///
//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "ppctti"
#include "PPC.h"
#include "PPCTargetMachine.h"
#include "llvm/Analysis/TargetTransformInfo.h"
+#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Target/CostTable.h"
#include "llvm/Target/TargetLowering.h"
using namespace llvm;
+#define DEBUG_TYPE "ppctti"
+
+static cl::opt<bool> DisablePPCConstHoist("disable-ppc-constant-hoisting",
+cl::desc("disable constant hoisting on PPC"), cl::init(false), cl::Hidden);
+
// Declare the pass initialization routine locally as target-specific passes
-// don't havve a target-wide initialization entry point, and so we rely on the
+// don't have a target-wide initialization entry point, and so we rely on the
// pass constructor initialization.
namespace llvm {
void initializePPCTTIPass(PassRegistry &);
namespace {
class PPCTTI final : public ImmutablePass, public TargetTransformInfo {
+ const TargetMachine *TM;
const PPCSubtarget *ST;
const PPCTargetLowering *TLI;
public:
- PPCTTI() : ImmutablePass(ID), ST(0), TLI(0) {
+ PPCTTI() : ImmutablePass(ID), ST(nullptr), TLI(nullptr) {
llvm_unreachable("This pass cannot be directly constructed");
}
PPCTTI(const PPCTargetMachine *TM)
- : ImmutablePass(ID), ST(TM->getSubtargetImpl()),
- TLI(TM->getTargetLowering()) {
+ : ImmutablePass(ID), TM(TM), ST(TM->getSubtargetImpl()),
+ TLI(TM->getSubtargetImpl()->getTargetLowering()) {
initializePPCTTIPass(*PassRegistry::getPassRegistry());
}
- virtual void initializePass() override {
+ void initializePass() override {
pushTTIStack(this);
}
- virtual void getAnalysisUsage(AnalysisUsage &AU) const override {
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
TargetTransformInfo::getAnalysisUsage(AU);
}
static char ID;
/// Provide necessary pointer adjustments for the two base classes.
- virtual void *getAdjustedAnalysisPointer(const void *ID) override {
+ void *getAdjustedAnalysisPointer(const void *ID) override {
if (ID == &TargetTransformInfo::ID)
return (TargetTransformInfo*)this;
return this;
/// \name Scalar TTI Implementations
/// @{
- virtual PopcntSupportKind
- getPopcntSupport(unsigned TyWidth) const override;
- virtual void getUnrollingPreferences(
- Loop *L, UnrollingPreferences &UP) const override;
+ unsigned getIntImmCost(const APInt &Imm, Type *Ty) const override;
+
+ unsigned getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm,
+ Type *Ty) const override;
+ unsigned getIntImmCost(Intrinsic::ID IID, unsigned Idx, const APInt &Imm,
+ Type *Ty) const override;
+
+ PopcntSupportKind getPopcntSupport(unsigned TyWidth) const override;
+ void getUnrollingPreferences(const Function *F, Loop *L,
+ UnrollingPreferences &UP) const override;
/// @}
/// \name Vector TTI Implementations
/// @{
- virtual unsigned getNumberOfRegisters(bool Vector) const override;
- virtual unsigned getRegisterBitWidth(bool Vector) const override;
- virtual unsigned getMaximumUnrollFactor() const override;
- virtual unsigned getArithmeticInstrCost(unsigned Opcode, Type *Ty,
- OperandValueKind,
- OperandValueKind) const override;
- virtual unsigned getShuffleCost(ShuffleKind Kind, Type *Tp,
- int Index, Type *SubTp) const override;
- virtual unsigned getCastInstrCost(unsigned Opcode, Type *Dst,
- Type *Src) const override;
