#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/GlobalVariable.h"
+#include "llvm/IR/LLVMContext.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCExpr.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Target/TargetLoweringObjectFile.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetRegisterInfo.h"
+#include "llvm/Target/TargetSubtargetInfo.h"
#include <cctype>
using namespace llvm;
-/// NOTE: The constructor takes ownership of TLOF.
-TargetLowering::TargetLowering(const TargetMachine &tm,
- const TargetLoweringObjectFile *tlof)
- : TargetLoweringBase(tm, tlof) {}
+/// NOTE: The TargetMachine owns TLOF.
+TargetLowering::TargetLowering(const TargetMachine &tm)
+ : TargetLoweringBase(tm) {}
const char *TargetLowering::getTargetNodeName(unsigned Opcode) const {
- return NULL;
+ return nullptr;
}
/// Check whether a given call node is in tail position within its function. If
return isUsedByReturnOnly(Node, Chain);
}
+/// \brief Set CallLoweringInfo attribute flags based on a call instruction
+/// and called function attributes.
+void TargetLowering::ArgListEntry::setAttributes(ImmutableCallSite *CS,
+ unsigned AttrIdx) {
+ isSExt = CS->paramHasAttr(AttrIdx, Attribute::SExt);
+ isZExt = CS->paramHasAttr(AttrIdx, Attribute::ZExt);
+ isInReg = CS->paramHasAttr(AttrIdx, Attribute::InReg);
+ isSRet = CS->paramHasAttr(AttrIdx, Attribute::StructRet);
+ isNest = CS->paramHasAttr(AttrIdx, Attribute::Nest);
+ isByVal = CS->paramHasAttr(AttrIdx, Attribute::ByVal);
+ isInAlloca = CS->paramHasAttr(AttrIdx, Attribute::InAlloca);
+ isReturned = CS->paramHasAttr(AttrIdx, Attribute::Returned);
+ Alignment = CS->getParamAlignment(AttrIdx);
+}
/// Generate a libcall taking the given operands as arguments and returning a
/// result of type RetVT.
-SDValue TargetLowering::makeLibCall(SelectionDAG &DAG,
- RTLIB::Libcall LC, EVT RetVT,
- const SDValue *Ops, unsigned NumOps,
- bool isSigned, DebugLoc dl) const {
+std::pair<SDValue, SDValue>
+TargetLowering::makeLibCall(SelectionDAG &DAG,
+ RTLIB::Libcall LC, EVT RetVT,
+ ArrayRef<SDValue> Ops,
+ bool isSigned, SDLoc dl,
+ bool doesNotReturn,
+ bool isReturnValueUsed) const {
TargetLowering::ArgListTy Args;
- Args.reserve(NumOps);
+ Args.reserve(Ops.size());
TargetLowering::ArgListEntry Entry;
- for (unsigned i = 0; i != NumOps; ++i) {
- Entry.Node = Ops[i];
+ for (SDValue Op : Ops) {
+ Entry.Node = Op;
Entry.Ty = Entry.Node.getValueType().getTypeForEVT(*DAG.getContext());
- Entry.isSExt = isSigned;
- Entry.isZExt = !isSigned;
+ Entry.isSExt = shouldSignExtendTypeInLibCall(Op.getValueType(), isSigned);
+ Entry.isZExt = !shouldSignExtendTypeInLibCall(Op.getValueType(), isSigned);
Args.push_back(Entry);
}
- SDValue Callee = DAG.getExternalSymbol(getLibcallName(LC), getPointerTy());
+
+ if (LC == RTLIB::UNKNOWN_LIBCALL)
+ report_fatal_error("Unsupported library call operation!");
+ SDValue Callee = DAG.getExternalSymbol(getLibcallName(LC),
+ getPointerTy(DAG.getDataLayout()));
Type *RetTy = RetVT.getTypeForEVT(*DAG.getContext());
- TargetLowering::
- CallLoweringInfo CLI(DAG.getEntryNode(), RetTy, isSigned, !isSigned, false,
- false, 0, getLibcallCallingConv(LC),
- /*isTailCall=*/false,
- /*doesNotReturn=*/false, /*isReturnValueUsed=*/true,
- Callee, Args, DAG, dl);
- std::pair<SDValue,SDValue> CallInfo = LowerCallTo(CLI);
-
- return CallInfo.first;
+ TargetLowering::CallLoweringInfo CLI(DAG);
+ bool signExtend = shouldSignExtendTypeInLibCall(RetVT, isSigned);
+ CLI.setDebugLoc(dl).setChain(DAG.getEntryNode())
+ .setCallee(getLibcallCallingConv(LC), RetTy, Callee, std::move(Args), 0)
+ .setNoReturn(doesNotReturn).setDiscardResult(!isReturnValueUsed)
+ .setSExtResult(signExtend).setZExtResult(!signExtend);
+ return LowerCallTo(CLI);
}
-
-/// SoftenSetCCOperands - Soften the operands of a comparison. This code is
-/// shared among BR_CC, SELECT_CC, and SETCC handlers.
