From 4b4ea012759882f8986aca045ea9b71f5621b379 Mon Sep 17 00:00:00 2001
From: xiaxia3 <wangxiaxia3@huawei.com>
Date: Thu, 20 Oct 2022 19:49:59 +0800
Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E5=A4=8Dindex=E7=AE=97=E5=AD=90?=
 =?UTF-8?q?=E9=87=8D=E5=A4=8D=E7=BC=96=E8=AF=91=E9=97=AE=E9=A2=98=20&&=20v?=
 =?UTF-8?q?iewcopy=E4=BF=AE=E6=94=B9=E5=85=A5=E5=8F=82=E7=BC=96=E8=AF=91?=
 =?UTF-8?q?=E9=80=89=E9=A1=B9?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 torch_npu/csrc/aten/common/CopyKernelNpu.cpp  | 12 ++++++------
 torch_npu/csrc/aten/npu_native_functions.yaml |  2 +-
 torch_npu/csrc/aten/ops/IndexKernelNpu.cpp    | 12 ++++--------
 3 files changed, 11 insertions(+), 15 deletions(-)

diff --git a/torch_npu/csrc/aten/common/CopyKernelNpu.cpp b/torch_npu/csrc/aten/common/CopyKernelNpu.cpp
index a4fac7ab3e..d4133b16fb 100644
--- a/torch_npu/csrc/aten/common/CopyKernelNpu.cpp
+++ b/torch_npu/csrc/aten/common/CopyKernelNpu.cpp
@@ -63,13 +63,13 @@ void copy_kernel_npu(
   OpCommand cmd;
   cmd.Name("ViewCopy")
       .InputWithoutContiguous(self)
-      .Input(self_size)
-      .Input(self_stride)
-      .InputScalarToNPUTensor(at::Scalar(0), at::kLong)
+      .Input(self_size, at::kLong, CompileType::MEMORY_HOST_COMPILE_INDEPENDENT)
+      .Input(self_stride, at::kLong, CompileType::MEMORY_HOST_COMPILE_INDEPENDENT)
+      .Input(at::Scalar(0), at::kLong)
       .InputWithoutContiguous(src)
-      .Input(src_size)
-      .Input(src_stride)
-      .InputScalarToNPUTensor(at::Scalar(0), at::kLong)
+      .Input(src_size, at::kLong, CompileType::MEMORY_HOST_COMPILE_INDEPENDENT)
+      .Input(src_stride, at::kLong, CompileType::MEMORY_HOST_COMPILE_INDEPENDENT)
+      .Input(at::Scalar(0), at::kLong)
       .Output(self)
       .Run();
 
diff --git a/torch_npu/csrc/aten/npu_native_functions.yaml b/torch_npu/csrc/aten/npu_native_functions.yaml
index 1ea5baaf5a..b177754c30 100644
--- a/torch_npu/csrc/aten/npu_native_functions.yaml
+++ b/torch_npu/csrc/aten/npu_native_functions.yaml
@@ -1841,7 +1841,6 @@ custom:
   - func: npu_conv3d.out(Tensor input, Tensor weight, Tensor? bias, int[] stride, int[] padding, int[] dilation, int groups, *, Tensor(a!) out) -> Tensor(a!)
   - func: npu_conv3d_backward(Tensor input, Tensor grad, Tensor weight, int[] stride, int[] padding, int[] dilation, int groups, bool[3] output_mask) -> (Tensor, Tensor, Tensor)
   - func: get_npu_format(Tensor self) -> int
-  - func: npu_format_cast(Tensor self, int acl_format) -> Tensor
   - func: npu_format_cast.Tensor(Tensor self, Tensor dst) -> Tensor
   - func: npu_format_cast_.acl_format(Tensor(a!) self, int acl_format) -> Tensor(a!)
   - func: npu_format_cast_(Tensor(a!) self, Tensor src) -> Tensor(a!)
@@ -1934,6 +1933,7 @@ custom_autograd:
   - func: npu_softmax_cross_entropy_with_logits(Tensor self, Tensor labels) -> Tensor
   - func: npu_max.dim(Tensor self, int dim, bool keepdim=False) -> (Tensor values, Tensor indices)
   - func: npu_max.names_dim(Tensor self, Dimname dim, bool keepdim=False) -> (Tensor values, Tensor indices)
+  - func: npu_format_cast(Tensor self, int acl_format) -> Tensor
   - func: npu_bmmV2(Tensor self, Tensor mat2, int[] output_sizes) -> Tensor
     variants: function, method
   - func: npu_dtype_cast(Tensor self, ScalarType dtype) -> Tensor
diff --git a/torch_npu/csrc/aten/ops/IndexKernelNpu.cpp b/torch_npu/csrc/aten/ops/IndexKernelNpu.cpp
index bf3b1d21ea..06aa1bf93a 100644
--- a/torch_npu/csrc/aten/ops/IndexKernelNpu.cpp
+++ b/torch_npu/csrc/aten/ops/IndexKernelNpu.cpp
@@ -21,14 +21,14 @@ namespace native {
 
 at::Tensor& index_out_nocheck_npu(
     const at::Tensor& self,
-    const at::Tensor& masksTensor,
+    const at::IntArrayRef masks,
     const at::TensorList& indices,
     at::Tensor& result) {
   OpCommand cmd;
   cmd.Name("Index")
       .Input(self)
-      .Input(masksTensor)
-      .Input(result.sizes());
+      .Input(masks, at::kLong, CompileType::MEMORY_HOST_COMPILE_INDEPENDENT)
+      .Input(result.sizes(), at::kLong, CompileType::MEMORY_HOST_COMPILE_INDEPENDENT);
   for (int i = 0; i < indices.size(); i++) {
     std::string name = "indices" + std::to_string(i);
     cmd.Input(indices[i], name);
@@ -61,12 +61,8 @@ at::Tensor NPUNativeFunctions::index(const at::Tensor& self, const torch::List<c
       masks.emplace_back(0);
     }
   }
-
-  at::Tensor masksTensor = CalcuOpUtil::copy_tensor_host_to_device(
-      at::from_blob(masks.data(), {masks.size()}, dtype(at::ScalarType::Long)));
-
   // calculate the output result of the NPU
-  index_out_nocheck_npu(formatCastOfSelf, masksTensor, allDefinedIndices, result);
+  index_out_nocheck_npu(formatCastOfSelf, masks, allDefinedIndices, result);
 
   return result;
 }
-- 
Gitee