[llvm] [AArch64][ISel] Fix interleave3 crash (PR #192046)
David Sherwood via llvm-commits
llvm-commits at lists.llvm.org
Tue Apr 14 08:11:26 PDT 2026
================
@@ -32322,6 +32322,63 @@ SDValue AArch64TargetLowering::LowerVECTOR_INTERLEAVE(SDValue Op,
return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, Op->getVTList(), Ops);
}
+ if (Op->getNumOperands() == 3 && (OpVT == MVT::v2f64 || OpVT == MVT::v2i64)) {
+ SDValue A = Op->getOperand(0);
+ SDValue B = Op->getOperand(1);
+ SDValue C = Op->getOperand(2);
+
+ // a0, a1 | b0, b1 | c0, c1
+ // ->
+ // a0, b0 | c0, a1 | b1, c1
+ SDValue Shuffle0 = DAG.getVectorShuffle(OpVT, DL, A, B, {0, 2});
+ SDValue Shuffle1 = DAG.getVectorShuffle(OpVT, DL, A, C, {2, 1});
+ SDValue Shuffle2 = DAG.getVectorShuffle(OpVT, DL, B, C, {1, 3});
+
+ return DAG.getMergeValues({Shuffle0, Shuffle1, Shuffle2}, DL);
+ }
+ if (Op->getNumOperands() == 3 && OpVT.isFixedLengthVector()) {
+ unsigned NElements = OpVT.getVectorNumElements();
+ SDValue A = Op->getOperand(0);
+ SDValue B = Op->getOperand(1);
+ SDValue C = Op->getOperand(2);
+
+ // Build two masks for two shuffles. The first mask shuffles elements from
+ // A and B, and with elements of C represented with -1 so the shuffle
+ // ignores them. The second shuffle then replaces -1 values with the actual
+ // indices from C.
+ SmallVector<int> ABMask;
+ for (unsigned E = 0; E < NElements; E++) {
+ for (unsigned F = 0; F < 3; F++) {
+ if (F == 2)
+ ABMask.push_back(-1);
+ else
+ ABMask.push_back(F * NElements + E);
+ }
+ }
+
+ std::vector<int> ABCMask;
+ for (unsigned F = 0, Idx = 0; F < 3; F++) {
+ for (unsigned E = 0; E < NElements; E++) {
+ if (ABMask[F * NElements + E] == -1)
+ ABCMask.push_back(NElements + Idx++);
+ else
+ ABCMask.push_back(Idx);
----------------
david-arm wrote:
As discussed privately, I think there is a problem here. When F=0, then for E=0,1 we push back the same index (0) to the ABC mask. That means in the shuffle below we're duplicating element 0 of `AB`, which leads to
```
[A0, A0, ...]
```
https://github.com/llvm/llvm-project/pull/192046
More information about the llvm-commits
mailing list