Skip to content

Bail out jump threading on indirect branches #103688

New issue

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

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

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 11, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 9 additions & 2 deletions llvm/lib/Transforms/Utils/Local.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -1028,7 +1028,14 @@ CanRedirectPredsOfEmptyBBToSucc(BasicBlock *BB, BasicBlock *Succ,
if (!BB->hasNPredecessorsOrMore(2))
return false;

// Get single common predecessors of both BB and Succ
if (any_of(BBPreds, [](const BasicBlock *Pred) {
return isa<PHINode>(Pred->begin()) &&
isa<IndirectBrInst>(Pred->getTerminator());
}))
return false;

// Get the single common predecessor of both BB and Succ. Return false
// when there are more than one common predecessors.
for (BasicBlock *SuccPred : SuccPreds) {
if (BBPreds.count(SuccPred)) {
if (CommonPred)
Expand Down Expand Up @@ -1133,7 +1140,7 @@ bool llvm::TryToSimplifyUncondBranchFromEmptyBlock(BasicBlock *BB,

bool BBKillable = CanPropagatePredecessorsForPHIs(BB, Succ, BBPreds);

// Even if we can not fold bB into Succ, we may be able to redirect the
// Even if we can not fold BB into Succ, we may be able to redirect the
// predecessors of BB to Succ.
bool BBPhisMergeable =
BBKillable ||
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,104 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt < %s -passes=simplifycfg -S | FileCheck %s

target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"

define dso_local noundef i32 @main() {
; CHECK-LABEL: define dso_local noundef i32 @main() {
; CHECK-NEXT: [[BB:.*]]:
; CHECK-NEXT: [[ALLOCA:%.*]] = alloca [2 x ptr], align 16
; CHECK-NEXT: store ptr blockaddress(@main, %[[BB4:.*]]), ptr [[ALLOCA]], align 16, !tbaa [[TBAA0:![0-9]+]]
; CHECK-NEXT: [[GETELEMENTPTR:%.*]] = getelementptr inbounds [2 x ptr], ptr [[ALLOCA]], i64 0, i64 1
; CHECK-NEXT: store ptr blockaddress(@main, %[[BB10:.*]]), ptr [[GETELEMENTPTR]], align 8, !tbaa [[TBAA0]]
; CHECK-NEXT: br label %[[BB1:.*]]
; CHECK: [[BB1]]:
; CHECK-NEXT: [[PHI:%.*]] = phi i32 [ 0, %[[BB]] ], [ [[PHI8:%.*]], %[[BB7:.*]] ]
; CHECK-NEXT: [[PHI2:%.*]] = phi i32 [ 0, %[[BB]] ], [ [[PHI9:%.*]], %[[BB7]] ]
; CHECK-NEXT: switch i32 [[PHI]], label %[[BB7]] [
; CHECK-NEXT: i32 0, label %[[BB12:.*]]
; CHECK-NEXT: i32 1, label %[[BB4]]
; CHECK-NEXT: i32 2, label %[[BB6:.*]]
; CHECK-NEXT: ]
; CHECK: [[BB4]]:
; CHECK-NEXT: [[PHI5:%.*]] = phi i32 [ [[PHI13:%.*]], %[[BB12]] ], [ [[PHI2]], %[[BB1]] ]
; CHECK-NEXT: br label %[[BB7]]
; CHECK: [[BB6]]:
; CHECK-NEXT: [[CALL:%.*]] = call i32 @foo(i32 noundef [[PHI2]])
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[PHI2]], 1
; CHECK-NEXT: br label %[[BB12]]
; CHECK: [[BB7]]:
; CHECK-NEXT: [[PHI8]] = phi i32 [ [[PHI]], %[[BB1]] ], [ 2, %[[BB4]] ]
; CHECK-NEXT: [[PHI9]] = phi i32 [ [[PHI2]], %[[BB1]] ], [ [[PHI5]], %[[BB4]] ]
; CHECK-NEXT: br label %[[BB1]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: [[BB10]]:
; CHECK-NEXT: [[CALL11:%.*]] = call i32 @foo(i32 noundef [[PHI13]])
; CHECK-NEXT: ret i32 0
; CHECK: [[BB12]]:
; CHECK-NEXT: [[PHI13]] = phi i32 [ [[ADD]], %[[BB6]] ], [ [[PHI2]], %[[BB1]] ]
; CHECK-NEXT: [[SEXT:%.*]] = sext i32 [[PHI13]] to i64
; CHECK-NEXT: [[GETELEMENTPTR14:%.*]] = getelementptr inbounds [2 x ptr], ptr [[ALLOCA]], i64 0, i64 [[SEXT]]
; CHECK-NEXT: [[LOAD:%.*]] = load ptr, ptr [[GETELEMENTPTR14]], align 8, !tbaa [[TBAA0]]
; CHECK-NEXT: indirectbr ptr [[LOAD]], [label %[[BB4]], label %bb10]
;
bb:
%alloca = alloca [2 x ptr], align 16
store ptr blockaddress(@main, %bb4), ptr %alloca, align 16, !tbaa !0
%getelementptr = getelementptr inbounds [2 x ptr], ptr %alloca, i64 0, i64 1
store ptr blockaddress(@main, %bb10), ptr %getelementptr, align 8, !tbaa !0
br label %bb1

