[DAGCombiner][RISCV] Enable reassociation for VP_FMA in visitFADDForF…

…MACombine. Reviewed By: fakepaper56 Differential Revision: https://reviews.llvm.org/D149911
llvm · May 5, 2023 · fe9f557 · fe9f557
1 parent 3b9ed6e
commit fe9f557
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 2 deletions.
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -15233,10 +15233,10 @@ SDValue DAGCombiner::visitFADDForFMACombine(SDNode *N) {
     SDValue TmpFMA = FMA;
     while (E && isFusedOp(TmpFMA) && TmpFMA.hasOneUse()) {
       SDValue FMul = TmpFMA->getOperand(2);
-      if (FMul.getOpcode() == ISD::FMUL && FMul.hasOneUse()) {
+      if (matcher.match(FMul, ISD::FMUL) && FMul.hasOneUse()) {
         SDValue C = FMul.getOperand(0);
         SDValue D = FMul.getOperand(1);
-        SDValue CDE = DAG.getNode(PreferredFusedOpcode, SL, VT, C, D, E);
+        SDValue CDE = matcher.getNode(PreferredFusedOpcode, SL, VT, C, D, E);
         DAG.ReplaceAllUsesOfValueWith(FMul, CDE);
         // Replacing the inner FMul could cause the outer FMA to be simplified
         // away.

diff --git a/llvm/test/CodeGen/RISCV/rvv/fold-vp-fadd-and-vp-fmul.ll b/llvm/test/CodeGen/RISCV/rvv/fold-vp-fadd-and-vp-fmul.ll
@@ -57,3 +57,18 @@ define <vscale x 1 x double> @fma_nonvp(<vscale x 1 x double> %x, <vscale x 1 x
   %2 = call fast <vscale x 1 x double> @llvm.vp.fadd.nxv1f64(<vscale x 1 x double> %1, <vscale x 1 x double> %z, <vscale x 1 x i1> %m, i32 %vl)
   ret <vscale x 1 x double> %2
 }
+
+define <vscale x 1 x double> @fma_reassociate(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x double> %c, <vscale x 1 x double> %d, <vscale x 1 x double> %e, <vscale x 1 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: fma_reassociate:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, mu
+; CHECK-NEXT:    vfmadd.vv v11, v10, v12, v0.t
+; CHECK-NEXT:    vfmadd.vv v9, v8, v11, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v9
+; CHECK-NEXT:    ret
+  %1 = call fast <vscale x 1 x double> @llvm.vp.fmul.nxv1f64(<vscale x 1 x double> %a, <vscale x 1 x double> %b, <vscale x 1 x i1> %m, i32 %vl)
+  %2 = call fast <vscale x 1 x double> @llvm.vp.fmul.nxv1f64(<vscale x 1 x double> %c, <vscale x 1 x double> %d, <vscale x 1 x i1> %m, i32 %vl)
+  %3 = call fast <vscale x 1 x double> @llvm.vp.fadd.nxv1f64(<vscale x 1 x double> %1, <vscale x 1 x double> %2, <vscale x 1 x i1> %m, i32 %vl)
+  %4 = call fast <vscale x 1 x double> @llvm.vp.fadd.nxv1f64(<vscale x 1 x double> %3, <vscale x 1 x double> %e, <vscale x 1 x i1> %m, i32 %vl)
+  ret <vscale x 1 x double> %4
+}