[llvm] [LV] Rewrite UDiv A, B -> UDiv A, UMax(B, 1) in trip counts if needed. (PR #92177)
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Fri Sep 6 13:15:56 PDT 2024
https://github.com/fhahn updated https://github.com/llvm/llvm-project/pull/92177
>From bb5f8e9fc1debebd998ecb43875c0c250379f028 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Tue, 14 May 2024 20:36:16 +0100
Subject: [PATCH 1/2] [LV] Don't vectorize if trip count expansion may
introduce UB.
Introduce a utility to check if a SCEV expansion may introduce UB
(couldn't find a similar utility after a quick glance) and use to the
avoid vectorizing when expanding the trip count introduces UB.
Fixes https://github.com/llvm/llvm-project/issues/89958.
---
.../Utils/ScalarEvolutionExpander.h | 7 +
.../Utils/ScalarEvolutionExpander.cpp | 48 +++-
.../Transforms/Vectorize/LoopVectorize.cpp | 3 +
.../trip-count-expansion-may-introduce-ub.ll | 255 +++++++++++-------
4 files changed, 220 insertions(+), 93 deletions(-)
diff --git a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
index 62c1e15a9a60e1..51a3153f5a7d82 100644
--- a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
+++ b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
@@ -418,6 +418,13 @@ class SCEVExpander : public SCEVVisitor<SCEVExpander, Value *> {
BasicBlock::iterator findInsertPointAfter(Instruction *I,
Instruction *MustDominate) const;
+ /// If \p L contains exits which may execute conditionally and contain UDiv
+ /// expressions with divisors that can be 0, expanding \p BTC may introduce
+ /// new UB. In that case, rewrite UDiv(A, B) to UDiv(A, UMAX(1, B)). If B is
+ /// 0, that exit cannot be taken.
+ static const SCEV *rewriteExpressionToRemoveUB(const SCEV *BTC, Loop *L,
+ ScalarEvolution &SE);
+
private:
LLVMContext &getContext() const { return SE.getContext(); }
diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index c7d758aa575e61..527dc8e448efe8 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -2341,12 +2341,17 @@ struct SCEVFindUnsafe {
};
} // namespace
-bool SCEVExpander::isSafeToExpand(const SCEV *S) const {
+static bool isSafeToExpand(const SCEV *S, bool CanonicalMode,
+ ScalarEvolution &SE) {
SCEVFindUnsafe Search(SE, CanonicalMode);
visitAll(S, Search);
return !Search.IsUnsafe;
}
+bool SCEVExpander::isSafeToExpand(const SCEV *S) const {
+ return ::isSafeToExpand(S, CanonicalMode, SE);
+}
+
bool SCEVExpander::isSafeToExpandAt(const SCEV *S,
const Instruction *InsertionPoint) const {
if (!isSafeToExpand(S))
@@ -2369,6 +2374,47 @@ bool SCEVExpander::isSafeToExpandAt(const SCEV *S,
return false;
}
+namespace {
+struct SCEVUDivRewriter : public SCEVRewriteVisitor<SCEVUDivRewriter> {
+ SCEVUDivRewriter(ScalarEvolution &SE) : SCEVRewriteVisitor(SE) {}
+
+ static const SCEV *rewrite(const SCEV *Scev, ScalarEvolution &SE) {
+ SCEVUDivRewriter Rewriter(SE);
+ return Rewriter.visit(Scev);
+ }
+
+ const SCEV *visitUDivExpr(const SCEVUDivExpr *Expr) {
+ auto *LHS = visit(Expr->getLHS());
+ auto *RHS = visit(Expr->getRHS());
+ return SE.getUDivExpr(LHS, SE.getUMaxExpr(SE.getOne(RHS->getType()), RHS));
+ }
+};
+} // namespace
+
+const SCEV *SCEVExpander::rewriteExpressionToRemoveUB(const SCEV *Expr, Loop *L,
+ ScalarEvolution &SE) {
+ SmallVector<BasicBlock *> Exiting;
+ L->getExitingBlocks(Exiting);
+ // Check if exit count for any exit that may execute unconditionally may in
+ // introduce UB. Note that we can skip checks in the header or if there's a
+ // single exit, as in those cases we know that the exit count will be
+ // evaluated in each loop iteration. There are other cases where the exiting
+ // block executes on each loop iteration, but we don't have a cheap way to
+ // check at the moment.
