Skip to content

[SCEV] Check whether the start is non-zero in ScalarEvolution::howFarToZero #131522

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.

Sign up for GitHub

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
Mar 17, 2025
Merged

[SCEV] Check whether the start is non-zero in ScalarEvolution::howFarToZero #131522

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ddb05e7
[SCEV][LoopUnroll] Add pre-commit tests. NFC.
dtcxzyw Mar 16, 2025
3282a05
[SCEV] Check whether the start is non-zero in `ScalarEvolution::howFa…
dtcxzyw Mar 16, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 5 additions & 4 deletions llvm/lib/Analysis/ScalarEvolution.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10635,10 +10635,11 @@ ScalarEvolution::ExitLimit ScalarEvolution::howFarToZero(const SCEV *V,
if (ControlsOnlyExit && AddRec->hasNoSelfWrap() &&
loopHasNoAbnormalExits(AddRec->getLoop())) {

// If the stride is zero, the loop must be infinite. In C++, most loops
// are finite by assumption, in which case the step being zero implies
// UB must execute if the loop is entered.
if (!loopIsFiniteByAssumption(L) && !isKnownNonZero(StepWLG))
// If the stride is zero and the start is non-zero, the loop must be
// infinite. In C++, most loops are finite by assumption, in which case the
// step being zero implies UB must execute if the loop is entered.
if (!(loopIsFiniteByAssumption(L) && isKnownNonZero(Start)) &&
!isKnownNonZero(StepWLG))
return getCouldNotCompute();

const SCEV *Exact =
Expand Down
34 changes: 26 additions & 8 deletions llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -329,10 +329,9 @@ define void @ne_nsw_nonneg_step(ptr nocapture %A, i32 %n, i32 %s) mustprogress {
;
; CHECK-LABEL: 'ne_nsw_nonneg_step'
; CHECK-NEXT: Determining loop execution counts for: @ne_nsw_nonneg_step
; CHECK-NEXT: Loop %for.body: backedge-taken count is (((-1 * %s) + %n) /u %s)
; CHECK-NEXT: Loop %for.body: constant max backedge-taken count is i32 -1
; CHECK-NEXT: Loop %for.body: symbolic max backedge-taken count is (((-1 * %s) + %n) /u %s)
; CHECK-NEXT: Loop %for.body: Trip multiple is 1
; CHECK-NEXT: Loop %for.body: Unpredictable backedge-taken count.
; CHECK-NEXT: Loop %for.body: Unpredictable constant max backedge-taken count.
; CHECK-NEXT: Loop %for.body: Unpredictable symbolic max backedge-taken count.
Copy link
Contributor

Choose a reason for hiding this comment

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

Could probably recover these if we used loop guards for start.

Copy link
Member Author

Choose a reason for hiding this comment

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

In this case we cannot recover these since the start value is zero :(
I will check if using applyLoopGuards(Start, Guards) is profitable.

Copy link
Contributor

Choose a reason for hiding this comment

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

Wouldn't the start value after subtraction here be something like -n?

Copy link
Member Author

Choose a reason for hiding this comment

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

Oh, sorry, the start value is (zext i1 %tobool2.not to i32).

Copy link
Member Author

Choose a reason for hiding this comment

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

https://github.com/dtcxzyw/llvm-opt-benchmark/pull/2212/files Looks like it is unprofitable.

;
entry:
%nonneg_step = icmp sge i32 %s, 0
Expand Down Expand Up @@ -442,10 +441,9 @@ define void @ne_nuw_nonneg_step(ptr nocapture %A, i32 %n, i32 %s) mustprogress {
;
; CHECK-LABEL: 'ne_nuw_nonneg_step'
; CHECK-NEXT: Determining loop execution counts for: @ne_nuw_nonneg_step
; CHECK-NEXT: Loop %for.body: backedge-taken count is (((-1 * %s) + %n) /u %s)
; CHECK-NEXT: Loop %for.body: constant max backedge-taken count is i32 -1
; CHECK-NEXT: Loop %for.body: symbolic max backedge-taken count is (((-1 * %s) + %n) /u %s)
; CHECK-NEXT: Loop %for.body: Trip multiple is 1
; CHECK-NEXT: Loop %for.body: Unpredictable backedge-taken count.
; CHECK-NEXT: Loop %for.body: Unpredictable constant max backedge-taken count.
; CHECK-NEXT: Loop %for.body: Unpredictable symbolic max backedge-taken count.
;
entry:
%nonneg_step = icmp sge i32 %s, 0
Expand Down Expand Up @@ -493,6 +491,26 @@ for.end: ; preds = %for.body, %entry
ret void
}

define i32 @pr131465(i1 %x) mustprogress {
; CHECK-LABEL: 'pr131465'
; CHECK-NEXT: Determining loop execution counts for: @pr131465
; CHECK-NEXT: Loop %for.body: Unpredictable backedge-taken count.
; CHECK-NEXT: Loop %for.body: Unpredictable constant max backedge-taken count.
; CHECK-NEXT: Loop %for.body: Unpredictable symbolic max backedge-taken count.
;
entry:
%inc = zext i1 %x to i32
br label %for.body

for.body:
%indvar = phi i32 [ 2, %entry ], [ %next, %for.body ]
%next = add nsw i32 %indvar, %inc
%exitcond = icmp eq i32 %next, 2
br i1 %exitcond, label %for.end, label %for.body

for.end:
ret i32 0
}

declare void @llvm.assume(i1)

Expand Down
43 changes: 43 additions & 0 deletions llvm/test/Transforms/LoopUnroll/pr131465.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,43 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S -passes=loop-unroll -unroll-runtime %s | FileCheck %s

define i32 @pr131465(i1 %x) mustprogress {
; CHECK-LABEL: define i32 @pr131465(
; CHECK-SAME: i1 [[X:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[INC:%.*]] = zext i1 [[X]] to i32
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INDVAR:%.*]] = phi i32 [ 2, %[[ENTRY]] ], [ [[NEXT_1:%.*]], %[[FOR_BODY_1:.*]] ]
; CHECK-NEXT: [[NEXT:%.*]] = add nsw i32 [[INDVAR]], [[INC]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[NEXT]], 2
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[FOR_END:.*]], label %[[FOR_BODY_1]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[FOR_BODY_1]]:
; CHECK-NEXT: [[NEXT_1]] = add nsw i32 [[NEXT]], [[INC]]
; CHECK-NEXT: [[EXITCOND_1:%.*]] = icmp eq i32 [[NEXT_1]], 2
; CHECK-NEXT: br i1 [[EXITCOND_1]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; CHECK: [[FOR_END]]:
; CHECK-NEXT: ret i32 0
;
entry:
%inc = zext i1 %x to i32
br label %for.body

for.body:
%indvar = phi i32 [ 2, %entry ], [ %next, %for.body ]
%next = add nsw i32 %indvar, %inc
%exitcond = icmp eq i32 %next, 2
br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0

for.end:
ret i32 0
}

; Force runtime unrolling.
!0 = !{!0, !{!"llvm.loop.unroll.count", i32 2}}
;.
; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]]}
; CHECK: [[META1]] = !{!"llvm.loop.unroll.count", i32 2}
; CHECK: [[LOOP2]] = distinct !{[[LOOP2]], [[META3:![0-9]+]]}
; CHECK: [[META3]] = !{!"llvm.loop.unroll.disable"}
;.