- virtual unsigned getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
- Type *CondTy) const override;
- virtual unsigned getVectorInstrCost(unsigned Opcode, Type *Val,
- unsigned Index) const override;
- virtual unsigned getMemoryOpCost(unsigned Opcode, Type *Src,
- unsigned Alignment,
- unsigned AddressSpace) const override;
+ unsigned getNumberOfRegisters(bool Vector) const override;
+ unsigned getRegisterBitWidth(bool Vector) const override;
+ unsigned getMaxInterleaveFactor() const override;
+ unsigned getArithmeticInstrCost(unsigned Opcode, Type *Ty, OperandValueKind,
+ OperandValueKind, OperandValueProperties,
+ OperandValueProperties) const override;
+ unsigned getShuffleCost(ShuffleKind Kind, Type *Tp,
+ int Index, Type *SubTp) const override;
+ unsigned getCastInstrCost(unsigned Opcode, Type *Dst,
+ Type *Src) const override;
+ unsigned getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
+ Type *CondTy) const override;
+ unsigned getVectorInstrCost(unsigned Opcode, Type *Val,
+ unsigned Index) const override;
+ unsigned getMemoryOpCost(unsigned Opcode, Type *Src, unsigned Alignment,
+ unsigned AddressSpace) const override;
/// @}
};
return PSK_Software;
}
-void PPCTTI::getUnrollingPreferences(Loop *L, UnrollingPreferences &UP) const {
- if (ST->getDarwinDirective() == PPC::DIR_A2) {
+unsigned PPCTTI::getIntImmCost(const APInt &Imm, Type *Ty) const {
+ if (DisablePPCConstHoist)
+ return TargetTransformInfo::getIntImmCost(Imm, Ty);
+
+ assert(Ty->isIntegerTy());
+
+ unsigned BitSize = Ty->getPrimitiveSizeInBits();
+ if (BitSize == 0)
+ return ~0U;
+
+ if (Imm == 0)
+ return TCC_Free;
+
+ if (Imm.getBitWidth() <= 64) {
+ if (isInt<16>(Imm.getSExtValue()))
+ return TCC_Basic;
+
+ if (isInt<32>(Imm.getSExtValue())) {
+ // A constant that can be materialized using lis.
+ if ((Imm.getZExtValue() & 0xFFFF) == 0)
+ return TCC_Basic;
+
+ return 2 * TCC_Basic;
+ }
+ }
+
+ return 4 * TCC_Basic;
+}
+
+unsigned PPCTTI::getIntImmCost(Intrinsic::ID IID, unsigned Idx,
+ const APInt &Imm, Type *Ty) const {
+ if (DisablePPCConstHoist)
+ return TargetTransformInfo::getIntImmCost(IID, Idx, Imm, Ty);
+
+ assert(Ty->isIntegerTy());
+
+ unsigned BitSize = Ty->getPrimitiveSizeInBits();
+ if (BitSize == 0)
+ return ~0U;
+
+ switch (IID) {
+ default: return TCC_Free;
+ case Intrinsic::sadd_with_overflow:
+ case Intrinsic::uadd_with_overflow:
+ case Intrinsic::ssub_with_overflow:
+ case Intrinsic::usub_with_overflow:
+ if ((Idx == 1) && Imm.getBitWidth() <= 64 && isInt<16>(Imm.getSExtValue()))
+ return TCC_Free;
+ break;
+ }
+ return PPCTTI::getIntImmCost(Imm, Ty);
+}
+
+unsigned PPCTTI::getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm,
+ Type *Ty) const {
+ if (DisablePPCConstHoist)
+ return TargetTransformInfo::getIntImmCost(Opcode, Idx, Imm, Ty);
+
+ assert(Ty->isIntegerTy());
+
+ unsigned BitSize = Ty->getPrimitiveSizeInBits();
+ if (BitSize == 0)
+ return ~0U;
+
+ unsigned ImmIdx = ~0U;
+ bool ShiftedFree = false, RunFree = false, UnsignedFree = false,
+ ZeroFree = false;
+ switch (Opcode) {
+ default: return TCC_Free;
+ case Instruction::GetElementPtr:
+ // Always hoist the base address of a GetElementPtr. This prevents the
+ // creation of new constants for every base constant that gets constant
+ // folded with the offset.
