This is an archive of the discontinued LLVM Phabricator instance.

Differential D71025

[LoopFusion] Restrict loop fusion to rotated loops.
ClosedPublic

Authored by kbarton on Dec 4 2019, 9:32 AM.

Details

Reviewers
jdoerfert
Meinersbur
dmgreen
etiotto
Whitney
fhahn
hfinkel
Commits
rGff07fc66d9ee: [LoopFusion] Restrict loop fusion to rotated loops.
Summary

This patch restricts loop fusion to only consider rotated loops as valid candidates.
This simplifies the analysis and transformation and aligns with other loop optimizations.

Diff Detail

Event Timeline

kbarton created this revision.Dec 4 2019, 9:32 AM
Herald added a project: Restricted Project. · View Herald TranscriptDec 4 2019, 9:32 AM
Herald added subscribers: llvm-commits, hiraditya. · View Herald Transcript
Harbormaster completed remote builds in B41867: Diff 232160.Dec 4 2019, 9:32 AM
Herald added a subscriber: ormris. · View Herald TranscriptDec 4 2019, 9:32 AM
Meinersbur accepted this revision.Dec 5 2019, 8:55 AM

[suggestion] Add a test case to check for "not rotated" to be rejected.

llvm/test/Transforms/LoopFusion/cannot_fuse.ll
270

Does this test something different now?

llvm/test/Transforms/LoopFusion/diagnostics_missed.ll
3–4

[nit] can be removed

This revision is now accepted and ready to land.Dec 5 2019, 8:55 AM
Whitney added a child revision: D71165: [LoopFusion] Move instructions from FC0.Latch to FC1.Latch..Dec 7 2019, 5:29 AM
Whitney added inline comments.Dec 12 2019, 4:12 PM
llvm/test/Transforms/LoopFusion/cannot_fuse.ll
42

This guard was not in the original test case, now the second loop has two guards, is that intensional?

78

bb34 is also a predecessor of bb33.

kbarton marked 4 inline comments as done.Dec 16 2019, 9:02 AM

Addressed all the review comments.
I'm ready to land this, unless @Whitney has concerns about the test case.

llvm/test/Transforms/LoopFusion/cannot_fuse.ll
42

Yes, this is intentional to make the two loops not control-flow-equivalent, which is what this specific test is checking.

270

This is a good catch. I missed this. I've fixed it to check for the memory dependencies again.

Whitney added inline comments.Dec 16 2019, 9:39 AM
llvm/test/Transforms/LoopFusion/cannot_fuse.ll
42

The two loops should already be not control-flow-equivalent, as bb7 is not guarded, while bb22 is guarded. Why is the extra guard needed?

kbarton marked 2 inline comments as done.Dec 16 2019, 9:56 AM
kbarton added inline comments.
llvm/test/Transforms/LoopFusion/cannot_fuse.ll
42

If the loop is not guarded, the code will use the preheader; if the loop is guarded, it uses the guard.
In this case, since one is guarded and the other is not guarded, it will compare the preheader for the first loop with the guard for the second loop and find them control flow equivalent. It will fail in a later check for fusion, however this specific test is meant to check that loops that are not control flow equivalent are put into different candidate sets. Putting an extra branch around the second loop satisfies that.

Whitney added inline comments.Dec 16 2019, 10:08 AM
llvm/test/Transforms/LoopFusion/cannot_fuse.ll
42

Notice that before changing the test case to rotated form, there is only one loop guard needed. I think we should change the way fusion compute the control flow equivalent between two loops. As a guarded loop (assuming the condition not always true) and an unguarded loop should be consider not control flow equivalent. I am ok to keep that as future improvement.

kbarton marked an inline comment as done.Dec 16 2019, 10:43 AM
kbarton added inline comments.
llvm/test/Transforms/LoopFusion/cannot_fuse.ll
42

Yes, this is an existing problem that came up recently. I don't think the current patch exacerbates the problem though.
I am currently working on a fix for that, but would prefer to land this patch as is and fix the existing problem separately.

Closed by commit rGff07fc66d9ee: [LoopFusion] Restrict loop fusion to rotated loops. (authored by kbarton). · Explain WhyDec 16 2019, 12:17 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
llvm/
lib/
Transforms/
Scalar/
6 lines
test/
Transforms/
LoopFusion/
307 lines
364 lines
140 lines
104 lines
309 lines

Diff 234118

llvm/lib/Transforms/Scalar/LoopFuse.cpp

Show First 20 Lines • Show All 85 Lines • ▼ Show 20 Lines
STATISTIC(UncomputableTripCount, "SCEV cannot compute trip count of loop"); STATISTIC(UncomputableTripCount, "SCEV cannot compute trip count of loop");
STATISTIC(NonEqualTripCount, "Loop trip counts are not the same"); STATISTIC(NonEqualTripCount, "Loop trip counts are not the same");
STATISTIC(NonAdjacent, "Loops are not adjacent"); STATISTIC(NonAdjacent, "Loops are not adjacent");
STATISTIC(NonEmptyPreheader, "Loop has a non-empty preheader"); STATISTIC(NonEmptyPreheader, "Loop has a non-empty preheader");
STATISTIC(FusionNotBeneficial, "Fusion is not beneficial"); STATISTIC(FusionNotBeneficial, "Fusion is not beneficial");
STATISTIC(NonIdenticalGuards, "Candidates have different guards"); STATISTIC(NonIdenticalGuards, "Candidates have different guards");
STATISTIC(NonEmptyExitBlock, "Candidate has a non-empty exit block"); STATISTIC(NonEmptyExitBlock, "Candidate has a non-empty exit block");
STATISTIC(NonEmptyGuardBlock, "Candidate has a non-empty guard block"); STATISTIC(NonEmptyGuardBlock, "Candidate has a non-empty guard block");
STATISTIC(NotRotated, "Candidate is not rotated");
enum FusionDependenceAnalysisChoice { enum FusionDependenceAnalysisChoice {
FUSION_DEPENDENCE_ANALYSIS_SCEV, FUSION_DEPENDENCE_ANALYSIS_SCEV,
FUSION_DEPENDENCE_ANALYSIS_DA, FUSION_DEPENDENCE_ANALYSIS_DA,
FUSION_DEPENDENCE_ANALYSIS_ALL, FUSION_DEPENDENCE_ANALYSIS_ALL,
}; };
static cl::opt<FusionDependenceAnalysisChoice> FusionDependenceAnalysis( static cl::opt<FusionDependenceAnalysisChoice> FusionDependenceAnalysis(
▲ Show 20 Lines • Show All 212 Lines • ▼ Show 20 Lines bool isEligibleForFusion(ScalarEvolution &SE) const {
} }
if (!L->isLoopSimplifyForm()) { if (!L->isLoopSimplifyForm()) {
LLVM_DEBUG(dbgs() << "Loop " << L->getName() LLVM_DEBUG(dbgs() << "Loop " << L->getName()
<< " is not in simplified form!\n"); << " is not in simplified form!\n");
return reportInvalidCandidate(NotSimplifiedForm); return reportInvalidCandidate(NotSimplifiedForm);
} }
if (!isRotated()) {
LLVM_DEBUG(dbgs() << "Loop " << L->getName() << " is not rotated!\n");
return reportInvalidCandidate(NotRotated);
}
return true; return true;
} }
private: private:
// This is only used internally for now, to clear the MemWrites and MemReads // This is only used internally for now, to clear the MemWrites and MemReads
// list and setting Valid to false. I can't envision other uses of this right // list and setting Valid to false. I can't envision other uses of this right
// now, since once FusionCandidates are put into the FusionCandidateSet they // now, since once FusionCandidates are put into the FusionCandidateSet they
// are immutable. Thus, any time we need to change/update a FusionCandidate, // are immutable. Thus, any time we need to change/update a FusionCandidate,
▲ Show 20 Lines • Show All 1,301 Lines • Show Last 20 Lines

llvm/test/Transforms/LoopFusion/cannot_fuse.ll

Show All 9 Lines
; CHECK: Fusion Candidates: ; CHECK: Fusion Candidates:
; CHECK: *** Fusion Candidate Set *** ; CHECK: *** Fusion Candidate Set ***
; CHECK: bb ; CHECK: bb
; CHECK: **************************** ; CHECK: ****************************
; CHECK: *** Fusion Candidate Set *** ; CHECK: *** Fusion Candidate Set ***
; CHECK: bb20.preheader ; CHECK: bb20.preheader
; CHECK: **************************** ; CHECK: ****************************
; CHECK: Loop Fusion complete ; CHECK: Loop Fusion complete
define void @non_cfe(i32* noalias %arg) { define void @non_cfe(i32* noalias %arg, i32 %N) {
bb: bb:
br label %bb5 br label %bb7
bb5: ; preds = %bb14, %bb
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb14 ], [ 0, %bb ]
%.01 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ]
%exitcond4 = icmp ne i64 %indvars.iv2, 100
br i1 %exitcond4, label %bb7, label %bb16
bb7: ; preds = %bb5 bb7: ; preds = %bb, %bb14
%tmp = add nsw i32 %.01, -3 %.014 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ]
%tmp8 = add nuw nsw i64 %indvars.iv2, 3 %indvars.iv23 = phi i64 [ 0, %bb ], [ %indvars.iv.next3, %bb14 ]
%tmp = add nsw i32 %.014, -3
%tmp8 = add nuw nsw i64 %indvars.iv23, 3
%tmp9 = trunc i64 %tmp8 to i32 %tmp9 = trunc i64 %tmp8 to i32
%tmp10 = mul nsw i32 %tmp, %tmp9 %tmp10 = mul nsw i32 %tmp, %tmp9
%tmp11 = trunc i64 %indvars.iv2 to i32 %tmp11 = trunc i64 %indvars.iv23 to i32
%tmp12 = srem i32 %tmp10, %tmp11 %tmp12 = srem i32 %tmp10, %tmp11
%tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2 %tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv23
store i32 %tmp12, i32* %tmp13, align 4 store i32 %tmp12, i32* %tmp13, align 4
br label %bb14 br label %bb14
bb14: ; preds = %bb7 bb14: ; preds = %bb7
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1 %indvars.iv.next3 = add nuw nsw i64 %indvars.iv23, 1
%tmp15 = add nuw nsw i32 %.01, 1 %tmp15 = add nuw nsw i32 %.014, 1
br label %bb5 %exitcond4 = icmp ne i64 %indvars.iv.next3, 100
br i1 %exitcond4, label %bb7, label %bb34
bb34:
%cmp = icmp slt i32 %N, 50
WhitneyUnsubmitted
Not Done
ReplyInline Actions

This guard was not in the original test case, now the second loop has two guards, is that intensional?

Whitney: This guard was not in the original test case, now the second loop has two guards, is that…
kbartonAuthorUnsubmitted
Not Done
ReplyInline Actions

Yes, this is intentional to make the two loops not control-flow-equivalent, which is what this specific test is checking.

kbarton: Yes, this is intentional to make the two loops not control-flow-equivalent, which is what this…
WhitneyUnsubmitted
Not Done
ReplyInline Actions

The two loops should already be not control-flow-equivalent, as bb7 is not guarded, while bb22 is guarded. Why is the extra guard needed?

Whitney: The two loops should already be not control-flow-equivalent, as bb7 is not guarded, while bb22…
kbartonAuthorUnsubmitted
Done
ReplyInline Actions

If the loop is not guarded, the code will use the preheader; if the loop is guarded, it uses the guard.
In this case, since one is guarded and the other is not guarded, it will compare the preheader for the first loop with the guard for the second loop and find them control flow equivalent. It will fail in a later check for fusion, however this specific test is meant to check that loops that are not control flow equivalent are put into different candidate sets. Putting an extra branch around the second loop satisfies that.

kbarton: If the loop is not guarded, the code will use the preheader; if the loop is guarded, it uses…
WhitneyUnsubmitted
Not Done
ReplyInline Actions

Notice that before changing the test case to rotated form, there is only one loop guard needed. I think we should change the way fusion compute the control flow equivalent between two loops. As a guarded loop (assuming the condition not always true) and an unguarded loop should be consider not control flow equivalent. I am ok to keep that as future improvement.

