commit 3b8e566e0f1900a2972bf8b0524a9b2faadc69ee
author Frank Barchard <fbarchard@google.com> Mon Apr 20 12:12:53 2020 -0700
committer XNNPACK Team <xnnpack-github-robot@google.com> Mon Apr 20 12:13:25 2020 -0700
tree 3c77ba613233722dbffff425ef34cfb35ed0bd25
parent 9b885f7ca22047ef88060473322b968dbcb93a8c

F16 8x8 GEMM ld64 microkernels

1.8% faster on Cortex A76
Was f16_gemm_6x8__aarch64_neonfp16arith_ld64 6410920
Now f16_gemm_8x8__aarch64_neonfp16arith_ld64 6294429

PiperOrigin-RevId: 307449974

BUILD.bazel

diff --git a/BUILD.bazel b/BUILD.bazel
index 1b2323c..ab09072 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -1471,9 +1471,11 @@
     "src/f16-gemm/gen/1x8-minmax-aarch64-neonfp16arith-ld64.S",
     "src/f16-gemm/gen/4x8-minmax-aarch64-neonfp16arith-ld64.S",
     "src/f16-gemm/gen/6x8-minmax-aarch64-neonfp16arith-ld64.S",
+    "src/f16-gemm/gen/8x8-minmax-aarch64-neonfp16arith-ld64.S",
     "src/f16-gemm/gen-inc/1x8inc-minmax-aarch64-neonfp16arith-ld64.S",
     "src/f16-gemm/gen-inc/4x8inc-minmax-aarch64-neonfp16arith-ld64.S",
     "src/f16-gemm/gen-inc/6x8inc-minmax-aarch64-neonfp16arith-ld64.S",
+    "src/f16-gemm/gen-inc/8x8inc-minmax-aarch64-neonfp16arith-ld64.S",
     "src/f32-dwconv/up4x9-minmax-aarch64-neonfma-cortex-a55.S",
     "src/f32-dwconv/up4x9-minmax-aarch64-neonfma.S",
     "src/f32-gemm/gen/1x8-minmax-aarch64-neonfma-ld64.S",