/third_party/openh264/codec/common/inc/ |
D | loongson_intrinsics.h | 81 static inline __m128i __lsx_vdp2add_h_b(__m128i in_c, __m128i in_h, __m128i in_l) in __lsx_vdp2add_h_b() argument 85 out = __lsx_vmaddwev_h_b(in_c, in_h, in_l); in __lsx_vdp2add_h_b() 86 out = __lsx_vmaddwod_h_b(out, in_h, in_l); in __lsx_vdp2add_h_b() 107 static inline __m128i __lsx_vdp2add_h_bu(__m128i in_c, __m128i in_h, __m128i in_l) in __lsx_vdp2add_h_bu() argument 111 out = __lsx_vmaddwev_h_bu(in_c, in_h, in_l); in __lsx_vdp2add_h_bu() 112 out = __lsx_vmaddwod_h_bu(out, in_h, in_l); in __lsx_vdp2add_h_bu() 133 static inline __m128i __lsx_vdp2add_w_h(__m128i in_c, __m128i in_h, __m128i in_l) in __lsx_vdp2add_w_h() argument 137 out = __lsx_vmaddwev_w_h(in_c, in_h, in_l); in __lsx_vdp2add_w_h() 138 out = __lsx_vmaddwod_w_h(out, in_h, in_l); in __lsx_vdp2add_w_h() 157 static inline __m128i __lsx_vdp2_h_b(__m128i in_h, __m128i in_l) in __lsx_vdp2_h_b() argument [all …]
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/base/ |
D | batch_to_space_base.c | 23 int in_h = in_shape[1]; in BatchToSpaceNoCropForNHWC() local 30 int in_stride_n = in_stride_h * in_h; in BatchToSpaceNoCropForNHWC() 32 for (int h = 0; h < in_h; ++h) { in BatchToSpaceNoCropForNHWC() 55 int in_h = in_shape[1]; in BatchToSpaceForNHWC() local 60 int h_end = MSMIN((in_h * block_h - crops[1]) / block_h + 1, in_h); in BatchToSpaceForNHWC() 61 int h_valid_end = in_h * block_h - crops[1] - 1; in BatchToSpaceForNHWC() 71 int in_stride_n = in_stride_h * in_h; in BatchToSpaceForNHWC()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/int8/ |
D | batch_to_space_int8.c | 23 int in_h = in_shape[1]; in BatchToSpaceNoCropForNHWCInt8() local 29 size_t in_stride_n = in_stride_h * in_h; in BatchToSpaceNoCropForNHWCInt8() 36 for (int h = 0; h < in_h; ++h) { in BatchToSpaceNoCropForNHWCInt8() 60 int in_h = in_shape[1]; in BatchToSpaceForNHWCInt8() local 65 int h_end = MSMIN((in_h * block_h - crops[1]) / block_h + 1, in_h); in BatchToSpaceForNHWCInt8() 66 int h_valid_end = in_h * block_h - crops[1] - 1; in BatchToSpaceForNHWCInt8() 75 size_t in_stride_n = in_stride_h * in_h; in BatchToSpaceForNHWCInt8()
|
D | resize_int8.c | 22 int ResizeBilinearInt8(const int8_t *input_ptr, int8_t *output_ptr, int batch, int in_h, int in_w, … in ResizeBilinearInt8() argument 27 int in_plane = in_h * in_w; in ResizeBilinearInt8() 70 …inearWithFloatScaleInt8(const int8_t *input_ptr, int8_t *output_ptr, int batch, int in_h, int in_w, in ResizeBilinearWithFloatScaleInt8() argument 76 int in_plane = in_h * in_w; in ResizeBilinearWithFloatScaleInt8() 163 int in_h, in_w, new_height, new_width; in ResizeNearestNeighborInt8Simple() local 164 in_h = input_shape[1]; in ResizeNearestNeighborInt8Simple() 172 ComputeNearestNeighborInt(y, in_h, new_height, align_corners, &input_y); in ResizeNearestNeighborInt8Simple() 203 int32_t in_h, in_w, new_height, new_width; in ResizeNearestNeighborInt8() local 204 in_h = input_shape[1]; in ResizeNearestNeighborInt8() 212 ComputeNearestNeighborInt(y, in_h, new_height, align_corners, &input_y); in ResizeNearestNeighborInt8()
|
D | space_to_batch_int8.c | 56 int in_h = param->input_shape_[1]; in DoSpaceToBatchPaddingNHWCInt8() local 70 int in_b_offset = in_batch * in_h * in_w * channel; in DoSpaceToBatchPaddingNHWCInt8() 78 …if (index_h < pad_t || index_h >= (pad_t + in_h) || index_w < pad_l || index_w >= (pad_l + in_w)) { in DoSpaceToBatchPaddingNHWCInt8()
|
