llvm-mirror/lib/Analysis/LoopUnrollAnalyzer.cpp
Michael Zolotukhin afd08c7313 [Unroll] Implement a conservative and monotonically increasing cost tracking system during the full unroll heuristic analysis that avoids counting any instruction cost until that instruction becomes "live" through a side-effect or use outside the...
Summary:
...loop after the last iteration.

This is really hard to do correctly. The core problem is that we need to
model liveness through the induction PHIs from iteration to iteration in
order to get the correct results, and we need to correctly de-duplicate
the common subgraphs of instructions feeding some subset of the
induction PHIs. All of this can be driven either from a side effect at
some iteration or from the loop values used after the loop finishes.

This patch implements this by storing the forward-propagating analysis
of each instruction in a cache to recall whether it was free and whether
it has become live and thus counted toward the total unroll cost. Then,
at each sink for a value in the loop, we recursively walk back through
every value that feeds the sink, including looping back through the
iterations as needed, until we have marked the entire input graph as
live. Because we cache this, we never visit instructions more than twice
-- once when we analyze them and put them into the cache, and once when
we count their cost towards the unrolled loop. Also, because the cache
is only two bits and because we are dealing with relatively small
iteration counts, we can store all of this very densely in memory to
avoid this from becoming an excessively slow analysis.

The code here is still pretty gross. I would appreciate suggestions
about better ways to factor or split this up, I've stared too long at
the algorithmic side to really have a good sense of what the design
should probably look at.

Also, it might seem like we should do all of this bottom-up, but I think
that is a red herring. Specifically, the simplification power is *much*
greater working top-down. We can forward propagate very effectively,
even across strange and interesting recurrances around the backedge.
Because we use data to propagate, this doesn't cause a state space
explosion. Doing this level of constant folding, etc, would be very
expensive to do bottom-up because it wouldn't be until the last moment
that you could collapse everything. The current solution is essentially
a top-down simplification with a bottom-up cost accounting which seems
to get the best of both worlds. It makes the simplification incremental
and powerful while leaving everything dead until we *know* it is needed.

Finally, a core property of this approach is its *monotonicity*. At all
times, the current UnrolledCost is a conservatively low estimate. This
ensures that we will never early-exit from the analysis due to exceeding
a threshold when if we had continued, the cost would have gone back
below the threshold. These kinds of bugs can cause incredibly hard to
track down random changes to behavior.

We could use a techinque similar (but much simpler) within the inliner
as well to avoid considering speculated code in the inline cost.

