The sign_bias argument is SIGN_BIAS or 0 and sets the sign to -1 or 1.  */
+static inline double exp_inline(double_t x, double_t xtail, uint32_t sign_bias)
+{
+	uint32_t abstop;
+	uint64_t ki, idx, top, sbits;
+	/* double_t for better performance on targets with FLT_EVAL_METHOD==2.  */
+	double_t kd, z, r, r2, scale, tail, tmp;
+
+	abstop = top12(x) & 0x7ff;
+	if (predict_false(abstop - top12(0x1p-54) >=
+			  top12(512.0) - top12(0x1p-54))) {
+		if (abstop - top12(0x1p-54) >= 0x80000000) {
+			/* Avoid spurious underflow for tiny x.  */
+			/* Note: 0 is common input.  */
+			double_t one = WANT_ROUNDING ? 1.0 + x : 1.0;
+			return sign_bias ? -one : one;
+		}
+		if (abstop >= top12(1024.0)) {
+			/* Note: inf and nan are already handled.  */
+			if (asuint64(x) >> 63)
+				return __math_uflow(sign_bias);
+			else
+				return __math_oflow(sign_bias);
 		}
+		/* Large x is special cased below.  */
+		abstop = 0;
 	}
 
-	s = 1.0; /* sign of result */
-	if (hx < 0) {
-		if (yisint == 0) /* (x<0)**(non-int) is NaN */
-			return (x-x)/(x-x);
-		if (yisint == 1) /* (x<0)**(odd int) */
-			s = -1.0;
-	}
+	/* exp(x) = 2^(k/N) * exp(r), with exp(r) in [2^(-1/2N),2^(1/2N)].  */
+	/* x = ln2/N*k + r, with int k and r in [-ln2/2N, ln2/2N].  */
+	z = InvLn2N * x;
+#if TOINT_INTRINSICS
+	kd = roundtoint(z);
+	ki = converttoint(z);
+#elif EXP_USE_TOINT_NARROW
+	/* z - kd is in [-0.5-2^-16, 0.5] in all rounding modes.  */
+	kd = eval_as_double(z + Shift);
+	ki = asuint64(kd) >> 16;
+	kd = (double_t)(int32_t)ki;
+#else
+	/* z - kd is in [-1, 1] in non-nearest rounding modes.  */
+	kd = eval_as_double(z + Shift);
+	ki = asuint64(kd);
+	kd -= Shift;
+#endif
+	r = x + kd * NegLn2hiN + kd * NegLn2loN;
+	/* The code assumes 2^-200 < |xtail| < 2^-8/N.  */
+	r += xtail;
+	/* 2^(k/N) ~= scale * (1 + tail).  */
+	idx = 2 * (ki % N);
+	top = (ki + sign_bias) << (52 - EXP_TABLE_BITS);
+	tail = asdouble(T[idx]);
+	/* This is only a valid scale when -1023*N < k < 1024*N.  */
+	sbits = T[idx + 1] + top;
+	/* exp(x) = 2^(k/N) * exp(r) ~= scale + scale * (tail + exp(r) - 1).  */
+	/* Evaluation is optimized assuming superscalar pipelined execution.  */
+	r2 = r * r;
+	/* Without fma the worst case error is 0.25/N ulp larger.  */
+	/* Worst case error is less than 0.5+1.11/N+(abs poly error * 2^53) ulp.  */
+	tmp = tail + r + r2 * (C2 + r * C3) + r2 * r2 * (C4 + r * C5);
+	if (predict_false(abstop == 0))
+		return specialcase(tmp, sbits, ki);
+	scale = asdouble(sbits);
+	/* Note: tmp == 0 or |tmp| > 2^-200 and scale > 2^-739, so there
+	   is no spurious underflow here even without fma.  */
+	return eval_as_double(scale + scale * tmp);
+}
+
+/* Returns 0 if not int, 1 if odd int, 2 if even int.  The argument is
+   the bit representation of a non-zero finite floating-point value.  */
+static inline int checkint(uint64_t iy)
+{
+	int e = iy >> 52 & 0x7ff;
+	if (e < 0x3ff)
+		return 0;
+	if (e > 0x3ff + 52)
+		return 2;
+	if (iy & ((1ULL << (0x3ff + 52 - e)) - 1))
+		return 0;
+	if (iy & (1ULL << (0x3ff + 52 - e)))
+		return 1;
+	return 2;
+}
+
+/* Returns 1 if input is the bit representation of 0, infinity or nan.  */
+static inline int zeroinfnan(uint64_t i)
+{
+	return 2 * i - 1 >= 2 * asuint64(INFINITY) - 1;
+}
 
-	/* |y| is huge */
-	if (iy > 0x41e00000) { /* if |y| > 2**31 */
-		if (iy > 0x43f00000) {  /* if |y| > 2**64, must o/uflow */
-			if (ix <= 0x3fefffff)
-				return hy < 0 ? huge*huge : tiny*tiny;
-			if (ix >= 0x3ff00000)
-				return hy > 0 ? huge*huge : tiny*tiny;
+double pow(double x, double y)
+{
+	uint32_t sign_bias = 0;
+	uint64_t ix, iy;
+	uint32_t topx, topy;
+
+	ix = asuint64(x);
+	iy = asuint64(y);
+	topx = top12(x);
+	topy = top12(y);
+	if (predict_false(topx - 0x001 >= 0x7ff - 0x001 ||
+			  (topy & 0x7ff) - 0x3be >= 0x43e - 0x3be)) {
+		/* Note: if |y| > 1075 * ln2 * 2^53 ~= 0x1.749p62 then pow(x,y) = inf/0
+		   and if |y| < 2^-54 / 1075 ~= 0x1.e7b6p-65 then pow(x,y) = +-1.  */
+		/* Special cases: (x < 0x1p-126 or inf or nan) or
+		   (|y| < 0x1p-65 or |y| >= 0x1p63 or nan).  */
+		if (predict_false(zeroinfnan(iy))) {
+			if (2 * iy == 0)
+				return issignaling_inline(x) ? x + y : 1.0;
+			if (ix == asuint64(1.0))
+				return issignaling_inline(y) ? x + y : 1.0;
+			if (2 * ix > 2 * asuint64(INFINITY) ||
+			    2 * iy > 2 * asuint64(INFINITY))
+				return x + y;
+			if (2 * ix == 2 * asuint64(1.0))
+				return 1.0;
+			if ((2 * ix < 2 * asuint64(1.0)) == !(iy >> 63))
+				return 0.0; /* |x|<1 && y==inf or |x|>1 && y==-inf.  */
+			return y * y;
 		}
-		/* over/underflow if x is not close to one */
-		if (ix < 0x3fefffff)
-			return hy < 0 ? s*huge*huge : s*tiny*tiny;
-		if (ix > 0x3ff00000)
-			return hy > 0 ? s*huge*huge : s*tiny*tiny;
-		/* now |1-x| is tiny <= 2**-20, suffice to compute
-		   log(x) by x-x^2/2+x^3/3-x^4/4 */
-		t = ax - 1.0;       /* t has 20 trailing zeros */
-		w = (t*t)*(0.5 - t*(0.3333333333333333333333-t*0.25));
-		u = ivln2_h*t;      /* ivln2_h has 21 sig. bits */
-		v = t*ivln2_l - w*ivln2;
-		t1 = u + v;
-		SET_LOW_WORD(t1, 0);
-		t2 = v - (t1-u);
-	} else {
-		double ss,s2,s_h,s_l,t_h,t_l;
-		n = 0;
-		/* take care subnormal number */
-		if (ix < 0x00100000) {
-			ax *= two53;
-			n -= 53;
-			GET_HIGH_WORD(ix,ax);
+		if (predict_false(zeroinfnan(ix))) {
+			double_t x2 = x * x;
+			if (ix >> 63 && checkint(iy) == 1)
+				x2 = -x2;
+			/* Without the barrier some versions of clang hoist the 1/x2 and
+			   thus division by zero exception can be signaled spuriously.  */
+			return iy >> 63 ? fp_barrier(1 / x2) : x2;
 		}
-		n += ((ix)>>20) - 0x3ff;
-		j = ix & 0x000fffff;
-		/* determine interval */
-		ix = j | 0x3ff00000;   /* normalize ix */
-		if (j <= 0x3988E)      /* |x|<sqrt(3/2) */
-			k = 0;
-		else if (j < 0xBB67A)  /* |x|<sqrt(3)   */
-			k = 1;
-		else {
-			k = 0;
-			n += 1;
-			ix -= 0x00100000;
+		/* Here x and y are non-zero finite.  */
+		if (ix >> 63) {
+			/* Finite x < 0.  */
+			int yint = checkint(iy);
+			if (yint == 0)
+				return __math_invalid(x);
+			if (yint == 1)
+				sign_bias = SIGN_BIAS;
+			ix &= 0x7fffffffffffffff;
+			topx &= 0x7ff;
+		}
+		if ((topy & 0x7ff) - 0x3be >= 0x43e - 0x3be) {
+			/* Note: sign_bias == 0 here because y is not odd.  */
+			if (ix == asuint64(1.0))
+				return 1.0;
+			if ((topy & 0x7ff) < 0x3be) {
+				/* |y| < 2^-65, x^y ~= 1 + y*log(x).  */
+				if (WANT_ROUNDING)
+					return ix > asuint64(1.0) ? 1.0 + y :
+								    1.0 - y;
+				else
+					return 1.0;
+			}
+			return (ix > asuint64(1.0)) == (topy < 0x800) ?
+				       __math_oflow(0) :
+				       __math_uflow(0);
+		}
+		if (topx == 0) {
+			/* Normalize subnormal x so exponent becomes negative.  */
+			ix = asuint64(x * 0x1p52);
+			ix &= 0x7fffffffffffffff;
+			ix -= 52ULL << 52;
 		}
-		SET_HIGH_WORD(ax, ix);
-
-		/* compute ss = s_h+s_l = (x-1)/(x+1) or (x-1.5)/(x+1.5) */
-		u = ax - bp[k];        /* bp[0]=1.0, bp[1]=1.5 */
-		v = 1.0/(ax+bp[k]);
-		ss = u*v;
-		s_h = ss;
-		SET_LOW_WORD(s_h, 0);
-		/* t_h=ax+bp[k] High */
-		t_h = 0.0;
-		SET_HIGH_WORD(t_h, ((ix>>1)|0x20000000) + 0x00080000 + (k<<18));
-		t_l = ax - (t_h-bp[k]);
-		s_l = v*((u-s_h*t_h)-s_h*t_l);
-		/* compute log(ax) */
-		s2 = ss*ss;
-		r = s2*s2*(L1+s2*(L2+s2*(L3+s2*(L4+s2*(L5+s2*L6)))));
-		r += s_l*(s_h+ss);
-		s2 = s_h*s_h;
-		t_h = 3.0 + s2 + r;
-		SET_LOW_WORD(t_h, 0);
-		t_l = r - ((t_h-3.0)-s2);
-		/* u+v = ss*(1+...) */
-		u = s_h*t_h;
-		v = s_l*t_h + t_l*ss;
-		/* 2/(3log2)*(ss+...) */
-		p_h = u + v;
-		SET_LOW_WORD(p_h, 0);
-		p_l = v - (p_h-u);
-		z_h = cp_h*p_h;        /* cp_h+cp_l = 2/(3*log2) */
-		z_l = cp_l*p_h+p_l*cp + dp_l[k];
-		/* log2(ax) = (ss+..)*2/(3*log2) = n + dp_h + z_h + z_l */
-		t = (double)n;
-		t1 = ((z_h + z_l) + dp_h[k]) + t;
-		SET_LOW_WORD(t1, 0);
-		t2 = z_l - (((t1 - t) - dp_h[k]) - z_h);
 	}
 
