Skip to content

[LoopIdiom] Update strlen idiom body loop condition to be clean up by LoopDeletion #134906

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Apr 11, 2025

Conversation

mustartt
Copy link
Member

@mustartt mustartt commented Apr 8, 2025

Fixes the case where subsequent passes were unable to find and delete the invariant loop left over by the strlen idiom conversion. Since loop-deletion only operate on computable loops, we can update the loop condition to something more easily picked up by loop-deletion

As pointed out in #134736

@llvmbot
Copy link
Member

llvmbot commented Apr 8, 2025

@llvm/pr-subscribers-llvm-transforms

Author: Henry Jiang (mustartt)

Changes

Fixes the case where subsequent passes were unable to find and delete the invariant loop left over by the strlen idiom conversion. Since loop-deletion only operate on computable loops, we can update the loop condition to something more easily picked up by loop-deletion

As pointed out in #134736


Full diff: https://github.com/llvm/llvm-project/pull/134906.diff

5 Files Affected:

  • (modified) llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp (+9)
  • (added) llvm/test/Transforms/LoopIdiom/strlen-cleanup.ll (+43)
  • (modified) llvm/test/Transforms/LoopIdiom/strlen.ll (+6-6)
  • (modified) llvm/test/Transforms/LoopIdiom/wcslen16.ll (+1-1)
  • (modified) llvm/test/Transforms/LoopIdiom/wcslen32.ll (+1-1)
diff --git a/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp b/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
index 4940dd8c47dde..d78a02381c68b 100644
--- a/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
@@ -1745,7 +1745,11 @@ bool LoopIdiomRecognize::recognizeAndInsertStrLen() {
     return false;
 
   BasicBlock *Preheader = CurLoop->getLoopPreheader();
+  BasicBlock *LoopBody = *CurLoop->block_begin();
   BasicBlock *LoopExitBB = CurLoop->getExitBlock();
+  BranchInst *LoopTerm = dyn_cast<BranchInst>(LoopBody->getTerminator());
+  assert(Preheader && LoopBody && LoopExitBB && LoopTerm &&
+         "Should be verified to be valid by StrlenVerifier");
 
   if (Verifier.OpWidth == 8) {
     if (DisableLIRP::Strlen)
@@ -1804,6 +1808,11 @@ bool LoopIdiomRecognize::recognizeAndInsertStrLen() {
   // up by later passes
   for (PHINode *PN : Cleanup)
     RecursivelyDeleteDeadPHINode(PN);
+
+  ConstantInt *NewLoopCond = LoopTerm->getSuccessor(0) == LoopBody
+                                 ? Builder.getFalse()
+                                 : Builder.getTrue();
+  LoopTerm->setCondition(NewLoopCond);
   SE->forgetLoop(CurLoop);
 
