summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--llvm/lib/Target/X86/X86ISelLowering.cpp2
-rw-r--r--llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll26
2 files changed, 27 insertions, 1 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 09f0be0e246..c74a84665fb 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -21012,7 +21012,7 @@ static SDValue PerformTargetShuffleCombine(SDValue N, SelectionDAG &DAG,
// See if this reduces to a PSHUFD which is no more expensive and can
// combine with more operations. Note that it has to at least flip the
// dwords as otherwise it would have been removed as a no-op.
- if (Mask[0] == 2 && Mask[1] == 3 && Mask[2] == 0 && Mask[3]) {
+ if (Mask[0] == 2 && Mask[1] == 3 && Mask[2] == 0 && Mask[3] == 1) {
int DMask[] = {0, 1, 2, 3};
int DOffset = N.getOpcode() == X86ISD::PSHUFLW ? 0 : 2;
DMask[DOffset + 0] = DOffset + 1;
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
index 79eecf32281..ddd3914d2cb 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
@@ -99,6 +99,32 @@ define <8 x i16> @shuffle_v8i16_44440000(<8 x i16> %a, <8 x i16> %b) {
%shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 4, i32 4, i32 4, i32 0, i32 0, i32 0, i32 0>
ret <8 x i16> %shuffle
}
+define <8 x i16> @shuffle_v8i16_23016745(<8 x i16> %a, <8 x i16> %b) {
+; ALL-LABEL: @shuffle_v8i16_23016745
+; ALL: # BB#0:
+; ALL-NEXT: pshufd {{.*}} # xmm0 = xmm0[1,0,3,2]
+; ALL-NEXT: retq
+ %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5>
+ ret <8 x i16> %shuffle
+}
+define <8 x i16> @shuffle_v8i16_23026745(<8 x i16> %a, <8 x i16> %b) {
+; ALL-LABEL: @shuffle_v8i16_23026745
+; ALL: # BB#0:
+; ALL-NEXT: pshuflw {{.*}} # xmm0 = xmm0[2,3,0,2,4,5,6,7]
+; ALL-NEXT: pshufd {{.*}} # xmm0 = xmm0[0,1,3,2]
+; ALL-NEXT: retq
+ %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 3, i32 0, i32 2, i32 6, i32 7, i32 4, i32 5>
+ ret <8 x i16> %shuffle
+}
+define <8 x i16> @shuffle_v8i16_23016747(<8 x i16> %a, <8 x i16> %b) {
+; ALL-LABEL: @shuffle_v8i16_23016747
+; ALL: # BB#0:
+; ALL-NEXT: pshufd {{.*}} # xmm0 = xmm0[1,0,2,3]
+; ALL-NEXT: pshufhw {{.*}} # xmm0 = xmm0[0,1,2,3,6,7,4,7]
+; ALL-NEXT: retq
+ %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 7>
+ ret <8 x i16> %shuffle
+}
define <8 x i16> @shuffle_v8i16_75643120(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: @shuffle_v8i16_75643120
; SSE2: # BB#0:
OpenPOWER on IntegriCloud