| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp16/ |
| D | group_convolution_fp16.cc | 43 float16_t *src_ptr = reinterpret_cast<float16_t *>(ori_in_data_) + group_id * sub_in_channel; in SeparateInput() local
45 MS_ASSERT(src_ptr); in SeparateInput()
48 memcpy(dst_ptr, src_ptr, sub_in_channel * sizeof(float16_t)); in SeparateInput()
49 src_ptr += ori_in_channel; in SeparateInput()
53 float *src_ptr = reinterpret_cast<float *>(ori_in_data_) + group_id * sub_in_channel; in SeparateInput() local
55 MS_ASSERT(src_ptr); in SeparateInput()
58 memcpy(dst_ptr, src_ptr, sub_in_channel * sizeof(float)); in SeparateInput()
59 src_ptr += ori_in_channel; in SeparateInput()
75 float16_t *src_ptr = sub_out_data; in PostConcat() local
78 memcpy(dst_ptr, src_ptr, sub_out_channel * sizeof(float16_t)); in PostConcat()
[all …]
|
| D | matmul_base_fp16.cc | 162 void MatmulBaseFP16CPUKernel::InitMatrixA(const void *src_ptr) { in InitMatrixA() argument
163 NNACL_CHECK_NULL_RETURN_VOID(src_ptr); in InitMatrixA()
168 …Float32ToFloat16(reinterpret_cast<const float *>(src_ptr), a_pack_ptr_, params_->batch * params_->… in InitMatrixA()
170 memcpy(a_pack_ptr_, src_ptr, params_->batch * params_->deep_ * sizeof(float16_t)); in InitMatrixA()
175 const int8_t *int8_src = reinterpret_cast<const int8_t *>(src_ptr); in InitMatrixA()
196 void MatmulBaseFP16CPUKernel::InitMatrixB(const void *src_ptr, TypeId src_data_type) { in InitMatrixB() argument
197 NNACL_CHECK_NULL_RETURN_VOID(src_ptr); in InitMatrixB()
198 const int8_t *int8_src = reinterpret_cast<const int8_t *>(src_ptr); in InitMatrixB()
203 Float32ToFloat16(reinterpret_cast<const float *>(src_ptr), b_pack_ptr_, in InitMatrixB()
208 …const auto *b_src = reinterpret_cast<const float16_t *>(src_ptr) + i * params_->col_align_ * param… in InitMatrixB()
[all …]
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32/ |
| D | matmul_fp32.h | 43 void RowMajor2ColMajor(const float *src_ptr, float *dst_ptr, int row, int col);
44 void RowMajor2Row4Major(const float *src_ptr, float *dst_ptr, int row, int col);
45 void RowMajor2Row6Major(const float *src_ptr, float *dst_ptr, int row, int col);
46 void RowMajor2Row8Major(const float *src_ptr, float *dst_ptr, int row, int col);
47 void RowMajor2Row12Major(const float *src_ptr, float *dst_ptr, int row, int col);
48 void RowMajor2Row16Major(const float *src_ptr, float *dst_ptr, int row, int col);
49 void RowMajor2Row32Major(const float *src_ptr, float *dst_ptr, int row, int col);
50 void RowMajor2Col4Major(const float *src_ptr, float *dst_ptr, int row, int col);
51 void RowMajor2Col6Major(const float *src_ptr, float *dst_ptr, int row, int col);
52 void RowMajor2Col8Major(const float *src_ptr, float *dst_ptr, int row, int col);
[all …]
|
| D | pack_fp32.c | 570 const float *src_ptr = src_batch + hw * channel + c; in PackNHWCToNCHWFp32() local
573 Transpose8X8Fp32Func_(src_ptr, dst_ptr, channel, plane); in PackNHWCToNCHWFp32()
577 dst_ptr[tc * plane + tr] = src_ptr[tr * channel + tc]; in PackNHWCToNCHWFp32()
583 const float *src_ptr = src_batch + hw * channel + c; in PackNHWCToNCHWFp32() local
586 dst_ptr[i] = src_ptr[i * channel]; in PackNHWCToNCHWFp32()
