mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-14 11:32:34 +00:00
revert r166264 because the LTO build is still failing
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@166340 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
parent
e06ce4c2c4
commit
a04a4a79ea
@ -22,7 +22,7 @@
|
||||
#include <set>
|
||||
|
||||
namespace llvm {
|
||||
class ScalarTargetTransformInfo;
|
||||
class TargetLowering;
|
||||
|
||||
/// Return true if the given expression is safe to expand in the sense that
|
||||
/// all materialized values are safe to speculate.
|
||||
@ -129,7 +129,7 @@ namespace llvm {
|
||||
/// representative. Return the number of phis eliminated.
|
||||
unsigned replaceCongruentIVs(Loop *L, const DominatorTree *DT,
|
||||
SmallVectorImpl<WeakVH> &DeadInsts,
|
||||
const ScalarTargetTransformInfo *STTI = NULL);
|
||||
const TargetLowering *TLI = NULL);
|
||||
|
||||
/// expandCodeFor - Insert code to directly compute the specified SCEV
|
||||
/// expression into the program. The inserted code is inserted into the
|
||||
|
@ -119,7 +119,7 @@ Pass *createLICMPass();
|
||||
// optional parameter used to consult the target machine whether certain
|
||||
// transformations are profitable.
|
||||
//
|
||||
Pass *createLoopStrengthReducePass();
|
||||
Pass *createLoopStrengthReducePass(const TargetLowering *TLI = 0);
|
||||
|
||||
Pass *createGlobalMergePass(const TargetLowering *TLI = 0);
|
||||
|
||||
@ -249,8 +249,9 @@ extern char &LowerSwitchID;
|
||||
// purpose "my LLVM-to-LLVM pass doesn't support the invoke instruction yet"
|
||||
// lowering pass.
|
||||
//
|
||||
FunctionPass *createLowerInvokePass();
|
||||
FunctionPass *createLowerInvokePass(bool useExpensiveEHSupport);
|
||||
FunctionPass *createLowerInvokePass(const TargetLowering *TLI = 0);
|
||||
FunctionPass *createLowerInvokePass(const TargetLowering *TLI,
|
||||
bool useExpensiveEHSupport);
|
||||
extern char &LowerInvokePassID;
|
||||
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
@ -19,8 +19,8 @@
|
||||
#include "llvm/LLVMContext.h"
|
||||
#include "llvm/Support/Debug.h"
|
||||
#include "llvm/DataLayout.h"
|
||||
#include "llvm/Target/TargetLowering.h"
|
||||
#include "llvm/ADT/STLExtras.h"
|
||||
#include "llvm/TargetTransformInfo.h"
|
||||
|
||||
using namespace llvm;
|
||||
|
||||
@ -1599,15 +1599,15 @@ static bool width_descending(Value *lhs, Value *rhs) {
|
||||
/// This does not depend on any SCEVExpander state but should be used in
|
||||
/// the same context that SCEVExpander is used.
|
||||
unsigned SCEVExpander::replaceCongruentIVs(Loop *L, const DominatorTree *DT,
|
||||
SmallVectorImpl<WeakVH> &DeadInsts,
|
||||
const ScalarTargetTransformInfo *STTI) {
|
||||
SmallVectorImpl<WeakVH> &DeadInsts,
|
||||
const TargetLowering *TLI) {
|
||||
// Find integer phis in order of increasing width.
|
||||
SmallVector<PHINode*, 8> Phis;
|
||||
for (BasicBlock::iterator I = L->getHeader()->begin();
|
||||
PHINode *Phi = dyn_cast<PHINode>(I); ++I) {
|
||||
Phis.push_back(Phi);
|
||||
}
|
||||
if (STTI)
|
||||
if (TLI)
|
||||
std::sort(Phis.begin(), Phis.end(), width_descending);
|
||||
|
||||
unsigned NumElim = 0;
|
||||
@ -1635,8 +1635,8 @@ unsigned SCEVExpander::replaceCongruentIVs(Loop *L, const DominatorTree *DT,
|
||||
PHINode *&OrigPhiRef = ExprToIVMap[SE.getSCEV(Phi)];
|
||||
if (!OrigPhiRef) {
|
||||
OrigPhiRef = Phi;
|
||||
if (Phi->getType()->isIntegerTy() && STTI &&
|
||||
STTI->isTruncateFree(Phi->getType(), Phis.back()->getType())) {
|
||||
if (Phi->getType()->isIntegerTy() && TLI
|
||||
&& TLI->isTruncateFree(Phi->getType(), Phis.back()->getType())) {
|
||||
// This phi can be freely truncated to the narrowest phi type. Map the
|
||||
// truncated expression to it so it will be reused for narrow types.
|
||||
const SCEV *TruncExpr =
|
||||
|
@ -359,7 +359,7 @@ void TargetPassConfig::addIRPasses() {
|
||||
|
||||
// Run loop strength reduction before anything else.