+
+ if (Exiting.size() == 1 || all_of(Exiting, [L, &SE](BasicBlock *E) {
+ if (L->getHeader() == E)
+ return true;
+ const SCEV *EC = SE.getExitCount(L, E);
+ if (isa<SCEVCouldNotCompute>(EC))
+ return true;
+ return ::isSafeToExpand(EC, true, SE);
+ }))
+ return Expr;
+
+ return SCEVUDivRewriter::rewrite(Expr, SE);
+}
+
void SCEVExpanderCleaner::cleanup() {
// Result is used, nothing to remove.
if (ResultUsed)
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index d059b576380127..1bbbab6a7778aa 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -916,6 +916,9 @@ const SCEV *createTripCountSCEV(Type *IdxTy, PredicatedScalarEvolution &PSE,
assert(!isa<SCEVCouldNotCompute>(BackedgeTakenCount) && "Invalid loop count");
ScalarEvolution &SE = *PSE.getSE();
+ if (OrigLoop)
+ BackedgeTakenCount = SCEVExpander::rewriteExpressionToRemoveUB(
+ BackedgeTakenCount, OrigLoop, SE);
return SE.getTripCountFromExitCount(BackedgeTakenCount, IdxTy, OrigLoop);
}
diff --git a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
index bf687a31d8d178..c82797c5d808f3 100644
--- a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
+++ b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
@@ -141,20 +141,21 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[N]]
+; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
; CHECK-NEXT: [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
; CHECK-NEXT: [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
-; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: vector.body:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
-; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[IV]], 0
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
@@ -169,7 +170,7 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i1> [[TMP7]], i32 1
; CHECK-NEXT: br i1 [[TMP10]], label [[PRED_STORE_IF1:%.*]], label [[PRED_STORE_CONTINUE2:%.*]]
; CHECK: pred.store.if1:
-; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[IV]], 1
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP11]]
; CHECK-NEXT: store i32 1, ptr [[TMP12]], align 4
; CHECK-NEXT: br label [[PRED_STORE_CONTINUE2]]
@@ -177,30 +178,30 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i1> [[TMP7]], i32 2
; CHECK-NEXT: br i1 [[TMP13]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4:%.*]]
; CHECK: pred.store.if3:
-; CHECK-NEXT: [[TMP14:%.*]] = add i64 [[INDEX]], 2
+; CHECK-NEXT: [[TMP14:%.*]] = add i64 [[IV]], 2
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]
; CHECK-NEXT: store i32 1, ptr [[TMP15]], align 4
; CHECK-NEXT: br label [[PRED_STORE_CONTINUE4]]
; CHECK: pred.store.continue4:
; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i1> [[TMP7]], i32 3
-; CHECK-NEXT: br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT: br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[LOOP_LATCH]]
; CHECK: pred.store.if5:
-; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 3
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[IV]], 3
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
; CHECK-NEXT: store i32 1, ptr [[TMP18]], align 4
-; CHECK-NEXT: br label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT: br label [[LOOP_LATCH]]
; CHECK: pred.store.continue6:
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 4
+; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT: br label [[LOOP_HEADER1:%.*]]
; CHECK: loop.header:
-; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP_LATCH1:%.*]] ]
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT: [[C_2:%.*]] = icmp eq i32 [[L]], 10
; CHECK-NEXT: br i1 [[C_2]], label [[THEN:%.*]], label [[CONTINUE:%.*]]
@@ -209,14 +210,14 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
; CHECK-NEXT: br label [[CONTINUE]]
; CHECK: continue:
; CHECK-NEXT: [[D:%.*]] = udiv i64 42, [[N]]
-; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_LATCH1]], label [[EXIT:%.*]]
; CHECK: loop.latch:
-; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
-; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
-; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT: [[IV_NEXT1]] = add i64 [[IV1]], 1
+; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
+; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK: exit:
-; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH1]] ]
; CHECK-NEXT: ret i64 [[P]]
;
entry:
@@ -257,16 +258,16 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
; CHECK-NEXT: [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
; CHECK-NEXT: [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
-; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: vector.body:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
-; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[IV]], 0
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
@@ -281,7 +282,7 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i1> [[TMP7]], i32 1
; CHECK-NEXT: br i1 [[TMP10]], label [[PRED_STORE_IF1:%.*]], label [[PRED_STORE_CONTINUE2:%.*]]
; CHECK: pred.store.if1:
-; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[IV]], 1
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP11]]
; CHECK-NEXT: store i32 1, ptr [[TMP12]], align 4
; CHECK-NEXT: br label [[PRED_STORE_CONTINUE2]]
@@ -289,30 +290,30 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i1> [[TMP7]], i32 2
; CHECK-NEXT: br i1 [[TMP13]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4:%.*]]
; CHECK: pred.store.if3:
-; CHECK-NEXT: [[TMP14:%.*]] = add i64 [[INDEX]], 2
+; CHECK-NEXT: [[TMP14:%.*]] = add i64 [[IV]], 2
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]
; CHECK-NEXT: store i32 1, ptr [[TMP15]], align 4
; CHECK-NEXT: br label [[PRED_STORE_CONTINUE4]]
; CHECK: pred.store.continue4:
; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i1> [[TMP7]], i32 3
-; CHECK-NEXT: br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT: br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[LOOP_LATCH]]
; CHECK: pred.store.if5:
-; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 3
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[IV]], 3
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
; CHECK-NEXT: store i32 1, ptr [[TMP18]], align 4
-; CHECK-NEXT: br label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT: br label [[LOOP_LATCH]]
; CHECK: pred.store.continue6:
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 4
+; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT: br label [[LOOP_HEADER1:%.*]]
; CHECK: loop.header:
-; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP_LATCH1:%.*]] ]
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT: [[C_2:%.*]] = icmp eq i32 [[L]], 10
; CHECK-NEXT: br i1 [[C_2]], label [[THEN:%.*]], label [[CONTINUE:%.*]]
@@ -321,14 +322,14 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
; CHECK-NEXT: br label [[CONTINUE]]
; CHECK: continue:
; CHECK-NEXT: [[D:%.*]] = udiv i64 [[N]], 42
-; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_LATCH1]], label [[EXIT:%.*]]
; CHECK: loop.latch:
-; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
-; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
-; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT: [[IV_NEXT1]] = add i64 [[IV1]], 1
+; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
+; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: exit:
-; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH1]] ]
; CHECK-NEXT: ret i64 [[P]]
;
entry:
@@ -459,14 +460,13 @@ exit:
ret i64 %p
}
-; FIXME: Currently miscompiled as we unconditionally execute udiv after
-; vectorization.
define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch(ptr %dst, i64 %N) {
; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch(
; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[N]]
+; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
; CHECK-NEXT: [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
@@ -542,34 +542,34 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(ptr %dst, i64
; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
; CHECK-NEXT: [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
-; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: vector.body:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP4]]
-; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_HEADER]] ]
+; CHECK-NEXT: [[IV:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT: br label [[LOOP_HEADER1:%.*]]
; CHECK: loop.header:
-; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
-; CHECK-NEXT: store i32 1, ptr [[GEP]], align 4
-; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV1]]
+; CHECK-NEXT: store i32 1, ptr [[GEP1]], align 4
+; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
; CHECK: loop.latch:
-; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV1]], 1
; CHECK-NEXT: [[D:%.*]] = udiv i64 [[N]], 42
-; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK: exit:
-; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER1]] ], [ 0, [[LOOP_LATCH]] ]
; CHECK-NEXT: ret i64 [[P]]
;
entry:
@@ -647,15 +647,14 @@ exit:
ret void
}
-; FIXME: Currently miscompiled as we unconditionally execute udiv after
-; vectorization.