Whitney: Notice that before changing the test case to rotated form, there is only one loop guard needed.
kbartonAuthorUnsubmitted
Done
ReplyInline Actions

Yes, this is an existing problem that came up recently. I don't think the current patch exacerbates the problem though.
I am currently working on a fix for that, but would prefer to land this patch as is and fix the existing problem separately.

kbarton: Yes, this is an existing problem that came up recently. I don't think the current patch…
br i1 %cmp, label %bb16, label %bb33
bb16: ; preds = %bb5 bb16: ; preds = %bb34
%tmp17 = load i32, i32* %arg, align 4 %tmp17 = load i32, i32* %arg, align 4
%tmp18 = icmp slt i32 %tmp17, 0 %tmp18 = icmp slt i32 %tmp17, 0
br i1 %tmp18, label %bb20, label %bb33 br i1 %tmp18, label %bb20.preheader, label %bb33
bb20: ; preds = %bb30, %bb16 bb20.preheader: ; preds = %bb16
%indvars.iv = phi i64 [ %indvars.iv.next, %bb30 ], [ 0, %bb16 ] br label %bb22
%.0 = phi i32 [ 0, %bb16 ], [ %tmp31, %bb30 ]
%exitcond = icmp ne i64 %indvars.iv, 100
br i1 %exitcond, label %bb22, label %bb33
bb22: ; preds = %bb20 bb22: ; preds = %bb20.preheader, %bb30
%tmp23 = add nsw i32 %.0, -3 %.02 = phi i32 [ 0, %bb20.preheader ], [ %tmp31, %bb30 ]
%tmp24 = add nuw nsw i64 %indvars.iv, 3 %indvars.iv1 = phi i64 [ 0, %bb20.preheader ], [ %indvars.iv.next, %bb30 ]
%tmp23 = add nsw i32 %.02, -3
%tmp24 = add nuw nsw i64 %indvars.iv1, 3
%tmp25 = trunc i64 %tmp24 to i32 %tmp25 = trunc i64 %tmp24 to i32
%tmp26 = mul nsw i32 %tmp23, %tmp25 %tmp26 = mul nsw i32 %tmp23, %tmp25
%tmp27 = trunc i64 %indvars.iv to i32 %tmp27 = trunc i64 %indvars.iv1 to i32
%tmp28 = srem i32 %tmp26, %tmp27 %tmp28 = srem i32 %tmp26, %tmp27
%tmp29 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %tmp29 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv1
store i32 %tmp28, i32* %tmp29, align 4 store i32 %tmp28, i32* %tmp29, align 4
br label %bb30 br label %bb30
bb30: ; preds = %bb22 bb30: ; preds = %bb22
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
%tmp31 = add nuw nsw i32 %.0, 1 %tmp31 = add nuw nsw i32 %.02, 1
br label %bb20 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb22, label %bb33.loopexit
bb33: ; preds = %bb20, %bb16 bb33.loopexit: ; preds = %bb30
br label %bb33
bb33: ; preds = %bb33.loopexit, %bb16, %bb34
ret void ret void
} }
WhitneyUnsubmitted
Done
ReplyInline Actions

bb34 is also a predecessor of bb33.

Whitney: bb34 is also a predecessor of bb33.
; Check that fusion detects the two canddates are not adjacent (the exit block ; Check that fusion detects the two canddates are not adjacent (the exit block
; of the first candidate is not the preheader of the second candidate). ; of the first candidate is not the preheader of the second candidate).
; CHECK: Performing Loop Fusion on function non_adjacent ; CHECK: Performing Loop Fusion on function non_adjacent
; CHECK: Fusion Candidates: ; CHECK: Fusion Candidates:
; CHECK: *** Fusion Candidate Set *** ; CHECK: *** Fusion Candidate Set ***
; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]]
; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]]
; CHECK-NEXT: **************************** ; CHECK-NEXT: ****************************
; CHECK: Attempting fusion on Candidate Set: ; CHECK: Attempting fusion on Candidate Set:
; CHECK-NEXT: [[LOOP1PREHEADER]] ; CHECK-NEXT: [[LOOP1PREHEADER]]
; CHECK-NEXT: [[LOOP2PREHEADER]] ; CHECK-NEXT: [[LOOP2PREHEADER]]
; CHECK: Fusion candidates are not adjacent. Not fusing. ; CHECK: Fusion candidates are not adjacent. Not fusing.
; CHECK: Loop Fusion complete ; CHECK: Loop Fusion complete
define void @non_adjacent(i32* noalias %arg) { define void @non_adjacent(i32* noalias %arg) {
bb: bb:
br label %bb3 br label %bb5
bb3: ; preds = %bb11, %bb
%.01 = phi i64 [ 0, %bb ], [ %tmp12, %bb11 ]
%exitcond2 = icmp ne i64 %.01, 100
br i1 %exitcond2, label %bb5, label %bb4
bb4: ; preds = %bb3 bb4: ; preds = %bb11
br label %bb13 br label %bb13
bb5: ; preds = %bb3 bb5: ; preds = %bb, %bb11
%tmp = add nsw i64 %.01, -3 %.013 = phi i64 [ 0, %bb ], [ %tmp12, %bb11 ]
%tmp6 = add nuw nsw i64 %.01, 3 %tmp = add nsw i64 %.013, -3
%tmp6 = add nuw nsw i64 %.013, 3
%tmp7 = mul nsw i64 %tmp, %tmp6 %tmp7 = mul nsw i64 %tmp, %tmp6
%tmp8 = srem i64 %tmp7, %.01 %tmp8 = srem i64 %tmp7, %.013
%tmp9 = trunc i64 %tmp8 to i32 %tmp9 = trunc i64 %tmp8 to i32
%tmp10 = getelementptr inbounds i32, i32* %arg, i64 %.01 %tmp10 = getelementptr inbounds i32, i32* %arg, i64 %.013
store i32 %tmp9, i32* %tmp10, align 4 store i32 %tmp9, i32* %tmp10, align 4
br label %bb11 br label %bb11
bb11: ; preds = %bb5 bb11: ; preds = %bb5
%tmp12 = add nuw nsw i64 %.01, 1 %tmp12 = add nuw nsw i64 %.013, 1
br label %bb3 %exitcond2 = icmp ne i64 %tmp12, 100
br i1 %exitcond2, label %bb5, label %bb4
bb13: ; preds = %bb4 bb13: ; preds = %bb4
br label %bb14 br label %bb16
bb14: ; preds = %bb23, %bb13
%.0 = phi i64 [ 0, %bb13 ], [ %tmp24, %bb23 ]
%exitcond = icmp ne i64 %.0, 100
br i1 %exitcond, label %bb16, label %bb15
bb15: ; preds = %bb14 bb15: ; preds = %bb23
br label %bb25 br label %bb25
bb16: ; preds = %bb14 bb16: ; preds = %bb13, %bb23
%tmp17 = add nsw i64 %.0, -3 %.02 = phi i64 [ 0, %bb13 ], [ %tmp24, %bb23 ]
%tmp18 = add nuw nsw i64 %.0, 3 %tmp17 = add nsw i64 %.02, -3
%tmp18 = add nuw nsw i64 %.02, 3
%tmp19 = mul nsw i64 %tmp17, %tmp18 %tmp19 = mul nsw i64 %tmp17, %tmp18
%tmp20 = srem i64 %tmp19, %.0 %tmp20 = srem i64 %tmp19, %.02
%tmp21 = trunc i64 %tmp20 to i32 %tmp21 = trunc i64 %tmp20 to i32
%tmp22 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %.0 %tmp22 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %.02
store i32 %tmp21, i32* %tmp22, align 4 store i32 %tmp21, i32* %tmp22, align 4
br label %bb23 br label %bb23
bb23: ; preds = %bb16 bb23: ; preds = %bb16
%tmp24 = add nuw nsw i64 %.0, 1 %tmp24 = add nuw nsw i64 %.02, 1
br label %bb14 %exitcond = icmp ne i64 %tmp24, 100
br i1 %exitcond, label %bb16, label %bb15
bb25: ; preds = %bb15 bb25: ; preds = %bb15
ret void ret void
} }
; Check that the different bounds are detected and prevent fusion. ; Check that the different bounds are detected and prevent fusion.
; CHECK: Performing Loop Fusion on function different_bounds ; CHECK: Performing Loop Fusion on function different_bounds
; CHECK: Fusion Candidates: ; CHECK: Fusion Candidates:
; CHECK: *** Fusion Candidate Set *** ; CHECK: *** Fusion Candidate Set ***
; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]]
; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]]
; CHECK-NEXT: **************************** ; CHECK-NEXT: ****************************
; CHECK: Attempting fusion on Candidate Set: ; CHECK: Attempting fusion on Candidate Set:
; CHECK-NEXT: [[LOOP1PREHEADER]] ; CHECK-NEXT: [[LOOP1PREHEADER]]
; CHECK-NEXT: [[LOOP2PREHEADER]] ; CHECK-NEXT: [[LOOP2PREHEADER]]
; CHECK: Fusion candidates do not have identical trip counts. Not fusing. ; CHECK: Fusion candidates do not have identical trip counts. Not fusing.
; CHECK: Loop Fusion complete ; CHECK: Loop Fusion complete
define void @different_bounds(i32* noalias %arg) { define void @different_bounds(i32* noalias %arg) {
bb: bb:
br label %bb3 br label %bb5
bb3: ; preds = %bb11, %bb
%.01 = phi i64 [ 0, %bb ], [ %tmp12, %bb11 ]
%exitcond2 = icmp ne i64 %.01, 100
br i1 %exitcond2, label %bb5, label %bb4
bb4: ; preds = %bb3 bb4: ; preds = %bb11
br label %bb13 br label %bb13
bb5: ; preds = %bb3 bb5: ; preds = %bb, %bb11
%tmp = add nsw i64 %.01, -3 %.013 = phi i64 [ 0, %bb ], [ %tmp12, %bb11 ]
%tmp6 = add nuw nsw i64 %.01, 3 %tmp = add nsw i64 %.013, -3
%tmp6 = add nuw nsw i64 %.013, 3
%tmp7 = mul nsw i64 %tmp, %tmp6 %tmp7 = mul nsw i64 %tmp, %tmp6
%tmp8 = srem i64 %tmp7, %.01 %tmp8 = srem i64 %tmp7, %.013
%tmp9 = trunc i64 %tmp8 to i32 %tmp9 = trunc i64 %tmp8 to i32
%tmp10 = getelementptr inbounds i32, i32* %arg, i64 %.01 %tmp10 = getelementptr inbounds i32, i32* %arg, i64 %.013
store i32 %tmp9, i32* %tmp10, align 4 store i32 %tmp9, i32* %tmp10, align 4
br label %bb11 br label %bb11
bb11: ; preds = %bb5 bb11: ; preds = %bb5
%tmp12 = add nuw nsw i64 %.01, 1 %tmp12 = add nuw nsw i64 %.013, 1
br label %bb3 %exitcond2 = icmp ne i64 %tmp12, 100
br i1 %exitcond2, label %bb5, label %bb4
bb13: ; preds = %bb4 bb13: ; preds = %bb4
br label %bb14 br label %bb16
bb14: ; preds = %bb23, %bb13
%.0 = phi i64 [ 0, %bb13 ], [ %tmp24, %bb23 ]
%exitcond = icmp ne i64 %.0, 200
br i1 %exitcond, label %bb16, label %bb15
bb15: ; preds = %bb14 bb15: ; preds = %bb23
br label %bb25 br label %bb25
bb16: ; preds = %bb14 bb16: ; preds = %bb13, %bb23
%tmp17 = add nsw i64 %.0, -3 %.02 = phi i64 [ 0, %bb13 ], [ %tmp24, %bb23 ]
%tmp18 = add nuw nsw i64 %.0, 3 %tmp17 = add nsw i64 %.02, -3
%tmp18 = add nuw nsw i64 %.02, 3
%tmp19 = mul nsw i64 %tmp17, %tmp18 %tmp19 = mul nsw i64 %tmp17, %tmp18
%tmp20 = srem i64 %tmp19, %.0 %tmp20 = srem i64 %tmp19, %.02
%tmp21 = trunc i64 %tmp20 to i32 %tmp21 = trunc i64 %tmp20 to i32
%tmp22 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %.0 %tmp22 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %.02
store i32 %tmp21, i32* %tmp22, align 4 store i32 %tmp21, i32* %tmp22, align 4
br label %bb23 br label %bb23
bb23: ; preds = %bb16 bb23: ; preds = %bb16
%tmp24 = add nuw nsw i64 %.0, 1 %tmp24 = add nuw nsw i64 %.02, 1
br label %bb14 %exitcond = icmp ne i64 %tmp24, 200
br i1 %exitcond, label %bb16, label %bb15
bb25: ; preds = %bb15 bb25: ; preds = %bb15
ret void ret void
} }
; Check that the negative dependence between the two candidates is identified ; Check that the negative dependence between the two candidates is identified
; and prevents fusion. ; and prevents fusion.
; CHECK: Performing Loop Fusion on function negative_dependence ; CHECK: Performing Loop Fusion on function negative_dependence
; CHECK: Fusion Candidates: ; CHECK: Fusion Candidates:
; CHECK: *** Fusion Candidate Set *** ; CHECK: *** Fusion Candidate Set ***
; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]]
; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]]
; CHECK-NEXT: **************************** ; CHECK-NEXT: ****************************
; CHECK: Attempting fusion on Candidate Set: ; CHECK: Attempting fusion on Candidate Set:
; CHECK-NEXT: [[LOOP1PREHEADER]] ; CHECK-NEXT: [[LOOP1PREHEADER]]
; CHECK-NEXT: [[LOOP2PREHEADER]] ; CHECK-NEXT: [[LOOP2PREHEADER]]
; CHECK: Memory dependencies do not allow fusion! ; CHECK: Memory dependencies do not allow fusion!
; CHECK: Loop Fusion complete ; CHECK: Loop Fusion complete
define void @negative_dependence(i32* noalias %arg) { define void @negative_dependence(i32* noalias %arg) {
bb: bb:
br label %bb5 br label %bb7
bb5: ; preds = %bb9, %bb bb11.preheader: ; preds = %bb9
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb9 ], [ 0, %bb ] br label %bb13
%exitcond4 = icmp ne i64 %indvars.iv2, 100
br i1 %exitcond4, label %bb7, label %bb11
bb7: ; preds = %bb5 bb7: ; preds = %bb, %bb9
%tmp = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2 %indvars.iv22 = phi i64 [ 0, %bb ], [ %indvars.iv.next3, %bb9 ]
%tmp8 = trunc i64 %indvars.iv2 to i32 %tmp = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv22
%tmp8 = trunc i64 %indvars.iv22 to i32
store i32 %tmp8, i32* %tmp, align 4 store i32 %tmp8, i32* %tmp, align 4
br label %bb9 br label %bb9
bb9: ; preds = %bb7 bb9: ; preds = %bb7
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1 %indvars.iv.next3 = add nuw nsw i64 %indvars.iv22, 1
br label %bb5 %exitcond4 = icmp ne i64 %indvars.iv.next3, 100
br i1 %exitcond4, label %bb7, label %bb11.preheader
bb11: ; preds = %bb18, %bb5
%indvars.iv = phi i64 [ %indvars.iv.next, %bb18 ], [ 0, %bb5 ] bb13: ; preds = %bb11.preheader, %bb18
%exitcond = icmp ne i64 %indvars.iv, 100 %indvars.iv1 = phi i64 [ 0, %bb11.preheader ], [ %indvars.iv.next, %bb18 ]
br i1 %exitcond, label %bb13, label %bb19 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
bb13: ; preds = %bb11
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%tmp14 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv.next %tmp14 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv.next
%tmp15 = load i32, i32* %tmp14, align 4 %tmp15 = load i32, i32* %tmp14, align 4
%tmp16 = shl nsw i32 %tmp15, 1 %tmp16 = shl nsw i32 %tmp15, 1
%tmp17 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %tmp17 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv1
store i32 %tmp16, i32* %tmp17, align 4 store i32 %tmp16, i32* %tmp17, align 4
br label %bb18 br label %bb18
bb18: ; preds = %bb13 bb18: ; preds = %bb13
br label %bb11 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb13, label %bb19
bb19: ; preds = %bb11 bb19: ; preds = %bb18
ret void ret void
} }
; Check for values defined in Loop 0 and used in Loop 1. ; Check for values defined in Loop 0 and used in Loop 1.
; It is not safe to fuse in this case, because the second loop has ; It is not safe to fuse in this case, because the second loop has
; a use of %.01.lcssa which is defined in the body of loop 0. The ; a use of %.01.lcssa which is defined in the body of loop 0. The
; first loop must execute completely in order to compute the correct ; first loop must execute completely in order to compute the correct
; value of %.01.lcssa to be used in the second loop. ; value of %.01.lcssa to be used in the second loop.
; CHECK: Performing Loop Fusion on function sumTest ; CHECK: Performing Loop Fusion on function sumTest
; CHECK: Fusion Candidates: ; CHECK: Fusion Candidates:
; CHECK: *** Fusion Candidate Set *** ; CHECK: *** Fusion Candidate Set ***
; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP1PREHEADER:bb[0-9]*]]
; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]] ; CHECK-NEXT: [[LOOP2PREHEADER:bb[0-9]*]]
; CHECK-NEXT: **************************** ; CHECK-NEXT: ****************************
; CHECK: Attempting fusion on Candidate Set: ; CHECK: Attempting fusion on Candidate Set:
; CHECK-NEXT: [[LOOP1PREHEADER]] ; CHECK-NEXT: [[LOOP1PREHEADER]]
; CHECK-NEXT: [[LOOP2PREHEADER]] ; CHECK-NEXT: [[LOOP2PREHEADER]]
; CHECK: Memory dependencies do not allow fusion! ; CHECK: Memory dependencies do not allow fusion!
MeinersburUnsubmitted
Done
ReplyInline Actions