|
||||
if (getOptLevel() != CodeGenOpt::None && !DisableLSR) {
|
||||
addPass(createLoopStrengthReducePass());
|
||||
addPass(createLoopStrengthReducePass(getTargetLowering()));
|
||||
if (PrintLSR)
|
||||
addPass(createPrintFunctionPass("\n\n*** Code after LSR ***\n", &dbgs()));
|
||||
}
|
||||
@ -389,7 +389,7 @@ void TargetPassConfig::addPassesToHandleExceptions() {
|
||||
addPass(createDwarfEHPass(TM));
|
||||
break;
|
||||
case ExceptionHandling::None:
|
||||
addPass(createLowerInvokePass());
|
||||
addPass(createLowerInvokePass(TM->getTargetLowering()));
|
||||
|
||||
// The lower invoke pass may create unreachable code. Remove it.
|
||||
addPass(createUnreachableBlockEliminationPass());
|
||||
|
@ -37,7 +37,7 @@
|
||||
//
|
||||
// TODO: Handle multiple loops at a time.
|
||||
//
|
||||
// TODO: Should AddrMode::BaseGV be changed to a ConstantExpr
|
||||
// TODO: Should TargetLowering::AddrMode::BaseGV be changed to a ConstantExpr
|
||||
// instead of a GlobalValue?
|
||||
//
|
||||
// TODO: When truncation is free, truncate ICmp users' operands to make it a
|
||||
@ -67,7 +67,6 @@
|
||||
#include "llvm/Transforms/Scalar.h"
|
||||
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
|
||||
#include "llvm/Transforms/Utils/Local.h"
|
||||
#include "llvm/TargetTransformInfo.h"
|
||||
#include "llvm/ADT/SmallBitVector.h"
|
||||
#include "llvm/ADT/SetVector.h"
|
||||
#include "llvm/ADT/DenseSet.h"
|
||||
@ -75,6 +74,7 @@
|
||||
#include "llvm/Support/CommandLine.h"
|
||||
#include "llvm/Support/ValueHandle.h"
|
||||
#include "llvm/Support/raw_ostream.h"
|
||||
#include "llvm/Target/TargetLowering.h"
|
||||
#include <algorithm>
|
||||
using namespace llvm;
|
||||
|
||||
@ -1118,7 +1118,7 @@ public:
|
||||
enum KindType {
|
||||
Basic, ///< A normal use, with no folding.
|
||||
Special, ///< A special case of basic, allowing -1 scales.
|
||||
Address, ///< An address use; folding according to ScalarTargetTransformInfo.
|
||||
Address, ///< An address use; folding according to TargetLowering
|
||||
ICmpZero ///< An equality icmp with both operands folded into one.
|
||||
// TODO: Add a generic icmp too?
|
||||
};
|
||||
@ -1272,12 +1272,12 @@ void LSRUse::dump() const {
|
||||
/// address-mode folding and special icmp tricks.
|
||||
static bool isLegalUse(const AddrMode &AM,
|
||||
LSRUse::KindType Kind, Type *AccessTy,
|
||||
const ScalarTargetTransformInfo *STTI) {
|
||||
const TargetLowering *TLI) {
|
||||
switch (Kind) {
|
||||
case LSRUse::Address:
|
||||
// If we have low-level target information, ask the target if it can
|
||||
// completely fold this address.
|
||||
if (STTI) return STTI->isLegalAddressingMode(AM, AccessTy);
|
||||
if (TLI) return TLI->isLegalAddressingMode(AM, AccessTy);
|
||||
|
||||
// Otherwise, just guess that reg+reg addressing is legal.
|
||||
return !AM.BaseGV && AM.BaseOffs == 0 && AM.Scale <= 1;
|
||||
@ -1300,7 +1300,7 @@ static bool isLegalUse(const AddrMode &AM,
|
||||
// If we have low-level target information, ask the target if it can fold an
|
||||
// integer immediate on an icmp.
|
||||
if (AM.BaseOffs != 0) {
|
||||
if (!STTI)
|
||||
if (!TLI)
|
||||
return false;
|
||||
// We have one of:
|
||||
// ICmpZero BaseReg + Offset => ICmp BaseReg, -Offset
|
||||
@ -1309,7 +1309,7 @@ static bool isLegalUse(const AddrMode &AM,
|
||||
int64_t Offs = AM.BaseOffs;
|
||||
if (AM.Scale == 0)
|
||||
Offs = -(uint64_t)Offs; // The cast does the right thing with INT64_MIN.
|
||||
return STTI->isLegalICmpImmediate(Offs);
|
||||
return TLI->isLegalICmpImmediate(Offs);
|
||||
}
|
||||
|
||||
// ICmpZero BaseReg + -1*ScaleReg => ICmp BaseReg, ScaleReg
|
||||
@ -1330,20 +1330,20 @@ static bool isLegalUse(const AddrMode &AM,
|
||||
static bool isLegalUse(AddrMode AM,
|
||||
int64_t MinOffset, int64_t MaxOffset,
|
||||
LSRUse::KindType Kind, Type *AccessTy,
|
||||
const ScalarTargetTransformInfo *LTTI) {
|
||||
const TargetLowering *TLI) {
|
||||
// Check for overflow.
