[PATCH] D134856: [AMDGPU] Add use check in v_fma combine.

[Thomas Symalla via Phabricator via llvm-commits]

tsymalla created this revision.
tsymalla added reviewers: foad, piotr.
Herald added subscribers: kosarev, ecnelises, kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, jvesely, kzhuravl.
Herald added a project: All.
tsymalla requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.

In D132837 <https://reviews.llvm.org/D132837>, an existing v_fma combine was extended to regard nested
fma instructions. Originally, the inner FMA was checked for being used
only once. In its current state, this check is missing, which causes
some regressions.

In this patch, this check was added.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D134856

Files:
  llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
  llvm/test/CodeGen/AMDGPU/dagcombine-fma-fmad.ll


Index: llvm/test/CodeGen/AMDGPU/dagcombine-fma-fmad.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/dagcombine-fma-fmad.ll
+++ llvm/test/CodeGen/AMDGPU/dagcombine-fma-fmad.ll
@@ -240,10 +240,11 @@
 define amdgpu_ps float @fmac_sequence_innermost_fmul_multiple_use(float inreg %a, float inreg %b, float inreg %c, float inreg %d, float inreg %e, float inreg %f, float %g) #0 {
 ; GCN-LABEL: fmac_sequence_innermost_fmul_multiple_use:
 ; GCN:       ; %bb.0:
-; GCN-NEXT:    v_mac_f32_e64 v0, s2, s3
-; GCN-NEXT:    v_fmac_f32_e64 v0, s0, s1
-; GCN-NEXT:    v_fma_f32 v1, s5, s4, v0
-; GCN-NEXT:    v_fmac_f32_e32 v0, s5, v1
+; GCN-NEXT:    v_mul_f32_e64 v1, s2, s3
+; GCN-NEXT:    v_fmac_f32_e64 v1, s0, s1
+; GCN-NEXT:    v_fma_f32 v2, s5, s4, v1
+; GCN-NEXT:    v_fmac_f32_e32 v1, s5, v2
+; GCN-NEXT:    v_add_f32_e32 v0, v1, v0
 ; GCN-NEXT:    ; return to shader part epilog
   %t0 = fmul fast float %a, %b
   %t1 = fmul fast float %c, %d
Index: llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
===================================================================
--- llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -14300,7 +14300,7 @@
     }
 
     SDValue TmpFMA = FMA;
-    while (E && isFusedOp(TmpFMA)) {
+    while (E && isFusedOp(TmpFMA) && TmpFMA.hasOneUse()) {
       SDValue FMul = TmpFMA->getOperand(2);
       if (FMul.getOpcode() == ISD::FMUL && FMul.hasOneUse()) {
         SDValue C = FMul.getOperand(0);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D134856.463777.patch
Type: text/x-patch
Size: 1520 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220929/d008c4dd/attachment.bin>