}
static bool MayFoldVectorLoad(SDValue V) {
- while (V.hasOneUse() && V.getOpcode() == ISD::BITCAST)
+ if (V.hasOneUse() && V.getOpcode() == ISD::BITCAST)
V = V.getOperand(0);
-
if (V.hasOneUse() && V.getOpcode() == ISD::SCALAR_TO_VECTOR)
V = V.getOperand(0);
if (V.hasOneUse() && V.getOpcode() == ISD::BUILD_VECTOR &&
V.getNumOperands() == 2 && V.getOperand(1).getOpcode() == ISD::UNDEF)
// BUILD_VECTOR (load), undef
V = V.getOperand(0);
-
- return MayFoldLoad(V);
+ if (MayFoldLoad(V))
+ return true;
+ return false;
}
// FIXME: the version above should always be used. Since there's
-; RUN: llc < %s -march=x86 -mattr=+avx | FileCheck %s
+; RUN: llc < %s -march=x86 -mattr=sse41 -o %t
+; RUN: grep pshufhw %t | grep -- -95 | count 1
+; RUN: grep shufps %t | count 1
+; RUN: not grep pslldq %t
-; CHECK: test
; Test case when creating pshufhw, we incorrectly set the higher order bit
; for an undef,
define void @test(<8 x i16>* %dest, <8 x i16> %in) nounwind {
entry:
-; CHECK-NOT: vmovaps
-; CHECK: vmovlpd
-; CHECK: vpshufhw $-95
%0 = load <8 x i16>* %dest
%1 = shufflevector <8 x i16> %0, <8 x i16> %in, <8 x i32> < i32 0, i32 1, i32 2, i32 3, i32 13, i32 undef, i32 14, i32 14>
store <8 x i16> %1, <8 x i16>* %dest
ret void
-}
+}
-; CHECK: test2
; A test case where we shouldn't generate a punpckldq but a pshufd and a pslldq
define void @test2(<4 x i32>* %dest, <4 x i32> %in) nounwind {
entry:
-; CHECK-NOT: pslldq
-; CHECK: shufps
%0 = shufflevector <4 x i32> %in, <4 x i32> <i32 0, i32 0, i32 0, i32 0>, <4 x i32> < i32 undef, i32 5, i32 undef, i32 2>
store <4 x i32> %0, <4 x i32>* %dest
ret void