|
||||
if (((int64_t)((uint64_t)AM.BaseOffs + MinOffset) > AM.BaseOffs) !=
|
||||
(MinOffset > 0))
|
||||
return false;
|
||||
AM.BaseOffs = (uint64_t)AM.BaseOffs + MinOffset;
|
||||
if (isLegalUse(AM, Kind, AccessTy, LTTI)) {
|
||||
if (isLegalUse(AM, Kind, AccessTy, TLI)) {
|
||||
AM.BaseOffs = (uint64_t)AM.BaseOffs - MinOffset;
|
||||
// Check for overflow.
|
||||
if (((int64_t)((uint64_t)AM.BaseOffs + MaxOffset) > AM.BaseOffs) !=
|
||||
(MaxOffset > 0))
|
||||
return false;
|
||||
AM.BaseOffs = (uint64_t)AM.BaseOffs + MaxOffset;
|
||||
return isLegalUse(AM, Kind, AccessTy, LTTI);
|
||||
return isLegalUse(AM, Kind, AccessTy, TLI);
|
||||
}
|
||||
return false;
|
||||
}
|
||||
@ -1352,7 +1352,7 @@ static bool isAlwaysFoldable(int64_t BaseOffs,
|
||||
GlobalValue *BaseGV,
|
||||
bool HasBaseReg,
|
||||
LSRUse::KindType Kind, Type *AccessTy,
|
||||
const ScalarTargetTransformInfo *LTTI) {
|
||||
const TargetLowering *TLI) {
|
||||
// Fast-path: zero is always foldable.
|
||||
if (BaseOffs == 0 && !BaseGV) return true;
|
||||
|
||||
@ -1371,14 +1371,14 @@ static bool isAlwaysFoldable(int64_t BaseOffs,
|
||||
AM.HasBaseReg = true;
|
||||
}
|
||||
|
||||
return isLegalUse(AM, Kind, AccessTy, LTTI);
|
||||
return isLegalUse(AM, Kind, AccessTy, TLI);
|
||||
}
|
||||
|
||||
static bool isAlwaysFoldable(const SCEV *S,
|
||||
int64_t MinOffset, int64_t MaxOffset,
|
||||
bool HasBaseReg,
|
||||
LSRUse::KindType Kind, Type *AccessTy,
|
||||
const ScalarTargetTransformInfo *LTTI,
|
||||
const TargetLowering *TLI,
|
||||
ScalarEvolution &SE) {
|
||||
// Fast-path: zero is always foldable.
|
||||
if (S->isZero()) return true;
|
||||
@ -1402,7 +1402,7 @@ static bool isAlwaysFoldable(const SCEV *S,
|
||||
AM.HasBaseReg = HasBaseReg;
|
||||
AM.Scale = Kind == LSRUse::ICmpZero ? -1 : 1;
|
||||
|
||||
return isLegalUse(AM, MinOffset, MaxOffset, Kind, AccessTy, LTTI);
|
||||
return isLegalUse(AM, MinOffset, MaxOffset, Kind, AccessTy, TLI);
|
||||
}
|
||||
|
||||
namespace {
|
||||
@ -1502,7 +1502,7 @@ class LSRInstance {
|
||||
ScalarEvolution &SE;
|
||||
DominatorTree &DT;
|
||||
LoopInfo &LI;
|
||||
const ScalarTargetTransformInfo *const STTI;
|
||||
const TargetLowering *const TLI;
|
||||
Loop *const L;
|
||||
bool Changed;
|
||||
|
||||
@ -1638,7 +1638,7 @@ class LSRInstance {
|
||||
Pass *P);
|
||||
|
||||
public:
|
||||
LSRInstance(const ScalarTargetTransformInfo *ltti, Loop *l, Pass *P);
|
||||
LSRInstance(const TargetLowering *tli, Loop *l, Pass *P);
|
||||
|
||||
bool getChanged() const { return Changed; }
|
||||
|
||||
@ -1688,10 +1688,11 @@ void LSRInstance::OptimizeShadowIV() {
|
||||
}
|
||||
if (!DestTy) continue;
|
||||
|
||||
if (STTI) {
|
||||
if (TLI) {
|
||||
// If target does not support DestTy natively then do not apply
|
||||
// this transformation.
|
||||
if (!STTI->isTypeLegal(DestTy)) continue;
|
||||
EVT DVT = TLI->getValueType(DestTy);
|
||||
if (!TLI->isTypeLegal(DVT)) continue;
|
||||
}
|
||||
|
||||
PHINode *PH = dyn_cast<PHINode>(ShadowUse->getOperand(0));
|
||||
@ -2014,18 +2015,18 @@ LSRInstance::OptimizeLoopTermCond() {
|
||||
if (C->getValue().getMinSignedBits() >= 64 ||
|
||||
C->getValue().isMinSignedValue())
|
||||
goto decline_post_inc;
|
||||
// Without STTI, assume that any stride might be valid, and so any
|
||||
// Without TLI, assume that any stride might be valid, and so any
|
||||
// use might be shared.
