Skip to content

[NFC][LoopVectorize] Avoid passing ScalarEvolution to VPlanTransforms::optimize #108380

New issue

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

Sign up for GitHub

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

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 13, 2024
Merged

[NFC][LoopVectorize] Avoid passing ScalarEvolution to VPlanTransforms::optimize #108380

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8553,9 +8553,9 @@ void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,
if (auto Plan = tryToBuildVPlanWithVPRecipes(SubRange)) {
// Now optimize the initial VPlan.
if (!Plan->hasVF(ElementCount::getFixed(1)))
VPlanTransforms::truncateToMinimalBitwidths(
*Plan, CM.getMinimalBitwidths(), PSE.getSE()->getContext());
VPlanTransforms::optimize(*Plan, *PSE.getSE());
VPlanTransforms::truncateToMinimalBitwidths(*Plan,
CM.getMinimalBitwidths());
VPlanTransforms::optimize(*Plan);
// TODO: try to put it close to addActiveLaneMask().
// Discard the plan if it is not EVL-compatible
if (CM.foldTailWithEVL() &&
Expand Down
2 changes: 1 addition & 1 deletion llvm/lib/Transforms/Vectorize/VPlan.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1625,7 +1625,7 @@ void LoopVectorizationPlanner::buildVPlans(ElementCount MinVF,
for (ElementCount VF = MinVF; ElementCount::isKnownLT(VF, MaxVFTimes2);) {
VFRange SubRange = {VF, MaxVFTimes2};
auto Plan = buildVPlan(SubRange);
VPlanTransforms::optimize(*Plan, *PSE.getSE());
VPlanTransforms::optimize(*Plan);
VPlans.push_back(std::move(Plan));
VF = SubRange.End;
}
Expand Down
33 changes: 17 additions & 16 deletions llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -523,9 +523,8 @@ static void removeDeadRecipes(VPlan &Plan) {
static VPScalarIVStepsRecipe *
createScalarIVSteps(VPlan &Plan, InductionDescriptor::InductionKind Kind,
Instruction::BinaryOps InductionOpcode,
FPMathOperator *FPBinOp, ScalarEvolution &SE,
Instruction *TruncI, VPValue *StartV, VPValue *Step,
VPBasicBlock::iterator IP) {
FPMathOperator *FPBinOp, Instruction *TruncI,
VPValue *StartV, VPValue *Step, VPBasicBlock::iterator IP) {
VPBasicBlock *HeaderVPBB = Plan.getVectorLoopRegion()->getEntryBasicBlock();
VPCanonicalIVPHIRecipe *CanonicalIV = Plan.getCanonicalIV();
VPSingleDefRecipe *BaseIV = CanonicalIV;
Expand All @@ -535,8 +534,8 @@ createScalarIVSteps(VPlan &Plan, InductionDescriptor::InductionKind Kind,
}

// Truncate base induction if needed.
VPTypeAnalysis TypeInfo(Plan.getCanonicalIV()->getScalarType(),
SE.getContext());
Type *CanonicalIVType = CanonicalIV->getScalarType();
VPTypeAnalysis TypeInfo(CanonicalIVType, CanonicalIVType->getContext());
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Post-commit nit, while we're here: is a context other than that of CanonicalIVType relevant, or can TypeInfo's constructor derive its second parameter from its first.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

#108540

Type *ResultTy = TypeInfo.inferScalarType(BaseIV);
if (TruncI) {
Type *TruncTy = TruncI->getType();
Expand Down Expand Up @@ -576,7 +575,7 @@ createScalarIVSteps(VPlan &Plan, InductionDescriptor::InductionKind Kind,
/// if any of its users needs scalar values, by providing them scalar steps
/// built on the canonical scalar IV and update the original IV's users. This is
/// an optional optimization to reduce the needs of vector extracts.
static void legalizeAndOptimizeInductions(VPlan &Plan, ScalarEvolution &SE) {
static void legalizeAndOptimizeInductions(VPlan &Plan) {
SmallVector<VPRecipeBase *> ToRemove;
VPBasicBlock *HeaderVPBB = Plan.getVectorLoopRegion()->getEntryBasicBlock();
bool HasOnlyVectorVFs = !Plan.hasVF(ElementCount::getFixed(1));
Expand All @@ -594,7 +593,7 @@ static void legalizeAndOptimizeInductions(VPlan &Plan, ScalarEvolution &SE) {
VPValue *StepV = PtrIV->getOperand(1);
VPScalarIVStepsRecipe *Steps = createScalarIVSteps(
Plan, InductionDescriptor::IK_IntInduction, Instruction::Add, nullptr,
SE, nullptr, StartV, StepV, InsertPt);
nullptr, StartV, StepV, InsertPt);

auto *Recipe = new VPInstruction(VPInstruction::PtrAdd,
{PtrIV->getStartValue(), Steps},
Expand All @@ -618,7 +617,7 @@ static void legalizeAndOptimizeInductions(VPlan &Plan, ScalarEvolution &SE) {
const InductionDescriptor &ID = WideIV->getInductionDescriptor();
VPScalarIVStepsRecipe *Steps = createScalarIVSteps(
Plan, ID.getKind(), ID.getInductionOpcode(),
dyn_cast_or_null<FPMathOperator>(ID.getInductionBinOp()), SE,
dyn_cast_or_null<FPMathOperator>(ID.getInductionBinOp()),
WideIV->getTruncInst(), WideIV->getStartValue(), WideIV->getStepValue(),
InsertPt);

Expand Down Expand Up @@ -976,10 +975,11 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
}

/// Try to simplify the recipes in \p Plan.
static void simplifyRecipes(VPlan &Plan, LLVMContext &Ctx) {
static void simplifyRecipes(VPlan &Plan) {
ReversePostOrderTraversal<VPBlockDeepTraversalWrapper<VPBlockBase *>> RPOT(
Plan.getEntry());
VPTypeAnalysis TypeInfo(Plan.getCanonicalIV()->getScalarType(), Ctx);
Type *CanonicalIVType = Plan.getCanonicalIV()->getScalarType();
VPTypeAnalysis TypeInfo(CanonicalIVType, CanonicalIVType->getContext());
for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(RPOT)) {
for (VPRecipeBase &R : make_early_inc_range(*VPBB)) {
simplifyRecipe(R, TypeInfo);
Expand All @@ -988,8 +988,7 @@ static void simplifyRecipes(VPlan &Plan, LLVMContext &Ctx) {
}

void VPlanTransforms::truncateToMinimalBitwidths(
VPlan &Plan, const MapVector<Instruction *, uint64_t> &MinBWs,
LLVMContext &Ctx) {
VPlan &Plan, const MapVector<Instruction *, uint64_t> &MinBWs) {
#ifndef NDEBUG
// Count the processed recipes and cross check the count later with MinBWs
// size, to make sure all entries in MinBWs have been handled.
Expand All @@ -1000,7 +999,9 @@ void VPlanTransforms::truncateToMinimalBitwidths(
// other uses have different types for their operands, making them invalidly
// typed.
DenseMap<VPValue *, VPWidenCastRecipe *> ProcessedTruncs;
VPTypeAnalysis TypeInfo(Plan.getCanonicalIV()->getScalarType(), Ctx);
Type *CanonicalIVType = Plan.getCanonicalIV()->getScalarType();
LLVMContext &Ctx = CanonicalIVType->getContext();
VPTypeAnalysis TypeInfo(CanonicalIVType, Ctx);
VPBasicBlock *PH = Plan.getEntry();
for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(
vp_depth_first_deep(Plan.getVectorLoopRegion()))) {
Expand Down Expand Up @@ -1119,12 +1120,12 @@ void VPlanTransforms::truncateToMinimalBitwidths(
"some entries in MinBWs haven't been processed");
}

void VPlanTransforms::optimize(VPlan &Plan, ScalarEvolution &SE) {
void VPlanTransforms::optimize(VPlan &Plan) {
removeRedundantCanonicalIVs(Plan);
removeRedundantInductionCasts(Plan);

simplifyRecipes(Plan, SE.getContext());
legalizeAndOptimizeInductions(Plan, SE);
simplifyRecipes(Plan);
legalizeAndOptimizeInductions(Plan);
removeDeadRecipes(Plan);

createAndOptimizeReplicateRegions(Plan);
Expand Down
5 changes: 2 additions & 3 deletions llvm/lib/Transforms/Vectorize/VPlanTransforms.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ struct VPlanTransforms {
/// Apply VPlan-to-VPlan optimizations to \p Plan, including induction recipe
/// optimizations, dead recipe removal, replicate region optimizations and
/// block merging.
static void optimize(VPlan &Plan, ScalarEvolution &SE);
static void optimize(VPlan &Plan);

/// Wrap predicated VPReplicateRecipes with a mask operand in an if-then
/// region block and remove the mask operand. Optimize the created regions by
Expand All @@ -82,8 +82,7 @@ struct VPlanTransforms {
/// will be folded later.
static void
truncateToMinimalBitwidths(VPlan &Plan,
const MapVector<Instruction *, uint64_t> &MinBWs,
LLVMContext &Ctx);
const MapVector<Instruction *, uint64_t> &MinBWs);

/// Drop poison flags from recipes that may generate a poison value that is
/// used after vectorization, even when their operands are not poison. Those
Expand Down
Loading