[x86] Fix another bug hit when bootstrapping with the new shuffle

lowering.

For maximum irony, I had already discovered this bug, diagnosed it, and
left FIXMEs about it in the test cases. =[ I just failed to go back over
those until after i had reduced a bootstrap miscompile down to a single
TU, stared at the assembly for an hour, and figured out the bug. Again.

Oh well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211955 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index fa80717f99479621a10d63e7c88a3476482f3608..b77a7fee0b717796b179f49889905f77913ff927 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -7105,7 +7105,7 @@ static SDValue lowerV4F32VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
 
       // Now we do a normal shuffle of V1 by giving V1 as both operands to
       // a blend.
-      HighV = V1;
+      LowV = HighV = V1;
       NewMask[0] = Mask[0] < 4 ? 0 : 2;
       NewMask[1] = Mask[0] < 4 ? 2 : 0;
       NewMask[2] = Mask[2] < 4 ? 1 : 3;

diff --git a/test/CodeGen/X86/vector-shuffle-128-v4.ll b/test/CodeGen/X86/vector-shuffle-128-v4.ll
index d7899f0db47f4f72f154af5baea9b2d8f485f14a..7d496fa19f15574b539f3fd3ef794a9bc97c8bab 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-128-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-128-v4.ll
@@ -146,10 +146,8 @@ define <4 x i32> @shuffle_v4i32_0145(<4 x i32> %a, <4 x i32> %b) {
 }
 define <4 x i32> @shuffle_v4i32_0451(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-SSE2-LABEL: @shuffle_v4i32_0451
-; CHECK-SSE2:         movaps %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*}} # xmm2 = xmm2[0,1],xmm1[0,1]
-; FIXME: This is wrong!!! xmm0 = xmm2[0,2],xmm2[3,1] would be correct....
-; CHECK-SSE2-NEXT:    shufps {{.*}} # xmm0 = xmm0[0,2],xmm2[3,1]
+; CHECK-SSE2:         shufps {{.*}} # xmm0 = xmm0[0,1],xmm1[0,1]
+; CHECK-SSE2-NEXT:    shufps {{.*}} # xmm0 = xmm0[0,2,3,1]
 ; CHECK-SSE2-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 4, i32 5, i32 1>
   ret <4 x i32> %shuffle
@@ -164,10 +162,8 @@ define <4 x i32> @shuffle_v4i32_4501(<4 x i32> %a, <4 x i32> %b) {
 }
 define <4 x i32> @shuffle_v4i32_4015(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-SSE2-LABEL: @shuffle_v4i32_4015
-; CHECK-SSE2:         movaps %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*}} # xmm2 = xmm2[0,1],xmm1[0,1]
-; FIXME: This is wrong!!! xmm0 = xmm2[0,2],xmm2[3,1] would be correct....
-; CHECK-SSE2-NEXT:    shufps {{.*}} # xmm0 = xmm0[2,0],xmm2[1,3]
+; CHECK-SSE2:         shufps {{.*}} # xmm0 = xmm0[0,1],xmm1[0,1]
+; CHECK-SSE2-NEXT:    shufps {{.*}} # xmm0 = xmm0[2,0,1,3]
 ; CHECK-SSE2-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 4, i32 0, i32 1, i32 5>
   ret <4 x i32> %shuffle