|
||||
if (!STTI)
|
||||
if (!TLI)
|
||||
goto decline_post_inc;
|
||||
// Check for possible scaled-address reuse.
|
||||
Type *AccessTy = getAccessType(UI->getUser());
|
||||
AddrMode AM;
|
||||
AM.Scale = C->getSExtValue();
|
||||
if (STTI->isLegalAddressingMode(AM, AccessTy))
|
||||
if (TLI->isLegalAddressingMode(AM, AccessTy))
|
||||
goto decline_post_inc;
|
||||
AM.Scale = -AM.Scale;
|
||||
if (STTI->isLegalAddressingMode(AM, AccessTy))
|
||||
if (TLI->isLegalAddressingMode(AM, AccessTy))
|
||||
goto decline_post_inc;
|
||||
}
|
||||
}
|
||||
@ -2096,12 +2097,12 @@ LSRInstance::reconcileNewOffset(LSRUse &LU, int64_t NewOffset, bool HasBaseReg,
|
||||
// Conservatively assume HasBaseReg is true for now.
|
||||
if (NewOffset < LU.MinOffset) {
|
||||
if (!isAlwaysFoldable(LU.MaxOffset - NewOffset, 0, HasBaseReg,
|
||||
Kind, AccessTy, STTI))
|
||||
Kind, AccessTy, TLI))
|
||||
return false;
|
||||
NewMinOffset = NewOffset;
|
||||
} else if (NewOffset > LU.MaxOffset) {
|
||||
if (!isAlwaysFoldable(NewOffset - LU.MinOffset, 0, HasBaseReg,
|
||||
Kind, AccessTy, STTI))
|
||||
Kind, AccessTy, TLI))
|
||||
return false;
|
||||
NewMaxOffset = NewOffset;
|
||||
}
|
||||
@ -2130,7 +2131,7 @@ LSRInstance::getUse(const SCEV *&Expr,
|
||||
int64_t Offset = ExtractImmediate(Expr, SE);
|
||||
|
||||
// Basic uses can't accept any offset, for example.
|
||||
if (!isAlwaysFoldable(Offset, 0, /*HasBaseReg=*/true, Kind, AccessTy, STTI)) {
|
||||
if (!isAlwaysFoldable(Offset, 0, /*HasBaseReg=*/true, Kind, AccessTy, TLI)) {
|
||||
Expr = Copy;
|
||||
Offset = 0;
|
||||
}
|
||||
@ -2395,7 +2396,7 @@ bool IVChain::isProfitableIncrement(const SCEV *OperExpr,
|
||||
/// TODO: Consider IVInc free if it's already used in another chains.
|
||||
static bool
|
||||
isProfitableChain(IVChain &Chain, SmallPtrSet<Instruction*, 4> &Users,
|
||||
ScalarEvolution &SE, const ScalarTargetTransformInfo *STTI) {
|
||||
ScalarEvolution &SE, const TargetLowering *TLI) {
|
||||
if (StressIVChain)
|
||||
return true;
|
||||
|
||||
@ -2653,7 +2654,7 @@ void LSRInstance::CollectChains() {
|
||||
for (unsigned UsersIdx = 0, NChains = IVChainVec.size();
|
||||
UsersIdx < NChains; ++UsersIdx) {
|
||||
if (!isProfitableChain(IVChainVec[UsersIdx],
|
||||
ChainUsersVec[UsersIdx].FarUsers, SE, STTI))
|
||||
ChainUsersVec[UsersIdx].FarUsers, SE, TLI))
|
||||
continue;
|
||||
// Preserve the chain at UsesIdx.
|
||||
if (ChainIdx != UsersIdx)
|
||||
@ -2680,8 +2681,7 @@ void LSRInstance::FinalizeChain(IVChain &Chain) {
|
||||
|
||||
/// Return true if the IVInc can be folded into an addressing mode.
|
||||
static bool canFoldIVIncExpr(const SCEV *IncExpr, Instruction *UserInst,
|
||||
Value *Operand,
|
||||
const ScalarTargetTransformInfo *STTI) {
|
||||
Value *Operand, const TargetLowering *TLI) {
|
||||
const SCEVConstant *IncConst = dyn_cast<SCEVConstant>(IncExpr);
|
||||
if (!IncConst || !isAddressUse(UserInst, Operand))
|
||||
return false;
|
||||
@ -2691,7 +2691,7 @@ static bool canFoldIVIncExpr(const SCEV *IncExpr, Instruction *UserInst,
|
||||
|
||||
int64_t IncOffset = IncConst->getValue()->getSExtValue();
|
||||
if (!isAlwaysFoldable(IncOffset, /*BaseGV=*/0, /*HaseBaseReg=*/false,
|
||||
LSRUse::Address, getAccessType(UserInst), STTI))
|
||||
LSRUse::Address, getAccessType(UserInst), TLI))
|
||||
return false;
|
||||
|
||||
return true;
|
||||
@ -2762,7 +2762,7 @@ void LSRInstance::GenerateIVChain(const IVChain &Chain, SCEVExpander &Rewriter,
|
||||
|
||||
// If an IV increment can't be folded, use it as the next IV value.