bb1: ; preds = %bb7, %bb
%phi = phi i32 [ 0, %bb ], [ %phi8, %bb7 ]
%phi2 = phi i32 [ 0, %bb ], [ %phi9, %bb7 ]
switch i32 %phi, label %bb7 [
i32 0, label %bb3
i32 1, label %bb4
i32 2, label %bb6
]

bb3: ; preds = %bb1
br label %bb12

bb4: ; preds = %bb12, %bb1
%phi5 = phi i32 [ %phi13, %bb12 ], [ %phi2, %bb1 ]
br label %bb7

bb6: ; preds = %bb1
%call = call i32 @foo(i32 noundef %phi2)
%add = add nsw i32 %phi2, 1
br label %bb12

bb7: ; preds = %bb4, %bb1
%phi8 = phi i32 [ %phi, %bb1 ], [ 2, %bb4 ]
%phi9 = phi i32 [ %phi2, %bb1 ], [ %phi5, %bb4 ]
br label %bb1, !llvm.loop !4

bb10: ; preds = %bb12
%call11 = call i32 @foo(i32 noundef %phi13)
ret i32 0

bb12: ; preds = %bb6, %bb3
%phi13 = phi i32 [ %add, %bb6 ], [ %phi2, %bb3 ]
%sext = sext i32 %phi13 to i64
%getelementptr14 = getelementptr inbounds [2 x ptr], ptr %alloca, i64 0, i64 %sext
%load = load ptr, ptr %getelementptr14, align 8, !tbaa !0
indirectbr ptr %load, [label %bb4, label %bb10]
}

declare i32 @foo(i32)

!0 = !{!1, !1, i64 0}
!1 = !{!"any pointer", !2, i64 0}
!2 = !{!"omnipotent char", !3, i64 0}
!3 = !{!"Simple C++ TBAA"}
!4 = !{!5, !5, i64 0}
!5 = !{!"int", !2, i64 0}
;.
; CHECK: [[TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
; CHECK: [[META1]] = !{!"any pointer", [[META2:![0-9]+]], i64 0}
; CHECK: [[META2]] = !{!"omnipotent char", [[META3:![0-9]+]], i64 0}
; CHECK: [[META3]] = !{!"Simple C++ TBAA"}
; CHECK: [[LOOP4]] = !{[[META5:![0-9]+]], [[META5]], i64 0}
; CHECK: [[META5]] = !{!"int", [[META2]], i64 0}
;.
Loading