Skip to content

Commit

Permalink
[LoopIdimo] Use tryZExtValue() instead of getZExtValue()
Browse files Browse the repository at this point in the history
To avoid an assertion for large BECounts.

I also suspect that this code is missing an overflow check.

Fixes llvm#70008.
  • Loading branch information
nikic committed Oct 24, 2023
1 parent 1e3a344 commit 97f1db2
Show file tree
Hide file tree
Showing 2 changed files with 46 additions and 3 deletions.
10 changes: 7 additions & 3 deletions llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -948,9 +948,13 @@ mayLoopAccessLocation(Value *Ptr, ModRefInfo Access, Loop *L,
// to be exactly the size of the memset, which is (BECount+1)*StoreSize
const SCEVConstant *BECst = dyn_cast<SCEVConstant>(BECount);
const SCEVConstant *ConstSize = dyn_cast<SCEVConstant>(StoreSizeSCEV);
if (BECst && ConstSize)
AccessSize = LocationSize::precise((BECst->getValue()->getZExtValue() + 1) *
ConstSize->getValue()->getZExtValue());
if (BECst && ConstSize) {
std::optional<uint64_t> BEInt = BECst->getAPInt().tryZExtValue();
std::optional<uint64_t> SizeInt = ConstSize->getAPInt().tryZExtValue();
// FIXME: Should this check for overflow?
if (BEInt && SizeInt)
AccessSize = LocationSize::precise((*BEInt + 1) * *SizeInt);
}

// TODO: For this to be really effective, we have to dive into the pointer
// operand in the store. Store to &A[i] of 100 will always return may alias
Expand Down
39 changes: 39 additions & 0 deletions llvm/test/Transforms/LoopIdiom/pr70008.ll
Original file line number Diff line number Diff line change
@@ -0,0 +1,39 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
; RUN: opt -S -passes=loop-idiom < %s | FileCheck %s

; Make sure we don't assert if the BECount is larger than 64 bits.

define void @test(ptr %p) {
; CHECK-LABEL: define void @test(
; CHECK-SAME: ptr [[P:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 4 [[P]], i8 0, i64 0, i1 false)
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[IV:%.*]] = phi i128 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[IV_TRUNC:%.*]] = trunc i128 [[IV]] to i64
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr { i64, i64 }, ptr [[P]], i64 [[IV_TRUNC]]
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr { i64, i64 }, ptr [[P]], i64 [[IV_TRUNC]], i32 1
; CHECK-NEXT: [[INC]] = add i128 [[IV]], 1
; CHECK-NEXT: [[TOBOOL_NOT:%.*]] = icmp eq i128 [[INC]], 0
; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label [[EXIT:%.*]], label [[FOR_BODY]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %for.body

for.body:
%iv = phi i128 [ 0, %entry ], [ %inc, %for.body ]
%iv.trunc = trunc i128 %iv to i64
%gep1 = getelementptr { i64, i64 }, ptr %p, i64 %iv.trunc
%gep2 = getelementptr { i64, i64 }, ptr %p, i64 %iv.trunc, i32 1
store i64 0, ptr %gep1
store i64 0, ptr %gep2
%inc = add i128 %iv, 1
%tobool.not = icmp eq i128 %inc, 0
br i1 %tobool.not, label %exit, label %for.body

exit:
ret void
}

0 comments on commit 97f1db2

Please sign in to comment.