define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(ptr %dst, i64 %N) {
; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(
; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[N]]
-; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[N]], [[TMP0]]
+; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
+; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[N]], [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = sub i64 42, [[TMP1]]
; CHECK-NEXT: [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP2]], i64 0)
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[SMAX1]])
@@ -723,45 +722,45 @@ define i64 @multi_exit_4_exit_count_with_urem_by_constant_in_latch(ptr %dst, i64
; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT: [[TMP2:%.*]] = urem i64 [[N]], 42
-; CHECK-NEXT: [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP2]], i64 0)
+; CHECK-NEXT: [[TMP0:%.*]] = urem i64 [[N]], 42
+; CHECK-NEXT: [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP0]], i64 0)
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[SMAX1]])
-; CHECK-NEXT: [[TMP3:%.*]] = add nuw i64 [[UMIN]], 1
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP3]], 4
+; CHECK-NEXT: [[TMP1:%.*]] = add nuw i64 [[UMIN]], 1
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4
-; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
-; CHECK-NEXT: [[TMP5:%.*]] = select i1 [[TMP4]], i64 4, i64 [[N_MOD_VF]]
-; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP3]], [[TMP5]]
-; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT: [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
+; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: vector.body:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP6]]
-; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP8]], align 4
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_HEADER]] ]
+; CHECK-NEXT: [[IV:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
+; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP18:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT: br label [[LOOP_HEADER1:%.*]]
; CHECK: loop.header:
-; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
-; CHECK-NEXT: store i32 1, ptr [[GEP]], align 4
-; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV1]]
+; CHECK-NEXT: store i32 1, ptr [[GEP1]], align 4
+; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
; CHECK: loop.latch:
-; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV1]], 1
; CHECK-NEXT: [[D:%.*]] = urem i64 [[N]], 42
-; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP19:![0-9]+]]
; CHECK: exit:
-; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER1]] ], [ 0, [[LOOP_LATCH]] ]
; CHECK-NEXT: ret i64 [[P]]
;
entry:
@@ -867,6 +866,76 @@ exit:
ret i64 %p
}
+define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(ptr %dst, i64 %N) {
+; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(
+; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
+; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
+; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
+; CHECK-NEXT: [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK: vector.ph:
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT: [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
+; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK: vector.body:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP4]]
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
+; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK: middle.block:
+; CHECK-NEXT: br label [[SCALAR_PH]]
+; CHECK: scalar.ph:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
+; CHECK: loop.header:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT: store i32 1, ptr [[GEP]], align 4
+; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK: loop.latch:
+; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT: [[D:%.*]] = udiv i64 42, [[N]]
+; CHECK-NEXT: [[X:%.*]] = sub i64 100, [[D]]
+; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK: exit:
+; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT: ret i64 [[P]]
+;
+entry:
+ br label %loop.header
+
+loop.header:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
+ %gep = getelementptr inbounds i32, ptr %dst, i64 %iv
+ store i32 1, ptr %gep
+ %c.0 = icmp slt i64 %iv, %N
+ br i1 %c.0, label %loop.latch, label %exit
+
+loop.latch:
+ %iv.next = add i64 %iv, 1
+ %d = udiv i64 42, %N
+ %x = sub i64 100, %d
+ %c.1 = icmp slt i64 %iv, %d
+ br i1 %c.1, label %loop.header, label %exit
+
+exit:
+ %p = phi i64 [ 1, %loop.header ], [ 0, %loop.latch]
+ ret i64 %p
+}
+
+
;.
; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
@@ -888,4 +957,6 @@ exit:
; CHECK: [[LOOP17]] = distinct !{[[LOOP17]], [[META2]], [[META1]]}
; CHECK: [[LOOP18]] = distinct !{[[LOOP18]], [[META1]], [[META2]]}
; CHECK: [[LOOP19]] = distinct !{[[LOOP19]], [[META2]], [[META1]]}
+; CHECK: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]}
+; CHECK: [[LOOP21]] = distinct !{[[LOOP21]], [[META2]], [[META1]]}
;.
>From 6ce1b5fac1d8a04956e2e9808ac2224dfc55218a Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Fri, 6 Sep 2024 21:13:59 +0100
Subject: [PATCH 2/2] !fixup introduce SafeUDivMode to SCEVExpander.
---
.../Utils/ScalarEvolutionExpander.h | 14 ++-
.../Utils/ScalarEvolutionExpander.cpp | 53 ++--------
.../Transforms/Vectorize/LoopVectorize.cpp | 3 -
llvm/lib/Transforms/Vectorize/VPlan.cpp | 27 +++++-
llvm/lib/Transforms/Vectorize/VPlan.h | 10 +-
.../lib/Transforms/Vectorize/VPlanRecipes.cpp | 2 +
llvm/lib/Transforms/Vectorize/VPlanUtils.cpp | 5 +-
llvm/lib/Transforms/Vectorize/VPlanUtils.h | 3 +-
.../trip-count-expansion-may-introduce-ub.ll | 97 ++++++++++++++++---
9 files changed, 144 insertions(+), 70 deletions(-)
diff --git a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
index 51a3153f5a7d82..cb3b7372f7e66e 100644
--- a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
+++ b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
@@ -124,6 +124,11 @@ class SCEVExpander : public SCEVVisitor<SCEVExpander, Value *> {
/// "expanded" form.
bool LSRMode;
+ /// When true, rewrite any divisors of UDiv expressions that may be 0 to
+ /// umax(Divisor, 1) to avoid introducing UB. If the divisor may be poison,
+ /// freeze it first.
