commit 914f57b3deca7dff46f762a4f3152f1bcdb98c65
author Frank Barchard <fbarchard@google.com> Mon Dec 13 12:31:42 2021 -0800
committer XNNPACK Team <xnnpack-github-robot@google.com> Mon Dec 13 12:32:36 2021 -0800
tree 2d14a4bd194c4086d3943bf79c5d2aa827d0371b
parent 77e9e655305c6dd1205c12a9133973f0b6612589

Aarch64 4x8 lane ld64 GEMM/IGEMM microkernels.

- Based on 4x16 microkernel, reduced to 4x8 size.
- Update register usage comments and push/pop for 4x16 IGEMM.

PiperOrigin-RevId: 416107685

CMakeLists.txt

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 9baf826..4c16652 100755
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -5354,6 +5354,8 @@
   src/qs8-gemm/gen/2x8c8-minmax-rndnu-aarch64-neon-mull.S
   src/qs8-gemm/gen/2x8c16-minmax-fp32-aarch64-neon-mlal.S
   src/qs8-gemm/gen/2x8c16-minmax-rndnu-aarch64-neon-mlal.S
+  src/qs8-gemm/gen/4x8-minmax-rndnu-aarch64-neon-mlal-lane-ld64.S
+  src/qs8-gemm/gen/4x8-minmax-rndnu-aarch64-neon-mlal-lane-prfm-ld64.S
   src/qs8-gemm/gen/4x16-minmax-fp32-aarch64-neon-mlal-lane-cortex-a53.S
   src/qs8-gemm/gen/4x16-minmax-fp32-aarch64-neon-mlal-lane-ld64.S
   src/qs8-gemm/gen/4x16-minmax-fp32-aarch64-neon-mlal-lane-prfm-cortex-a53.S
@@ -5388,6 +5390,8 @@
   src/qs8-igemm/gen/2x8c8-minmax-rndnu-aarch64-neon-mlal.S
   src/qs8-igemm/gen/2x8c16-minmax-fp32-aarch64-neon-mlal.S
   src/qs8-igemm/gen/2x8c16-minmax-rndnu-aarch64-neon-mlal.S
+  src/qs8-igemm/gen/4x8-minmax-rndnu-aarch64-neon-mlal-lane-ld64.S
+  src/qs8-igemm/gen/4x8-minmax-rndnu-aarch64-neon-mlal-lane-prfm-ld64.S
   src/qs8-igemm/gen/4x16-minmax-fp32-aarch64-neon-mlal-lane-cortex-a53.S
   src/qs8-igemm/gen/4x16-minmax-fp32-aarch64-neon-mlal-lane-ld64.S
   src/qs8-igemm/gen/4x16-minmax-fp32-aarch64-neon-mlal-lane-prfm-cortex-a53.S