+/// Soften the operands of a comparison. This code is shared among BR_CC,
+/// SELECT_CC, and SETCC handlers.
void TargetLowering::softenSetCCOperands(SelectionDAG &DAG, EVT VT,
SDValue &NewLHS, SDValue &NewRHS,
ISD::CondCode &CCCode,
- DebugLoc dl) const {
+ SDLoc dl) const {
assert((VT == MVT::f32 || VT == MVT::f64 || VT == MVT::f128)
&& "Unsupported setcc type!");
// Expand into one or more soft-fp libcall(s).
RTLIB::Libcall LC1 = RTLIB::UNKNOWN_LIBCALL, LC2 = RTLIB::UNKNOWN_LIBCALL;
+ bool ShouldInvertCC = false;
switch (CCCode) {
case ISD::SETEQ:
case ISD::SETOEQ:
LC1 = (VT == MVT::f32) ? RTLIB::O_F32 :
(VT == MVT::f64) ? RTLIB::O_F64 : RTLIB::O_F128;
break;
- default:
+ case ISD::SETONE:
+ // SETONE = SETOLT | SETOGT
+ LC1 = (VT == MVT::f32) ? RTLIB::OLT_F32 :
+ (VT == MVT::f64) ? RTLIB::OLT_F64 : RTLIB::OLT_F128;
+ LC2 = (VT == MVT::f32) ? RTLIB::OGT_F32 :
+ (VT == MVT::f64) ? RTLIB::OGT_F64 : RTLIB::OGT_F128;
+ break;
+ case ISD::SETUEQ:
LC1 = (VT == MVT::f32) ? RTLIB::UO_F32 :
(VT == MVT::f64) ? RTLIB::UO_F64 : RTLIB::UO_F128;
+ LC2 = (VT == MVT::f32) ? RTLIB::OEQ_F32 :
+ (VT == MVT::f64) ? RTLIB::OEQ_F64 : RTLIB::OEQ_F128;
+ break;
+ default:
+ // Invert CC for unordered comparisons
+ ShouldInvertCC = true;
switch (CCCode) {
- case ISD::SETONE:
- // SETONE = SETOLT | SETOGT
- LC1 = (VT == MVT::f32) ? RTLIB::OLT_F32 :
- (VT == MVT::f64) ? RTLIB::OLT_F64 : RTLIB::OLT_F128;
- // Fallthrough
- case ISD::SETUGT:
- LC2 = (VT == MVT::f32) ? RTLIB::OGT_F32 :
- (VT == MVT::f64) ? RTLIB::OGT_F64 : RTLIB::OGT_F128;
- break;
- case ISD::SETUGE:
- LC2 = (VT == MVT::f32) ? RTLIB::OGE_F32 :
- (VT == MVT::f64) ? RTLIB::OGE_F64 : RTLIB::OGE_F128;
- break;
case ISD::SETULT:
- LC2 = (VT == MVT::f32) ? RTLIB::OLT_F32 :
- (VT == MVT::f64) ? RTLIB::OLT_F64 : RTLIB::OLT_F128;
+ LC1 = (VT == MVT::f32) ? RTLIB::OGE_F32 :
+ (VT == MVT::f64) ? RTLIB::OGE_F64 : RTLIB::OGE_F128;
break;
case ISD::SETULE:
- LC2 = (VT == MVT::f32) ? RTLIB::OLE_F32 :
+ LC1 = (VT == MVT::f32) ? RTLIB::OGT_F32 :
+ (VT == MVT::f64) ? RTLIB::OGT_F64 : RTLIB::OGT_F128;
+ break;
+ case ISD::SETUGT:
+ LC1 = (VT == MVT::f32) ? RTLIB::OLE_F32 :
(VT == MVT::f64) ? RTLIB::OLE_F64 : RTLIB::OLE_F128;
break;
- case ISD::SETUEQ:
- LC2 = (VT == MVT::f32) ? RTLIB::OEQ_F32 :