Reviewers: chandlerc

Subscribers: sanjoy, mzolotukhin, llvm-commits

Differential Revision: http://reviews.llvm.org/D11758

llvm-svn: 269388
2016-05-13 01:42:39 +00:00

202 lines
6.6 KiB
C++

//===- LoopUnrollAnalyzer.cpp - Unrolling Effect Estimation -----*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements UnrolledInstAnalyzer class. It's used for predicting
// potential effects that loop unrolling might have, such as enabling constant
// propagation and other optimizations.
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/LoopUnrollAnalyzer.h"
#include "llvm/IR/Dominators.h"
using namespace llvm;
/// \brief Try to simplify instruction \param I using its SCEV expression.
///
/// The idea is that some AddRec expressions become constants, which then
/// could trigger folding of other instructions. However, that only happens
/// for expressions whose start value is also constant, which isn't always the
/// case. In another common and important case the start value is just some
/// address (i.e. SCEVUnknown) - in this case we compute the offset and save
/// it along with the base address instead.
bool UnrolledInstAnalyzer::simplifyInstWithSCEV(Instruction *I) {
if (!SE.isSCEVable(I->getType()))
return false;
const SCEV *S = SE.getSCEV(I);
if (auto *SC = dyn_cast<SCEVConstant>(S)) {
SimplifiedValues[I] = SC->getValue();
return true;
}
auto *AR = dyn_cast<SCEVAddRecExpr>(S);
if (!AR || AR->getLoop() != L)
return false;
const SCEV *ValueAtIteration = AR->evaluateAtIteration(IterationNumber, SE);
// Check if the AddRec expression becomes a constant.
if (auto *SC = dyn_cast<SCEVConstant>(ValueAtIteration)) {
SimplifiedValues[I] = SC->getValue();
return true;
}
// Check if the offset from the base address becomes a constant.
auto *Base = dyn_cast<SCEVUnknown>(SE.getPointerBase(S));
if (!Base)
return false;
auto *Offset =
dyn_cast<SCEVConstant>(SE.getMinusSCEV(ValueAtIteration, Base));
if (!Offset)
return false;
SimplifiedAddress Address;
Address.Base = Base->getValue();
Address.Offset = Offset->getValue();
SimplifiedAddresses[I] = Address;
return false;
}
/// Try to simplify binary operator I.
///
/// TODO: Probably it's worth to hoist the code for estimating the
/// simplifications effects to a separate class, since we have a very similar
/// code in InlineCost already.
bool UnrolledInstAnalyzer::visitBinaryOperator(BinaryOperator &I) {
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
if (!isa<Constant>(LHS))
if (Constant *SimpleLHS = SimplifiedValues.lookup(LHS))
LHS = SimpleLHS;
if (!isa<Constant>(RHS))
if (Constant *SimpleRHS = SimplifiedValues.lookup(RHS))
RHS = SimpleRHS;
Value *SimpleV = nullptr;
const DataLayout &DL = I.getModule()->getDataLayout();
if (auto FI = dyn_cast<FPMathOperator>(&I))
SimpleV =
SimplifyFPBinOp(I.getOpcode(), LHS, RHS, FI->getFastMathFlags(), DL);
else
SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS, DL);
if (Constant *C = dyn_cast_or_null<Constant>(SimpleV))
SimplifiedValues[&I] = C;
if (SimpleV)
return true;
return Base::visitBinaryOperator(I);
}
/// Try to fold load I.
bool UnrolledInstAnalyzer::visitLoad(LoadInst &I) {
Value *AddrOp = I.getPointerOperand();
auto AddressIt = SimplifiedAddresses.find(AddrOp);
if (AddressIt == SimplifiedAddresses.end())
return false;
ConstantInt *SimplifiedAddrOp = AddressIt->second.Offset;
auto *GV = dyn_cast<GlobalVariable>(AddressIt->second.Base);
// We're only interested in loads that can be completely folded to a
// constant.
if (!GV || !GV->hasDefinitiveInitializer() || !GV->isConstant())
return false;
ConstantDataSequential *CDS =
dyn_cast<ConstantDataSequential>(GV->getInitializer());
if (!CDS)
return false;
// We might have a vector load from an array. FIXME: for now we just bail
// out in this case, but we should be able to resolve and simplify such
// loads.
if(!CDS->isElementTypeCompatible(I.getType()))
return false;
int ElemSize = CDS->getElementType()->getPrimitiveSizeInBits() / 8U;
assert(SimplifiedAddrOp->getValue().getActiveBits() < 64 &&
"Unexpectedly large index value.");
int64_t Index = SimplifiedAddrOp->getSExtValue() / ElemSize;
if (Index >= CDS->getNumElements()) {
// FIXME: For now we conservatively ignore out of bound accesses, but
// we're allowed to perform the optimization in this case.
return false;
}
Constant *CV = CDS->getElementAsConstant(Index);
assert(CV && "Constant expected.");
SimplifiedValues[&I] = CV;
return true;
}
/// Try to simplify cast instruction.
bool UnrolledInstAnalyzer::visitCastInst(CastInst &I) {
// Propagate constants through casts.
Constant *COp = dyn_cast<Constant>(I.getOperand(0));
if (!COp)
COp = SimplifiedValues.lookup(I.getOperand(0));
if (COp)
if (Constant *C =
ConstantExpr::getCast(I.getOpcode(), COp, I.getType())) {
SimplifiedValues[&I] = C;
return true;
}
return Base::visitCastInst(I);
}
/// Try to simplify cmp instruction.
bool UnrolledInstAnalyzer::visitCmpInst(CmpInst &I) {
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
// First try to handle simplified comparisons.
if (!isa<Constant>(LHS))
if (Constant *SimpleLHS = SimplifiedValues.lookup(LHS))
LHS = SimpleLHS;
if (!isa<Constant>(RHS))
if (Constant *SimpleRHS = SimplifiedValues.lookup(RHS))
RHS = SimpleRHS;
if (!isa<Constant>(LHS) && !isa<Constant>(RHS)) {
auto SimplifiedLHS = SimplifiedAddresses.find(LHS);
if (SimplifiedLHS != SimplifiedAddresses.end()) {
auto SimplifiedRHS = SimplifiedAddresses.find(RHS);
if (SimplifiedRHS != SimplifiedAddresses.end()) {
SimplifiedAddress &LHSAddr = SimplifiedLHS->second;
SimplifiedAddress &RHSAddr = SimplifiedRHS->second;
if (LHSAddr.Base == RHSAddr.Base) {
LHS = LHSAddr.Offset;
RHS = RHSAddr.Offset;
}
}
}
}
if (Constant *CLHS = dyn_cast<Constant>(LHS)) {
if (Constant *CRHS = dyn_cast<Constant>(RHS)) {
if (Constant *C = ConstantExpr::getCompare(I.getPredicate(), CLHS, CRHS)) {
SimplifiedValues[&I] = C;
return true;
}
}
}
return Base::visitCmpInst(I);
}
bool UnrolledInstAnalyzer::visitPHINode(PHINode &PN) {
// Run base visitor first. This way we can gather some useful for later
// analysis information.
if (Base::visitPHINode(PN))
return true;
// The loop induction PHI nodes are definitionally free.
return PN.getParent() == L->getHeader();
}