-	/* split up y into y1+y2 and compute (y1+y2)*(t1+t2) */
-	y1 = y;
-	SET_LOW_WORD(y1, 0);
-	p_l = (y-y1)*t1 + y*t2;
-	p_h = y1*t1;
-	z = p_l + p_h;
-	EXTRACT_WORDS(j, i, z);
-	if (j >= 0x40900000) {                      /* z >= 1024 */
-		if (((j-0x40900000)|i) != 0)        /* if z > 1024 */
-			return s*huge*huge;         /* overflow */
-		if (p_l + ovt > z - p_h)
-			return s*huge*huge;         /* overflow */
-	} else if ((j&0x7fffffff) >= 0x4090cc00) {  /* z <= -1075 */
-		if (((j-0xc090cc00)|i) != 0)        /* z < -1075 */
-			return s*tiny*tiny;         /* underflow */
-		if (p_l <= z - p_h)
-			return s*tiny*tiny;         /* underflow */
-	}
-	/*
-	 * compute 2**(p_h+p_l)
-	 */
-	i = j & 0x7fffffff;
-	k = (i>>20) - 0x3ff;
-	n = 0;
-	if (i > 0x3fe00000) {  /* if |z| > 0.5, set n = [z+0.5] */
-		n = j + (0x00100000>>(k+1));
-		k = ((n&0x7fffffff)>>20) - 0x3ff;  /* new k for n */
-		t = 0.0;
-		SET_HIGH_WORD(t, n & ~(0x000fffff>>k));
-		n = ((n&0x000fffff)|0x00100000)>>(20-k);
-		if (j < 0)
-			n = -n;
-		p_h -= t;
-	}
-	t = p_l + p_h;
-	SET_LOW_WORD(t, 0);
-	u = t*lg2_h;
-	v = (p_l-(t-p_h))*lg2 + t*lg2_l;
-	z = u + v;
-	w = v - (z-u);
-	t = z*z;
-	t1 = z - t*(P1+t*(P2+t*(P3+t*(P4+t*P5))));
-	r = (z*t1)/(t1-2.0) - (w + z*w);
-	z = 1.0 - (r-z);
-	GET_HIGH_WORD(j, z);
-	j += n<<20;
-	if ((j>>20) <= 0)  /* subnormal output */
-		z = scalbn(z,n);
-	else
-		SET_HIGH_WORD(z, j);
-	return s*z;
+	double_t lo;
+	double_t hi = log_inline(ix, &lo);
+	double_t ehi, elo;
+#if __FP_FAST_FMA
+	ehi = y * hi;
+	elo = y * lo + __builtin_fma(y, hi, -ehi);
+#else
+	double_t yhi = asdouble(iy & -1ULL << 27);
+	double_t ylo = y - yhi;
+	double_t lhi = asdouble(asuint64(hi) & -1ULL << 27);
+	double_t llo = hi - lhi + lo;
+	ehi = yhi * lhi;
+	elo = ylo * lhi + y * llo; /* |elo| < |ehi| * 2^-25.  */
+#endif
+	return exp_inline(ehi, elo, sign_bias);
 }
diff --git a/src/math/pow_data.c b/src/math/pow_data.c
new file mode 100644
index 0000000..81e760d
--- /dev/null
+++ b/src/math/pow_data.c
@@ -0,0 +1,180 @@
+/*
+ * Data for the log part of pow.
+ *
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+
+#include "pow_data.h"
+
+#define N (1 << POW_LOG_TABLE_BITS)
+
+const struct pow_log_data __pow_log_data = {
+.ln2hi = 0x1.62e42fefa3800p-1,
+.ln2lo = 0x1.ef35793c76730p-45,
+.poly = {
+// relative error: 0x1.11922ap-70
+// in -0x1.6bp-8 0x1.6bp-8
+// Coefficients are scaled to match the scaling during evaluation.
+-0x1p-1,
+0x1.555555555556p-2 * -2,
+-0x1.0000000000006p-2 * -2,
+0x1.999999959554ep-3 * 4,
+-0x1.555555529a47ap-3 * 4,
+0x1.2495b9b4845e9p-3 * -8,
+-0x1.0002b8b263fc3p-3 * -8,
+},
+/* Algorithm:
+
+	x = 2^k z
+	log(x) = k ln2 + log(c) + log(z/c)
+	log(z/c) = poly(z/c - 1)
+
+where z is in [0x1.69555p-1; 0x1.69555p0] which is split into N subintervals
+and z falls into the ith one, then table entries are computed as
+
+	tab[i].invc = 1/c
+	tab[i].logc = round(0x1p43*log(c))/0x1p43
+	tab[i].logctail = (double)(log(c) - logc)
+
+where c is chosen near the center of the subinterval such that 1/c has only a
+few precision bits so z/c - 1 is exactly representible as double:
+
+	1/c = center < 1 ? round(N/center)/N : round(2*N/center)/N/2
+
+Note: |z/c - 1| < 1/N for the chosen c, |log(c) - logc - logctail| < 0x1p-97,
+the last few bits of logc are rounded away so k*ln2hi + logc has no rounding
+error and the interval for z is selected such that near x == 1, where log(x)
+is tiny, large cancellation error is avoided in logc + poly(z/c - 1).  */
+.tab = {
+#define A(a, b, c) {a, 0, b, c},
+A(0x1.6a00000000000p+0, -0x1.62c82f2b9c800p-2, 0x1.ab42428375680p-48)
+A(0x1.6800000000000p+0, -0x1.5d1bdbf580800p-2, -0x1.ca508d8e0f720p-46)
+A(0x1.6600000000000p+0, -0x1.5767717455800p-2, -0x1.362a4d5b6506dp-45)
+A(0x1.6400000000000p+0, -0x1.51aad872df800p-2, -0x1.684e49eb067d5p-49)
+A(0x1.6200000000000p+0, -0x1.4be5f95777800p-2, -0x1.41b6993293ee0p-47)
+A(0x1.6000000000000p+0, -0x1.4618bc21c6000p-2, 0x1.3d82f484c84ccp-46)
+A(0x1.5e00000000000p+0, -0x1.404308686a800p-2, 0x1.c42f3ed820b3ap-50)
+A(0x1.5c00000000000p+0, -0x1.3a64c55694800p-2, 0x1.0b1c686519460p-45)
+A(0x1.5a00000000000p+0, -0x1.347dd9a988000p-2, 0x1.5594dd4c58092p-45)
+A(0x1.5800000000000p+0, -0x1.2e8e2bae12000p-2, 0x1.67b1e99b72bd8p-45)
+A(0x1.5600000000000p+0, -0x1.2895a13de8800p-2, 0x1.5ca14b6cfb03fp-46)
+A(0x1.5600000000000p+0, -0x1.2895a13de8800p-2, 0x1.5ca14b6cfb03fp-46)
+A(0x1.5400000000000p+0, -0x1.22941fbcf7800p-2, -0x1.65a242853da76p-46)
+A(0x1.5200000000000p+0, -0x1.1c898c1699800p-2, -0x1.fafbc68e75404p-46)
+A(0x1.5000000000000p+0, -0x1.1675cababa800p-2, 0x1.f1fc63382a8f0p-46)
+A(0x1.4e00000000000p+0, -0x1.1058bf9ae4800p-2, -0x1.6a8c4fd055a66p-45)
+A(0x1.4c00000000000p+0, -0x1.0a324e2739000p-2, -0x1.c6bee7ef4030ep-47)
+A(0x1.4a00000000000p+0, -0x1.0402594b4d000p-2, -0x1.036b89ef42d7fp-48)
+A(0x1.4a00000000000p+0, -0x1.0402594b4d000p-2, -0x1.036b89ef42d7fp-48)
+A(0x1.4800000000000p+0, -0x1.fb9186d5e4000p-3, 0x1.d572aab993c87p-47)
+A(0x1.4600000000000p+0, -0x1.ef0adcbdc6000p-3, 0x1.b26b79c86af24p-45)
+A(0x1.4400000000000p+0, -0x1.e27076e2af000p-3, -0x1.72f4f543fff10p-46)
+A(0x1.4200000000000p+0, -0x1.d5c216b4fc000p-3, 0x1.1ba91bbca681bp-45)
+A(0x1.4000000000000p+0, -0x1.c8ff7c79aa000p-3, 0x1.7794f689f8434p-45)
+A(0x1.4000000000000p+0, -0x1.c8ff7c79aa000p-3, 0x1.7794f689f8434p-45)
+A(0x1.3e00000000000p+0, -0x1.bc286742d9000p-3, 0x1.94eb0318bb78fp-46)
+A(0x1.3c00000000000p+0, -0x1.af3c94e80c000p-3, 0x1.a4e633fcd9066p-52)
+A(0x1.3a00000000000p+0, -0x1.a23bc1fe2b000p-3, -0x1.58c64dc46c1eap-45)
+A(0x1.3a00000000000p+0, -0x1.a23bc1fe2b000p-3, -0x1.58c64dc46c1eap-45)
+A(0x1.3800000000000p+0, -0x1.9525a9cf45000p-3, -0x1.ad1d904c1d4e3p-45)
+A(0x1.3600000000000p+0, -0x1.87fa06520d000p-3, 0x1.bbdbf7fdbfa09p-45)
+A(0x1.3400000000000p+0, -0x1.7ab890210e000p-3, 0x1.bdb9072534a58p-45)
+A(0x1.3400000000000p+0, -0x1.7ab890210e000p-3, 0x1.bdb9072534a58p-45)
+A(0x1.3200000000000p+0, -0x1.6d60fe719d000p-3, -0x1.0e46aa3b2e266p-46)
+A(0x1.3000000000000p+0, -0x1.5ff3070a79000p-3, -0x1.e9e439f105039p-46)
+A(0x1.3000000000000p+0, -0x1.5ff3070a79000p-3, -0x1.e9e439f105039p-46)
+A(0x1.2e00000000000p+0, -0x1.526e5e3a1b000p-3, -0x1.0de8b90075b8fp-45)
+A(0x1.2c00000000000p+0, -0x1.44d2b6ccb8000p-3, 0x1.70cc16135783cp-46)
+A(0x1.2c00000000000p+0, -0x1.44d2b6ccb8000p-3, 0x1.70cc16135783cp-46)
+A(0x1.2a00000000000p+0, -0x1.371fc201e9000p-3, 0x1.178864d27543ap-48)
+A(0x1.2800000000000p+0, -0x1.29552f81ff000p-3, -0x1.48d301771c408p-45)
+A(0x1.2600000000000p+0, -0x1.1b72ad52f6000p-3, -0x1.e80a41811a396p-45)
+A(0x1.2600000000000p+0, -0x1.1b72ad52f6000p-3, -0x1.e80a41811a396p-45)
+A(0x1.2400000000000p+0, -0x1.0d77e7cd09000p-3, 0x1.a699688e85bf4p-47)
+A(0x1.2400000000000p+0, -0x1.0d77e7cd09000p-3, 0x1.a699688e85bf4p-47)
+A(0x1.2200000000000p+0, -0x1.fec9131dbe000p-4, -0x1.575545ca333f2p-45)
+A(0x1.2000000000000p+0, -0x1.e27076e2b0000p-4, 0x1.a342c2af0003cp-45)
+A(0x1.2000000000000p+0, -0x1.e27076e2b0000p-4, 0x1.a342c2af0003cp-45)
+A(0x1.1e00000000000p+0, -0x1.c5e548f5bc000p-4, -0x1.d0c57585fbe06p-46)
+A(0x1.1c00000000000p+0, -0x1.a926d3a4ae000p-4, 0x1.53935e85baac8p-45)
+A(0x1.1c00000000000p+0, -0x1.a926d3a4ae000p-4, 0x1.53935e85baac8p-45)
+A(0x1.1a00000000000p+0, -0x1.8c345d631a000p-4, 0x1.37c294d2f5668p-46)
+A(0x1.1a00000000000p+0, -0x1.8c345d631a000p-4, 0x1.37c294d2f5668p-46)
+A(0x1.1800000000000p+0, -0x1.6f0d28ae56000p-4, -0x1.69737c93373dap-45)
+A(0x1.1600000000000p+0, -0x1.51b073f062000p-4, 0x1.f025b61c65e57p-46)
+A(0x1.1600000000000p+0, -0x1.51b073f062000p-4, 0x1.f025b61c65e57p-46)
+A(0x1.1400000000000p+0, -0x1.341d7961be000p-4, 0x1.c5edaccf913dfp-45)
+A(0x1.1400000000000p+0, -0x1.341d7961be000p-4, 0x1.c5edaccf913dfp-45)
+A(0x1.1200000000000p+0, -0x1.16536eea38000p-4, 0x1.47c5e768fa309p-46)
+A(0x1.1000000000000p+0, -0x1.f0a30c0118000p-5, 0x1.d599e83368e91p-45)
+A(0x1.1000000000000p+0, -0x1.f0a30c0118000p-5, 0x1.d599e83368e91p-45)
+A(0x1.0e00000000000p+0, -0x1.b42dd71198000p-5, 0x1.c827ae5d6704cp-46)
+A(0x1.0e00000000000p+0, -0x1.b42dd71198000p-5, 0x1.c827ae5d6704cp-46)
+A(0x1.0c00000000000p+0, -0x1.77458f632c000p-5, -0x1.cfc4634f2a1eep-45)
+A(0x1.0c00000000000p+0, -0x1.77458f632c000p-5, -0x1.cfc4634f2a1eep-45)
+A(0x1.0a00000000000p+0, -0x1.39e87b9fec000p-5, 0x1.502b7f526feaap-48)
+A(0x1.0a00000000000p+0, -0x1.39e87b9fec000p-5, 0x1.502b7f526feaap-48)
+A(0x1.0800000000000p+0, -0x1.f829b0e780000p-6, -0x1.980267c7e09e4p-45)
+A(0x1.0800000000000p+0, -0x1.f829b0e780000p-6, -0x1.980267c7e09e4p-45)
+A(0x1.0600000000000p+0, -0x1.7b91b07d58000p-6, -0x1.88d5493faa639p-45)
+A(0x1.0400000000000p+0, -0x1.fc0a8b0fc0000p-7, -0x1.f1e7cf6d3a69cp-50)
+A(0x1.0400000000000p+0, -0x1.fc0a8b0fc0000p-7, -0x1.f1e7cf6d3a69cp-50)
+A(0x1.0200000000000p+0, -0x1.fe02a6b100000p-8, -0x1.9e23f0dda40e4p-46)
+A(0x1.0200000000000p+0, -0x1.fe02a6b100000p-8, -0x1.9e23f0dda40e4p-46)
+A(0x1.0000000000000p+0, 0x0.0000000000000p+0, 0x0.0000000000000p+0)
+A(0x1.0000000000000p+0, 0x0.0000000000000p+0, 0x0.0000000000000p+0)
+A(0x1.fc00000000000p-1, 0x1.0101575890000p-7, -0x1.0c76b999d2be8p-46)
+A(0x1.f800000000000p-1, 0x1.0205658938000p-6, -0x1.3dc5b06e2f7d2p-45)
+A(0x1.f400000000000p-1, 0x1.8492528c90000p-6, -0x1.aa0ba325a0c34p-45)
+A(0x1.f000000000000p-1, 0x1.0415d89e74000p-5, 0x1.111c05cf1d753p-47)
+A(0x1.ec00000000000p-1, 0x1.466aed42e0000p-5, -0x1.c167375bdfd28p-45)
+A(0x1.e800000000000p-1, 0x1.894aa149fc000p-5, -0x1.97995d05a267dp-46)
+A(0x1.e400000000000p-1, 0x1.ccb73cdddc000p-5, -0x1.a68f247d82807p-46)
+A(0x1.e200000000000p-1, 0x1.eea31c006c000p-5, -0x1.e113e4fc93b7bp-47)
+A(0x1.de00000000000p-1, 0x1.1973bd1466000p-4, -0x1.5325d560d9e9bp-45)
+A(0x1.da00000000000p-1, 0x1.3bdf5a7d1e000p-4, 0x1.cc85ea5db4ed7p-45)
+A(0x1.d600000000000p-1, 0x1.5e95a4d97a000p-4, -0x1.c69063c5d1d1ep-45)
+A(0x1.d400000000000p-1, 0x1.700d30aeac000p-4, 0x1.c1e8da99ded32p-49)
+A(0x1.d000000000000p-1, 0x1.9335e5d594000p-4, 0x1.3115c3abd47dap-45)
+A(0x1.cc00000000000p-1, 0x1.b6ac88dad6000p-4, -0x1.390802bf768e5p-46)
+A(0x1.ca00000000000p-1, 0x1.c885801bc4000p-4, 0x1.646d1c65aacd3p-45)
+A(0x1.c600000000000p-1, 0x1.ec739830a2000p-4, -0x1.dc068afe645e0p-45)
+A(0x1.c400000000000p-1, 0x1.fe89139dbe000p-4, -0x1.534d64fa10afdp-45)
+A(0x1.c000000000000p-1, 0x1.1178e8227e000p-3, 0x1.1ef78ce2d07f2p-45)
+A(0x1.be00000000000p-1, 0x1.1aa2b7e23f000p-3, 0x1.ca78e44389934p-45)
+A(0x1.ba00000000000p-1, 0x1.2d1610c868000p-3, 0x1.39d6ccb81b4a1p-47)
+A(0x1.b800000000000p-1, 0x1.365fcb0159000p-3, 0x1.62fa8234b7289p-51)
+A(0x1.b400000000000p-1, 0x1.4913d8333b000p-3, 0x1.5837954fdb678p-45)
+A(0x1.b200000000000p-1, 0x1.527e5e4a1b000p-3, 0x1.633e8e5697dc7p-45)
+A(0x1.ae00000000000p-1, 0x1.6574ebe8c1000p-3, 0x1.9cf8b2c3c2e78p-46)
+A(0x1.ac00000000000p-1, 0x1.6f0128b757000p-3, -0x1.5118de59c21e1p-45)
+A(0x1.aa00000000000p-1, 0x1.7898d85445000p-3, -0x1.c661070914305p-46)
+A(0x1.a600000000000p-1, 0x1.8beafeb390000p-3, -0x1.73d54aae92cd1p-47)
+A(0x1.a400000000000p-1, 0x1.95a5adcf70000p-3, 0x1.7f22858a0ff6fp-47)
+A(0x1.a000000000000p-1, 0x1.a93ed3c8ae000p-3, -0x1.8724350562169p-45)
+A(0x1.9e00000000000p-1, 0x1.b31d8575bd000p-3, -0x1.c358d4eace1aap-47)
+A(0x1.9c00000000000p-1, 0x1.bd087383be000p-3, -0x1.d4bc4595412b6p-45)
+A(0x1.9a00000000000p-1, 0x1.c6ffbc6f01000p-3, -0x1.1ec72c5962bd2p-48)
+A(0x1.9600000000000p-1, 0x1.db13db0d49000p-3, -0x1.aff2af715b035p-45)
+A(0x1.9400000000000p-1, 0x1.e530effe71000p-3, 0x1.212276041f430p-51)
+A(0x1.9200000000000p-1, 0x1.ef5ade4dd0000p-3, -0x1.a211565bb8e11p-51)
+A(0x1.9000000000000p-1, 0x1.f991c6cb3b000p-3, 0x1.bcbecca0cdf30p-46)
+A(0x1.8c00000000000p-1, 0x1.07138604d5800p-2, 0x1.89cdb16ed4e91p-48)
+A(0x1.8a00000000000p-1, 0x1.0c42d67616000p-2, 0x1.7188b163ceae9p-45)
+A(0x1.8800000000000p-1, 0x1.1178e8227e800p-2, -0x1.c210e63a5f01cp-45)
+A(0x1.8600000000000p-1, 0x1.16b5ccbacf800p-2, 0x1.b9acdf7a51681p-45)
+A(0x1.8400000000000p-1, 0x1.1bf99635a6800p-2, 0x1.ca6ed5147bdb7p-45)
+A(0x1.8200000000000p-1, 0x1.214456d0eb800p-2, 0x1.a87deba46baeap-47)
+A(0x1.7e00000000000p-1, 0x1.2bef07cdc9000p-2, 0x1.a9cfa4a5004f4p-45)
+A(0x1.7c00000000000p-1, 0x1.314f1e1d36000p-2, -0x1.8e27ad3213cb8p-45)
+A(0x1.7a00000000000p-1, 0x1.36b6776be1000p-2, 0x1.16ecdb0f177c8p-46)
+A(0x1.7800000000000p-1, 0x1.3c25277333000p-2, 0x1.83b54b606bd5cp-46)
+A(0x1.7600000000000p-1, 0x1.419b423d5e800p-2, 0x1.8e436ec90e09dp-47)
+A(0x1.7400000000000p-1, 0x1.4718dc271c800p-2, -0x1.f27ce0967d675p-45)
+A(0x1.7200000000000p-1, 0x1.4c9e09e173000p-2, -0x1.e20891b0ad8a4p-45)
+A(0x1.7000000000000p-1, 0x1.522ae0738a000p-2, 0x1.ebe708164c759p-45)
+A(0x1.6e00000000000p-1, 0x1.57bf753c8d000p-2, 0x1.fadedee5d40efp-46)
+A(0x1.6c00000000000p-1, 0x1.5d5bddf596000p-2, -0x1.a0b2a08a465dcp-47)
+},
+};
diff --git a/src/math/pow_data.h b/src/math/pow_data.h
new file mode 100644
index 0000000..5d609ae
--- /dev/null
+++ b/src/math/pow_data.h
@@ -0,0 +1,22 @@
+/*
+ * Copyright (c) 2018, Arm Limited.
+ * SPDX-License-Identifier: MIT
+ */
+#ifndef _POW_DATA_H
+#define _POW_DATA_H
+
+#include <features.h>
+
+#define POW_LOG_TABLE_BITS 7
+#define POW_LOG_POLY_ORDER 8
+extern hidden const struct pow_log_data {
+	double ln2hi;
+	double ln2lo;
+	double poly[POW_LOG_POLY_ORDER - 1]; /* First coefficient is 1.  */
+	/* Note: the pad field is unused, but allows slightly faster indexing.  */
+	struct {
+		double invc, pad, logc, logctail;
+	} tab[1 << POW_LOG_TABLE_BITS];
+} __pow_log_data;
+
+#endif
diff --git a/src/math/powf.c b/src/math/powf.c
index 7d94c23..732f3b3 100644
--- a/src/math/powf.c
+++ b/src/math/powf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float
 bp[]   = {1.0, 1.5,},
diff --git a/src/math/remainder.c b/src/math/remainder.c
new file mode 100644
index 0000000..612155f
--- /dev/null
+++ b/src/math/remainder.c
@@ -0,0 +1,9 @@
+#include <math.h>
+
+double remainder(double x, double y)
+{
+	int q;
+	return remquo(x, y, &q);
+}
+
+weak_alias(remainder, drem);
diff --git a/src/math/remainderf.c b/src/math/remainderf.c
new file mode 100644
index 0000000..bf1d7b2
--- /dev/null
+++ b/src/math/remainderf.c
@@ -0,0 +1,9 @@
+#include <math.h>
+
+float remainderf(float x, float y)
+{
+	int q;
+	return remquof(x, y, &q);
+}
+
+weak_alias(remainderf, dremf);
diff --git a/src/math/remainderl.c b/src/math/remainderl.c
new file mode 100644
index 0000000..2a13c1d
--- /dev/null
+++ b/src/math/remainderl.c
@@ -0,0 +1,15 @@
+#include <math.h>
+#include <float.h>
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double remainderl(long double x, long double y)
+{
+	return remainder(x, y);
+}
+#else
+long double remainderl(long double x, long double y)
+{
+	int q;
+	return remquol(x, y, &q);
+}
+#endif
diff --git a/src/math/remquo.c b/src/math/remquo.c
new file mode 100644
index 0000000..59d5ad5
--- /dev/null
+++ b/src/math/remquo.c
@@ -0,0 +1,82 @@
+#include <math.h>
+#include <stdint.h>
+
+double remquo(double x, double y, int *quo)
+{
+	union {double f; uint64_t i;} ux = {x}, uy = {y};
+	int ex = ux.i>>52 & 0x7ff;
+	int ey = uy.i>>52 & 0x7ff;
+	int sx = ux.i>>63;
+	int sy = uy.i>>63;
+	uint32_t q;
+	uint64_t i;
+	uint64_t uxi = ux.i;
+
+	*quo = 0;
+	if (uy.i<<1 == 0 || isnan(y) || ex == 0x7ff)
+		return (x*y)/(x*y);
+	if (ux.i<<1 == 0)
+		return x;
+
+	/* normalize x and y */
+	if (!ex) {
+		for (i = uxi<<12; i>>63 == 0; ex--, i <<= 1);
+		uxi <<= -ex + 1;
+	} else {
+		uxi &= -1ULL >> 12;
+		uxi |= 1ULL << 52;
+	}
+	if (!ey) {
+		for (i = uy.i<<12; i>>63 == 0; ey--, i <<= 1);
+		uy.i <<= -ey + 1;
+	} else {
+		uy.i &= -1ULL >> 12;
+		uy.i |= 1ULL << 52;
+	}
+
+	q = 0;
+	if (ex < ey) {
+		if (ex+1 == ey)
+			goto end;
+		return x;
+	}
+
+	/* x mod y */
+	for (; ex > ey; ex--) {
+		i = uxi - uy.i;
+		if (i >> 63 == 0) {
+			uxi = i;
+			q++;
+		}
+		uxi <<= 1;
+		q <<= 1;
+	}
+	i = uxi - uy.i;
+	if (i >> 63 == 0) {
+		uxi = i;
+		q++;
+	}
+	if (uxi == 0)
+		ex = -60;
+	else
+		for (; uxi>>52 == 0; uxi <<= 1, ex--);
+end:
+	/* scale result and decide between |x| and |x|-|y| */
+	if (ex > 0) {
+		uxi -= 1ULL << 52;
+		uxi |= (uint64_t)ex << 52;
+	} else {
+		uxi >>= -ex + 1;
+	}
+	ux.i = uxi;
+	x = ux.f;
+	if (sy)
+		y = -y;
+	if (ex == ey || (ex+1 == ey && (2*x > y || (2*x == y && q%2)))) {
+		x -= y;
+		q++;
+	}
+	q &= 0x7fffffff;
+	*quo = sx^sy ? -(int)q : (int)q;
+	return sx ? -x : x;
+}
diff --git a/src/math/remquof.c b/src/math/remquof.c
new file mode 100644
index 0000000..2f41ff7
--- /dev/null
+++ b/src/math/remquof.c
@@ -0,0 +1,82 @@
+#include <math.h>
+#include <stdint.h>
+
+float remquof(float x, float y, int *quo)
+{
+	union {float f; uint32_t i;} ux = {x}, uy = {y};
+	int ex = ux.i>>23 & 0xff;
+	int ey = uy.i>>23 & 0xff;
+	int sx = ux.i>>31;
+	int sy = uy.i>>31;
+	uint32_t q;
+	uint32_t i;
+	uint32_t uxi = ux.i;
+
+	*quo = 0;
+	if (uy.i<<1 == 0 || isnan(y) || ex == 0xff)
+		return (x*y)/(x*y);
+	if (ux.i<<1 == 0)
+		return x;
+
+	/* normalize x and y */
+	if (!ex) {
+		for (i = uxi<<9; i>>31 == 0; ex--, i <<= 1);
+		uxi <<= -ex + 1;
+	} else {
+		uxi &= -1U >> 9;
+		uxi |= 1U << 23;
+	}
+	if (!ey) {
+		for (i = uy.i<<9; i>>31 == 0; ey--, i <<= 1);
+		uy.i <<= -ey + 1;
+	} else {
+		uy.i &= -1U >> 9;
+		uy.i |= 1U << 23;
+	}
+
+	q = 0;
+	if (ex < ey) {
+		if (ex+1 == ey)
+			goto end;
+		return x;
+	}
+
+	/* x mod y */
+	for (; ex > ey; ex--) {
+		i = uxi - uy.i;
+		if (i >> 31 == 0) {
+			uxi = i;
+			q++;
+		}
+		uxi <<= 1;
+		q <<= 1;
+	}
+	i = uxi - uy.i;
+	if (i >> 31 == 0) {
+		uxi = i;
+		q++;
+	}
+	if (uxi == 0)
+		ex = -30;
+	else
+		for (; uxi>>23 == 0; uxi <<= 1, ex--);
+end:
+	/* scale result and decide between |x| and |x|-|y| */
+	if (ex > 0) {
+		uxi -= 1U << 23;
+		uxi |= (uint32_t)ex << 23;
+	} else {
+		uxi >>= -ex + 1;
+	}
+	ux.i = uxi;
+	x = ux.f;
+	if (sy)
+		y = -y;
+	if (ex == ey || (ex+1 == ey && (2*x > y || (2*x == y && q%2)))) {
+		x -= y;
+		q++;
+	}
+	q &= 0x7fffffff;
+	*quo = sx^sy ? -(int)q : (int)q;
+	return sx ? -x : x;
+}
diff --git a/src/math/remquol.c b/src/math/remquol.c
new file mode 100644
index 0000000..9b065c0
--- /dev/null
+++ b/src/math/remquol.c
@@ -0,0 +1,124 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double remquol(long double x, long double y, int *quo)
+{
+	return remquo(x, y, quo);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double remquol(long double x, long double y, int *quo)
+{
+	union ldshape ux = {x}, uy = {y};
+	int ex = ux.i.se & 0x7fff;
+	int ey = uy.i.se & 0x7fff;
+	int sx = ux.i.se >> 15;
+	int sy = uy.i.se >> 15;
+	uint32_t q;
+
+	*quo = 0;
+	if (y == 0 || isnan(y) || ex == 0x7fff)
+		return (x*y)/(x*y);
+	if (x == 0)
+		return x;
+
+	/* normalize x and y */
+	if (!ex) {
+		ux.i.se = ex;
+		ux.f *= 0x1p120f;
+		ex = ux.i.se - 120;
+	}
+	if (!ey) {
+		uy.i.se = ey;
+		uy.f *= 0x1p120f;
+		ey = uy.i.se - 120;
+	}
+
+	q = 0;
+	if (ex >= ey) {
+		/* x mod y */
+#if LDBL_MANT_DIG == 64
+		uint64_t i, mx, my;
+		mx = ux.i.m;
+		my = uy.i.m;
+		for (; ex > ey; ex--) {
+			i = mx - my;
+			if (mx >= my) {
+				mx = 2*i;
+				q++;
+				q <<= 1;
+			} else if (2*mx < mx) {
+				mx = 2*mx - my;
+				q <<= 1;
+				q++;
+			} else {
+				mx = 2*mx;
+				q <<= 1;
+			}
+		}
+		i = mx - my;
+		if (mx >= my) {
+			mx = i;
+			q++;
+		}
+		if (mx == 0)
+			ex = -120;
+		else
+			for (; mx >> 63 == 0; mx *= 2, ex--);
+		ux.i.m = mx;
+#elif LDBL_MANT_DIG == 113
+		uint64_t hi, lo, xhi, xlo, yhi, ylo;
+		xhi = (ux.i2.hi & -1ULL>>16) | 1ULL<<48;
+		yhi = (uy.i2.hi & -1ULL>>16) | 1ULL<<48;
+		xlo = ux.i2.lo;
+		ylo = ux.i2.lo;
+		for (; ex > ey; ex--) {
+			hi = xhi - yhi;
+			lo = xlo - ylo;
+			if (xlo < ylo)
+				hi -= 1;
+			if (hi >> 63 == 0) {
+				xhi = 2*hi + (lo>>63);
+				xlo = 2*lo;
+				q++;
+			} else {
+				xhi = 2*xhi + (xlo>>63);
+				xlo = 2*xlo;
+			}
+			q <<= 1;
+		}
+		hi = xhi - yhi;
+		lo = xlo - ylo;
+		if (xlo < ylo)
+			hi -= 1;
+		if (hi >> 63 == 0) {
+			xhi = hi;
+			xlo = lo;
+			q++;
+		}
+		if ((xhi|xlo) == 0)
+			ex = -120;
+		else
+			for (; xhi >> 48 == 0; xhi = 2*xhi + (xlo>>63), xlo = 2*xlo, ex--);
+		ux.i2.hi = xhi;
+		ux.i2.lo = xlo;
+#endif
+	}
+
+	/* scale result and decide between |x| and |x|-|y| */
+	if (ex <= 0) {
+		ux.i.se = ex + 120;
+		ux.f *= 0x1p-120f;
+	} else
+		ux.i.se = ex;
+	x = ux.f;
+	if (sy)
+		y = -y;
+	if (ex == ey || (ex+1 == ey && (2*x > y || (2*x == y && q%2)))) {
+		x -= y;
+		q++;
+	}
+	q &= 0x7fffffff;
+	*quo = sx^sy ? -(int)q : (int)q;
+	return sx ? -x : x;
+}
+#endif
diff --git a/src/math/rint.c b/src/math/rint.c
index 2d4b230..2954e15 100644
--- a/src/math/rint.c
+++ b/src/math/rint.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define EPS DBL_EPSILON
 static const double_t toint = 1/EPS;
diff --git a/src/math/rintf.c b/src/math/rintf.c
index e24178b..7a0ec93 100644
--- a/src/math/rintf.c
+++ b/src/math/rintf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define EPS FLT_EPSILON
 static const float_t toint = 1/EPS;
diff --git a/src/math/rintl.c b/src/math/rintl.c
new file mode 100644
index 0000000..374327d
--- /dev/null
+++ b/src/math/rintl.c
@@ -0,0 +1,29 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double rintl(long double x)
+{
+	return rint(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+
+static const long double toint = 1/LDBL_EPSILON;
+
+long double rintl(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	int s = u.i.se >> 15;
+	long double y;
+
+	if (e >= 0x3fff+LDBL_MANT_DIG-1)
+		return x;
+	if (s)
+		y = x - toint + toint;
+	else
+		y = x + toint - toint;
+	if (y == 0)
+		return 0*x;
+	return y;
+}
+#endif
diff --git a/src/math/round.c b/src/math/round.c
index 6a13a7b..e6b9b1d 100644
--- a/src/math/round.c
+++ b/src/math/round.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define EPS DBL_EPSILON
 static const double_t toint = 1/EPS;
diff --git a/src/math/roundf.c b/src/math/roundf.c
index 73a4bc1..e51bf72 100644
--- a/src/math/roundf.c
+++ b/src/math/roundf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 #define EPS FLT_EPSILON
 static const float_t toint = 1/EPS;
diff --git a/src/math/scalb.c b/src/math/scalb.c
new file mode 100644
index 0000000..efe69e6
--- /dev/null
+++ b/src/math/scalb.c
@@ -0,0 +1,35 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_scalb.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunSoft, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+/*
+ * scalb(x, fn) is provide for
+ * passing various standard test suite. One
+ * should use scalbn() instead.
+ */
+
+#define _GNU_SOURCE
+#include <math.h>
+
+double scalb(double x, double fn)
+{
+	if (isnan(x) || isnan(fn))
+		return x*fn;
+	if (!isfinite(fn)) {
+		if (fn > 0.0)
+			return x*fn;
+		else
+			return x/(-fn);
+	}
+	if (rint(fn) != fn) return (fn-fn)/(fn-fn);
+	if ( fn > 65000.0) return scalbn(x, 65000);
+	if (-fn > 65000.0) return scalbn(x,-65000);
+	return scalbn(x,(int)fn);
+}
diff --git a/src/math/scalbf.c b/src/math/scalbf.c
new file mode 100644
index 0000000..f44ed5b
--- /dev/null
+++ b/src/math/scalbf.c
@@ -0,0 +1,32 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/e_scalbf.c */
+/*
+ * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
+ */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+#define _GNU_SOURCE
+#include <math.h>
+
+float scalbf(float x, float fn)
+{
+	if (isnan(x) || isnan(fn)) return x*fn;
+	if (!isfinite(fn)) {
+		if (fn > 0.0f)
+			return x*fn;
+		else
+			return x/(-fn);
+	}
+	if (rintf(fn) != fn) return (fn-fn)/(fn-fn);
+	if ( fn > 65000.0f) return scalbnf(x, 65000);
+	if (-fn > 65000.0f) return scalbnf(x,-65000);
+	return scalbnf(x,(int)fn);
+}
diff --git a/src/math/scalbln.c b/src/math/scalbln.c
index 1fdb3e8..e23c345 100644
--- a/src/math/scalbln.c
+++ b/src/math/scalbln.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double scalbln(double x, long n)
 {
diff --git a/src/math/scalblnf.c b/src/math/scalblnf.c
index 5291812..fd95fd8 100644
--- a/src/math/scalblnf.c
+++ b/src/math/scalblnf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float scalblnf(float x, long n)
 {
diff --git a/src/math/scalbn.c b/src/math/scalbn.c
index c5191f8..5044ee2 100644
--- a/src/math/scalbn.c
+++ b/src/math/scalbn.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 double scalbn(double x, int n)
 {
diff --git a/src/math/scalbnf.c b/src/math/scalbnf.c
index eb9165d..56b9f92 100644
--- a/src/math/scalbnf.c
+++ b/src/math/scalbnf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float scalbnf(float x, int n)
 {
diff --git a/src/math/signgam.c b/src/math/signgam.c
new file mode 100644
index 0000000..ee331b2
--- /dev/null
+++ b/src/math/signgam.c
@@ -0,0 +1,6 @@
+#include <math.h>
+#include "libm.h"
+
+int __signgam = 0;
+
+weak_alias(__signgam, signgam);
diff --git a/src/math/significand.c b/src/math/significand.c
new file mode 100644
index 0000000..40d9aa9
--- /dev/null
+++ b/src/math/significand.c
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <math.h>
+
+double significand(double x)
+{
+	return scalbn(x, -ilogb(x));
+}
diff --git a/src/math/significandf.c b/src/math/significandf.c
new file mode 100644
index 0000000..8a697e1
--- /dev/null
+++ b/src/math/significandf.c
@@ -0,0 +1,7 @@
+#define _GNU_SOURCE
+#include <math.h>
+
+float significandf(float x)
+{
+	return scalbnf(x, -ilogbf(x));
+}
diff --git a/src/math/sin.c b/src/math/sin.c
index a5562b8..7627d6a 100644
--- a/src/math/sin.c
+++ b/src/math/sin.c
@@ -29,8 +29,7 @@
  *      TRIG(x) returns trig(x) nearly rounded
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 double sin(double x)
 {
diff --git a/src/math/sincos.c b/src/math/sincos.c
new file mode 100644
index 0000000..35b2d92
--- /dev/null
+++ b/src/math/sincos.c
@@ -0,0 +1,69 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_sin.c */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+#define _GNU_SOURCE
+#include "libm.h"
+
+void sincos(double x, double *sin, double *cos)
+{
+	double y[2], s, c;
+	uint32_t ix;
+	unsigned n;
+
+	GET_HIGH_WORD(ix, x);
+	ix &= 0x7fffffff;
+
+	/* |x| ~< pi/4 */
+	if (ix <= 0x3fe921fb) {
+		/* if |x| < 2**-27 * sqrt(2) */
+		if (ix < 0x3e46a09e) {
+			/* raise inexact if x!=0 and underflow if subnormal */
+			FORCE_EVAL(ix < 0x00100000 ? x/0x1p120f : x+0x1p120f);
+			*sin = x;
+			*cos = 1.0;
+			return;
+		}
+		*sin = __sin(x, 0.0, 0);
+		*cos = __cos(x, 0.0);
+		return;
+	}
+
+	/* sincos(Inf or NaN) is NaN */
+	if (ix >= 0x7ff00000) {
+		*sin = *cos = x - x;
+		return;
+	}
+
+	/* argument reduction needed */
+	n = __rem_pio2(x, y);
+	s = __sin(y[0], y[1], 1);
+	c = __cos(y[0], y[1]);
+	switch (n&3) {
+	case 0:
+		*sin = s;
+		*cos = c;
+		break;
+	case 1:
+		*sin = c;
+		*cos = -s;
+		break;
+	case 2:
+		*sin = -s;
+		*cos = -c;
+		break;
+	case 3:
+	default:
+		*sin = -c;
+		*cos = s;
+		break;
+	}
+}
diff --git a/src/math/sincosf.c b/src/math/sincosf.c
new file mode 100644
index 0000000..f8ca723
--- /dev/null
+++ b/src/math/sincosf.c
@@ -0,0 +1,117 @@
+/* origin: FreeBSD /usr/src/lib/msun/src/s_sinf.c */
+/*
+ * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
+ * Optimized by Bruce D. Evans.
+ */
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+#define _GNU_SOURCE
+#include "libm.h"
+
+/* Small multiples of pi/2 rounded to double precision. */
+static const double
+s1pio2 = 1*M_PI_2, /* 0x3FF921FB, 0x54442D18 */
+s2pio2 = 2*M_PI_2, /* 0x400921FB, 0x54442D18 */
+s3pio2 = 3*M_PI_2, /* 0x4012D97C, 0x7F3321D2 */
+s4pio2 = 4*M_PI_2; /* 0x401921FB, 0x54442D18 */
+
+void sincosf(float x, float *sin, float *cos)
+{
+	double y;
+	float_t s, c;
+	uint32_t ix;
+	unsigned n, sign;
+
+	GET_FLOAT_WORD(ix, x);
+	sign = ix >> 31;
+	ix &= 0x7fffffff;
+
+	/* |x| ~<= pi/4 */
+	if (ix <= 0x3f490fda) {
+		/* |x| < 2**-12 */
+		if (ix < 0x39800000) {
+			/* raise inexact if x!=0 and underflow if subnormal */
+			FORCE_EVAL(ix < 0x00100000 ? x/0x1p120f : x+0x1p120f);
+			*sin = x;
+			*cos = 1.0f;
+			return;
+		}
+		*sin = __sindf(x);
+		*cos = __cosdf(x);
+		return;
+	}
+
+	/* |x| ~<= 5*pi/4 */
+	if (ix <= 0x407b53d1) {
+		if (ix <= 0x4016cbe3) {  /* |x| ~<= 3pi/4 */
+			if (sign) {
+				*sin = -__cosdf(x + s1pio2);
+				*cos = __sindf(x + s1pio2);
+			} else {
+				*sin = __cosdf(s1pio2 - x);
+				*cos = __sindf(s1pio2 - x);
+			}
+			return;
+		}
+		/* -sin(x+c) is not correct if x+c could be 0: -0 vs +0 */
+		*sin = -__sindf(sign ? x + s2pio2 : x - s2pio2);
+		*cos = -__cosdf(sign ? x + s2pio2 : x - s2pio2);
+		return;
+	}
+
+	/* |x| ~<= 9*pi/4 */
+	if (ix <= 0x40e231d5) {
+		if (ix <= 0x40afeddf) {  /* |x| ~<= 7*pi/4 */
+			if (sign) {
+				*sin = __cosdf(x + s3pio2);
+				*cos = -__sindf(x + s3pio2);
+			} else {
+				*sin = -__cosdf(x - s3pio2);
+				*cos = __sindf(x - s3pio2);
+			}
+			return;
+		}
+		*sin = __sindf(sign ? x + s4pio2 : x - s4pio2);
+		*cos = __cosdf(sign ? x + s4pio2 : x - s4pio2);
+		return;
+	}
+
+	/* sin(Inf or NaN) is NaN */
+	if (ix >= 0x7f800000) {
+		*sin = *cos = x - x;
+		return;
+	}
+
+	/* general argument reduction needed */
+	n = __rem_pio2f(x, &y);
+	s = __sindf(y);
+	c = __cosdf(y);
+	switch (n&3) {
+	case 0:
+		*sin = s;
+		*cos = c;
+		break;
+	case 1:
+		*sin = c;
+		*cos = -s;
+		break;
+	case 2:
+		*sin = -s;
+		*cos = -c;
+		break;
+	case 3:
+	default:
+		*sin = -c;
+		*cos = s;
+		break;
+	}
+}
diff --git a/src/math/sincosl.c b/src/math/sincosl.c
new file mode 100644
index 0000000..d3ac1c4
--- /dev/null
+++ b/src/math/sincosl.c
@@ -0,0 +1,60 @@
+#define _GNU_SOURCE
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+void sincosl(long double x, long double *sin, long double *cos)
+{
+	double sind, cosd;
+	sincos(x, &sind, &cosd);
+	*sin = sind;
+	*cos = cosd;
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+void sincosl(long double x, long double *sin, long double *cos)
+{
+	union ldshape u = {x};
+	unsigned n;
+	long double y[2], s, c;
+
+	u.i.se &= 0x7fff;
+	if (u.i.se == 0x7fff) {
+		*sin = *cos = x - x;
+		return;
+	}
+	if (u.f < M_PI_4) {
+		if (u.i.se < 0x3fff - LDBL_MANT_DIG) {
+			/* raise underflow if subnormal */
+			if (u.i.se == 0) FORCE_EVAL(x*0x1p-120f);
+			*sin = x;
+			/* raise inexact if x!=0 */
+			*cos = 1.0 + x;
+			return;
+		}
+		*sin = __sinl(x, 0, 0);
+		*cos = __cosl(x, 0);
+		return;
+	}
+	n = __rem_pio2l(x, y);
+	s = __sinl(y[0], y[1], 1);
+	c = __cosl(y[0], y[1]);
+	switch (n & 3) {
+	case 0:
+		*sin = s;
+		*cos = c;
+		break;
+	case 1:
+		*sin = c;
+		*cos = -s;
+		break;
+	case 2:
+		*sin = -s;
+		*cos = -c;
+		break;
+	case 3:
+	default:
+		*sin = -c;
+		*cos = s;
+		break;
+	}
+}
+#endif
diff --git a/src/math/sinf.c b/src/math/sinf.c
index bb0b81f..d25b130 100644
--- a/src/math/sinf.c
+++ b/src/math/sinf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* Small multiples of pi/2 rounded to double precision. */
 static const double
diff --git a/src/math/sinh.c b/src/math/sinh.c
index 4c797a2..a01951a 100644
--- a/src/math/sinh.c
+++ b/src/math/sinh.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* sinh(x) = (exp(x) - 1/exp(x))/2
  *         = (exp(x)-1 + (exp(x)-1)/exp(x))/2
@@ -35,6 +34,6 @@ double sinh(double x)
 