|
||||
if (!canFoldIVIncExpr(LeftOverExpr, IncI->UserInst, IncI->IVOperand,
|
||||
STTI)) {
|
||||
TLI)) {
|
||||
assert(IVTy == IVOper->getType() && "inconsistent IV increment type");
|
||||
IVSrc = IVOper;
|
||||
LeftOverExpr = 0;
|
||||
@ -3108,7 +3108,7 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
|
||||
// into an immediate field.
|
||||
if (isAlwaysFoldable(*J, LU.MinOffset, LU.MaxOffset,
|
||||
Base.getNumRegs() > 1,
|
||||
LU.Kind, LU.AccessTy, STTI, SE))
|
||||
LU.Kind, LU.AccessTy, TLI, SE))
|
||||
continue;
|
||||
|
||||
// Collect all operands except *J.
|
||||
@ -3122,7 +3122,7 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
|
||||
if (InnerAddOps.size() == 1 &&
|
||||
isAlwaysFoldable(InnerAddOps[0], LU.MinOffset, LU.MaxOffset,
|
||||
Base.getNumRegs() > 1,
|
||||
LU.Kind, LU.AccessTy, STTI, SE))
|
||||
LU.Kind, LU.AccessTy, TLI, SE))
|
||||
continue;
|
||||
|
||||
const SCEV *InnerSum = SE.getAddExpr(InnerAddOps);
|
||||
@ -3132,9 +3132,9 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
|
||||
|
||||
// Add the remaining pieces of the add back into the new formula.
|
||||
const SCEVConstant *InnerSumSC = dyn_cast<SCEVConstant>(InnerSum);
|
||||
if (STTI && InnerSumSC &&
|
||||
if (TLI && InnerSumSC &&
|
||||
SE.getTypeSizeInBits(InnerSumSC->getType()) <= 64 &&
|
||||
STTI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
|
||||
TLI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
|
||||
InnerSumSC->getValue()->getZExtValue())) {
|
||||
F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset +
|
||||
InnerSumSC->getValue()->getZExtValue();
|
||||
@ -3144,8 +3144,8 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
|
||||
|
||||
// Add J as its own register, or an unfolded immediate.
|
||||
const SCEVConstant *SC = dyn_cast<SCEVConstant>(*J);
|
||||
if (STTI && SC && SE.getTypeSizeInBits(SC->getType()) <= 64 &&
|
||||
STTI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
|
||||
if (TLI && SC && SE.getTypeSizeInBits(SC->getType()) <= 64 &&
|
||||
TLI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
|
||||
SC->getValue()->getZExtValue()))
|
||||
F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset +
|
||||
SC->getValue()->getZExtValue();
|
||||
@ -3205,7 +3205,7 @@ void LSRInstance::GenerateSymbolicOffsets(LSRUse &LU, unsigned LUIdx,
|
||||
Formula F = Base;
|
||||
F.AM.BaseGV = GV;
|
||||
if (!isLegalUse(F.AM, LU.MinOffset, LU.MaxOffset,
|
||||
LU.Kind, LU.AccessTy, STTI))
|
||||
LU.Kind, LU.AccessTy, TLI))
|
||||
continue;
|
||||
F.BaseRegs[i] = G;
|
||||
(void)InsertFormula(LU, LUIdx, F);
|
||||
@ -3230,7 +3230,7 @@ void LSRInstance::GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx,
|
||||
Formula F = Base;
|
||||
F.AM.BaseOffs = (uint64_t)Base.AM.BaseOffs - *I;
|
||||
if (isLegalUse(F.AM, LU.MinOffset - *I, LU.MaxOffset - *I,
|
||||
LU.Kind, LU.AccessTy, STTI)) {
|
||||
LU.Kind, LU.AccessTy, TLI)) {
|
||||
// Add the offset to the base register.
|
||||
const SCEV *NewG = SE.getAddExpr(SE.getConstant(G->getType(), *I), G);
|
||||
// If it cancelled out, drop the base register, otherwise update it.
|
||||
@ -3250,7 +3250,7 @@ void LSRInstance::GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx,
|
||||
Formula F = Base;
|
||||
F.AM.BaseOffs = (uint64_t)F.AM.BaseOffs + Imm;
|
||||
if (!isLegalUse(F.AM, LU.MinOffset, LU.MaxOffset,
|
||||
LU.Kind, LU.AccessTy, STTI))
|
||||
LU.Kind, LU.AccessTy, TLI))
|
||||
continue;
|
||||
F.BaseRegs[i] = G;
|
||||
(void)InsertFormula(LU, LUIdx, F);
|
||||
@ -3297,7 +3297,7 @@ void LSRInstance::GenerateICmpZeroScales(LSRUse &LU, unsigned LUIdx,
|
||||
F.AM.BaseOffs = NewBaseOffs;
|
||||
|
||||
// Check that this scale is legal.
