[llvm] [LV][VPlan] Implement VPlan-based cost for exit condition. (PR #125640)
Elvis Wang via llvm-commits
llvm-commits at lists.llvm.org
Wed Aug 6 17:52:45 PDT 2025
https://github.com/ElvisWang123 updated https://github.com/llvm/llvm-project/pull/125640
>From 35def52194c409d8f4d43a601c6da9696a82c3af Mon Sep 17 00:00:00 2001
From: Elvis Wang <elvis.wang at sifive.com>
Date: Wed, 6 Aug 2025 17:35:53 -0700
Subject: [PATCH] [LV][VPlan] Implement VPlan-based cost for exit condition.
This patch tried to model the cost of exit conditions through
vplan-based cost model.
* `BranchOnCount` will generate icmp + br.
The branch instruction is already implemented by the VPRegionBlock so
we only need to calculate the cost of icmp.
If the VF is same as the trip count of the loop, the cost of the
BranchOnCount is free.
This patch is not quite NFC for following reasons.
* Some of the BranchOnCount could be optimized to BranchOnCond, which is
free.
* Some of the instructions calculated in the exit condition in legacy
cost model will be optimized out.
---
.../Transforms/Vectorize/LoopVectorize.cpp | 60 ++----
.../lib/Transforms/Vectorize/VPlanRecipes.cpp | 31 +++
.../AArch64/conditional-branches-cost.ll | 56 ++++--
.../LoopVectorize/AArch64/early_exit_costs.ll | 8 +-
.../AArch64/fully-unrolled-cost.ll | 14 +-
.../AArch64/induction-costs-sve.ll | 176 +++++++-----------
.../AArch64/reduction-recurrence-costs-sve.ll | 32 ++--
.../LoopVectorize/ARM/mve-icmpcost.ll | 21 +--
.../X86/CostModel/vpinstruction-cost.ll | 7 +-
.../LoopVectorize/X86/cost-model.ll | 70 +++----
.../LoopVectorize/X86/multi-exit-cost.ll | 21 ++-
.../LoopVectorize/X86/reduction-small-size.ll | 4 +-
12 files changed, 249 insertions(+), 251 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index a52aa8420b301..bf78704ce65a4 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -6333,6 +6333,7 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,
case Instruction::ICmp:
case Instruction::FCmp: {
Type *ValTy = I->getOperand(0)->getType();
+ InstructionCost Cost = 0;
if (canTruncateToMinimalBitwidth(I, VF)) {
[[maybe_unused]] Instruction *Op0AsInstruction =
@@ -6344,11 +6345,22 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,
ValTy = IntegerType::get(ValTy->getContext(), MinBWs[I]);
}
+ // If the Cmp instruction has multiple uses in the loop, it
+ // will generate a scalar Cmp for latch and a vector Cmp for other uses.
+ if (I == TheLoop->getLatchCmpInst() && !I->hasOneUse())
+ Cost += TTI.getCmpSelInstrCost(I->getOpcode(), ValTy,
+ CmpInst::makeCmpResultType(ValTy),
+ cast<CmpInst>(I)->getPredicate(), CostKind,
+ {TTI::OK_AnyValue, TTI::OP_None},
+ {TTI::OK_AnyValue, TTI::OP_None}, I);
+
VectorTy = toVectorTy(ValTy, VF);
- return TTI.getCmpSelInstrCost(
- I->getOpcode(), VectorTy, CmpInst::makeCmpResultType(VectorTy),
- cast<CmpInst>(I)->getPredicate(), CostKind,
- {TTI::OK_AnyValue, TTI::OP_None}, {TTI::OK_AnyValue, TTI::OP_None}, I);
+ return Cost + TTI.getCmpSelInstrCost(I->getOpcode(), VectorTy,
+ CmpInst::makeCmpResultType(VectorTy),
+ cast<CmpInst>(I)->getPredicate(),
+ CostKind,
+ {TTI::OK_AnyValue, TTI::OP_None},
+ {TTI::OK_AnyValue, TTI::OP_None}, I);
}
case Instruction::Store:
case Instruction::Load: {
@@ -6891,46 +6903,6 @@ LoopVectorizationPlanner::precomputeCosts(VPlan &Plan, ElementCount VF,
}
}
- /// Compute the cost of all exiting conditions of the loop using the legacy
- /// cost model. This is to match the legacy behavior, which adds the cost of
- /// all exit conditions. Note that this over-estimates the cost, as there will
- /// be a single condition to control the vector loop.
- SmallVector<BasicBlock *> Exiting;
- CM.TheLoop->getExitingBlocks(Exiting);
- SetVector<Instruction *> ExitInstrs;
- // Collect all exit conditions.
- for (BasicBlock *EB : Exiting) {
- auto *Term = dyn_cast<BranchInst>(EB->getTerminator());
- if (!Term || CostCtx.skipCostComputation(Term, VF.isVector()))
- continue;
- if (auto *CondI = dyn_cast<Instruction>(Term->getOperand(0))) {
- ExitInstrs.insert(CondI);
- }
- }
- // Compute the cost of all instructions only feeding the exit conditions.
- for (unsigned I = 0; I != ExitInstrs.size(); ++I) {
- Instruction *CondI = ExitInstrs[I];
- if (!OrigLoop->contains(CondI) ||
- !CostCtx.SkipCostComputation.insert(CondI).second)
- continue;
- InstructionCost CondICost = CostCtx.getLegacyCost(CondI, VF);
- LLVM_DEBUG({
- dbgs() << "Cost of " << CondICost << " for VF " << VF
- << ": exit condition instruction " << *CondI << "\n";
- });
- Cost += CondICost;
- for (Value *Op : CondI->operands()) {
- auto *OpI = dyn_cast<Instruction>(Op);
- if (!OpI || CostCtx.skipCostComputation(OpI, VF.isVector()) ||
- any_of(OpI->users(), [&ExitInstrs, this](User *U) {
- return OrigLoop->contains(cast<Instruction>(U)->getParent()) &&
- !ExitInstrs.contains(cast<Instruction>(U));
- }))
- continue;
- ExitInstrs.insert(OpI);
- }
- }
-
// Pre-compute the costs for branches except for the backedge, as the number
// of replicate regions in a VPlan may not directly match the number of
// branches, which would lead to different decisions.
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index e971ba1aac15c..43db0ce2cca8a 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -964,6 +964,37 @@ InstructionCost VPInstruction::computeCost(ElementCount VF,
return Ctx.TTI.getArithmeticReductionCost(
Instruction::Or, cast<VectorType>(VecTy), std::nullopt, Ctx.CostKind);
}
+ case VPInstruction::BranchOnCount: {
+ Type *ValTy = Ctx.Types.inferScalarType(getOperand(0));
+
+ // If the vector loop only executed once (VF == original trip count), ignore
+ // the cost of cmp.
+ // TODO: We can remove this after hoist `unrollByUF` and
+ // `optimizeForVFandUF` which will optimize BranchOnCount out.
+ auto TC = dyn_cast_if_present<ConstantInt>(
+ getParent()->getPlan()->getTripCount()->getUnderlyingValue());
+ if (TC && VF.isFixed() && TC->getSExtValue() == VF.getFixedValue())
+ return 0;
+
+ // BranchOnCount will generate icmp_eq + br instructions and the
+ // cost of branch will be calculated in VPRegionBlock.
+ return Ctx.TTI.getCmpSelInstrCost(Instruction::ICmp, ValTy, nullptr,
+ CmpInst::ICMP_EQ, Ctx.CostKind);
+ }
+ case VPInstruction::BranchOnCond: {
+ // BranchOnCond is free since the branch cost is already
+ // calculated by VPBB.
+ if (vputils::onlyFirstLaneUsed(getOperand(0)))
+ return 0;
+
+ // Otherwise, BranchOnCond will generate `extractelement` to extract the
+ // condition from vector type.
