| /dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/x64/ |
| H A D | jit_avx512_core_bf16_conv_kernel.cpp | 1872 auto zmm_ddst = [=](int i_iw) { in compute_ic_block_step_extern()
1903 int i_iw_shift = jcp.tr_ow - ur_w - ((jcp.l_pad != pad_l) ? jcp.l_pad : 0); in compute_ic_block_step_extern()
1916 int i_iw = 2 * i_ur + (i_kw * (jcp.dilate_w + 1)) / str_w in compute_ic_block_step_extern() local
2006 auto zmm_ddst = [=](int i_iw) { in compute_ic_block_step_interleave()
2009 auto zmm_in = [=](int i_iw, int i_ic, bool stride_reg) { in compute_ic_block_step_interleave()
2028 auto load_src_to_stack = [=](int i_iw, int i_ic, Opmask mask, in compute_ic_block_step_interleave()
2030 bool stride_mask_empty) { in compute_ic_block_step_interleave()
2116 for (int i_iw = reorder_start; i_iw < reorder_end; i_iw += reorder_block) { in compute_ic_block_step_interleave() local
2180 int i_iw = 2 * i_ur * str_w + i_kw; in compute_ic_block_step_interleave() local
2322 auto get_src_reg_idx = [=](int i_iw, int i_ic) { in compute_ic_block_step_vpermw_expl()
[all …]
|
| H A D | jit_avx512_common_conv_kernel.cpp | 3336 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step_fma() local
3376 auto get_src_reg_idx = [=](int i_iw, int i_ic) { in compute_ic_block_step_fma_expl()
3401 int i_iw = get_iw_idx(0, i_kw, pad_l); in compute_ic_block_step_fma_expl() local
3422 int i_iw = get_iw_idx(i_ur, kw - 1, pad_l); in compute_ic_block_step_fma_expl() local
3433 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step_fma_expl() local
3481 auto inp_addr = [=](int i_iw, int i_ic, ptrdiff_t extra_offset = 0) { in compute_ic_block_step_4fma()
3488 auto zmm_out = [=](int i_iw) { in compute_ic_block_step_4fma()
3539 int i_iw = i_ur + i_kw; in compute_ic_block_step_4fma() local
5426 auto src_offset = [&](int i_ic, int i_iw) { in generate_microkernel()
|
| H A D | jit_avx2_conv_kernel_f32.cpp | 1513 int i_iw = i_ur * jcp.stride_w + i_kw; in compute_ic_block_step() local
|
| H A D | jit_avx512_common_convolution.cpp | 1336 = [&](int g, int i_mb, int i_id, int i_ih, int i_ic, int i_iw) { in compute_diff_weights_nxc()
1418 int i_iw = i_ow * jcp.stride_w - jcp.l_pad; in compute_diff_weights_nxc() local
|
| /dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/ |
| H A D | jit_avx512_core_bf16_conv_kernel.cpp | 1863 auto zmm_ddst = [=](int i_iw) { in compute_ic_block_step_extern()
1894 int i_iw_shift = jcp.tr_ow - ur_w - ((jcp.l_pad != pad_l) ? jcp.l_pad : 0); in compute_ic_block_step_extern()
1907 int i_iw = 2 * i_ur + (i_kw * (jcp.dilate_w + 1)) / str_w in compute_ic_block_step_extern() local
1997 auto zmm_ddst = [=](int i_iw) { in compute_ic_block_step_interleave()
2000 auto zmm_in = [=](int i_iw, int i_ic, bool stride_reg) { in compute_ic_block_step_interleave()
2019 auto load_src_to_stack = [=](int i_iw, int i_ic, Opmask mask, in compute_ic_block_step_interleave()
2021 bool stride_mask_empty) { in compute_ic_block_step_interleave()
2107 for (int i_iw = reorder_start; i_iw < reorder_end; i_iw += reorder_block) { in compute_ic_block_step_interleave() local
2171 int i_iw = 2 * i_ur * str_w + i_kw; in compute_ic_block_step_interleave() local
2313 auto get_src_reg_idx = [=](int i_iw, int i_ic) { in compute_ic_block_step_vpermw_expl()
[all …]
|
| H A D | jit_avx512_common_conv_kernel.cpp | 3324 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step_fma() local
3364 auto get_src_reg_idx = [=](int i_iw, int i_ic) { in compute_ic_block_step_fma_expl()
3389 int i_iw = get_iw_idx(0, i_kw, pad_l); in compute_ic_block_step_fma_expl() local
3410 int i_iw = get_iw_idx(i_ur, kw - 1, pad_l); in compute_ic_block_step_fma_expl() local
3421 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step_fma_expl() local
3469 auto inp_addr = [=](int i_iw, int i_ic, ptrdiff_t extra_offset = 0) { in compute_ic_block_step_4fma()
3476 auto zmm_out = [=](int i_iw) { in compute_ic_block_step_4fma()
3527 int i_iw = i_ur + i_kw; in compute_ic_block_step_4fma() local
5427 auto src_offset = [&](int i_ic, int i_iw) { in generate_microkernel()
|
| H A D | jit_avx2_conv_kernel_f32.cpp | 1503 int i_iw = i_ur * jcp.stride_w + i_kw; in compute_ic_block_step() local
|
| H A D | jit_avx512_common_convolution.cpp | 1336 = [&](int g, int i_mb, int i_id, int i_ih, int i_ic, int i_iw) { in compute_diff_weights_nxc()
1418 int i_iw = i_ow * jcp.stride_w - jcp.l_pad; in compute_diff_weights_nxc() local
|
| /dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/aarch64/ |
| H A D | jit_sve_512_conv_kernel.cpp | 2598 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step() local
2635 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step() local
|
| /dports/math/onednn/oneDNN-2.5.1/src/cpu/aarch64/ |
| H A D | jit_sve_512_conv_kernel.cpp | 2598 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step() local
2635 int i_iw = get_iw_idx(i_ur, i_kw, pad_l); in compute_ic_block_step() local
|
| /dports/math/onednn/oneDNN-2.5.1/src/gpu/jit/ |
| H A D | xe_hp_conv_data_kernel.cpp | 647 int i_iw = -conf.l_pad + i_ow * conf.stride_w; in xe_hp_conv_data_kernel_t() local
|