/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/aarch64/ |
H A D | jit_sve_512_convolution.cpp | 55 PIPELINE(reduce_work); in jit_conv_ker_pipeline() 68 reduce_work, load_work); in jit_conv_ker_pipeline_ow_thr() 77 reduce_work, load_work); in jit_conv_ker_pipeline_iw_thr() 83 int reduce_work, int load_work) { in jit_sve_512_conv_3d_ker_pipeline() argument 93 PIPELINE(reduce_work); in jit_sve_512_conv_3d_ker_pipeline() 103 int reduce_work, int load_work, int flags) { in jit_sve_512_conv_3d_ker_pipeline_ow_thr() argument 108 kh_padding, kd_padding, reduce_work, load_work); in jit_sve_512_conv_3d_ker_pipeline_ow_thr() 115 reduce_work, load_work); in jit_conv_ker_pipeline_bwd_w() 134 PIPELINE(reduce_work); in jit_sve_512_conv_2d_ker_bwd_w_pipeline() 156 PIPELINE(reduce_work); in jit_sve_512_conv_3d_ker_bwd_w_pipeline() [all …]
|
H A D | jit_primitive_conf.hpp | 273 size_t reduce_work; member
|
H A D | jit_sve_512_conv_kernel.cpp | 559 ldr(reg_channel, ptr(param, GET_OFF(reduce_work))); in compute_loop() 1798 ldr(reg_channel, ptr(param, GET_OFF(reduce_work))); in compute_loop() 2763 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in compute_oh_step_unroll_ow_icblock() 2888 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in compute_oh_step_unroll_ow() 3101 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in compute_oh_step_common() 3278 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in maybe_zero_kernel()
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/aarch64/ |
H A D | jit_sve_512_convolution.cpp | 55 PIPELINE(reduce_work); in jit_conv_ker_pipeline() 68 reduce_work, load_work); in jit_conv_ker_pipeline_ow_thr() 77 reduce_work, load_work); in jit_conv_ker_pipeline_iw_thr() 83 int reduce_work, int load_work) { in jit_sve_512_conv_3d_ker_pipeline() argument 93 PIPELINE(reduce_work); in jit_sve_512_conv_3d_ker_pipeline() 103 int reduce_work, int load_work, int flags) { in jit_sve_512_conv_3d_ker_pipeline_ow_thr() argument 108 kh_padding, kd_padding, reduce_work, load_work); in jit_sve_512_conv_3d_ker_pipeline_ow_thr() 115 reduce_work, load_work); in jit_conv_ker_pipeline_bwd_w() 134 PIPELINE(reduce_work); in jit_sve_512_conv_2d_ker_bwd_w_pipeline() 156 PIPELINE(reduce_work); in jit_sve_512_conv_3d_ker_bwd_w_pipeline() [all …]
|
H A D | jit_primitive_conf.hpp | 273 size_t reduce_work; member
|
H A D | jit_sve_512_conv_kernel.cpp | 559 ldr(reg_channel, ptr(param, GET_OFF(reduce_work))); in compute_loop() 1798 ldr(reg_channel, ptr(param, GET_OFF(reduce_work))); in compute_loop() 2763 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in compute_oh_step_unroll_ow_icblock() 2888 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in compute_oh_step_unroll_ow() 3101 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in compute_oh_step_common() 3278 ldr(reg_icb, ptr(param, GET_OFF(reduce_work))); in maybe_zero_kernel()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/x64/ |
H A D | jit_avx512_common_convolution.cpp | 54 PIPELINE(reduce_work); in jit_conv_ker_pipeline() 67 reduce_work, load_work); in jit_conv_ker_pipeline_iw_thr() 73 int reduce_work, int load_work) { in jit_conv_3d_ker_pipeline() argument 83 PIPELINE(reduce_work); in jit_conv_3d_ker_pipeline() 102 reduce_work, load_work); in jit_conv_ker_pipeline_ow_thr() 109 int reduce_work, int load_work, int flags) { in jit_conv_3d_ker_pipeline_ow_thr() argument 114 kd_padding, reduce_work, load_work); in jit_conv_3d_ker_pipeline_ow_thr() 138 reduce_work, load_work); in jit_conv_ker_pipeline_bwd_w() 157 PIPELINE(reduce_work); in jit_conv_2d_ker_bwd_w_pipeline() 179 PIPELINE(reduce_work); in jit_conv_3d_ker_bwd_w_pipeline() [all …]
|
H A D | jit_brgemm_inner_product_utils.cpp | 351 float reduce_work = 0.5f * num_min_chunk_sz * jbgp.nb_os in init_ip_conf_fwd() local 356 int(reduce_work)); in init_ip_conf_fwd() 492 float reduce_work = 0.5f * num_min_chunk_sz * jbgp.nb_os in init_ip_conf_bwd_d() local 499 int(reduce_work)); in init_ip_conf_bwd_d()
|
H A D | jit_avx2_convolution.cpp | 146 par_conv.reduce_work = this_block_size( in execute_forward() 331 par_conv.reduce_work in execute_backward_data()
|
H A D | jit_primitive_conf.hpp | 470 size_t reduce_work; member
|
H A D | jit_avx512_core_bf16_conv_kernel.cpp | 2730 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_unroll_ow_icblock() 2833 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_unroll_ow() 3006 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_common() 3028 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_common() 3164 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_common() 3272 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in maybe_zero_kernel()
|