+ return Ctx.TTI.getVectorInstrCost(
+ Instruction::ExtractElement,
+ cast<VectorType>(
+ toVectorTy(Ctx.Types.inferScalarType(getOperand(0)), VF)),
+ Ctx.CostKind, 0, nullptr, nullptr);
+ }
case VPInstruction::FirstActiveLane: {
// Calculate the cost of determining the lane index.
auto *PredTy = toVectorTy(Ctx.Types.inferScalarType(getOperand(0)), VF);
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll
index 0232d88347d0a..36ad157c3e584 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll
@@ -657,28 +657,58 @@ define void @multiple_exit_conditions(ptr %src, ptr noalias %dst) #1 {
; DEFAULT-LABEL: define void @multiple_exit_conditions(
; DEFAULT-SAME: ptr [[SRC:%.*]], ptr noalias [[DST:%.*]]) #[[ATTR2:[0-9]+]] {
; DEFAULT-NEXT: [[ENTRY:.*]]:
-; DEFAULT-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; DEFAULT-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; DEFAULT-NEXT: [[TMP9:%.*]] = mul nuw i64 [[TMP7]], 8
+; DEFAULT-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 257, [[TMP9]]
+; DEFAULT-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; DEFAULT: [[VECTOR_PH]]:
-; DEFAULT-NEXT: [[IND_END:%.*]] = getelementptr i8, ptr [[DST]], i64 2048
+; DEFAULT-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; DEFAULT-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; DEFAULT-NEXT: [[N_MOD_VF:%.*]] = urem i64 257, [[TMP3]]
+; DEFAULT-NEXT: [[N_VEC:%.*]] = sub i64 257, [[N_MOD_VF]]
+; DEFAULT-NEXT: [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; DEFAULT-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP10]], 8
+; DEFAULT-NEXT: [[TMP6:%.*]] = mul i64 [[N_VEC]], 8
+; DEFAULT-NEXT: [[IND_END:%.*]] = getelementptr i8, ptr [[DST]], i64 [[TMP6]]
+; DEFAULT-NEXT: [[TMP8:%.*]] = mul i64 [[N_VEC]], 2
; DEFAULT-NEXT: br label %[[VECTOR_BODY:.*]]
; DEFAULT: [[VECTOR_BODY]]:
; DEFAULT-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; DEFAULT-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
; DEFAULT-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[DST]], i64 [[OFFSET_IDX]]
; DEFAULT-NEXT: [[TMP1:%.*]] = load i16, ptr [[SRC]], align 2
-; DEFAULT-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i16> poison, i16 [[TMP1]], i64 0
-; DEFAULT-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i16> [[BROADCAST_SPLATINSERT]], <8 x i16> poison, <8 x i32> zeroinitializer
-; DEFAULT-NEXT: [[TMP2:%.*]] = or <8 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
-; DEFAULT-NEXT: [[TMP3:%.*]] = uitofp <8 x i16> [[TMP2]] to <8 x double>
-; DEFAULT-NEXT: store <8 x double> [[TMP3]], ptr [[NEXT_GEP]], align 8
-; DEFAULT-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; DEFAULT-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
-; DEFAULT-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; DEFAULT-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i16> poison, i16 [[TMP1]], i64 0
+; DEFAULT-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i16> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+; DEFAULT-NEXT: [[TMP28:%.*]] = or <vscale x 2 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
+; DEFAULT-NEXT: [[TMP11:%.*]] = or <vscale x 2 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
+; DEFAULT-NEXT: [[TMP12:%.*]] = or <vscale x 2 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
+; DEFAULT-NEXT: [[TMP13:%.*]] = or <vscale x 2 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
+; DEFAULT-NEXT: [[TMP14:%.*]] = uitofp <vscale x 2 x i16> [[TMP28]] to <vscale x 2 x double>
+; DEFAULT-NEXT: [[TMP15:%.*]] = uitofp <vscale x 2 x i16> [[TMP11]] to <vscale x 2 x double>
+; DEFAULT-NEXT: [[TMP16:%.*]] = uitofp <vscale x 2 x i16> [[TMP12]] to <vscale x 2 x double>
+; DEFAULT-NEXT: [[TMP17:%.*]] = uitofp <vscale x 2 x i16> [[TMP13]] to <vscale x 2 x double>
+; DEFAULT-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; DEFAULT-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 2
+; DEFAULT-NEXT: [[TMP20:%.*]] = getelementptr double, ptr [[NEXT_GEP]], i64 [[TMP19]]
+; DEFAULT-NEXT: [[TMP21:%.*]] = call i64 @llvm.vscale.i64()
+; DEFAULT-NEXT: [[TMP22:%.*]] = mul nuw i64 [[TMP21]], 4
+; DEFAULT-NEXT: [[TMP23:%.*]] = getelementptr double, ptr [[NEXT_GEP]], i64 [[TMP22]]
+; DEFAULT-NEXT: [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; DEFAULT-NEXT: [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 6
+; DEFAULT-NEXT: [[TMP26:%.*]] = getelementptr double, ptr [[NEXT_GEP]], i64 [[TMP25]]
+; DEFAULT-NEXT: store <vscale x 2 x double> [[TMP14]], ptr [[NEXT_GEP]], align 8
+; DEFAULT-NEXT: store <vscale x 2 x double> [[TMP15]], ptr [[TMP20]], align 8
+; DEFAULT-NEXT: store <vscale x 2 x double> [[TMP16]], ptr [[TMP23]], align 8
+; DEFAULT-NEXT: store <vscale x 2 x double> [[TMP17]], ptr [[TMP26]], align 8
+; DEFAULT-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; DEFAULT-NEXT: [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; DEFAULT-NEXT: br i1 [[TMP27]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; DEFAULT: [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT: br label %[[SCALAR_PH]]
+; DEFAULT-NEXT: [[CMP_N:%.*]] = icmp eq i64 257, [[N_VEC]]
+; DEFAULT-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; DEFAULT: [[SCALAR_PH]]:
; DEFAULT-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[DST]], %[[ENTRY]] ]
-; DEFAULT-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ 512, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[TMP8]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; DEFAULT-NEXT: br label %[[LOOP:.*]]
; DEFAULT: [[LOOP]]:
; DEFAULT-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[PTR_IV_NEXT:%.*]], %[[LOOP]] ]
@@ -691,7 +721,7 @@ define void @multiple_exit_conditions(ptr %src, ptr noalias %dst) #1 {
; DEFAULT-NEXT: [[PTR_IV_NEXT]] = getelementptr i8, ptr [[PTR_IV]], i64 8
; DEFAULT-NEXT: [[IV_CLAMP:%.*]] = and i64 [[IV]], 4294967294
; DEFAULT-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_CLAMP]], 512
-; DEFAULT-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
+; DEFAULT-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
; DEFAULT: [[EXIT]]:
; DEFAULT-NEXT: ret void
;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/early_exit_costs.ll b/llvm/test/Transforms/LoopVectorize/AArch64/early_exit_costs.ll
index 19ee763e6ffae..a58e3511aad87 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/early_exit_costs.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/early_exit_costs.ll
@@ -92,9 +92,11 @@ loop.end:
define i64 @vectorization_not_profitable_due_to_trunc(ptr dereferenceable(800) %src) {
; CHECK-LABEL: LV: Checking a loop in 'vectorization_not_profitable_due_to_trunc'
-; CHECK: LV: Selecting VF: 1.
-; CHECK-NEXT: Calculating cost of work in exit block vector.early.exit:
-; CHECK-NEXT: LV: Vectorization is possible but not beneficial.