D | resize_int8.h | 30 int ResizeBilinearInt8(const int8_t *input_ptr, int8_t *output_ptr, int batch, int in_h, int in_w, … 33 …inearWithFloatScaleInt8(const int8_t *input_ptr, int8_t *output_ptr, int batch, int in_h, int in_w,
|
D | pooling_int8.c | 30 int in_h = pooling_param->input_h_; in AvgPoolingInt8() local 44 int in_batch_offset = batch * in_h * in_w * channel; in AvgPoolingInt8() 59 …if ((in_h_index + h) < 0 || (in_h_index + h) >= in_h || (in_w_index + w) < 0 || (in_w_index + w) >… in AvgPoolingInt8() 286 int in_h = pooling_param->input_h_; in MaxPoolingInt8() local 299 int in_batch_offset = batch * in_h * in_w * channel; in MaxPoolingInt8() 313 …if ((in_h_index + h) < 0 || (in_h_index + h) >= in_h || (in_w_index + w) < 0 || (in_w_index + w) >… in MaxPoolingInt8() 331 int in_h = pooling_param->input_h_; in MaxPoolingWithQuantInt8() local 346 int in_batch_offset = batch * in_h * in_w * channel; in MaxPoolingWithQuantInt8() 371 if ((in_h_index + h) < 0 || (in_h_index + h) >= in_h || (in_w_index + w) < 0 || in MaxPoolingWithQuantInt8() 407 if ((in_h_index + h) < 0 || (in_h_index + h) >= in_h || (in_w_index + w) < 0 || in MaxPoolingWithQuantInt8()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp16_grad/ |
D | convolution_grad_filter.c | 28 int in_h = conv_param->input_h_; in FilterGrad32Arm() local 39 int x_size = in_h * in_w * in_ch; in FilterGrad32Arm() 63 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in FilterGrad32Arm() 106 int in_h = conv_param->input_h_; in FilterGrad16Arm() local 117 int x_size = in_h * in_w * in_ch; in FilterGrad16Arm() 136 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in FilterGrad16Arm() 164 int in_h = conv_param->input_h_; in FilterGrad8Arm() local 175 int x_size = in_h * in_w * in_ch; in FilterGrad8Arm() 193 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in FilterGrad8Arm() 213 int in_h = conv_param->input_h_; in FilterGrad4Arm() local [all …]
|
D | convolution_grad_input.c | 25 int in_h = conv_param->input_h_; in ConvDwInputGrad16() local 36 int in_size = in_h * in_w * in_ch; in ConvDwInputGrad16() 57 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in ConvDwInputGrad16() 130 int in_h = conv_param->input_h_; in ConvDwInputGrad8() local 141 int in_size = in_h * in_w * in_ch; in ConvDwInputGrad8() 163 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in ConvDwInputGrad8() 213 int in_h = conv_param->input_h_; in ConvDwInputGrad4() local 224 int in_size = in_h * in_w * in_ch; in ConvDwInputGrad4() 246 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in ConvDwInputGrad4() 289 int in_h = conv_param->input_h_; in ConvDwInputGradFp16() local [all …]
|
D | pooling_grad.c | 30 int in_h = pooling_param->input_h_; in AvgPoolingFp16Grad() local 39 float16_t *out = &output_ptr[(ib * in_h * in_w * channel)]; in AvgPoolingFp16Grad() 45 int kh_e = MSMIN(win_h, in_h + over_h); in AvgPoolingFp16Grad() 115 int in_h = pooling_param->input_h_; in MaxPoolingFp16Grad() local 120 float16_t *out = &output_ptr[(ib * in_h * in_w * channel)]; in MaxPoolingFp16Grad() 121 const float16_t *inPtr = &input_ptr[(ib * in_h * in_w * channel)]; in MaxPoolingFp16Grad() 126 int kh_e = MSMIN(win_h, in_h + over_h); in MaxPoolingFp16Grad()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32_grad/ |