Does this test something different now?

Meinersbur: Does this test something different now?
kbartonAuthorUnsubmitted
Done
ReplyInline Actions

This is a good catch. I missed this. I've fixed it to check for the memory dependencies again.

kbarton: This is a good catch. I missed this. I've fixed it to check for the memory dependencies again.
; CHECK: Loop Fusion complete ; CHECK: Loop Fusion complete
define i32 @sumTest(i32* noalias %arg) { define i32 @sumTest(i32* noalias %arg) {
bb: bb:
br label %bb6 br label %bb9
bb6: ; preds = %bb9, %bb bb13.preheader: ; preds = %bb9
%indvars.iv3 = phi i64 [ %indvars.iv.next4, %bb9 ], [ 0, %bb ] br label %bb15
%.01 = phi i32 [ 0, %bb ], [ %tmp11, %bb9 ]
%exitcond5 = icmp ne i64 %indvars.iv3, 100
br i1 %exitcond5, label %bb9, label %bb13
bb9: ; preds = %bb6 bb9: ; preds = %bb, %bb9
%tmp = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv3 %.01.lcssa = phi i32 [ 0, %bb ], [ %tmp11, %bb9 ]
%.013 = phi i32 [ 0, %bb ], [ %tmp11, %bb9 ]
%indvars.iv32 = phi i64 [ 0, %bb ], [ %indvars.iv.next4, %bb9 ]
%tmp = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv32
%tmp10 = load i32, i32* %tmp, align 4 %tmp10 = load i32, i32* %tmp, align 4
%tmp11 = add nsw i32 %.01, %tmp10 %tmp11 = add nsw i32 %.013, %tmp10
%indvars.iv.next4 = add nuw nsw i64 %indvars.iv3, 1 %indvars.iv.next4 = add nuw nsw i64 %indvars.iv32, 1
br label %bb6 %exitcond5 = icmp ne i64 %indvars.iv.next4, 100
br i1 %exitcond5, label %bb9, label %bb13.preheader
bb13: ; preds = %bb20, %bb6
%.01.lcssa = phi i32 [ %.01, %bb6 ], [ %.01.lcssa, %bb20 ]
%indvars.iv = phi i64 [ %indvars.iv.next, %bb20 ], [ 0, %bb6 ]
%exitcond = icmp ne i64 %indvars.iv, 100
br i1 %exitcond, label %bb15, label %bb14
bb14: ; preds = %bb13 bb14: ; preds = %bb20
br label %bb21 br label %bb21
bb15: ; preds = %bb13 bb15: ; preds = %bb13.preheader, %bb20
%tmp16 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv %indvars.iv1 = phi i64 [ 0, %bb13.preheader ], [ %indvars.iv.next, %bb20 ]
%tmp16 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv1
%tmp17 = load i32, i32* %tmp16, align 4 %tmp17 = load i32, i32* %tmp16, align 4
%tmp18 = sdiv i32 %tmp17, %.01.lcssa %tmp18 = sdiv i32 %tmp17, %.01.lcssa
%tmp19 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %tmp19 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv1
store i32 %tmp18, i32* %tmp19, align 4 store i32 %tmp18, i32* %tmp19, align 4
br label %bb20 br label %bb20
bb20: ; preds = %bb15 bb20: ; preds = %bb15
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
br label %bb13 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb15, label %bb14
bb21: ; preds = %bb14 bb21: ; preds = %bb14
ret i32 %.01.lcssa ret i32 %.01.lcssa
} }
; Similar to sumTest above. The first loop computes %add and must ; Similar to sumTest above. The first loop computes %add and must
; complete before it is used in the second loop. Thus, these two loops ; complete before it is used in the second loop. Thus, these two loops
; also cannot be fused. ; also cannot be fused.
Show All 36 Linesfor.body8: ; preds = %for.body, %for.body8
%inc14 = add nuw nsw i64 %i2.031, 1 %inc14 = add nuw nsw i64 %i2.031, 1
%cmp5 = icmp ult i64 %inc14, %conv %cmp5 = icmp ult i64 %inc14, %conv
br i1 %cmp5, label %for.body8, label %for.cond.cleanup7 br i1 %cmp5, label %for.body8, label %for.cond.cleanup7
for.cond.cleanup7: ; preds = %for.body8, %entry for.cond.cleanup7: ; preds = %for.body8, %entry
%sum1.0.lcssa36 = phi float [ 0.000000e+00, %entry ], [ %add, %for.body8 ] %sum1.0.lcssa36 = phi float [ 0.000000e+00, %entry ], [ %add, %for.body8 ]
ret float %sum1.0.lcssa36 ret float %sum1.0.lcssa36
} }
; Check that non-rotated loops are not considered for fusion.
; CHECK: Performing Loop Fusion on function notRotated
; CHECK: Loop bb{{.*}} is not rotated!
; CHECK: Loop bb{{.*}} is not rotated!
define void @notRotated(i32* noalias %arg) {
bb:
br label %bb5
bb5: ; preds = %bb14, %bb
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb14 ], [ 0, %bb ]
%.01 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ]
%exitcond4 = icmp ne i64 %indvars.iv2, 100
br i1 %exitcond4, label %bb7, label %bb17
bb7: ; preds = %bb5
%tmp = add nsw i32 %.01, -3
%tmp8 = add nuw nsw i64 %indvars.iv2, 3
%tmp9 = trunc i64 %tmp8 to i32
%tmp10 = mul nsw i32 %tmp, %tmp9
%tmp11 = trunc i64 %indvars.iv2 to i32
%tmp12 = srem i32 %tmp10, %tmp11
%tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2
store i32 %tmp12, i32* %tmp13, align 4
br label %bb14
bb14: ; preds = %bb7
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1
%tmp15 = add nuw nsw i32 %.01, 1
br label %bb5
bb17: ; preds = %bb27, %bb5
%indvars.iv = phi i64 [ %indvars.iv.next, %bb27 ], [ 0, %bb5 ]
%.0 = phi i32 [ 0, %bb5 ], [ %tmp28, %bb27 ]
%exitcond = icmp ne i64 %indvars.iv, 100
br i1 %exitcond, label %bb19, label %bb18
bb18: ; preds = %bb17
br label %bb29
bb19: ; preds = %bb17
%tmp20 = add nsw i32 %.0, -3
%tmp21 = add nuw nsw i64 %indvars.iv, 3
%tmp22 = trunc i64 %tmp21 to i32
%tmp23 = mul nsw i32 %tmp20, %tmp22
%tmp24 = trunc i64 %indvars.iv to i32
%tmp25 = srem i32 %tmp23, %tmp24
%tmp26 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv
store i32 %tmp25, i32* %tmp26, align 4
br label %bb27
bb27: ; preds = %bb19
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%tmp28 = add nuw nsw i32 %.0, 1
br label %bb17
bb29: ; preds = %bb18
ret void
}

llvm/test/Transforms/LoopFusion/diagnostics_missed.ll

; RUN: opt -S -loop-fusion -pass-remarks-missed=loop-fusion -disable-output < %s 2>&1 | FileCheck %s ; RUN: opt -S -loop-fusion -pass-remarks-missed=loop-fusion -disable-output < %s 2>&1 | FileCheck %s
;
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128" target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
MeinersburUnsubmitted
Done
ReplyInline Actions