+; CHECK: Calculating cost of work in exit block vector.early.exit:
+; CHECK-NEXT: Cost of 6 for VF 2: EMIT vp<{{.*}}> = first-active-lane ir<{{.*}}>
+; CHECK-NEXT: Cost of 2 for VF 2: EMIT vp<{{.*}}> = extract-lane vp<{{.*}}>, ir<{{.*}}>
+; CHECK: LV: Minimum required TC for runtime checks to be profitable:28
+; CHECK: LV: Found a vectorizable loop (2)
entry:
br label %loop.header
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/fully-unrolled-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/fully-unrolled-cost.ll
index c3b0bc8c00a74..5b755d89f4fb5 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/fully-unrolled-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/fully-unrolled-cost.ll
@@ -9,11 +9,12 @@ define i64 @test(ptr %a, ptr %b) #0 {
; CHECK-LABEL: LV: Checking a loop in 'test'
; CHECK: Cost of 1 for VF 8: induction instruction %i.iv.next = add nuw nsw i64 %i.iv, 1
; CHECK-NEXT: Cost of 0 for VF 8: induction instruction %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]
-; CHECK-NEXT: Cost of 1 for VF 8: exit condition instruction %exitcond.not = icmp eq i64 %i.iv.next, 16
; CHECK-NEXT: Cost of 0 for VF 8: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 1 for VF 8: EMIT branch-on-count vp<%{{.+}}>, vp<%{{.+}}>
; CHECK: Cost for VF 8: 30
; CHECK-NEXT: Cost of 0 for VF 16: induction instruction %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]
; CHECK-NEXT: Cost of 0 for VF 16: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 0 for VF 16: EMIT branch-on-count vp<%{{.+}}>, vp<%{{.+}}>
; CHECK: Cost for VF 16: 56
; CHECK: LV: Selecting VF: 16
entry:
@@ -43,12 +44,13 @@ define i64 @test_external_iv_user(ptr %a, ptr %b) #0 {
; CHECK-LABEL: LV: Checking a loop in 'test_external_iv_user'
; CHECK: Cost of 1 for VF 8: induction instruction %i.iv.next = add nuw nsw i64 %i.iv, 1
; CHECK-NEXT: Cost of 0 for VF 8: induction instruction %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]
-; CHECK-NEXT: Cost of 1 for VF 8: exit condition instruction %exitcond.not = icmp eq i64 %i.iv.next, 16
; CHECK-NEXT: Cost of 0 for VF 8: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 1 for VF 8: EMIT branch-on-count vp<{{.+}}>, vp<{{.+}}>
; CHECK: Cost for VF 8: 30
; CHECK-NEXT: Cost of 1 for VF 16: induction instruction %i.iv.next = add nuw nsw i64 %i.iv, 1
; CHECK-NEXT: Cost of 0 for VF 16: induction instruction %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]
; CHECK-NEXT: Cost of 0 for VF 16: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 0 for VF 16: EMIT branch-on-count vp<{{.+}}>, vp<{{.+}}>
; CHECK: Cost for VF 16: 57
; CHECK: LV: Selecting VF: vscale x 2
entry:
@@ -80,12 +82,13 @@ define i64 @test_two_ivs(ptr %a, ptr %b, i64 %start) #0 {
; CHECK-NEXT: Cost of 0 for VF 8: induction instruction %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]
; CHECK-NEXT: Cost of 1 for VF 8: induction instruction %j.iv.next = add nuw nsw i64 %j.iv, 1
; CHECK-NEXT: Cost of 0 for VF 8: induction instruction %j.iv = phi i64 [ %start, %entry ], [ %j.iv.next, %for.body ]
-; CHECK-NEXT: Cost of 1 for VF 8: exit condition instruction %exitcond.not = icmp eq i64 %i.iv.next, 16
; CHECK-NEXT: Cost of 0 for VF 8: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 1 for VF 8: EMIT branch-on-count vp<%{{.+}}>, vp<%{{.+}}>
; CHECK: Cost for VF 8: 27
; CHECK-NEXT: Cost of 0 for VF 16: induction instruction %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]
; CHECK-NEXT: Cost of 0 for VF 16: induction instruction %j.iv = phi i64 [ %start, %entry ], [ %j.iv.next, %for.body ]
; CHECK-NEXT: Cost of 0 for VF 16: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 0 for VF 16: EMIT branch-on-count vp<%{{.+}}>, vp<%{{.+}}>
; CHECK: Cost for VF 16: 48
; CHECK: LV: Selecting VF: 16
entry:
@@ -116,11 +119,12 @@ define i1 @test_extra_cmp_user(ptr nocapture noundef %dst, ptr nocapture noundef
; CHECK-LABEL: LV: Checking a loop in 'test_extra_cmp_user'
; CHECK: Cost of 4 for VF 8: induction instruction %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
; CHECK-NEXT: Cost of 0 for VF 8: induction instruction %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
-; CHECK-NEXT: Cost of 4 for VF 8: exit condition instruction %exitcond.not = icmp eq i64 %indvars.iv.next, 16
; CHECK-NEXT: Cost of 0 for VF 8: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
-; CHECK: Cost for VF 8: 12
+; CHECK: Cost of 1 for VF 8: EMIT branch-on-count vp<%{{.+}}>, vp<%{{.+}}>
+; CHECK: Cost for VF 8: 13
; CHECK-NEXT: Cost of 0 for VF 16: induction instruction %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
; CHECK-NEXT: Cost of 0 for VF 16: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
+; CHECK: Cost of 0 for VF 16: EMIT branch-on-count vp<%{{.+}}>, vp<%{{.+}}>
; CHECK: Cost for VF 16: 4
; CHECK: LV: Selecting VF: 16
entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll b/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll
index 8b354d91909b1..87a5fb28f58f5 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll
@@ -96,46 +96,46 @@ define void @iv_casts(ptr %dst, ptr %src, i32 %x, i64 %N) #0 {
; PRED-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; PRED: [[VECTOR_MEMCHECK]]:
; PRED-NEXT: [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP2:%.*]] = mul nuw i64 [[TMP1]], 16
+; PRED-NEXT: [[TMP2:%.*]] = mul nuw i64 [[TMP1]], 8
; PRED-NEXT: [[TMP3:%.*]] = sub i64 [[DST1]], [[SRC2]]
; PRED-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP3]], [[TMP2]]
; PRED-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; PRED: [[VECTOR_PH]]:
; PRED-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 16
+; PRED-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
; PRED-NEXT: [[TMP8:%.*]] = sub i64 [[TMP5]], 1
; PRED-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], [[TMP8]]
; PRED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
; PRED-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; PRED-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 16
-; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[X]], i64 0
-; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
+; PRED-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 8
+; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[X]], i64 0
+; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
; PRED-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
+; PRED-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 8
; PRED-NEXT: [[TMP13:%.*]] = sub i64 [[TMP0]], [[TMP12]]
; PRED-NEXT: [[TMP14:%.*]] = icmp ugt i64 [[TMP0]], [[TMP12]]
; PRED-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i64 [[TMP13]], i64 0
-; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[TMP0]])
-; PRED-NEXT: [[TMP16:%.*]] = trunc <vscale x 16 x i32> [[BROADCAST_SPLAT]] to <vscale x 16 x i16>
+; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[TMP0]])
+; PRED-NEXT: [[TMP16:%.*]] = trunc <vscale x 8 x i32> [[BROADCAST_SPLAT]] to <vscale x 8 x i16>
; PRED-NEXT: br label %[[VECTOR_BODY:.*]]
; PRED: [[VECTOR_BODY]]:
; PRED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PRED-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[INDEX]]
-; PRED-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP18]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
-; PRED-NEXT: [[TMP17:%.*]] = zext <vscale x 16 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 16 x i16>
-; PRED-NEXT: [[TMP22:%.*]] = mul <vscale x 16 x i16> [[TMP17]], [[TMP16]]
-; PRED-NEXT: [[TMP24:%.*]] = zext <vscale x 16 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 16 x i16>
-; PRED-NEXT: [[TMP20:%.*]] = or <vscale x 16 x i16> [[TMP22]], [[TMP24]]
-; PRED-NEXT: [[TMP21:%.*]] = lshr <vscale x 16 x i16> [[TMP20]], splat (i16 1)
-; PRED-NEXT: [[TMP23:%.*]] = trunc <vscale x 16 x i16> [[TMP21]] to <vscale x 16 x i8>
+; PRED-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x i8> @llvm.masked.load.nxv8i8.p0(ptr [[TMP18]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i8> poison)
+; PRED-NEXT: [[TMP22:%.*]] = zext <vscale x 8 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 8 x i16>
+; PRED-NEXT: [[TMP17:%.*]] = mul <vscale x 8 x i16> [[TMP22]], [[TMP16]]
+; PRED-NEXT: [[TMP23:%.*]] = zext <vscale x 8 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 8 x i16>
+; PRED-NEXT: [[TMP19:%.*]] = or <vscale x 8 x i16> [[TMP17]], [[TMP23]]
+; PRED-NEXT: [[TMP20:%.*]] = lshr <vscale x 8 x i16> [[TMP19]], splat (i16 1)
+; PRED-NEXT: [[TMP21:%.*]] = trunc <vscale x 8 x i16> [[TMP20]] to <vscale x 8 x i8>
; PRED-NEXT: [[TMP26:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; PRED-NEXT: call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP23]], ptr [[TMP26]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; PRED-NEXT: call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP21]], ptr [[TMP26]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
; PRED-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP10]]
-; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP15]])
-; PRED-NEXT: [[TMP28:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; PRED-NEXT: [[TMP29:%.*]] = extractelement <vscale x 16 x i1> [[TMP28]], i32 0
+; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP15]])
+; PRED-NEXT: [[TMP28:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; PRED-NEXT: [[TMP29:%.*]] = extractelement <vscale x 8 x i1> [[TMP28]], i32 0
; PRED-NEXT: br i1 [[TMP29]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; PRED: [[MIDDLE_BLOCK]]:
; PRED-NEXT: br label %[[EXIT:.*]]
@@ -441,74 +441,56 @@ define void @trunc_ivs_and_store(i32 %x, ptr %dst, i64 %N) #0 {
; PRED-NEXT: [[TMP13:%.*]] = or i1 [[TMP9]], [[TMP12]]
; PRED-NEXT: br i1 [[TMP13]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; PRED: [[VECTOR_PH]]:
-; PRED-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], 3
-; PRED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4
+; PRED-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], 1
+; PRED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 2
; PRED-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; PRED-NEXT: [[TMP14:%.*]] = sub i64 [[TMP0]], 4
-; PRED-NEXT: [[TMP15:%.*]] = icmp ugt i64 [[TMP0]], 4
+; PRED-NEXT: [[TMP14:%.*]] = sub i64 [[TMP0]], 2
+; PRED-NEXT: [[TMP15:%.*]] = icmp ugt i64 [[TMP0]], 2
; PRED-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i64 [[TMP14]], i64 0
-; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 0, i64 [[TMP0]])
-; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[MUL]], i64 0
-; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <2 x i1> @llvm.get.active.lane.mask.v2i1.i64(i64 0, i64 [[TMP0]])
+; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[MUL]], i64 0
+; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; PRED-NEXT: br label %[[VECTOR_BODY:.*]]
; PRED: [[VECTOR_BODY]]:
-; PRED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE7:.*]] ]
-; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[PRED_STORE_CONTINUE7]] ]
-; PRED-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE7]] ]
+; PRED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE3:.*]] ]
+; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[PRED_STORE_CONTINUE3]] ]
+; PRED-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE3]] ]
; PRED-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
-; PRED-NEXT: [[TMP17:%.*]] = mul <4 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
-; PRED-NEXT: [[TMP18:%.*]] = zext <4 x i32> [[TMP17]] to <4 x i64>
-; PRED-NEXT: [[TMP19:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 0
+; PRED-NEXT: [[TMP18:%.*]] = mul <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
+; PRED-NEXT: [[TMP23:%.*]] = zext <2 x i32> [[TMP18]] to <2 x i64>
+; PRED-NEXT: [[TMP19:%.*]] = extractelement <2 x i1> [[ACTIVE_LANE_MASK]], i32 0
; PRED-NEXT: br i1 [[TMP19]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
; PRED: [[PRED_STORE_IF]]:
-; PRED-NEXT: [[TMP20:%.*]] = extractelement <4 x i64> [[TMP18]], i32 0
+; PRED-NEXT: [[TMP20:%.*]] = extractelement <2 x i64> [[TMP23]], i32 0
; PRED-NEXT: [[TMP21:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP20]]
; PRED-NEXT: [[TMP22:%.*]] = add i32 [[OFFSET_IDX]], 0
; PRED-NEXT: store i32 [[TMP22]], ptr [[TMP21]], align 4
; PRED-NEXT: br label %[[PRED_STORE_CONTINUE]]
; PRED: [[PRED_STORE_CONTINUE]]:
-; PRED-NEXT: [[TMP23:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 1
-; PRED-NEXT: br i1 [[TMP23]], label %[[PRED_STORE_IF2:.*]], label %[[PRED_STORE_CONTINUE3:.*]]
+; PRED-NEXT: [[TMP27:%.*]] = extractelement <2 x i1> [[ACTIVE_LANE_MASK]], i32 1
+; PRED-NEXT: br i1 [[TMP27]], label %[[PRED_STORE_IF2:.*]], label %[[PRED_STORE_CONTINUE3]]
; PRED: [[PRED_STORE_IF2]]:
-; PRED-NEXT: [[TMP24:%.*]] = extractelement <4 x i64> [[TMP18]], i32 1
+; PRED-NEXT: [[TMP24:%.*]] = extractelement <2 x i64> [[TMP23]], i32 1
; PRED-NEXT: [[TMP25:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP24]]
; PRED-NEXT: [[TMP26:%.*]] = add i32 [[OFFSET_IDX]], 1
; PRED-NEXT: store i32 [[TMP26]], ptr [[TMP25]], align 4
; PRED-NEXT: br label %[[PRED_STORE_CONTINUE3]]
; PRED: [[PRED_STORE_CONTINUE3]]:
-; PRED-NEXT: [[TMP27:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 2
-; PRED-NEXT: br i1 [[TMP27]], label %[[PRED_STORE_IF4:.*]], label %[[PRED_STORE_CONTINUE5:.*]]
-; PRED: [[PRED_STORE_IF4]]:
-; PRED-NEXT: [[TMP28:%.*]] = extractelement <4 x i64> [[TMP18]], i32 2
-; PRED-NEXT: [[TMP29:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP28]]
-; PRED-NEXT: [[TMP30:%.*]] = add i32 [[OFFSET_IDX]], 2
-; PRED-NEXT: store i32 [[TMP30]], ptr [[TMP29]], align 4
-; PRED-NEXT: br label %[[PRED_STORE_CONTINUE5]]
-; PRED: [[PRED_STORE_CONTINUE5]]:
-; PRED-NEXT: [[TMP31:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 3
-; PRED-NEXT: br i1 [[TMP31]], label %[[PRED_STORE_IF6:.*]], label %[[PRED_STORE_CONTINUE7]]
-; PRED: [[PRED_STORE_IF6]]:
-; PRED-NEXT: [[TMP32:%.*]] = extractelement <4 x i64> [[TMP18]], i32 3
-; PRED-NEXT: [[TMP33:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP32]]
-; PRED-NEXT: [[TMP34:%.*]] = add i32 [[OFFSET_IDX]], 3
-; PRED-NEXT: store i32 [[TMP34]], ptr [[TMP33]], align 4
-; PRED-NEXT: br label %[[PRED_STORE_CONTINUE7]]
-; PRED: [[PRED_STORE_CONTINUE7]]:
-; PRED-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
-; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 [[INDEX]], i64 [[TMP16]])
-; PRED-NEXT: [[TMP35:%.*]] = xor <4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; PRED-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
-; PRED-NEXT: [[TMP36:%.*]] = extractelement <4 x i1> [[TMP35]], i32 0
-; PRED-NEXT: br i1 [[TMP36]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; PRED-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 2
+; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <2 x i1> @llvm.get.active.lane.mask.v2i1.i64(i64 [[INDEX]], i64 [[TMP16]])
+; PRED-NEXT: [[TMP28:%.*]] = xor <2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; PRED-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
+; PRED-NEXT: [[TMP29:%.*]] = extractelement <2 x i1> [[TMP28]], i32 0
+; PRED-NEXT: br i1 [[TMP29]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; PRED: [[MIDDLE_BLOCK]]:
; PRED-NEXT: br label %[[EXIT:.*]]
; PRED: [[SCALAR_PH]]:
; PRED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
-; PRED-NEXT: [[BC_RESUME_VAL8:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; PRED-NEXT: [[BC_RESUME_VAL4:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
; PRED-NEXT: br label %[[LOOP:.*]]
; PRED: [[LOOP]]:
; PRED-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], %[[LOOP]] ]
-; PRED-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL8]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP]] ]
+; PRED-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL4]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP]] ]
; PRED-NEXT: [[IV_1_TRUNC:%.*]] = trunc i64 [[IV_1]] to i32
; PRED-NEXT: [[IV_1_MUL:%.*]] = mul i32 [[MUL]], [[IV_1_TRUNC]]
; PRED-NEXT: [[IV_2_NEXT]] = add i32 [[IV_2]], 1
@@ -635,74 +617,56 @@ define void @ivs_trunc_and_ext(i32 %x, ptr %dst, i64 %N) #0 {
; PRED-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
; PRED-NEXT: br i1 [[TMP12]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; PRED: [[VECTOR_PH]]:
-; PRED-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], 3
-; PRED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4
+; PRED-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], 1
+; PRED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 2
; PRED-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; PRED-NEXT: [[TMP13:%.*]] = sub i64 [[TMP0]], 4
-; PRED-NEXT: [[TMP14:%.*]] = icmp ugt i64 [[TMP0]], 4
+; PRED-NEXT: [[TMP13:%.*]] = sub i64 [[TMP0]], 2
+; PRED-NEXT: [[TMP14:%.*]] = icmp ugt i64 [[TMP0]], 2
; PRED-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i64 [[TMP13]], i64 0
-; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 0, i64 [[TMP0]])
-; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[ADD]], i64 0
-; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <2 x i1> @llvm.get.active.lane.mask.v2i1.i64(i64 0, i64 [[TMP0]])
+; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[ADD]], i64 0
+; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; PRED-NEXT: br label %[[VECTOR_BODY:.*]]
; PRED: [[VECTOR_BODY]]:
-; PRED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE6:.*]] ]
-; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[PRED_STORE_CONTINUE6]] ]
-; PRED-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE6]] ]
+; PRED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE2:.*]] ]
+; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[PRED_STORE_CONTINUE2]] ]
+; PRED-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE2]] ]
; PRED-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
-; PRED-NEXT: [[TMP16:%.*]] = mul <4 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