 	/* |x| > log(DBL_MAX) or nan */
 	/* note: the result is stored to handle overflow */
-	t = 2*h*__expo2(absx);
+	t = __expo2(absx, 2*h);
 	return t;
 }
diff --git a/src/math/sinhf.c b/src/math/sinhf.c
index 04a3ce9..b9caa79 100644
--- a/src/math/sinhf.c
+++ b/src/math/sinhf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float sinhf(float x)
 {
@@ -27,6 +26,6 @@ float sinhf(float x)
 	}
 
 	/* |x| > logf(FLT_MAX) or nan */
-	t = 2*h*__expo2f(absx);
+	t = __expo2f(absx, 2*h);
 	return t;
 }
diff --git a/src/math/sinhl.c b/src/math/sinhl.c
new file mode 100644
index 0000000..b305d4d
--- /dev/null
+++ b/src/math/sinhl.c
@@ -0,0 +1,43 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double sinhl(long double x)
+{
+	return sinh(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+long double sinhl(long double x)
+{
+	union ldshape u = {x};
+	unsigned ex = u.i.se & 0x7fff;
+	long double h, t, absx;
+
+	h = 0.5;
+	if (u.i.se & 0x8000)
+		h = -h;
+	/* |x| */
+	u.i.se = ex;
+	absx = u.f;
+
+	/* |x| < log(LDBL_MAX) */
+	if (ex < 0x3fff+13 || (ex == 0x3fff+13 && u.i.m>>32 < 0xb17217f7)) {
+		t = expm1l(absx);
+		if (ex < 0x3fff) {
+			if (ex < 0x3fff-32)
+				return x;
+			return h*(2*t - t*t/(1+t));
+		}
+		return h*(t + t/(t+1));
+	}
+
+	/* |x| > log(LDBL_MAX) or nan */
+	t = expl(0.5*absx);
+	return h*t*t;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double sinhl(long double x)
+{
+	return sinh(x);
+}
+#endif
diff --git a/src/math/sinl.c b/src/math/sinl.c
new file mode 100644
index 0000000..9c0b16e
--- /dev/null
+++ b/src/math/sinl.c
@@ -0,0 +1,41 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double sinl(long double x)
+{
+	return sin(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double sinl(long double x)
+{
+	union ldshape u = {x};
+	unsigned n;
+	long double y[2], hi, lo;
+
+	u.i.se &= 0x7fff;
+	if (u.i.se == 0x7fff)
+		return x - x;
+	if (u.f < M_PI_4) {
+		if (u.i.se < 0x3fff - LDBL_MANT_DIG/2) {
+			/* raise inexact if x!=0 and underflow if subnormal */
+			FORCE_EVAL(u.i.se == 0 ? x*0x1p-120f : x+0x1p120f);
+			return x;
+		}
+		return __sinl(x, 0.0, 0);
+	}
+	n = __rem_pio2l(x, y);
+	hi = y[0];
+	lo = y[1];
+	switch (n & 3) {
+	case 0:
+		return __sinl(hi, lo, 1);
+	case 1:
+		return __cosl(hi, lo);
+	case 2:
+		return -__sinl(hi, lo, 1);
+	case 3:
+	default:
+		return -__cosl(hi, lo);
+	}
+}
+#endif
diff --git a/src/math/sqrt.c b/src/math/sqrt.c
index ae27975..8c1237e 100644
--- a/src/math/sqrt.c
+++ b/src/math/sqrt.c
@@ -65,8 +65,7 @@
  *      sqrt(NaN) = NaN         ... with invalid signal for signaling NaN
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 static const double tiny = 1.0e-300;
 
diff --git a/src/math/sqrtf.c b/src/math/sqrtf.c
index 18f0a2d..aad2c08 100644
--- a/src/math/sqrtf.c
+++ b/src/math/sqrtf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static const float tiny = 1.0e-30;
 
diff --git a/src/math/tan.c b/src/math/tan.c
index a64b0f8..1a7c813 100644
--- a/src/math/tan.c
+++ b/src/math/tan.c
@@ -28,8 +28,7 @@
  *      TRIG(x) returns trig(x) nearly rounded
  */
 
-#include <math.h>
-
+#include "libm.h"
 
