Marat Dukhan | f948068 | 2020-07-31 14:50:24 -0700 | [diff] [blame] | 1 | #!/bin/sh |
| 2 | # Copyright 2020 Google LLC |
| 3 | # |
| 4 | # This source code is licensed under the BSD-style license found in the |
| 5 | # LICENSE file in the root directory of this source tree. |
| 6 | |
Marat Dukhan | 27203da | 2020-08-05 15:19:03 -0700 | [diff] [blame] | 7 | ################################## WAsm SIMD ################################## |
Marat Dukhan | ef4ce31 | 2020-09-10 12:29:08 -0700 | [diff] [blame] | 8 | ### C8 micro-kernels |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 9 | tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=1 -D VARIANT=LD64 -o src/qs8-igemm/gen/1x4c8-minmax-wasmsimd-ld64.c |
| 10 | tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=2 -D VARIANT=LD64 -o src/qs8-igemm/gen/2x4c8-minmax-wasmsimd-ld64.c |
| 11 | tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=3 -D VARIANT=LD64 -o src/qs8-igemm/gen/3x4c8-minmax-wasmsimd-ld64.c |
Marat Dukhan | 27203da | 2020-08-05 15:19:03 -0700 | [diff] [blame] | 12 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 13 | tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=1 -D VARIANT=LD128 -o src/qs8-igemm/gen/1x4c8-minmax-wasmsimd-ld128.c |
| 14 | tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=2 -D VARIANT=LD128 -o src/qs8-igemm/gen/2x4c8-minmax-wasmsimd-ld128.c |
| 15 | tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=3 -D VARIANT=LD128 -o src/qs8-igemm/gen/3x4c8-minmax-wasmsimd-ld128.c |
Marat Dukhan | 27203da | 2020-08-05 15:19:03 -0700 | [diff] [blame] | 16 | |
Marat Dukhan | 40bbafe | 2020-08-04 02:04:22 -0700 | [diff] [blame] | 17 | ################################### ARM NEON ################################## |
Frank Barchard | da78da1 | 2021-03-02 14:28:00 -0800 | [diff] [blame] | 18 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=1 -D NR=8 -o src/qs8-igemm/gen/1x8-minmax-neon-mlal-lane.c |
| 19 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=2 -D NR=8 -o src/qs8-igemm/gen/2x8-minmax-neon-mlal-lane.c |
| 20 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=3 -D NR=8 -o src/qs8-igemm/gen/3x8-minmax-neon-mlal-lane.c |
| 21 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=4 -D NR=8 -o src/qs8-igemm/gen/4x8-minmax-neon-mlal-lane.c |
| 22 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 23 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=1 -D NR=16 -o src/qs8-igemm/gen/1x16-minmax-neon-mlal-lane.c |
| 24 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=2 -D NR=16 -o src/qs8-igemm/gen/2x16-minmax-neon-mlal-lane.c |
Frank Barchard | ec0bf14 | 2021-01-15 10:14:05 -0800 | [diff] [blame] | 25 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=3 -D NR=16 -o src/qs8-igemm/gen/3x16-minmax-neon-mlal-lane.c |
| 26 | tools/xngen src/qs8-igemm/neon-mlal-lane.c.in -D MR=4 -D NR=16 -o src/qs8-igemm/gen/4x16-minmax-neon-mlal-lane.c |
Marat Dukhan | 40bbafe | 2020-08-04 02:04:22 -0700 | [diff] [blame] | 27 | |
Frank Barchard | da78da1 | 2021-03-02 14:28:00 -0800 | [diff] [blame] | 28 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=1 -D NR=8 -o src/qs8-igemm/gen/1x8-minmax-neon-mull-addw-dup.c |
| 29 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=2 -D NR=8 -o src/qs8-igemm/gen/2x8-minmax-neon-mull-addw-dup.c |
| 30 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=3 -D NR=8 -o src/qs8-igemm/gen/3x8-minmax-neon-mull-addw-dup.c |
| 31 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=4 -D NR=8 -o src/qs8-igemm/gen/4x8-minmax-neon-mull-addw-dup.c |
| 32 | |
| 33 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=1 -D NR=16 -o src/qs8-igemm/gen/1x16-minmax-neon-mull-addw-dup.c |
| 34 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=2 -D NR=16 -o src/qs8-igemm/gen/2x16-minmax-neon-mull-addw-dup.c |
| 35 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=3 -D NR=16 -o src/qs8-igemm/gen/3x16-minmax-neon-mull-addw-dup.c |
| 36 | tools/xngen src/qs8-igemm/neon-mull-addw-dup.c.in -D MR=4 -D NR=16 -o src/qs8-igemm/gen/4x16-minmax-neon-mull-addw-dup.c |
