[llvm] [LV] Tidy up early bail outs related to runtime checks (PR #228077)

[via llvm-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-vectorizers

Author: David Sherwood (david-arm)

<details>
<summary>Changes</summary>

There was no debug or error reporting when bailing out due to proving at compile time that we will fail the SCEV checks. I've added debug and a report for this. Also, the early bail out for runtime memory checks has no corresponding test and I cannot find any test that breaks when I remove it. I also cannot expose this code path when building the LLVM test suite, so I've removed it as dead code.

---
Full diff: https://github.com/llvm/llvm-project/pull/228077.diff


2 Files Affected:

- (modified) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp (+6-4) 
- (modified) llvm/test/Transforms/LoopVectorize/AArch64/runtime-check-size-based-threshold.ll (+5-1) 


``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 45edf702d7092..c6edd933b17ff 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -7953,11 +7953,13 @@ bool LoopVectorizePass::processLoop(Loop *L) {
       // fail. In that case, the vector loop would never execute.
       using namespace llvm::PatternMatch;
       if (Checks.getSCEVChecks().first &&
-          match(Checks.getSCEVChecks().first, m_One()))
-        return false;
-      if (Checks.getMemRuntimeChecks().first &&
-          match(Checks.getMemRuntimeChecks().first, m_One()))
+          match(Checks.getSCEVChecks().first, m_One())) {
+        reportVectorizationFailure(
+            "runtime checks are known to fail, so we will never enter the "
+            "vector loop",
+            "RuntimeChecksNeverEnterVectorLoop", ORE, L);
         return false;
+      }
     }
 
     // Check if it is profitable to vectorize with runtime checks.
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/runtime-check-size-based-threshold.ll b/llvm/test/Transforms/LoopVectorize/AArch64/runtime-check-size-based-threshold.ll
index 9ded4ccc2822c..df4750abd6b4a 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/runtime-check-size-based-threshold.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/runtime-check-size-based-threshold.ll
@@ -1,7 +1,9 @@
 ; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-iphoneos -vectorizer-min-trip-count=8 \
 ; RUN:   -enable-epilogue-vectorization=false -S %s | FileCheck --check-prefixes=CHECK,DEFAULT %s
 ; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-iphoneos -vectorizer-min-trip-count=8 \
-; RUN:   -enable-epilogue-vectorization=false -vectorize-memory-check-threshold=1 -S %s | FileCheck --check-prefixes=CHECK,THRESHOLD %s
+; RUN:   -enable-epilogue-vectorization=false -vectorize-memory-check-threshold=1 -S %s \
+; RUN:   -debug-only=loop-vectorize 2> %t | FileCheck --check-prefixes=CHECK,THRESHOLD %s
+; RUN: cat %t | FileCheck --check-prefix=DEBUG %s
 
 ; Tests for loops with large numbers of runtime checks. Check that loops are
 ; vectorized, if the loop trip counts are large and the impact of the runtime
@@ -12,6 +14,8 @@
 ; CHECK-LABEL: define {{.*}} @test_tc_too_small
 ; CHECK-NOT: vector.memcheck
 ; CHECK-NOT: vector.body
+; DEBUG-LABEL: Checking a loop in 'test_tc_too_small'
+; DEBUG: LV: Not vectorizing: runtime checks are known to fail, so we will never enter the vector loop.
 define void @test_tc_too_small(ptr %ptr.1, ptr %ptr.2, ptr %ptr.3, ptr %ptr.4, i64 %off.1, i64 %off.2) {
 entry:
   br label %loop

``````````

</details>


https://github.com/llvm/llvm-project/pull/228077