[llvm] X86: implement lowerings for shuffles on `bf16` element type. (PR #76076)

Phoebe Wang via llvm-commits llvm-commits at lists.llvm.org
Wed Dec 20 16:56:13 PST 2023


================
@@ -17112,6 +17149,9 @@ static SDValue lower512BitShuffle(const SDLoc &DL, ArrayRef<int> Mask,
     return lowerV8F64Shuffle(DL, Mask, Zeroable, V1, V2, Subtarget, DAG);
   case MVT::v16f32:
     return lowerV16F32Shuffle(DL, Mask, Zeroable, V1, V2, Subtarget, DAG);
+  case MVT::v32f16:
+  case MVT::v32bf16:
+    return lowerV32F16Shuffle(DL, Mask, Zeroable, V1, V2, Subtarget, DAG);
----------------
phoebewang wrote:

You just need to add `|| VT == MVT::v32bf16` to line 17132.

https://github.com/llvm/llvm-project/pull/76076


More information about the llvm-commits mailing list