RFR (M): 7023898: Intrinsify AtomicLongFieldUpdater.getAndIncrement() (original) (raw)

Vladimir Kozlov vladimir.kozlov at oracle.com
Mon Sep 17 09:46:56 PDT 2012

My fault, missed it when I added MulVI case. I can add the break into my vectors safepoint changes you reviewed last week. Do you agree?

Thanks, Vladimir

Roland Westrelin wrote:

When I merged this with the current code base, I noticed a missing break in Matcher::matchrulesupported() on x86. See below.

Roland. diff --git a/src/cpu/x86/vm/x86.ad b/src/cpu/x86/vm/x86.ad --- a/src/cpu/x86/vm/x86.ad +++ b/src/cpu/x86/vm/x86.ad @@ -509,10 +509,18 @@ case OpPopCountL: if (!UsePopCountInstruction) return false; + break; case OpMulVI: if ((UseSSE < 4) && (UseAVX < 1)) // only with SSE41 or AVX return false; break; + case OpCompareAndSwapL: +#ifdef LP64 + case OpCompareAndSwapP: +#endif + if (!VMVersion::supportscx8()) + return false; + break; } return true; // Per default match rules are supported.

More information about the hotspot-compiler-dev mailing list