+ if (Idx == 0)
+ return 2 * TCC_Basic;
+ return TCC_Free;
+ case Instruction::And:
+ RunFree = true; // (for the rotate-and-mask instructions)
+ // Fallthrough...
+ case Instruction::Add:
+ case Instruction::Or:
+ case Instruction::Xor:
+ ShiftedFree = true;
+ // Fallthrough...
+ case Instruction::Sub:
+ case Instruction::Mul:
+ case Instruction::Shl:
+ case Instruction::LShr:
+ case Instruction::AShr:
+ ImmIdx = 1;
+ break;
+ case Instruction::ICmp:
+ UnsignedFree = true;
+ ImmIdx = 1;
+ // Fallthrough... (zero comparisons can use record-form instructions)
+ case Instruction::Select:
+ ZeroFree = true;
+ break;
+ case Instruction::PHI:
+ case Instruction::Call:
+ case Instruction::Ret:
+ case Instruction::Load:
+ case Instruction::Store:
+ break;
+ }
+
+ if (ZeroFree && Imm == 0)
+ return TCC_Free;
+
+ if (Idx == ImmIdx && Imm.getBitWidth() <= 64) {
+ if (isInt<16>(Imm.getSExtValue()))
+ return TCC_Free;
+
+ if (RunFree) {
+ if (Imm.getBitWidth() <= 32 &&
+ (isShiftedMask_32(Imm.getZExtValue()) ||
+ isShiftedMask_32(~Imm.getZExtValue())))
+ return TCC_Free;
+
+
+ if (ST->isPPC64() &&
+ (isShiftedMask_64(Imm.getZExtValue()) ||
+ isShiftedMask_64(~Imm.getZExtValue())))
+ return TCC_Free;
+ }
+
+ if (UnsignedFree && isUInt<16>(Imm.getZExtValue()))
+ return TCC_Free;
+
+ if (ShiftedFree && (Imm.getZExtValue() & 0xFFFF) == 0)
+ return TCC_Free;
+ }
+
+ return PPCTTI::getIntImmCost(Imm, Ty);
+}
+
+void PPCTTI::getUnrollingPreferences(const Function *F, Loop *L,
+ UnrollingPreferences &UP) const {
+ if (TM->getSubtarget<PPCSubtarget>(F).getDarwinDirective() == PPC::DIR_A2) {
// The A2 is in-order with a deep pipeline, and concatenation unrolling
// helps expose latency-hiding opportunities to the instruction scheduler.
UP.Partial = UP.Runtime = true;
}
-unsigned PPCTTI::getMaximumUnrollFactor() const {
+unsigned PPCTTI::getMaxInterleaveFactor() const {
unsigned Directive = ST->getDarwinDirective();
// The 440 has no SIMD support, but floating-point instructions
// have a 5-cycle latency, so unroll by 5x for latency hiding.
return 2;
}
-unsigned PPCTTI::getArithmeticInstrCost(unsigned Opcode, Type *Ty,
- OperandValueKind Op1Info,
- OperandValueKind Op2Info) const {
+unsigned PPCTTI::getArithmeticInstrCost(
+ unsigned Opcode, Type *Ty, OperandValueKind Op1Info,
+ OperandValueKind Op2Info, OperandValueProperties Opd1PropInfo,
+ OperandValueProperties Opd2PropInfo) const {
assert(TLI->InstructionOpcodeToISD(Opcode) && "Invalid opcode");
// Fallback to the default implementation.
- return TargetTransformInfo::getArithmeticInstrCost(Opcode, Ty, Op1Info,
- Op2Info);
+ return TargetTransformInfo::getArithmeticInstrCost(
+ Opcode, Ty, Op1Info, Op2Info, Opd1PropInfo, Opd2PropInfo);
}
unsigned PPCTTI::getShuffleCost(ShuffleKind Kind, Type *Tp, int Index,