 double tan(double x)
 {
diff --git a/src/math/tanf.c b/src/math/tanf.c
index 20563be..461eb1f 100644
--- a/src/math/tanf.c
+++ b/src/math/tanf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* Small multiples of pi/2 rounded to double precision. */
 static const double
diff --git a/src/math/tanh.c b/src/math/tanh.c
index 1bdc299..20d6dbc 100644
--- a/src/math/tanh.c
+++ b/src/math/tanh.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 /* tanh(x) = (exp(x) - exp(-x))/(exp(x) + exp(-x))
  *         = (exp(2*x) - 1)/(exp(2*x) - 1 + 2)
diff --git a/src/math/tanhf.c b/src/math/tanhf.c
index fd6125a..10636fb 100644
--- a/src/math/tanhf.c
+++ b/src/math/tanhf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 float tanhf(float x)
 {
diff --git a/src/math/tanhl.c b/src/math/tanhl.c
new file mode 100644
index 0000000..4e1aa9f
--- /dev/null
+++ b/src/math/tanhl.c
@@ -0,0 +1,48 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double tanhl(long double x)
+{
+	return tanh(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+long double tanhl(long double x)
+{
+	union ldshape u = {x};
+	unsigned ex = u.i.se & 0x7fff;
+	unsigned sign = u.i.se & 0x8000;
+	uint32_t w;
+	long double t;
+
+	/* x = |x| */
+	u.i.se = ex;
+	x = u.f;
+	w = u.i.m >> 32;
+
+	if (ex > 0x3ffe || (ex == 0x3ffe && w > 0x8c9f53d5)) {
+		/* |x| > log(3)/2 ~= 0.5493 or nan */
+		if (ex >= 0x3fff+5) {
+			/* |x| >= 32 */
+			t = 1 + 0/(x + 0x1p-120f);
+		} else {
+			t = expm1l(2*x);
+			t = 1 - 2/(t+2);
+		}
+	} else if (ex > 0x3ffd || (ex == 0x3ffd && w > 0x82c577d4)) {
+		/* |x| > log(5/3)/2 ~= 0.2554 */
+		t = expm1l(2*x);
+		t = t/(t+2);
+	} else {
+		/* |x| is small */
+		t = expm1l(-2*x);
+		t = -t/(t+2);
+	}
+	return sign ? -t : t;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double tanhl(long double x)
+{
+	return tanh(x);
+}
+#endif
diff --git a/src/math/tanl.c b/src/math/tanl.c
new file mode 100644
index 0000000..6af0671
--- /dev/null
+++ b/src/math/tanl.c
@@ -0,0 +1,29 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double tanl(long double x)
+{
+	return tan(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+long double tanl(long double x)
+{
+	union ldshape u = {x};
+	long double y[2];
+	unsigned n;
+
+	u.i.se &= 0x7fff;
+	if (u.i.se == 0x7fff)
+		return x - x;
+	if (u.f < M_PI_4) {
+		if (u.i.se < 0x3fff - LDBL_MANT_DIG/2) {
+			/* raise inexact if x!=0 and underflow if subnormal */
+			FORCE_EVAL(u.i.se == 0 ? x*0x1p-120f : x+0x1p120f);
+			return x;
+		}
+		return __tanl(x, 0, 0);
+	}
+	n = __rem_pio2l(x, y);
+	return __tanl(y[0], y[1], n&1);
+}
+#endif
diff --git a/src/math/tgamma.c b/src/math/tgamma.c
new file mode 100644
index 0000000..28f6e0f
--- /dev/null
+++ b/src/math/tgamma.c
@@ -0,0 +1,222 @@
+/*
+"A Precision Approximation of the Gamma Function" - Cornelius Lanczos (1964)
+"Lanczos Implementation of the Gamma Function" - Paul Godfrey (2001)
+"An Analysis of the Lanczos Gamma Approximation" - Glendon Ralph Pugh (2004)
+
+approximation method:
+
+                        (x - 0.5)         S(x)
+Gamma(x) = (x + g - 0.5)         *  ----------------
+                                    exp(x + g - 0.5)
+
+with
+                 a1      a2      a3            aN
+S(x) ~= [ a0 + ----- + ----- + ----- + ... + ----- ]
+               x + 1   x + 2   x + 3         x + N
+
+with a0, a1, a2, a3,.. aN constants which depend on g.
+
+for x < 0 the following reflection formula is used:
+
+Gamma(x)*Gamma(-x) = -pi/(x sin(pi x))
+
+most ideas and constants are from boost and python
+*/
+#include "libm.h"
+
+static const double pi = 3.141592653589793238462643383279502884;
+
+/* sin(pi x) with x > 0x1p-100, if sin(pi*x)==0 the sign is arbitrary */
+static double sinpi(double x)
+{
+	int n;
+
+	/* argument reduction: x = |x| mod 2 */
+	/* spurious inexact when x is odd int */
+	x = x * 0.5;
+	x = 2 * (x - floor(x));
+
+	/* reduce x into [-.25,.25] */
+	n = 4 * x;
+	n = (n+1)/2;
+	x -= n * 0.5;
+
+	x *= pi;
+	switch (n) {
+	default: /* case 4 */
+	case 0:
+		return __sin(x, 0, 0);
+	case 1:
+		return __cos(x, 0);
+	case 2:
+		return __sin(-x, 0, 0);
+	case 3:
+		return -__cos(x, 0);
+	}
+}
+
+#define N 12
+//static const double g = 6.024680040776729583740234375;
+static const double gmhalf = 5.524680040776729583740234375;
+static const double Snum[N+1] = {
+	23531376880.410759688572007674451636754734846804940,
+	42919803642.649098768957899047001988850926355848959,
+	35711959237.355668049440185451547166705960488635843,
+	17921034426.037209699919755754458931112671403265390,
+	6039542586.3520280050642916443072979210699388420708,
+	1439720407.3117216736632230727949123939715485786772,
+	248874557.86205415651146038641322942321632125127801,
+	31426415.585400194380614231628318205362874684987640,
+	2876370.6289353724412254090516208496135991145378768,
+	186056.26539522349504029498971604569928220784236328,
+	8071.6720023658162106380029022722506138218516325024,
+	210.82427775157934587250973392071336271166969580291,
+	2.5066282746310002701649081771338373386264310793408,
+};
+static const double Sden[N+1] = {
+	0, 39916800, 120543840, 150917976, 105258076, 45995730, 13339535,
+	2637558, 357423, 32670, 1925, 66, 1,
+};
+/* n! for small integer n */
+static const double fact[] = {
+	1, 1, 2, 6, 24, 120, 720, 5040.0, 40320.0, 362880.0, 3628800.0, 39916800.0,
+	479001600.0, 6227020800.0, 87178291200.0, 1307674368000.0, 20922789888000.0,
+	355687428096000.0, 6402373705728000.0, 121645100408832000.0,
+	2432902008176640000.0, 51090942171709440000.0, 1124000727777607680000.0,
+};
+
+/* S(x) rational function for positive x */
+static double S(double x)
+{
+	double_t num = 0, den = 0;
+	int i;
+
+	/* to avoid overflow handle large x differently */
+	if (x < 8)
+		for (i = N; i >= 0; i--) {
+			num = num * x + Snum[i];
+			den = den * x + Sden[i];
+		}
+	else
+		for (i = 0; i <= N; i++) {
+			num = num / x + Snum[i];
+			den = den / x + Sden[i];
+		}
+	return num/den;
+}
+
+double tgamma(double x)
+{
+	union {double f; uint64_t i;} u = {x};
+	double absx, y;
+	double_t dy, z, r;
+	uint32_t ix = u.i>>32 & 0x7fffffff;
+	int sign = u.i>>63;
+
+	/* special cases */
+	if (ix >= 0x7ff00000)
+		/* tgamma(nan)=nan, tgamma(inf)=inf, tgamma(-inf)=nan with invalid */
+		return x + INFINITY;
+	if (ix < (0x3ff-54)<<20)
+		/* |x| < 2^-54: tgamma(x) ~ 1/x, +-0 raises div-by-zero */
+		return 1/x;
+
+	/* integer arguments */
+	/* raise inexact when non-integer */
+	if (x == floor(x)) {
+		if (sign)
+			return 0/0.0;
+		if (x <= sizeof fact/sizeof *fact)
+			return fact[(int)x - 1];
+	}
+
+	/* x >= 172: tgamma(x)=inf with overflow */
+	/* x =< -184: tgamma(x)=+-0 with underflow */
+	if (ix >= 0x40670000) { /* |x| >= 184 */
+		if (sign) {
+			FORCE_EVAL((float)(0x1p-126/x));
+			if (floor(x) * 0.5 == floor(x * 0.5))
+				return 0;
+			return -0.0;
+		}
+		x *= 0x1p1023;
+		return x;
+	}
+
+	absx = sign ? -x : x;
+
+	/* handle the error of x + g - 0.5 */
+	y = absx + gmhalf;
+	if (absx > gmhalf) {
+		dy = y - absx;
+		dy -= gmhalf;
+	} else {
+		dy = y - gmhalf;
+		dy -= absx;
+	}
+
+	z = absx - 0.5;
+	r = S(absx) * exp(-y);
+	if (x < 0) {
+		/* reflection formula for negative x */
+		/* sinpi(absx) is not 0, integers are already handled */
+		r = -pi / (sinpi(absx) * absx * r);
+		dy = -dy;
+		z = -z;
+	}
+	r += dy * (gmhalf+0.5) * r / y;
+	z = pow(y, 0.5*z);
+	y = r * z * z;
+	return y;
+}
+
+#if 0
+double __lgamma_r(double x, int *sign)
+{
+	double r, absx;
+
+	*sign = 1;
+
+	/* special cases */
+	if (!isfinite(x))
+		/* lgamma(nan)=nan, lgamma(+-inf)=inf */
+		return x*x;
+
+	/* integer arguments */
+	if (x == floor(x) && x <= 2) {
+		/* n <= 0: lgamma(n)=inf with divbyzero */
+		/* n == 1,2: lgamma(n)=0 */
+		if (x <= 0)
+			return 1/0.0;
+		return 0;
+	}
+
+	absx = fabs(x);
+
+	/* lgamma(x) ~ -log(|x|) for tiny |x| */
+	if (absx < 0x1p-54) {
+		*sign = 1 - 2*!!signbit(x);
+		return -log(absx);
+	}
+
+	/* use tgamma for smaller |x| */
+	if (absx < 128) {
+		x = tgamma(x);
+		*sign = 1 - 2*!!signbit(x);
+		return log(fabs(x));
+	}
+
+	/* second term (log(S)-g) could be more precise here.. */
+	/* or with stirling: (|x|-0.5)*(log(|x|)-1) + poly(1/|x|) */
+	r = (absx-0.5)*(log(absx+gmhalf)-1) + (log(S(absx)) - (gmhalf+0.5));
+	if (x < 0) {
+		/* reflection formula for negative x */
+		x = sinpi(absx);
+		*sign = 2*!!signbit(x) - 1;
+		r = log(pi/(fabs(x)*absx)) - r;
+	}
+	return r;
+}
+
+weak_alias(__lgamma_r, lgamma_r);
+#endif
diff --git a/src/math/tgammaf.c b/src/math/tgammaf.c
new file mode 100644
index 0000000..b4ca51c
--- /dev/null
+++ b/src/math/tgammaf.c
@@ -0,0 +1,6 @@
+#include <math.h>
+
+float tgammaf(float x)
+{
+	return tgamma(x);
+}
diff --git a/src/math/tgammal.c b/src/math/tgammal.c
new file mode 100644
index 0000000..5336c5b
--- /dev/null
+++ b/src/math/tgammal.c
@@ -0,0 +1,281 @@
+/* origin: OpenBSD /usr/src/lib/libm/src/ld80/e_tgammal.c */
+/*
+ * Copyright (c) 2008 Stephen L. Moshier <steve@moshier.net>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+/*
+ *      Gamma function
+ *
+ *
+ * SYNOPSIS:
+ *
+ * long double x, y, tgammal();
+ *
+ * y = tgammal( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns gamma function of the argument.  The result is
+ * correctly signed.
+ *
+ * Arguments |x| <= 13 are reduced by recurrence and the function
+ * approximated by a rational function of degree 7/8 in the
+ * interval (2,3).  Large arguments are handled by Stirling's
+ * formula. Large negative arguments are made positive using
+ * a reflection formula.
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE     -40,+40      10000       3.6e-19     7.9e-20
+ *    IEEE    -1755,+1755   10000       4.8e-18     6.5e-19
+ *
+ * Accuracy for large arguments is dominated by error in powl().
+ *
+ */
+
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double tgammal(long double x)
+{
+	return tgamma(x);
+}
+#elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
+/*
+tgamma(x+2) = tgamma(x+2) P(x)/Q(x)
+0 <= x <= 1
+Relative error
+n=7, d=8
+Peak error =  1.83e-20
+Relative error spread =  8.4e-23
+*/
+static const long double P[8] = {
+ 4.212760487471622013093E-5L,
+ 4.542931960608009155600E-4L,
+ 4.092666828394035500949E-3L,
+ 2.385363243461108252554E-2L,
+ 1.113062816019361559013E-1L,
+ 3.629515436640239168939E-1L,
+ 8.378004301573126728826E-1L,
+ 1.000000000000000000009E0L,
+};
+static const long double Q[9] = {
+-1.397148517476170440917E-5L,
+ 2.346584059160635244282E-4L,
+-1.237799246653152231188E-3L,
+-7.955933682494738320586E-4L,
+ 2.773706565840072979165E-2L,
+-4.633887671244534213831E-2L,
+-2.243510905670329164562E-1L,
+ 4.150160950588455434583E-1L,
+ 9.999999999999999999908E-1L,
+};
+
+/*
+static const long double P[] = {
+-3.01525602666895735709e0L,
+-3.25157411956062339893e1L,
+-2.92929976820724030353e2L,
+-1.70730828800510297666e3L,
+-7.96667499622741999770e3L,
+-2.59780216007146401957e4L,
+-5.99650230220855581642e4L,
+-7.15743521530849602425e4L
+};
+static const long double Q[] = {
+ 1.00000000000000000000e0L,
+-1.67955233807178858919e1L,
+ 8.85946791747759881659e1L,
+ 5.69440799097468430177e1L,
+-1.98526250512761318471e3L,
+ 3.31667508019495079814e3L,
+ 1.60577839621734713377e4L,
+-2.97045081369399940529e4L,
+-7.15743521530849602412e4L
+};
+*/
+#define MAXGAML 1755.455L
+/*static const long double LOGPI = 1.14472988584940017414L;*/
+
+/* Stirling's formula for the gamma function
+tgamma(x) = sqrt(2 pi) x^(x-.5) exp(-x) (1 + 1/x P(1/x))
+z(x) = x
+13 <= x <= 1024
+Relative error
+n=8, d=0
+Peak error =  9.44e-21
+Relative error spread =  8.8e-4
+*/
+static const long double STIR[9] = {
+ 7.147391378143610789273E-4L,
+-2.363848809501759061727E-5L,
+-5.950237554056330156018E-4L,
+ 6.989332260623193171870E-5L,
+ 7.840334842744753003862E-4L,
+-2.294719747873185405699E-4L,
+-2.681327161876304418288E-3L,
+ 3.472222222230075327854E-3L,
+ 8.333333333333331800504E-2L,
+};
+
+#define MAXSTIR 1024.0L
+static const long double SQTPI = 2.50662827463100050242E0L;
+
+/* 1/tgamma(x) = z P(z)
+ * z(x) = 1/x
+ * 0 < x < 0.03125
+ * Peak relative error 4.2e-23
+ */
+static const long double S[9] = {
+-1.193945051381510095614E-3L,
+ 7.220599478036909672331E-3L,
+-9.622023360406271645744E-3L,
+-4.219773360705915470089E-2L,
+ 1.665386113720805206758E-1L,
+-4.200263503403344054473E-2L,
+-6.558780715202540684668E-1L,
+ 5.772156649015328608253E-1L,
+ 1.000000000000000000000E0L,
+};
+
+/* 1/tgamma(-x) = z P(z)
+ * z(x) = 1/x
+ * 0 < x < 0.03125
+ * Peak relative error 5.16e-23
+ * Relative error spread =  2.5e-24
+ */
+static const long double SN[9] = {
+ 1.133374167243894382010E-3L,
+ 7.220837261893170325704E-3L,
+ 9.621911155035976733706E-3L,
+-4.219773343731191721664E-2L,
+-1.665386113944413519335E-1L,
+-4.200263503402112910504E-2L,
+ 6.558780715202536547116E-1L,
+ 5.772156649015328608727E-1L,
+-1.000000000000000000000E0L,
+};
+
+static const long double PIL = 3.1415926535897932384626L;
+
+/* Gamma function computed by Stirling's formula.
+ */
+static long double stirf(long double x)
+{
+	long double y, w, v;
+
+	w = 1.0/x;
+	/* For large x, use rational coefficients from the analytical expansion.  */
+	if (x > 1024.0)
+		w = (((((6.97281375836585777429E-5L * w
+		 + 7.84039221720066627474E-4L) * w
+		 - 2.29472093621399176955E-4L) * w
+		 - 2.68132716049382716049E-3L) * w
+		 + 3.47222222222222222222E-3L) * w
+		 + 8.33333333333333333333E-2L) * w
+		 + 1.0;
+	else
+		w = 1.0 + w * __polevll(w, STIR, 8);
+	y = expl(x);
+	if (x > MAXSTIR) { /* Avoid overflow in pow() */
+		v = powl(x, 0.5L * x - 0.25L);
+		y = v * (v / y);
+	} else {
+		y = powl(x, x - 0.5L) / y;
+	}
+	y = SQTPI * y * w;
+	return y;
+}
+
+long double tgammal(long double x)
+{
+	long double p, q, z;
+
+	if (!isfinite(x))
+		return x + INFINITY;
+
+	q = fabsl(x);
+	if (q > 13.0) {
+		if (x < 0.0) {
+			p = floorl(q);
+			z = q - p;
+			if (z == 0)
+				return 0 / z;
+			if (q > MAXGAML) {
+				z = 0;
+			} else {
+				if (z > 0.5) {
+					p += 1.0;
+					z = q - p;
+				}
+				z = q * sinl(PIL * z);
+				z = fabsl(z) * stirf(q);
+				z = PIL/z;
+			}
+			if (0.5 * p == floorl(q * 0.5))
+				z = -z;
+		} else if (x > MAXGAML) {
+			z = x * 0x1p16383L;
+		} else {
+			z = stirf(x);
+		}
+		return z;
+	}
+
+	z = 1.0;
+	while (x >= 3.0) {
+		x -= 1.0;
+		z *= x;
+	}
+	while (x < -0.03125L) {
+		z /= x;
+		x += 1.0;
+	}
+	if (x <= 0.03125L)
+		goto small;
+	while (x < 2.0) {
+		z /= x;
+		x += 1.0;
+	}
+	if (x == 2.0)
+		return z;
+
+	x -= 2.0;
+	p = __polevll(x, P, 7);
+	q = __polevll(x, Q, 8);
+	z = z * p / q;
+	return z;
+
+small:
+	/* z==1 if x was originally +-0 */
+	if (x == 0 && z != 1)
+		return x / x;
+	if (x < 0.0) {
+		x = -x;
+		q = z / (x * __polevll(x, SN, 8));
+	} else
+		q = z / (x * __polevll(x, S, 8));
+	return q;
+}
+#elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
+// TODO: broken implementation to make things compile
+long double tgammal(long double x)
+{
+	return tgamma(x);
+}
+#endif
diff --git a/src/math/trunc.c b/src/math/trunc.c
index dce9a53..c5a3f91 100644
--- a/src/math/trunc.c
+++ b/src/math/trunc.c
@@ -1,5 +1,5 @@
-#include <math.h>
-
+#include "libm.h"
+#include "libc.h"
 
 static double __trunc(double x)
 {
diff --git a/src/math/truncf.c b/src/math/truncf.c
index d11e9fe..e3de74b 100644
--- a/src/math/truncf.c
+++ b/src/math/truncf.c
@@ -1,5 +1,4 @@
-#include <math.h>
-
+#include "libm.h"
 