|
||||
if (!isLegalUse(F.AM, Offset, Offset, LU.Kind, LU.AccessTy, STTI))
|
||||
if (!isLegalUse(F.AM, Offset, Offset, LU.Kind, LU.AccessTy, TLI))
|
||||
continue;
|
||||
|
||||
// Compensate for the use having MinOffset built into it.
|
||||
@ -3353,12 +3353,12 @@ void LSRInstance::GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base) {
|
||||
Base.AM.HasBaseReg = Base.BaseRegs.size() > 1;
|
||||
// Check whether this scale is going to be legal.
|
||||
if (!isLegalUse(Base.AM, LU.MinOffset, LU.MaxOffset,
|
||||
LU.Kind, LU.AccessTy, STTI)) {
|
||||
LU.Kind, LU.AccessTy, TLI)) {
|
||||
// As a special-case, handle special out-of-loop Basic users specially.
|
||||
// TODO: Reconsider this special case.
|
||||
if (LU.Kind == LSRUse::Basic &&
|
||||
isLegalUse(Base.AM, LU.MinOffset, LU.MaxOffset,
|
||||
LSRUse::Special, LU.AccessTy, STTI) &&
|
||||
LSRUse::Special, LU.AccessTy, TLI) &&
|
||||
LU.AllFixupsOutsideLoop)
|
||||
LU.Kind = LSRUse::Special;
|
||||
else
|
||||
@ -3391,8 +3391,8 @@ void LSRInstance::GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base) {
|
||||
|
||||
/// GenerateTruncates - Generate reuse formulae from different IV types.
|
||||
void LSRInstance::GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base) {
|
||||
// This requires ScalarTargetTransformInfo to tell us which truncates are free.
|
||||
if (!STTI) return;
|
||||
// This requires TargetLowering to tell us which truncates are free.
|
||||
if (!TLI) return;
|
||||
|
||||
// Don't bother truncating symbolic values.
|
||||
if (Base.AM.BaseGV) return;
|
||||
@ -3405,7 +3405,7 @@ void LSRInstance::GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base) {
|
||||
for (SmallSetVector<Type *, 4>::const_iterator
|
||||
I = Types.begin(), E = Types.end(); I != E; ++I) {
|
||||
Type *SrcTy = *I;
|
||||
if (SrcTy != DstTy && STTI->isTruncateFree(SrcTy, DstTy)) {
|
||||
if (SrcTy != DstTy && TLI->isTruncateFree(SrcTy, DstTy)) {
|
||||
Formula F = Base;
|
||||
|
||||
if (F.ScaledReg) F.ScaledReg = SE.getAnyExtendExpr(F.ScaledReg, *I);
|
||||
@ -3561,7 +3561,7 @@ void LSRInstance::GenerateCrossUseConstantOffsets() {
|
||||
Formula NewF = F;
|
||||
NewF.AM.BaseOffs = Offs;
|
||||
if (!isLegalUse(NewF.AM, LU.MinOffset, LU.MaxOffset,
|
||||
LU.Kind, LU.AccessTy, STTI))
|
||||
LU.Kind, LU.AccessTy, TLI))
|
||||
continue;
|
||||
NewF.ScaledReg = SE.getAddExpr(NegImmS, NewF.ScaledReg);
|
||||
|
||||
@ -3586,9 +3586,9 @@ void LSRInstance::GenerateCrossUseConstantOffsets() {
|
||||
Formula NewF = F;
|
||||
NewF.AM.BaseOffs = (uint64_t)NewF.AM.BaseOffs + Imm;
|
||||
if (!isLegalUse(NewF.AM, LU.MinOffset, LU.MaxOffset,
|
||||
LU.Kind, LU.AccessTy, STTI)) {
|
||||
if (!STTI ||
|
||||
!STTI->isLegalAddImmediate((uint64_t)NewF.UnfoldedOffset + Imm))
|
||||
LU.Kind, LU.AccessTy, TLI)) {
|
||||
if (!TLI ||
|
||||
!TLI->isLegalAddImmediate((uint64_t)NewF.UnfoldedOffset + Imm))
|
||||
continue;
|
||||
NewF = F;
|
||||
NewF.UnfoldedOffset = (uint64_t)NewF.UnfoldedOffset + Imm;
|
||||
@ -3900,7 +3900,7 @@ void LSRInstance::NarrowSearchSpaceByCollapsingUnrolledCode() {
|
||||
Formula &F = LUThatHas->Formulae[i];
|
||||
if (!isLegalUse(F.AM,
|
||||
LUThatHas->MinOffset, LUThatHas->MaxOffset,
|
||||
LUThatHas->Kind, LUThatHas->AccessTy, STTI)) {
|
||||
LUThatHas->Kind, LUThatHas->AccessTy, TLI)) {
|
||||
DEBUG(dbgs() << " Deleting "; F.print(dbgs());
|
||||
dbgs() << '\n');
|
||||
LUThatHas->DeleteFormula(F);
|
||||
@ -4589,12 +4589,12 @@ LSRInstance::ImplementSolution(const SmallVectorImpl<const Formula *> &Solution,
|
||||
Changed |= DeleteTriviallyDeadInstructions(DeadInsts);
|
||||
}
|
||||
|
||||
LSRInstance::LSRInstance(const ScalarTargetTransformInfo *stti, Loop *l, Pass *P)
|
||||
LSRInstance::LSRInstance(const TargetLowering *tli, Loop *l, Pass *P)
|
||||
: IU(P->getAnalysis<IVUsers>()),
|
||||
SE(P->getAnalysis<ScalarEvolution>()),
|
||||
DT(P->getAnalysis<DominatorTree>()),
|
||||
LI(P->getAnalysis<LoopInfo>()),
|
||||
STTI(stti), L(l), Changed(false), IVIncInsertPos(0) {
|
||||
TLI(tli), L(l), Changed(false), IVIncInsertPos(0) {
|
||||
|
||||
// If LoopSimplify form is not available, stay out of trouble.
