Searched refs:fwd_ofw_rb (Results 1 – 10 of 10) sorted by relevance
/dports/math/libxsmm/libxsmm-1.16.3/src/ |
H A D | libxsmm_dnn_convolution_forward.c | 39 …dispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in LIBXSMM_INTRINSICS() 40 …ispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*(handle->fwd_ofw_rb-1), handle->ifmb… in LIBXSMM_INTRINSICS() 70 …dispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in LIBXSMM_INTRINSICS() 71 …ispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*(handle->fwd_ofw_rb-1), handle->ifmb… in LIBXSMM_INTRINSICS() 107 …dispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in LIBXSMM_INTRINSICS() 108 …ispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*(handle->fwd_ofw_rb-1), handle->ifmb… in LIBXSMM_INTRINSICS() 184 …dispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in LIBXSMM_INTRINSICS() 185 …ispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*(handle->fwd_ofw_rb-1), handle->ifmb… in LIBXSMM_INTRINSICS() 212 …dispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in LIBXSMM_INTRINSICS() 213 …ispatch_reducebatch_addr(handle->ofmblock, handle->fwd_ofh_rb*(handle->fwd_ofw_rb-1), handle->ifmb… in LIBXSMM_INTRINSICS() [all …]
|
H A D | libxsmm_dnn_convolution.c | 339 handle->fwd_lp_output_block_scratch_size = (size_t)handle->desc.threads * handle->fwd_ofw_rb * in libxsmm_dnn_convolution_setup_fwd_scratch() 918 handle->fwd_ofw_rb = libxsmm_dnn_convolution_setup_fwd_ofw_rb(handle); in libxsmm_dnn_convolution_setup() 941 …kernel = libxsmm_dispatch_meltw_cvtfp32bf16(handle->ofmblock * handle->fwd_ofw_rb, handle->fwd_ofh… in libxsmm_dnn_convolution_setup() 956 …dispatch_reducebatch_strd(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in libxsmm_dnn_convolution_setup() 978 …dispatch_reducebatch_offs(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in libxsmm_dnn_convolution_setup() 982 …dispatch_reducebatch_strd(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in libxsmm_dnn_convolution_setup() 983 …ispatch_reducebatch_strd(handle->ofmblock, handle->fwd_ofh_rb*(handle->fwd_ofw_rb-1), handle->ifmb… in libxsmm_dnn_convolution_setup() 997 …dispatch_reducebatch_strd(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in libxsmm_dnn_convolution_setup() 1018 …dispatch_reducebatch_offs(handle->ofmblock, handle->fwd_ofh_rb*handle->fwd_ofw_rb, handle->ifmbloc… in libxsmm_dnn_convolution_setup() 1024 printf("Fwd_ofw_rb = %d\n", handle->fwd_ofw_rb); in libxsmm_dnn_convolution_setup() [all …]
|
H A D | libxsmm_main.h | 386 int fwd_ofw_rb; member
|
/dports/math/libxsmm/libxsmm-1.16.3/src/template/ |
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic_i8i32.tpl.c | 17 const int w_tasks = handle->ofw/handle->fwd_ofw_rb; 76 oi = (((imgofm1ofhofw % work_KHW)%work_HW)%w_tasks)*handle->fwd_ofw_rb; 82 for (oii = 0; oii < handle->fwd_ofw_rb; ++oii) { 101 } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 122 oi = (((imgofm1ofhofw % work_KHW)%work_HW)%w_tasks)*handle->fwd_ofw_rb; 128 for (oii = 0; oii < handle->fwd_ofw_rb; ++oii) { 147 oi = (((imgofm1ofhofw % work_KHW)%work_HW)%w_tasks)*handle->fwd_ofw_rb; 153 for (oii = 0; oii < handle->fwd_ofw_rb; ++oii) {
|
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic_i8i8.tpl.c | 16 const int w_tasks = handle->ofw/handle->fwd_ofw_rb; 40 oi = (((imgofm1ofhofw % work_KHW)%work_HW)%w_tasks)*handle->fwd_ofw_rb; 52 oi = (((imgofm1ofhofw % work_KHW)%work_HW)%w_tasks)*handle->fwd_ofw_rb;
|
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic_bf16.tpl.c | 34 …handle->fwd_lp_output_block_scratch_offset) + ((size_t) ltid * handle->fwd_ofw_rb * handle->fwd_of… 38 LIBXSMM_VLA_DECL(3, float, scratch_fp32, out_scratch, handle->fwd_ofw_rb, handle->ofmblock); 133 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 178 } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 251 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 279 …cc_load == 1) ? &LIBXSMM_VLA_ACCESS( 3, scratch_fp32, 0, 0, 0, handle->fwd_ofw_rb, handle->ofmbloc… 317 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 364 … } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 442 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 496 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) {
|
H A D | libxsmm_dnn_convolve_st_fwd_custom_custom_generic.tpl.c | 165 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 192 } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 243 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 297 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 327 … } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 379 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 420 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) {
|
H A D | libxsmm_dnn_convolve_st_fwd_nhwc_custom-rsck_generic.tpl.c | 171 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 203 } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 264 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 323 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 358 … } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) { 420 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) { 466 for (oi = 0; oi < handle->ofw; oi += handle->fwd_ofw_rb) {
|
H A D | libxsmm_dnn_convolve_st_upd_custom_custom_generic.tpl.c | 259 } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) {
|
H A D | libxsmm_dnn_convolve_st_upd_nhwc_custom-rsck_generic.tpl.c | 308 } else if (oi == handle->ofw-handle->fwd_ofw_rb && ki == handle->desc.S-1) {
|