 static float __truncf(float x)
 {
diff --git a/src/math/truncl.c b/src/math/truncl.c
new file mode 100644
index 0000000..f07b193
--- /dev/null
+++ b/src/math/truncl.c
@@ -0,0 +1,34 @@
+#include "libm.h"
+
+#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
+long double truncl(long double x)
+{
+	return trunc(x);
+}
+#elif (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
+
+static const long double toint = 1/LDBL_EPSILON;
+
+long double truncl(long double x)
+{
+	union ldshape u = {x};
+	int e = u.i.se & 0x7fff;
+	int s = u.i.se >> 15;
+	long double y;
+
+	if (e >= 0x3fff+LDBL_MANT_DIG-1)
+		return x;
+	if (e <= 0x3fff-1) {
+		FORCE_EVAL(x + 0x1p120f);
+		return x*0;
+	}
+	/* y = int(|x|) - |x|, where int(|x|) is an integer neighbor of |x| */
+	if (s)
+		x = -x;
+	y = x + toint - toint - x;
+	if (y > 0)
+		y -= 1;
+	x += y;
+	return s ? -x : x;
+}
+#endif
diff --git a/src/multibyte/btowc.c b/src/multibyte/btowc.c
new file mode 100644
index 0000000..8acd0a2
--- /dev/null
+++ b/src/multibyte/btowc.c
@@ -0,0 +1,10 @@
+#include <stdio.h>
+#include <wchar.h>
+#include <stdlib.h>
+#include "internal.h"
+
+wint_t btowc(int c)
+{
+	int b = (unsigned char)c;
+	return b<128U ? b : (MB_CUR_MAX==1 && c!=EOF) ? CODEUNIT(c) : WEOF;
+}
diff --git a/src/multibyte/c16rtomb.c b/src/multibyte/c16rtomb.c
new file mode 100644
index 0000000..39ca375
--- /dev/null
+++ b/src/multibyte/c16rtomb.c
@@ -0,0 +1,35 @@
+#include <uchar.h>
+#include <errno.h>
+#include <wchar.h>
+
+size_t c16rtomb(char *restrict s, char16_t c16, mbstate_t *restrict ps)
+{
+	static unsigned internal_state;
+	if (!ps) ps = (void *)&internal_state;
+	unsigned *x = (unsigned *)ps;
+	wchar_t wc;
+
+	if (!s) {
+		if (*x) goto ilseq;
+		return 1;
+	}
+
+	if (!*x && c16 - 0xd800u < 0x400) {
+		*x = c16 - 0xd7c0 << 10;
+		return 0;
+	}
+
+	if (*x) {
+		if (c16 - 0xdc00u >= 0x400) goto ilseq;
+		else wc = *x + c16 - 0xdc00;
+		*x = 0;
+	} else {
+		wc = c16;
+	}
+	return wcrtomb(s, wc, 0);
+
+ilseq:
+	*x = 0;
+	errno = EILSEQ;
+	return -1;
+}
diff --git a/src/multibyte/c32rtomb.c b/src/multibyte/c32rtomb.c
new file mode 100644
index 0000000..6785132
--- /dev/null
+++ b/src/multibyte/c32rtomb.c
@@ -0,0 +1,7 @@
+#include <uchar.h>
+#include <wchar.h>
+
+size_t c32rtomb(char *restrict s, char32_t c32, mbstate_t *restrict ps)
+{
+	return wcrtomb(s, c32, ps);
+}
diff --git a/src/multibyte/internal.c b/src/multibyte/internal.c
new file mode 100644
index 0000000..2f5aaa9
--- /dev/null
+++ b/src/multibyte/internal.c
@@ -0,0 +1,26 @@
+#include "internal.h"
+
+#define C(x) ( x<2 ? -1 : ( R(0x80,0xc0) | x ) )
+#define D(x) C((x+16))
+#define E(x) ( ( x==0 ? R(0xa0,0xc0) : \
+                 x==0xd ? R(0x80,0xa0) : \
+                 R(0x80,0xc0) ) \
+             | ( R(0x80,0xc0) >> 6 ) \
+             | x )
+#define F(x) ( ( x>=5 ? 0 : \
+                 x==0 ? R(0x90,0xc0) : \
+                 x==4 ? R(0x80,0x90) : \
+                 R(0x80,0xc0) ) \
+             | ( R(0x80,0xc0) >> 6 ) \
+             | ( R(0x80,0xc0) >> 12 ) \
+             | x )
+
+const uint32_t bittab[] = {
+	              C(0x2),C(0x3),C(0x4),C(0x5),C(0x6),C(0x7),
+	C(0x8),C(0x9),C(0xa),C(0xb),C(0xc),C(0xd),C(0xe),C(0xf),
+	D(0x0),D(0x1),D(0x2),D(0x3),D(0x4),D(0x5),D(0x6),D(0x7),
+	D(0x8),D(0x9),D(0xa),D(0xb),D(0xc),D(0xd),D(0xe),D(0xf),
+	E(0x0),E(0x1),E(0x2),E(0x3),E(0x4),E(0x5),E(0x6),E(0x7),
+	E(0x8),E(0x9),E(0xa),E(0xb),E(0xc),E(0xd),E(0xe),E(0xf),
+	F(0x0),F(0x1),F(0x2),F(0x3),F(0x4)
+};
diff --git a/src/multibyte/internal.h b/src/multibyte/internal.h
new file mode 100644
index 0000000..45bbc6d
--- /dev/null
+++ b/src/multibyte/internal.h
@@ -0,0 +1,24 @@
+#define bittab __fsmu8
+
+#include <stdint.h>
+#include <features.h>
+
+extern hidden const uint32_t bittab[];
+
+/* Upper 6 state bits are a negative integer offset to bound-check next byte */
+/*    equivalent to: ( (b-0x80) | (b+offset) ) & ~0x3f      */
+#define OOB(c,b) (((((b)>>3)-0x10)|(((b)>>3)+((int32_t)(c)>>26))) & ~7)
+
+/* Interval [a,b). Either a must be 80 or b must be c0, lower 3 bits clear. */
+#define R(a,b) ((uint32_t)((a==0x80 ? 0x40u-b : 0u-a) << 23))
+#define FAILSTATE R(0x80,0x80)
+
+#define SA 0xc2u
+#define SB 0xf4u
+
+/* Arbitrary encoding for representing code units instead of characters. */
+#define CODEUNIT(c) (0xdfff & (signed char)(c))
+#define IS_CODEUNIT(c) ((unsigned)(c)-0xdf80 < 0x80)
+
+/* Get inline definition of MB_CUR_MAX. */
+#include "locale_impl.h"
diff --git a/src/multibyte/mblen.c b/src/multibyte/mblen.c
new file mode 100644
index 0000000..a4304bf
--- /dev/null
+++ b/src/multibyte/mblen.c
@@ -0,0 +1,6 @@
+#include <stdlib.h>
+
+int mblen(const char *s, size_t n)
+{
+	return mbtowc(0, s, n);
+}
diff --git a/src/multibyte/mbrlen.c b/src/multibyte/mbrlen.c
new file mode 100644
index 0000000..accf4b3
--- /dev/null
+++ b/src/multibyte/mbrlen.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+size_t mbrlen(const char *restrict s, size_t n, mbstate_t *restrict st)
+{
+	static unsigned internal;
+	return mbrtowc(0, s, n, st ? st : (mbstate_t *)&internal);
+}
diff --git a/src/multibyte/mbrtowc.c b/src/multibyte/mbrtowc.c
new file mode 100644
index 0000000..c94819e
--- /dev/null
+++ b/src/multibyte/mbrtowc.c
@@ -0,0 +1,51 @@
+#include <stdlib.h>
+#include <wchar.h>
+#include <errno.h>
+#include "internal.h"
+
+size_t mbrtowc(wchar_t *restrict wc, const char *restrict src, size_t n, mbstate_t *restrict st)
+{
+	static unsigned internal_state;
+	unsigned c;
+	const unsigned char *s = (const void *)src;
+	const unsigned N = n;
+	wchar_t dummy;
+
+	if (!st) st = (void *)&internal_state;
+	c = *(unsigned *)st;
+	
+	if (!s) {
+		if (c) goto ilseq;
+		return 0;
+	} else if (!wc) wc = &dummy;
+
+	if (!n) return -2;
+	if (!c) {
+		if (*s < 0x80) return !!(*wc = *s);
+		if (MB_CUR_MAX==1) return (*wc = CODEUNIT(*s)), 1;
+		if (*s-SA > SB-SA) goto ilseq;
+		c = bittab[*s++-SA]; n--;
+	}
+
+	if (n) {
+		if (OOB(c,*s)) goto ilseq;
+loop:
+		c = c<<6 | *s++-0x80; n--;
+		if (!(c&(1U<<31))) {
+			*(unsigned *)st = 0;
+			*wc = c;
+			return N-n;
+		}
+		if (n) {
+			if (*s-0x80u >= 0x40) goto ilseq;
+			goto loop;
+		}
+	}
+
+	*(unsigned *)st = c;
+	return -2;
+ilseq:
+	*(unsigned *)st = 0;
+	errno = EILSEQ;
+	return -1;
+}
diff --git a/src/multibyte/mbsinit.c b/src/multibyte/mbsinit.c
new file mode 100644
index 0000000..c608194
--- /dev/null
+++ b/src/multibyte/mbsinit.c
@@ -0,0 +1,6 @@
+#include <wchar.h>
+
+int mbsinit(const mbstate_t *st)
+{
+	return !st || !*(unsigned *)st;
+}
diff --git a/src/multibyte/mbsnrtowcs.c b/src/multibyte/mbsnrtowcs.c
new file mode 100644
index 0000000..931192e
--- /dev/null
+++ b/src/multibyte/mbsnrtowcs.c
@@ -0,0 +1,55 @@
+#include <wchar.h>
+
+size_t mbsnrtowcs(wchar_t *restrict wcs, const char **restrict src, size_t n, size_t wn, mbstate_t *restrict st)
+{
+	size_t l, cnt=0, n2;
+	wchar_t *ws, wbuf[256];
+	const char *s = *src;
+	const char *tmp_s;
+
+	if (!wcs) ws = wbuf, wn = sizeof wbuf / sizeof *wbuf;
+	else ws = wcs;
+
+	/* making sure output buffer size is at most n/4 will ensure
+	 * that mbsrtowcs never reads more than n input bytes. thus
+	 * we can use mbsrtowcs as long as it's practical.. */
+
+	while ( s && wn && ( (n2=n/4)>=wn || n2>32 ) ) {
+		if (n2>=wn) n2=wn;
+		tmp_s = s;
+		l = mbsrtowcs(ws, &s, n2, st);
+		if (!(l+1)) {
+			cnt = l;
+			wn = 0;
+			break;
+		}
+		if (ws != wbuf) {
+			ws += l;
+			wn -= l;
+		}
+		n = s ? n - (s - tmp_s) : 0;
+		cnt += l;
+	}
+	if (s) while (wn && n) {
+		l = mbrtowc(ws, s, n, st);
+		if (l+2<=2) {
+			if (!(l+1)) {
+				cnt = l;
+				break;
+			}
+			if (!l) {
+				s = 0;
+				break;
+			}
+			/* have to roll back partial character */
+			*(unsigned *)st = 0;
+			break;
+		}
+		s += l; n -= l;
+		/* safe - this loop runs fewer than sizeof(wbuf)/8 times */
+		ws++; wn--;
+		cnt++;
+	}
+	if (wcs) *src = s;
+	return cnt;
+}
diff --git a/src/multibyte/mbsrtowcs.c b/src/multibyte/mbsrtowcs.c
new file mode 100644
index 0000000..9b2f2df
--- /dev/null
+++ b/src/multibyte/mbsrtowcs.c
@@ -0,0 +1,120 @@
+#include <stdint.h>
+#include <wchar.h>
+#include <errno.h>
+#include <string.h>
+#include <stdlib.h>
+#include "internal.h"
+
+size_t mbsrtowcs(wchar_t *restrict ws, const char **restrict src, size_t wn, mbstate_t *restrict st)
+{
+	const unsigned char *s = (const void *)*src;
+	size_t wn0 = wn;
+	unsigned c = 0;
+
+	if (st && (c = *(unsigned *)st)) {
+		if (ws) {
+			*(unsigned *)st = 0;
+			goto resume;
+		} else {
+			goto resume0;
+		}
+	}
+
+	if (MB_CUR_MAX==1) {
+		if (!ws) return strlen((const char *)s);
+		for (;;) {
+			if (!wn) {
+				*src = (const void *)s;
+				return wn0;
+			}
+			if (!*s) break;
+			c = *s++;
+			*ws++ = CODEUNIT(c);
+			wn--;
+		}
+		*ws = 0;
+		*src = 0;
+		return wn0-wn;
+	}
+
+	if (!ws) for (;;) {
+#ifdef __GNUC__
+		typedef uint32_t __attribute__((__may_alias__)) w32;
+		if (*s-1u < 0x7f && (uintptr_t)s%4 == 0) {
+			while (!(( *(w32*)s | *(w32*)s-0x01010101) & 0x80808080)) {
+				s += 4;
+				wn -= 4;
+			}
+		}
+#endif
+		if (*s-1u < 0x7f) {
+			s++;
+			wn--;
+			continue;
+		}
+		if (*s-SA > SB-SA) break;
+		c = bittab[*s++-SA];
+resume0:
+		if (OOB(c,*s)) { s--; break; }
+		s++;
+		if (c&(1U<<25)) {
+			if (*s-0x80u >= 0x40) { s-=2; break; }
+			s++;
+			if (c&(1U<<19)) {
+				if (*s-0x80u >= 0x40) { s-=3; break; }
+				s++;
+			}
+		}
+		wn--;
+		c = 0;
+	} else for (;;) {
+		if (!wn) {
+			*src = (const void *)s;
+			return wn0;
+		}
+#ifdef __GNUC__
+		typedef uint32_t __attribute__((__may_alias__)) w32;
+		if (*s-1u < 0x7f && (uintptr_t)s%4 == 0) {
+			while (wn>=5 && !(( *(w32*)s | *(w32*)s-0x01010101) & 0x80808080)) {
+				*ws++ = *s++;
+				*ws++ = *s++;
+				*ws++ = *s++;
+				*ws++ = *s++;
+				wn -= 4;
+			}
+		}
+#endif
+		if (*s-1u < 0x7f) {
+			*ws++ = *s++;
+			wn--;
+			continue;
+		}
+		if (*s-SA > SB-SA) break;
+		c = bittab[*s++-SA];
+resume:
+		if (OOB(c,*s)) { s--; break; }
+		c = (c<<6) | *s++-0x80;
+		if (c&(1U<<31)) {
+			if (*s-0x80u >= 0x40) { s-=2; break; }
+			c = (c<<6) | *s++-0x80;
+			if (c&(1U<<31)) {
+				if (*s-0x80u >= 0x40) { s-=3; break; }
+				c = (c<<6) | *s++-0x80;
+			}
+		}
+		*ws++ = c;
+		wn--;
+		c = 0;
+	}
+
+	if (!c && !*s) {
+		if (ws) {
+			*ws = 0;
+			*src = 0;
+		}
+		return wn0-wn;
+	}
+	errno = EILSEQ;
+	if (ws) *src = (const void *)s;
+	return -1;
+}
diff --git a/src/multibyte/mbstowcs.c b/src/multibyte/mbstowcs.c
new file mode 100644
index 0000000..dc0d459
--- /dev/null
+++ b/src/multibyte/mbstowcs.c
@@ -0,0 +1,7 @@
+#include <stdlib.h>
+#include <wchar.h>
+
+size_t mbstowcs(wchar_t *restrict ws, const char *restrict s, size_t wn)
+{
+	return mbsrtowcs(ws, (void*)&s, wn, 0);
+}
diff --git a/src/multibyte/mbtowc.c b/src/multibyte/mbtowc.c
new file mode 100644
index 0000000..c191bb0
--- /dev/null
+++ b/src/multibyte/mbtowc.c
@@ -0,0 +1,47 @@
+#include <stdlib.h>
+#include <wchar.h>
+#include <errno.h>
+#include "internal.h"
+
+int mbtowc(wchar_t *restrict wc, const char *restrict src, size_t n)
+{
+	unsigned c;
+	const unsigned char *s = (const void *)src;
+	wchar_t dummy;
+
+	if (!s) return 0;
+	if (!n) goto ilseq;
+	if (!wc) wc = &dummy;
+
+	if (*s < 0x80) return !!(*wc = *s);
+	if (MB_CUR_MAX==1) return (*wc = CODEUNIT(*s)), 1;
+	if (*s-SA > SB-SA) goto ilseq;
+	c = bittab[*s++-SA];
+
+	/* Avoid excessive checks against n: If shifting the state n-1
+	 * times does not clear the high bit, then the value of n is
+	 * insufficient to read a character */
+	if (n<4 && ((c<<(6*n-6)) & (1U<<31))) goto ilseq;
+
+	if (OOB(c,*s)) goto ilseq;
+	c = c<<6 | *s++-0x80;
+	if (!(c&(1U<<31))) {
+		*wc = c;
+		return 2;
+	}
+
+	if (*s-0x80u >= 0x40) goto ilseq;
+	c = c<<6 | *s++-0x80;
+	if (!(c&(1U<<31))) {
+		*wc = c;
+		return 3;
+	}
+
+	if (*s-0x80u >= 0x40) goto ilseq;
+	*wc = c<<6 | *s++-0x80;
+	return 4;
+
+ilseq:
+	errno = EILSEQ;
+	return -1;
+}
diff --git a/src/multibyte/wcsnrtombs.c b/src/multibyte/wcsnrtombs.c
new file mode 100644
index 0000000..95e25e7
--- /dev/null
+++ b/src/multibyte/wcsnrtombs.c
@@ -0,0 +1,35 @@
+#include <wchar.h>
+#include <limits.h>
+#include <string.h>
+
+size_t wcsnrtombs(char *restrict dst, const wchar_t **restrict wcs, size_t wn, size_t n, mbstate_t *restrict st)
+{
+	const wchar_t *ws = *wcs;
+	size_t cnt = 0;
+	if (!dst) n=0;
+	while (ws && wn) {
+		char tmp[MB_LEN_MAX];
+		size_t l = wcrtomb(n<MB_LEN_MAX ? tmp : dst, *ws, 0);
+		if (l==-1) {
+			cnt = -1;
+			break;
+		}
+		if (dst) {
+			if (n<MB_LEN_MAX) {
+				if (l>n) break;
+				memcpy(dst, tmp, l);
+			}
+			dst += l;
+			n -= l;
+		}
+		if (!*ws) {
+			ws = 0;
+			break;
+		}
+		ws++;
+		wn--;
+		cnt += l;
+	}
+	if (dst) *wcs = ws;
+	return cnt;
+}
diff --git a/src/multibyte/wcsrtombs.c b/src/multibyte/wcsrtombs.c
new file mode 100644
index 0000000..b5713ae
--- /dev/null
+++ b/src/multibyte/wcsrtombs.c
@@ -0,0 +1,55 @@
+#include <wchar.h>
+
+size_t wcsrtombs(char *restrict s, const wchar_t **restrict ws, size_t n, mbstate_t *restrict st)
+{
+	const wchar_t *ws2;
+	char buf[4];
+	size_t N = n, l;
+	if (!s) {
+		for (n=0, ws2=*ws; *ws2; ws2++) {
+			if (*ws2 >= 0x80u) {
+				l = wcrtomb(buf, *ws2, 0);
+				if (!(l+1)) return -1;
+				n += l;
+			} else n++;
+		}
+		return n;
+	}
+	while (n>=4) {
+		if (**ws-1u >= 0x7fu) {
+			if (!**ws) {
+				*s = 0;
+				*ws = 0;
+				return N-n;
+			}
+			l = wcrtomb(s, **ws, 0);
+			if (!(l+1)) return -1;
+			s += l;
+			n -= l;
+		} else {
+			*s++ = **ws;
+			n--;
+		}
+		(*ws)++;
+	}
+	while (n) {
+		if (**ws-1u >= 0x7fu) {
+			if (!**ws) {
+				*s = 0;
+				*ws = 0;
+				return N-n;
+			}
+			l = wcrtomb(buf, **ws, 0);
+			if (!(l+1)) return -1;
+			if (l>n) return N-n;
+			wcrtomb(s, **ws, 0);
+			s += l;
+			n -= l;
+		} else {
+			*s++ = **ws;
+			n--;
+		}
+		(*ws)++;
+	}
+	return N;
+}
diff --git a/src/multibyte/wcstombs.c b/src/multibyte/wcstombs.c
new file mode 100644
index 0000000..ab15287
--- /dev/null
+++ b/src/multibyte/wcstombs.c
@@ -0,0 +1,7 @@
+#include <stdlib.h>
+#include <wchar.h>
+
+size_t wcstombs(char *restrict s, const wchar_t *restrict ws, size_t n)
+{
+	return wcsrtombs(s, &(const wchar_t *){ws}, n, 0);
+}
diff --git a/src/multibyte/wctob.c b/src/multibyte/wctob.c
new file mode 100644
index 0000000..b484a3f
--- /dev/null
+++ b/src/multibyte/wctob.c
@@ -0,0 +1,11 @@
+#include <wchar.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include "internal.h"
+
+int wctob(wint_t c)
+{
+	if (c < 128U) return c;
+	if (MB_CUR_MAX==1 && IS_CODEUNIT(c)) return (unsigned char)c;
+	return EOF;
+}
diff --git a/src/multibyte/wctomb.c b/src/multibyte/wctomb.c
new file mode 100644
index 0000000..bad41c5
--- /dev/null
+++ b/src/multibyte/wctomb.c
@@ -0,0 +1,8 @@
+#include <stdlib.h>
+#include <wchar.h>
+
+int wctomb(char *s, wchar_t wc)
+{
+	if (!s) return 0;
+	return wcrtomb(s, wc, 0);
+}
diff --git a/src/stdlib/wcstod.c b/src/stdlib/wcstod.c
new file mode 100644
index 0000000..0deb701
--- /dev/null
+++ b/src/stdlib/wcstod.c
@@ -0,0 +1,64 @@
+#include "shgetc.h"
+#include "floatscan.h"
+#include "stdio_impl.h"
+#include <wchar.h>
+#include <wctype.h>
+
+/* This read function heavily cheats. It knows:
+ *  (1) len will always be 1
+ *  (2) non-ascii characters don't matter */
+
+static size_t do_read(FILE *f, unsigned char *buf, size_t len)
+{
+	size_t i;
+	const wchar_t *wcs = f->cookie;
+
+	if (!wcs[0]) wcs=L"@";
+	for (i=0; i<f->buf_size && wcs[i]; i++)
+		f->buf[i] = wcs[i] < 128 ? wcs[i] : '@';
+	f->rpos = f->buf;
+	f->rend = f->buf + i;
+	f->cookie = (void *)(wcs+i);
+
+	if (i && len) {
+		*buf = *f->rpos++;
+		return 1;
+	}
+	return 0;
+}
+
+static long double wcstox(const wchar_t *s, wchar_t **p, int prec)
+{
+	wchar_t *t = (wchar_t *)s;
+	unsigned char buf[64];
+	FILE f = {0};
+	f.flags = 0;
+	f.rpos = f.rend = f.buf = buf + 4;
+	f.buf_size = sizeof buf - 4;
+	f.lock = -1;
+	f.read = do_read;
+	while (iswspace(*t)) t++;
+	f.cookie = (void *)t;
+	shlim(&f, 0);
+	long double y = __floatscan(&f, prec, 1);
+	if (p) {
+		size_t cnt = shcnt(&f);
+		*p = cnt ? t + cnt : (wchar_t *)s;
+	}
+	return y;
+}
+
+float wcstof(const wchar_t *restrict s, wchar_t **restrict p)
+{
+	return wcstox(s, p, 0);
+}
+
+double wcstod(const wchar_t *restrict s, wchar_t **restrict p)
+{
+	return wcstox(s, p, 1);
+}
+
+long double wcstold(const wchar_t *restrict s, wchar_t **restrict p)
+{
+	return wcstox(s, p, 2);
+}
diff --git a/src/stdlib/wcstol.c b/src/stdlib/wcstol.c
new file mode 100644
index 0000000..1eeb495
--- /dev/null
+++ b/src/stdlib/wcstol.c
@@ -0,0 +1,81 @@
+#include "stdio_impl.h"
+#include "intscan.h"
+#include "shgetc.h"
+#include <inttypes.h>
+#include <limits.h>
+#include <wctype.h>
+#include <wchar.h>
+
+/* This read function heavily cheats. It knows:
+ *  (1) len will always be 1
+ *  (2) non-ascii characters don't matter */
+
+static size_t do_read(FILE *f, unsigned char *buf, size_t len)
+{
+	size_t i;
+	const wchar_t *wcs = f->cookie;
+
+	if (!wcs[0]) wcs=L"@";
+	for (i=0; i<f->buf_size && wcs[i]; i++)
+		f->buf[i] = wcs[i] < 128 ? wcs[i] : '@';
+	f->rpos = f->buf;
+	f->rend = f->buf + i;
+	f->cookie = (void *)(wcs+i);
+
+	if (i && len) {
+		*buf = *f->rpos++;
+		return 1;
+	}
+	return 0;
+}
+
+static unsigned long long wcstox(const wchar_t *s, wchar_t **p, int base, unsigned long long lim)
+{
+	wchar_t *t = (wchar_t *)s;
+	unsigned char buf[64];
+	FILE f = {0};
+	f.flags = 0;
+	f.rpos = f.rend = f.buf = buf + 4;
+	f.buf_size = sizeof buf - 4;
+	f.lock = -1;
+	f.read = do_read;
+	while (iswspace(*t)) t++;
+	f.cookie = (void *)t;
+	shlim(&f, 0);
+	unsigned long long y = __intscan(&f, base, 1, lim);
+	if (p) {
+		size_t cnt = shcnt(&f);
+		*p = cnt ? t + cnt : (wchar_t *)s;
+	}
+	return y;
+}
+
+unsigned long long wcstoull(const wchar_t *restrict s, wchar_t **restrict p, int base)
+{
+	return wcstox(s, p, base, ULLONG_MAX);
+}
+
+long long wcstoll(const wchar_t *restrict s, wchar_t **restrict p, int base)
+{
+	return wcstox(s, p, base, LLONG_MIN);
+}
+
+unsigned long wcstoul(const wchar_t *restrict s, wchar_t **restrict p, int base)
+{
+	return wcstox(s, p, base, ULONG_MAX);
+}
+
+long wcstol(const wchar_t *restrict s, wchar_t **restrict p, int base)
+{
+	return wcstox(s, p, base, 0UL+LONG_MIN);
+}
+
+intmax_t wcstoimax(const wchar_t *restrict s, wchar_t **restrict p, int base)
+{
+	return wcstoll(s, p, base);
+}
+
+uintmax_t wcstoumax(const wchar_t *restrict s, wchar_t **restrict p, int base)
+{
+	return wcstoull(s, p, base);
+}
diff --git a/src/string/wcpcpy.c b/src/string/wcpcpy.c
new file mode 100644
index 0000000..ef40134
--- /dev/null
+++ b/src/string/wcpcpy.c
@@ -0,0 +1,6 @@
+#include <wchar.h>
+
+wchar_t *wcpcpy(wchar_t *restrict d, const wchar_t *restrict s)
+{
+	return wcscpy(d, s) + wcslen(s);
+}
diff --git a/src/string/wcpncpy.c b/src/string/wcpncpy.c
new file mode 100644
index 0000000..d7a09ea
--- /dev/null
+++ b/src/string/wcpncpy.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+#include "wchar_impl.h"
+
+wchar_t *wcpncpy(wchar_t *restrict d, const wchar_t *restrict s, size_t n)
+{
+	return wcsncpy(d, s, n) + wcsnlen(s, n);
+}
diff --git a/src/string/wcscasecmp.c b/src/string/wcscasecmp.c
new file mode 100644
index 0000000..df87325
--- /dev/null
+++ b/src/string/wcscasecmp.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+#include <wctype.h>
+#include "wchar_impl.h"
+
+int wcscasecmp(const wchar_t *l, const wchar_t *r)
+{
+	return wcsncasecmp(l, r, -1);
+}
diff --git a/src/string/wcscasecmp_l.c b/src/string/wcscasecmp_l.c
new file mode 100644
index 0000000..a3c6cbf
--- /dev/null
+++ b/src/string/wcscasecmp_l.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+#include "wchar_impl.h"
+
+int wcscasecmp_l(const wchar_t *l, const wchar_t *r, locale_t locale)
+{
+	return wcscasecmp(l, r);
+}
diff --git a/src/string/wcscat.c b/src/string/wcscat.c
new file mode 100644
index 0000000..d4f00eb
--- /dev/null
+++ b/src/string/wcscat.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+wchar_t *wcscat(wchar_t *restrict dest, const wchar_t *restrict src)
+{
+	wcscpy(dest + wcslen(dest), src);
+	return dest;
+}
diff --git a/src/string/wcscmp.c b/src/string/wcscmp.c
new file mode 100644
index 0000000..26eeee7
--- /dev/null
+++ b/src/string/wcscmp.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+int wcscmp(const wchar_t *l, const wchar_t *r)
+{
+	for (; *l==*r && *l && *r; l++, r++);
+	return *l - *r;
+}
diff --git a/src/string/wcscpy.c b/src/string/wcscpy.c
new file mode 100644
index 0000000..625bf53
--- /dev/null
+++ b/src/string/wcscpy.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wcscpy(wchar_t *restrict d, const wchar_t *restrict s)
+{
+	wchar_t *a = d;
+	while ((*d++ = *s++));
+	return a;
+}
diff --git a/src/string/wcscspn.c b/src/string/wcscspn.c
new file mode 100644
index 0000000..c4e5272
--- /dev/null
+++ b/src/string/wcscspn.c
@@ -0,0 +1,10 @@
+#include <wchar.h>
+
+size_t wcscspn(const wchar_t *s, const wchar_t *c)
+{
+	const wchar_t *a;
+	if (!c[0]) return wcslen(s);
+	if (!c[1]) return (s=wcschr(a=s, *c)) ? s-a : wcslen(a);
+	for (a=s; *s && !wcschr(c, *s); s++);
+	return s-a;
+}
diff --git a/src/string/wcsdup.c b/src/string/wcsdup.c
new file mode 100644
index 0000000..f398e80
--- /dev/null
+++ b/src/string/wcsdup.c
@@ -0,0 +1,10 @@
+#include <stdlib.h>
+#include <wchar.h>
+
+wchar_t *wcsdup(const wchar_t *s)
+{
+	size_t l = wcslen(s);
+	wchar_t *d = malloc((l+1)*sizeof(wchar_t));
+	if (!d) return NULL;
+	return wmemcpy(d, s, l+1);
+}
diff --git a/src/string/wcsncasecmp.c b/src/string/wcsncasecmp.c
new file mode 100644
index 0000000..8fefe79
--- /dev/null
+++ b/src/string/wcsncasecmp.c
@@ -0,0 +1,9 @@
+#include <wchar.h>
+#include <wctype.h>
+
+int wcsncasecmp(const wchar_t *l, const wchar_t *r, size_t n)
+{
+	if (!n--) return 0;
+	for (; *l && *r && n && (*l == *r || towlower(*l) == towlower(*r)); l++, r++, n--);
+	return towlower(*l) - towlower(*r);
+}
diff --git a/src/string/wcsncasecmp_l.c b/src/string/wcsncasecmp_l.c
new file mode 100644
index 0000000..dcc9cab
--- /dev/null
+++ b/src/string/wcsncasecmp_l.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+#include "wchar_impl.h"
+
+int wcsncasecmp_l(const wchar_t *l, const wchar_t *r, size_t n, locale_t locale)
+{
+	return wcsncasecmp(l, r, n);
+}
diff --git a/src/string/wcsncat.c b/src/string/wcsncat.c
new file mode 100644
index 0000000..8563f1a
--- /dev/null
+++ b/src/string/wcsncat.c
@@ -0,0 +1,10 @@
+#include <wchar.h>
+
+wchar_t *wcsncat(wchar_t *restrict d, const wchar_t *restrict s, size_t n)
+{
+	wchar_t *a = d;
+	d += wcslen(d);
+	while (n && *s) n--, *d++ = *s++;
+	*d++ = 0;
+	return a;
+}
diff --git a/src/string/wcsncmp.c b/src/string/wcsncmp.c
new file mode 100644
index 0000000..4ab32a9
--- /dev/null
+++ b/src/string/wcsncmp.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+int wcsncmp(const wchar_t *l, const wchar_t *r, size_t n)
+{
+	for (; n && *l==*r && *l && *r; n--, l++, r++);
+	return n ? *l - *r : 0;
+}
diff --git a/src/string/wcsncpy.c b/src/string/wcsncpy.c
new file mode 100644
index 0000000..4bede04
--- /dev/null
+++ b/src/string/wcsncpy.c
@@ -0,0 +1,9 @@
+#include <wchar.h>
+
+wchar_t *wcsncpy(wchar_t *restrict d, const wchar_t *restrict s, size_t n)
+{
+	wchar_t *a = d;
+	while (n && *s) n--, *d++ = *s++;
+	wmemset(d, 0, n);
+	return a;
+}
diff --git a/src/string/wcsnlen.c b/src/string/wcsnlen.c
new file mode 100644
index 0000000..a776337
--- /dev/null
+++ b/src/string/wcsnlen.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+size_t wcsnlen(const wchar_t *s, size_t n)
+{
+	const wchar_t *z = wmemchr(s, 0, n);
+	if (z) n = z-s;
+	return n;
+}
diff --git a/src/string/wcspbrk.c b/src/string/wcspbrk.c
new file mode 100644
index 0000000..0c72c19
--- /dev/null
+++ b/src/string/wcspbrk.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+wchar_t *wcspbrk(const wchar_t *s, const wchar_t *b)
+{
+	s += wcscspn(s, b);
+	return *s ? (wchar_t *)s : NULL;
+}
diff --git a/src/string/wcsspn.c b/src/string/wcsspn.c
new file mode 100644
index 0000000..4320d8f
--- /dev/null
+++ b/src/string/wcsspn.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+size_t wcsspn(const wchar_t *s, const wchar_t *c)
+{
+	const wchar_t *a;
+	for (a=s; *s && wcschr(c, *s); s++);
+	return s-a;
+}
diff --git a/src/string/wcsstr.c b/src/string/wcsstr.c
new file mode 100644
index 0000000..4caaef3
--- /dev/null
+++ b/src/string/wcsstr.c
@@ -0,0 +1,105 @@
+#include <wchar.h>
+
+#define MAX(a,b) ((a)>(b)?(a):(b))
+#define MIN(a,b) ((a)<(b)?(a):(b))
+
+static wchar_t *twoway_wcsstr(const wchar_t *h, const wchar_t *n)
+{
+	const wchar_t *z;
+	size_t l, ip, jp, k, p, ms, p0, mem, mem0;
+
+	/* Computing length of needle */
+	for (l=0; n[l] && h[l]; l++);
+	if (n[l]) return 0; /* hit the end of h */
+
+	/* Compute maximal suffix */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] > n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	ms = ip;
+	p0 = p;
+
+	/* And with the opposite comparison */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] < n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	if (ip+1 > ms+1) ms = ip;
+	else p = p0;
+
+	/* Periodic needle? */
+	if (wmemcmp(n, n+p, ms+1)) {
+		mem0 = 0;
+		p = MAX(ms, l-ms-1) + 1;
+	} else mem0 = l-p;
+	mem = 0;
+
+	/* Initialize incremental end-of-haystack pointer */
+	z = h;
+
+	/* Search loop */
+	for (;;) {
+		/* Update incremental end-of-haystack pointer */
+		if (z-h < l) {
+			/* Fast estimate for MIN(l,63) */
+			size_t grow = l | 63;
+			const wchar_t *z2 = wmemchr(z, 0, grow);
+			if (z2) {
+				z = z2;
+				if (z-h < l) return 0;
+			} else z += grow;
+		}
+
+		/* Compare right half */
+		for (k=MAX(ms+1,mem); n[k] && n[k] == h[k]; k++);
+		if (n[k]) {
+			h += k-ms;
+			mem = 0;
+			continue;
+		}
+		/* Compare left half */
+		for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
+		if (k <= mem) return (wchar_t *)h;
+		h += p;
+		mem = mem0;
+	}
+}
+
+wchar_t *wcsstr(const wchar_t *restrict h, const wchar_t *restrict n)
+{
+	/* Return immediately on empty needle or haystack */
+	if (!n[0]) return (wchar_t *)h;
+	if (!h[0]) return 0;
+
+	/* Use faster algorithms for short needles */
+	h = wcschr(h, *n);
+	if (!h || !n[1]) return (wchar_t *)h;
+	if (!h[1]) return 0;
+
+	return twoway_wcsstr(h, n);
+}
diff --git a/src/string/wcstok.c b/src/string/wcstok.c
new file mode 100644
index 0000000..ecc8033
--- /dev/null
+++ b/src/string/wcstok.c
@@ -0,0 +1,12 @@
+#include <wchar.h>
+
+wchar_t *wcstok(wchar_t *restrict s, const wchar_t *restrict sep, wchar_t **restrict p)
+{
+	if (!s && !(s = *p)) return NULL;
+	s += wcsspn(s, sep);
+	if (!*s) return *p = 0;
+	*p = s + wcscspn(s, sep);
+	if (**p) *(*p)++ = 0;
+	else *p = 0;
+	return s;
+}
diff --git a/src/string/wcswcs.c b/src/string/wcswcs.c
new file mode 100644
index 0000000..9cfe4ac
--- /dev/null
+++ b/src/string/wcswcs.c
@@ -0,0 +1,6 @@
+#include <wchar.h>
+
+wchar_t *wcswcs(const wchar_t *haystack, const wchar_t *needle)
+{
+	return wcsstr(haystack, needle);
+}
diff --git a/src/string/wmemchr.c b/src/string/wmemchr.c
new file mode 100644
index 0000000..2bc2c27
--- /dev/null
+++ b/src/string/wmemchr.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+wchar_t *wmemchr(const wchar_t *s, wchar_t c, size_t n)
+{
+	for (; n && *s != c; n--, s++);
+	return n ? (wchar_t *)s : 0;
+}
diff --git a/src/string/wmemcmp.c b/src/string/wmemcmp.c
new file mode 100644
index 0000000..2a19326
--- /dev/null
+++ b/src/string/wmemcmp.c
@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+int wmemcmp(const wchar_t *l, const wchar_t *r, size_t n)
+{
+	for (; n && *l==*r; n--, l++, r++);
+	return n ? *l-*r : 0;
+}
diff --git a/src/string/wmemcpy.c b/src/string/wmemcpy.c
new file mode 100644
index 0000000..52e6e6e
--- /dev/null
+++ b/src/string/wmemcpy.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wmemcpy(wchar_t *restrict d, const wchar_t *restrict s, size_t n)
+{
+	wchar_t *a = d;
+	while (n--) *d++ = *s++;
+	return a;
+}
diff --git a/src/string/wmemmove.c b/src/string/wmemmove.c
new file mode 100644
index 0000000..964c903
--- /dev/null
+++ b/src/string/wmemmove.c
@@ -0,0 +1,13 @@
+#include <wchar.h>
+#include <stdint.h>
+
+wchar_t *wmemmove(wchar_t *d, const wchar_t *s, size_t n)
+{
+	wchar_t *d0 = d;
+	if (d == s) return d;
+	if ((uintptr_t)d-(uintptr_t)s < n * sizeof *d)
+		while (n--) d[n] = s[n];
+	else
+		while (n--) *d++ = *s++;
+	return d0;
+}
diff --git a/src/string/wmemset.c b/src/string/wmemset.c
new file mode 100644
index 0000000..07a037a
--- /dev/null
+++ b/src/string/wmemset.c
@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wmemset(wchar_t *d, wchar_t c, size_t n)
+{
+	wchar_t *ret = d;
+	while (n--) *d++ = c;
+	return ret;
+}
diff --git a/src/time/wcsftime.c b/src/time/wcsftime.c
new file mode 100644
index 0000000..98ddd8d
--- /dev/null
+++ b/src/time/wcsftime.c
@@ -0,0 +1,11 @@
+#include <wchar.h>
+#include <time.h>
+#include <locale.h>
+#include "locale_impl.h"
+#include "time_impl.h"
+
+size_t wcsftime(wchar_t *restrict wcs, size_t n, const wchar_t *restrict f, const struct tm *restrict tm)
+{
+    /* FIXME: no support */
+	return 0;
+}
-- 
Gitee