591 const float *src_ptr = src_batch + hw * channel; in PackNHWCToNCHWFp32() local
594 dst_ptr[i * plane] = src_ptr[i]; in PackNHWCToNCHWFp32()
605 inline void Transpose8X8Fp32Arm64(const float *src_ptr, float *dst_ptr, int src_stride, int dst_str… in Transpose8X8Fp32Arm64() argument
674 …: [ dst_ptr ] "r"(dst_ptr), [ src_ptr ] "r"(src_ptr), [ srcStride ] "r"(srcStride), [ dstStride ] … in Transpose8X8Fp32Arm64()
682 inline void Transpose8X8Fp32Arm32(const float *src_ptr, float *dst_ptr, int src_stride, int dst_str… in Transpose8X8Fp32Arm32() argument
[all …]
|
| D | pack_fp32.h | 56 typedef void (*Transpose8X8Fp32Func)(const float *src_ptr, float *dst_ptr, int src_stride, int dst_…
58 void Transpose8X8Fp32Arm64(const float *src_ptr, float *dst_ptr, int src_stride, int dst_stride);
61 void Transpose8X8Fp32Arm32(const float *src_ptr, float *dst_ptr, int src_stride, int dst_stride);
70 void Transpose8X8Fp32Avx(const float *src_ptr, float *dst_ptr, int src_stride, int dst_stride);
73 void Transpose8X8Fp32Sse(const float *src_ptr, float *dst_ptr, int src_stride, int dst_stride);
|
| D | deconv_winograd_fp32.c | 166 void DeConvWgInputPack(const float *src_ptr, float *dst_ptr, int channel, int stride) { in DeConvWgInputPack() argument
169 const float *src = src_ptr; in DeConvWgInputPack()
224 void DeConvWgMergeArm32(const float *src_ptr, float *dst_ptr, size_t src_step, size_t dst_step) { in DeConvWgMergeArm32() argument
278 …: [ src_ptr ] "r"(src_ptr), [ dst_ptr ] "r"(dst_ptr), [ src_step ] "r"(src_step), [ dst_step ] "r"… in DeConvWgMergeArm32()
283 void DeConvWgMergeArm32(const float *src_ptr, float *dst_ptr, size_t src_step, size_t dst_step) { in DeConvWgMergeArm32() argument
337 …: [ src_ptr ] "r"(src_ptr), [ dst_ptr ] "r"(dst_ptr), [ src_step ] "r"(src_step), [ dst_step ] "r"… in DeConvWgMergeArm32()
345 const float *src_ptr = src; in DeConvWgMerge() local
407 …: [ src_ptr ] "r"(src_ptr), [ dst_ptr ] "r"(dst_ptr), [ src_step ] "r"(src_step), [ dst_step ] "r"… in DeConvWgMerge()
412 DeConvWgMergeArm32(src_ptr, dst_ptr, src_step, dst_step); in DeConvWgMerge()
415 const float *s = src_ptr + j * src_stride; in DeConvWgMerge()
[all …]
|
| D | winograd_transform.c | 135 const float *src_ptr = gemm_out + src_oc4_offset; in WinogradOutputNHWCTransform() local
138 func(src_ptr, dst_ptr, bias_ptr, C8NUM, output_w, output_channel, r_w, r_h, r_c); in WinogradOutputNHWCTransform()
147 const float *src_ptr = gemm_out + src_oc8_offset; in WinogradOutputNHWCTransform() local
150 func(src_ptr, dst_ptr, bias_ptr, C8NUM, output_w, output_channel, r_w, r_h, r_c); in WinogradOutputNHWCTransform()
192 const float *src_ptr = gemm_out + src_oc8_offset; in WinogradOutputNC4HW4Transform() local
195 func(src_ptr, dst_ptr, bias_ptr, C8NUM, output_w, r_c, r_w, r_h, r_c); in WinogradOutputNC4HW4Transform()
205 const float *src_ptr = gemm_out + src_oc4_offset; in WinogradOutputNC4HW4Transform() local
208 func(src_ptr, dst_ptr, bias_ptr, C8NUM, output_w, r_c, r_w, r_h, r_c); in WinogradOutputNC4HW4Transform()
|
| D | matmul_fp32.c | 29 void RowMajor2ColMajor(const float *src_ptr, float *dst_ptr, int row, int col) { in RowMajor2ColMajor() argument
