[llvm] e32374e - [X86][SSE] canonicalizeShuffleMaskWithHorizOp - add TODO for better 256/512-bit shuffle+hop folding support. NFC.

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2021-05-10T18:43:16+01:00
New Revision: e32374ed5cb27494c67817f12d1e1cec05486f40

URL: https://github.com/llvm/llvm-project/commit/e32374ed5cb27494c67817f12d1e1cec05486f40
DIFF: https://github.com/llvm/llvm-project/commit/e32374ed5cb27494c67817f12d1e1cec05486f40.diff

LOG: [X86][SSE] canonicalizeShuffleMaskWithHorizOp - add TODO for better 256/512-bit shuffle+hop folding support. NFC.

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ISelLowering.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index dce05267267d..72ad5ff70d44 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -36228,6 +36228,7 @@ static SDValue canonicalizeShuffleMaskWithHorizOp(
   // See if we can remove the shuffle by resorting the HOP chain so that
   // the HOP args are pre-shuffled.
   // TODO: Generalize to any sized/depth chain.
+  // TODO: Add support for 256/512-bit vectors.
   // TODO: Add support for PACKSS/PACKUS.
   if (isHoriz && NumEltsPerLane == 4 && VT0.is128BitVector() &&
       shouldUseHorizontalOp(Ops.size() == 1, DAG, Subtarget)) {