+ bool SafeUDivMode = false;
+
typedef IRBuilder<InstSimplifyFolder, IRBuilderCallbackInserter> BuilderType;
BuilderType Builder;
@@ -300,6 +305,9 @@ class SCEVExpander : public SCEVVisitor<SCEVExpander, Value *> {
/// location and their operands are defined at this location.
bool isSafeToExpandAt(const SCEV *S, const Instruction *InsertionPoint) const;
+ static bool isSafeToExpand(const SCEV *S, bool CanonicalMode,
+ ScalarEvolution &SE);
+
/// Insert code to directly compute the specified SCEV expression into the
/// program. The code is inserted into the specified block.
Value *expandCodeFor(const SCEV *SH, Type *Ty, BasicBlock::iterator I);
@@ -418,13 +426,11 @@ class SCEVExpander : public SCEVVisitor<SCEVExpander, Value *> {
BasicBlock::iterator findInsertPointAfter(Instruction *I,
Instruction *MustDominate) const;
- /// If \p L contains exits which may execute conditionally and contain UDiv
- /// expressions with divisors that can be 0, expanding \p BTC may introduce
- /// new UB. In that case, rewrite UDiv(A, B) to UDiv(A, UMAX(1, B)). If B is
- /// 0, that exit cannot be taken.
static const SCEV *rewriteExpressionToRemoveUB(const SCEV *BTC, Loop *L,
ScalarEvolution &SE);
+ void setSafeUDivMode() { SafeUDivMode = true; }
+
private:
LLVMContext &getContext() const { return SE.getContext(); }
diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index 527dc8e448efe8..28ba21759a8567 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -665,6 +665,12 @@ Value *SCEVExpander::visitUDivExpr(const SCEVUDivExpr *S) {
}
Value *RHS = expand(S->getRHS());
+ if (SafeUDivMode && !SE.isKnownNonZero(S->getRHS())) {
+ if (!isGuaranteedNotToBePoison(RHS))
+ RHS = Builder.CreateFreeze(RHS);
+ RHS = Builder.CreateIntrinsic(RHS->getType(), Intrinsic::umax,
+ {RHS, ConstantInt::get(RHS->getType(), 1)});
+ }
return InsertBinop(Instruction::UDiv, LHS, RHS, SCEV::FlagAnyWrap,
/*IsSafeToHoist*/ SE.isKnownNonZero(S->getRHS()));
}
@@ -2341,15 +2347,15 @@ struct SCEVFindUnsafe {
};
} // namespace
-static bool isSafeToExpand(const SCEV *S, bool CanonicalMode,
- ScalarEvolution &SE) {
+bool SCEVExpander::isSafeToExpand(const SCEV *S, bool CanonicalMode,
+ ScalarEvolution &SE) {
SCEVFindUnsafe Search(SE, CanonicalMode);
visitAll(S, Search);
return !Search.IsUnsafe;
}
bool SCEVExpander::isSafeToExpand(const SCEV *S) const {
- return ::isSafeToExpand(S, CanonicalMode, SE);
+ return isSafeToExpand(S, CanonicalMode, SE);
}
bool SCEVExpander::isSafeToExpandAt(const SCEV *S,
@@ -2374,47 +2380,6 @@ bool SCEVExpander::isSafeToExpandAt(const SCEV *S,
return false;
}
-namespace {
-struct SCEVUDivRewriter : public SCEVRewriteVisitor<SCEVUDivRewriter> {
- SCEVUDivRewriter(ScalarEvolution &SE) : SCEVRewriteVisitor(SE) {}
-
- static const SCEV *rewrite(const SCEV *Scev, ScalarEvolution &SE) {
- SCEVUDivRewriter Rewriter(SE);
- return Rewriter.visit(Scev);
- }
-
- const SCEV *visitUDivExpr(const SCEVUDivExpr *Expr) {
- auto *LHS = visit(Expr->getLHS());
- auto *RHS = visit(Expr->getRHS());
- return SE.getUDivExpr(LHS, SE.getUMaxExpr(SE.getOne(RHS->getType()), RHS));
- }
-};
-} // namespace
-
-const SCEV *SCEVExpander::rewriteExpressionToRemoveUB(const SCEV *Expr, Loop *L,
- ScalarEvolution &SE) {
- SmallVector<BasicBlock *> Exiting;
- L->getExitingBlocks(Exiting);
- // Check if exit count for any exit that may execute unconditionally may in
- // introduce UB. Note that we can skip checks in the header or if there's a
- // single exit, as in those cases we know that the exit count will be
- // evaluated in each loop iteration. There are other cases where the exiting
- // block executes on each loop iteration, but we don't have a cheap way to
- // check at the moment.
