[LoopCacheAnalysis] Fix crash after #164798

[kasuga-fj]

Fix the assertion failure after #164798. The issue is that the comparison Sizes.back() == ElementSize can fail when their types are different. We should cast them to the wider type before the comparison.

[llvmbot]

@llvm/pr-subscribers-llvm-analysis

Author: Ryotaro Kasuga (kasuga-fj)

Changes

Fix the assertion failure after #164798. The issue is that the comparison Sizes.back() == ElementSize can fail when their types are different. We should cast them to the wider type before the comparison.

---

Full diff: https://github.com/llvm/llvm-project/pull/169486.diff

2 Files Affected:

• (modified) llvm/lib/Analysis/LoopCacheAnalysis.cpp (+9-1)
• (added) llvm/test/Analysis/LoopCacheAnalysis/crash-after-pr164798.ll (+33)

diff --git a/llvm/lib/Analysis/LoopCacheAnalysis.cpp b/llvm/lib/Analysis/LoopCacheAnalysis.cpp
index e0e2be8e35929..3bba2e8c0d8ad 100644
--- a/llvm/lib/Analysis/LoopCacheAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopCacheAnalysis.cpp
@@ -368,8 +368,16 @@ bool IndexedReference::tryDelinearizeFixedSize(
   // the load/store instruction being analyzed. It is not needed for further
   // analysis.
   // TODO: Maybe this property should be enforced in delinearizeFixedSizeArray.
+#ifndef NDEBUG
   assert(!Sizes.empty() && Subscripts.size() == Sizes.size() &&
-         Sizes.back() == ElementSize && "Unexpected delinearization result");
+         "Inconsistent length of Sizes and Subscripts");
+  Type *WideTy =
+      SE.getWiderType(ElementSize->getType(), Sizes.back()->getType());
+  const SCEV *ElemSizeExt = SE.getNoopOrZeroExtend(ElementSize, WideTy);
+  const SCEV *LastSizeExt = SE.getNoopOrZeroExtend(Sizes.back(), WideTy);
+  assert(ElemSizeExt == LastSizeExt && "Unexpected last element of Sizes");
+#endif
+
   Sizes.pop_back();
   return true;
 }
diff --git a/llvm/test/Analysis/LoopCacheAnalysis/crash-after-pr164798.ll b/llvm/test/Analysis/LoopCacheAnalysis/crash-after-pr164798.ll
new file mode 100644
index 0000000000000..e6b6d1753adb7
--- /dev/null
+++ b/llvm/test/Analysis/LoopCacheAnalysis/crash-after-pr164798.ll
@@ -0,0 +1,33 @@
+; RUN: opt < %s -passes='print<loop-cache-cost>' -disable-output
+
+; Ensure no crash happens after PR #164798
+
+target datalayout = "p21:32:16"
+
+define i16 @f() {
+entry:
+  br label %for.cond1.preheader
+
+for.cond1.preheader:
+  %i.02 = phi i16 [ 0, %entry ], [ %inc8, %for.cond.cleanup3 ]
+  %idxprom = zext i16 %i.02 to i32
+  %arrayidx = getelementptr [18 x i16], ptr addrspace(21) null, i32 %idxprom
+  br label %for.body4
+
+for.cond.cleanup:
+  ret i16 0
+
+for.cond.cleanup3:
+  %inc8 = add i16 %i.02, 1
+  %exitcond3.not = icmp eq i16 %inc8, 0
+  br i1 %exitcond3.not, label %for.cond.cleanup, label %for.cond1.preheader
+
+for.body4:
+  %j.01 = phi i16 [ 0, %for.cond1.preheader ], [ %inc.2, %for.body4 ]
+  %idxprom5 = zext i16 %j.01 to i32
+  %arrayidx6 = getelementptr i16, ptr addrspace(21) %arrayidx, i32 %idxprom5
+  store i16 0, ptr addrspace(21) %arrayidx6, align 1
+  %inc.2 = add i16 %j.01, 1
+  %exitcond.not.2 = icmp eq i16 %inc.2, 18
+  br i1 %exitcond.not.2, label %for.cond.cleanup3, label %for.body4
+}

[mikaelholmen]

LGTM and I've verified that this solves the crash I saw with #164798

Thanks!