| 37 | |
| 38 | ### C2 micro-kernels |
| 39 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=1 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/1x8c2-minmax-neon-mull-padal-dup.c |
| 40 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=2 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/2x8c2-minmax-neon-mull-padal-dup.c |
| 41 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=3 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/3x8c2-minmax-neon-mull-padal-dup.c |
| 42 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=4 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/4x8c2-minmax-neon-mull-padal-dup.c |
| 43 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=1 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/1x16c2-minmax-neon-mull-padal-dup.c |
| 44 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=2 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/2x16c2-minmax-neon-mull-padal-dup.c |
| 45 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=3 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/3x16c2-minmax-neon-mull-padal-dup.c |
| 46 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=4 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/4x16c2-minmax-neon-mull-padal-dup.c |
| 47 | |
| 48 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=1 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/1x8c2-minmax-neon-mlal-padal-dup.c |
| 49 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=2 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/2x8c2-minmax-neon-mlal-padal-dup.c |
| 50 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=3 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/3x8c2-minmax-neon-mlal-padal-dup.c |
| 51 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=4 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/4x8c2-minmax-neon-mlal-padal-dup.c |
| 52 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=1 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/1x16c2-minmax-neon-mlal-padal-dup.c |
| 53 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=2 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/2x16c2-minmax-neon-mlal-padal-dup.c |
| 54 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=3 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/3x16c2-minmax-neon-mlal-padal-dup.c |
| 55 | tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MR=4 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/4x16c2-minmax-neon-mlal-padal-dup.c |
| 56 | |
Frank Barchard | d79391d | 2021-02-15 14:22:32 -0800 | [diff] [blame] | 57 | ### C8 micro-kernels |
Frank Barchard | da78da1 | 2021-03-02 14:28:00 -0800 | [diff] [blame] | 58 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=1 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/1x8c8-minmax-neon-mull-padal.c |
| 59 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=2 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/2x8c8-minmax-neon-mull-padal.c |
| 60 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=3 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/3x8c8-minmax-neon-mull-padal.c |
| 61 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=4 -D NR=8 -D MLA=0 -o src/qs8-igemm/gen/4x8c8-minmax-neon-mull-padal.c |
| 62 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=1 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/1x16c8-minmax-neon-mull-padal.c |
| 63 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=2 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/2x16c8-minmax-neon-mull-padal.c |
| 64 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=3 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/3x16c8-minmax-neon-mull-padal.c |
| 65 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=4 -D NR=16 -D MLA=0 -o src/qs8-igemm/gen/4x16c8-minmax-neon-mull-padal.c |
| 66 | |
| 67 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=1 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/1x8c8-minmax-neon-mlal-padal.c |