-; PRED-NEXT: [[TMP17:%.*]] = zext <4 x i32> [[TMP16]] to <4 x i64>
-; PRED-NEXT: [[TMP18:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 0
+; PRED-NEXT: [[TMP17:%.*]] = mul <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
+; PRED-NEXT: [[TMP22:%.*]] = zext <2 x i32> [[TMP17]] to <2 x i64>
+; PRED-NEXT: [[TMP18:%.*]] = extractelement <2 x i1> [[ACTIVE_LANE_MASK]], i32 0
; PRED-NEXT: br i1 [[TMP18]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
; PRED: [[PRED_STORE_IF]]:
-; PRED-NEXT: [[TMP19:%.*]] = extractelement <4 x i64> [[TMP17]], i32 0
+; PRED-NEXT: [[TMP19:%.*]] = extractelement <2 x i64> [[TMP22]], i32 0
; PRED-NEXT: [[TMP20:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP19]]
; PRED-NEXT: [[TMP21:%.*]] = add i32 [[OFFSET_IDX]], 0
; PRED-NEXT: store i32 [[TMP21]], ptr [[TMP20]], align 4
; PRED-NEXT: br label %[[PRED_STORE_CONTINUE]]
; PRED: [[PRED_STORE_CONTINUE]]:
-; PRED-NEXT: [[TMP22:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 1
-; PRED-NEXT: br i1 [[TMP22]], label %[[PRED_STORE_IF1:.*]], label %[[PRED_STORE_CONTINUE2:.*]]
+; PRED-NEXT: [[TMP26:%.*]] = extractelement <2 x i1> [[ACTIVE_LANE_MASK]], i32 1
+; PRED-NEXT: br i1 [[TMP26]], label %[[PRED_STORE_IF1:.*]], label %[[PRED_STORE_CONTINUE2]]
; PRED: [[PRED_STORE_IF1]]:
-; PRED-NEXT: [[TMP23:%.*]] = extractelement <4 x i64> [[TMP17]], i32 1
+; PRED-NEXT: [[TMP23:%.*]] = extractelement <2 x i64> [[TMP22]], i32 1
; PRED-NEXT: [[TMP24:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP23]]
; PRED-NEXT: [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], 1
; PRED-NEXT: store i32 [[TMP25]], ptr [[TMP24]], align 4
; PRED-NEXT: br label %[[PRED_STORE_CONTINUE2]]
; PRED: [[PRED_STORE_CONTINUE2]]:
-; PRED-NEXT: [[TMP26:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 2
-; PRED-NEXT: br i1 [[TMP26]], label %[[PRED_STORE_IF3:.*]], label %[[PRED_STORE_CONTINUE4:.*]]
-; PRED: [[PRED_STORE_IF3]]:
-; PRED-NEXT: [[TMP27:%.*]] = extractelement <4 x i64> [[TMP17]], i32 2
-; PRED-NEXT: [[TMP28:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP27]]
-; PRED-NEXT: [[TMP29:%.*]] = add i32 [[OFFSET_IDX]], 2
-; PRED-NEXT: store i32 [[TMP29]], ptr [[TMP28]], align 4
-; PRED-NEXT: br label %[[PRED_STORE_CONTINUE4]]
-; PRED: [[PRED_STORE_CONTINUE4]]:
-; PRED-NEXT: [[TMP30:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 3
-; PRED-NEXT: br i1 [[TMP30]], label %[[PRED_STORE_IF5:.*]], label %[[PRED_STORE_CONTINUE6]]
-; PRED: [[PRED_STORE_IF5]]:
-; PRED-NEXT: [[TMP31:%.*]] = extractelement <4 x i64> [[TMP17]], i32 3
-; PRED-NEXT: [[TMP32:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP31]]
-; PRED-NEXT: [[TMP33:%.*]] = add i32 [[OFFSET_IDX]], 3
-; PRED-NEXT: store i32 [[TMP33]], ptr [[TMP32]], align 4
-; PRED-NEXT: br label %[[PRED_STORE_CONTINUE6]]
-; PRED: [[PRED_STORE_CONTINUE6]]:
-; PRED-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
-; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 [[INDEX]], i64 [[TMP15]])
-; PRED-NEXT: [[TMP34:%.*]] = xor <4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; PRED-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
-; PRED-NEXT: [[TMP35:%.*]] = extractelement <4 x i1> [[TMP34]], i32 0
-; PRED-NEXT: br i1 [[TMP35]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; PRED-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 2
+; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <2 x i1> @llvm.get.active.lane.mask.v2i1.i64(i64 [[INDEX]], i64 [[TMP15]])
+; PRED-NEXT: [[TMP27:%.*]] = xor <2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; PRED-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
+; PRED-NEXT: [[TMP28:%.*]] = extractelement <2 x i1> [[TMP27]], i32 0
+; PRED-NEXT: br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; PRED: [[MIDDLE_BLOCK]]:
; PRED-NEXT: br label %[[EXIT:.*]]
; PRED: [[SCALAR_PH]]:
; PRED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
-; PRED-NEXT: [[BC_RESUME_VAL7:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; PRED-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
; PRED-NEXT: br label %[[LOOP:.*]]
; PRED: [[LOOP]]:
; PRED-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], %[[LOOP]] ]
-; PRED-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL7]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP]] ]
+; PRED-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP]] ]
; PRED-NEXT: [[IV_TRUNC:%.*]] = trunc i64 [[IV_1]] to i32
; PRED-NEXT: [[IV_MUL:%.*]] = mul i32 [[ADD]], [[IV_TRUNC]]
; PRED-NEXT: [[IV_2_NEXT]] = add i32 [[IV_2]], 1
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll b/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
index 08d35f71e7cc3..3c6ff99efa14a 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
@@ -453,38 +453,38 @@ define i16 @reduce_udiv(ptr %src, i16 %x, i64 %N) #0 {
; PRED-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; PRED: [[VECTOR_PH]]:
; PRED-NEXT: [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP2:%.*]] = mul nuw i64 [[TMP1]], 8
+; PRED-NEXT: [[TMP2:%.*]] = mul nuw i64 [[TMP1]], 4
; PRED-NEXT: [[TMP5:%.*]] = sub i64 [[TMP2]], 1
; PRED-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], [[TMP5]]
; PRED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP2]]
; PRED-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; PRED-NEXT: [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP7:%.*]] = mul nuw i64 [[TMP6]], 8
+; PRED-NEXT: [[TMP7:%.*]] = mul nuw i64 [[TMP6]], 4
; PRED-NEXT: [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; PRED-NEXT: [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
+; PRED-NEXT: [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 4
; PRED-NEXT: [[TMP10:%.*]] = sub i64 [[TMP0]], [[TMP9]]
; PRED-NEXT: [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], [[TMP9]]
; PRED-NEXT: [[TMP12:%.*]] = select i1 [[TMP11]], i64 [[TMP10]], i64 0
-; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[TMP0]])
-; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i16> poison, i16 [[X]], i64 0
-; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i16> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer
+; PRED-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[TMP0]])
+; PRED-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i16> poison, i16 [[X]], i64 0
+; PRED-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i16> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer
; PRED-NEXT: br label %[[VECTOR_BODY:.*]]
; PRED: [[VECTOR_BODY]]:
; PRED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PRED-NEXT: [[VEC_PHI:%.*]] = phi <vscale x 8 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
+; PRED-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; PRED-NEXT: [[VEC_PHI:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
; PRED-NEXT: [[TMP14:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[INDEX]]
-; PRED-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr [[TMP14]], i32 2, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> poison)
-; PRED-NEXT: [[TMP20:%.*]] = udiv <vscale x 8 x i16> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
-; PRED-NEXT: [[TMP21:%.*]] = or <vscale x 8 x i16> [[TMP20]], [[VEC_PHI]]
-; PRED-NEXT: [[TMP16]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> [[TMP21]], <vscale x 8 x i16> [[VEC_PHI]]
+; PRED-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[TMP14]], i32 2, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i16> poison)
+; PRED-NEXT: [[TMP15:%.*]] = udiv <vscale x 4 x i16> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
+; PRED-NEXT: [[TMP13:%.*]] = or <vscale x 4 x i16> [[TMP15]], [[VEC_PHI]]
+; PRED-NEXT: [[TMP16]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i16> [[TMP13]], <vscale x 4 x i16> [[VEC_PHI]]
; PRED-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP7]]
-; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP12]])
-; PRED-NEXT: [[TMP17:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; PRED-NEXT: [[TMP18:%.*]] = extractelement <vscale x 8 x i1> [[TMP17]], i32 0
+; PRED-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP12]])
+; PRED-NEXT: [[TMP17:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; PRED-NEXT: [[TMP18:%.*]] = extractelement <vscale x 4 x i1> [[TMP17]], i32 0
; PRED-NEXT: br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; PRED: [[MIDDLE_BLOCK]]:
-; PRED-NEXT: [[TMP19:%.*]] = call i16 @llvm.vector.reduce.or.nxv8i16(<vscale x 8 x i16> [[TMP16]])
+; PRED-NEXT: [[TMP19:%.*]] = call i16 @llvm.vector.reduce.or.nxv4i16(<vscale x 4 x i16> [[TMP16]])
; PRED-NEXT: br label %[[EXIT:.*]]
; PRED: [[SCALAR_PH]]:
; PRED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll b/llvm/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll
index bdf832f32964f..13a686251daef 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll
@@ -21,7 +21,6 @@ target triple = "thumbv8.1m.main-arm-none-eabi"
; CHECK: LV: Scalar loop costs: 5.
