[SCEV] Check whether the start is non-zero in ScalarEvolution::howFarToZero
#131522
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
llvmbot
added
llvm:analysis
|
@llvm/pr-subscribers-llvm-transforms Author: Yingwei Zheng (dtcxzyw) Changes#94525 assumes that the loop will be infinite when the stride is zero. However, it doesn't hold when the start value of addrec is also zero. Closes #131465. Full diff: https://github.com/llvm/llvm-project/pull/131522.diff 3 Files Affected:
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 8f74c1c398ced..314baa7c7aee1 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -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 =
diff --git a/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll b/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
index 2d02cb6194f4c..1f08a620b2e15 100644
--- a/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
+++ b/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
@@ -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.
;
entry:
%nonneg_step = icmp sge i32 %s, 0
@@ -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
@@ -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)
diff --git a/llvm/test/Transforms/LoopUnroll/pr131465.ll b/llvm/test/Transforms/LoopUnroll/pr131465.ll
new file mode 100644
index 0000000000000..643b020c6c110
--- /dev/null
+++ b/llvm/test/Transforms/LoopUnroll/pr131465.ll
@@ -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"}
+;.
|
|
@llvm/pr-subscribers-llvm-analysis Author: Yingwei Zheng (dtcxzyw) Changes#94525 assumes that the loop will be infinite when the stride is zero. However, it doesn't hold when the start value of addrec is also zero. Closes #131465. Full diff: https://github.com/llvm/llvm-project/pull/131522.diff 3 Files Affected:
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 8f74c1c398ced..314baa7c7aee1 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -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 =
diff --git a/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll b/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
index 2d02cb6194f4c..1f08a620b2e15 100644
--- a/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
+++ b/llvm/test/Analysis/ScalarEvolution/trip-count-unknown-stride.ll
@@ -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.
;
entry:
%nonneg_step = icmp sge i32 %s, 0
@@ -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
@@ -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)
diff --git a/llvm/test/Transforms/LoopUnroll/pr131465.ll b/llvm/test/Transforms/LoopUnroll/pr131465.ll
new file mode 100644
index 0000000000000..643b020c6c110
--- /dev/null
+++ b/llvm/test/Transforms/LoopUnroll/pr131465.ll
@@ -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"}
+;.
|
nikic
approved these changes
Mar 16, 2025
| ; 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. |
There was a problem hiding this comment.
Could probably recover these if we used loop guards for start.
There was a problem hiding this comment.
In this case we cannot recover these since the start value is zero :(
I will check if using applyLoopGuards(Start, Guards) is profitable.
There was a problem hiding this comment.
Wouldn't the start value after subtraction here be something like -n?
There was a problem hiding this comment.
Oh, sorry, the start value is (zext i1 %tobool2.not to i32).
There was a problem hiding this comment.
https://github.com/dtcxzyw/llvm-opt-benchmark/pull/2212/files Looks like it is unprofitable.
This was referenced Mar 16, 2025
|
/cherry-pick c5a491e |
|
/pull-request #131568 |
|
Thank you for finding and fixing this one. I'm a bit shocked I managed to miss that condition and it took this long to find. Initially, I'd suspected we had some kind of partial guard somewhere, but after some quick digging, I couldn't find one. So, this appears to have just been wrong for this case from the start. |
swift-ci
pushed a commit
to swiftlang/llvm-project
that referenced
this pull request
Mar 18, 2025
…rToZero` (llvm#131522) llvm#94525 assumes that the loop will be infinite when the stride is zero. However, it doesn't hold when the start value of addrec is also zero. Closes llvm#131465. (cherry picked from commit c5a491e)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
#94525 assumes that the loop will be infinite when the stride is zero. However, it doesn't hold when the start value of addrec is also zero.
Closes #131465.