32 dst_ptr[c * row + r] = src_ptr[r * col + c]; in RowMajor2ColMajor()
37 void RowMajor2Row4Major(const float *src_ptr, float *dst_ptr, int row, int col) { in RowMajor2Row4Major() argument
39 const float *src = src_ptr + r * col; in RowMajor2Row4Major()
55 void RowMajor2Row6Major(const float *src_ptr, float *dst_ptr, int row, int col) { in RowMajor2Row6Major() argument
57 const float *src = src_ptr + r * col; in RowMajor2Row6Major()
73 void RowMajor2Row8Major(const float *src_ptr, float *dst_ptr, int row, int col) { in RowMajor2Row8Major() argument
75 const float *src = src_ptr + r * col; in RowMajor2Row8Major()
91 void RowMajor2Row12Major(const float *src_ptr, float *dst_ptr, int row, int col) { in RowMajor2Row12Major() argument
93 const float *src = src_ptr + r * col; in RowMajor2Row12Major()
[all …]
|
| /third_party/mindspore/mindspore/lite/micro/coder/wrapper/fp32/ |
| D | matmul_fp32_wrapper.c | 18 void InitMatrixA(const float *src_ptr, float *dst_ptr, const MatMulParameter *params_, bool is_vect… in InitMatrixA() argument
20 memcpy(dst_ptr, src_ptr, (size_t)(params_->batch * params_->deep_) * sizeof(float)); in InitMatrixA()
24 const float *src = src_ptr + i * params_->deep_ * params_->row_; in InitMatrixA()
34 void InitMatrixB(const float *src_ptr, float *dst_ptr, const MatMulParameter *params_, bool is_vect… in InitMatrixB() argument
37 …memcpy(dst_ptr, src_ptr, (size_t)(params_->batch * params_->col_ * params_->deep_) * sizeof(float)… in InitMatrixB()
40 const float *src = src_ptr + i * params_->deep_ * params_->col_; in InitMatrixB()
48 const float *src = src_ptr + i * params_->deep_ * params_->col_; in InitMatrixB()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/control/ |
| D | tensorlist_getitem.cc | 49 auto src_ptr = input0->GetTensor(index_); in Run() local
50 MS_ASSERT(src_ptr != nullptr); in Run()
51 if (src_ptr->data_type() != kTypeUnknown) { in Run()
52 if (src_ptr->ElementsNum() != out_tensors_.at(0)->ElementsNum()) { in Run()
53 MS_LOG(ERROR) << "src_ptr->ElementsNum():" << src_ptr->ElementsNum() in Run()
57 auto status = lite::Tensor::CopyTensorData(*src_ptr, out_tensors_.at(0)); in Run()
|
| /third_party/flutter/skia/third_party/externals/libjpeg-turbo/ |
| D | transupp.c | 206 JCOEFPTR src_ptr, dst_ptr; in do_flip_h() local
238 src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1]; in do_flip_h()
241 *dst_ptr++ = *src_ptr++; /* copy even column */ in do_flip_h()
242 *dst_ptr++ = - *src_ptr++; /* copy odd column with sign change */ in do_flip_h()
268 JCOEFPTR src_ptr, dst_ptr; in do_flip_v() local
314 src_ptr = src_row_ptr[dst_blk_x]; in do_flip_v()
318 *dst_ptr++ = *src_ptr++; in do_flip_v()
321 *dst_ptr++ = - *src_ptr++; in do_flip_v()
346 JCOEFPTR src_ptr, dst_ptr; in do_transpose() local
372 src_ptr = in do_transpose()
[all …]
|
| /third_party/libjpeg-turbo/ |
| D | transupp.c | 206 JCOEFPTR src_ptr, dst_ptr; in do_flip_h() local
238 src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1]; in do_flip_h()
241 *dst_ptr++ = *src_ptr++; /* copy even column */ in do_flip_h()
