[llvm] LSV: document hang reported in #37865 (PR #102479)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-transforms

Author: Ramkumar Ramachandra (artagnon)

<details>
<summary>Changes</summary>

LoadStoreVectorizer hangs on certain examples, when its reorder function goes into a cycle. Detect this cycle and explicitly forbid it, using an assert, and document the resulting crash in a test-case under AArch64.

-- 8< --
It is unclear if this code-path will ever be hit when running the full opt pipeline. Moreover, the fix is also very unclear to me.

---
Full diff: https://github.com/llvm/llvm-project/pull/102479.diff


2 Files Affected:

- (modified) llvm/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp (+2) 
- (added) llvm/test/Transforms/LoadStoreVectorizer/AArch64/pr37865.ll (+13) 


``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp b/llvm/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp
index c91911ecad7456..c35ea431296b70 100644
--- a/llvm/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp
@@ -216,6 +216,8 @@ void reorder(Instruction *I) {
       if (IM->getParent() != I->getParent())
         continue;
 
+      assert(IM != I && "Unexpected cycle while re-ordering instructions");
+
       if (!IM->comesBefore(I)) {
         InstructionsToMove.insert(IM);
         Worklist.push_back(IM);
diff --git a/llvm/test/Transforms/LoadStoreVectorizer/AArch64/pr37865.ll b/llvm/test/Transforms/LoadStoreVectorizer/AArch64/pr37865.ll
new file mode 100644
index 00000000000000..d77c7256bfadbd
--- /dev/null
+++ b/llvm/test/Transforms/LoadStoreVectorizer/AArch64/pr37865.ll
@@ -0,0 +1,13 @@
+; REQUIRES: asserts
+; RUN: not --crash opt -mtriple=aarch64 -passes=load-store-vectorizer -disable-output %s
+
+define i32 @load_cycle(ptr %x) {
+entry:
+  %gep.x.1 = getelementptr inbounds [2 x i32], ptr %x, i32 0, i32 1
+  %load.x.1 = load i32, ptr %gep.x.1
+  %rem = urem i32 %load.x.1, 1
+  %gep.x.2 = getelementptr inbounds [2 x i32], ptr %x, i32 %rem, i32 0
+  %load.x.2 = load i32, ptr %gep.x.2
+  %ret = add i32 %load.x.2, %load.x.1
+  ret i32 %ret
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/102479