/dports/math/libxsmm/libxsmm-1.16.3/src/template/ |
H A D | libxsmm_dnn_fullyconnected_st_fwd_ncnc_kcck_generic.tpl.c | 31 int mb2 = 0, ofm2 = 0; variable 77 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 83 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 96 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 119 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 133 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 160 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 166 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 179 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 202 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { [all …]
|
H A D | libxsmm_dnn_fullyconnected_st_fwd_ncnc_kcck_generic_bf16.tpl.c | 36 int mb2 = 0, ofm2 = 0; variable 108 for ( ofm2 = 0; ofm2 < handle->bk; ofm2 += 32 ) { 130 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 155 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 180 for ( ofm2 = 0; ofm2 < handle->bk; ofm2 += 32 ) { 201 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 259 for ( ofm2 = 0; ofm2 < handle->bk; ofm2 += 32 ) { 281 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 306 for ( ofm2 = 0; ofm2 <handle->bk; ++ofm2 ) { 331 for ( ofm2 = 0; ofm2 < handle->bk; ofm2 += 32 ) { [all …]
|
H A D | libxsmm_dnn_fullyconnected_st_bwdupd_custom_generic.tpl.c | 40 int ofm2 = 0; variable 73 for (ofm2 = 0; ofm2 < nOFmBlock; ++ofm2) { 79 …LIBXSMM_VLA_ACCESS(4, filter_tr, ifm1, ofm1, ofm2, ifm2, nBlocksOFm, nOFmBlock, nIFmBlock) = filte… 81 LIBXSMM_VLA_ACCESS(4, filter_tr, ifm1, ofm1, ofm2, ifm2, nBlocksOFm, nOFmBlock, nIFmBlock) = 82 LIBXSMM_VLA_ACCESS(4, filter, ofm1, ifm1, ifm2, ofm2, nBlocksIFm, nIFmBlock, nOFmBlock); 108 for ( ofm2 = 0; ofm2 < nOFmBlock; ++ofm2 ) { /* GEMM K-loop */ 113 …SS(4, filter_tr, ifm1, ofm1, ofm2, ifm2, nBlocksOFm, nOFmBlock, nIFmBlock) * LIBXSMM_VLA_ACCESS(3,… 217 int ifm2, ofm2; variable 222 for ( ofm2 = 0; ofm2 < nOFmBlock; ++ofm2 ) { 229 for ( ofm2 = 0; ofm2 < nOFmBlock; ++ofm2 ) { /* GEMM m-loop */ [all …]
|
H A D | libxsmm_dnn_fullyconnected_st_bwdupd_ncnc_kcck_generic.tpl.c | 48 int ofm1 = 0, mb1 = 0, ofm2 = 0, mb2 = 0; variable 74 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 93 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 94 LIBXSMM_VLA_ACCESS( 2, dbias, ofm1, ofm2, handle->bk ) = 0.0f; 99 for ( ofm2 = 0; ofm2 < handle->bk; ++ofm2 ) { 100 …LIBXSMM_VLA_ACCESS( 2, dbias, ofm1, ofm2, handle->bk ) += LIBXSMM_VLA_ACCESS(4, doutput, mb1, ofm… 166 for (ofm2 = 0; ofm2 < bk; ++ofm2) { 168 LIBXSMM_VLA_ACCESS(4, filter_tr, ifm1, ofm1, ofm2, ifm2, nBlocksOFm, bk, bc) = 169 LIBXSMM_VLA_ACCESS(4, filter, ofm1, ifm1, ifm2, ofm2, nBlocksIFm, bc, bk); 314 ofm2 = (ifm1ofm1 % Cck_work) / Cc_work; [all …]
|
H A D | libxsmm_dnn_fullyconnected_st_fwd_custom_generic.tpl.c | 77 int img2, ifm1, ifm2, ofm2; 82 for ( ofm2 = 0; ofm2 < nOFmBlock; ++ofm2 ) { 83 … LIBXSMM_VLA_ACCESS(3, output, img2, ofm1, ofm2, nBlocksOFm, nOFmBlock) = (element_output_type)0; 90 for ( ofm2 = 0; ofm2 < nOFmBlock; ++ofm2 ) { /* GEMM m-loop */ 91 LIBXSMM_VLA_ACCESS(3, output, img2, ofm1, ofm2, nBlocksOFm, nOFmBlock) += 92 …LIBXSMM_VLA_ACCESS(4, filter, ofm1, ifm1, ifm2, ofm2, nBlocksIFm, nIFmBlock, nOFmBlock) * LIBXSMM_…
|