-
- if (Exiting.size() == 1 || all_of(Exiting, [L, &SE](BasicBlock *E) {
- if (L->getHeader() == E)
- return true;
- const SCEV *EC = SE.getExitCount(L, E);
- if (isa<SCEVCouldNotCompute>(EC))
- return true;
- return ::isSafeToExpand(EC, true, SE);
- }))
- return Expr;
-
- return SCEVUDivRewriter::rewrite(Expr, SE);
-}
-
void SCEVExpanderCleaner::cleanup() {
// Result is used, nothing to remove.
if (ResultUsed)
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 1bbbab6a7778aa..d059b576380127 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -916,9 +916,6 @@ const SCEV *createTripCountSCEV(Type *IdxTy, PredicatedScalarEvolution &PSE,
assert(!isa<SCEVCouldNotCompute>(BackedgeTakenCount) && "Invalid loop count");
ScalarEvolution &SE = *PSE.getSE();
- if (OrigLoop)
- BackedgeTakenCount = SCEVExpander::rewriteExpressionToRemoveUB(
- BackedgeTakenCount, OrigLoop, SE);
return SE.getTripCountFromExitCount(BackedgeTakenCount, IdxTy, OrigLoop);
}
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp
index ac322638056879..d318639f928d4b 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp
@@ -866,8 +866,31 @@ VPlanPtr VPlan::createInitialVPlan(const SCEV *TripCount, ScalarEvolution &SE,
VPIRBasicBlock *Entry = new VPIRBasicBlock(TheLoop->getLoopPreheader());
VPBasicBlock *VecPreheader = new VPBasicBlock("vector.ph");
auto Plan = std::make_unique<VPlan>(Entry, VecPreheader);
- Plan->TripCount =
- vputils::getOrCreateVPValueForSCEVExpr(*Plan, TripCount, SE);
+
+ bool NeedsSafeUDivMode = false;
+ {
+ SmallVector<BasicBlock *> Exiting;
+ TheLoop->getExitingBlocks(Exiting);
+
+ // Check if exit count for any exit that may execute unconditionally may in
+ // introduce UB. Note that we can skip checks in the header or if there's a
+ // single exit, as in those cases we know that the exit count will be
+ // evaluated in each loop iteration. There are other cases where the exiting
+ // block executes on each loop iteration, but we don't have a cheap way to
+ // check at the moment.
+ NeedsSafeUDivMode =
+ Exiting.size() != 1 && any_of(Exiting, [TheLoop, &SE](BasicBlock *E) {
+ if (TheLoop->getHeader() == E)
+ return false;
+ const SCEV *EC = SE.getExitCount(TheLoop, E);
+ if (isa<SCEVCouldNotCompute>(EC))
+ return false;
+ return !SCEVExpander::isSafeToExpand(EC, true, SE);
+ });
+ }
+
+ Plan->TripCount = vputils::getOrCreateVPValueForSCEVExpr(*Plan, TripCount, SE,
+ NeedsSafeUDivMode);
// Create VPRegionBlock, with empty header and latch blocks, to be filled
// during processing later.
VPBasicBlock *HeaderVPBB = new VPBasicBlock("vector.body");
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index cec3135395483b..64aa1a02c0c3ce 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -2771,9 +2771,15 @@ class VPExpandSCEVRecipe : public VPSingleDefRecipe {
const SCEV *Expr;
ScalarEvolution &SE;
+ /// When set to true, set SafeUDivMode when expanding the SCEV to avoid
+ /// introducing UB.