   ++NumStrLen;
diff --git a/llvm/test/Transforms/LoopIdiom/strlen-cleanup.ll b/llvm/test/Transforms/LoopIdiom/strlen-cleanup.ll
new file mode 100644
index 0000000000000..60b5ac6ce1d78
--- /dev/null
+++ b/llvm/test/Transforms/LoopIdiom/strlen-cleanup.ll
@@ -0,0 +1,43 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt < %s -S -passes=loop-idiom,loop-deletion | FileCheck %s
+
+define [2 x i64] @convert_null_terminated_to_bounded_ptr(ptr noundef %p) {
+; CHECK-LABEL: define [2 x i64] @convert_null_terminated_to_bounded_ptr(
+; CHECK-SAME: ptr noundef [[P:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[WCSLEN:%.*]] = call i64 @wcslen(ptr [[P]])
+; CHECK-NEXT:    br label %[[TERMINATED_BY_LOOP_END:.*]]
+; CHECK:       [[TERMINATED_BY_LOOP_END]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[P]], i64 [[WCSLEN]]
+; CHECK-NEXT:    [[TERMINATED_BY_UPPER:%.*]] = getelementptr i8, ptr [[TMP0]], i64 4
+; CHECK-NEXT:    [[TMP1:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT:    [[DOTFCA_0_INSERT:%.*]] = insertvalue [2 x i64] poison, i64 [[TMP1]], 0
+; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[TERMINATED_BY_UPPER]] to i64
+; CHECK-NEXT:    [[DOTFCA_1_INSERT:%.*]] = insertvalue [2 x i64] [[DOTFCA_0_INSERT]], i64 [[TMP2]], 1
+; CHECK-NEXT:    ret [2 x i64] [[DOTFCA_1_INSERT]]
+;
+entry:
+  br label %terminated_by.loop_cond
+
+terminated_by.loop_cond:                          ; preds = %terminated_by.loop_cond, %entry
+  %terminated_by.len.0 = phi i64 [ 0, %entry ], [ %terminated_by.new_len, %terminated_by.loop_cond ]
+  %0 = getelementptr inbounds i32, ptr %p, i64 %terminated_by.len.0
+  %terminated_by.elem = load i32, ptr %0, align 4
+  %terminted_by.check_terminator = icmp eq i32 %terminated_by.elem, 0
+  %terminated_by.new_len = add i64 %terminated_by.len.0, 1
+  br i1 %terminted_by.check_terminator, label %terminated_by.loop_end, label %terminated_by.loop_cond
+
+terminated_by.loop_end:                           ; preds = %terminated_by.loop_cond
+  %1 = getelementptr inbounds i32, ptr %p, i64 %terminated_by.len.0
+  %terminated_by.upper = getelementptr i8, ptr %1, i64 4
+  %2 = ptrtoint ptr %p to i64
+  %.fca.0.insert = insertvalue [2 x i64] poison, i64 %2, 0
+  %3 = ptrtoint ptr %terminated_by.upper to i64
+  %.fca.1.insert = insertvalue [2 x i64] %.fca.0.insert, i64 %3, 1
+  ret [2 x i64] %.fca.1.insert
+}
+
+!llvm.module.flags = !{!0}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+
diff --git a/llvm/test/Transforms/LoopIdiom/strlen.ll b/llvm/test/Transforms/LoopIdiom/strlen.ll
index c1141177c659f..3b9d33ebdc74e 100644
--- a/llvm/test/Transforms/LoopIdiom/strlen.ll
+++ b/llvm/test/Transforms/LoopIdiom/strlen.ll
@@ -24,7 +24,7 @@ define i64 @valid_basic_strlen(ptr %str) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[STR_ADDR_0]], align 1
 ; CHECK-NEXT:    [[CMP_NOT:%.*]] = icmp eq i8 [[TMP0]], 0
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr i8, ptr [[STR_ADDR_0]], i64 1
-; CHECK-NEXT:    br i1 [[CMP_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_COND]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[WHILE_COND]]
 ; CHECK:       [[WHILE_END]]:
 ; CHECK-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[SCEVGEP]] to i64
 ; CHECK-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[STR]] to i64
@@ -74,7 +74,7 @@ define i32 @valid_basic_strlen_rotated(ptr %str) {
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds nuw i8, ptr [[STR_ADDR_0]], i64 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i8, ptr [[INCDEC_PTR]], align 1
 ; CHECK-NEXT:    [[TOBOOL1_NOT:%.*]] = icmp eq i8 [[TMP2]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL1_NOT]], label %[[DO_END:.*]], label %[[DO_BODY]]
+; CHECK-NEXT:    br i1 true, label %[[DO_END:.*]], label %[[DO_BODY]]
 ; CHECK:       [[DO_END]]:
 ; CHECK-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[SCEVGEP1]] to i64
 ; CHECK-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[STR]] to i64
@@ -163,7 +163,7 @@ define dso_local void @valid_strlen_with_aux_indvar(ptr noundef %str, ptr nounde
 ; CHECK-NEXT:    [[INCDEC_PTR2]] = getelementptr inbounds nuw i8, ptr [[FOO_ADDR_011]], i64 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = load i8, ptr [[INCDEC_PTR]], align 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i8 [[TMP10]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY]]
 ; CHECK:       [[WHILE_END_LOOPEXIT]]:
 ; CHECK-NEXT:    br label %[[WHILE_END]]
 ; CHECK:       [[WHILE_END]]:
@@ -232,7 +232,7 @@ define i32 @valid_strlen_index(ptr %str) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i8 [[TMP0]], 0
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_COND]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[WHILE_COND]]
 ; CHECK:       [[WHILE_END]]:
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i64 [[STRLEN]] to i32
 ; CHECK-NEXT:    ret i32 [[TMP1]]
@@ -290,7 +290,7 @@ define dso_local void @valid_strlen_offset(ptr noundef %str) local_unnamed_addr
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i8, ptr [[STR_ADDR_0]], align 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i8 [[TMP4]], 0
 ; CHECK-NEXT:    [[INCDEC_PTR14]] = getelementptr inbounds nuw i8, ptr [[STR_ADDR_0]], i64 1
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_COND]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[WHILE_COND]]
 ; CHECK:       [[WHILE_END]]:
 ; CHECK-NEXT:    tail call void @use(ptr noundef nonnull [[SCEVGEP]])
 ; CHECK-NEXT:    br label %[[RETURN]]
@@ -377,7 +377,7 @@ define void @valid_nested_idiom(ptr %strs, i32 %n) {
 ; CHECK-NEXT:    [[INC]] = add nuw nsw i32 [[COUNT_08]], 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i8, ptr [[INCDEC_PTR]], align 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i8 [[TMP4]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY]]
 ; CHECK:       [[WHILE_END_LOOPEXIT]]:
 ; CHECK-NEXT:    br label %[[WHILE_END]]
 ; CHECK:       [[WHILE_END]]:
diff --git a/llvm/test/Transforms/LoopIdiom/wcslen16.ll b/llvm/test/Transforms/LoopIdiom/wcslen16.ll
index 8be51869f1f2e..cd8a31f4af3dd 100644
--- a/llvm/test/Transforms/LoopIdiom/wcslen16.ll
+++ b/llvm/test/Transforms/LoopIdiom/wcslen16.ll
@@ -26,7 +26,7 @@ define i64 @valid_strlen16(ptr %src) {
 ; CHECK-NEXT:    [[CURR_0]] = getelementptr inbounds i8, ptr [[SRC_PN]], i64 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = load i16, ptr [[CURR_0]], align 2
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i16 [[TMP3]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_COND]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[WHILE_COND]]
 ; CHECK:       [[WHILE_END]]:
 ; CHECK-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[END]] to i64
 ; CHECK-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[SRC]] to i64
diff --git a/llvm/test/Transforms/LoopIdiom/wcslen32.ll b/llvm/test/Transforms/LoopIdiom/wcslen32.ll
index 2c7ceb4d187e6..f1cddf9424ffd 100644
--- a/llvm/test/Transforms/LoopIdiom/wcslen32.ll
+++ b/llvm/test/Transforms/LoopIdiom/wcslen32.ll
@@ -26,7 +26,7 @@ define i64 @valid_wcslen32(ptr %src) {
 ; CHECK-NEXT:    [[CURR_0]] = getelementptr inbounds i8, ptr [[SRC_PN]], i64 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[CURR_0]], align 4
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i32 [[TMP3]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_COND]]
+; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[WHILE_COND]]
 ; CHECK:       [[WHILE_END]]:
 ; CHECK-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[END]] to i64
 ; CHECK-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[SRC]] to i64