|
||||
if (!L->isLoopSimplifyForm())
|
||||
@ -4684,7 +4684,7 @@ LSRInstance::LSRInstance(const ScalarTargetTransformInfo *stti, Loop *l, Pass *P
|
||||
for (SmallVectorImpl<Formula>::const_iterator J = LU.Formulae.begin(),
|
||||
JE = LU.Formulae.end(); J != JE; ++J)
|
||||
assert(isLegalUse(J->AM, LU.MinOffset, LU.MaxOffset,
|
||||
LU.Kind, LU.AccessTy, STTI) &&
|
||||
LU.Kind, LU.AccessTy, TLI) &&
|
||||
"Illegal formula generated!");
|
||||
};
|
||||
#endif
|
||||
@ -4757,13 +4757,13 @@ void LSRInstance::dump() const {
|
||||
namespace {
|
||||
|
||||
class LoopStrengthReduce : public LoopPass {
|
||||
/// ScalarTargetTransformInfo provides target information that is needed
|
||||
/// for strength reducing loops.
|
||||
const ScalarTargetTransformInfo *STTI;
|
||||
/// TLI - Keep a pointer of a TargetLowering to consult for determining
|
||||
/// transformation profitability.
|
||||
const TargetLowering *const TLI;
|
||||
|
||||
public:
|
||||
static char ID; // Pass ID, replacement for typeid
|
||||
LoopStrengthReduce();
|
||||
explicit LoopStrengthReduce(const TargetLowering *tli = 0);
|
||||
|
||||
private:
|
||||
bool runOnLoop(Loop *L, LPPassManager &LPM);
|
||||
@ -4783,12 +4783,13 @@ INITIALIZE_PASS_DEPENDENCY(LoopSimplify)
|
||||
INITIALIZE_PASS_END(LoopStrengthReduce, "loop-reduce",
|
||||
"Loop Strength Reduction", false, false)
|
||||
|
||||
Pass *llvm::createLoopStrengthReducePass() {
|
||||
return new LoopStrengthReduce();
|
||||
|
||||
Pass *llvm::createLoopStrengthReducePass(const TargetLowering *TLI) {
|
||||
return new LoopStrengthReduce(TLI);
|
||||
}
|
||||
|
||||
LoopStrengthReduce::LoopStrengthReduce()
|
||||
: LoopPass(ID), STTI(0) {
|
||||
LoopStrengthReduce::LoopStrengthReduce(const TargetLowering *tli)
|
||||
: LoopPass(ID), TLI(tli) {
|
||||
initializeLoopStrengthReducePass(*PassRegistry::getPassRegistry());
|
||||
}
|
||||
|
||||
@ -4814,13 +4815,8 @@ void LoopStrengthReduce::getAnalysisUsage(AnalysisUsage &AU) const {
|
||||
bool LoopStrengthReduce::runOnLoop(Loop *L, LPPassManager & /*LPM*/) {
|
||||
bool Changed = false;
|
||||
|
||||
TargetTransformInfo *TTI = getAnalysisIfAvailable<TargetTransformInfo>();
|
||||
|
||||
if (TTI)
|
||||
STTI = TTI->getScalarTargetTransformInfo();
|
||||
|
||||
// Run the main LSR transformation.
|
||||
Changed |= LSRInstance(STTI, L, this).getChanged();
|
||||
Changed |= LSRInstance(TLI, L, this).getChanged();
|
||||
|
||||
// Remove any extra phis created by processing inner loops.
|
||||
Changed |= DeleteDeadPHIs(L->getHeader());
|
||||
@ -4831,7 +4827,7 @@ bool LoopStrengthReduce::runOnLoop(Loop *L, LPPassManager & /*LPM*/) {
|
||||
Rewriter.setDebugType(DEBUG_TYPE);
|
||||
#endif
|
||||
unsigned numFolded = Rewriter.