| 68 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=2 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/2x8c8-minmax-neon-mlal-padal.c |
| 69 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=3 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/3x8c8-minmax-neon-mlal-padal.c |
| 70 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=4 -D NR=8 -D MLA=1 -o src/qs8-igemm/gen/4x8c8-minmax-neon-mlal-padal.c |
| 71 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=1 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/1x16c8-minmax-neon-mlal-padal.c |
| 72 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=2 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/2x16c8-minmax-neon-mlal-padal.c |
| 73 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=3 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/3x16c8-minmax-neon-mlal-padal.c |
| 74 | tools/xngen src/qs8-igemm/c8-neon-mull-padal.c.in -D MR=4 -D NR=16 -D MLA=1 -o src/qs8-igemm/gen/4x16c8-minmax-neon-mlal-padal.c |
Frank Barchard | d79391d | 2021-02-15 14:22:32 -0800 | [diff] [blame] | 75 | |
Frank Barchard | 36f95cf | 2021-02-23 03:12:52 -0800 | [diff] [blame] | 76 | ### C16 micro-kernels |
| 77 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=1 -D NR=8 -o src/qs8-igemm/gen/1x8c16-minmax-neon-mlal-padal.c |
| 78 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=2 -D NR=8 -o src/qs8-igemm/gen/2x8c16-minmax-neon-mlal-padal.c |
| 79 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=3 -D NR=8 -o src/qs8-igemm/gen/3x8c16-minmax-neon-mlal-padal.c |
| 80 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=4 -D NR=8 -o src/qs8-igemm/gen/4x8c16-minmax-neon-mlal-padal.c |
| 81 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=1 -D NR=16 -o src/qs8-igemm/gen/1x16c16-minmax-neon-mlal-padal.c |
| 82 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=2 -D NR=16 -o src/qs8-igemm/gen/2x16c16-minmax-neon-mlal-padal.c |
| 83 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=3 -D NR=16 -o src/qs8-igemm/gen/3x16c16-minmax-neon-mlal-padal.c |
| 84 | tools/xngen src/qs8-igemm/c16-neon-mlal-padal.c.in -D MR=4 -D NR=16 -o src/qs8-igemm/gen/4x16c16-minmax-neon-mlal-padal.c |
| 85 | |
Benoit Jacob | a05487f | 2020-08-14 13:37:21 -0700 | [diff] [blame] | 86 | ### C4 micro-kernels |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 87 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=1 -D NR=8 -o src/qs8-igemm/gen/1x8c4-minmax-neondot.c |
| 88 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=4 -D NR=8 -o src/qs8-igemm/gen/4x8c4-minmax-neondot.c |
| 89 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=6 -D NR=8 -o src/qs8-igemm/gen/6x8c4-minmax-neondot.c |
| 90 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=8 -D NR=8 -o src/qs8-igemm/gen/8x8c4-minmax-neondot.c |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 91 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=1 -D NR=16 -o src/qs8-igemm/gen/1x16c4-minmax-neondot.c |
| 92 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=4 -D NR=16 -o src/qs8-igemm/gen/4x16c4-minmax-neondot.c |
| 93 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=6 -D NR=16 -o src/qs8-igemm/gen/6x16c4-minmax-neondot.c |
| 94 | tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=8 -D NR=16 -o src/qs8-igemm/gen/8x16c4-minmax-neondot.c |
| 95 | |
Marat Dukhan | f948068 | 2020-07-31 14:50:24 -0700 | [diff] [blame] | 96 | ################################### x86 SSE ################################### |
| 97 | ### C2 micro-kernels |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 98 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=2 -D LD128=0 -o src/qs8-igemm/gen/1x4c2-minmax-sse2-ld64.c |
| 99 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=2 -D LD128=0 -o src/qs8-igemm/gen/4x4c2-minmax-sse2-ld64.c |
Marat Dukhan | f948068 | 2020-07-31 14:50:24 -0700 | [diff] [blame] | 100 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 101 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=3 -D LD128=0 -o src/qs8-igemm/gen/1x4c2-minmax-ssse3-ld64.c |