- (VT == MVT::f64) ? RTLIB::OEQ_F64 : RTLIB::OEQ_F128;
+ case ISD::SETUGE:
+ LC1 = (VT == MVT::f32) ? RTLIB::OLT_F32 :
+ (VT == MVT::f64) ? RTLIB::OLT_F64 : RTLIB::OLT_F128;
break;
default: llvm_unreachable("Do not know how to soften this setcc!");
}
// Use the target specific return value for comparions lib calls.
EVT RetVT = getCmpLibcallReturnType();
- SDValue Ops[2] = { NewLHS, NewRHS };
- NewLHS = makeLibCall(DAG, LC1, RetVT, Ops, 2, false/*sign irrelevant*/, dl);
- NewRHS = DAG.getConstant(0, RetVT);
+ SDValue Ops[2] = {NewLHS, NewRHS};
+ NewLHS = makeLibCall(DAG, LC1, RetVT, Ops, false /*sign irrelevant*/,
+ dl).first;
+ NewRHS = DAG.getConstant(0, dl, RetVT);
+
CCCode = getCmpLibcallCC(LC1);
+ if (ShouldInvertCC)
+ CCCode = getSetCCInverse(CCCode, /*isInteger=*/true);
+
if (LC2 != RTLIB::UNKNOWN_LIBCALL) {
- SDValue Tmp = DAG.getNode(ISD::SETCC, dl, getSetCCResultType(RetVT),
- NewLHS, NewRHS, DAG.getCondCode(CCCode));
- NewLHS = makeLibCall(DAG, LC2, RetVT, Ops, 2, false/*sign irrelevant*/, dl);
- NewLHS = DAG.getNode(ISD::SETCC, dl, getSetCCResultType(RetVT), NewLHS,
- NewRHS, DAG.getCondCode(getCmpLibcallCC(LC2)));
+ SDValue Tmp = DAG.getNode(
+ ISD::SETCC, dl,
+ getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), RetVT),
+ NewLHS, NewRHS, DAG.getCondCode(CCCode));
+ NewLHS = makeLibCall(DAG, LC2, RetVT, Ops, false/*sign irrelevant*/,
+ dl).first;
+ NewLHS = DAG.getNode(
+ ISD::SETCC, dl,
+ getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), RetVT),
+ NewLHS, NewRHS, DAG.getCondCode(getCmpLibcallCC(LC2)));
NewLHS = DAG.getNode(ISD::OR, dl, Tmp.getValueType(), Tmp, NewLHS);
NewRHS = SDValue();
}
}
-/// getJumpTableEncoding - Return the entry encoding for a jump table in the
-/// current function. The returned value is a member of the
-/// MachineJumpTableInfo::JTEntryKind enum.
+/// Return the entry encoding for a jump table in the current function. The
+/// returned value is a member of the MachineJumpTableInfo::JTEntryKind enum.
unsigned TargetLowering::getJumpTableEncoding() const {
// In non-pic modes, just use the address of a block.
if (getTargetMachine().getRelocationModel() != Reloc::PIC_)
return MachineJumpTableInfo::EK_BlockAddress;
// In PIC mode, if the target supports a GPRel32 directive, use it.