[nit] can be removed

Meinersbur: [nit] can be removed
@B = common global [1024 x i32] zeroinitializer, align 16, !dbg !0 @B = common global [1024 x i32] zeroinitializer, align 16, !dbg !0
; CHECK: remark: diagnostics_missed.c:18:3: [non_adjacent]: entry and for.end: Loops are not adjacent ; CHECK: remark: diagnostics_missed.c:18:3: [non_adjacent]: entry and for.end: Loops are not adjacent
define void @non_adjacent(i32* noalias %A) !dbg !67 { define void @non_adjacent(i32* noalias %A) !dbg !14 {
entry: entry:
br label %for.cond br label %for.body
for.cond: ; preds = %for.inc, %entry
%i.0 = phi i64 [ 0, %entry ], [ %inc, %for.inc ]
%exitcond1 = icmp ne i64 %i.0, 100
br i1 %exitcond1, label %for.body, label %for.cond.cleanup
for.cond.cleanup: ; preds = %for.cond for.cond.cleanup: ; preds = %for.inc
br label %for.end br label %for.end
for.body: ; preds = %for.cond for.body: ; preds = %entry, %for.inc
%sub = add nsw i64 %i.0, -3 %i.02 = phi i64 [ 0, %entry ], [ %inc, %for.inc ]
%add = add nuw nsw i64 %i.0, 3 %sub = add nsw i64 %i.02, -3
%add = add nuw nsw i64 %i.02, 3
%mul = mul nsw i64 %sub, %add %mul = mul nsw i64 %sub, %add
%rem = srem i64 %mul, %i.0 %rem = srem i64 %mul, %i.02
%conv = trunc i64 %rem to i32 %conv = trunc i64 %rem to i32
%arrayidx = getelementptr inbounds i32, i32* %A, i64 %i.0 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %i.02
store i32 %conv, i32* %arrayidx, align 4 store i32 %conv, i32* %arrayidx, align 4
br label %for.inc br label %for.inc
for.inc: ; preds = %for.body for.inc: ; preds = %for.body
%inc = add nuw nsw i64 %i.0, 1, !dbg !86 %inc = add nuw nsw i64 %i.02, 1, !dbg !26
br label %for.cond, !dbg !87, !llvm.loop !88 %exitcond1 = icmp ne i64 %inc, 100
br i1 %exitcond1, label %for.body, label %for.cond.cleanup, !llvm.loop !28
for.end: ; preds = %for.cond.cleanup for.end: ; preds = %for.cond.cleanup
br label %for.cond2 br label %for.body6
for.cond2: ; preds = %for.inc13, %for.end for.cond.cleanup5: ; preds = %for.inc13
%i1.0 = phi i64 [ 0, %for.end ], [ %inc14, %for.inc13 ]
%exitcond = icmp ne i64 %i1.0, 100
br i1 %exitcond, label %for.body6, label %for.cond.cleanup5
for.cond.cleanup5: ; preds = %for.cond2
br label %for.end15 br label %for.end15
for.body6: ; preds = %for.cond2 for.body6: ; preds = %for.end, %for.inc13
%sub7 = add nsw i64 %i1.0, -3 %i1.01 = phi i64 [ 0, %for.end ], [ %inc14, %for.inc13 ]
%add8 = add nuw nsw i64 %i1.0, 3 %sub7 = add nsw i64 %i1.01, -3
%add8 = add nuw nsw i64 %i1.01, 3
%mul9 = mul nsw i64 %sub7, %add8 %mul9 = mul nsw i64 %sub7, %add8
%rem10 = srem i64 %mul9, %i1.0 %rem10 = srem i64 %mul9, %i1.01
%conv11 = trunc i64 %rem10 to i32 %conv11 = trunc i64 %rem10 to i32
%arrayidx12 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %i1.0 %arrayidx12 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %i1.01
store i32 %conv11, i32* %arrayidx12, align 4 store i32 %conv11, i32* %arrayidx12, align 4
br label %for.inc13 br label %for.inc13
for.inc13: ; preds = %for.body6 for.inc13: ; preds = %for.body6
%inc14 = add nuw nsw i64 %i1.0, 1, !dbg !100 %inc14 = add nuw nsw i64 %i1.01, 1, !dbg !31
br label %for.cond2, !dbg !101, !llvm.loop !102 %exitcond = icmp ne i64 %inc14, 100
br i1 %exitcond, label %for.body6, label %for.cond.cleanup5, !llvm.loop !33
for.end15: ; preds = %for.cond.cleanup5 for.end15: ; preds = %for.cond.cleanup5
ret void ret void
} }
; CHECK: remark: diagnostics_missed.c:28:3: [different_bounds]: entry and for.end: Loop trip counts are not the same ; CHECK: remark: diagnostics_missed.c:28:3: [different_bounds]: entry and for.end: Loop trip counts are not the same
define void @different_bounds(i32* noalias %A) !dbg !105 { define void @different_bounds(i32* noalias %A) !dbg !36 {
entry: entry:
br label %for.cond br label %for.body
for.cond: ; preds = %for.inc, %entry for.cond.cleanup: ; preds = %for.inc
%i.0 = phi i64 [ 0, %entry ], [ %inc, %for.inc ]
%exitcond1 = icmp ne i64 %i.0, 100
br i1 %exitcond1, label %for.body, label %for.cond.cleanup
for.cond.cleanup: ; preds = %for.cond
br label %for.end br label %for.end
for.body: ; preds = %for.cond for.body: ; preds = %entry, %for.inc
%sub = add nsw i64 %i.0, -3 %i.02 = phi i64 [ 0, %entry ], [ %inc, %for.inc ]
%add = add nuw nsw i64 %i.0, 3 %sub = add nsw i64 %i.02, -3
%add = add nuw nsw i64 %i.02, 3
%mul = mul nsw i64 %sub, %add %mul = mul nsw i64 %sub, %add
%rem = srem i64 %mul, %i.0 %rem = srem i64 %mul, %i.02
%conv = trunc i64 %rem to i32 %conv = trunc i64 %rem to i32
%arrayidx = getelementptr inbounds i32, i32* %A, i64 %i.0 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %i.02
store i32 %conv, i32* %arrayidx, align 4 store i32 %conv, i32* %arrayidx, align 4
br label %for.inc br label %for.inc
for.inc: ; preds = %for.body for.inc: ; preds = %for.body
%inc = add nuw nsw i64 %i.0, 1, !dbg !123 %inc = add nuw nsw i64 %i.02, 1, !dbg !43
br label %for.cond, !dbg !124, !llvm.loop !125 %exitcond1 = icmp ne i64 %inc, 100
br i1 %exitcond1, label %for.body, label %for.cond.cleanup, !llvm.loop !45
for.end: ; preds = %for.cond.cleanup for.end: ; preds = %for.cond.cleanup
br label %for.cond2 br label %for.body6
for.cond2: ; preds = %for.inc13, %for.end
%i1.0 = phi i64 [ 0, %for.end ], [ %inc14, %for.inc13 ]
%exitcond = icmp ne i64 %i1.0, 200
br i1 %exitcond, label %for.body6, label %for.cond.cleanup5
for.cond.cleanup5: ; preds = %for.cond2 for.cond.cleanup5: ; preds = %for.inc13
br label %for.end15 br label %for.end15
for.body6: ; preds = %for.cond2 for.body6: ; preds = %for.end, %for.inc13
%sub7 = add nsw i64 %i1.0, -3 %i1.01 = phi i64 [ 0, %for.end ], [ %inc14, %for.inc13 ]
%add8 = add nuw nsw i64 %i1.0, 3 %sub7 = add nsw i64 %i1.01, -3
%add8 = add nuw nsw i64 %i1.01, 3
%mul9 = mul nsw i64 %sub7, %add8 %mul9 = mul nsw i64 %sub7, %add8
%rem10 = srem i64 %mul9, %i1.0 %rem10 = srem i64 %mul9, %i1.01
%conv11 = trunc i64 %rem10 to i32 %conv11 = trunc i64 %rem10 to i32
%arrayidx12 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %i1.0 %arrayidx12 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %i1.01
store i32 %conv11, i32* %arrayidx12, align 4 store i32 %conv11, i32* %arrayidx12, align 4
br label %for.inc13 br label %for.inc13
for.inc13: ; preds = %for.body6 for.inc13: ; preds = %for.body6
%inc14 = add nuw nsw i64 %i1.0, 1 %inc14 = add nuw nsw i64 %i1.01, 1
br label %for.cond2, !dbg !138, !llvm.loop !139 %exitcond = icmp ne i64 %inc14, 200
br i1 %exitcond, label %for.body6, label %for.cond.cleanup5, !llvm.loop !48
for.end15: ; preds = %for.cond.cleanup5 for.end15: ; preds = %for.cond.cleanup5
ret void ret void
} }
; CHECK: remark: diagnostics_missed.c:38:3: [negative_dependence]: entry and for.end: Loop has a non-empty preheader ; CHECK: remark: diagnostics_missed.c:38:3: [negative_dependence]: entry and for.end: Loop has a non-empty preheader
define void @negative_dependence(i32* noalias %A) !dbg !142 { define void @negative_dependence(i32* noalias %A) !dbg !51 {
entry: entry:
br label %for.cond br label %for.body
for.cond: ; preds = %for.inc, %entry for.body: ; preds = %entry, %for.inc
%indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc ], [ 0, %entry ] %indvars.iv13 = phi i64 [ 0, %entry ], [ %indvars.iv.next2, %for.inc ]
%exitcond3 = icmp ne i64 %indvars.iv1, 100 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv13
br i1 %exitcond3, label %for.body, label %for.end %tmp = trunc i64 %indvars.iv13 to i32
for.body: ; preds = %for.cond
%arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv1
%tmp = trunc i64 %indvars.iv1 to i32
store i32 %tmp, i32* %arrayidx, align 4 store i32 %tmp, i32* %arrayidx, align 4
br label %for.inc br label %for.inc
for.inc: ; preds = %for.body for.inc: ; preds = %for.body
%indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1 %indvars.iv.next2 = add nuw nsw i64 %indvars.iv13, 1
br label %for.cond, !dbg !160, !llvm.loop !161 %exitcond3 = icmp ne i64 %indvars.iv.next2, 100
br i1 %exitcond3, label %for.body, label %for.end, !llvm.loop !58
for.end: ; preds = %for.cond
call void @llvm.dbg.value(metadata i32 0, metadata !147, metadata !DIExpression()), !dbg !163 for.end: ; preds = %for.inc
br label %for.cond2, !dbg !164 call void @llvm.dbg.value(metadata i32 0, metadata !56, metadata !DIExpression()), !dbg !61
br label %for.body5
for.cond2: ; preds = %for.inc10, %for.end
%indvars.iv = phi i64 [ %indvars.iv.next, %for.inc10 ], [ 0, %for.end ] for.body5: ; preds = %for.end, %for.inc10
%exitcond = icmp ne i64 %indvars.iv, 100 %indvars.iv2 = phi i64 [ 0, %for.end ], [ %indvars.iv.next, %for.inc10 ]
br i1 %exitcond, label %for.body5, label %for.end12 %indvars.iv.next = add nuw nsw i64 %indvars.iv2, 1
for.body5: ; preds = %for.cond2
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%arrayidx7 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv.next %arrayidx7 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv.next
%tmp4 = load i32, i32* %arrayidx7, align 4 %tmp4 = load i32, i32* %arrayidx7, align 4
%mul = shl nsw i32 %tmp4, 1 %mul = shl nsw i32 %tmp4, 1
%arrayidx9 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %arrayidx9 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv2
store i32 %mul, i32* %arrayidx9, align 4 store i32 %mul, i32* %arrayidx9, align 4
br label %for.inc10 br label %for.inc10
for.inc10: ; preds = %for.body5 for.inc10: ; preds = %for.body5
br label %for.cond2 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %for.body5, label %for.end12
for.end12: ; preds = %for.cond. for.end12: ; preds = %for.inc10
ret void, !dbg !178 ret void, !dbg !62
} }
; CHECK: remark: diagnostics_missed.c:51:3: [sumTest]: entry and for.cond2.preheader: Dependencies prevent fusion ; CHECK: remark: diagnostics_missed.c:51:3: [sumTest]: entry and for.cond2.preheader: Dependencies prevent fusion
define i32 @sumTest(i32* noalias %A) !dbg !179 { define i32 @sumTest(i32* noalias %A) !dbg !63 {
entry: entry:
br label %for.cond br label %for.body
for.cond: ; preds = %for.inc, %entry for.cond2.preheader: ; preds = %for.inc
%indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc ], [ 0, %entry ] br label %for.body5
%sum.0 = phi i32 [ 0, %entry ], [ %add, %for.inc ]
%exitcond3 = icmp ne i64 %indvars.iv1, 100
br i1 %exitcond3, label %for.body, label %for.cond2
for.body: ; preds = %for.cond for.body: ; preds = %entry, %for.inc
%sum.04 = phi i32 [ 0, %entry ], [ %add, %for.inc ]
%indvars.iv13 = phi i64 [ 0, %entry ], [ %indvars.iv.next2, %for.inc ]
br label %for.inc br label %for.inc
for.inc: ; preds = %for.body for.inc: ; preds = %for.body
%arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv1 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv13
%tmp = load i32, i32* %arrayidx, align 4 %tmp = load i32, i32* %arrayidx, align 4
%add = add nsw i32 %sum.0, %tmp %add = add nsw i32 %sum.04, %tmp
%indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1 %indvars.iv.next2 = add nuw nsw i64 %indvars.iv13, 1
br label %for.cond, !dbg !199, !llvm.loop !200 %exitcond3 = icmp ne i64 %indvars.iv.next2, 100
br i1 %exitcond3, label %for.body, label %for.cond2.preheader, !llvm.loop !73
for.cond2: ; preds = %for.inc10, %for.cond
%sum.0.lcssa = phi i32 [ %sum.0, %for.cond ], [ %sum.0.lcssa, %for.inc10 ] for.body5: ; preds = %for.cond2.preheader, %for.inc10
%indvars.iv = phi i64 [ %indvars.iv.next, %for.inc10 ], [ 0, %for.cond ] %indvars.iv2 = phi i64 [ 0, %for.cond2.preheader ], [ %indvars.iv.next, %for.inc10 ]
%exitcond = icmp ne i64 %indvars.iv, 100 %arrayidx7 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv2
br i1 %exitcond, label %for.body5, label %for.end12
for.body5: ; preds = %for.cond2
%arrayidx7 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