D | convolution_grad_filter.c | 26 int in_h = conv_param->input_h_; in FilterGrad16Arm() local 37 int x_size = in_h * in_w * in_ch; in FilterGrad16Arm() 56 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in FilterGrad16Arm() 103 int in_h = conv_param->input_h_; in FilterGrad12Arm() local 114 int x_size = in_h * in_w * in_ch; in FilterGrad12Arm() 133 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in FilterGrad12Arm() 173 int in_h = conv_param->input_h_; in FilterGrad8Arm() local 184 int x_size = in_h * in_w * in_ch; in FilterGrad8Arm() 203 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in FilterGrad8Arm() 232 int in_h = conv_param->input_h_; in FilterGrad4Arm() local [all …]
|
D | pooling_grad.c | 30 int in_h = pooling_param->input_h_; in AvgPoolingGrad() local 39 float *out = output_ptr + ib * in_h * in_w * channel; in AvgPoolingGrad() 45 int kh_e = MSMIN(win_h, in_h + over_h); in AvgPoolingGrad() 113 int in_h = pooling_param->input_h_; in MaxPoolingGrad() local 117 float *out = output_ptr + ib * in_h * in_w * channel; in MaxPoolingGrad() 118 const float *inPtr = input_ptr + ib * in_h * in_w * channel; in MaxPoolingGrad() 123 int kh_e = MSMIN(win_h, in_h + over_h); in MaxPoolingGrad()
|
D | convolution_grad_input.c | 24 int in_h = conv_param->input_h_; in ConvDwInputGrad() local 59 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in ConvDwInputGrad() 91 … if (((unsigned)(input_row) < (unsigned)(in_h)) && ((unsigned)(input_col) < (unsigned)(in_w))) { in ConvDwInputGrad()
|
/third_party/mindspore/mindspore/core/ops/ |
D | avg_pool_3d.cc | 62 auto in_h = in_shape[3]; in GetOutputShape() local 71 …static_cast<int64_t>(std::floor((in_h + pad_list[2] + pad_list[3] - kernel_h + stride_h - 1) / str… in GetOutputShape() 77 if ((out_h - 1) * stride_h >= in_h + pad_list[2]) { in GetOutputShape() 85 …out_h = static_cast<int64_t>(std::floor((in_h + pad_list[2] + pad_list[3] - kernel_h) / stride_h +… in GetOutputShape() 92 void GetPadsByPadding(int64_t in_d, int64_t in_h, int64_t in_w, int64_t kernel_d, int64_t kernel_h,… in GetPadsByPadding() argument 102 int64_t tail_h = in_h % stride_h; in GetPadsByPadding() 135 auto in_h = in_shape[3]; in InferShape() local 144 …GetPadsByPadding(in_d, in_h, in_w, kernel_d, kernel_h, kernel_w, stride_d, stride_h, stride_w, pad… in InferShape()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/ |
D | resize_nearest_neighbor_grad_impl.cu | 46 int in_h; in ResizeNearestNeighborGrad() local 54 in_h = pos_array[2]; in ResizeNearestNeighborGrad() 57 …min((align_corners) ? static_cast<int>(roundf(in_h * h_scale)) : static_cast<int>(floorf(in_h * h_… in ResizeNearestNeighborGrad()
|
/third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp32_grad/ |
D | convolution_grad_input.cc | 92 int in_h = conv_param->input_h_; in Execute() local 119 …ConvDwInputGrad(dy_addr + (i * groups) * m * k, w_addr, dx_addr + (i * groups) * in_h * in_w, star… in Execute() 145 …rolling_col2im_hwc(mat_c, dx_addr + (i * groups) * (in_ch / groups) * in_h * in_w + j * (in_ch / g… in Execute() 149 …dx_addr + (i * groups) * (in_ch / groups) * in_h * in_w + j * (in_ch / groups) + ci * (in_ch / gro… in Execute() 174 int in_h = conv_param->input_h_; in Run() local 178 memset(dx_addr, 0, sizeof(float) * batch * in_ch * in_h * in_w); in Run()
|
D | deconvolution_grad_filter.cc | 88 int in_h = conv_param->input_h_; in Execute() local 109 for (int ci = 0; ci < in_h; ci += chunk) { in Execute() 110 int real_chunk = MSMIN(in_h - ci, chunk); in Execute() 111 … float *mat_a = x_addr + (i * (in_ch * in_h * in_w) + j * (in_ch / groups)) + ci * in_w * in_ch; in Execute()
|
/third_party/mindspore/mindspore/lite/test/ut/src/runtime/kernel/arm/common/ |