- if (getTargetMachine().getMCAsmInfo()->getGPRel32Directive() != 0)
+ if (getTargetMachine().getMCAsmInfo()->getGPRel32Directive() != nullptr)
return MachineJumpTableInfo::EK_GPRel32BlockAddress;
// Otherwise, use a label difference.
if ((JTEncoding == MachineJumpTableInfo::EK_GPRel64BlockAddress) ||
(JTEncoding == MachineJumpTableInfo::EK_GPRel32BlockAddress))
- return DAG.getGLOBAL_OFFSET_TABLE(getPointerTy(0));
+ return DAG.getGLOBAL_OFFSET_TABLE(getPointerTy(DAG.getDataLayout()));
return Table;
}
-/// getPICJumpTableRelocBaseExpr - This returns the relocation base for the
-/// given PIC jumptable, the same as getPICJumpTableRelocBase, but as an
-/// MCExpr.
+/// This returns the relocation base for the given PIC jumptable, the same as
+/// getPICJumpTableRelocBase, but as an MCExpr.
const MCExpr *
TargetLowering::getPICJumpTableRelocBaseExpr(const MachineFunction *MF,
unsigned JTI,MCContext &Ctx) const{
// The normal PIC reloc base is the label at the start of the jump table.
- return MCSymbolRefExpr::Create(MF->getJTISymbol(JTI, Ctx), Ctx);
+ return MCSymbolRefExpr::create(MF->getJTISymbol(JTI, Ctx), Ctx);
}
bool
// In dynamic-no-pic mode, assume that known defined values are safe.
if (getTargetMachine().getRelocationModel() == Reloc::DynamicNoPIC &&
- GA &&
- !GA->getGlobal()->isDeclaration() &&
- !GA->getGlobal()->isWeakForLinker())
+ GA && GA->getGlobal()->isStrongDefinitionForLinker())
return true;
// Otherwise assume nothing is safe.
// Optimization Methods
//===----------------------------------------------------------------------===//
-/// ShrinkDemandedConstant - Check to see if the specified operand of the
-/// specified instruction is a constant integer. If so, check to see if there
-/// are any bits set in the constant that are not demanded. If so, shrink the
-/// constant and return true.
+/// Check to see if the specified operand of the specified instruction is a
+/// constant integer. If so, check to see if there are any bits set in the
+/// constant that are not demanded. If so, shrink the constant and return true.
bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant(SDValue Op,
const APInt &Demanded) {
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
// FIXME: ISD::SELECT, ISD::SELECT_CC
switch (Op.getOpcode()) {
SDValue New = DAG.getNode(Op.getOpcode(), dl, VT, Op.getOperand(0),
DAG.getConstant(Demanded &
C->getAPIntValue(),
- VT));
+ dl, VT));
return CombineTo(Op, New);
}
return false;
}
-/// ShrinkDemandedOp - Convert x+y to (VT)((SmallVT)x+(SmallVT)y) if the
-/// casts are free. This uses isZExtFree and ZERO_EXTEND for the widening
-/// cast, but it could be generalized for targets with other types of
-/// implicit widening casts.
+/// Convert x+y to (VT)((SmallVT)x+(SmallVT)y) if the casts are free.
+/// This uses isZExtFree and ZERO_EXTEND for the widening cast, but it could be
+/// generalized for targets with other types of implicit widening casts.
bool
TargetLowering::TargetLoweringOpt::ShrinkDemandedOp(SDValue Op,
unsigned BitWidth,
const APInt &Demanded,
- DebugLoc dl) {
+ SDLoc dl) {
assert(Op.getNumOperands() == 2 &&
"ShrinkDemandedOp only supports binary operators!");
assert(Op.getNode()->getNumValues() == 1 &&
"ShrinkDemandedOp only supports nodes with one result!");
+ // Early return, as this function cannot handle vector types.