%tmp4 = load i32, i32* %arrayidx7, align 4 %tmp4 = load i32, i32* %arrayidx7, align 4
%div = sdiv i32 %tmp4, %sum.0.lcssa %div = sdiv i32 %tmp4, %add
%arrayidx9 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %arrayidx9 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv2
store i32 %div, i32* %arrayidx9, align 4 store i32 %div, i32* %arrayidx9, align 4
br label %for.inc10 br label %for.inc10
for.inc10: ; preds = %for.body5 for.inc10: ; preds = %for.body5
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv2, 1
br label %for.cond2 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %for.body5, label %for.end12
for.end12: ; preds = %for.cond2 for.end12: ; preds = %for.inc10
ret i32 %sum.0.lcssa, !dbg !215 ret i32 %add, !dbg !76
} }
declare void @llvm.dbg.value(metadata, metadata, metadata) ; Function Attrs: nounwind readnone speculatable willreturn
declare void @llvm.dbg.value(metadata, metadata, metadata) #0
attributes #0 = { nounwind readnone speculatable willreturn }
!llvm.dbg.cu = !{!2} !llvm.dbg.cu = !{!2}
!llvm.module.flags = !{!11, !12, !13, !14} !llvm.module.flags = !{!10, !11, !12, !13}
!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression()) !0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
!1 = distinct !DIGlobalVariable(name: "B", scope: !2, file: !6, line: 46, type: !7, isLocal: false, isDefinition: true) !1 = distinct !DIGlobalVariable(name: "B", scope: !2, file: !3, line: 46, type: !6, isLocal: false, isDefinition: true)
!2 = distinct !DICompileUnit(language: DW_LANG_C99, file: !3, producer: "clang version 9.0.0 (git@github.ibm.com:compiler/llvm-project.git 23c4baaa9f5b33d2d52eda981d376c6b0a7a3180)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !4, globals: !5, nameTableKind: GNU) !2 = distinct !DICompileUnit(language: DW_LANG_C99, file: !3, producer: "clang version 9.0.0 (git@github.ibm.com:compiler/llvm-project.git 23c4baaa9f5b33d2d52eda981d376c6b0a7a3180)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !4, globals: !5, nameTableKind: GNU)
!3 = !DIFile(filename: "diagnostics_missed.c", directory: "/tmp") !3 = !DIFile(filename: "diagnostics_missed.c", directory: "/tmp")
!4 = !{} !4 = !{}
!5 = !{!0} !5 = !{!0}
!6 = !DIFile(filename: "diagnostics_missed.c", directory: "/tmp") !6 = !DICompositeType(tag: DW_TAG_array_type, baseType: !7, size: 32768, elements: !8)
!7 = !DICompositeType(tag: DW_TAG_array_type, baseType: !8, size: 32768, elements: !9) !7 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
!8 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed) !8 = !{!9}
!9 = !{!10} !9 = !DISubrange(count: 1024)
!10 = !DISubrange(count: 1024) !10 = !{i32 2, !"Dwarf Version", i32 4}
!11 = !{i32 2, !"Dwarf Version", i32 4} !11 = !{i32 2, !"Debug Info Version", i32 3}
!12 = !{i32 2, !"Debug Info Version", i32 3} !12 = !{i32 1, !"wchar_size", i32 4}
!13 = !{i32 1, !"wchar_size", i32 4} !13 = !{i32 7, !"PIC Level", i32 2}
!14 = !{i32 7, !"PIC Level", i32 2} !14 = distinct !DISubprogram(name: "non_adjacent", scope: !3, file: !3, line: 17, type: !15, scopeLine: 17, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !19)
!17 = !DISubroutineType(types: !18) !15 = !DISubroutineType(types: !16)
!18 = !{null, !19} !16 = !{null, !17}
!19 = !DIDerivedType(tag: DW_TAG_restrict_type, baseType: !20) !17 = !DIDerivedType(tag: DW_TAG_restrict_type, baseType: !18)
!20 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !8, size: 64) !18 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !7, size: 64)
!67 = distinct !DISubprogram(name: "non_adjacent", scope: !6, file: !6, line: 17, type: !17, scopeLine: 17, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !68) !19 = !{!20, !21, !24}
!68 = !{!69, !70, !73} !20 = !DILocalVariable(name: "A", arg: 1, scope: !14, file: !3, line: 17, type: !17)
!69 = !DILocalVariable(name: "A", arg: 1, scope: !67, file: !6, line: 17, type: !19) !21 = !DILocalVariable(name: "i", scope: !22, file: !3, line: 18, type: !23)
!70 = !DILocalVariable(name: "i", scope: !71, file: !6, line: 18, type: !72) !22 = distinct !DILexicalBlock(scope: !14, file: !3, line: 18, column: 3)
!71 = distinct !DILexicalBlock(scope: !67, file: !6, line: 18, column: 3) !23 = !DIBasicType(name: "long int", size: 64, encoding: DW_ATE_signed)
!72 = !DIBasicType(name: "long int", size: 64, encoding: DW_ATE_signed) !24 = !DILocalVariable(name: "i", scope: !25, file: !3, line: 22, type: !23)
!73 = !DILocalVariable(name: "i", scope: !74, file: !6, line: 22, type: !72) !25 = distinct !DILexicalBlock(scope: !14, file: !3, line: 22, column: 3)
!74 = distinct !DILexicalBlock(scope: !67, file: !6, line: 22, column: 3) !26 = !DILocation(line: 18, column: 30, scope: !27)
!79 = distinct !DILexicalBlock(scope: !71, file: !6, line: 18, column: 3) !27 = distinct !DILexicalBlock(scope: !22, file: !3, line: 18, column: 3)
!80 = !DILocation(line: 18, column: 3, scope: !71) !28 = distinct !{!28, !29, !30}
!86 = !DILocation(line: 18, column: 30, scope: !79) !29 = !DILocation(line: 18, column: 3, scope: !22)
!87 = !DILocation(line: 18, column: 3, scope: !79) !30 = !DILocation(line: 20, column: 3, scope: !22)
!88 = distinct !{!88, !80, !89} !31 = !DILocation(line: 22, column: 30, scope: !32)
!89 = !DILocation(line: 20, column: 3, scope: !71) !32 = distinct !DILexicalBlock(scope: !25, file: !3, line: 22, column: 3)
!93 = distinct !DILexicalBlock(scope: !74, file: !6, line: 22, column: 3) !33 = distinct !{!33, !34, !35}
!94 = !DILocation(line: 22, column: 3, scope: !74) !34 = !DILocation(line: 22, column: 3, scope: !25)
!100 = !DILocation(line: 22, column: 30, scope: !93) !35 = !DILocation(line: 24, column: 3, scope: !25)
!101 = !DILocation(line: 22, column: 3, scope: !93) !36 = distinct !DISubprogram(name: "different_bounds", scope: !3, file: !3, line: 27, type: !15, scopeLine: 27, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !37)
!102 = distinct !{!102, !94, !103} !37 = !{!38, !39, !41}
!103 = !DILocation(line: 24, column: 3, scope: !74) !38 = !DILocalVariable(name: "A", arg: 1, scope: !36, file: !3, line: 27, type: !17)
!105 = distinct !DISubprogram(name: "different_bounds", scope: !6, file: !6, line: 27, type: !17, scopeLine: 27, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !106) !39 = !DILocalVariable(name: "i", scope: !40, file: !3, line: 28, type: !23)
!106 = !{!107, !108, !110} !40 = distinct !DILexicalBlock(scope: !36, file: !3, line: 28, column: 3)
!107 = !DILocalVariable(name: "A", arg: 1, scope: !105, file: !6, line: 27, type: !19) !41 = !DILocalVariable(name: "i", scope: !42, file: !3, line: 32, type: !23)
!108 = !DILocalVariable(name: "i", scope: !109, file: !6, line: 28, type: !72) !42 = distinct !DILexicalBlock(scope: !36, file: !3, line: 32, column: 3)
!109 = distinct !DILexicalBlock(scope: !105, file: !6, line: 28, column: 3) !43 = !DILocation(line: 28, column: 30, scope: !44)
!110 = !DILocalVariable(name: "i", scope: !111, file: !6, line: 32, type: !72) !44 = distinct !DILexicalBlock(scope: !40, file: !3, line: 28, column: 3)
!111 = distinct !DILexicalBlock(scope: !105, file: !6, line: 32, column: 3) !45 = distinct !{!45, !46, !47}
!116 = distinct !DILexicalBlock(scope: !109, file: !6, line: 28, column: 3) !46 = !DILocation(line: 28, column: 3, scope: !40)
!117 = !DILocation(line: 28, column: 3, scope: !109) !47 = !DILocation(line: 30, column: 3, scope: !40)
!123 = !DILocation(line: 28, column: 30, scope: !116) !48 = distinct !{!48, !49, !50}
!124 = !DILocation(line: 28, column: 3, scope: !116) !49 = !DILocation(line: 32, column: 3, scope: !42)
!125 = distinct !{!125, !117, !126} !50 = !DILocation(line: 34, column: 3, scope: !42)
!126 = !DILocation(line: 30, column: 3, scope: !109) !51 = distinct !DISubprogram(name: "negative_dependence", scope: !3, file: !3, line: 37, type: !15, scopeLine: 37, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !52)
!130 = distinct !DILexicalBlock(scope: !111, file: !6, line: 32, column: 3) !52 = !{!53, !54, !56}
!131 = !DILocation(line: 32, column: 3, scope: !111) !53 = !DILocalVariable(name: "A", arg: 1, scope: !51, file: !3, line: 37, type: !17)
!138 = !DILocation(line: 32, column: 3, scope: !130) !54 = !DILocalVariable(name: "i", scope: !55, file: !3, line: 38, type: !7)
!139 = distinct !{!139, !131, !140} !55 = distinct !DILexicalBlock(scope: !51, file: !3, line: 38, column: 3)
!140 = !DILocation(line: 34, column: 3, scope: !111) !56 = !DILocalVariable(name: "i", scope: !57, file: !3, line: 42, type: !7)
!142 = distinct !DISubprogram(name: "negative_dependence", scope: !6, file: !6, line: 37, type: !17, scopeLine: 37, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !143) !57 = distinct !DILexicalBlock(scope: !51, file: !3, line: 42, column: 3)
!143 = !{!144, !145, !147} !58 = distinct !{!58, !59, !60}
!144 = !DILocalVariable(name: "A", arg: 1, scope: !142, file: !6, line: 37, type: !19) !59 = !DILocation(line: 38, column: 3, scope: !55)
!145 = !DILocalVariable(name: "i", scope: !146, file: !6, line: 38, type: !8) !60 = !DILocation(line: 40, column: 3, scope: !55)
!146 = distinct !DILexicalBlock(scope: !142, file: !6, line: 38, column: 3) !61 = !DILocation(line: 0, scope: !57)
!147 = !DILocalVariable(name: "i", scope: !148, file: !6, line: 42, type: !8) !62 = !DILocation(line: 45, column: 1, scope: !51)
!148 = distinct !DILexicalBlock(scope: !142, file: !6, line: 42, column: 3) !63 = distinct !DISubprogram(name: "sumTest", scope: !3, file: !3, line: 48, type: !64, scopeLine: 48, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !66)
!153 = distinct !DILexicalBlock(scope: !146, file: !6, line: 38, column: 3) !64 = !DISubroutineType(types: !65)
!154 = !DILocation(line: 38, column: 3, scope: !146) !65 = !{!7, !17}
!160 = !DILocation(line: 38, column: 3, scope: !153) !66 = !{!67, !68, !69, !71}
!161 = distinct !{!161, !154, !162} !67 = !DILocalVariable(name: "A", arg: 1, scope: !63, file: !3, line: 48, type: !17)
!162 = !DILocation(line: 40, column: 3, scope: !146) !68 = !DILocalVariable(name: "sum", scope: !63, file: !3, line: 49, type: !7)
!163 = !DILocation(line: 0, scope: !148) !69 = !DILocalVariable(name: "i", scope: !70, file: !3, line: 51, type: !7)
!164 = !DILocation(line: 42, column: 8, scope: !148) !70 = distinct !DILexicalBlock(scope: !63, file: !3, line: 51, column: 3)
!178 = !DILocation(line: 45, column: 1, scope: !142) !71 = !DILocalVariable(name: "i", scope: !72, file: !3, line: 54, type: !7)
!179 = distinct !DISubprogram(name: "sumTest", scope: !6, file: !6, line: 48, type: !180, scopeLine: 48, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !182) !72 = distinct !DILexicalBlock(scope: !63, file: !3, line: 54, column: 3)
!180 = !DISubroutineType(types: !181) !73 = distinct !{!73, !74, !75}
!181 = !{!8, !19} !74 = !DILocation(line: 51, column: 3, scope: !70)
!182 = !{!183, !184, !185, !187} !75 = !DILocation(line: 52, column: 15, scope: !70)
!183 = !DILocalVariable(name: "A", arg: 1, scope: !179, file: !6, line: 48, type: !19) !76 = !DILocation(line: 57, column: 3, scope: !63)
!184 = !DILocalVariable(name: "sum", scope: !179, file: !6, line: 49, type: !8)
!185 = !DILocalVariable(name: "i", scope: !186, file: !6, line: 51, type: !8)
!186 = distinct !DILexicalBlock(scope: !179, file: !6, line: 51, column: 3)
!187 = !DILocalVariable(name: "i", scope: !188, file: !6, line: 54, type: !8)
!188 = distinct !DILexicalBlock(scope: !179, file: !6, line: 54, column: 3)
!193 = distinct !DILexicalBlock(scope: !186, file: !6, line: 51, column: 3)
!194 = !DILocation(line: 51, column: 3, scope: !186)
!199 = !DILocation(line: 51, column: 3, scope: !193)
!200 = distinct !{!200, !194, !201}
!201 = !DILocation(line: 52, column: 15, scope: !186)
!215 = !DILocation(line: 57, column: 3, scope: !179)

