[DFAJumpThreading] Handle circular determinator

[XChy]

Fixes the buildbot failure in #78134 (comment)
When we meet the path with single determinator, the determinator actually takes itself as a predecessor. Thus, we need to let Prev be the determinator when PathBBs has only one element.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: XChy (XChy)

Changes

Fixes the buildbot failure in #78134 (comment)
When we meet the path with single determinator, the determinator actually takes itself as a predecessor. Thus, we need to let Prev be the determinator when PathBBs has only one element.

---

Full diff: https://github.com/llvm/llvm-project/pull/78177.diff

2 Files Affected:

• (modified) llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp (+3-2)
• (modified) llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll (+35)

diff --git a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
index 0219f65618d8d8..85d4065286e41f 100644
--- a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
+++ b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
@@ -910,8 +910,9 @@ struct TransformDFA {
       PathBBs.pop_front();
 
     auto DetIt = llvm::find(PathBBs, Determinator);
-    auto Prev = std::prev(DetIt);
-    BasicBlock *PrevBB = *Prev;
+    // When there is only one BB in PathBBs, the determinator takes itself as a
+    // direct predecessor.
+    BasicBlock *PrevBB = PathBBs.size() == 1 ? *DetIt : *std::prev(DetIt);
     for (auto BBIt = DetIt; BBIt != PathBBs.end(); BBIt++) {
       BasicBlock *BB = *BBIt;
       BlocksToClean.insert(BB);
diff --git a/llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll b/llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll
index f40d4853d8a2fb..7cc82386aa1714 100644
--- a/llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll
+++ b/llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll
@@ -267,3 +267,38 @@ entry:
 end:
   ret void
 }
+
+define void @self-reference(i1 %c) {
+; CHECK-LABEL: @self-reference(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[C:%.*]], label [[DOTSPLIT_PREHEADER:%.*]], label [[DOTSPLIT_PREHEADER]]
+; CHECK:       .split.preheader:
+; CHECK-NEXT:    br label [[DOTSPLIT:%.*]]
+; CHECK:       .split:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ 0, [[DOTSPLIT_PREHEADER]] ]
+; CHECK-NEXT:    switch i32 [[TMP0]], label [[END:%.*]] [
+; CHECK-NEXT:      i32 -1, label [[END]]
+; CHECK-NEXT:      i32 0, label [[DOTSPLIT_JT4294967295:%.*]]
+; CHECK-NEXT:    ]
+; CHECK:       .split.jt4294967295:
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ -1, [[DOTSPLIT]] ]
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br i1 %c, label %.split.preheader, label %.split.preheader
+
+.split.preheader:
+  br label %.split
+
+.split:
+  %0 = phi i32 [ 0, %.split.preheader ], [ -1, %.split ]
+  switch i32 %0, label %end [
+  i32 -1, label %end
+  i32 0, label %.split
+  ]
+
+end:
+  ret void
+}

[steelannelida]

I've checked this PR - it indeed fixes the asan failure in our setup as well. Would anyone mind merging if I merge it?

[nikic]

LGTM

[nikic]

Is this split preheader somehow relevant?

[XChy]

43414e7 reduces it now.

[vitalybuka]

I am going to merge to fix the bot.