|
||||
replaceCongruentIVs(L, &getAnalysis<DominatorTree>(), DeadInsts, STTI);
|
||||
replaceCongruentIVs(L, &getAnalysis<DominatorTree>(), DeadInsts, TLI);
|
||||
if (numFolded) {
|
||||
Changed = true;
|
||||
DeleteTriviallyDeadInstructions(DeadInsts);
|
||||
|
@ -45,10 +45,10 @@
|
||||
#include "llvm/Pass.h"
|
||||
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
|
||||
#include "llvm/Transforms/Utils/Local.h"
|
||||
#include "llvm/TargetTransformInfo.h"
|
||||
#include "llvm/ADT/SmallVector.h"
|
||||
#include "llvm/ADT/Statistic.h"
|
||||
#include "llvm/Support/CommandLine.h"
|
||||
#include "llvm/Target/TargetLowering.h"
|
||||
#include <csetjmp>
|
||||
#include <set>
|
||||
using namespace llvm;
|
||||
@ -70,14 +70,15 @@ namespace {
|
||||
Constant *SetJmpFn, *LongJmpFn, *StackSaveFn, *StackRestoreFn;
|
||||
bool useExpensiveEHSupport;
|
||||
|
||||
// We peek in STTI to grab the target's jmp_buf size and alignment
|
||||
const ScalarTargetTransformInfo *STTI;
|
||||
// We peek in TLI to grab the target's jmp_buf size and alignment
|
||||
const TargetLowering *TLI;
|
||||
|
||||
public:
|
||||
static char ID; // Pass identification, replacement for typeid
|
||||
explicit LowerInvoke(bool useExpensiveEHSupport = ExpensiveEHSupport)
|
||||
explicit LowerInvoke(const TargetLowering *tli = NULL,
|
||||
bool useExpensiveEHSupport = ExpensiveEHSupport)
|
||||
: FunctionPass(ID), useExpensiveEHSupport(useExpensiveEHSupport),
|
||||
STTI(0) {
|
||||
TLI(tli) {
|
||||
initializeLowerInvokePass(*PassRegistry::getPassRegistry());
|
||||
}
|
||||
bool doInitialization(Module &M);
|
||||
@ -107,24 +108,21 @@ INITIALIZE_PASS(LowerInvoke, "lowerinvoke",
|
||||
char &llvm::LowerInvokePassID = LowerInvoke::ID;
|
||||
|
||||
// Public Interface To the LowerInvoke pass.
|
||||
FunctionPass *llvm::createLowerInvokePass() {
|
||||
return new LowerInvoke(ExpensiveEHSupport);
|
||||
FunctionPass *llvm::createLowerInvokePass(const TargetLowering *TLI) {
|
||||
return new LowerInvoke(TLI, ExpensiveEHSupport);
|
||||
}
|
||||
FunctionPass *llvm::createLowerInvokePass(bool useExpensiveEHSupport) {
|
||||
return new LowerInvoke(useExpensiveEHSupport);
|
||||
FunctionPass *llvm::createLowerInvokePass(const TargetLowering *TLI,
|
||||
bool useExpensiveEHSupport) {
|
||||
return new LowerInvoke(TLI, useExpensiveEHSupport);
|
||||
}
|
||||
|
||||
// doInitialization - Make sure that there is a prototype for abort in the
|
||||
// current module.
|
||||
bool LowerInvoke::doInitialization(Module &M) {
|
||||
TargetTransformInfo *TTI = getAnalysisIfAvailable<TargetTransformInfo>();
|
||||
if (TTI)
|
||||
STTI = TTI->getScalarTargetTransformInfo();
|
||||
|
||||
Type *VoidPtrTy = Type::getInt8PtrTy(M.getContext());
|
||||
if (useExpensiveEHSupport) {
|
||||
// Insert a type for the linked list of jump buffers.
|
||||
unsigned JBSize = STTI ? STTI->getJumpBufSize() : 0;
|
||||
unsigned JBSize = TLI ? TLI->getJumpBufSize() : 0;
|
||||
JBSize = JBSize ? JBSize : 200;
|
||||
Type *JmpBufTy = ArrayType::get(VoidPtrTy, JBSize);
|
||||
|
||||
@ -432,7 +430,7 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) {
|
||||
// Create an alloca for the incoming jump buffer ptr and the new jump buffer
|
||||
// that needs to be restored on all exits from the function. This is an
|
||||
// alloca because the value needs to be live across invokes.
|
||||
unsigned Align = STTI ? STTI->getJumpBufAlignment() : 0;
|
||||
unsigned Align = TLI ? TLI->getJumpBufAlignment() : 0;
|
||||
AllocaInst *JmpBuf =
|
||||
new AllocaInst(JBLinkTy, 0, Align,
|
||||
"jblink", F.begin()->begin());
|
||||
@ -577,10 +575,6 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) {
|
||||
}
|
||||
|
||||
bool LowerInvoke::runOnFunction(Function &F) {
|
||||
TargetTransformInfo *TTI = getAnalysisIfAvailable<TargetTransformInfo>();
|
||||
if (TTI)
|
||||
STTI = TTI->getScalarTargetTransformInfo();
|
||||
|
||||
if (useExpensiveEHSupport)
|
||||
return insertExpensiveEHSupport(F);
|
||||
else
|
||||
|
Loading…
Reference in New Issue
Block a user