[llvm] relaxed simd fma (PR #147487)

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Tue Jul 8 02:12:51 PDT 2025


================
@@ -3412,6 +3418,37 @@ static SDValue performSETCCCombine(SDNode *N,
   return SDValue();
 }
 
+static SDValue performFAddCombine(SDNode *N, SelectionDAG &DAG) {
+  assert(N->getOpcode() == ISD::FADD);
+  using namespace llvm::SDPatternMatch;
+  if (!N->getFlags().hasFastMath())
+    return SDValue();
----------------
arsenm wrote:

This does not require all fast math flags

https://github.com/llvm/llvm-project/pull/147487


More information about the llvm-commits mailing list