diff options
| author | Duncan Sands <baldrick@free.fr> | 2011-09-17 16:49:39 +0000 |
|---|---|---|
| committer | Duncan Sands <baldrick@free.fr> | 2011-09-17 16:49:39 +0000 |
| commit | 6bcd2196e5febb97efe188d0d6de403c577d96fc (patch) | |
| tree | b147a99b1f7f3b65365ba8535f8466ab53f5d766 /test/CodeGen/X86/sse41-blend.ll | |
| parent | 2c693dc1261d8b8d92b5c96cf00ec873dde86bb2 (diff) | |
| download | llvm-6bcd2196e5febb97efe188d0d6de403c577d96fc.tar.gz llvm-6bcd2196e5febb97efe188d0d6de403c577d96fc.tar.bz2 llvm-6bcd2196e5febb97efe188d0d6de403c577d96fc.tar.xz | |
Synthesize x86 max/min instructions also for vectors (i.e. produce
maxps and maxpd). This broke the sse41-blend.ll testcase by causing
maxpd to be produced rather than a cmp+blend pair, which is the reason
I tweaked it. Gives a small speedup on doduc with dragonegg when the
GCC vectorizer is used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139986 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/sse41-blend.ll')
| -rw-r--r-- | test/CodeGen/X86/sse41-blend.ll | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/test/CodeGen/X86/sse41-blend.ll b/test/CodeGen/X86/sse41-blend.ll index a21416b7a8..78604a0e96 100644 --- a/test/CodeGen/X86/sse41-blend.ll +++ b/test/CodeGen/X86/sse41-blend.ll @@ -73,10 +73,10 @@ define <2 x double> @A(<2 x double> %x, <2 x double> %y) { ; CHECK: B define <2 x double> @B(<2 x double> %x, <2 x double> %y) { - ; CHECK: cmpltpd + ; CHECK: cmpnlepd ; CHECK: blendvpd - %max_is_x = fcmp ogt <2 x double> %x, %y - %max = select <2 x i1> %max_is_x, <2 x double> %x, <2 x double> %y - ret <2 x double> %max + %min_is_x = fcmp ult <2 x double> %x, %y + %min = select <2 x i1> %min_is_x, <2 x double> %x, <2 x double> %y + ret <2 x double> %min } |