H A D | jit_avx512_core_amx_convolution.cpp | 1268 p.reduce_work = ic_to_compute; in compute_diff_weights_2d() 1504 p.reduce_work = ic_to_compute; in compute_diff_weights_3d() 1756 p.reduce_work = ic_to_compute; in compute_diff_weights()
|
H A D | jit_avx512_core_bf16_convolution.cpp | 1224 p.reduce_work = ic_to_compute; in compute_diff_weights_2d() 1472 p.reduce_work = ic_to_compute; in compute_diff_weights_3d() 1767 p.reduce_work = ic_to_compute; in compute_diff_weights()
|
H A D | jit_avx2_conv_kernel_f32.cpp | 555 mov(reg_channel, ptr[param1 + GET_OFF(reduce_work)]); in generate() 852 mov(reg_reduce_work, ptr[param1 + GET_OFF(reduce_work)]); in compute_loop()
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/ |
H A D | jit_avx512_common_convolution.cpp | 54 PIPELINE(reduce_work); in jit_conv_ker_pipeline() 67 reduce_work, load_work); in jit_conv_ker_pipeline_iw_thr() 73 int reduce_work, int load_work) { in jit_conv_3d_ker_pipeline() argument 83 PIPELINE(reduce_work); in jit_conv_3d_ker_pipeline() 102 reduce_work, load_work); in jit_conv_ker_pipeline_ow_thr() 109 int reduce_work, int load_work, int flags) { in jit_conv_3d_ker_pipeline_ow_thr() argument 114 kd_padding, reduce_work, load_work); in jit_conv_3d_ker_pipeline_ow_thr() 138 reduce_work, load_work); in jit_conv_ker_pipeline_bwd_w() 157 PIPELINE(reduce_work); in jit_conv_2d_ker_bwd_w_pipeline() 179 PIPELINE(reduce_work); in jit_conv_3d_ker_bwd_w_pipeline() [all …]
|
H A D | jit_brgemm_inner_product_utils.cpp | 370 float reduce_work = 0.5f * num_min_chunk_sz * jbgp.nb_os in init_ip_conf_fwd() local 375 int(reduce_work)); in init_ip_conf_fwd() 511 float reduce_work = 0.5f * num_min_chunk_sz * jbgp.nb_os in init_ip_conf_bwd_d() local 518 int(reduce_work)); in init_ip_conf_bwd_d()
|
H A D | jit_avx2_convolution.cpp | 146 par_conv.reduce_work = this_block_size( in execute_forward() 331 par_conv.reduce_work in execute_backward_data()
|
H A D | jit_primitive_conf.hpp | 472 size_t reduce_work; member
|
H A D | jit_avx512_core_bf16_conv_kernel.cpp | 2721 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_unroll_ow_icblock() 2824 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_unroll_ow() 2997 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_common() 3019 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_common() 3155 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in compute_oh_step_common() 3263 mov(reg_icb, ptr[param + GET_OFF(reduce_work)]); in maybe_zero_kernel()
|
H A D | jit_avx512_core_amx_convolution.cpp | 1249 p.reduce_work = ic_to_compute; in compute_diff_weights_2d() 1485 p.reduce_work = ic_to_compute; in compute_diff_weights_3d() 1737 p.reduce_work = ic_to_compute; in compute_diff_weights()
|
H A D | jit_avx512_core_bf16_convolution.cpp | 1224 p.reduce_work = ic_to_compute; in compute_diff_weights_2d() 1472 p.reduce_work = ic_to_compute; in compute_diff_weights_3d() 1767 p.reduce_work = ic_to_compute; in compute_diff_weights()
|
H A D | jit_avx2_conv_kernel_f32.cpp | 544 mov(reg_channel, ptr[param1 + GET_OFF(reduce_work)]); in generate() 842 mov(reg_reduce_work, ptr[param1 + GET_OFF(reduce_work)]); in compute_loop()
|
/dports/math/libxsmm/libxsmm-1.16.3/src/template/ |
H A D | libxsmm_dnn_convolve_st_upd_custom_custom_generic.tpl.c | 523 …const int reduce_work = handle->blocksofm * handle->blocksifm * handle->desc.R * handle->desc.S * … variable 524 …onst int reduce_chunksize = (reduce_work % handle->desc.threads == 0) ? (reduce_work / handle->des… 525 …nt reduce_thr_begin = (ltid * reduce_chunksize < reduce_work) ? (ltid * reduce_chunksize) : reduce… 526 …ce_thr_end = ((ltid + 1) * reduce_chunksize < reduce_work) ? ((ltid + 1) * reduce_chunksize) : red…
|
H A D | libxsmm_dnn_convolve_st_upd_custom_custom_generic_bf16.tpl.c | 103 const int reduce_work = (handle->desc.C * handle->desc.K * handle->desc.R * handle->desc.S)/16; variable 104 …onst int reduce_chunksize = (reduce_work % handle->desc.threads == 0) ? (reduce_work / handle->des… 105 …nt reduce_thr_begin = (ltid * reduce_chunksize < reduce_work) ? (ltid * reduce_chunksize) : reduce… 106 …ce_thr_end = ((ltid + 1) * reduce_chunksize < reduce_work) ? ((ltid + 1) * reduce_chunksize) : red…
|
H A D | libxsmm_dnn_convolve_st_upd_nhwc_custom-rsck_generic.tpl.c | 621 …const int reduce_work = handle->blocksofm * handle->blocksifm * handle->desc.R * handle->desc.S * … variable 622 …onst int reduce_chunksize = (reduce_work % handle->desc.threads == 0) ? (reduce_work / handle->des… 623 …nt reduce_thr_begin = (ltid * reduce_chunksize < reduce_work) ? (ltid * reduce_chunksize) : reduce… 624 …ce_thr_end = ((ltid + 1) * reduce_chunksize < reduce_work) ? ((ltid + 1) * reduce_chunksize) : red…
|