projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 60dda38) | patch
Use sse_load_f32/64 for scalar FMA3 intrinsic patterns instead of 128-bit loads to...
authorCraig Topper <craig.topper@gmail.com>
Sun, 3 Jun 2012 01:40:43 +0000 (01:40 +0000)
committerCraig Topper <craig.topper@gmail.com>
Sun, 3 Jun 2012 01:40:43 +0000 (01:40 +0000)
commit57ae246a6a5a2bce6997cdf1d1e41b568109a469
treef1a4653525f9150821c6a39782436e20e3622c39tree | snapshot
parent60dda3800861d31bf67130a16f1d62105254851acommit | diff
Use sse_load_f32/64 for scalar FMA3 intrinsic patterns instead of 128-bit loads to match instruction behavior.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@157895 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/X86/X86InstrFMA.td diff | blob | history
test/CodeGen/X86/fma3-intrinsics.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors