Kerry McLaughlin f0e9bba024
[LoopVectorize] Generate wide active lane masks (#147535)
This patch adds a new flag (-enable-wide-lane-mask) which allows
LoopVectorize to generate wider-than-VF active lane masks when it
is safe to do so (i.e. the mask is used for data and control flow).

The transform in extractFromWideActiveLaneMask creates vector
extracts from the first active lane mask in the header & loop body,
modifying the active lane mask phi operands to use the extracts.

An additional operand is passed to the ActiveLaneMask instruction,
the value of which is used as a multiplier of VF when generating the
mask.
By default this is 1, and is updated to UF by
extractFromWideActiveLaneMask.

The motivation for this change is to improve interleaved loops when
SVE2.1 is available, where we can make use of the whilelo instruction
which returns a predicate pair.

This is based on a PR that was created by @momchil-velikov (#81140)
and contains tests which were added there.
2025-09-01 13:53:30 +01:00

145 lines
5.4 KiB
C++

//===- VPlanUtils.cpp - VPlan-related utilities ---------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "VPlanUtils.h"
#include "VPlanCFG.h"
#include "VPlanPatternMatch.h"
#include "llvm/ADT/TypeSwitch.h"
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
using namespace llvm;
bool vputils::onlyFirstLaneUsed(const VPValue *Def) {
return all_of(Def->users(),
[Def](const VPUser *U) { return U->onlyFirstLaneUsed(Def); });
}
bool vputils::onlyFirstPartUsed(const VPValue *Def) {
return all_of(Def->users(),
[Def](const VPUser *U) { return U->onlyFirstPartUsed(Def); });
}
bool vputils::onlyScalarValuesUsed(const VPValue *Def) {
return all_of(Def->users(),
[Def](const VPUser *U) { return U->usesScalars(Def); });
}
VPValue *vputils::getOrCreateVPValueForSCEVExpr(VPlan &Plan, const SCEV *Expr) {
if (auto *Expanded = Plan.getSCEVExpansion(Expr))
return Expanded;
VPValue *Expanded = nullptr;
if (auto *E = dyn_cast<SCEVConstant>(Expr))
Expanded = Plan.getOrAddLiveIn(E->getValue());
else {
auto *U = dyn_cast<SCEVUnknown>(Expr);
// Skip SCEV expansion if Expr is a SCEVUnknown wrapping a non-instruction
// value. Otherwise the value may be defined in a loop and using it directly
// will break LCSSA form. The SCEV expansion takes care of preserving LCSSA
// form.
if (U && !isa<Instruction>(U->getValue())) {
Expanded = Plan.getOrAddLiveIn(U->getValue());
} else {
Expanded = new VPExpandSCEVRecipe(Expr);
Plan.getEntry()->appendRecipe(Expanded->getDefiningRecipe());
}
}
Plan.addSCEVExpansion(Expr, Expanded);
return Expanded;
}
bool vputils::isHeaderMask(const VPValue *V, VPlan &Plan) {
if (isa<VPActiveLaneMaskPHIRecipe>(V))
return true;
auto IsWideCanonicalIV = [](VPValue *A) {
return isa<VPWidenCanonicalIVRecipe>(A) ||
(isa<VPWidenIntOrFpInductionRecipe>(A) &&
cast<VPWidenIntOrFpInductionRecipe>(A)->isCanonical());
};
VPValue *A, *B;
using namespace VPlanPatternMatch;
if (match(V, m_ActiveLaneMask(m_VPValue(A), m_VPValue(B), m_SpecificInt(1))))
return B == Plan.getTripCount() &&
(match(A, m_ScalarIVSteps(m_Specific(Plan.getCanonicalIV()),
m_SpecificInt(1),
m_Specific(&Plan.getVF()))) ||
IsWideCanonicalIV(A));
return match(V, m_Binary<Instruction::ICmp>(m_VPValue(A), m_VPValue(B))) &&
IsWideCanonicalIV(A) && B == Plan.getOrCreateBackedgeTakenCount();
}
const SCEV *vputils::getSCEVExprForVPValue(VPValue *V, ScalarEvolution &SE) {
if (V->isLiveIn()) {
if (Value *LiveIn = V->getLiveInIRValue())
return SE.getSCEV(LiveIn);
return SE.getCouldNotCompute();
}
// TODO: Support constructing SCEVs for more recipes as needed.
return TypeSwitch<const VPRecipeBase *, const SCEV *>(V->getDefiningRecipe())
.Case<VPExpandSCEVRecipe>(
[](const VPExpandSCEVRecipe *R) { return R->getSCEV(); })
.Default([&SE](const VPRecipeBase *) { return SE.getCouldNotCompute(); });
}
bool vputils::isUniformAcrossVFsAndUFs(VPValue *V) {
using namespace VPlanPatternMatch;
// Live-ins are uniform.
if (V->isLiveIn())
return true;
VPRecipeBase *R = V->getDefiningRecipe();
if (R && V->isDefinedOutsideLoopRegions()) {
if (match(V->getDefiningRecipe(),
m_VPInstruction<VPInstruction::CanonicalIVIncrementForPart>(
m_VPValue())))
return false;
return all_of(R->operands(), isUniformAcrossVFsAndUFs);
}
auto *CanonicalIV = R->getParent()->getPlan()->getCanonicalIV();
// Canonical IV chain is uniform.
if (V == CanonicalIV || V == CanonicalIV->getBackedgeValue())
return true;
return TypeSwitch<const VPRecipeBase *, bool>(R)
.Case<VPDerivedIVRecipe>([](const auto *R) { return true; })
.Case<VPReplicateRecipe>([](const auto *R) {
// Loads and stores that are uniform across VF lanes are handled by
// VPReplicateRecipe.IsUniform. They are also uniform across UF parts if
// all their operands are invariant.
// TODO: Further relax the restrictions.
return R->isSingleScalar() &&
(isa<LoadInst, StoreInst>(R->getUnderlyingValue())) &&
all_of(R->operands(), isUniformAcrossVFsAndUFs);
})
.Case<VPInstruction>([](const auto *VPI) {
return VPI->isScalarCast() &&
isUniformAcrossVFsAndUFs(VPI->getOperand(0));
})
.Case<VPWidenCastRecipe>([](const auto *R) {
// A cast is uniform according to its operand.
return isUniformAcrossVFsAndUFs(R->getOperand(0));
})
.Default([](const VPRecipeBase *) { // A value is considered non-uniform
// unless proven otherwise.
return false;
});
}
VPBasicBlock *vputils::getFirstLoopHeader(VPlan &Plan, VPDominatorTree &VPDT) {
auto DepthFirst = vp_depth_first_shallow(Plan.getEntry());
auto I = find_if(DepthFirst, [&VPDT](VPBlockBase *VPB) {
return VPBlockUtils::isHeader(VPB, VPDT);
});
return I == DepthFirst.end() ? nullptr : cast<VPBasicBlock>(*I);
}