aboutsummaryrefslogtreecommitdiff
path: root/scripts
diff options
context:
space:
mode:
authorFrank Barchard <fbarchard@google.com>2021-02-26 15:32:28 -0800
committerXNNPACK Team <xnnpack-github-robot@google.com>2021-02-26 15:33:22 -0800
commit02121caa363ea04fda5f79ef073cf4884ab35279 (patch)
tree93be69c19107d5c19a2190c452f6da4671117d96 /scripts
parent8f6a1ed35c3feb6e6cccadee56d3f104924babc9 (diff)
downloadXNNPACK-02121caa363ea04fda5f79ef073cf4884ab35279.tar.gz
QS8 Neon IGEMM microkernels with 8 bit MUL using DUP
PiperOrigin-RevId: 359852046
Diffstat (limited to 'scripts')
-rwxr-xr-xscripts/generate-qs8-igemm.sh18
1 files changed, 13 insertions, 5 deletions
diff --git a/scripts/generate-qs8-igemm.sh b/scripts/generate-qs8-igemm.sh
index 227ed32d4..6751f4744 100755
--- a/scripts/generate-qs8-igemm.sh
+++ b/scripts/generate-qs8-igemm.sh
@@ -15,11 +15,10 @@ tools/xngen src/qs8-igemm/MRx4c8-wasmsimd.c.in -D MR=2 -D VARIANT=LD128 -o src/q
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
################################### ARM NEON ##################################
-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
-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
-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
-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
-
+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
+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
+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
+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
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
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
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
@@ -64,6 +63,15 @@ tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MLA=1 -D MR=2 -D NR=16
tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MLA=1 -D MR=3 -D NR=16 -o src/qs8-igemm/gen/3x16c2-minmax-neon-mlal-padal-dup.c
tools/xngen src/qs8-igemm/c2-neon-mull-padal-dup.c.in -D MLA=1 -D MR=4 -D NR=16 -o src/qs8-igemm/gen/4x16c2-minmax-neon-mlal-padal-dup.c
+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
+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
+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
+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
+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
+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
+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
+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
+
### C4 micro-kernels
tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=1 -D NR=8 -o src/qs8-igemm/gen/1x8c4-minmax-neondot.c
tools/xngen src/qs8-igemm/MRxNRc4-neondot.c.in -D MR=4 -D NR=8 -o src/qs8-igemm/gen/4x8c4-minmax-neondot.c