H A D | libxsmm_dnn_convolve_st_bwd_custom_custom_generic_bf16.tpl.c | 70 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 72 …(7, tr_wt, ifm1, ofm1, handle->desc.R-1-kj , handle->desc.S-1-ki, ofm2/lpb, ifm2, ofm2%lpb, handle… 157 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 168 if (ofm2 == handle->blocksofm && 181 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 192 if (ofm2 == handle->blocksofm && 205 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 216 if (ofm2 == handle->blocksofm && 270 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 334 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { [all …]
|
H A D | libxsmm_dnn_convolve_st_bwd_nhwc_custom-rsck_generic.tpl.c | 11 int img, ofm1, ofm2, ifm1, ifm2, oj, oi, kj, ki, oi_use, oj_use, ii_use, ij_use, ofmb, ifmb, ojb, m… variable 92 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 96 …LIBXSMM_VLA_ACCESS(6, wt, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->desc.R, handl… 100 …LIBXSMM_VLA_ACCESS(6, wt, kj, ki, ifm1, ifm2, ofm1, ofm2, handle->desc.S, handle->blocksifm, handl… 187 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 196 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 205 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 253 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 257 …rs[ind] = &LIBXSMM_VLA_ACCESS(5, output, img, oj_use + kj, oi_use + ki, ofm2, 0, handle->ofhp, h… 303 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { [all …]
|
H A D | libxsmm_dnn_fullyconnected_st_bwdupd_ncnc_kcck_generic_bf16.tpl.c | 24 int mb1ofm1 = 0, mb1 = 0, ofm1 = 0, mb2 = 0, ofm2 = 0; variable 139 for (ofm2 = 0; ofm2 < bk; ofm2++) { 140 …put_tr, ofm1, mb1, mb2/lpb, ofm2, mb2%lpb, nBlocksMB, bn_lp, bk, lpb) = LIBXSMM_VLA_ACCESS(4, dout… 264 for (ofm2 = 0; ofm2 < bk; ++ofm2) { 266 …ter_tr, ifm1, ofm1, ofm2/lpb, ifm2, ofm2%lpb, nBlocksOFm, bk_lp, bc, lpb) = LIBXSMM_VLA_ACCESS(5, … 348 ofm2 = 0; 350 …r_padded, ifm1, ofm1, ofm2/lpb, ifm2, ofm2%lpb, nBlocksOFm, 1, bc, lpb) = LIBXSMM_VLA_ACCESS(5, fi… 358 ofm2 = 0; 478 for (ofm2 = 0; ofm2 < bk; ofm2++) { 557 ofm2 = (ifm1ofm1 % Cck_work) / Cc_work; [all …]
|
H A D | libxsmm_dnn_convolve_st_bwd_custom_custom_generic.tpl.c | 11 int img, ofm1, ofm2, ifm1, ifm2, oj, oi, kj, ki, oi_use, oj_use, ii_use, ij_use, ofmb, ifmb, ojb, m… variable 85 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 88 …LIBXSMM_VLA_ACCESS(6, wt, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->desc.R, handl… 172 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 181 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 183 …B_ptrs[ind] = &LIBXSMM_VLA_ACCESS(5, output, img, ofm2, oj_use + kj, oi_use + ki, 0, handle->blo… 190 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 238 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { 242 …B_ptrs[ind] = &LIBXSMM_VLA_ACCESS(5, output, img, ofm2, oj_use + kj, oi_use + ki, 0, handle->blo… 288 for (ofm2 = ofm1; ofm2 < ofm1 + handle->blocksofm_blocking; ofm2++) { [all …]
|
H A D | libxsmm_dnn_convolve_st_upd_custom_custom_generic_bf16.tpl.c | 17 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2+=32) {\ 18 pixel_0 = _mm512_loadu_si512((element_output_type*)src_out+ofm2);\ 22 _mm512_storeu_si512(tr_out+ofm2*2, ofms_lo);\ 29 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2+=32) {\ 34 _mm512_storeu_si512(tr_out+ofm2*2, ofms_lo);\ 39 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2+=32) {\ 53 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2+=32) {\ 58 _mm512_storeu_si512(tr_out+ofm2*2, ofms_lo);\ 205 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++) { 238 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++) { [all …]
|
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic_i8i32.tpl.c | 12 int img, ofm1, ofm2, ifm1, ifm2, oj, oi, kj, ki, ii_use, ij_use, oii, spread_out = 1; variable 84 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 85 temp_ptr[ofm2] = (element_output_type)0; 130 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 131 temp_ptr[ofm2] = (element_output_type)0; 155 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 156 temp_ptr[ofm2] = (element_output_type)0;
|
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic.tpl.c | 12 int img, ofm1, ofm2 = 0, ifm1, ifm2 = 0, oj, oi, kj, ki, oi_use, oj_use, ii_use, ij_use, ofmb, ifmb… variable 158 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 159 temp_ptr[ofm2] = (element_output_type)0; 235 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 236 temp_ptr[ofm2] = (element_output_type)0; 287 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 288 temp_ptr[ofm2] = (element_output_type)0; 369 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 370 temp_ptr[ofm2] = (element_output_type)0; 428 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { [all …]
|
H A D | libxsmm_dnn_convolve_st_fwd_nhwc_custom-rsck_generic.tpl.c | 12 int img, ofm1, ofm2 = 0, ifm1, ifm2 = 0, oj, oi, kj, ki, oi_use, oj_use, ii_use, ij_use, ofmb, ifmb… variable 164 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 165 temp_ptr[ofm2] = (element_output_type)0; 256 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 257 temp_ptr[ofm2] = (element_output_type)0; 313 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 314 temp_ptr[ofm2] = (element_output_type)0; 410 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 411 temp_ptr[ofm2] = (element_output_type)0; 474 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { [all …]
|
H A D | libxsmm_dnn_convolve_st_bwd_nhwc_custom-rsck_fallback_generic.tpl.c | 12 int imgifm1, img, ofm1, ifm1, oj, ij, oi, ii, kj, ki, ifm2, ofm2, ifm1ofm1; variable 65 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 68 …A_ACCESS(6, tr_wt, ifm1, ofm1, handle->desc.R-1-kj , handle->desc.S-1-ki, ofm2, ifm2, handle->bloc… 69 …LIBXSMM_VLA_ACCESS(6, wt, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->desc.R, handl… 72 …A_ACCESS(6, tr_wt, ifm1, ofm1, handle->desc.R-1-kj , handle->desc.S-1-ki, ofm2, ifm2, handle->bloc… 73 …LIBXSMM_VLA_ACCESS(6, wt, kj, ki, ifm1, ifm2, ofm1, ofm2, handle->desc.S, handle->blocksifm, handl…
|
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic_bf16.tpl.c | 12 int img, ofm1, ofm2 = 0, ifm1, ifm2 = 0, oj, oi, kj, ki, oi_use, oj_use, ii_use, ij_use, ofmb, ifmb… variable 126 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 127 temp_ptr[ofm2] = (float)0; 243 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 244 temp_ptr[ofm2] = (float)0; 307 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 308 temp_ptr[ofm2] = (float)0; 432 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 433 temp_ptr[ofm2] = (float)0; 504 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { [all …]
|
H A D | libxsmm_dnn_convolve_st_bwd_custom_custom_fallback_generic.tpl.c | 12 int imgifm1, img, ofm1, ifm1, oj, ij, oi, ii, kj, ki, ifm2, ofm2, ifm1ofm1; variable 60 for (ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2) { 62 …A_ACCESS(6, tr_wt, ifm1, ofm1, handle->desc.R-1-kj , handle->desc.S-1-ki, ofm2, ifm2, handle->bloc… 63 …LIBXSMM_VLA_ACCESS(6, wt, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->desc.R, handl…
|