llvm/test/Transforms/LoopFusion/four_loops.ll

; RUN: opt -S -loop-fusion < %s | FileCheck %s ; RUN: opt -S -loop-fusion < %s | FileCheck %s
@A = common global [1024 x i32] zeroinitializer, align 16 @A = common global [1024 x i32] zeroinitializer, align 16
@B = common global [1024 x i32] zeroinitializer, align 16 @B = common global [1024 x i32] zeroinitializer, align 16
@C = common global [1024 x i32] zeroinitializer, align 16 @C = common global [1024 x i32] zeroinitializer, align 16
@D = common global [1024 x i32] zeroinitializer, align 16 @D = common global [1024 x i32] zeroinitializer, align 16
; CHECK: void @dep_free ; CHECK: void @dep_free
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]+]] ; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]+]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %exitcond12, label %[[LOOP1BODY:bb[0-9]+]], label %[[LOOP2PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP1BODY]]
; CHECK: br label %[[LOOP1LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP1LATCH:bb[0-9]+]]
; CHECK: [[LOOP1LATCH]] ; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2PREHEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]+]], label %[[LOOP2BODY]]
; CHECK: [[LOOP2PREHEADER]] ; CHECK: [[LOOP2BODY]]
; CHECK: br i1 %exitcond9, label %[[LOOP2HEADER:bb[0-9]+]], label %[[LOOP3PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP2HEADER]]
; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]] ; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP3PREHEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP3BODY:bb[0-9]+]], label %[[LOOP3BODY]]
; CHECK: [[LOOP3PREHEADER]] ; CHECK: [[LOOP3BODY]]
; CHECK: br i1 %exitcond6, label %[[LOOP3HEADER:bb[0-9]+]], label %[[LOOP4PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP3HEADER]]
; CHECK: br label %[[LOOP3LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP3LATCH:bb[0-9]+]]
; CHECK: [[LOOP3LATCH]] ; CHECK: [[LOOP3LATCH]]
; CHECK: br label %[[LOOP4PREHEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP4BODY:bb[0-9]+]], label %[[LOOP4BODY]]
; CHECK: [[LOOP4PREHEADER]] ; CHECK: [[LOOP4BODY]]
; CHECK: br i1 %exitcond, label %[[LOOP4HEADER:bb[0-9]+]], label %[[LOOP4EXIT:bb[0-9]+]]
; CHECK: [[LOOP4EXIT]]
; CHECK: br label %[[FUNCEXIT:bb[0-9]+]]
; CHECK: [[LOOP4HEADER]]
; CHECK: br label %[[LOOP4LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP4LATCH:bb[0-9]+]]
; CHECK: [[LOOP4LATCH]] ; CHECK: [[LOOP4LATCH]]
; CHECK: br label %[[LOOP1HEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %[[LOOPEXIT:bb[0-9]+]]
; CHECK: [[FUNCEXIT]]
; CHECK: ret void ; CHECK: ret void
define void @dep_free() { define void @dep_free() {
bb: bb:
br label %bb13 br label %bb15
bb13: ; preds = %bb22, %bb bb25.preheader: ; preds = %bb22
%indvars.iv10 = phi i64 [ %indvars.iv.next11, %bb22 ], [ 0, %bb ] br label %bb27
%.0 = phi i32 [ 0, %bb ], [ %tmp23, %bb22 ]
%exitcond12 = icmp ne i64 %indvars.iv10, 100 bb15: ; preds = %bb, %bb22
br i1 %exitcond12, label %bb15, label %bb25 %.08 = phi i32 [ 0, %bb ], [ %tmp23, %bb22 ]
%indvars.iv107 = phi i64 [ 0, %bb ], [ %indvars.iv.next11, %bb22 ]
bb15: ; preds = %bb13 %tmp = add nsw i32 %.08, -3
%tmp = add nsw i32 %.0, -3 %tmp16 = add nuw nsw i64 %indvars.iv107, 3
%tmp16 = add nuw nsw i64 %indvars.iv10, 3
%tmp17 = trunc i64 %tmp16 to i32 %tmp17 = trunc i64 %tmp16 to i32
%tmp18 = mul nsw i32 %tmp, %tmp17 %tmp18 = mul nsw i32 %tmp, %tmp17
%tmp19 = trunc i64 %indvars.iv10 to i32 %tmp19 = trunc i64 %indvars.iv107 to i32
%tmp20 = srem i32 %tmp18, %tmp19 %tmp20 = srem i32 %tmp18, %tmp19
%tmp21 = getelementptr inbounds [1024 x i32], [1024 x i32]* @A, i64 0, i64 %indvars.iv10 %tmp21 = getelementptr inbounds [1024 x i32], [1024 x i32]* @A, i64 0, i64 %indvars.iv107
store i32 %tmp20, i32* %tmp21, align 4 store i32 %tmp20, i32* %tmp21, align 4
br label %bb22 br label %bb22
bb22: ; preds = %bb15 bb22: ; preds = %bb15
%indvars.iv.next11 = add nuw nsw i64 %indvars.iv10, 1 %indvars.iv.next11 = add nuw nsw i64 %indvars.iv107, 1
%tmp23 = add nuw nsw i32 %.0, 1 %tmp23 = add nuw nsw i32 %.08, 1
br label %bb13 %exitcond12 = icmp ne i64 %indvars.iv.next11, 100
br i1 %exitcond12, label %bb15, label %bb25.preheader
bb25: ; preds = %bb35, %bb13
%indvars.iv7 = phi i64 [ %indvars.iv.next8, %bb35 ], [ 0, %bb13 ] bb38.preheader: ; preds = %bb35
%.01 = phi i32 [ 0, %bb13 ], [ %tmp36, %bb35 ] br label %bb40
%exitcond9 = icmp ne i64 %indvars.iv7, 100
br i1 %exitcond9, label %bb27, label %bb38 bb27: ; preds = %bb25.preheader, %bb35
%.016 = phi i32 [ 0, %bb25.preheader ], [ %tmp36, %bb35 ]
bb27: ; preds = %bb25 %indvars.iv75 = phi i64 [ 0, %bb25.preheader ], [ %indvars.iv.next8, %bb35 ]
%tmp28 = add nsw i32 %.01, -3 %tmp28 = add nsw i32 %.016, -3
%tmp29 = add nuw nsw i64 %indvars.iv7, 3 %tmp29 = add nuw nsw i64 %indvars.iv75, 3
%tmp30 = trunc i64 %tmp29 to i32 %tmp30 = trunc i64 %tmp29 to i32
%tmp31 = mul nsw i32 %tmp28, %tmp30 %tmp31 = mul nsw i32 %tmp28, %tmp30
%tmp32 = trunc i64 %indvars.iv7 to i32 %tmp32 = trunc i64 %indvars.iv75 to i32
%tmp33 = srem i32 %tmp31, %tmp32 %tmp33 = srem i32 %tmp31, %tmp32
%tmp34 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv7 %tmp34 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv75
store i32 %tmp33, i32* %tmp34, align 4 store i32 %tmp33, i32* %tmp34, align 4
br label %bb35 br label %bb35
bb35: ; preds = %bb27 bb35: ; preds = %bb27
%indvars.iv.next8 = add nuw nsw i64 %indvars.iv7, 1 %indvars.iv.next8 = add nuw nsw i64 %indvars.iv75, 1
%tmp36 = add nuw nsw i32 %.01, 1 %tmp36 = add nuw nsw i32 %.016, 1
br label %bb25 %exitcond9 = icmp ne i64 %indvars.iv.next8, 100
br i1 %exitcond9, label %bb27, label %bb38.preheader
bb38: ; preds = %bb48, %bb25
%indvars.iv4 = phi i64 [ %indvars.iv.next5, %bb48 ], [ 0, %bb25 ] bb51.preheader: ; preds = %bb48
%.02 = phi i32 [ 0, %bb25 ], [ %tmp49, %bb48 ] br label %bb53
%exitcond6 = icmp ne i64 %indvars.iv4, 100
br i1 %exitcond6, label %bb40, label %bb51 bb40: ; preds = %bb38.preheader, %bb48
%.024 = phi i32 [ 0, %bb38.preheader ], [ %tmp49, %bb48 ]
bb40: ; preds = %bb38 %indvars.iv43 = phi i64 [ 0, %bb38.preheader ], [ %indvars.iv.next5, %bb48 ]
%tmp41 = add nsw i32 %.02, -3 %tmp41 = add nsw i32 %.024, -3
%tmp42 = add nuw nsw i64 %indvars.iv4, 3 %tmp42 = add nuw nsw i64 %indvars.iv43, 3
%tmp43 = trunc i64 %tmp42 to i32 %tmp43 = trunc i64 %tmp42 to i32
%tmp44 = mul nsw i32 %tmp41, %tmp43 %tmp44 = mul nsw i32 %tmp41, %tmp43
%tmp45 = trunc i64 %indvars.iv4 to i32 %tmp45 = trunc i64 %indvars.iv43 to i32
%tmp46 = srem i32 %tmp44, %tmp45 %tmp46 = srem i32 %tmp44, %tmp45
%tmp47 = getelementptr inbounds [1024 x i32], [1024 x i32]* @C, i64 0, i64 %indvars.iv4 %tmp47 = getelementptr inbounds [1024 x i32], [1024 x i32]* @C, i64 0, i64 %indvars.iv43
store i32 %tmp46, i32* %tmp47, align 4 store i32 %tmp46, i32* %tmp47, align 4
br label %bb48 br label %bb48
bb48: ; preds = %bb40 bb48: ; preds = %bb40
%indvars.iv.next5 = add nuw nsw i64 %indvars.iv4, 1 %indvars.iv.next5 = add nuw nsw i64 %indvars.iv43, 1
%tmp49 = add nuw nsw i32 %.02, 1 %tmp49 = add nuw nsw i32 %.024, 1
br label %bb38 %exitcond6 = icmp ne i64 %indvars.iv.next5, 100
br i1 %exitcond6, label %bb40, label %bb51.preheader
bb51: ; preds = %bb61, %bb38
%indvars.iv = phi i64 [ %indvars.iv.next, %bb61 ], [ 0, %bb38 ]
%.03 = phi i32 [ 0, %bb38 ], [ %tmp62, %bb61 ]
%exitcond = icmp ne i64 %indvars.iv, 100
br i1 %exitcond, label %bb53, label %bb52
bb52: ; preds = %bb51 bb52: ; preds = %bb61
br label %bb63 br label %bb63
bb53: ; preds = %bb51 bb53: ; preds = %bb51.preheader, %bb61
%tmp54 = add nsw i32 %.03, -3 %.032 = phi i32 [ 0, %bb51.preheader ], [ %tmp62, %bb61 ]
%tmp55 = add nuw nsw i64 %indvars.iv, 3 %indvars.iv1 = phi i64 [ 0, %bb51.preheader ], [ %indvars.iv.next, %bb61 ]
%tmp54 = add nsw i32 %.032, -3
%tmp55 = add nuw nsw i64 %indvars.iv1, 3
%tmp56 = trunc i64 %tmp55 to i32 %tmp56 = trunc i64 %tmp55 to i32
%tmp57 = mul nsw i32 %tmp54, %tmp56 %tmp57 = mul nsw i32 %tmp54, %tmp56
%tmp58 = trunc i64 %indvars.iv to i32 %tmp58 = trunc i64 %indvars.iv1 to i32
%tmp59 = srem i32 %tmp57, %tmp58 %tmp59 = srem i32 %tmp57, %tmp58
%tmp60 = getelementptr inbounds [1024 x i32], [1024 x i32]* @D, i64 0, i64 %indvars.iv %tmp60 = getelementptr inbounds [1024 x i32], [1024 x i32]* @D, i64 0, i64 %indvars.iv1
store i32 %tmp59, i32* %tmp60, align 4 store i32 %tmp59, i32* %tmp60, align 4
br label %bb61 br label %bb61
bb61: ; preds = %bb53 bb61: ; preds = %bb53
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
%tmp62 = add nuw nsw i32 %.03, 1 %tmp62 = add nuw nsw i32 %.032, 1
br label %bb51 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb53, label %bb52
bb63: ; preds = %bb52 bb63: ; preds = %bb52
ret void ret void
} }

