diff options
author | Keichi Takahashi <keichi.t@me.com> | 2021-09-13 17:25:41 +0300 |
---|---|---|
committer | Copybara-Service <copybara-worker@google.com> | 2021-09-13 17:26:05 +0300 |
commit | c31af31ad62cc74650b04368cd9369a389d23248 (patch) | |
tree | a8a4c228138af3135d10c7efe417acd27edaa11a | |
parent | 34f1aa7f3b8712aee21883da774cf11020d09b38 (diff) |
Add missing volatile qualifier in Pack8bitRowMajorForNeonDotprod
I was getting incorrect results on some environments and this turned out to be the cause.
Closes https://github.com/google/ruy/pull/276
COPYBARA_INTEGRATE_REVIEW=https://github.com/google/ruy/pull/276 from keichi:add-missing-volatile e2d89fe29ce36510a08c704b603f513729713faf
PiperOrigin-RevId: 396351130
-rw-r--r-- | ruy/pack_arm.cc | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/ruy/pack_arm.cc b/ruy/pack_arm.cc index c337986..91f823b 100644 --- a/ruy/pack_arm.cc +++ b/ruy/pack_arm.cc @@ -1592,7 +1592,7 @@ void Pack8bitRowMajorForNeonDotprod(const void* src_ptr0, const void* src_ptr1, int packed_stride, std::int32_t* sums_ptr, int input_xor) { profiler::ScopeLabel label("Pack (kNeonDotprod, from row-major)"); - asm( + asm volatile( // clang-format off // Prefetch data. This was tuned on Cortex-A55-rev1 cores. RUY_PREFETCH_LOAD("prfm pldl1strm, [%[src_ptr0]]\n") |