; CHECK: Cost of 1 for VF 2: induction instruction %inc = add nuw nsw i32 %i.016, 1
; CHECK: Cost of 0 for VF 2: induction instruction %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
-; CHECK: Cost of 1 for VF 2: exit condition instruction %exitcond.not = icmp eq i32 %inc, %n
; CHECK: Cost of 0 for VF 2: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 2: vp<{{.+}}> = SCALAR-STEPS vp<{{.+}}>, ir<1>
; CHECK: Cost of 0 for VF 2: CLONE ir<%arrayidx> = getelementptr inbounds ir<%s>, vp<{{.+}}>
@@ -34,11 +33,10 @@ target triple = "thumbv8.1m.main-arm-none-eabi"
; CHECK: Cost of 0 for VF 2: vp<{{.+}}> = vector-pointer ir<%arrayidx7>
; CHECK: Cost of 16 for VF 2: WIDEN store vp<{{.+}}>, ir<%conv6>, ir<%cmp2>
; CHECK: Cost of 0 for VF 2: EMIT vp<%index.next> = add nuw vp<{{.+}}>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 2: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 2: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 2: 86 (Estimated cost per lane: 43.
; CHECK: Cost of 1 for VF 4: induction instruction %inc = add nuw nsw i32 %i.016, 1
; CHECK: Cost of 0 for VF 4: induction instruction %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
-; CHECK: Cost of 1 for VF 4: exit condition instruction %exitcond.not = icmp eq i32 %inc, %n
; CHECK: Cost of 0 for VF 4: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 4: vp<{{.+}}> = SCALAR-STEPS vp<{{.+}}>, ir<1>
; CHECK: Cost of 0 for VF 4: CLONE ir<%arrayidx> = getelementptr inbounds ir<%s>, vp<{{.+}}>
@@ -51,11 +49,10 @@ target triple = "thumbv8.1m.main-arm-none-eabi"
; CHECK: Cost of 0 for VF 4: vp<{{.+}}> = vector-pointer ir<%arrayidx7>
; CHECK: Cost of 2 for VF 4: WIDEN store vp<{{.+}}>, ir<%conv6>, ir<%cmp2>
; CHECK: Cost of 0 for VF 4: EMIT vp<%index.next> = add nuw vp<{{.+}}>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 4: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 4: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 4: 10 (Estimated cost per lane: 2.
; CHECK: Cost of 1 for VF 8: induction instruction %inc = add nuw nsw i32 %i.016, 1
; CHECK: Cost of 0 for VF 8: induction instruction %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
-; CHECK: Cost of 1 for VF 8: exit condition instruction %exitcond.not = icmp eq i32 %inc, %n
; CHECK: Cost of 0 for VF 8: EMIT vp<{{.+}}> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 8: vp<{{.+}}> = SCALAR-STEPS vp<{{.+}}>, ir<1>
; CHECK: Cost of 0 for VF 8: CLONE ir<%arrayidx> = getelementptr inbounds ir<%s>, vp<{{.+}}>
@@ -68,7 +65,7 @@ target triple = "thumbv8.1m.main-arm-none-eabi"
; CHECK: Cost of 0 for VF 8: vp<{{.+}}> = vector-pointer ir<%arrayidx7>
; CHECK: Cost of 2 for VF 8: WIDEN store vp<{{.+}}>, ir<%conv6>, ir<%cmp2>
; CHECK: Cost of 0 for VF 8: EMIT vp<%index.next> = add nuw vp<{{.+}}>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 8: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 8: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 8: 46 (Estimated cost per lane: 5.
; CHECK: LV: Selecting VF: 4.
define void @expensive_icmp(ptr noalias nocapture %d, ptr nocapture readonly %s, i32 %n, i16 zeroext %m) #0 {
@@ -134,7 +131,6 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 2: induction instruction %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: Cost of 0 for VF 2: induction instruction %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: Cost of 0 for VF 2: induction instruction %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
-; CHECK: Cost of 1 for VF 2: exit condition instruction %cmp.not = icmp eq i32 %dec, 0
; CHECK: Cost of 0 for VF 2: EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 2: vp<[[STEPS1:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>
; CHECK: Cost of 0 for VF 2: EMIT vp<%next.gep> = ptradd ir<%pSrcA>, vp<[[STEPS1]]>
@@ -156,7 +152,7 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 2: vp<[[VEC_PTR3:%.+]]> = vector-pointer vp<%next.gep>.1
; CHECK: Cost of 18 for VF 2: WIDEN store vp<[[VEC_PTR3]]>, ir<%conv4>
; CHECK: Cost of 0 for VF 2: EMIT vp<%index.next> = add nuw vp<[[CAN_IV]]>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 2: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 2: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 2: 130 (Estimated cost per lane: 65.
; CHECK: Cost of 1 for VF 4: induction instruction %dec = add i32 %blkCnt.012, -1
; CHECK: Cost of 0 for VF 4: induction instruction %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
@@ -166,7 +162,6 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 4: induction instruction %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: Cost of 0 for VF 4: induction instruction %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: Cost of 0 for VF 4: induction instruction %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
-; CHECK: Cost of 1 for VF 4: exit condition instruction %cmp.not = icmp eq i32 %dec, 0
; CHECK: Cost of 0 for VF 4: EMIT vp<[[CAN_IV:%.]]> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 4: vp<[[STEPS1:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>
; CHECK: Cost of 0 for VF 4: EMIT vp<%next.gep> = ptradd ir<%pSrcA>, vp<[[STEPS1]]>
@@ -188,7 +183,7 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 4: vp<[[VEC_PTR2:%.+]]> = vector-pointer vp<%next.gep>.1
; CHECK: Cost of 2 for VF 4: WIDEN store vp<[[VEC_PTR2]]>, ir<%conv4>
; CHECK: Cost of 0 for VF 4: EMIT vp<%index.next> = add nuw vp<[[CAN_IV]]>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 4: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 4: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 4: 14 (Estimated cost per lane: 3.
; CHECK: Cost of 1 for VF 8: induction instruction %dec = add i32 %blkCnt.012, -1
; CHECK: Cost of 0 for VF 8: induction instruction %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
@@ -198,7 +193,6 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 8: induction instruction %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: Cost of 0 for VF 8: induction instruction %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: Cost of 0 for VF 8: induction instruction %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
-; CHECK: Cost of 1 for VF 8: exit condition instruction %cmp.not = icmp eq i32 %dec, 0
; CHECK: Cost of 0 for VF 8: EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 8: vp<[[STEPS1:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>
; CHECK: Cost of 0 for VF 8: EMIT vp<%next.gep> = ptradd ir<%pSrcA>, vp<[[STEPS1]]>
@@ -220,7 +214,7 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 8: vp<[[VEC_PTR3:%.+]]> = vector-pointer vp<%next.gep>.1
; CHECK: Cost of 2 for VF 8: WIDEN store vp<[[VEC_PTR3]]>, ir<%conv4>
; CHECK: Cost of 0 for VF 8: EMIT vp<%index.next> = add nuw vp<[[CAN_IV]]>, vp<{{.+}}
-; CHECK: Cost of 0 for VF 8: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 8: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 8: 26 (Estimated cost per lane: 3.
