Add a triple to the test to fix it on some hosts.
authorRafael Espindola <rafael.espindola@gmail.com>
Mon, 22 Jun 2015 15:44:20 +0000 (15:44 +0000)
committerRafael Espindola <rafael.espindola@gmail.com>
Mon, 22 Jun 2015 15:44:20 +0000 (15:44 +0000)
The slp vectorizer doesn't optimize this case in 32 bits.

Fixes PR23453.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240289 91177308-0d34-0410-b5e6-96231b3b80d8

test/tools/gold/slp-vectorize.ll

index a75f0b3..102f5fa 100644 (file)
@@ -8,7 +8,7 @@
 ; test that the vectorizer is run.
 ; CHECK: fadd <4 x float>
 
-target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
 
 define void @f(float* nocapture %x) {
   %tmp = load float, float* %x, align 4