From 1fcfbb9caea1332df3b5f706f405c26f30d95ac5 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Thu, 8 Sep 2022 00:23:12 +0800
Subject: [PATCH 09/13] fix: update some FIXME

---
 src/stdio/vfprintf.c | 5 ++---
 src/stdio/vfscanf.c  | 2 +-
 src/time/wcsftime.c  | 4 +++-
 3 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/src/stdio/vfprintf.c b/src/stdio/vfprintf.c
index 67bc086..9b7e2c2 100644
--- a/src/stdio/vfprintf.c
+++ b/src/stdio/vfprintf.c
@@ -602,8 +602,7 @@ static int printf_core(FILE *f, const char *fmt, va_list *ap, union arg *nl_arg,
 			arg.p = wc;
 			p = -1;
 		case 'S':
-            /* FIXME: !!!
-			ws = arg.p;
+            ws = arg.p;
 			for (i=l=0; i<p && *ws && (l=wctomb(mb, *ws++))>=0 && l<=p-i; i+=l);
 			if (l<0) return -1;
 			if (i > INT_MAX) goto overflow;
@@ -613,7 +612,7 @@ static int printf_core(FILE *f, const char *fmt, va_list *ap, union arg *nl_arg,
 			for (i=0; i<0U+p && *ws && i+(l=wctomb(mb, *ws++))<=p; i+=l)
 				out(f, mb, l);
 			pad(f, ' ', w, p, fl^LEFT_ADJ);
-			l = w>p ? w : p;*/
+			l = w>p ? w : p;
 			continue;
 		case 'e': case 'f': case 'g': case 'a':
 		case 'E': case 'F': case 'G': case 'A':
diff --git a/src/stdio/vfscanf.c b/src/stdio/vfscanf.c
index 87bda14..780f6e4 100644
--- a/src/stdio/vfscanf.c
+++ b/src/stdio/vfscanf.c
@@ -248,7 +248,7 @@ int vfscanf(FILE *restrict f, const char *restrict fmt, va_list ap)
 						wcs = tmp;
 					}
 				}
-				/* FIXME:!!! if (!mbsinit(&st)) goto input_fail;*/
+				if (!mbsinit(&st)) goto input_fail;
 			} else if (alloc) {
 				s = malloc(k);
 				if (!s) goto alloc_fail;
diff --git a/src/time/wcsftime.c b/src/time/wcsftime.c
index 98ddd8d..feee58c 100644
--- a/src/time/wcsftime.c
+++ b/src/time/wcsftime.c
@@ -1,11 +1,13 @@
 #include <wchar.h>
 #include <time.h>
+#include <stdlib.h>
 #include <locale.h>
 #include "locale_impl.h"
 #include "time_impl.h"
 
 size_t wcsftime(wchar_t *restrict wcs, size_t n, const wchar_t *restrict f, const struct tm *restrict tm)
 {
-    /* FIXME: no support */
+    printf("wcsftime not support!\n");
+    abort();
 	return 0;
 }
-- 
Gitee


From 31ba3d34df80271c0b057ca6a00ea5552535e978 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Thu, 8 Sep 2022 01:00:05 +0800
Subject: [PATCH 10/13] feat: add stdio wide char support

---
 src/arch/generic/bits/errno.h           |   3 +-
 src/ctype/isascii.c                     |   7 ++
 src/ctype/toascii.c                     |   7 ++
 src/env/__libc_start_main.c             |   1 +
 src/include/ctype.h                     |   6 ++
 src/internal/libc.h                     |   1 +
 src/internal/locale_impl.h              |   2 +-
 src/stdio/asprintf.c_                   |  13 ---
 src/stdio/dprintf.c_                    |  12 ---
 src/stdio/ext.c_                        |  57 ----------
 src/stdio/ext2.c_                       |  24 -----
 src/stdio/fgetln.c_                     |  21 ----
 src/stdio/{fgetwc.c_ => fgetwc.c}       |   0
 src/stdio/{fgetws.c_ => fgetws.c}       |   0
 src/stdio/fileno.c_                     |  16 ---
 src/stdio/flockfile.c_                  |   9 --
 src/stdio/fmemopen.c_                   | 127 ----------------------
 src/stdio/fopencookie.c_                | 135 ------------------------
 src/stdio/{fputwc.c_ => fputwc.c}       |   0
 src/stdio/{fputws.c_ => fputws.c}       |   0
 src/stdio/ftrylockfile.c_               |  46 --------
 src/stdio/funlockfile.c_                |  13 ---
 src/stdio/{fwide.c_ => fwide.c}         |   0
 src/stdio/{fwprintf.c_ => fwprintf.c}   |   0
 src/stdio/{fwscanf.c_ => fwscanf.c}     |   0
 src/stdio/getc_unlocked.c_              |   9 --
 src/stdio/getchar_unlocked.c_           |   6 --
 src/stdio/{getwc.c_ => getwc.c}         |   0
 src/stdio/{getwchar.c_ => getwchar.c}   |   0
 src/stdio/open_memstream.c_             |  99 -----------------
 src/stdio/open_wmemstream.c_            | 102 ------------------
 src/stdio/putc_unlocked.c_              |   9 --
 src/stdio/putchar_unlocked.c_           |   6 --
 src/stdio/{putwc.c_ => putwc.c}         |   0
 src/stdio/{putwchar.c_ => putwchar.c}   |   0
 src/stdio/setbuffer.c_                  |   7 --
 src/stdio/setlinebuf.c_                 |   7 --
 src/stdio/{swprintf.c_ => swprintf.c}   |   0
 src/stdio/{swscanf.c_ => swscanf.c}     |   0
 src/stdio/tempnam.c_                    |  49 ---------
 src/stdio/{ungetwc.c_ => ungetwc.c}     |   0
 src/stdio/vasprintf.c_                  |  15 ---
 src/stdio/vdprintf.c_                   |  11 --
 src/stdio/{vfwprintf.c_ => vfwprintf.c} |   0
 src/stdio/{vfwscanf.c_ => vfwscanf.c}   |   0
 src/stdio/{vswprintf.c_ => vswprintf.c} |   0
 src/stdio/{vswscanf.c_ => vswscanf.c}   |   0
 src/stdio/{vwprintf.c_ => vwprintf.c}   |   0
 src/stdio/{vwscanf.c_ => vwscanf.c}     |   0
 src/stdio/{wprintf.c_ => wprintf.c}     |   0
 src/stdio/{wscanf.c_ => wscanf.c}       |   0
 51 files changed, 25 insertions(+), 795 deletions(-)
 create mode 100644 src/ctype/isascii.c
 create mode 100644 src/ctype/toascii.c
 delete mode 100644 src/stdio/asprintf.c_
 delete mode 100644 src/stdio/dprintf.c_
 delete mode 100644 src/stdio/ext.c_
 delete mode 100644 src/stdio/ext2.c_
 delete mode 100644 src/stdio/fgetln.c_
 rename src/stdio/{fgetwc.c_ => fgetwc.c} (100%)
 rename src/stdio/{fgetws.c_ => fgetws.c} (100%)
 delete mode 100644 src/stdio/fileno.c_
 delete mode 100644 src/stdio/flockfile.c_
 delete mode 100644 src/stdio/fmemopen.c_
 delete mode 100644 src/stdio/fopencookie.c_
 rename src/stdio/{fputwc.c_ => fputwc.c} (100%)
 rename src/stdio/{fputws.c_ => fputws.c} (100%)
 delete mode 100644 src/stdio/ftrylockfile.c_
 delete mode 100644 src/stdio/funlockfile.c_
 rename src/stdio/{fwide.c_ => fwide.c} (100%)
 rename src/stdio/{fwprintf.c_ => fwprintf.c} (100%)
 rename src/stdio/{fwscanf.c_ => fwscanf.c} (100%)
 delete mode 100644 src/stdio/getc_unlocked.c_
 delete mode 100644 src/stdio/getchar_unlocked.c_
 rename src/stdio/{getwc.c_ => getwc.c} (100%)
 rename src/stdio/{getwchar.c_ => getwchar.c} (100%)
 delete mode 100644 src/stdio/open_memstream.c_
 delete mode 100644 src/stdio/open_wmemstream.c_
 delete mode 100644 src/stdio/putc_unlocked.c_
 delete mode 100644 src/stdio/putchar_unlocked.c_
 rename src/stdio/{putwc.c_ => putwc.c} (100%)
 rename src/stdio/{putwchar.c_ => putwchar.c} (100%)
 delete mode 100644 src/stdio/setbuffer.c_
 delete mode 100644 src/stdio/setlinebuf.c_
 rename src/stdio/{swprintf.c_ => swprintf.c} (100%)
 rename src/stdio/{swscanf.c_ => swscanf.c} (100%)
 delete mode 100644 src/stdio/tempnam.c_
 rename src/stdio/{ungetwc.c_ => ungetwc.c} (100%)
 delete mode 100644 src/stdio/vasprintf.c_
 delete mode 100644 src/stdio/vdprintf.c_
 rename src/stdio/{vfwprintf.c_ => vfwprintf.c} (100%)
 rename src/stdio/{vfwscanf.c_ => vfwscanf.c} (100%)
 rename src/stdio/{vswprintf.c_ => vswprintf.c} (100%)
 rename src/stdio/{vswscanf.c_ => vswscanf.c} (100%)
 rename src/stdio/{vwprintf.c_ => vwprintf.c} (100%)
 rename src/stdio/{vwscanf.c_ => vwscanf.c} (100%)
 rename src/stdio/{wprintf.c_ => wprintf.c} (100%)
 rename src/stdio/{wscanf.c_ => wscanf.c} (100%)

diff --git a/src/arch/generic/bits/errno.h b/src/arch/generic/bits/errno.h
index fb7901d..f6b8e0a 100644
--- a/src/arch/generic/bits/errno.h
+++ b/src/arch/generic/bits/errno.h
@@ -4,4 +4,5 @@
 #define EILSEQ          3 /* Illegal byte sequence */
 
 #define EINVAL          4 /* Invalid value */
-#define EOVERFLOW       5 /* Overflow */
\ No newline at end of file
+#define EOVERFLOW       5 /* Overflow */
+#define EAGAIN          6 /* Error Again */
\ No newline at end of file
diff --git a/src/ctype/isascii.c b/src/ctype/isascii.c
new file mode 100644
index 0000000..54ad3bf
--- /dev/null
+++ b/src/ctype/isascii.c
@@ -0,0 +1,7 @@
+#include <ctype.h>
+#undef isascii
+
+int isascii(int c)
+{
+	return !(c&~0x7f);
+}
diff --git a/src/ctype/toascii.c b/src/ctype/toascii.c
new file mode 100644
index 0000000..f0e48e8
--- /dev/null
+++ b/src/ctype/toascii.c
@@ -0,0 +1,7 @@
+#include <ctype.h>
+
+/* nonsense function that should NEVER be used! */
+int toascii(int c)
+{
+	return c & 0x7f;
+}
diff --git a/src/env/__libc_start_main.c b/src/env/__libc_start_main.c
index 0c9b4ed..236e4eb 100644
--- a/src/env/__libc_start_main.c
+++ b/src/env/__libc_start_main.c
@@ -15,6 +15,7 @@ void __init_libc(char * envline, char ** envp)
     libc.environ = envp;
 	libc.secure = 1;
     NX_MemSet(&libc.global_locale, 0, sizeof(libc.global_locale));
+    libc.locale = &libc.global_locale;
     __init_stdio();
 }
 
diff --git a/src/include/ctype.h b/src/include/ctype.h
index 51631c9..cff55a3 100644
--- a/src/include/ctype.h
+++ b/src/include/ctype.h
@@ -22,6 +22,12 @@ int   isxdigit(int);
 int   tolower(int);
 int   toupper(int);
 
+int   isascii(int);
+int   toascii(int);
+#define _tolower(a) ((a)|0x20)
+#define _toupper(a) ((a)&0x5f)
+#define isascii(a) (0 ? isascii(a) : (unsigned)(a) < 128)
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/internal/libc.h b/src/internal/libc.h
index 21adc89..71b8044 100644
--- a/src/internal/libc.h
+++ b/src/internal/libc.h
@@ -14,6 +14,7 @@ struct __locale_struct {
 struct __libc {
 	size_t page_size;
 	struct __locale_struct global_locale;
+	struct __locale_struct * locale;
     char * envline; /* nxos envline */
     char **environ; /* unix environ */
 	char secure;
diff --git a/src/internal/locale_impl.h b/src/internal/locale_impl.h
index 0b26439..d95fea3 100644
--- a/src/internal/locale_impl.h
+++ b/src/internal/locale_impl.h
@@ -39,7 +39,7 @@ hidden char *__gettextdomain(void);
 #define C_LOCALE ((locale_t)&__c_locale)
 #define UTF8_LOCALE ((locale_t)&__c_dot_utf8_locale)
 
-#define CURRENT_LOCALE (&libc.global_locale)
+#define CURRENT_LOCALE (libc.locale)
 
 #define CURRENT_UTF8 (!!libc.global_locale.cat[LC_CTYPE])
 
diff --git a/src/stdio/asprintf.c_ b/src/stdio/asprintf.c_
deleted file mode 100644
index 4ec8353..0000000
--- a/src/stdio/asprintf.c_
+++ /dev/null
@@ -1,13 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-#include <stdarg.h>
-
-int asprintf(char **s, const char *fmt, ...)
-{
-	int ret;
-	va_list ap;
-	va_start(ap, fmt);
-	ret = vasprintf(s, fmt, ap);
-	va_end(ap);
-	return ret;
-}
diff --git a/src/stdio/dprintf.c_ b/src/stdio/dprintf.c_
deleted file mode 100644
index 93082ee..0000000
--- a/src/stdio/dprintf.c_
+++ /dev/null
@@ -1,12 +0,0 @@
-#include <stdio.h>
-#include <stdarg.h>
-
-int dprintf(int fd, const char *restrict fmt, ...)
-{
-	int ret;
-	va_list ap;
-	va_start(ap, fmt);
-	ret = vdprintf(fd, fmt, ap);
-	va_end(ap);
-	return ret;
-}
diff --git a/src/stdio/ext.c_ b/src/stdio/ext.c_
deleted file mode 100644
index 1fd9549..0000000
--- a/src/stdio/ext.c_
+++ /dev/null
@@ -1,57 +0,0 @@
-#define _GNU_SOURCE
-#include "stdio_impl.h"
-#include <stdio_ext.h>
-
-void _flushlbf(void)
-{
-	fflush(0);
-}
-
-int __fsetlocking(FILE *f, int type)
-{
-	return 0;
-}
-
-int __fwriting(FILE *f)
-{
-	return (f->flags & F_NORD) || f->wend;
-}
-
-int __freading(FILE *f)
-{
-	return (f->flags & F_NOWR) || f->rend;
-}
-
-int __freadable(FILE *f)
-{
-	return !(f->flags & F_NORD);
-}
-
-int __fwritable(FILE *f)
-{
-	return !(f->flags & F_NOWR);
-}
-
-int __flbf(FILE *f)
-{
-	return f->lbf >= 0;
-}
-
-size_t __fbufsize(FILE *f)
-{
-	return f->buf_size;
-}
-
-size_t __fpending(FILE *f)
-{
-	return f->wend ? f->wpos - f->wbase : 0;
-}
-
-int __fpurge(FILE *f)
-{
-	f->wpos = f->wbase = f->wend = 0;
-	f->rpos = f->rend = 0;
-	return 0;
-}
-
-weak_alias(__fpurge, fpurge);
diff --git a/src/stdio/ext2.c_ b/src/stdio/ext2.c_
deleted file mode 100644
index 3416278..0000000
--- a/src/stdio/ext2.c_
+++ /dev/null
@@ -1,24 +0,0 @@
-#include "stdio_impl.h"
-#include <stdio_ext.h>
-
-size_t __freadahead(FILE *f)
-{
-	return f->rend ? f->rend - f->rpos : 0;
-}
-
-const char *__freadptr(FILE *f, size_t *sizep)
-{
-	if (f->rpos == f->rend) return 0;
-	*sizep = f->rend - f->rpos;
-	return (const char *)f->rpos;
-}
-
-void __freadptrinc(FILE *f, size_t inc)
-{
-	f->rpos += inc;
-}
-
-void __fseterr(FILE *f)
-{
-	f->flags |= F_ERR;
-}
diff --git a/src/stdio/fgetln.c_ b/src/stdio/fgetln.c_
deleted file mode 100644
index 5748435..0000000
--- a/src/stdio/fgetln.c_
+++ /dev/null
@@ -1,21 +0,0 @@
-#define _GNU_SOURCE
-#include "stdio_impl.h"
-#include <string.h>
-
-char *fgetln(FILE *f, size_t *plen)
-{
-	char *ret = 0, *z;
-	ssize_t l;
-	FLOCK(f);
-	ungetc(getc_unlocked(f), f);
-	if (f->rend && (z=memchr(f->rpos, '\n', f->rend - f->rpos))) {
-		ret = (char *)f->rpos;
-		*plen = ++z - ret;
-		f->rpos = (void *)z;
-	} else if ((l = getline(&f->getln_buf, (size_t[]){0}, f)) > 0) {
-		*plen = l;
-		ret = f->getln_buf;
-	}
-	FUNLOCK(f);
-	return ret;
-}
diff --git a/src/stdio/fgetwc.c_ b/src/stdio/fgetwc.c
similarity index 100%
rename from src/stdio/fgetwc.c_
rename to src/stdio/fgetwc.c
diff --git a/src/stdio/fgetws.c_ b/src/stdio/fgetws.c
similarity index 100%
rename from src/stdio/fgetws.c_
rename to src/stdio/fgetws.c
diff --git a/src/stdio/fileno.c_ b/src/stdio/fileno.c_
deleted file mode 100644
index 0bd0e98..0000000
--- a/src/stdio/fileno.c_
+++ /dev/null
@@ -1,16 +0,0 @@
-#include "stdio_impl.h"
-#include <errno.h>
-
-int fileno(FILE *f)
-{
-	FLOCK(f);
-	int fd = f->fd;
-	FUNLOCK(f);
-	if (fd < 0) {
-		errno = EBADF;
-		return -1;
-	}
-	return fd;
-}
-
-weak_alias(fileno, fileno_unlocked);
diff --git a/src/stdio/flockfile.c_ b/src/stdio/flockfile.c_
deleted file mode 100644
index 8e22065..0000000
--- a/src/stdio/flockfile.c_
+++ /dev/null
@@ -1,9 +0,0 @@
-#include "stdio_impl.h"
-#include "pthread_impl.h"
-
-void flockfile(FILE *f)
-{
-	if (!ftrylockfile(f)) return;
-	__lockfile(f);
-	__register_locked_file(f, __pthread_self());
-}
diff --git a/src/stdio/fmemopen.c_ b/src/stdio/fmemopen.c_
deleted file mode 100644
index 343e3e3..0000000
--- a/src/stdio/fmemopen.c_
+++ /dev/null
@@ -1,127 +0,0 @@
-#include "stdio_impl.h"
-#include <errno.h>
-#include <string.h>
-#include <stdlib.h>
-#include <stddef.h>
-#include <inttypes.h>
-#include "libc.h"
-
-struct cookie {
-	size_t pos, len, size;
-	unsigned char *buf;
-	int mode;
-};
-
-struct mem_FILE {
-	FILE f;
-	struct cookie c;
-	unsigned char buf[UNGET+BUFSIZ], buf2[];
-};
-
-static off_t mseek(FILE *f, off_t off, int whence)
-{
-	ssize_t base;
-	struct cookie *c = f->cookie;
-	if (whence>2U) {
-fail:
-		errno = EINVAL;
-		return -1;
-	}
-	base = (size_t [3]){0, c->pos, c->len}[whence];
-	if (off < -base || off > (ssize_t)c->size-base) goto fail;
-	return c->pos = base+off;
-}
-
-static size_t mread(FILE *f, unsigned char *buf, size_t len)
-{
-	struct cookie *c = f->cookie;
-	size_t rem = c->len - c->pos;
-	if (c->pos > c->len) rem = 0;
-	if (len > rem) {
-		len = rem;
-		f->flags |= F_EOF;
-	}
-	memcpy(buf, c->buf+c->pos, len);
-	c->pos += len;
-	rem -= len;
-	if (rem > f->buf_size) rem = f->buf_size;
-	f->rpos = f->buf;
-	f->rend = f->buf + rem;
-	memcpy(f->rpos, c->buf+c->pos, rem);
-	c->pos += rem;
-	return len;
-}
-
-static size_t mwrite(FILE *f, const unsigned char *buf, size_t len)
-{
-	struct cookie *c = f->cookie;
-	size_t rem;
-	size_t len2 = f->wpos - f->wbase;
-	if (len2) {
-		f->wpos = f->wbase;
-		if (mwrite(f, f->wpos, len2) < len2) return 0;
-	}
-	if (c->mode == 'a') c->pos = c->len;
-	rem = c->size - c->pos;
-	if (len > rem) len = rem;
-	memcpy(c->buf+c->pos, buf, len);
-	c->pos += len;
-	if (c->pos > c->len) {
-		c->len = c->pos;
-		if (c->len < c->size) c->buf[c->len] = 0;
-		else if ((f->flags&F_NORD) && c->size) c->buf[c->size-1] = 0;
-	}
-	return len;
-}
-
-static int mclose(FILE *m)
-{
-	return 0;
-}
-
-FILE *fmemopen(void *restrict buf, size_t size, const char *restrict mode)
-{
-	struct mem_FILE *f;
-	int plus = !!strchr(mode, '+');
-	
-	if (!strchr("rwa", *mode)) {
-		errno = EINVAL;
-		return 0;
-	}
-
-	if (!buf && size > PTRDIFF_MAX) {
-		errno = ENOMEM;
-		return 0;
-	}
-
-	f = malloc(sizeof *f + (buf?0:size));
-	if (!f) return 0;
-	memset(f, 0, offsetof(struct mem_FILE, buf));
-	f->f.cookie = &f->c;
-	f->f.fd = -1;
-	f->f.lbf = EOF;
-	f->f.buf = f->buf + UNGET;
-	f->f.buf_size = sizeof f->buf - UNGET;
-	if (!buf) {
-		buf = f->buf2;
-		memset(buf, 0, size);
-	}
-
-	f->c.buf = buf;
-	f->c.size = size;
-	f->c.mode = *mode;
-	
-	if (!plus) f->f.flags = (*mode == 'r') ? F_NOWR : F_NORD;
-	if (*mode == 'r') f->c.len = size;
-	else if (*mode == 'a') f->c.len = f->c.pos = strnlen(buf, size);
-	else if (plus) *f->c.buf = 0;
-
-	f->f.read = mread;
-	f->f.write = mwrite;
-	f->f.seek = mseek;
-	f->f.close = mclose;
-
-	if (!libc.threaded) f->f.lock = -1;
-
-	return __ofl_add(&f->f);
-}
diff --git a/src/stdio/fopencookie.c_ b/src/stdio/fopencookie.c_
deleted file mode 100644
index da042fe..0000000
--- a/src/stdio/fopencookie.c_
+++ /dev/null
@@ -1,135 +0,0 @@
-#define _GNU_SOURCE
-#include "stdio_impl.h"
-#include <stdlib.h>
-#include <sys/ioctl.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <string.h>
-
-struct fcookie {
-	void *cookie;
-	cookie_io_functions_t iofuncs;
-};
-
-struct cookie_FILE {
-	FILE f;
-	struct fcookie fc;
-	unsigned char buf[UNGET+BUFSIZ];
-};
-
-static size_t cookieread(FILE *f, unsigned char *buf, size_t len)
-{
-	struct fcookie *fc = f->cookie;
-	ssize_t ret = -1;
-	size_t remain = len, readlen = 0;
-	size_t len2 = len - !!f->buf_size;
-
-	if (!fc->iofuncs.read) goto bail;
-
-	if (len2) {
-		ret = fc->iofuncs.read(fc->cookie, (char *) buf, len2);
-		if (ret <= 0) goto bail;
-
-		readlen += ret;
-		remain -= ret;
-	}
-
-	if (!f->buf_size || remain > !!f->buf_size) return readlen;
-
-	f->rpos = f->buf;
-	ret = fc->iofuncs.read(fc->cookie, (char *) f->rpos, f->buf_size);
-	if (ret <= 0) goto bail;
-	f->rend = f->rpos + ret;
-
-	buf[readlen++] = *f->rpos++;
-
-	return readlen;
-
-bail:
-	f->flags |= ret == 0 ? F_EOF : F_ERR;
-	f->rpos = f->rend = f->buf;
-	return readlen;
-}
-
-static size_t cookiewrite(FILE *f, const unsigned char *buf, size_t len)
-{
-	struct fcookie *fc = f->cookie;
-	ssize_t ret;
-	size_t len2 = f->wpos - f->wbase;
-	if (!fc->iofuncs.write) return len;
-	if (len2) {
-		f->wpos = f->wbase;
-		if (cookiewrite(f, f->wpos, len2) < len2) return 0;
-	}
-	ret = fc->iofuncs.write(fc->cookie, (const char *) buf, len);
-	if (ret < 0) {
-		f->wpos = f->wbase = f->wend = 0;
-		f->flags |= F_ERR;
-		return 0;
-	}
-	return ret;
-}
-
-static off_t cookieseek(FILE *f, off_t off, int whence)
-{
-	struct fcookie *fc = f->cookie;
-	int res;
-	if (whence > 2U) {
-		errno = EINVAL;
-		return -1;
-	}
-	if (!fc->iofuncs.seek) {
-		errno = ENOTSUP;
-		return -1;
-	}
-	res = fc->iofuncs.seek(fc->cookie, &off, whence);
-	if (res < 0)
-		return res;
-	return off;
-}
-
-static int cookieclose(FILE *f)
-{
-	struct fcookie *fc = f->cookie;
-	if (fc->iofuncs.close) return fc->iofuncs.close(fc->cookie);
-	return 0;
-}
-
-FILE *fopencookie(void *cookie, const char *mode, cookie_io_functions_t iofuncs)
-{
-	struct cookie_FILE *f;
-
-	/* Check for valid initial mode character */
-	if (!strchr("rwa", *mode)) {
-		errno = EINVAL;
-		return 0;
-	}
-
-	/* Allocate FILE+fcookie+buffer or fail */
-	if (!(f=malloc(sizeof *f))) return 0;
-
-	/* Zero-fill only the struct, not the buffer */
-	memset(&f->f, 0, sizeof f->f);
-
-	/* Impose mode restrictions */
-	if (!strchr(mode, '+')) f->f.flags = (*mode == 'r') ? F_NOWR : F_NORD;
-
-	/* Set up our fcookie */
-	f->fc.cookie = cookie;
-	f->fc.iofuncs = iofuncs;
-
-	f->f.fd = -1;
-	f->f.cookie = &f->fc;
-	f->f.buf = f->buf + UNGET;
-	f->f.buf_size = sizeof f->buf - UNGET;
-	f->f.lbf = EOF;
-
-	/* Initialize op ptrs. No problem if some are unneeded. */
-	f->f.read = cookieread;
-	f->f.write = cookiewrite;
-	f->f.seek = cookieseek;
-	f->f.close = cookieclose;
-
-	/* Add new FILE to open file list */
-	return __ofl_add(&f->f);
-}
diff --git a/src/stdio/fputwc.c_ b/src/stdio/fputwc.c
similarity index 100%
rename from src/stdio/fputwc.c_
rename to src/stdio/fputwc.c
diff --git a/src/stdio/fputws.c_ b/src/stdio/fputws.c
similarity index 100%
rename from src/stdio/fputws.c_
rename to src/stdio/fputws.c
diff --git a/src/stdio/ftrylockfile.c_ b/src/stdio/ftrylockfile.c_
deleted file mode 100644
index 5065058..0000000
--- a/src/stdio/ftrylockfile.c_
+++ /dev/null
@@ -1,46 +0,0 @@
-#include "stdio_impl.h"
-#include "pthread_impl.h"
-#include <limits.h>
-
-void __do_orphaned_stdio_locks()
-{
-	FILE *f;
-	for (f=__pthread_self()->stdio_locks; f; f=f->next_locked)
-		a_store(&f->lock, 0x40000000);
-}
-
-void __unlist_locked_file(FILE *f)
-{
-	if (f->lockcount) {
-		if (f->next_locked) f->next_locked->prev_locked = f->prev_locked;
-		if (f->prev_locked) f->prev_locked->next_locked = f->next_locked;
-		else __pthread_self()->stdio_locks = f->next_locked;
-	}
-}
-
-void __register_locked_file(FILE *f, pthread_t self)
-{
-	f->lockcount = 1;
-	f->prev_locked = 0;
-	f->next_locked = self->stdio_locks;
-	if (f->next_locked) f->next_locked->prev_locked = f;
-	self->stdio_locks = f;
-}
-
-int ftrylockfile(FILE *f)
-{
-	pthread_t self = __pthread_self();
-	int tid = self->tid;
-	int owner = f->lock;
-	if ((owner & ~MAYBE_WAITERS) == tid) {
-		if (f->lockcount == LONG_MAX)
-			return -1;
-		f->lockcount++;
-		return 0;
-	}
-	if (owner < 0) f->lock = owner = 0;
-	if (owner || a_cas(&f->lock, 0, tid))
-		return -1;
-	__register_locked_file(f, self);
-	return 0;
-}
diff --git a/src/stdio/funlockfile.c_ b/src/stdio/funlockfile.c_
deleted file mode 100644
index 44d8b0d..0000000
--- a/src/stdio/funlockfile.c_
+++ /dev/null
@@ -1,13 +0,0 @@
-#include "stdio_impl.h"
-#include "pthread_impl.h"
-
-void funlockfile(FILE *f)
-{
-	if (f->lockcount == 1) {
-		__unlist_locked_file(f);
-		f->lockcount = 0;
-		__unlockfile(f);
-	} else {
-		f->lockcount--;
-	}
-}
diff --git a/src/stdio/fwide.c_ b/src/stdio/fwide.c
similarity index 100%
rename from src/stdio/fwide.c_
rename to src/stdio/fwide.c
diff --git a/src/stdio/fwprintf.c_ b/src/stdio/fwprintf.c
similarity index 100%
rename from src/stdio/fwprintf.c_
rename to src/stdio/fwprintf.c
diff --git a/src/stdio/fwscanf.c_ b/src/stdio/fwscanf.c
similarity index 100%
rename from src/stdio/fwscanf.c_
rename to src/stdio/fwscanf.c
diff --git a/src/stdio/getc_unlocked.c_ b/src/stdio/getc_unlocked.c_
deleted file mode 100644
index b38dad1..0000000
--- a/src/stdio/getc_unlocked.c_
+++ /dev/null
@@ -1,9 +0,0 @@
-#include "stdio_impl.h"
-
-int (getc_unlocked)(FILE *f)
-{
-	return getc_unlocked(f);
-}
-
-weak_alias (getc_unlocked, fgetc_unlocked);
-weak_alias (getc_unlocked, _IO_getc_unlocked);
diff --git a/src/stdio/getchar_unlocked.c_ b/src/stdio/getchar_unlocked.c_
deleted file mode 100644
index 355ac31..0000000
--- a/src/stdio/getchar_unlocked.c_
+++ /dev/null
@@ -1,6 +0,0 @@
-#include "stdio_impl.h"
-
-int getchar_unlocked(void)
-{
-	return getc_unlocked(stdin);
-}
diff --git a/src/stdio/getwc.c_ b/src/stdio/getwc.c
similarity index 100%
rename from src/stdio/getwc.c_
rename to src/stdio/getwc.c
diff --git a/src/stdio/getwchar.c_ b/src/stdio/getwchar.c
similarity index 100%
rename from src/stdio/getwchar.c_
rename to src/stdio/getwchar.c
diff --git a/src/stdio/open_memstream.c_ b/src/stdio/open_memstream.c_
deleted file mode 100644
index 600d277..0000000
--- a/src/stdio/open_memstream.c_
+++ /dev/null
@@ -1,99 +0,0 @@
-#include "stdio_impl.h"
-#include <errno.h>
-#include <limits.h>
-#include <string.h>
-#include <stdlib.h>
-#include "libc.h"
-
-struct cookie {
-	char **bufp;
-	size_t *sizep;
-	size_t pos;
-	char *buf;
-	size_t len;
-	size_t space;
-};
-
-struct ms_FILE {
-	FILE f;
-	struct cookie c;
-	unsigned char buf[BUFSIZ];
-};
-
-static off_t ms_seek(FILE *f, off_t off, int whence)
-{
-	ssize_t base;
-	struct cookie *c = f->cookie;
-	if (whence>2U) {
-fail:
-		errno = EINVAL;
-		return -1;
-	}
-	base = (size_t [3]){0, c->pos, c->len}[whence];
-	if (off < -base || off > SSIZE_MAX-base) goto fail;
-	return c->pos = base+off;
-}
-
-static size_t ms_write(FILE *f, const unsigned char *buf, size_t len)
-{
-	struct cookie *c = f->cookie;
-	size_t len2 = f->wpos - f->wbase;
-	char *newbuf;
-	if (len2) {
-		f->wpos = f->wbase;
-		if (ms_write(f, f->wbase, len2) < len2) return 0;
-	}
-	if (len + c->pos >= c->space) {
-		len2 = 2*c->space+1 | c->pos+len+1;
-		newbuf = realloc(c->buf, len2);
-		if (!newbuf) return 0;
-		*c->bufp = c->buf = newbuf;
-		memset(c->buf + c->space, 0, len2 - c->space);
-		c->space = len2;
-	}
-	memcpy(c->buf+c->pos, buf, len);
-	c->pos += len;
-	if (c->pos >= c->len) c->len = c->pos;
-	*c->sizep = c->pos;
-	return len;
-}
-
-static int ms_close(FILE *f)
-{
-	return 0;
-}
-
-FILE *open_memstream(char **bufp, size_t *sizep)
-{
-	struct ms_FILE *f;
-	char *buf;
-
-	if (!(f=malloc(sizeof *f))) return 0;
-	if (!(buf=malloc(sizeof *buf))) {
-		free(f);
-		return 0;
-	}
-	memset(&f->f, 0, sizeof f->f);
-	memset(&f->c, 0, sizeof f->c);
-	f->f.cookie = &f->c;
-
-	f->c.bufp = bufp;
-	f->c.sizep = sizep;
-	f->c.pos = f->c.len = f->c.space = *sizep = 0;
-	f->c.buf = *bufp = buf;
-	*buf = 0;
-
-	f->f.flags = F_NORD;
-	f->f.fd = -1;
-	f->f.buf = f->buf;
-	f->f.buf_size = sizeof f->buf;
-	f->f.lbf = EOF;
-	f->f.write = ms_write;
-	f->f.seek = ms_seek;
-	f->f.close = ms_close;
-	f->f.mode = -1;
-
-	if (!libc.threaded) f->f.lock = -1;
-
-	return __ofl_add(&f->f);
-}
diff --git a/src/stdio/open_wmemstream.c_ b/src/stdio/open_wmemstream.c_
deleted file mode 100644
index ed1b561..0000000
--- a/src/stdio/open_wmemstream.c_
+++ /dev/null
@@ -1,102 +0,0 @@
-#include "stdio_impl.h"
-#include <wchar.h>
-#include <errno.h>
-#include <limits.h>
-#include <string.h>
-#include <stdlib.h>
-#include "libc.h"
-
-struct cookie {
-	wchar_t **bufp;
-	size_t *sizep;
-	size_t pos;
-	wchar_t *buf;
-	size_t len;
-	size_t space;
-	mbstate_t mbs;
-};
-
-struct wms_FILE {
-	FILE f;
-	struct cookie c;
-	unsigned char buf[1];
-};
-
-static off_t wms_seek(FILE *f, off_t off, int whence)
-{
-	ssize_t base;
-	struct cookie *c = f->cookie;
-	if (whence>2U) {
-fail:
-		errno = EINVAL;
-		return -1;
-	}
-	base = (size_t [3]){0, c->pos, c->len}[whence];
-	if (off < -base || off > SSIZE_MAX/4-base) goto fail;
-	memset(&c->mbs, 0, sizeof c->mbs);
-	return c->pos = base+off;
-}
-
-static size_t wms_write(FILE *f, const unsigned char *buf, size_t len)
-{
-	struct cookie *c = f->cookie;
-	size_t len2;
-	wchar_t *newbuf;
-	if (len + c->pos >= c->space) {
-		len2 = 2*c->space+1 | c->pos+len+1;
-		if (len2 > SSIZE_MAX/4) return 0;
-		newbuf = realloc(c->buf, len2*4);
-		if (!newbuf) return 0;
-		*c->bufp = c->buf = newbuf;
-		memset(c->buf + c->space, 0, 4*(len2 - c->space));
-		c->space = len2;
-	}
-	
-	len2 = mbsnrtowcs(c->buf+c->pos, (void *)&buf, len, c->space-c->pos, &c->mbs);
-	if (len2 == -1) return 0;
-	c->pos += len2;
-	if (c->pos >= c->len) c->len = c->pos;
-	*c->sizep = c->pos;
-	return len;
-}
-
-static int wms_close(FILE *f)
-{
-	return 0;
-}
-
-FILE *open_wmemstream(wchar_t **bufp, size_t *sizep)
-{
-	struct wms_FILE *f;
-	wchar_t *buf;
-
-	if (!(f=malloc(sizeof *f))) return 0;
-	if (!(buf=malloc(sizeof *buf))) {
-		free(f);
-		return 0;
-	}
-	memset(&f->f, 0, sizeof f->f);
-	memset(&f->c, 0, sizeof f->c);
-	f->f.cookie = &f->c;
-
-	f->c.bufp = bufp;
-	f->c.sizep = sizep;
-	f->c.pos = f->c.len = f->c.space = *sizep = 0;
-	f->c.buf = *bufp = buf;
-	*buf = 0;
-
-	f->f.flags = F_NORD;
-	f->f.fd = -1;
-	f->f.buf = f->buf;
-	f->f.buf_size = 0;
-	f->f.lbf = EOF;
-	f->f.write = wms_write;
-	f->f.seek = wms_seek;
-	f->f.close = wms_close;
-
-	if (!libc.threaded) f->f.lock = -1;
-
-	fwide(&f->f, 1);
-
-	return __ofl_add(&f->f);
-}
diff --git a/src/stdio/putc_unlocked.c_ b/src/stdio/putc_unlocked.c_
deleted file mode 100644
index 1007131..0000000
--- a/src/stdio/putc_unlocked.c_
+++ /dev/null
@@ -1,9 +0,0 @@
-#include "stdio_impl.h"
-
-int (putc_unlocked)(int c, FILE *f)
-{
-	return putc_unlocked(c, f);
-}
-
-weak_alias(putc_unlocked, fputc_unlocked);
-weak_alias(putc_unlocked, _IO_putc_unlocked);
diff --git a/src/stdio/putchar_unlocked.c_ b/src/stdio/putchar_unlocked.c_
deleted file mode 100644
index 8b5d060..0000000
--- a/src/stdio/putchar_unlocked.c_
+++ /dev/null
@@ -1,6 +0,0 @@
-#include "stdio_impl.h"
-
-int putchar_unlocked(int c)
-{
-	return putc_unlocked(c, stdout);
-}
diff --git a/src/stdio/putwc.c_ b/src/stdio/putwc.c
similarity index 100%
rename from src/stdio/putwc.c_
rename to src/stdio/putwc.c
diff --git a/src/stdio/putwchar.c_ b/src/stdio/putwchar.c
similarity index 100%
rename from src/stdio/putwchar.c_
rename to src/stdio/putwchar.c
diff --git a/src/stdio/setbuffer.c_ b/src/stdio/setbuffer.c_
deleted file mode 100644
index 71233d2..0000000
--- a/src/stdio/setbuffer.c_
+++ /dev/null
@@ -1,7 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-
-void setbuffer(FILE *f, char *buf, size_t size)
-{
-	setvbuf(f, buf, buf ? _IOFBF : _IONBF, size);
-}
diff --git a/src/stdio/setlinebuf.c_ b/src/stdio/setlinebuf.c_
deleted file mode 100644
index b93c4d6..0000000
--- a/src/stdio/setlinebuf.c_
+++ /dev/null
@@ -1,7 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-
-void setlinebuf(FILE *f)
-{
-	setvbuf(f, 0, _IOLBF, 0);
-}
diff --git a/src/stdio/swprintf.c_ b/src/stdio/swprintf.c
similarity index 100%
rename from src/stdio/swprintf.c_
rename to src/stdio/swprintf.c
diff --git a/src/stdio/swscanf.c_ b/src/stdio/swscanf.c
similarity index 100%
rename from src/stdio/swscanf.c_
rename to src/stdio/swscanf.c
diff --git a/src/stdio/tempnam.c_ b/src/stdio/tempnam.c_
deleted file mode 100644
index 565df6b..0000000
--- a/src/stdio/tempnam.c_
+++ /dev/null
@@ -1,49 +0,0 @@
-#include <stdio.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/stat.h>
-#include <limits.h>
-#include <string.h>
-#include <stdlib.h>
-#include "syscall.h"
-#include "kstat.h"
-
-#define MAXTRIES 100
-
-char *tempnam(const char *dir, const char *pfx)
-{
-	char s[PATH_MAX];
-	size_t l, dl, pl;
-	int try;
-	int r;
-
-	if (!dir) dir = P_tmpdir;
-	if (!pfx) pfx = "temp";
-
-	dl = strlen(dir);
-	pl = strlen(pfx);
-	l = dl + 1 + pl + 1 + 6;
-
-	if (l >= PATH_MAX) {
-		errno = ENAMETOOLONG;
-		return 0;
-	}
-
-	memcpy(s, dir, dl);
-	s[dl] = '/';
-	memcpy(s+dl+1, pfx, pl);
-	s[dl+1+pl] = '_';
-	s[l] = 0;
-
-	for (try=0; try<MAXTRIES; try++) {
-		__randname(s+l-6);
-#ifdef SYS_lstat
-		r = __syscall(SYS_lstat, s, &(struct kstat){0});
-#else
-		r = __syscall(SYS_fstatat, AT_FDCWD, s,
-			&(struct kstat){0}, AT_SYMLINK_NOFOLLOW);
-#endif
-		if (r == -ENOENT) return strdup(s);
-	}
-	return 0;
-}
diff --git a/src/stdio/ungetwc.c_ b/src/stdio/ungetwc.c
similarity index 100%
rename from src/stdio/ungetwc.c_
rename to src/stdio/ungetwc.c
diff --git a/src/stdio/vasprintf.c_ b/src/stdio/vasprintf.c_
deleted file mode 100644
index 08251bc..0000000
--- a/src/stdio/vasprintf.c_
+++ /dev/null
@@ -1,15 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-#include <stdarg.h>
-#include <stdlib.h>
-
-int vasprintf(char **s, const char *fmt, va_list ap)
-{
-	va_list ap2;
-	va_copy(ap2, ap);
-	int l = vsnprintf(0, 0, fmt, ap2);
-	va_end(ap2);
-
-	if (l<0 || !(*s=malloc(l+1U))) return -1;
-	return vsnprintf(*s, l+1U, fmt, ap);
-}
diff --git a/src/stdio/vdprintf.c_ b/src/stdio/vdprintf.c_
deleted file mode 100644
index 3b9c093..0000000
--- a/src/stdio/vdprintf.c_
+++ /dev/null
@@ -1,11 +0,0 @@
-#include "stdio_impl.h"
-
-int vdprintf(int fd, const char *restrict fmt, va_list ap)
-{
-	FILE f = {
-		.fd = fd, .lbf = EOF, .write = __stdio_write,
-		.buf = (void *)fmt, .buf_size = 0,
-		.lock = -1
-	};
-	return vfprintf(&f, fmt, ap);
-}
diff --git a/src/stdio/vfwprintf.c_ b/src/stdio/vfwprintf.c
similarity index 100%
rename from src/stdio/vfwprintf.c_
rename to src/stdio/vfwprintf.c
diff --git a/src/stdio/vfwscanf.c_ b/src/stdio/vfwscanf.c
similarity index 100%
rename from src/stdio/vfwscanf.c_
rename to src/stdio/vfwscanf.c
diff --git a/src/stdio/vswprintf.c_ b/src/stdio/vswprintf.c
similarity index 100%
rename from src/stdio/vswprintf.c_
rename to src/stdio/vswprintf.c
diff --git a/src/stdio/vswscanf.c_ b/src/stdio/vswscanf.c
similarity index 100%
rename from src/stdio/vswscanf.c_
rename to src/stdio/vswscanf.c
diff --git a/src/stdio/vwprintf.c_ b/src/stdio/vwprintf.c
similarity index 100%
rename from src/stdio/vwprintf.c_
rename to src/stdio/vwprintf.c
diff --git a/src/stdio/vwscanf.c_ b/src/stdio/vwscanf.c
similarity index 100%
rename from src/stdio/vwscanf.c_
rename to src/stdio/vwscanf.c
diff --git a/src/stdio/wprintf.c_ b/src/stdio/wprintf.c
similarity index 100%
rename from src/stdio/wprintf.c_
rename to src/stdio/wprintf.c
diff --git a/src/stdio/wscanf.c_ b/src/stdio/wscanf.c
similarity index 100%
rename from src/stdio/wscanf.c_
rename to src/stdio/wscanf.c
-- 
Gitee