242 *dst_ptr++ = - *src_ptr++; /* copy odd column with sign change */ in do_flip_h()
268 JCOEFPTR src_ptr, dst_ptr; in do_flip_v() local
314 src_ptr = src_row_ptr[dst_blk_x]; in do_flip_v()
318 *dst_ptr++ = *src_ptr++; in do_flip_v()
321 *dst_ptr++ = - *src_ptr++; in do_flip_v()
346 JCOEFPTR src_ptr, dst_ptr; in do_transpose() local
372 src_ptr = in do_transpose()
[all …]
|
| /third_party/ffmpeg/libavfilter/ |
| D | vf_deband.c | 143 const uint8_t *src_ptr = (const uint8_t *)in->data[p]; in deband_8_c() local
159 … const int ref0 = src_ptr[av_clip(y + y_pos, 0, h) * src_linesize + av_clip(x + x_pos, 0, w)]; in deband_8_c()
160 … const int ref1 = src_ptr[av_clip(y + -y_pos, 0, h) * src_linesize + av_clip(x + x_pos, 0, w)]; in deband_8_c()
161 … const int ref2 = src_ptr[av_clip(y + -y_pos, 0, h) * src_linesize + av_clip(x + -x_pos, 0, w)]; in deband_8_c()
162 … const int ref3 = src_ptr[av_clip(y + y_pos, 0, h) * src_linesize + av_clip(x + -x_pos, 0, w)]; in deband_8_c()
163 const int src0 = src_ptr[y * src_linesize + x]; in deband_8_c()
202 const uint8_t *src_ptr = (const uint8_t *)in->data[p]; in deband_8_coupling_c() local
207 … const int ref0 = src_ptr[av_clip(y + y_pos, 0, h) * src_linesize + av_clip(x + x_pos, 0, w)]; in deband_8_coupling_c()
208 … const int ref1 = src_ptr[av_clip(y + -y_pos, 0, h) * src_linesize + av_clip(x + x_pos, 0, w)]; in deband_8_coupling_c()
209 … const int ref2 = src_ptr[av_clip(y + -y_pos, 0, h) * src_linesize + av_clip(x + -x_pos, 0, w)]; in deband_8_coupling_c()
[all …]
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32_grad/ |
| D | gemm.c | 29 const float *src_ptr = v1; local
33 dst_ptr[c] += beta * src_ptr[c];
35 src_ptr += stride;
56 static void RowMajor2Row4MajorStride(const float *src_ptr, float *dst_ptr, int row, int col, int le… argument
58 const float *src = src_ptr + r * lead;
68 static void RowMajor2Row8MajorStride(const float *src_ptr, float *dst_ptr, int row, int col, int le… argument
70 const float *src = src_ptr + r * lead;
81 static void RowMajor2Row12MajorStride(const float *src_ptr, float *dst_ptr, int row, int col, int l… argument
83 const float *src = src_ptr + r * lead;
93 static void RowMajor2Col12MajorStride(const float *src_ptr, float *dst_ptr, size_t row, size_t col,… argument
[all …]
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/int8/ |
| D | group_convolution_int8.cc | 29 int8_t *src_ptr = reinterpret_cast<int8_t *>(ori_in_data_) + group_id * sub_in_channel; in SeparateInput() local
32 memcpy(dst_ptr, src_ptr, static_cast<size_t>(sub_in_channel) * sizeof(int8_t)); in SeparateInput()
33 src_ptr += ori_in_channel; in SeparateInput()
45 int8_t *src_ptr = sub_out_data; in PostConcat() local
48 memcpy(dst_ptr, src_ptr, static_cast<size_t>(sub_out_channel) * sizeof(int8_t)); in PostConcat()
49 src_ptr += sub_out_channel; in PostConcat()
|
| /third_party/mindspore/mindspore/lite/src/delegate/npu/ |
| D | transpose_kernel.cc | 34 const float *src_ptr = src_batch + hw * channel + c; in PackNHWCToNCHWFp32() local
36 Transpose8X8Fp32Arm64(src_ptr, dst_ptr, channel, plane); in PackNHWCToNCHWFp32()
40 const float *src_ptr = src_batch + hw * channel + c; in PackNHWCToNCHWFp32() local
