1584 lines
58 KiB
YAML
1584 lines
58 KiB
YAML
# Copyright 2023 Google LLC
|
|
#
|
|
# This source code is licensed under the BSD-style license found in the
|
|
# LICENSE file in the root directory of this source tree.
|
|
|
|
# SSSE3 / SSE4 MADD
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__ssse3_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__ssse3_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__ssse3_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__ssse3_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__sse41_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__sse41_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__sse41_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__sse41_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__ssse3_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__ssse3_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__ssse3_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__ssse3_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__sse41_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__sse41_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__sse41_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__sse41_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX256SKX MADD
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX2 MADD
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX512SKX MADD
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c4__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c4__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c8__avx512skx_madd
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c8__avx512skx_madd_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4uw_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX512AMX
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x64c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x64c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_16x64c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_16x64c4__avx512amx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x32c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x32c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_16x32c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_16x32c4__avx512amx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_16x16c4__avx512amx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_16x16c4__avx512amx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 64
|
|
|
|
# ARM NEONDOT
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x16c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x16c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x16c4__neondot
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
|
|
# ARM NEONI8MM
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x32c8__neoni8mm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# ARM NEON
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16__neon_mlal_lane
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16__neon_mlal_lane_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x16__neon_mlal_lane
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x16__neon_mlal_lane_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x16__neon_mlal_lane
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x16__neon_mlal_lane_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16__neon_mlal_lane
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16__neon_mlal_lane_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x16__neon_mlal_lane
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x16__neon_mlal_lane_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
|
|
# AVX512SKX
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512skx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512skx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512skx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512skx_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# AVX512 VNNI
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c4__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c4__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c8__avx512vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c8__avx512vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX512 VNNI GFNI
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c4__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c4__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 8
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c8__avx512vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x16c8__avx512vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX256 VNNI
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVX256 VNNI GFNI
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# AVXVNNI
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni_prfm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
unsigned-inputs: True
|
|
|
|
# x86 AVX2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx2
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# x86 AVX256SKX
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# AVX
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__avx_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__avx_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__avx_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__avx_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__avx_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__avx_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__avx_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__avx_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# SSE4
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__sse41_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__sse41_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__sse41_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__sse41_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__sse41_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__sse41_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__sse41_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__sse41_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# SSE2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__sse2_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__sse2_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__sse2_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__sse2_ld128
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__sse2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__sse2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__sse2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__sse2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# RISC-V Vector
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_5x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_6x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_7x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_8x4v__rvv
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
|
|
# WAsmSIMD
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4c8__wasmsimd_dot16x2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4c8__wasmsimd_dot16x2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_3x4c8__wasmsimd_dot16x2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4c8__wasmsimd_dot16x2_ld64
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 16
|
|
|
|
# Scalar
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x1__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x2__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x2__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4__scalar
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
|
|
# WAsm
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x2__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x4__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_1x8__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x2__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x4__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|
|
- name: xnn_qd8_f32_qc4w_gemm_minmax_ukernel_4x4__wasm
|
|
init: xnn_init_f32_qc4w_minmax_scalar_params
|
|
pack: xnn_pack_qs8_qc4w_gemm_goi_w
|
|
k-block: 2
|