| 102 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=3 -D LD128=0 -o src/qs8-igemm/gen/4x4c2-minmax-ssse3-ld64.c |
Marat Dukhan | f948068 | 2020-07-31 14:50:24 -0700 | [diff] [blame] | 103 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 104 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=4 -D LD128=0 -o src/qs8-igemm/gen/1x4c2-minmax-sse41-ld64.c |
| 105 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=4 -D LD128=0 -o src/qs8-igemm/gen/4x4c2-minmax-sse41-ld64.c |
Marat Dukhan | f948068 | 2020-07-31 14:50:24 -0700 | [diff] [blame] | 106 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 107 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=5 -D LD128=0 -o src/qs8-igemm/gen/1x4c2-minmax-xop-ld64.c |
| 108 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=5 -D LD128=0 -o src/qs8-igemm/gen/4x4c2-minmax-xop-ld64.c |
Marat Dukhan | 1566fee | 2020-08-02 21:55:41 -0700 | [diff] [blame] | 109 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 110 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=2 -D LD128=1 -o src/qs8-igemm/gen/1x4c2-minmax-sse2-ld128.c |
| 111 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=2 -D LD128=1 -o src/qs8-igemm/gen/4x4c2-minmax-sse2-ld128.c |
Marat Dukhan | dee732b | 2020-07-31 17:32:04 -0700 | [diff] [blame] | 112 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 113 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=3 -D LD128=1 -o src/qs8-igemm/gen/1x4c2-minmax-ssse3-ld128.c |
| 114 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=3 -D LD128=1 -o src/qs8-igemm/gen/4x4c2-minmax-ssse3-ld128.c |
Marat Dukhan | dee732b | 2020-07-31 17:32:04 -0700 | [diff] [blame] | 115 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 116 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=4 -D LD128=1 -o src/qs8-igemm/gen/1x4c2-minmax-sse41-ld128.c |
| 117 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=4 -D LD128=1 -o src/qs8-igemm/gen/4x4c2-minmax-sse41-ld128.c |
Marat Dukhan | dee732b | 2020-07-31 17:32:04 -0700 | [diff] [blame] | 118 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 119 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=1 -D SSE=5 -D LD128=1 -o src/qs8-igemm/gen/1x4c2-minmax-xop-ld128.c |
| 120 | tools/xngen src/qs8-igemm/MRx4c2-sse.c.in -D MR=4 -D SSE=5 -D LD128=1 -o src/qs8-igemm/gen/4x4c2-minmax-xop-ld128.c |
Marat Dukhan | 1566fee | 2020-08-02 21:55:41 -0700 | [diff] [blame] | 121 | |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 122 | ### C8 micro-kernels |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 123 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=2 -D LD128=0 -o src/qs8-igemm/gen/1x4c8-minmax-sse2-ld64.c |
| 124 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=2 -D LD128=0 -o src/qs8-igemm/gen/2x4c8-minmax-sse2-ld64.c |
| 125 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=2 -D LD128=0 -o src/qs8-igemm/gen/3x4c8-minmax-sse2-ld64.c |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 126 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 127 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=3 -D LD128=0 -o src/qs8-igemm/gen/1x4c8-minmax-ssse3-ld64.c |
| 128 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=3 -D LD128=0 -o src/qs8-igemm/gen/2x4c8-minmax-ssse3-ld64.c |
| 129 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=3 -D LD128=0 -o src/qs8-igemm/gen/3x4c8-minmax-ssse3-ld64.c |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 130 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 131 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=4 -D LD128=0 -o src/qs8-igemm/gen/1x4c8-minmax-sse41-ld64.c |
| 132 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=4 -D LD128=0 -o src/qs8-igemm/gen/2x4c8-minmax-sse41-ld64.c |
| 133 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=4 -D LD128=0 -o src/qs8-igemm/gen/3x4c8-minmax-sse41-ld64.c |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 134 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 135 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=5 -D LD128=0 -o src/qs8-igemm/gen/1x4c8-minmax-xop-ld64.c |
