1 files changed, 4 insertions, 0 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 57bf86e051f..1ca5c6acc6c 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -13125,6 +13125,10 @@ static SDValue lowerV16I8VectorShuffle(const SDLoc &DL, ArrayRef<int> Mask,
             DL, MVT::v16i8, V1, V2, Mask, Subtarget, DAG))
       return Broadcast;
 
+    if (SDValue V =
+            lowerVectorShuffleWithUNPCK(DL, MVT::v16i8, Mask, V1, V2, DAG))
+      return V;
+
     // Check whether we can widen this to an i16 shuffle by duplicating bytes.
     // Notably, this handles splat and partial-splat shuffles more efficiently.
     // However, it only makes sense if the pre-duplication shuffle simplifies