[llvm] [AArch64] Fix #94909: Optimize vector fmul(sitofp(x), 0.5) -> scvtf(x, 2) (PR #141480)
JP Hafer via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 24 08:19:28 PDT 2025
================
@@ -3952,6 +3960,129 @@ static bool checkCVTFixedPointOperandWithFBits(SelectionDAG *CurDAG, SDValue N,
return true;
}
+static bool checkCVTFixedPointOperandWithFBitsForVectors(SelectionDAG *CurDAG,
+ SDValue N,
+ SDValue &FixedPos,
+ unsigned FloatWidth,
+ bool isReciprocal) {
+
+ // N must be a bitcast/nvcast of a vector float type.
+ if (!((N.getOpcode() == ISD::BITCAST ||
+ N.getOpcode() == AArch64ISD::NVCAST) &&
+ N.getValueType().isVector() && N.getValueType().isFloatingPoint())) {
----------------
jph-13 wrote:
updated as:
```
SDValue ImmediateNode;
// N must be a bitcast or nvcast
if (N.getOpcode() == ISD::BITCAST || N.getOpcode() == AArch64ISD::NVCAST) {
ImmediateNode = N.getOperand(0);
} else {
return false;
}
EVT NodeVT = N.getValueType();
EVT InputVT = ImmediateNode.getValueType();
// The input must me a vector of int but N must be a floating point
if (!InputVT.isVector() || !InputVT.isInteger() ||
!NodeVT.isFloatingPoint())
return false;
```
https://github.com/llvm/llvm-project/pull/141480
More information about the llvm-commits
mailing list