Normalize splat 256bit vectors with 8 elements.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@168600 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 276859b..f08477c 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -6689,7 +6689,7 @@
// Handle splats by matching through known shuffle masks
if ((Size == 128 && NumElem <= 4) ||
- (Size == 256 && NumElem < 8))
+ (Size == 256 && NumElem <= 8))
return SDValue();
// All remaning splats are promoted to target supported vector shuffles.
diff --git a/test/CodeGen/X86/avx-splat.ll b/test/CodeGen/X86/avx-splat.ll
index 94bcddd..5ad7523 100644
--- a/test/CodeGen/X86/avx-splat.ll
+++ b/test/CodeGen/X86/avx-splat.ll
@@ -47,9 +47,9 @@
; shuffle (scalar_to_vector (load (ptr + 4))), undef, <0, 0, 0, 0>
; To:
; shuffle (vload ptr)), undef, <1, 1, 1, 1>
-; CHECK: vmovaps
+; CHECK: vmovdqa
+; CHECK-NEXT: vpshufd $-1
; CHECK-NEXT: vinsertf128 $1
-; CHECK-NEXT: vpermilps $-1
define <8 x float> @funcE() nounwind {
allocas:
%udx495 = alloca [18 x [18 x float]], align 32
@@ -75,8 +75,8 @@
ret <8 x float> %load_broadcast12281250
}
-; CHECK: vinsertf128 $1
-; CHECK-NEXT: vpermilps $0
+; CHECK: vpshufd $0
+; CHECK-NEXT: vinsertf128 $1
define <8 x float> @funcF(i32 %val) nounwind {
%ret6 = insertelement <8 x i32> undef, i32 %val, i32 6
%ret7 = insertelement <8 x i32> %ret6, i32 %val, i32 7
@@ -84,8 +84,8 @@
ret <8 x float> %tmp
}
-; CHECK: vinsertf128 $1
-; CHECK-NEXT: vpermilps $0
+; CHECK: vpermilps $0
+; CHECK-NEXT: vinsertf128 $1
define <8 x float> @funcG(<8 x float> %a) nounwind uwtable readnone ssp {
entry:
%shuffle = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -93,8 +93,8 @@
}
; CHECK: vextractf128 $1
-; CHECK-NEXT: vinsertf128 $1
; CHECK-NEXT: vpermilps $85
+; CHECK-NEXT: vinsertf128 $1
define <8 x float> @funcH(<8 x float> %a) nounwind uwtable readnone ssp {
entry:
%shuffle = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>