@delcypher delcypher requested a review from fhahn April 8, 2025 21:56
@mustartt mustartt requested review from Meinersbur and maryammo April 8, 2025 21:57
Copy link
Member

@Meinersbur Meinersbur left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@@ -1804,6 +1808,11 @@ bool LoopIdiomRecognize::recognizeAndInsertStrLen() {
// up by later passes
for (PHINode *PN : Cleanup)
RecursivelyDeleteDeadPHINode(PN);

ConstantInt *NewLoopCond = LoopTerm->getSuccessor(0) == LoopBody
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Consider adding a comment explaining that this intends to make the old code unreachable/dead.

Add assert that at least one of the successors is LoopBody?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for taking a look. I added the asserts and comments.

I will take a look to see if there's any further testing required.

@mustartt mustartt merged commit 68b7cba into llvm:main Apr 11, 2025
11 checks passed
@mustartt mustartt deleted the strlen-simplfy-delete branch April 11, 2025 16:55
@llvm-ci
Copy link
Collaborator

llvm-ci commented Apr 11, 2025

LLVM Buildbot has detected a new failure on builder clang-ppc64le-linux-test-suite running on ppc64le-clang-test-suite while building llvm at step 6 "test-build-unified-tree-check-all".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/95/builds/11892

Here is the relevant piece of the build log for the reference
Step 6 (test-build-unified-tree-check-all) failure: test (failure)
******************** TEST 'LLVM-Unit :: Support/./SupportTests/943/1509' FAILED ********************
Script(shard):
--
GTEST_OUTPUT=json:/home/buildbots/llvm-external-buildbots/workers/ppc64le-clang-test-suite/clang-ppc64le-test-suite/build/unittests/Support/./SupportTests-LLVM-Unit-3440064-943-1509.json GTEST_SHUFFLE=0 GTEST_TOTAL_SHARDS=1509 GTEST_SHARD_INDEX=943 /home/buildbots/llvm-external-buildbots/workers/ppc64le-clang-test-suite/clang-ppc64le-test-suite/build/unittests/Support/./SupportTests
--

Script:
--
/home/buildbots/llvm-external-buildbots/workers/ppc64le-clang-test-suite/clang-ppc64le-test-suite/build/unittests/Support/./SupportTests --gtest_filter=ProgramEnvTest.TestLockFile
--
/home/buildbots/llvm-external-buildbots/workers/ppc64le-clang-test-suite/clang-ppc64le-test-suite/llvm-project/llvm/unittests/Support/ProgramTest.cpp:570: Failure
Value of: Error.empty()
  Actual: false
Expected: true


/home/buildbots/llvm-external-buildbots/workers/ppc64le-clang-test-suite/clang-ppc64le-test-suite/llvm-project/llvm/unittests/Support/ProgramTest.cpp:570
Value of: Error.empty()
  Actual: false
Expected: true



********************


var-const pushed a commit to ldionne/llvm-project that referenced this pull request Apr 17, 2025
… LoopDeletion (llvm#134906)

Fixes the case where subsequent passes were unable to find and delete
the invariant loop left over by the strlen idiom conversion. Since
`loop-deletion` only operate on computable loops, we can update the loop
condition to something more easily picked up by `loop-deletion`

As pointed out in llvm#134736
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

#108985 optimization doesn't remove strlen like loop body in some cases
4 participants