[LoopFusion] Exit early if one of fusion candidate has guarded branch but the another has not

[StephenFan]

Summary

Fixes: https://github.com/llvm/llvm-project/issues/59024

[nikic]

llvm/test/Transforms/LoopFusion/guarded.ll

70

Please precommit the test regeneration.

[StephenFan]

Pre-commit test check regeneration.

[nikic]

Not familiar with this pass, but it looks reasonable.

llvm/lib/Transforms/Scalar/LoopFuse.cpp

899–902

nit: I'd prefer doing a (bool) cast or something, the ! != ! is pretty weird.

llvm/lib/Transforms/Scalar/LoopFuse.cpp

for (auto &CandidateSet : FusionCandidates) {
 
           if (!isAdjacent(*FC0, *FC1)) {
             LLVM_DEBUG(dbgs()
                        << "Fusion candidates are not adjacent. Not fusing.\n");
             reportLoopFusion<OptimizationRemarkMissed>(*FC0, *FC1, NonAdjacent);
             continue;
           }
 
-          if (!FC0->GuardBranch && FC1->GuardBranch) {
-            LLVM_DEBUG(dbgs() << "The second candidate is guarded while the "
-                                 "first one is not. Not fusing.\n");
+          if ((!FC0->GuardBranch && FC1->GuardBranch) ||
+              (FC0->GuardBranch && !FC1->GuardBranch)) {
+            LLVM_DEBUG(dbgs() << "The one of candidate is guarded while the "
+                                 "another one is not. Not fusing.\n");
             reportLoopFusion<OptimizationRemarkMissed>(
                 *FC0, *FC1, OnlySecondCandidateIsGuarded);
             continue;
           }
 
           // Ensure that FC0 and FC1 have identical guards.
           // If one (or both) are not guarded, this check is not necessary.
           if (FC0->GuardBranch && FC1->GuardBranch &&

llvm/test/Transforms/LoopFusion/guarded.ll

bb10:                                 ; preds = %bb5
   br label %bb14
 
 bb14:                                          ; preds = %bb10, %entry
   %cmp31 = icmp slt i64 0, %N
   br i1 %cmp31, label %bb8, label %bb12
 
 bb8:                              ; preds = %bb14
   br label %bb9
 
 bb9:                                        ; preds = %bb8, %bb9
   %i1.02 = phi i64 [ %inc14, %bb9 ], [ 0, %bb8 ]
   %sub7 = sub nsw i64 %i1.02, 3
   %add8 = add nsw i64 %i1.02, 3
   %mul9 = mul nsw i64 %sub7, %add8
   %rem10 = srem i64 %mul9, %i1.02
   %conv11 = trunc i64 %rem10 to i32
   %arrayidx12 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %i1.02

 
 for.second.exit:
   br label %for.end
 
 for.end:
   %j.lcssa = phi i64 [ 0, %for.second.guard ], [ %j.02, %for.second.exit ]
   ret i64 %j.lcssa
 }
+
+define void @pr59024() {
+; CHECK-LABEL: @pr59024(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[FOR_2_PREHEADER:%.*]], label [[FOR_1_PREHEADER:%.*]]
+; CHECK:       for.1.preheader:
+; CHECK-NEXT:    br label [[FOR_1:%.*]]
+; CHECK:       for.1:
+; CHECK-NEXT:    br i1 true, label [[FOR_2_PREHEADER_LOOPEXIT:%.*]], label [[FOR_1]]
+; CHECK:       for.2.preheader.loopexit:
+; CHECK-NEXT:    br label [[FOR_2_PREHEADER]]
+; CHECK:       for.2.preheader:
+; CHECK-NEXT:    br label [[FOR_2:%.*]]
+; CHECK:       for.2:
+; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[FOR_2]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br i1 false, label %for.2, label %for.1
+
+for.1:                                        ; preds = %for.body6, %entry
+  br i1 true, label %for.2, label %for.1
+
+for.2:                                       ; preds = %for.cond13, %for.body6, %entry
+  br i1 true, label %exit, label %for.2
+
+exit:                                          ; preds = %for.cond13
+  ret void
+}