+ if (Op.getValueType().isVector())
+ return false;
+
// Don't do this if the node has another user, which may require the
// full value.
if (!Op.getNode()->hasOneUse())
return false;
}
-/// SimplifyDemandedBits - Look at Op. At this point, we know that only the
-/// DemandedMask bits of the result of Op are ever used downstream. If we can
-/// use this information to simplify Op, create a new simplified DAG node and
-/// return true, returning the original and new nodes in Old and New. Otherwise,
-/// analyze the expression and return a mask of KnownOne and KnownZero bits for
-/// the expression (used to simplify the caller). The KnownZero/One bits may
-/// only be accurate for those bits in the DemandedMask.
+/// Look at Op. At this point, we know that only the DemandedMask bits of the
+/// result of Op are ever used downstream. If we can use this information to
+/// simplify Op, create a new simplified DAG node and return true, returning the
+/// original and new nodes in Old and New. Otherwise, analyze the expression and
+/// return a mask of KnownOne and KnownZero bits for the expression (used to
+/// simplify the caller). The KnownZero/One bits may only be accurate for those
+/// bits in the DemandedMask.
bool TargetLowering::SimplifyDemandedBits(SDValue Op,
const APInt &DemandedMask,
APInt &KnownZero,
assert(Op.getValueType().getScalarType().getSizeInBits() == BitWidth &&
"Mask size mismatches value type size!");
APInt NewMask = DemandedMask;
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
+ auto &DL = TLO.DAG.getDataLayout();
// Don't know anything.
KnownZero = KnownOne = APInt(BitWidth, 0);
if (Depth != 0) {
// If not at the root, Just compute the KnownZero/KnownOne bits to
// simplify things downstream.
- TLO.DAG.ComputeMaskedBits(Op, KnownZero, KnownOne, Depth);
+ TLO.DAG.computeKnownBits(Op, KnownZero, KnownOne, Depth);
return false;
}
// If this is the root being simplified, allow it to have multiple uses,
if (ConstantSDNode *RHSC = dyn_cast<ConstantSDNode>(Op.getOperand(1))) {
APInt LHSZero, LHSOne;
// Do not increment Depth here; that can cause an infinite loop.
- TLO.DAG.ComputeMaskedBits(Op.getOperand(0), LHSZero, LHSOne, Depth);
+ TLO.DAG.computeKnownBits(Op.getOperand(0), LHSZero, LHSOne, Depth);
// If the LHS already has zeros where RHSC does, this and is dead.
if ((LHSZero & NewMask) == (~RHSC->getAPIntValue() & NewMask))
return TLO.CombineTo(Op, Op.getOperand(0));
return TLO.CombineTo(Op, Op.getOperand(1));
// If all of the demanded bits in the inputs are known zeros, return zero.
if ((NewMask & (KnownZero|KnownZero2)) == NewMask)
- return TLO.CombineTo(Op, TLO.DAG.getConstant(0, Op.getValueType()));
+ return TLO.CombineTo(Op, TLO.DAG.getConstant(0, dl, Op.getValueType()));
// If the RHS is a constant, see if we can simplify it.
if (TLO.ShrinkDemandedConstant(Op, ~KnownZero2 & NewMask))
return true;
// into an AND, as we know the bits will be cleared.