llvm/test/Transforms/LoopFusion/loop_nest.ll

Show All 17 Lines
; ;
@A = common global [1024 x [1024 x i32]] zeroinitializer, align 16 @A = common global [1024 x [1024 x i32]] zeroinitializer, align 16
@B = common global [1024 x [1024 x i32]] zeroinitializer, align 16 @B = common global [1024 x [1024 x i32]] zeroinitializer, align 16
; CHECK: void @dep_free ; CHECK: void @dep_free
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]+]] ; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]+]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %exitcond12, label %[[LOOP3PREHEADER:bb[0-9]+.preheader]], label %[[LOOP2HEADER:bb[0-9]+]]
; CHECK: [[LOOP3PREHEADER]]
; CHECK: br label %[[LOOP3HEADER:bb[0-9]+]] ; CHECK: br label %[[LOOP3HEADER:bb[0-9]+]]
; CHECK: [[LOOP3HEADER]] ; CHECK: [[LOOP3HEADER]]
; CHECK: br i1 %exitcond9, label %[[LOOP3BODY:bb[0-9]+]], label %[[LOOP1LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP3LATCH:bb[0-9]+]]
; CHECK: [[LOOP3LATCH]]
; CHECK: br i1 %{{.*}}, label %[[LOOP3HEADER]], label %[[LOOP1LATCH:bb[0-9]+]]
; CHECK: [[LOOP1LATCH]] ; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2HEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2PREHEADER:bb[0-9]+]], label %[[LOOP2PREHEADER]]
; CHECK: [[LOOP2HEADER]] ; CHECK: [[LOOP2PREHEADER]]
; CHECK: br i1 %exitcond6, label %[[LOOP4PREHEADER:bb[0-9]+.preheader]], label %[[LOOP2EXITBLOCK:bb[0-9]+]]
; CHECK: [[LOOP4PREHEADER]]
; CHECK: br label %[[LOOP4HEADER:bb[0-9]+]] ; CHECK: br label %[[LOOP4HEADER:bb[0-9]+]]
; CHECK: [[LOOP2EXITBLOCK]]
; CHECK-NEXT: br label %[[FUNCEXIT:bb[0-9]+]]
; CHECK: [[LOOP4HEADER]] ; CHECK: [[LOOP4HEADER]]
; CHECK: br i1 %exitcond, label %[[LOOP4BODY:bb[0-9]+]], label %[[LOOP2LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP4LATCH:bb[0-9]+]]
; CHECK: [[LOOP4LATCH]]
; CHECK: br i1 %{{.*}}, label %[[LOOP4HEADER]], label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]] ; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP1HEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %[[LOOP1EXIT:bb[0-9]*]]
; CHECK: [[FUNCEXIT]]
; CHECK: ret void ; CHECK: ret void
; TODO: The current version of loop fusion does not allow the inner loops to be ; TODO: The current version of loop fusion does not allow the inner loops to be
; fused because they are not control flow equivalent and adjacent. These are ; fused because they are not control flow equivalent and adjacent. These are
; limitations that can be addressed in future improvements to fusion. ; limitations that can be addressed in future improvements to fusion.
define void @dep_free() { define void @dep_free() {
bb: bb:
br label %bb13 br label %bb16
bb13: ; preds = %bb27, %bb bb16: ; preds = %bb, %bb27
%indvars.iv10 = phi i64 [ %indvars.iv.next11, %bb27 ], [ 0, %bb ] %.06 = phi i32 [ 0, %bb ], [ %tmp28, %bb27 ]
%.0 = phi i32 [ 0, %bb ], [ %tmp28, %bb27 ] %indvars.iv105 = phi i64 [ 0, %bb ], [ %indvars.iv.next11, %bb27 ]
%exitcond12 = icmp ne i64 %indvars.iv10, 100 br label %bb18
br i1 %exitcond12, label %bb16, label %bb30
bb16: ; preds = %bb25, %bb13 bb30: ; preds = %bb27
%indvars.iv7 = phi i64 [ %indvars.iv.next8, %bb25 ], [ 0, %bb13 ] br label %bb33
%exitcond9 = icmp ne i64 %indvars.iv7, 100
br i1 %exitcond9, label %bb18, label %bb27
bb18: ; preds = %bb16 bb18: ; preds = %bb16, %bb25
%tmp = add nsw i32 %.0, -3 %indvars.iv74 = phi i64 [ 0, %bb16 ], [ %indvars.iv.next8, %bb25 ]
%tmp19 = add nuw nsw i64 %indvars.iv10, 3 %tmp = add nsw i32 %.06, -3
%tmp19 = add nuw nsw i64 %indvars.iv105, 3
%tmp20 = trunc i64 %tmp19 to i32 %tmp20 = trunc i64 %tmp19 to i32
%tmp21 = mul nsw i32 %tmp, %tmp20 %tmp21 = mul nsw i32 %tmp, %tmp20
%tmp22 = trunc i64 %indvars.iv10 to i32 %tmp22 = trunc i64 %indvars.iv105 to i32
%tmp23 = srem i32 %tmp21, %tmp22 %tmp23 = srem i32 %tmp21, %tmp22
%tmp24 = getelementptr inbounds [1024 x [1024 x i32]], [1024 x [1024 x i32]]* @A, i64 0, i64 %indvars.iv10, i64 %indvars.iv7 %tmp24 = getelementptr inbounds [1024 x [1024 x i32]], [1024 x [1024 x i32]]* @A, i64 0, i64 %indvars.iv105, i64 %indvars.iv74
store i32 %tmp23, i32* %tmp24, align 4 store i32 %tmp23, i32* %tmp24, align 4
br label %bb25 br label %bb25
bb25: ; preds = %bb18 bb25: ; preds = %bb18
%indvars.iv.next8 = add nuw nsw i64 %indvars.iv7, 1 %indvars.iv.next8 = add nuw nsw i64 %indvars.iv74, 1
br label %bb16 %exitcond9 = icmp ne i64 %indvars.iv.next8, 100
br i1 %exitcond9, label %bb18, label %bb27
bb27: ; preds = %bb16 bb27: ; preds = %bb25
%indvars.iv.next11 = add nuw nsw i64 %indvars.iv10, 1 %indvars.iv.next11 = add nuw nsw i64 %indvars.iv105, 1
%tmp28 = add nuw nsw i32 %.0, 1 %tmp28 = add nuw nsw i32 %.06, 1
br label %bb13 %exitcond12 = icmp ne i64 %indvars.iv.next11, 100
br i1 %exitcond12, label %bb16, label %bb30
bb30: ; preds = %bb45, %bb13
%indvars.iv4 = phi i64 [ %indvars.iv.next5, %bb45 ], [ 0, %bb13 ]
%.02 = phi i32 [ 0, %bb13 ], [ %tmp46, %bb45 ]
%exitcond6 = icmp ne i64 %indvars.iv4, 100
br i1 %exitcond6, label %bb33, label %bb31
bb31: ; preds = %bb30 bb33: ; preds = %bb30, %bb45
br label %bb47 %.023 = phi i32 [ 0, %bb30 ], [ %tmp46, %bb45 ]
%indvars.iv42 = phi i64 [ 0, %bb30 ], [ %indvars.iv.next5, %bb45 ]
br label %bb35
bb33: ; preds = %bb43, %bb30 bb31: ; preds = %bb45
%indvars.iv = phi i64 [ %indvars.iv.next, %bb43 ], [ 0, %bb30 ] br label %bb47
%exitcond = icmp ne i64 %indvars.iv, 100
br i1 %exitcond, label %bb35, label %bb45
bb35: ; preds = %bb33 bb35: ; preds = %bb33, %bb43
%tmp36 = add nsw i32 %.02, -3 %indvars.iv1 = phi i64 [ 0, %bb33 ], [ %indvars.iv.next, %bb43 ]
%tmp37 = add nuw nsw i64 %indvars.iv4, 3 %tmp36 = add nsw i32 %.023, -3
%tmp37 = add nuw nsw i64 %indvars.iv42, 3
%tmp38 = trunc i64 %tmp37 to i32 %tmp38 = trunc i64 %tmp37 to i32
%tmp39 = mul nsw i32 %tmp36, %tmp38 %tmp39 = mul nsw i32 %tmp36, %tmp38
%tmp40 = trunc i64 %indvars.iv4 to i32 %tmp40 = trunc i64 %indvars.iv42 to i32
%tmp41 = srem i32 %tmp39, %tmp40 %tmp41 = srem i32 %tmp39, %tmp40
%tmp42 = getelementptr inbounds [1024 x [1024 x i32]], [1024 x [1024 x i32]]* @B, i64 0, i64 %indvars.iv4, i64 %indvars.iv %tmp42 = getelementptr inbounds [1024 x [1024 x i32]], [1024 x [1024 x i32]]* @B, i64 0, i64 %indvars.iv42, i64 %indvars.iv1
store i32 %tmp41, i32* %tmp42, align 4 store i32 %tmp41, i32* %tmp42, align 4
br label %bb43 br label %bb43
bb43: ; preds = %bb35 bb43: ; preds = %bb35
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
br label %bb33 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb35, label %bb45
bb45: ; preds = %bb33 bb45: ; preds = %bb43
%indvars.iv.next5 = add nuw nsw i64 %indvars.iv4, 1 %indvars.iv.next5 = add nuw nsw i64 %indvars.iv42, 1
%tmp46 = add nuw nsw i32 %.02, 1 %tmp46 = add nuw nsw i32 %.023, 1
br label %bb30 %exitcond6 = icmp ne i64 %indvars.iv.next5, 100
br i1 %exitcond6, label %bb33, label %bb31
bb47: ; preds = %bb31 bb47: ; preds = %bb31
ret void ret void
} }

