summaryrefslogtreecommitdiffstats
path: root/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
diff options
context:
space:
mode:
authorChandler Carruth <chandlerc@gmail.com>2014-09-27 04:42:39 +0000
committerChandler Carruth <chandlerc@gmail.com>2014-09-27 04:42:39 +0000
commit81e6b29f03af3b7b33a6fb01500488c3a8ced811 (patch)
tree07064e1869615f55d9fb4d0e2ea5df8d56d82b49 /llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
parent5996da203214fce8ae313bc0a86a1413de2749c5 (diff)
downloadbcm5719-llvm-81e6b29f03af3b7b33a6fb01500488c3a8ced811.tar.gz
bcm5719-llvm-81e6b29f03af3b7b33a6fb01500488c3a8ced811.zip
[x86] Flip the sentinel values used in the target shuffle mask decoding
to significantly more sane sentinels. Notably, everywhere else in the backend's representation of shuffles uses '-1' to represent undef. The target shuffle masks really shouldn't diverge from that, especially as in a few places they are manipulated by shared code. This causes us to lose some undef lanes in various test masks. I want to get these back, but technically it isn't invalid and there are a *lot* of bugs here so I want to try to establish a saner baseline for fixing some of the bugs by aligning the specific senitnel values used. llvm-svn: 218561
Diffstat (limited to 'llvm/lib/Target/X86/Utils/X86ShuffleDecode.h')
-rw-r--r--llvm/lib/Target/X86/Utils/X86ShuffleDecode.h2
1 files changed, 1 insertions, 1 deletions
diff --git a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
index ece895d77cc..af26d0a509b 100644
--- a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -26,7 +26,7 @@ namespace llvm {
class Constant;
class MVT;
-enum { SM_SentinelZero = -1, SM_SentinelUndef = -2 };
+enum { SM_SentinelUndef = -1, SM_SentinelZero = -2 };
void DecodeINSERTPSMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
OpenPOWER on IntegriCloud