H A D | libxsmm_dnn_convolve_st_upd_nhwc_custom-rsck_generic.tpl.c | 12 int img, my_img_start, my_img_end, ofmb, ifmb, ojb, ofm1, ifm1, ifm2 = 0, ofm2 = 0, oj, oi, ii, ij,… variable 224 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++) { 226 …LIBXSMM_VLA_ACCESS(6, weight_global, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->de… 229 …LIBXSMM_VLA_ACCESS(6, weight_global, kj, ki, ifm1, ifm2, ofm1, ofm2, handle->desc.S, handle->block… 468 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++) { 470 …LIBXSMM_VLA_ACCESS(6, weight_current, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->d… 473 …LIBXSMM_VLA_ACCESS(6, weight_current, kj, ki, ifm1, ifm2, ofm1, ofm2, handle->desc.S, handle->bloc… 521 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++ ) { 524 …LIBXSMM_VLA_ACCESS(6, weight_private_group, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, han… 527 …LIBXSMM_VLA_ACCESS(6, weight_private_group, kj, ki, ifm1, ifm2, ofm1, ofm2, handle->desc.S, handle…
|
H A D | libxsmm_dnn_convolve_st_upd_custom_custom_generic.tpl.c | 12 int img, my_img_start, my_img_end, ofmb, ifmb, ojb, ofm1, ifm1, ifm2 = 0, ofm2 = 0, oj, oi, ii, ij,… variable 197 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++) { 198 …LIBXSMM_VLA_ACCESS(6, weight_global, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->de… 400 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++) { 401 …LIBXSMM_VLA_ACCESS(6, weight_current, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->d… 438 for (ofm2 = 0; ofm2 < handle->ofmblock; ofm2++ ) { 440 …LIBXSMM_VLA_ACCESS(6, weight_private_group, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, han…
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/x64/ |
H A D | jit_avx512_core_f32_wino_conv_4x3.cpp | 337 [&](int ofm1, int ifm1, int ofm2, int ifm2) { in _execute_data_W_S_G_D() argument 343 + ofm2, in _execute_data_W_S_G_D() 434 [&](int ofm1, int ifm1, int ofm2, int ifm2) { in _execute_data_W_SGD() argument 440 + ofm2, in _execute_data_W_SGD() 723 for (int ofm2 = 0; ofm2 < jcp.oc_block; ++ofm2) { in _execute_backward_weights_SDGtWo() local 748 for (int ofm2 = 0; ofm2 < jcp.oc_block; ++ofm2) { in _execute_backward_weights_SDGtWo() local 881 [&](int ofm1, int ofm2, int img) { in _execute_backward_weights_S_D_Giot_W() argument 974 for (int ofm2 = 0; ofm2 < blk_sz; ++ofm2) { in _execute_backward_weights_S_D_Giot_W() local 975 pbias[ofm2] = pbias_prv[ofm2]; in _execute_backward_weights_S_D_Giot_W() 981 for (int ofm2 = 0; ofm2 < blk_sz; ++ofm2) { in _execute_backward_weights_S_D_Giot_W() local [all …]
|
H A D | jit_avx512_common_convolution_winograd.cpp | 977 [&](int ofm1, int ifm1, int ofm2, int ifm2) { in _execute_data_W_S_G_D() argument 979 ? &(U(ofm1, 0, 0, ifm1, ofm2, ifm2, 0, 0)) in _execute_data_W_S_G_D() 980 : &(U(ifm1, 0, 0, ofm1, ifm2, ofm2, 0, 0)); in _execute_data_W_S_G_D() 982 &(weights(ofm1 * jcp.oc_block + ofm2, in _execute_data_W_S_G_D() 1114 [&](int img, int ofm1, int ofm2) { in _execute_backward_weights_S_D_G_W() argument 1122 &(M(ofm1, 0, 0, 0, ofm2, 0, 0, 0)), dbias); in _execute_backward_weights_S_D_G_W() 1151 [&](int ifm1, int ofm1, int ofm2, int ifm2) { in _execute_backward_weights_S_D_G_W() argument 1153 &(diff_weights(ofm1 * jcp.oc_block + ofm2, in _execute_backward_weights_S_D_G_W() 1155 &(U(ifm1, ofm1, 0, 0, ofm2, ifm2, 0, 0))); in _execute_backward_weights_S_D_G_W() 1166 for (int ofm2 = 0; ofm2 < simd_w; ofm2++) { in _execute_backward_weights_S_D_G_W() local [all …]
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/ |
H A D | jit_avx512_core_f32_wino_conv_4x3.cpp | 343 + ofm2, in _execute_data_W_S_G_D() 440 + ofm2, in _execute_data_W_SGD() 723 for (int ofm2 = 0; ofm2 < jcp.oc_block; ++ofm2) { in _execute_backward_weights_SDGtWo() local 724 int ofm = (ofm1 * jcp.oc_block + ofm2) in _execute_backward_weights_SDGtWo() 748 for (int ofm2 = 0; ofm2 < jcp.oc_block; ++ofm2) { in _execute_backward_weights_SDGtWo() local 750 int ofm = (ofm1 * jcp.oc_block + ofm2) in _execute_backward_weights_SDGtWo() 882 [&](dim_t ofm1, dim_t ofm2, dim_t img) { in _execute_backward_weights_S_D_Giot_W() argument 976 for (int ofm2 = 0; ofm2 < blk_sz; ++ofm2) { in _execute_backward_weights_S_D_Giot_W() local 977 pbias[ofm2] = pbias_prv[ofm2]; in _execute_backward_weights_S_D_Giot_W() 983 for (int ofm2 = 0; ofm2 < blk_sz; ++ofm2) { in _execute_backward_weights_S_D_Giot_W() local [all …]
|
H A D | jit_avx512_common_convolution_winograd.cpp | 977 [&](dim_t ofm1, dim_t ifm1, dim_t ofm2, dim_t ifm2) { in _execute_data_W_S_G_D() argument 979 ? &(U(ofm1, 0, 0, ifm1, ofm2, ifm2, 0, 0)) in _execute_data_W_S_G_D() 980 : &(U(ifm1, 0, 0, ofm1, ifm2, ofm2, 0, 0)); in _execute_data_W_S_G_D() 982 &(weights(ofm1 * jcp.oc_block + ofm2, in _execute_data_W_S_G_D() 1114 [&](dim_t img, dim_t ofm1, dim_t ofm2) { in _execute_backward_weights_S_D_G_W() argument 1122 &(M(ofm1, 0, 0, 0, ofm2, 0, 0, 0)), dbias); in _execute_backward_weights_S_D_G_W() 1151 [&](dim_t ifm1, dim_t ofm1, dim_t ofm2, dim_t ifm2) { in _execute_backward_weights_S_D_G_W() argument 1153 &(diff_weights(ofm1 * jcp.oc_block + ofm2, in _execute_backward_weights_S_D_G_W() 1155 &(U(ifm1, ofm1, 0, 0, ofm2, ifm2, 0, 0))); in _execute_backward_weights_S_D_G_W() 1166 for (int ofm2 = 0; ofm2 < simd_w; ofm2++) { in _execute_backward_weights_S_D_G_W() local [all …]
|
/dports/math/libxsmm/libxsmm-1.16.3/src/ |
H A D | libxsmm_dnn_convolution.c | 1819 int ifm1, ifm2, kj, ki, ofm1, ofm2; in libxsmm_dnn_trans_reg_bf16_filter() local 1831 for ( ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2 ) { in libxsmm_dnn_trans_reg_bf16_filter() 1833 …(7, tr_wt, ifm1, ofm1, handle->desc.R-1-kj , handle->desc.S-1-ki, ofm2/lpb, ifm2, ofm2%lpb, handle… in libxsmm_dnn_trans_reg_bf16_filter() 1834 …LIBXSMM_VLA_ACCESS(7, wt, ofm1, ifm1, kj, ki, ifm2/lpb, ofm2, ifm2%lpb, handle->blocksifm, handle-… in libxsmm_dnn_trans_reg_bf16_filter() 1857 int ifm1, ifm2, kj, ki, ofm1, ofm2; in libxsmm_dnn_trans_reg_filter() local 1866 for ( ofm2 = 0; ofm2 < handle->ofmblock; ++ofm2 ) { in libxsmm_dnn_trans_reg_filter() 1868 …LA_ACCESS(6, tr_wt, ifm1, ofm1, handle->desc.R-1-kj, handle->desc.S-1-ki, ofm2, ifm2, handle->bloc… in libxsmm_dnn_trans_reg_filter() 1869 …LIBXSMM_VLA_ACCESS(6, wt, ofm1, ifm1, kj, ki, ifm2, ofm2, handle->blocksifm, handle->desc.R, handl… in libxsmm_dnn_trans_reg_filter()
|
/dports/x11-wm/enlightenment/enlightenment-0.24.2/src/bin/ |
H A D | e_test.c | 522 Evas_Object *ofm, *ofm2, *of, *ob, *ot; in _e_test_internal() local 569 ofm2 = ofm; in _e_test_internal() 594 evas_object_smart_callback_add(ofm, "files_changed", _e_test_cb_favorites_files_changed, ofm2); in _e_test_internal() 595 evas_object_smart_callback_add(ofm, "selected", _e_test_cb_favorites_selected, ofm2); in _e_test_internal()
|
/dports/net/ns3/ns-allinone-3.35/ns-3.35/src/openflow/model/ |
H A D | openflow-interface.cc | 886 …ofp_flow_mod* ofm2 = BuildFlow (key, -1, OFPFC_MODIFY, x2, sizeof(x2), OFP_FLOW_PERMANENT, m_expir… in ReceiveFromSwitch() local 887 SendToSwitch (swtch, ofm2, ofm2->header.length); in ReceiveFromSwitch()
|