diff options
author | Frank Barchard <fbarchard@google.com> | 2021-01-22 14:08:02 -0800 |
---|---|---|
committer | XNNPACK Team <xnnpack-github-robot@google.com> | 2021-01-22 14:09:03 -0800 |
commit | 2302ffd7d9b2a68ae622dffb2aaa73dc57d49915 (patch) | |
tree | 1302d362131207b636ddea56f4aa0a6df86712df /scripts | |
parent | cadd422eb2642e6d1e9c73bead5ca3cde8c4e011 (diff) | |
download | XNNPACK-2302ffd7d9b2a68ae622dffb2aaa73dc57d49915.tar.gz |
QS8 Neon GEMM microkernel with 8 bit multiply and vpadal to accumulate
PiperOrigin-RevId: 353315852
Diffstat (limited to 'scripts')
-rwxr-xr-x | scripts/generate-qs8-gemm.sh | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/scripts/generate-qs8-gemm.sh b/scripts/generate-qs8-gemm.sh index f6cf0bec8..4cf3fc5fb 100755 --- a/scripts/generate-qs8-gemm.sh +++ b/scripts/generate-qs8-gemm.sh @@ -44,6 +44,16 @@ tools/xngen src/qs8-gemm/neon-mull-addw-dup.c.in -D MR=2 -D NR=16 -o src/qs8-gem tools/xngen src/qs8-gemm/neon-mull-addw-dup.c.in -D MR=3 -D NR=16 -o src/qs8-gemm/gen/3x16-minmax-neon-mull-addw-dup.c tools/xngen src/qs8-gemm/neon-mull-addw-dup.c.in -D MR=4 -D NR=16 -o src/qs8-gemm/gen/4x16-minmax-neon-mull-addw-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=1 -D NR=8 -o src/qs8-gemm/gen/1x8c2-minmax-neon-mull-padal-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=2 -D NR=8 -o src/qs8-gemm/gen/2x8c2-minmax-neon-mull-padal-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=3 -D NR=8 -o src/qs8-gemm/gen/3x8c2-minmax-neon-mull-padal-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=4 -D NR=8 -o src/qs8-gemm/gen/4x8c2-minmax-neon-mull-padal-dup.c + +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=1 -D NR=16 -o src/qs8-gemm/gen/1x16c2-minmax-neon-mull-padal-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=2 -D NR=16 -o src/qs8-gemm/gen/2x16c2-minmax-neon-mull-padal-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=3 -D NR=16 -o src/qs8-gemm/gen/3x16c2-minmax-neon-mull-padal-dup.c +tools/xngen src/qs8-gemm/c2-neon-mull-padal-dup.c.in -D MR=4 -D NR=16 -o src/qs8-gemm/gen/4x16c2-minmax-neon-mull-padal-dup.c + ### C4 micro-kernels tools/xngen src/qs8-gemm/MRxNRc4-neondot.c.in -D MR=1 -D NR=8 -o src/qs8-gemm/gen/1x8c4-minmax-neondot.c tools/xngen src/qs8-gemm/MRxNRc4-neondot.c.in -D MR=4 -D NR=8 -o src/qs8-gemm/gen/4x8c4-minmax-neondot.c |