D | pack_tests.cc | 67 int in_h = conv_param->input_h_; in TEST_F() local 87 int in_batch_offset = b * in_channel * in_h * in_w; in TEST_F() 130 int in_h = conv_param->input_h_; in TEST_F() local 150 int in_batch_offset = b * in_channel * in_h * in_w; in TEST_F()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32/ |
D | pooling_fp32.c | 28 int in_h = pooling_param->input_h_; in AvgPooling() local 46 const float *src_b_ptr = input_ptr + batch * in_h * in_w * channel; in AvgPooling() 62 int real_win_h_end = MSMIN(win_h, in_h - in_h_index); in AvgPooling() 142 int in_h = pooling_param->input_h_; in MaxPooling() local 161 const float *src_b_ptr = input_ptr + batch * in_h * in_w * channel; in MaxPooling() 177 int real_win_h_end = MSMIN(win_h, in_h - in_h_index); in MaxPooling()
|
D | resize_fp32.c | 74 int in_h = input_shape[1]; in PrepareResizeBilinear() local 81 float actual_y = calculate(h, in_h, new_height); in PrepareResizeBilinear() 82 CalculateCoordinate(actual_y, in_h, y_bottoms + h, y_tops + h, y_bottom_weights + h); in PrepareResizeBilinear() 98 int in_h = input_shape[1]; in PrepareResizeBicubic() local 104 float actual_y = calculate(h, in_h, new_height); in PrepareResizeBicubic() 105 CalculateWeightForBicubic(actual_y, in_h, y_tops + 4 * h, y_weights + 4 * h, cubic_coeff); in PrepareResizeBicubic() 121 int in_h = input_shape[1]; in PrepareCropAndResizeBilinear() local 144 actual_y = start_h * (in_h - 1) + h * (end_h - start_h) * (in_h - 1) / (new_height - 1); in PrepareCropAndResizeBilinear() 146 actual_y = 0.5 * (end_h + start_h) * (in_h - 1); in PrepareCropAndResizeBilinear() 148 CalculateCoordinate(actual_y, in_h, y_bottom + h, y_top + h, y_bottom_weight + h); in PrepareCropAndResizeBilinear() [all …]
|
D | space_to_batch_fp32.c | 54 int in_h = (out_h * block_shape_height + shift_h) - padding_top; in DoSpaceToBatch() local 56 … const float *in = input + input_batch * in_stride[0] + in_h * in_stride[1] + in_w * in_stride[2]; in DoSpaceToBatch()
|
/third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp16_grad/ |
D | convolution_fp16_grad_input.cc | 91 int in_h = conv_param->input_h_; in DoExecute() local 142 … dx_addr + (i * groups) * (in_ch / groups) * in_h * in_w + j * (in_ch / groups), in DoExecute() 146 …dx_addr + (i * groups) * (in_ch / groups) * in_h * in_w + j * (in_ch / groups) + ci * (in_ch / gro… in DoExecute() 171 int in_h = conv_param->input_h_; in Run() local 175 std::memset(dx_addr, 0, sizeof(float16_t) * batch * in_ch * in_h * in_w); in Run()
|
/third_party/mindspore/mindspore/core/ops/fusion/ |
D | avg_pool_fusion.cc | 71 auto in_h = in_shape[2]; in InferShape() local 83 out_h = static_cast<int64_t>(ceil((in_h - (kernel_h - 1)) / static_cast<float>(stride_h))); in InferShape() 86 out_h = static_cast<int64_t>(ceil(in_h / static_cast<float>(stride_h))); in InferShape()
|
D | max_pool_fusion.cc | 69 auto in_h = in_shape[2]; in InferShape() local 80 out_h = static_cast<int64_t>(ceil((in_h - (kernel_h - 1)) / static_cast<float>(stride_h))); in InferShape() 83 out_h = static_cast<int64_t>(ceil(in_h / static_cast<float>(stride_h))); in InferShape()
|
/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp16/ |
D | pooling_fp16.c | 27 int in_h = pooling_param->input_h_; in AvgPoolingFp16() local 40 const float16_t *src_b_ptr = input_ptr + batch * in_h * in_w * channel; in AvgPoolingFp16() 56 int real_win_h_end = MSMIN(win_h, in_h - in_h_index); in AvgPoolingFp16() 263 int in_h = pooling_param->input_h_; in MaxPoolingFp16() local 274 int in_batch_offset = batch * in_h * in_w * channel; in MaxPoolingFp16() 287 int real_win_h_end = MSMIN(win_h, in_h - in_h_index); in MaxPoolingFp16()
|