43 dst_ptr[i] = src_ptr[i * channel]; in PackNHWCToNCHWFp32()
48 const float *src_ptr = src_batch + hw * channel; in PackNHWCToNCHWFp32() local
51 dst_ptr[i * plane] = src_ptr[i]; in PackNHWCToNCHWFp32()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp32/ |
| D | group_convolution_fp32.cc | 35 float *src_ptr = reinterpret_cast<float *>(ori_in_data_) + group_id * sub_in_channel; in SeparateInput() local
38 memcpy(dst_ptr, src_ptr, sub_in_channel * sizeof(float)); in SeparateInput()
39 src_ptr += ori_in_channel; in SeparateInput()
56 float *src_ptr = sub_out_data; in PostConcat() local
59 memcpy(dst_ptr, src_ptr, sub_out_channel * sizeof(float)); in PostConcat()
60 src_ptr += sub_out_channel; in PostConcat()
|
| /third_party/mindspore/mindspore/ccsrc/minddata/dataset/kernels/image/lite_cv/ |
| D | warp_affine.cc | 132 size_t src_step, const uint8_t *src_ptr, uint8_t *dst_ptr) { in RemapBilinearNotCur1C() argument
136 const uint8_t *t_src_ptr = src_ptr + shy * src_step + shx; in RemapBilinearNotCur1C()
142 size_t src_step, const uint8_t *src_ptr, uint8_t *dst_ptr) { in RemapBilinearNotCur2C() argument
146 const uint8_t *t_src_ptr = src_ptr + shy * src_step + shx * 2; in RemapBilinearNotCur2C()
156 size_t src_step, const uint8_t *src_ptr, uint8_t *dst_ptr) { in RemapBilinearNotCur3C() argument
160 const uint8_t *t_src_ptr = src_ptr + shy * src_step + shx * 3; in RemapBilinearNotCur3C()
173 size_t src_step, const uint8_t *src_ptr, uint8_t *dst_ptr) { in RemapBilinearNotCur4C() argument
177 const uint8_t *t_src_ptr = src_ptr + shy * src_step + shx * 4; in RemapBilinearNotCur4C()
193 … size_t src_step, int cn, const uint8_t *src_ptr, uint8_t *dst_ptr) { in RemapBilinearNotCurMoreC() argument
197 const uint8_t *t_src_ptr = src_ptr + shy * src_step + shx * cn; in RemapBilinearNotCurMoreC()
[all …]
|
| /third_party/skia/third_party/externals/libjpeg-turbo/ |
| D | transupp.c | 482 JCOEFPTR src_ptr, dst_ptr; in do_crop_ext_reflect() local
532 src_ptr = *src_row_ptr++; /* source goes right */ in do_crop_ext_reflect()
535 *dst_ptr++ = *src_ptr++; /* copy even column */ in do_crop_ext_reflect()
536 *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign in do_crop_ext_reflect()
551 src_ptr = *(--src_row_ptr); /* source goes left */ in do_crop_ext_reflect()
554 *dst_ptr++ = *src_ptr++; /* copy even column */ in do_crop_ext_reflect()
555 *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign in do_crop_ext_reflect()
665 JCOEFPTR src_ptr, dst_ptr; in do_reflect() local
687 src_ptr = *(--src_row_ptr); /* source goes left */ in do_reflect()
690 *dst_ptr++ = *src_ptr++; /* copy even column */ in do_reflect()
[all …]
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp16/ |
| D | pack_fp16.c | 217 const float16_t *src_ptr = src_batch + hw * channel + c; in PackNHWCToNCHWFp16() local
222 Transpose16x8ARM64Fp16(src_ptr, dst_ptr, src_stride, dst_stride); in PackNHWCToNCHWFp16()
226 Transpose8x8A32Fp16(src_ptr, dst_ptr, src_stride, dst_stride); in PackNHWCToNCHWFp16()
230 dst_ptr[tc * plane + tr] = src_ptr[tr * channel + tc]; in PackNHWCToNCHWFp16()
