diff options
| author | Ben Gamari <ben@smart-cactus.org> | 2023-01-16 12:21:30 -0500 |
|---|---|---|
| committer | Ben Gamari <ben@smart-cactus.org> | 2023-01-16 12:27:41 -0500 |
| commit | af54cf7ddfdd68aa3e628aa63835fb77a7904e1e (patch) | |
| tree | c9a081f2a748deabbb6b818b76163f6e5cfc9298 | |
| parent | c79b2b65fff93b4f4e38e5d124f0cb8c6ef704c0 (diff) | |
| download | haskell-wip/T22764.tar.gz | |
nativeGen/X86: MFENCE is unnecessary for release semanticswip/T22764
In #22764 a user noticed that a program implementing a simple atomic
counter via an STRef regressed significantly due to the introduction of
necessary atomic operations in the MutVar# primops (#22468). This
regression was caused by a bug in the NCG, which emitted an unnecessary
MFENCE instruction for a release-ordered atomic write. MFENCE is rather
only needed to achieve sequentially consistent ordering.
Fixes #22764.
| -rw-r--r-- | compiler/GHC/CmmToAsm/X86/CodeGen.hs | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/compiler/GHC/CmmToAsm/X86/CodeGen.hs b/compiler/GHC/CmmToAsm/X86/CodeGen.hs index d407a8a86a..bc0135b920 100644 --- a/compiler/GHC/CmmToAsm/X86/CodeGen.hs +++ b/compiler/GHC/CmmToAsm/X86/CodeGen.hs @@ -3936,7 +3936,7 @@ genAtomicWrite width mord addr val = do code <- assignMem_IntCode (intFormat width) addr val let needs_fence = case mord of MemOrderSeqCst -> True - MemOrderRelease -> True + MemOrderRelease -> False MemOrderAcquire -> pprPanic "genAtomicWrite: acquire ordering on write" empty MemOrderRelaxed -> False return $ if needs_fence then code `snocOL` MFENCE else code |