From 20d103738b1ed4826c7612d60b4e14d2cf4e2823 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Thu, 8 Sep 2022 01:26:08 +0800
Subject: [PATCH 11/13] fix: clear some compile warning

---
 src/include/wchar.h    |  2 +-
 src/include/wctype.h   |  2 +-
 src/math/__rem_pio2.c  |  2 +-
 src/math/__rem_pio2f.c |  2 +-
 src/math/fma.c         |  2 +-
 src/math/fmaf.c        |  2 +-
 src/math/fmal.c        |  2 +-
 src/math/ilogb.c       |  2 +-
 src/math/ilogbf.c      |  2 +-
 src/math/ilogbl.c      |  4 +--
 src/math/lrint.c       |  2 +-
 src/math/lrintl.c      |  2 +-
 src/math/nearbyint.c   |  2 +-
 src/math/nearbyintf.c  |  2 +-
 src/math/nearbyintl.c  |  2 +-
 src/stdio/getdelim.c_  | 81 ------------------------------------------
 src/stdio/getline.c_   |  6 ----
 src/stdio/getw.c_      |  8 -----
 src/stdio/pclose.c_    | 13 -------
 src/stdio/popen.c_     | 73 -------------------------------------
 src/stdio/putw.c_      |  7 ----
 src/stdio/vfwprintf.c  |  1 +
 src/string/strncpy.c   |  1 +
 23 files changed, 18 insertions(+), 204 deletions(-)
 delete mode 100644 src/stdio/getdelim.c_
 delete mode 100644 src/stdio/getline.c_
 delete mode 100644 src/stdio/getw.c_
 delete mode 100644 src/stdio/pclose.c_
 delete mode 100644 src/stdio/popen.c_
 delete mode 100644 src/stdio/putw.c_

diff --git a/src/include/wchar.h b/src/include/wchar.h
index 83390c6..95e592e 100644
--- a/src/include/wchar.h
+++ b/src/include/wchar.h
@@ -25,7 +25,7 @@ typedef int mbstate_t;
 #endif
 
 #undef WEOF
-#define WEOF 0xffffffffU
+#define WEOF 0xFFFFU
 
 wchar_t *wcscpy (wchar_t *__restrict, const wchar_t *__restrict);
 wchar_t *wcsncpy (wchar_t *__restrict, const wchar_t *__restrict, size_t);
diff --git a/src/include/wctype.h b/src/include/wctype.h
index 2d78ee3..cae21df 100644
--- a/src/include/wctype.h
+++ b/src/include/wctype.h
@@ -12,7 +12,7 @@ extern "C" {
 typedef const int * wctrans_t;
 
 #undef WEOF
-#define WEOF 0xffffffffU
+#define WEOF 0xFFFFU
 
 #undef iswdigit
 
diff --git a/src/math/__rem_pio2.c b/src/math/__rem_pio2.c
index ee43b25..6d6e5ab 100644
--- a/src/math/__rem_pio2.c
+++ b/src/math/__rem_pio2.c
@@ -4,7 +4,7 @@
  * use __rem_pio2_large() for large x
  */
 
-#include <math.h>
+#include "libm.h"
 
 #define EPS DBL_EPSILON
 /*
diff --git a/src/math/__rem_pio2f.c b/src/math/__rem_pio2f.c
index 50fdd09..2db493e 100644
--- a/src/math/__rem_pio2f.c
+++ b/src/math/__rem_pio2f.c
@@ -5,7 +5,7 @@
  * use __rem_pio2_large() for large x
  */
 
-#include <math.h>
+#include "libm.h"
 
 #define EPS DBL_EPSILON
 /*
diff --git a/src/math/fma.c b/src/math/fma.c
index 0c6f90c..4a33a41 100644
--- a/src/math/fma.c
+++ b/src/math/fma.c
@@ -41,7 +41,7 @@ static void mul(uint64_t *hi, uint64_t *lo, uint64_t x, uint64_t y)
 
 double fma(double x, double y, double z)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 
 	/* normalize so top 10bits and last bit are 0 */
 	struct num nx, ny, nz;
diff --git a/src/math/fmaf.c b/src/math/fmaf.c
index 80f5cd8..7d0e343 100644
--- a/src/math/fmaf.c
+++ b/src/math/fmaf.c
@@ -38,7 +38,7 @@
  */
 float fmaf(float x, float y, float z)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	double xy, result;
 	union {double f; uint64_t i;} u;
 	int e;
diff --git a/src/math/fmal.c b/src/math/fmal.c
index 4506aac..4f56e63 100644
--- a/src/math/fmal.c
+++ b/src/math/fmal.c
@@ -164,7 +164,7 @@ static inline struct dd dd_mul(long double a, long double b)
  */
 long double fmal(long double x, long double y, long double z)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	long double xs, ys, zs, adj;
 	struct dd xy, r;
 	int oround;
diff --git a/src/math/ilogb.c b/src/math/ilogb.c
index 64d4015..b7783fe 100644
--- a/src/math/ilogb.c
+++ b/src/math/ilogb.c
@@ -3,7 +3,7 @@
 
 int ilogb(double x)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	union {double f; uint64_t i;} u = {x};
 	uint64_t i = u.i;
 	int e = i>>52 & 0x7ff;
diff --git a/src/math/ilogbf.c b/src/math/ilogbf.c
index e23ba20..4c844a6 100644
--- a/src/math/ilogbf.c
+++ b/src/math/ilogbf.c
@@ -3,7 +3,7 @@
 
 int ilogbf(float x)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	union {float f; uint32_t i;} u = {x};
 	uint32_t i = u.i;
 	int e = i>>23 & 0xff;
diff --git a/src/math/ilogbl.c b/src/math/ilogbl.c
index 7b1a9cf..5151712 100644
--- a/src/math/ilogbl.c
+++ b/src/math/ilogbl.c
@@ -9,7 +9,7 @@ int ilogbl(long double x)
 #elif LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384
 int ilogbl(long double x)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	union ldshape u = {x};
 	uint64_t m = u.i.m;
 	int e = u.i.se & 0x7fff;
@@ -32,7 +32,7 @@ int ilogbl(long double x)
 #elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384
 int ilogbl(long double x)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	union ldshape u = {x};
 	int e = u.i.se & 0x7fff;
 
diff --git a/src/math/lrint.c b/src/math/lrint.c
index ddee7a0..8aee280 100644
--- a/src/math/lrint.c
+++ b/src/math/lrint.c
@@ -40,7 +40,7 @@ __attribute__((noinline))
 #endif
 static long lrint_slow(double x)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	int e;
 
 	e = fetestexcept(FE_INEXACT);
diff --git a/src/math/lrintl.c b/src/math/lrintl.c
index b2a8106..ea8802e 100644
--- a/src/math/lrintl.c
+++ b/src/math/lrintl.c
@@ -18,7 +18,7 @@ raises inexact (with tonearest or upward rounding mode)
 */
 long lrintl(long double x)
 {
-	#pragma STDC FENV_ACCESS ON
+	
 	int e;
 
 	e = fetestexcept(FE_INEXACT);
diff --git a/src/math/nearbyint.c b/src/math/nearbyint.c
index f4e8aac..2a46d41 100644
--- a/src/math/nearbyint.c
+++ b/src/math/nearbyint.c
@@ -6,7 +6,7 @@
 double nearbyint(double x)
 {
 #ifdef FE_INEXACT
-	#pragma STDC FENV_ACCESS ON
+	
 	int e;
 
 	e = fetestexcept(FE_INEXACT);
diff --git a/src/math/nearbyintf.c b/src/math/nearbyintf.c
index 092e9ff..6d39815 100644
--- a/src/math/nearbyintf.c
+++ b/src/math/nearbyintf.c
@@ -4,7 +4,7 @@
 float nearbyintf(float x)
 {
 #ifdef FE_INEXACT
-	#pragma STDC FENV_ACCESS ON
+	
 	int e;
 
 	e = fetestexcept(FE_INEXACT);
diff --git a/src/math/nearbyintl.c b/src/math/nearbyintl.c
index 8285249..337ffc1 100644
--- a/src/math/nearbyintl.c
+++ b/src/math/nearbyintl.c
@@ -11,7 +11,7 @@ long double nearbyintl(long double x)
 long double nearbyintl(long double x)
 {
 #ifdef FE_INEXACT
-	#pragma STDC FENV_ACCESS ON
+	
 	int e;
 
 	e = fetestexcept(FE_INEXACT);
diff --git a/src/stdio/getdelim.c_ b/src/stdio/getdelim.c_
deleted file mode 100644
index d2f5b15..0000000
--- a/src/stdio/getdelim.c_
+++ /dev/null
@@ -1,81 +0,0 @@
-#include "stdio_impl.h"
-#include <string.h>
-#include <stdlib.h>
-#include <inttypes.h>
-#include <errno.h>
-
-ssize_t getdelim(char **restrict s, size_t *restrict n, int delim, FILE *restrict f)
-{
-	char *tmp;
-	unsigned char *z;
-	size_t k;
-	size_t i=0;
-	int c;
-
-	FLOCK(f);
-
-	if (!n || !s) {
-		f->mode |= f->mode-1;
-		f->flags |= F_ERR;
-		FUNLOCK(f);
-		errno = EINVAL;
-		return -1;
-	}
-
-	if (!*s) *n=0;
-
-	for (;;) {
-		if (f->rpos != f->rend) {
-			z = memchr(f->rpos, delim, f->rend - f->rpos);
-			k = z ? z - f->rpos + 1 : f->rend - f->rpos;
-		} else {
-			z = 0;
-			k = 0;
-		}
-		if (i+k >= *n) {
-			size_t m = i+k+2;
-			if (!z && m < SIZE_MAX/4) m += m/2;
-			tmp = realloc(*s, m);
-			if (!tmp) {
-				m = i+k+2;
-				tmp = realloc(*s, m);
-				if (!tmp) {
-					/* Copy as much as fits and ensure no
-					 * pushback remains in the FILE buf. */
-					k = *n-i;
-					memcpy(*s+i, f->rpos, k);
-					f->rpos += k;
-					f->mode |= f->mode-1;
-					f->flags |= F_ERR;
-					FUNLOCK(f);
-					errno = ENOMEM;
-					return -1;
-				}
-			}
-			*s = tmp;
-			*n = m;
-		}
-		memcpy(*s+i, f->rpos, k);
-		f->rpos += k;
-		i += k;
-		if (z) break;
-		if ((c = getc_unlocked(f)) == EOF) {
-			if (!i || !feof(f)) {
-				FUNLOCK(f);
-				return -1;
-			}
-			break;
-		}
-		/* If the byte read by getc won't fit without growing the
-		 * output buffer, push it back for next iteration. */
-		if (i+1 >= *n) *--f->rpos = c;
-		else if (((*s)[i++] = c) == delim) break;
-	}
-	(*s)[i] = 0;
-
-	FUNLOCK(f);
-
-	return i;
-}
-
-weak_alias(getdelim, __getdelim);
diff --git a/src/stdio/getline.c_ b/src/stdio/getline.c_
deleted file mode 100644
index 476d0b0..0000000
--- a/src/stdio/getline.c_
+++ /dev/null
@@ -1,6 +0,0 @@
-#include <stdio.h>
-
-ssize_t getline(char **restrict s, size_t *restrict n, FILE *restrict f)
-{
-	return getdelim(s, n, '\n', f);
-}
diff --git a/src/stdio/getw.c_ b/src/stdio/getw.c_
deleted file mode 100644
index 73d2c0d..0000000
--- a/src/stdio/getw.c_
+++ /dev/null
@@ -1,8 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-
-int getw(FILE *f)
-{
-	int x;
-	return fread(&x, sizeof x, 1, f) ? x : EOF;
-}
diff --git a/src/stdio/pclose.c_ b/src/stdio/pclose.c_
deleted file mode 100644
index 080a426..0000000
--- a/src/stdio/pclose.c_
+++ /dev/null
@@ -1,13 +0,0 @@
-#include "stdio_impl.h"
-#include <errno.h>
-#include <unistd.h>
-
-int pclose(FILE *f)
-{
-	int status, r;
-	pid_t pid = f->pipe_pid;
-	fclose(f);
-	while ((r=__syscall(SYS_wait4, pid, &status, 0, 0)) == -EINTR);
-	if (r<0) return __syscall_ret(r);
-	return status;
-}
diff --git a/src/stdio/popen.c_ b/src/stdio/popen.c_
deleted file mode 100644
index 92cb57e..0000000
--- a/src/stdio/popen.c_
+++ /dev/null
@@ -1,73 +0,0 @@
-#include <fcntl.h>
-#include <unistd.h>
-#include <errno.h>
-#include <string.h>
-#include <spawn.h>
-#include "stdio_impl.h"
-#include "syscall.h"
-
-extern char **__environ;
-
-FILE *popen(const char *cmd, const char *mode)
-{
-	int p[2], op, e;
-	pid_t pid;
-	FILE *f;
-	posix_spawn_file_actions_t fa;
-
-	if (*mode == 'r') {
-		op = 0;
-	} else if (*mode == 'w') {
-		op = 1;
-	} else {
-		errno = EINVAL;
-		return 0;
-	}
-	
-	if (pipe2(p, O_CLOEXEC)) return NULL;
-	f = fdopen(p[op], mode);
-	if (!f) {
-		__syscall(SYS_close, p[0]);
-		__syscall(SYS_close, p[1]);
-		return NULL;
-	}
-	FLOCK(f);
-
-	/* If the child's end of the pipe happens to already be on the final
-	 * fd number to which it will be assigned (either 0 or 1), it must
-	 * be moved to a different fd. Otherwise, there is no safe way to
-	 * remove the close-on-exec flag in the child without also creating
-	 * a file descriptor leak race condition in the parent. */
-	if (p[1-op] == 1-op) {
-		int tmp = fcntl(1-op, F_DUPFD_CLOEXEC, 0);
-		if (tmp < 0) {
-			e = errno;
-			goto fail;
-		}
-		__syscall(SYS_close, p[1-op]);
-		p[1-op] = tmp;
-	}
-
-	e = ENOMEM;
-	if (!posix_spawn_file_actions_init(&fa)) {
-		if (!posix_spawn_file_actions_adddup2(&fa, p[1-op], 1-op)) {
-			if (!(e = posix_spawn(&pid, "/bin/sh", &fa, 0,
-			    (char *[]){ "sh", "-c", (char *)cmd, 0 }, __environ))) {
-				posix_spawn_file_actions_destroy(&fa);
-				f->pipe_pid = pid;
-				if (!strchr(mode, 'e'))
-					fcntl(p[op], F_SETFD, 0);
-				__syscall(SYS_close, p[1-op]);
-				FUNLOCK(f);
-				return f;
-			}
-		}
-		posix_spawn_file_actions_destroy(&fa);
-	}
-fail:
-	fclose(f);
-	__syscall(SYS_close, p[1-op]);
-
-	errno = e;
-	return 0;
-}
diff --git a/src/stdio/putw.c_ b/src/stdio/putw.c_
deleted file mode 100644
index 0ff9d7f..0000000
--- a/src/stdio/putw.c_
+++ /dev/null
@@ -1,7 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-
-int putw(int x, FILE *f)
-{
-	return (int)fwrite(&x, sizeof x, 1, f)-1;
-}
diff --git a/src/stdio/vfwprintf.c b/src/stdio/vfwprintf.c
index 85b036c..e1934cc 100644
--- a/src/stdio/vfwprintf.c
+++ b/src/stdio/vfwprintf.c
@@ -7,6 +7,7 @@
 #include <stddef.h>
 #include <stdlib.h>
 #include <wchar.h>
+#include <wctype.h>
 #include <inttypes.h>
 
 /* Convenient bit representation for modifier flags, which all fall
diff --git a/src/string/strncpy.c b/src/string/strncpy.c
index 545892e..cdffa11 100644
--- a/src/string/strncpy.c
+++ b/src/string/strncpy.c
@@ -1,4 +1,5 @@
 #include <string.h>
+#include "string_impl.h"
 
 char *strncpy(char *restrict d, const char *restrict s, size_t n)
 {
-- 
Gitee


From 5ad713e00dcaa4fb41c58f309399089473107063 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Thu, 8 Sep 2022 01:33:05 +0800
Subject: [PATCH 12/13] refactor: no compiler warning in c code

---
 src/Makefile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Makefile b/src/Makefile
index e1ed3ef..e813b8e 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -21,9 +21,9 @@ $(error unsupported arch!)
 endif
 
 X_CFLAGS	:= $(MCFLAGS) -fno-builtin -fno-stack-protector
-X_CFLAGS	+= -nostdlib -nostdinc -Wall -O3 -ffunction-sections -fdata-sections -ffreestanding -std=gnu99
+X_CFLAGS	+= -nostdlib -nostdinc -w -O3 -ffunction-sections -fdata-sections -ffreestanding -std=gnu99
 
-X_ACFLAGS	+= $(MCFLAGS) -Wall -O3 -ffunction-sections -fdata-sections -ffreestanding -std=gnu99
+X_ACFLAGS	+= $(MCFLAGS) -O3 -ffunction-sections -fdata-sections -ffreestanding -std=gnu99
 
 X_LDFLAGS 	:= -no-pie -nostartfile -n
 X_LDFLAGS 	+= -nostdlib
-- 
Gitee


From b89cd675a13f202c6d8a7ac8cff5fcd0893cf3d5 Mon Sep 17 00:00:00 2001
From: hzc1998 <2323168280@qq.com>
Date: Thu, 8 Sep 2022 02:49:46 +0800
Subject: [PATCH 13/13] fix(riscv): type no  off_t

---
 src/arch/riscv64/include/bits/types.h | 3 +++
 src/include/stdio.h                   | 2 --
 src/stdio/__stdio_close.c             | 1 +
 3 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/src/arch/riscv64/include/bits/types.h b/src/arch/riscv64/include/bits/types.h
index eab49eb..dbb3441 100644
--- a/src/arch/riscv64/include/bits/types.h
+++ b/src/arch/riscv64/include/bits/types.h
@@ -20,6 +20,9 @@ typedef unsigned long long		uintptr_t;
 typedef unsigned long long		size_t;
 typedef signed long	long		ssize_t;
 
+typedef signed int				off_t;
+typedef signed long long		loff_t;
+
 typedef signed int				bool_t;
 
 typedef int wchar_t;   
diff --git a/src/include/stdio.h b/src/include/stdio.h
index 1a4aff8..abdf893 100644
--- a/src/include/stdio.h
+++ b/src/include/stdio.h
@@ -3,8 +3,6 @@
 
 #include <features.h>
 
-#include <nxbase.h> // NX_Printf
-
 #include <bits/types.h>
 #include <stdarg.h>
 
diff --git a/src/stdio/__stdio_close.c b/src/stdio/__stdio_close.c
index 6958367..44502b8 100644
--- a/src/stdio/__stdio_close.c
+++ b/src/stdio/__stdio_close.c
@@ -1,4 +1,5 @@
 #include "stdio_impl.h"
+#include <nxbase.h>
 
 static int dummy(int fd)
 {
-- 
Gitee