236 const float16_t *src_ptr = src_batch + hw * channel + c; in PackNHWCToNCHWFp16() local
239 dst_ptr[i] = src_ptr[i * channel]; in PackNHWCToNCHWFp16()
244 const float16_t *src_ptr = src_batch + hw * channel; in PackNHWCToNCHWFp16() local
247 dst_ptr[i * plane] = src_ptr[i]; in PackNHWCToNCHWFp16()
662 inline void Transpose4x8ARM64Fp16(const float16_t *src_ptr, float16_t *dst_ptr, size_t src_stride, … in Transpose4x8ARM64Fp16() argument
697 : [ dst_ptr ] "r"(dst_ptr), [ src_ptr ] "r"(src_ptr), [ src_stride ] "r"(src_stride), in Transpose4x8ARM64Fp16()
[all …]
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/base/ |
| D | split_with_over_lap_base.c | 30 const char *src_ptr = in_data + start_index * param->inner_stride_ * param->element_bytes_; in DoSplitWithOverlapParallel() local
34 (void)memcpy(dst_ptr + i * out_stride, src_ptr, out_stride); in DoSplitWithOverlapParallel()
35 src_ptr += input_stride; in DoSplitWithOverlapParallel()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/control/ |
| D | tensorlist_setitem_infer.c | 92 TensorC *src_ptr = &input0->tensors_[i]; in TensorListSetItemInferShape() local
93 if (src_ptr == NULL) { in TensorListSetItemInferShape()
98 if (src_ptr->data_type_ != kTypeUnknown) { in TensorListSetItemInferShape()
99 out_shape.shape_[out_shape.size_] = src_ptr->shape_; in TensorListSetItemInferShape()
100 out_shape.shape_size_[out_shape.size_] = (int)(src_ptr->shape_size_); in TensorListSetItemInferShape()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/int8/ |
| D | matmul_int8.h | 31 void RowMajor2Row16x4MajorInt8(const int8_t *src_ptr, int8_t *dst_ptr, int row, int col);
43 void RowMajor2Row8x4MajorInt8(const int8_t *src_ptr, int8_t *dst_ptr, int row, int col);
51 void RowMajor2Row2x16MajorInt8(const int8_t *src_ptr, int8_t *dst_ptr, int row, int col);
52 void RowMajor2Col16x2MajorInt8(const int8_t *src_ptr, int8_t *dst_ptr, int row, int col);
60 void RowMajor2Row4x16MajorInt8(const int8_t *src_ptr, int8_t *dst_ptr, int row, int col);
|
| /third_party/mesa3d/src/gallium/auxiliary/util/ |
| D | u_linear.c | 37 pipe_linear_to_tile(size_t src_stride, const void *src_ptr, in pipe_linear_to_tile() argument
51 ptr = (char*)src_ptr + src_stride * t->rows * y + bytes * x; in pipe_linear_to_tile()
61 void pipe_linear_from_tile(struct pipe_tile_info *t, const void *src_ptr, in pipe_linear_from_tile() argument
67 const char *src_ptr2 = (const char *) src_ptr; in pipe_linear_from_tile()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp16_grad/ |
| D | gemm_fp16.c | 26 static void Row2Col16Block16(const float16_t *src_ptr, float16_t *dst_ptr, size_t col) { in Row2Col16Block16() argument
128 : [ dst_c ] "r"(dst_ptr), [ src_c ] "r"(src_ptr), [ stride ] "r"(stride) in Row2Col16Block16()
136 const float16_t *src_ptr = v1; in AddMatrixFp16() local
146 float16x8_t src_0 = vld1q_f16(src_ptr + c); in AddMatrixFp16()
152 dst_ptr[c] += beta * src_ptr[c]; in AddMatrixFp16()
154 src_ptr += stride; in AddMatrixFp16()
306 const float16_t *src_ptr = src + r * stride; in RowMajor2Row8MajorStrideFp16() local
311 dst[cd8 * C8NUM * row + r * C8NUM + cm8] = src_ptr[c]; in RowMajor2Row8MajorStrideFp16()
|