realize T_matmul_input_0_input_1<float32>([0, 1908], [0, 114], [0, 16], [0, 16]) {
  3   produce T_matmul_input_0_input_1 {
  4     for (ax0, 0, 1908) {
  5       for (ax1, 0, 114) {
  6         for (ax2, 0, 16) {
  7           for (ax3, 0, 16) {
  8             T_matmul_input_0_input_1(ax0, ax1, ax2, ax3) = 0f
  9             for (ko, 0, 64) {
 10               for (ki, 0, 16) {
 11                 // attr [[iter_var(ko, range(min=0, ext=64)), iter_var(ki, range(min=0, ext=16))]] reduce_update = ""
 12                 T_matmul_input_0_input_1(ax0, ax1, ax2, ax3) = mad(T_matmul_input_0_input_1(ax0, ax1, ax2, ax3), float32((input_0(ko, ax1, ax2, ki)*input_1(ko, ax0, ax3, ki)))):float32:PI
 13               }
 14             }
 15           }
 16         }
 17       }
 18     }
 19   }
 20   // attr [extern(T_transdata_T_matmul_input_0_input_1, 0x12a07d0)] realize_scope = ""
 21   realize T_transdata_T_matmul_input_0_input_1<float32>([0, 1824], [0, 30522]) {
 22     produce T_transdata_T_matmul_input_0_input_1 {
 23       // attr [[buffer(T_matmul_input_0_input_1, 0x129aa70), Tensor(shape=[1908, 114, 16, 16], op.name=T_matmul_input_0_input_1)]] buffer_bind_scope = tvm_tuple(0, 1908, 0, 114, 0, 16, 0, 16):handle:I
 24       // attr [[buffer(T_transdata_T_matmul_input_0_input_1, 0x129e600), Tensor(shape=[1824, 30522], op.name=T_transdata_T_matmul_input_0_input_1)]] buffer_bind_scope = tvm_tuple(0, 1824, 0, 30522):ha    ndle:I
 25       // attr [0] extern_scope = 0
 26       for (i, 0, 1908) {
 27         for (j, 0, 114) {
 28           for (k, 0, 16) {
 29             for (i_0, 0, 16) {
 30               if (((((j*16) + k) < 1824) && (((i*16) + i_0) < 30522))) {
 31                 tensor_store(T_transdata_T_matmul_input_0_input_1, tensor_load(T_matmul_input_0_input_1, i, j, k, i_0):float32:PI, ((j*16) + k), ((i*16) + i_0)):float32:PI
 32               }
 33             }
 34           }
 35         }
 36       }
 37     }
 38   }
 39 }

76         // attr [0] pragma_cube_l0write = 1
 77         for (cc10, 0, 16) {
 78           for (cc11, 0, 16) {
 79             T_matmul_input_0_input_1_local_UB(0, 0, cc10, cc11) = T_matmul_input_0_input_1_local_UB_local_L0C(0, 0, cc10, cc11)
 80           }
 81         }
 82         // attr [0] pragma_fuse_vector = 1
 83         if ((cc0 <= 1906)) {
 84           // attr [placeholder(T_transdata_T_matmul_input_0_input_1_local_UB, 0x146cae0)] realize_scope = "local.UB"
 85           realize T_transdata_T_matmul_input_0_input_1_local_UB<float32>([0, 16], [0, 16]) {
 86             for (cc12, 0, 16) {
 87               for (cc13, 0, 16) {
 88                 T_transdata_T_matmul_input_0_input_1_local_UB(cc12, cc13) = T_matmul_input_0_input_1_local_UB(0, 0, cc12, cc13)
 89               }
 90             }
 91             for (cc10, 0, 16) {
 92               for (cc11, 0, 16) {
 93                 T_transdata_T_matmul_input_0_input_1(((16*cc1) + cc10), ((16*cc0) + cc11)) = T_transdata_T_matmul_input_0_input_1_local_UB(cc10, cc11)
 94               }
 95             }
 96           }
 97         } else {
 98           // attr [placeholder(T_transdata_T_matmul_input_0_input_1_local_UB, 0x1356490)] realize_scope = "local.UB"
 99           realize T_transdata_T_matmul_input_0_input_1_local_UB<float32>([0, 16], [0, 16]) {
100             for (cc12, 0, 16) {
101               for (cc13, 0, 10) {
102                 T_transdata_T_matmul_input_0_input_1_local_UB(cc12, cc13) = T_matmul_input_0_input_1_local_UB(0, 0, cc12, cc13)
103               }
104             }
105             for (cc10, 0, 16) {
106               for (cc11, 0, 10) {
107                 T_transdata_T_matmul_input_0_input_1(((16*cc1) + cc10), (cc11 + 30512)) = T_transdata_T_matmul_input_0_input_1_local_UB(cc10, cc11)
108               }
109             }
110           }
111         }
112       }
113     }
114   }