| 136 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=5 -D LD128=0 -o src/qs8-igemm/gen/2x4c8-minmax-xop-ld64.c |
| 137 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=5 -D LD128=0 -o src/qs8-igemm/gen/3x4c8-minmax-xop-ld64.c |
Marat Dukhan | 1566fee | 2020-08-02 21:55:41 -0700 | [diff] [blame] | 138 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 139 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=2 -D LD128=1 -o src/qs8-igemm/gen/1x4c8-minmax-sse2-ld128.c |
| 140 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=2 -D LD128=1 -o src/qs8-igemm/gen/2x4c8-minmax-sse2-ld128.c |
| 141 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=2 -D LD128=1 -o src/qs8-igemm/gen/3x4c8-minmax-sse2-ld128.c |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 142 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 143 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=3 -D LD128=1 -o src/qs8-igemm/gen/1x4c8-minmax-ssse3-ld128.c |
| 144 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=3 -D LD128=1 -o src/qs8-igemm/gen/2x4c8-minmax-ssse3-ld128.c |
| 145 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=3 -D LD128=1 -o src/qs8-igemm/gen/3x4c8-minmax-ssse3-ld128.c |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 146 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 147 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=4 -D LD128=1 -o src/qs8-igemm/gen/1x4c8-minmax-sse41-ld128.c |
| 148 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=4 -D LD128=1 -o src/qs8-igemm/gen/2x4c8-minmax-sse41-ld128.c |
| 149 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=4 -D LD128=1 -o src/qs8-igemm/gen/3x4c8-minmax-sse41-ld128.c |
Marat Dukhan | 07bd252 | 2020-07-31 19:12:39 -0700 | [diff] [blame] | 150 | |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 151 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=1 -D SSE=5 -D LD128=1 -o src/qs8-igemm/gen/1x4c8-minmax-xop-ld128.c |
| 152 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=2 -D SSE=5 -D LD128=1 -o src/qs8-igemm/gen/2x4c8-minmax-xop-ld128.c |
| 153 | tools/xngen src/qs8-igemm/MRx4c8-sse.c.in -D MR=3 -D SSE=5 -D LD128=1 -o src/qs8-igemm/gen/3x4c8-minmax-xop-ld128.c |
Marat Dukhan | 1566fee | 2020-08-02 21:55:41 -0700 | [diff] [blame] | 154 | |
Marat Dukhan | 1280952 | 2020-08-02 22:23:51 -0700 | [diff] [blame] | 155 | ################################### x86 AVX2 ################################## |
| 156 | ### C8 micro-kernels |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 157 | tools/xngen src/qs8-igemm/MRx8c8-avx2.c.in -D MR=1 -o src/qs8-igemm/gen/1x8c8-minmax-avx2.c |
| 158 | tools/xngen src/qs8-igemm/MRx8c8-avx2.c.in -D MR=2 -o src/qs8-igemm/gen/2x8c8-minmax-avx2.c |
| 159 | tools/xngen src/qs8-igemm/MRx8c8-avx2.c.in -D MR=3 -o src/qs8-igemm/gen/3x8c8-minmax-avx2.c |
Marat Dukhan | 1280952 | 2020-08-02 22:23:51 -0700 | [diff] [blame] | 160 | |
Marat Dukhan | bb00b1d | 2020-08-10 11:37:23 -0700 | [diff] [blame] | 161 | ################################## x86 AVX512 ################################# |
| 162 | ### C8 micro-kernels |
Marat Dukhan | 66ccf64 | 2020-09-28 16:23:42 -0700 | [diff] [blame] | 163 | tools/xngen src/qs8-igemm/MRx16c8-avx512skx.c.in -D MR=1 -D VARIANT=LD256 -o src/qs8-igemm/gen/1x16c8-minmax-avx512skx.c |
| 164 | tools/xngen src/qs8-igemm/MRx16c8-avx512skx.c.in -D MR=2 -D VARIANT=LD256 -o src/qs8-igemm/gen/2x16c8-minmax-avx512skx.c |
| 165 | tools/xngen src/qs8-igemm/MRx16c8-avx512skx.c.in -D MR=3 -D VARIANT=LD256 -o src/qs8-igemm/gen/3x16c8-minmax-avx512skx.c |
| 166 | tools/xngen src/qs8-igemm/MRx16c8-avx512skx.c.in -D MR=4 -D VARIANT=LD256 -o src/qs8-igemm/gen/4x16c8-minmax-avx512skx.c |
Marat Dukhan | bb00b1d | 2020-08-10 11:37:23 -0700 | [diff] [blame] | 167 | |
Marat Dukhan | f948068 | 2020-07-31 14:50:24 -0700 | [diff] [blame] | 168 | ################################## Unit tests ################################# |
| 169 | tools/generate-gemm-test.py --spec test/qs8-igemm-minmax.yaml --output test/qs8-igemm-minmax.cc |