/third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32/ |
D | conv_1x1_x86_fp32.h | 26 …1x1SWKernel)(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 38 …WOWxOCKernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 43 …SW3x32Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 47 …SW1x32Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 51 …SW4x24Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 55 …SW1x24Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 59 …SW6x16Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 63 …SW1x16Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 67 …SW12x8Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 71 …1SW8x8Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, [all …]
|
D | conv_1x1_x86_fp32.c | 102 …SW3x32Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in Conv1x1SW3x32Kernel() argument 372 …: "r"(src), "r"(weight), "r"(bias), "r"(ic_align), "r"(in_sw_step), "r"(act_flag), "r"(oc_align), … in Conv1x1SW3x32Kernel() 378 …SW1x32Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in Conv1x1SW1x32Kernel() argument 519 …: "r"(src), "r"(weight), "r"(bias), "r"(ic_align), "r"(in_sw_step), "r"(act_flag), "r"(oc_align), … in Conv1x1SW1x32Kernel() 525 …SW4x24Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in Conv1x1SW4x24Kernel() argument 797 …src), "r"(weight), "r"(bias), "r"(ic_align), "r"(in_sw_step), "r"(src_3_step), "r"(act_flag), // 6 in Conv1x1SW4x24Kernel() 803 …SW1x24Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in Conv1x1SW1x24Kernel() argument 923 …: "r"(src), "r"(weight), "r"(bias), "r"(ic_align), "r"(in_sw_step), "r"(act_flag), "r"(oc_align), … in Conv1x1SW1x24Kernel() 928 …SW6x16Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in Conv1x1SW6x16Kernel() argument 1212 …src), "r"(weight), "r"(bias), "r"(ic_align), "r"(in_sw_step), "r"(src_3_step), "r"(act_flag), // 6 in Conv1x1SW6x16Kernel() [all …]
|
D | conv_common_fp32.h | 44 size_t out_step, size_t act_flag, size_t real_cal_row); 47 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 62 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 68 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 73 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 78 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 83 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 88 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 93 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 98 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, [all …]
|
D | conv_depthwise_fp32.h | 68 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 72 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 76 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 80 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 84 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 88 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 92 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 96 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 100 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, 117 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin,
|
D | conv_common_fp32.c | 199 const size_t out_step, const size_t act_flag, const size_t real_cal_row) { in CommonConv6x16Kernel() argument 263 if (0x1 & act_flag) { // relu6 in CommonConv6x16Kernel() 277 if (0x2 & act_flag) { // relu in CommonConv6x16Kernel() 434 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t out_step, in SWConv3x32Kernel() argument 587 : "a"(act_flag), "r"(out_step), "r"(dst), "r"(write_mode), "r"(dst_4) in SWConv3x32Kernel() 593 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t out_step, in SWConv1x32Kernel() argument 683 : "a"(act_flag), "r"(out_step), "r"(dst), "r"(write_mode), "r"(dst_4) in SWConv1x32Kernel() 688 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t out_step, in SWConv4x24Kernel() argument 857 : "a"(act_flag), "r"(out_step), "r"(dst), "r"(dst_3), "r"(write_mode) in SWConv4x24Kernel() 863 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t out_step, in SWConv1x24Kernel() argument [all …]
|
D | matmul_fp32.h | 85 …ecMulKernel)(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 89 …ul1x32Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 91 …ul1x24Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 93 …ul1x16Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 95 …Mul1x8Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, 100 …owxColKernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag,
|
D | matmul_fp32.c | 1076 int act_flag = 0; in MatVecMulAvxFp32() local 1078 act_flag += 1; in MatVecMulAvxFp32() 1081 act_flag += 2; in MatVecMulAvxFp32() 1087 …kernel[(col_block >> 3) - 1](c + col_index, a, b + col_index * depth, bias_data, act_flag, 1, col_… in MatVecMulAvxFp32() 1095 …ul1x32Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in MatVecMul1x32Kernel() argument 1206 : "r"(src), "r"(weight), "r"(bias), "r"(deep), "a"(act_flag), "r"(dst) // 5 in MatVecMul1x32Kernel() 1210 …ul1x24Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in MatVecMul1x24Kernel() argument 1311 : "r"(src), "r"(weight), "r"(bias), "r"(deep), "a"(act_flag), "r"(dst) // 5 in MatVecMul1x24Kernel() 1315 …ul1x16Kernel(float *dst, const float *src, const float *weight, const float *bias, size_t act_flag, in MatVecMul1x16Kernel() argument 1401 : "r"(src), "r"(weight), "r"(bias), "r"(deep), "a"(act_flag), "r"(dst) // 5 in MatVecMul1x16Kernel() [all …]
|
D | conv_depthwise_fp32.c | 1138 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, in DepthwiseSWWxKKernel() argument 1186 if (0x1 & act_flag) { // relu6 in DepthwiseSWWxKKernel() 1189 if (0x2 & act_flag) { // relu in DepthwiseSWWxKKernel() 1199 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, in DepthwiseSW3x32Kernel() argument 1338 : "a"(act_flag), "r"(oc_algin), "r"(dst) in DepthwiseSW3x32Kernel() 1344 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, in DepthwiseSW1x32Kernel() argument 1418 : "a"(act_flag), "r"(oc_algin), "r"(dst) in DepthwiseSW1x32Kernel() 1423 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, in DepthwiseSW4x24Kernel() argument 1562 : "a"(act_flag), "r"(oc_algin), "r"(dst), "r"(dst_3) in DepthwiseSW4x24Kernel() 1568 … size_t kernel_w, size_t act_flag, size_t ow_block, size_t oc_block, size_t oc_algin, in DepthwiseSW1x24Kernel() argument [all …]
|