+ bool SafeUDivMode;
+
public:
- VPExpandSCEVRecipe(const SCEV *Expr, ScalarEvolution &SE)
- : VPSingleDefRecipe(VPDef::VPExpandSCEVSC, {}), Expr(Expr), SE(SE) {}
+ VPExpandSCEVRecipe(const SCEV *Expr, ScalarEvolution &SE,
+ bool SafeUDivMode = false)
+ : VPSingleDefRecipe(VPDef::VPExpandSCEVSC, {}), Expr(Expr), SE(SE),
+ SafeUDivMode(SafeUDivMode) {}
~VPExpandSCEVRecipe() override = default;
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index e94e7561f3c2da..2ec1163d901177 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -2932,6 +2932,8 @@ void VPExpandSCEVRecipe::execute(VPTransformState &State) {
assert(!State.Instance && "cannot be used in per-lane");
const DataLayout &DL = State.CFG.PrevBB->getDataLayout();
SCEVExpander Exp(SE, DL, "induction");
+ if (SafeUDivMode)
+ Exp.setSafeUDivMode();
Value *Res = Exp.expandCodeFor(Expr, Expr->getType(),
&*State.Builder.GetInsertPoint());
diff --git a/llvm/lib/Transforms/Vectorize/VPlanUtils.cpp b/llvm/lib/Transforms/Vectorize/VPlanUtils.cpp
index c18bea4f4c5926..fa666b28a2525b 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanUtils.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanUtils.cpp
@@ -23,7 +23,8 @@ bool vputils::onlyFirstPartUsed(const VPValue *Def) {
}
VPValue *vputils::getOrCreateVPValueForSCEVExpr(VPlan &Plan, const SCEV *Expr,
- ScalarEvolution &SE) {
+ ScalarEvolution &SE,
+ bool SafeUDivMode) {
if (auto *Expanded = Plan.getSCEVExpansion(Expr))
return Expanded;
VPValue *Expanded = nullptr;
@@ -32,7 +33,7 @@ VPValue *vputils::getOrCreateVPValueForSCEVExpr(VPlan &Plan, const SCEV *Expr,
else if (auto *E = dyn_cast<SCEVUnknown>(Expr))
Expanded = Plan.getOrAddLiveIn(E->getValue());
else {
- Expanded = new VPExpandSCEVRecipe(Expr, SE);
+ Expanded = new VPExpandSCEVRecipe(Expr, SE, SafeUDivMode);
Plan.getPreheader()->appendRecipe(Expanded->getDefiningRecipe());
}
Plan.addSCEVExpansion(Expr, Expanded);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanUtils.h b/llvm/lib/Transforms/Vectorize/VPlanUtils.h
index fc11208a433961..b23f5a66a710c9 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanUtils.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanUtils.h
@@ -24,7 +24,8 @@ bool onlyFirstPartUsed(const VPValue *Def);
/// pre-header already contains a recipe expanding \p Expr, return it. If not,
/// create a new one.
VPValue *getOrCreateVPValueForSCEVExpr(VPlan &Plan, const SCEV *Expr,
- ScalarEvolution &SE);
+ ScalarEvolution &SE,
+ bool SaveUDivMode = false);
/// Returns true if \p VPV is uniform after vectorization.