; CHECK: Cost of 1 for VF 16: induction instruction %dec = add i32 %blkCnt.012, -1
; CHECK: Cost of 0 for VF 16: induction instruction %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
@@ -230,7 +224,6 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 16: induction instruction %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: Cost of 0 for VF 16: induction instruction %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: Cost of 0 for VF 16: induction instruction %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
-; CHECK: Cost of 1 for VF 16: exit condition instruction %cmp.not = icmp eq i32 %dec, 0
; CHECK: Cost of 0 for VF 16: EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 16: vp<[[STEPS1:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>
; CHECK: Cost of 0 for VF 16: EMIT vp<%next.gep> = ptradd ir<%pSrcA>, vp<[[STEPS1]]>
@@ -252,7 +245,7 @@ for.inc: ; preds = %for.body, %if.then
; CHECK: Cost of 0 for VF 16: vp<[[VEC_PTR2:%.+]]> = vector-pointer vp<%next.gep>.1
; CHECK: Cost of 2 for VF 16: WIDEN store vp<[[VEC_PTR2]]>, ir<%conv4>
; CHECK: Cost of 0 for VF 16: EMIT vp<%index.next> = add nuw vp<[[CAN_IV]]>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 16: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 16: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
; CHECK: Cost for VF 16: 50
; CHECK: LV: Selecting VF: 16.
define void @cheap_icmp(ptr nocapture readonly %pSrcA, ptr nocapture readonly %pSrcB, ptr nocapture %pDst, i32 %blockSize) #0 {
diff --git a/llvm/test/Transforms/LoopVectorize/X86/CostModel/vpinstruction-cost.ll b/llvm/test/Transforms/LoopVectorize/X86/CostModel/vpinstruction-cost.ll
index bb85b88f181f7..ec01bd068f9fc 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/CostModel/vpinstruction-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/CostModel/vpinstruction-cost.ll
@@ -9,7 +9,6 @@ define void @wide_or_replaced_with_add_vpinstruction(ptr %src, ptr noalias %dst)
; CHECK-LABEL: 'wide_or_replaced_with_add_vpinstruction'
; CHECK: Cost of 1 for VF 2: induction instruction %iv.next = add nuw nsw i64 %iv, 1
; CHECK: Cost of 0 for VF 2: induction instruction %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
-; CHECK: Cost of 1 for VF 2: exit condition instruction %exitcond = icmp eq i64 %iv.next, 32
; CHECK: Cost of 0 for VF 2: EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 2: ir<%iv> = WIDEN-INDUCTION ir<0>, ir<1>, vp<%0>
; CHECK: Cost of 0 for VF 2: vp<%4> = SCALAR-STEPS vp<%3>, ir<1>
@@ -23,11 +22,10 @@ define void @wide_or_replaced_with_add_vpinstruction(ptr %src, ptr noalias %dst)
; CHECK: Cost of 0 for VF 2: vp<%6> = vector-pointer ir<%g.dst>
; CHECK: Cost of 1 for VF 2: WIDEN store vp<%6>, ir<%iv.4>, ir<%c>
; CHECK: Cost of 0 for VF 2: EMIT vp<%index.next> = add nuw vp<%3>, vp<%1>
-; CHECK: Cost of 0 for VF 2: EMIT branch-on-count vp<%index.next>, vp<%2>
+; CHECK: Cost of 1 for VF 2: EMIT branch-on-count vp<%index.next>, vp<%2>
; CHECK: Cost of 0 for VF 2: vector loop backedge
; CHECK: Cost of 1 for VF 4: induction instruction %iv.next = add nuw nsw i64 %iv, 1
; CHECK: Cost of 0 for VF 4: induction instruction %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
-; CHECK: Cost of 1 for VF 4: exit condition instruction %exitcond = icmp eq i64 %iv.next, 32
; CHECK: Cost of 0 for VF 4: EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 0 for VF 4: ir<%iv> = WIDEN-INDUCTION ir<0>, ir<1>, vp<%0>
; CHECK: Cost of 0 for VF 4: vp<%4> = SCALAR-STEPS vp<%3>, ir<1>
@@ -41,11 +39,10 @@ define void @wide_or_replaced_with_add_vpinstruction(ptr %src, ptr noalias %dst)
; CHECK: Cost of 0 for VF 4: vp<%6> = vector-pointer ir<%g.dst>
; CHECK: Cost of 1 for VF 4: WIDEN store vp<%6>, ir<%iv.4>, ir<%c>
; CHECK: Cost of 0 for VF 4: EMIT vp<%index.next> = add nuw vp<%3>, vp<%1>
-; CHECK: Cost of 0 for VF 4: EMIT branch-on-count vp<%index.next>, vp<%2>
+; CHECK: Cost of 1 for VF 4: EMIT branch-on-count vp<%index.next>, vp<%2>
; CHECK: Cost of 0 for VF 4: vector loop backedge
; CHECK: Cost of 1 for VF 4: induction instruction %iv.next = add nuw nsw i64 %iv, 1
; CHECK: Cost of 0 for VF 4: induction instruction %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
-; CHECK: Cost of 1 for VF 4: exit condition instruction %exitcond = icmp eq i64 %iv.next, 32
;
entry:
br label %loop.header
diff --git a/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll b/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
index 85b475c996c79..42aa6be3af116 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
@@ -335,7 +335,7 @@ define void @multi_exit(ptr %dst, ptr %src.1, ptr %src.2, i64 %A, i64 %B) #0 {
; CHECK-NEXT: [[TMP1:%.*]] = freeze i64 [[TMP0]]
; CHECK-NEXT: [[UMIN7:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 [[A:%.*]])
; CHECK-NEXT: [[TMP2:%.*]] = add nuw i64 [[UMIN7]], 1
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP2]], 28
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP2]], 14
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[B]], i64 1)
@@ -1216,49 +1216,49 @@ define i32 @g(i64 %n) {
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i64 [[N]], 4294967295
; CHECK-NEXT: br i1 [[TMP2]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK: vector.main.loop.iter.check:
-; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i32 [[TMP1]], 16
+; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i32 [[TMP1]], 32
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 16
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 32
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP1]], [[N_MOD_VF]]
-; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[N]], i64 0
-; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i64> poison, i64 [[N]], i64 0
+; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT]], <8 x i64> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI3:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI4:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
-; CHECK-NEXT: [[STEP_ADD_2:%.*]] = add <4 x i32> [[STEP_ADD]], splat (i32 4)
-; CHECK-NEXT: [[STEP_ADD_3:%.*]] = add <4 x i32> [[STEP_ADD_2]], splat (i32 4)
-; CHECK-NEXT: [[TMP3:%.*]] = zext <4 x i32> [[VEC_IND]] to <4 x i64>
-; CHECK-NEXT: [[TMP4:%.*]] = zext <4 x i32> [[STEP_ADD]] to <4 x i64>
-; CHECK-NEXT: [[TMP5:%.*]] = zext <4 x i32> [[STEP_ADD_2]] to <4 x i64>
-; CHECK-NEXT: [[TMP6:%.*]] = zext <4 x i32> [[STEP_ADD_3]] to <4 x i64>
-; CHECK-NEXT: [[TMP7:%.*]] = icmp eq <4 x i64> [[BROADCAST_SPLAT]], [[TMP3]]
-; CHECK-NEXT: [[TMP8:%.*]] = icmp eq <4 x i64> [[BROADCAST_SPLAT]], [[TMP4]]
-; CHECK-NEXT: [[TMP9:%.*]] = icmp eq <4 x i64> [[BROADCAST_SPLAT]], [[TMP5]]
-; CHECK-NEXT: [[TMP10:%.*]] = icmp eq <4 x i64> [[BROADCAST_SPLAT]], [[TMP6]]
-; CHECK-NEXT: [[TMP11:%.*]] = select <4 x i1> [[TMP7]], <4 x i32> zeroinitializer, <4 x i32> splat (i32 2)
-; CHECK-NEXT: [[TMP12:%.*]] = select <4 x i1> [[TMP8]], <4 x i32> zeroinitializer, <4 x i32> splat (i32 2)
-; CHECK-NEXT: [[TMP13:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> zeroinitializer, <4 x i32> splat (i32 2)
-; CHECK-NEXT: [[TMP14:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> zeroinitializer, <4 x i32> splat (i32 2)
-; CHECK-NEXT: [[TMP15]] = or <4 x i32> [[TMP11]], [[VEC_PHI]]
-; CHECK-NEXT: [[TMP16]] = or <4 x i32> [[TMP12]], [[VEC_PHI2]]
-; CHECK-NEXT: [[TMP17]] = or <4 x i32> [[TMP13]], [[VEC_PHI3]]