llvm/test/Transforms/LoopFusion/simple.ll

; RUN: opt -S -loop-fusion < %s | FileCheck %s ; RUN: opt -S -loop-fusion < %s | FileCheck %s
@B = common global [1024 x i32] zeroinitializer, align 16 @B = common global [1024 x i32] zeroinitializer, align 16
; CHECK: void @dep_free ; CHECK: void @dep_free
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]*]] ; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]*]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP1BODY:bb[0-9]*]], label %[[LOOP2PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP1BODY]]
; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]] ; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]]
; CHECK: [[LOOP1LATCH]] ; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2PREHEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2HEADER:bb[0-9]+]], label %[[LOOP2HEADER]]
; CHECK: [[LOOP2PREHEADER]] ; CHECK: [[LOOP2HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]*]], label %[[LOOP2EXIT:bb[0-9]*]]
; CHECK: [[LOOP2BODY]]
; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]] ; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP1HEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %{{.*}}
; CHECK: ret void ; CHECK: ret void
define void @dep_free(i32* noalias %arg) { define void @dep_free(i32* noalias %arg) {
bb: bb:
br label %bb5 br label %bb7
bb5: ; preds = %bb14, %bb bb7: ; preds = %bb, %bb14
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb14 ], [ 0, %bb ] %.014 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ]
%.01 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ] %indvars.iv23 = phi i64 [ 0, %bb ], [ %indvars.iv.next3, %bb14 ]
%exitcond4 = icmp ne i64 %indvars.iv2, 100 %tmp = add nsw i32 %.014, -3
br i1 %exitcond4, label %bb7, label %bb17 %tmp8 = add nuw nsw i64 %indvars.iv23, 3
bb7: ; preds = %bb5
%tmp = add nsw i32 %.01, -3
%tmp8 = add nuw nsw i64 %indvars.iv2, 3
%tmp9 = trunc i64 %tmp8 to i32 %tmp9 = trunc i64 %tmp8 to i32
%tmp10 = mul nsw i32 %tmp, %tmp9 %tmp10 = mul nsw i32 %tmp, %tmp9
%tmp11 = trunc i64 %indvars.iv2 to i32 %tmp11 = trunc i64 %indvars.iv23 to i32
%tmp12 = srem i32 %tmp10, %tmp11 %tmp12 = srem i32 %tmp10, %tmp11
%tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2 %tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv23
store i32 %tmp12, i32* %tmp13, align 4 store i32 %tmp12, i32* %tmp13, align 4
br label %bb14 br label %bb14
bb14: ; preds = %bb7 bb14: ; preds = %bb7
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1 %indvars.iv.next3 = add nuw nsw i64 %indvars.iv23, 1
%tmp15 = add nuw nsw i32 %.01, 1 %tmp15 = add nuw nsw i32 %.014, 1
br label %bb5 %exitcond4 = icmp ne i64 %indvars.iv.next3, 100
br i1 %exitcond4, label %bb7, label %bb17.preheader
bb17: ; preds = %bb27, %bb5
%indvars.iv = phi i64 [ %indvars.iv.next, %bb27 ], [ 0, %bb5 ] bb17.preheader: ; preds = %bb14
%.0 = phi i32 [ 0, %bb5 ], [ %tmp28, %bb27 ] br label %bb19
%exitcond = icmp ne i64 %indvars.iv, 100
br i1 %exitcond, label %bb19, label %bb18 bb19: ; preds = %bb17.preheader, %bb27
%.02 = phi i32 [ 0, %bb17.preheader ], [ %tmp28, %bb27 ]
bb18: ; preds = %bb17 %indvars.iv1 = phi i64 [ 0, %bb17.preheader ], [ %indvars.iv.next, %bb27 ]
br label %bb29 %tmp20 = add nsw i32 %.02, -3
%tmp21 = add nuw nsw i64 %indvars.iv1, 3
bb19: ; preds = %bb17
%tmp20 = add nsw i32 %.0, -3
%tmp21 = add nuw nsw i64 %indvars.iv, 3
%tmp22 = trunc i64 %tmp21 to i32 %tmp22 = trunc i64 %tmp21 to i32
%tmp23 = mul nsw i32 %tmp20, %tmp22 %tmp23 = mul nsw i32 %tmp20, %tmp22
%tmp24 = trunc i64 %indvars.iv to i32 %tmp24 = trunc i64 %indvars.iv1 to i32
%tmp25 = srem i32 %tmp23, %tmp24 %tmp25 = srem i32 %tmp23, %tmp24
%tmp26 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %tmp26 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv1
store i32 %tmp25, i32* %tmp26, align 4 store i32 %tmp25, i32* %tmp26, align 4
br label %bb27 br label %bb27
bb27: ; preds = %bb19 bb27: ; preds = %bb19
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
%tmp28 = add nuw nsw i32 %.0, 1 %tmp28 = add nuw nsw i32 %.02, 1
br label %bb17 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb19, label %bb18
bb18: ; preds = %bb27
br label %bb29
bb29: ; preds = %bb18 bb29: ; preds = %bb18
ret void ret void
} }
; CHECK: void @dep_free_parametric ; CHECK: void @dep_free_parametric
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]*]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1PREHEADER:bb[0-9.a-z]*]], label %[[EXITBLOCK:bb[0-9]*]]
; CHECK: [[LOOP1PREHEADER]]
; CHECK: br label %[[LOOP1HEADER:bb[0-9]*]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP1BODY:bb[0-9]*]], label %[[LOOP2PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP1BODY]]
; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]] ; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]]
; CHECK: [[LOOP1LATCH]] ; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2PREHEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2HEADER:bb[0-9]*]], label %[[LOOP2HEADER]]
; CHECK: [[LOOP2PREHEADER]] ; CHECK: [[LOOP2HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]*]], label %[[LOOP2EXIT:bb[0-9]*]]
; CHECK: [[LOOP2BODY]]
; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]] ; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP1HEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %[[EXITBLOCK]]
; CHECK: ret void ; CHECK: ret void
define void @dep_free_parametric(i32* noalias %arg, i64 %arg2) { define void @dep_free_parametric(i32* noalias %arg, i64 %arg2) {
bb: bb:
br label %bb3 %tmp3 = icmp slt i64 0, %arg2
br i1 %tmp3, label %bb5, label %bb15.preheader
bb3: ; preds = %bb12, %bb bb5: ; preds = %bb5, %bb12
%.01 = phi i64 [ 0, %bb ], [ %tmp13, %bb12 ] %.014 = phi i64 [ 0, %bb ], [ %tmp13, %bb12 ]
%tmp = icmp slt i64 %.01, %arg2 %tmp6 = add nsw i64 %.014, -3
br i1 %tmp, label %bb5, label %bb15 %tmp7 = add nuw nsw i64 %.014, 3
bb5: ; preds = %bb3
%tmp6 = add nsw i64 %.01, -3
%tmp7 = add nuw nsw i64 %.01, 3
%tmp8 = mul nsw i64 %tmp6, %tmp7 %tmp8 = mul nsw i64 %tmp6, %tmp7
%tmp9 = srem i64 %tmp8, %.01 %tmp9 = srem i64 %tmp8, %.014
%tmp10 = trunc i64 %tmp9 to i32 %tmp10 = trunc i64 %tmp9 to i32
%tmp11 = getelementptr inbounds i32, i32* %arg, i64 %.01 %tmp11 = getelementptr inbounds i32, i32* %arg, i64 %.014
store i32 %tmp10, i32* %tmp11, align 4 store i32 %tmp10, i32* %tmp11, align 4
br label %bb12 br label %bb12
bb12: ; preds = %bb5 bb12: ; preds = %bb5
%tmp13 = add nuw nsw i64 %.01, 1 %tmp13 = add nuw nsw i64 %.014, 1
br label %bb3 %tmp = icmp slt i64 %tmp13, %arg2
br i1 %tmp, label %bb5, label %bb15.preheader
bb15: ; preds = %bb25, %bb3
%.0 = phi i64 [ 0, %bb3 ], [ %tmp26, %bb25 ] bb15.preheader: ; preds = %bb12, %bb
%tmp16 = icmp slt i64 %.0, %arg2 %tmp161 = icmp slt i64 0, %arg2
br i1 %tmp16, label %bb18, label %bb17 br i1 %tmp161, label %bb18, label %bb27
bb17: ; preds = %bb15 bb18: ; preds = %bb15.preheader, %bb25
br label %bb27 %.02 = phi i64 [ 0, %bb15.preheader ], [ %tmp26, %bb25 ]
%tmp19 = add nsw i64 %.02, -3
bb18: ; preds = %bb15 %tmp20 = add nuw nsw i64 %.02, 3
%tmp19 = add nsw i64 %.0, -3
%tmp20 = add nuw nsw i64 %.0, 3
%tmp21 = mul nsw i64 %tmp19, %tmp20 %tmp21 = mul nsw i64 %tmp19, %tmp20
%tmp22 = srem i64 %tmp21, %.0 %tmp22 = srem i64 %tmp21, %.02
%tmp23 = trunc i64 %tmp22 to i32 %tmp23 = trunc i64 %tmp22 to i32
%tmp24 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %.0 %tmp24 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %.02
store i32 %tmp23, i32* %tmp24, align 4 store i32 %tmp23, i32* %tmp24, align 4
br label %bb25 br label %bb25
bb25: ; preds = %bb18 bb25: ; preds = %bb18
%tmp26 = add nuw nsw i64 %.0, 1 %tmp26 = add nuw nsw i64 %.02, 1
br label %bb15 %tmp16 = icmp slt i64 %tmp26, %arg2
br i1 %tmp16, label %bb18, label %bb27
bb27: ; preds = %bb17 bb27: ; preds = %bb17
ret void ret void
} }
; CHECK: void @raw_only ; CHECK: void @raw_only
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]*]] ; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]*]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP1BODY:bb[0-9]*]], label %[[LOOP2PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP1BODY]]
; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]] ; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]]
; CHECK: [[LOOP1LATCH]] ; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2PREHEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2HEADER:bb[0-9]+]], label %[[LOOP2HEADER]]
; CHECK: [[LOOP2PREHEADER]] ; CHECK: [[LOOP2HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]*]], label %[[LOOP2EXIT:bb[0-9]*]]
; CHECK: [[LOOP2BODY]]
; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]] ; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP1HEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %{{.*}}
; CHECK: ret void ; CHECK: ret void
define void @raw_only(i32* noalias %arg) { define void @raw_only(i32* noalias %arg) {
bb: bb:
br label %bb5 br label %bb7
bb11.preheader: ; preds = %bb9
br label %bb13
bb5: ; preds = %bb9, %bb bb7: ; preds = %bb, %bb9
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb9 ], [ 0, %bb ] %indvars.iv22 = phi i64 [ 0, %bb ], [ %indvars.iv.next3, %bb9 ]
%exitcond4 = icmp ne i64 %indvars.iv2, 100 %tmp = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv22
br i1 %exitcond4, label %bb7, label %bb11 %tmp8 = trunc i64 %indvars.iv22 to i32
bb7: ; preds = %bb5
%tmp = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2
%tmp8 = trunc i64 %indvars.iv2 to i32
store i32 %tmp8, i32* %tmp, align 4 store i32 %tmp8, i32* %tmp, align 4
br label %bb9 br label %bb9
bb9: ; preds = %bb7 bb9: ; preds = %bb7
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1 %indvars.iv.next3 = add nuw nsw i64 %indvars.iv22, 1
br label %bb5 %exitcond4 = icmp ne i64 %indvars.iv.next3, 100
br i1 %exitcond4, label %bb7, label %bb11.preheader
bb11: ; preds = %bb18, %bb5
%indvars.iv = phi i64 [ %indvars.iv.next, %bb18 ], [ 0, %bb5 ] bb13: ; preds = %bb11.preheader, %bb18
%exitcond = icmp ne i64 %indvars.iv, 100 %indvars.iv1 = phi i64 [ 0, %bb11.preheader ], [ %indvars.iv.next, %bb18 ]
br i1 %exitcond, label %bb13, label %bb19 %tmp14 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv1
bb13: ; preds = %bb11
%tmp14 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv
%tmp15 = load i32, i32* %tmp14, align 4 %tmp15 = load i32, i32* %tmp14, align 4
%tmp16 = shl nsw i32 %tmp15, 1 %tmp16 = shl nsw i32 %tmp15, 1
%tmp17 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %tmp17 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv1
store i32 %tmp16, i32* %tmp17, align 4 store i32 %tmp16, i32* %tmp17, align 4
br label %bb18 br label %bb18
bb18: ; preds = %bb13 bb18: ; preds = %bb13
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
br label %bb11 %exitcond = icmp ne i64 %indvars.iv.next, 100 br i1 %exitcond, label %bb13, label %bb19
bb19: ; preds = %bb11 bb19: ; preds = %bb18
ret void ret void
} }
; CHECK: void @raw_only_parametric ; CHECK: void @raw_only_parametric
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK: br i1 %{{.*}}, label %[[LOOP1PREHEADER:bb[0-9.a-z]*]], label %[[EXITBLOCK:bb[0-9]*]]
; CHECK: [[LOOP1PREHEADER]]
; CHECK: br label %[[LOOP1HEADER:bb[0-9]*]] ; CHECK: br label %[[LOOP1HEADER:bb[0-9]*]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP1BODY:bb[0-9]*]], label %[[LOOP2PREHEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2HEADER:bb[0-9]*]], label %[[LOOP2HEADER]]
; CHECK: [[LOOP1BODY]] ; CHECK: [[LOOP2HEADER]]
; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %[[EXITBLOCK]]
; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP2PREHEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]*]], label %[[LOOP2EXIT:bb[0-9]*]]
; CHECK: [[LOOP2BODY]]
; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP1HEADER]]
; CHECK: ret void ; CHECK: ret void
define void @raw_only_parametric(i32* noalias %arg, i32 %arg4) { define void @raw_only_parametric(i32* noalias %arg, i32 %arg4) {
bb: bb:
br label %bb5
bb5: ; preds = %bb11, %bb
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb11 ], [ 0, %bb ]
%tmp = sext i32 %arg4 to i64 %tmp = sext i32 %arg4 to i64
%tmp6 = icmp slt i64 %indvars.iv2, %tmp %tmp64 = icmp sgt i32 %arg4, 0
br i1 %tmp6, label %bb8, label %bb14 br i1 %tmp64, label %bb8, label %bb23
bb8: ; preds = %bb5 bb8: ; preds = %bb, %bb8
%tmp9 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2 %indvars.iv25 = phi i64 [ %indvars.iv.next3, %bb8 ], [ 0, %bb ]
%tmp10 = trunc i64 %indvars.iv2 to i32 %tmp9 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv25
%tmp10 = trunc i64 %indvars.iv25 to i32
store i32 %tmp10, i32* %tmp9, align 4 store i32 %tmp10, i32* %tmp9, align 4
br label %bb11 %indvars.iv.next3 = add nuw nsw i64 %indvars.iv25, 1
%tmp6 = icmp slt i64 %indvars.iv.next3, %tmp
bb11: ; preds = %bb8 br i1 %tmp6, label %bb8, label %bb17
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1
br label %bb5 bb17: ; preds = %bb8, %bb17
%indvars.iv3 = phi i64 [ %indvars.iv.next, %bb17 ], [ 0, %bb8 ]
bb14: ; preds = %bb22, %bb5 %tmp18 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv3
%indvars.iv = phi i64 [ %indvars.iv.next, %bb22 ], [ 0, %bb5 ]
%tmp13 = sext i32 %arg4 to i64
%tmp15 = icmp slt i64 %indvars.iv, %tmp13
br i1 %tmp15, label %bb17, label %bb23
bb17: ; preds = %bb14
%tmp18 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv
%tmp19 = load i32, i32* %tmp18, align 4 %tmp19 = load i32, i32* %tmp18, align 4
%tmp20 = shl nsw i32 %tmp19, 1 %tmp20 = shl nsw i32 %tmp19, 1
%tmp21 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv %tmp21 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv3
store i32 %tmp20, i32* %tmp21, align 4 store i32 %tmp20, i32* %tmp21, align 4
br label %bb22 %indvars.iv.next = add nuw nsw i64 %indvars.iv3, 1
%tmp15 = icmp slt i64 %indvars.iv.next, %tmp
bb22: ; preds = %bb17 br i1 %tmp15, label %bb17, label %bb23
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
br label %bb14
bb23: ; preds = %bb14 bb23: ; preds = %bb17, %bb
ret void ret void
} }
; CHECK: void @forward_dep ; CHECK: void @forward_dep
; CHECK-NEXT: bb: ; CHECK-NEXT: bb:
; CHECK: br label %[[LOOP1HEADER:bb[0-9]*]] ; CHECK: br label %[[LOOP1HEADER:bb[0-9]*]]
; CHECK: [[LOOP1HEADER]] ; CHECK: [[LOOP1HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP1BODY:bb[0-9]*]], label %[[LOOP2PREHEADER:bb[0-9]+]]
; CHECK: [[LOOP1BODY]]
; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]] ; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]]
; CHECK: [[LOOP1LATCH]] ; CHECK: [[LOOP1LATCH]]
; CHECK: br label %[[LOOP2PREHEADER:bb[0-9]+]] ; CHECK: br i1 %{{.*}}, label %[[LOOP2HEADER:bb[0-9]+]], label %[[LOOP2HEADER]]
; CHECK: [[LOOP2PREHEADER]] ; CHECK: [[LOOP2HEADER]]
; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]*]], label %[[LOOP2EXIT:bb[0-9]*]]
; CHECK: [[LOOP2BODY]]
; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]] ; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
; CHECK: [[LOOP2LATCH]] ; CHECK: [[LOOP2LATCH]]
; CHECK: br label %[[LOOP1HEADER]] ; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %{{.*}}
; CHECK: ret void ; CHECK: ret void
define void @forward_dep(i32* noalias %arg) { define void @forward_dep(i32* noalias %arg) {
bb: bb:
br label %bb5 br label %bb7
bb5: ; preds = %bb14, %bb bb7: ; preds = %bb, %bb14
%indvars.iv2 = phi i64 [ %indvars.iv.next3, %bb14 ], [ 0, %bb ] %.013 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ]
%.01 = phi i32 [ 0, %bb ], [ %tmp15, %bb14 ] %indvars.iv22 = phi i64 [ 0, %bb ], [ %indvars.iv.next3, %bb14 ]
%exitcond4 = icmp ne i64 %indvars.iv2, 100 %tmp = add nsw i32 %.013, -3
br i1 %exitcond4, label %bb7, label %bb17 %tmp8 = add nuw nsw i64 %indvars.iv22, 3
bb7: ; preds = %bb5
%tmp = add nsw i32 %.01, -3
%tmp8 = add nuw nsw i64 %indvars.iv2, 3
%tmp9 = trunc i64 %tmp8 to i32 %tmp9 = trunc i64 %tmp8 to i32
%tmp10 = mul nsw i32 %tmp, %tmp9 %tmp10 = mul nsw i32 %tmp, %tmp9
%tmp11 = trunc i64 %indvars.iv2 to i32 %tmp11 = trunc i64 %indvars.iv22 to i32
%tmp12 = srem i32 %tmp10, %tmp11 %tmp12 = srem i32 %tmp10, %tmp11
%tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv2 %tmp13 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv22
store i32 %tmp12, i32* %tmp13, align 4 store i32 %tmp12, i32* %tmp13, align 4
br label %bb14 br label %bb14
bb14: ; preds = %bb7 bb14: ; preds = %bb7
%indvars.iv.next3 = add nuw nsw i64 %indvars.iv2, 1 %indvars.iv.next3 = add nuw nsw i64 %indvars.iv22, 1
%tmp15 = add nuw nsw i32 %.01, 1 %tmp15 = add nuw nsw i32 %.013, 1
br label %bb5 %exitcond4 = icmp ne i64 %indvars.iv.next3, 100
br i1 %exitcond4, label %bb7, label %bb19
bb17: ; preds = %bb25, %bb5
%indvars.iv = phi i64 [ %indvars.iv.next, %bb25 ], [ 0, %bb5 ] bb19: ; preds = %bb14, %bb25
%exitcond = icmp ne i64 %indvars.iv, 100 %indvars.iv1 = phi i64 [ 0, %bb14 ], [ %indvars.iv.next, %bb25 ]
br i1 %exitcond, label %bb19, label %bb26 %tmp20 = add nsw i64 %indvars.iv1, -3
bb19: ; preds = %bb17
%tmp20 = add nsw i64 %indvars.iv, -3
%tmp21 = getelementptr inbounds i32, i32* %arg, i64 %tmp20 %tmp21 = getelementptr inbounds i32, i32* %arg, i64 %tmp20
%tmp22 = load i32, i32* %tmp21, align 4 %tmp22 = load i32, i32* %tmp21, align 4
%tmp23 = mul nsw i32 %tmp22, 3 %tmp23 = mul nsw i32 %tmp22, 3
%tmp24 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv %tmp24 = getelementptr inbounds i32, i32* %arg, i64 %indvars.iv1
store i32 %tmp23, i32* %tmp24, align 4 store i32 %tmp23, i32* %tmp24, align 4
br label %bb25 br label %bb25
bb25: ; preds = %bb19 bb25: ; preds = %bb19
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 %indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
br label %bb17 %exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb19, label %bb26
bb26: ; preds = %bb17 bb26: ; preds = %bb25
ret void ret void
} }