inline bool isUniformAfterVectorization(const VPValue *VPV) {
diff --git a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
index c82797c5d808f3..44d1282b8f0ed8 100644
--- a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
+++ b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
@@ -141,7 +141,8 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT: [[TMP20:%.*]] = freeze i64 [[N]]
+; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[TMP20]], i64 1)
; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
; CHECK-NEXT: [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
@@ -527,6 +528,76 @@ exit:
ret i64 %p
}
+define i64 @multi_exit_4_exit_count_with_udiv_by_frozen_value_in_latch(ptr %dst, i64 %N) {
+; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_udiv_by_frozen_value_in_latch(
+; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[FR_N:%.*]] = freeze i64 [[N]]
+; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
+; CHECK-NEXT: [[TMP1:%.*]] = call i64 @llvm.umax.i64(i64 [[FR_N]], i64 1)
+; CHECK-NEXT: [[TMP2:%.*]] = udiv i64 42, [[TMP1]]
+; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP2]])
+; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[UMIN]], 1
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP3]], 4
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK: vector.ph:
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4
+; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = select i1 [[TMP4]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP3]], [[TMP5]]
+; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK: vector.body:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP6]]
+; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
+; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP8]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK: middle.block:
+; CHECK-NEXT: br label [[SCALAR_PH]]
+; CHECK: scalar.ph:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
+; CHECK: loop.header:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT: store i32 1, ptr [[GEP]], align 4
+; CHECK-NEXT: [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK: loop.latch:
+; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT: [[D:%.*]] = udiv i64 42, [[FR_N]]
+; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK: exit:
+; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT: ret i64 [[P]]
+;
+entry:
+ %fr.N = freeze i64 %N
+ br label %loop.header
+
+loop.header:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
+ %gep = getelementptr inbounds i32, ptr %dst, i64 %iv
+ store i32 1, ptr %gep
+ %c.0 = icmp slt i64 %iv, %N
+ br i1 %c.0, label %loop.latch, label %exit
+
+loop.latch:
+ %iv.next = add i64 %iv, 1
+ %d = udiv i64 42, %fr.N
+ %c.1 = icmp slt i64 %iv, %d
+ br i1 %c.1, label %loop.header, label %exit
+
+exit:
+ %p = phi i64 [ 1, %loop.header ], [ 0, %loop.latch]
+ ret i64 %p
+}
+
+
define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(ptr %dst, i64 %N) {
; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(
; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
@@ -551,7 +622,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(ptr %dst, i64
; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP14:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
@@ -567,7 +638,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(ptr %dst, i64
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV1]], 1
; CHECK-NEXT: [[D:%.*]] = udiv i64 [[N]], 42
; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER1]] ], [ 0, [[LOOP_LATCH]] ]
; CHECK-NEXT: ret i64 [[P]]
@@ -613,7 +684,7 @@ define void @single_exit_tc_with_udiv(ptr %dst, i64 %N) {
; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP4]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP1]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
@@ -627,7 +698,7 @@ define void @single_exit_tc_with_udiv(ptr %dst, i64 %N) {
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
; CHECK-NEXT: [[D:%.*]] = udiv i64 42, [[N]]
; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP15:![0-9]+]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP17:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
@@ -654,7 +725,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(ptr %dst, i64 %N
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
; CHECK-NEXT: [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
-; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[N]], [[TMP0]]
+; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[N]], [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = sub i64 42, [[TMP1]]
; CHECK-NEXT: [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP2]], i64 0)
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[SMAX1]])
@@ -675,7 +746,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(ptr %dst, i64 %N
; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP8]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
@@ -691,7 +762,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(ptr %dst, i64 %N
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
; CHECK-NEXT: [[D:%.*]] = urem i64 42, [[N]]
; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP17:![0-9]+]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP19:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
; CHECK-NEXT: ret i64 [[P]]
@@ -742,7 +813,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_constant_in_latch(ptr %dst, i64
; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP20:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
@@ -758,7 +829,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_constant_in_latch(ptr %dst, i64
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV1]], 1
; CHECK-NEXT: [[D:%.*]] = urem i64 [[N]], 42
; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP21:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER1]] ], [ 0, [[LOOP_LATCH]] ]
; CHECK-NEXT: ret i64 [[P]]
@@ -891,7 +962,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(ptr %dst, i64 %
; CHECK-NEXT: store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br label [[SCALAR_PH]]
; CHECK: scalar.ph:
@@ -908,7 +979,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(ptr %dst, i64 %
; CHECK-NEXT: [[D:%.*]] = udiv i64 42, [[N]]
; CHECK-NEXT: [[X:%.*]] = sub i64 100, [[D]]
; CHECK-NEXT: [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP23:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
; CHECK-NEXT: ret i64 [[P]]
@@ -959,4 +1030,6 @@ exit:
; CHECK: [[LOOP19]] = distinct !{[[LOOP19]], [[META2]], [[META1]]}
; CHECK: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]}
; CHECK: [[LOOP21]] = distinct !{[[LOOP21]], [[META2]], [[META1]]}
+; CHECK: [[LOOP22]] = distinct !{[[LOOP22]], [[META1]], [[META2]]}
+; CHECK: [[LOOP23]] = distinct !{[[LOOP23]], [[META2]], [[META1]]}
;.
More information about the llvm-commits
mailing list