-; CHECK-NEXT: [[TMP18]] = or <4 x i32> [[TMP14]], [[VEC_PHI4]]
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 16
-; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD_3]], splat (i32 4)
+; CHECK-NEXT: [[VEC_IND:%.*]] = phi <8 x i32> [ <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI2:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI3:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI4:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[STEP_ADD:%.*]] = add <8 x i32> [[VEC_IND]], splat (i32 8)
+; CHECK-NEXT: [[STEP_ADD_2:%.*]] = add <8 x i32> [[STEP_ADD]], splat (i32 8)
+; CHECK-NEXT: [[STEP_ADD_3:%.*]] = add <8 x i32> [[STEP_ADD_2]], splat (i32 8)
+; CHECK-NEXT: [[TMP3:%.*]] = zext <8 x i32> [[VEC_IND]] to <8 x i64>
+; CHECK-NEXT: [[TMP4:%.*]] = zext <8 x i32> [[STEP_ADD]] to <8 x i64>
+; CHECK-NEXT: [[TMP5:%.*]] = zext <8 x i32> [[STEP_ADD_2]] to <8 x i64>
+; CHECK-NEXT: [[TMP6:%.*]] = zext <8 x i32> [[STEP_ADD_3]] to <8 x i64>
+; CHECK-NEXT: [[TMP7:%.*]] = icmp eq <8 x i64> [[BROADCAST_SPLAT]], [[TMP3]]
+; CHECK-NEXT: [[TMP8:%.*]] = icmp eq <8 x i64> [[BROADCAST_SPLAT]], [[TMP4]]
+; CHECK-NEXT: [[TMP9:%.*]] = icmp eq <8 x i64> [[BROADCAST_SPLAT]], [[TMP5]]
+; CHECK-NEXT: [[TMP10:%.*]] = icmp eq <8 x i64> [[BROADCAST_SPLAT]], [[TMP6]]
+; CHECK-NEXT: [[TMP11:%.*]] = select <8 x i1> [[TMP7]], <8 x i32> zeroinitializer, <8 x i32> splat (i32 2)
+; CHECK-NEXT: [[TMP12:%.*]] = select <8 x i1> [[TMP8]], <8 x i32> zeroinitializer, <8 x i32> splat (i32 2)
+; CHECK-NEXT: [[TMP13:%.*]] = select <8 x i1> [[TMP9]], <8 x i32> zeroinitializer, <8 x i32> splat (i32 2)
+; CHECK-NEXT: [[TMP14:%.*]] = select <8 x i1> [[TMP10]], <8 x i32> zeroinitializer, <8 x i32> splat (i32 2)
+; CHECK-NEXT: [[TMP15]] = or <8 x i32> [[TMP11]], [[VEC_PHI]]
+; CHECK-NEXT: [[TMP16]] = or <8 x i32> [[TMP12]], [[VEC_PHI2]]
+; CHECK-NEXT: [[TMP17]] = or <8 x i32> [[TMP13]], [[VEC_PHI3]]
+; CHECK-NEXT: [[TMP18]] = or <8 x i32> [[TMP14]], [[VEC_PHI4]]
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 32
+; CHECK-NEXT: [[VEC_IND_NEXT]] = add <8 x i32> [[STEP_ADD_3]], splat (i32 8)
; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; CHECK: middle.block:
-; CHECK-NEXT: [[BIN_RDX:%.*]] = or <4 x i32> [[TMP16]], [[TMP15]]
-; CHECK-NEXT: [[BIN_RDX5:%.*]] = or <4 x i32> [[TMP17]], [[BIN_RDX]]
-; CHECK-NEXT: [[BIN_RDX6:%.*]] = or <4 x i32> [[TMP18]], [[BIN_RDX5]]
-; CHECK-NEXT: [[TMP20:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[BIN_RDX6]])
+; CHECK-NEXT: [[BIN_RDX:%.*]] = or <8 x i32> [[TMP16]], [[TMP15]]
+; CHECK-NEXT: [[BIN_RDX5:%.*]] = or <8 x i32> [[TMP17]], [[BIN_RDX]]
+; CHECK-NEXT: [[BIN_RDX6:%.*]] = or <8 x i32> [[TMP18]], [[BIN_RDX5]]
+; CHECK-NEXT: [[TMP20:%.*]] = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> [[BIN_RDX6]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP1]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK: vec.epilog.iter.check:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/multi-exit-cost.ll b/llvm/test/Transforms/LoopVectorize/X86/multi-exit-cost.ll
index de6418066dea0..7b0b68b294f75 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/multi-exit-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/multi-exit-cost.ll
@@ -27,23 +27,30 @@ define i64 @test_value_in_exit_compare_chain_used_outside(ptr %src, i64 %x, i64
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[TMP10:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <8 x i8> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP29:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i8> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[VEC_PHI3:%.*]] = phi <4 x i8> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP19:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP10]], 1
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP18]]
; CHECK-NEXT: [[TMP27:%.*]] = getelementptr i8, ptr [[TMP26]], i32 0
-; CHECK-NEXT: [[TMP28:%.*]] = getelementptr i8, ptr [[TMP27]], i32 -7
-; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP28]], align 1
-; CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <8 x i8> [[WIDE_LOAD]], <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT: [[TMP29]] = xor <8 x i8> [[REVERSE]], [[VEC_PHI]]
+; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP27]], i32 -3
+; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[TMP26]], i32 -4
+; CHECK-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[TMP14]], i32 -3
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP13]], align 1
+; CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
+; CHECK-NEXT: [[REVERSE5:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD4]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: [[TMP17]] = xor <4 x i8> [[REVERSE]], [[VEC_PHI]]
+; CHECK-NEXT: [[TMP19]] = xor <4 x i8> [[REVERSE5]], [[VEC_PHI3]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[TMP10]], 8
; CHECK-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP30]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
-; CHECK-NEXT: [[TMP31:%.*]] = call i8 @llvm.vector.reduce.xor.v8i8(<8 x i8> [[TMP29]])
+; CHECK-NEXT: [[BIN_RDX:%.*]] = xor <4 x i8> [[TMP19]], [[TMP17]]
+; CHECK-NEXT: [[TMP20:%.*]] = call i8 @llvm.vector.reduce.xor.v4i8(<4 x i8> [[BIN_RDX]])
; CHECK-NEXT: br label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
-; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i8 [ [[TMP31]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i8 [ [[TMP20]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: br label %[[LOOP_HEADER:.*]]
; CHECK: [[LOOP_HEADER]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/reduction-small-size.ll b/llvm/test/Transforms/LoopVectorize/X86/reduction-small-size.ll
index 2cda2533e80e0..d905d925dc834 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/reduction-small-size.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/reduction-small-size.ll
@@ -30,8 +30,6 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
; CHECK: LV: Found an estimated cost of {{[0-9]+}} for VF 1 For instruction: br
; CHECK: Cost of 1 for VF 2: induction instruction %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
; CHECK: Cost of 1 for VF 2: induction instruction %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
-; CHECK: Cost of 1 for VF 2: exit condition instruction %exitcond = icmp eq i32 %lftr.wideiv, %n
-; CHECK: Cost of 0 for VF 2: exit condition instruction %lftr.wideiv = trunc i64 %indvars.iv.next to i32
; CHECK: Cost of 0 for VF 2: EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION ir<0>, vp<%index.next>
; CHECK: Cost of 1 for VF 2: WIDEN-REDUCTION-PHI ir<%sum.013> = phi vp<{{.+}}>, vp<[[EXT:%.+]]>
; CHECK: Cost of 0 for VF 2: vp<[[STEPS:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>
@@ -49,7 +47,7 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
; CHECK: Cost of 0 for VF 2: WIDEN-CAST vp<[[TRUNC:%.+]]> = trunc ir<%add5> to i8
; CHECK: Cost of 0 for VF 2: WIDEN-CAST vp<[[EXT]]> = zext vp<[[TRUNC]]> to i32
; CHECK: Cost of 0 for VF 2: EMIT vp<%index.next> = add nuw vp<[[CAN_IV]]>, vp<{{.+}}>
-; CHECK: Cost of 0 for VF 2: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
+; CHECK: Cost of 1 for VF 2: EMIT branch-on-count vp<%index.next>, vp<{{.+}}>
;
define i8 @reduction_i8(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %n) {
entry:
More information about the llvm-commits
mailing list