// e.g. (X | C1) ^ C2 --> (X | C1) & ~C2 iff (C1&C2) == C2
// NB: it is okay if more bits are known than are requested
- if ((NewMask & (KnownZero|KnownOne)) == NewMask) { // all known on one side
+ if ((NewMask & (KnownZero|KnownOne)) == NewMask) { // all known on one side
if (KnownOne == KnownOne2) { // set bits are the same on both sides
EVT VT = Op.getValueType();
- SDValue ANDC = TLO.DAG.getConstant(~KnownOne & NewMask, VT);
+ SDValue ANDC = TLO.DAG.getConstant(~KnownOne & NewMask, dl, VT);
return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::AND, dl, VT,
Op.getOperand(0), ANDC));
}
if (Expanded != C->getAPIntValue()) {
EVT VT = Op.getValueType();
SDValue New = TLO.DAG.getNode(Op.getOpcode(), dl,VT, Op.getOperand(0),
- TLO.DAG.getConstant(Expanded, VT));
+ TLO.DAG.getConstant(Expanded, dl, VT));
return TLO.CombineTo(Op, New);
}
// if it already has all the bits set, nothing to change
}
SDValue NewSA =
- TLO.DAG.getConstant(Diff, Op.getOperand(1).getValueType());
+ TLO.DAG.getConstant(Diff, dl, Op.getOperand(1).getValueType());
EVT VT = Op.getValueType();
return TLO.CombineTo(Op, TLO.DAG.getNode(Opc, dl, VT,
InOp.getOperand(0), NewSA));
unsigned InnerBits = InnerVT.getSizeInBits();
if (ShAmt < InnerBits && NewMask.lshr(InnerBits) == 0 &&
isTypeDesirableForOp(ISD::SHL, InnerVT)) {
- EVT ShTy = getShiftAmountTy(InnerVT);
+ EVT ShTy = getShiftAmountTy(InnerVT, DL);
if (!APInt(BitWidth, ShAmt).isIntN(ShTy.getSizeInBits()))
ShTy = InnerVT;
SDValue NarrowShl =
TLO.DAG.getNode(ISD::SHL, dl, InnerVT, InnerOp,
- TLO.DAG.getConstant(ShAmt, ShTy));
+ TLO.DAG.getConstant(ShAmt, dl, ShTy));
return
TLO.CombineTo(Op,
TLO.DAG.getNode(ISD::ANY_EXTEND, dl, Op.getValueType(),
NarrowShl));
}
+ // Repeat the SHL optimization above in cases where an extension
+ // intervenes: (shl (anyext (shr x, c1)), c2) to
+ // (shl (anyext x), c2-c1). This requires that the bottom c1 bits
+ // aren't demanded (as above) and that the shifted upper c1 bits of
+ // x aren't demanded.
+ if (InOp.hasOneUse() &&
+ InnerOp.getOpcode() == ISD::SRL &&
+ InnerOp.hasOneUse() &&
+ isa<ConstantSDNode>(InnerOp.getOperand(1))) {
+ uint64_t InnerShAmt = cast<ConstantSDNode>(InnerOp.getOperand(1))
+ ->getZExtValue();
+ if (InnerShAmt < ShAmt &&
+ InnerShAmt < InnerBits &&
+ NewMask.lshr(InnerBits - InnerShAmt + ShAmt) == 0 &&
+ NewMask.trunc(ShAmt) == 0) {
+ SDValue NewSA =
+ TLO.DAG.getConstant(ShAmt - InnerShAmt, dl,
+ Op.getOperand(1).getValueType());
+ EVT VT = Op.getValueType();
+ SDValue NewExt = TLO.DAG.getNode(ISD::ANY_EXTEND, dl, VT,
+ InnerOp.getOperand(0));
+ return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SHL, dl, VT,
+ NewExt, NewSA));
+ }
+ }
}
KnownZero <<= SA->getZExtValue();
if (ShAmt >= BitWidth)
break;
+ APInt InDemandedMask = (NewMask << ShAmt);
+
+ // If the shift is exact, then it does demand the low bits (and knows that
+ // they are zero).
+ if (cast<BinaryWithFlagsSDNode>(Op)->Flags.hasExact())
+ InDemandedMask |= APInt::getLowBitsSet(BitWidth, ShAmt);
+
// If this is ((X << C1) >>u ShAmt), see if we can simplify this into a
// single shift. We can do this if the top bits (which are shifted out)
// are never demanded.
}
SDValue NewSA =
- TLO.DAG.getConstant(Diff, Op.getOperand(1).getValueType());
+ TLO.DAG.getConstant(Diff, dl, Op.getOperand(1).getValueType());
return TLO.CombineTo(Op, TLO.DAG.getNode(Opc, dl, VT,
InOp.getOperand(0), NewSA));
}
}
// Compute the new bits that are at the top now.
- if (SimplifyDemandedBits(InOp, (NewMask << ShAmt),
+ if (SimplifyDemandedBits(InOp, InDemandedMask,
KnownZero, KnownOne, TLO, Depth+1))
return true;
assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?");
APInt InDemandedMask = (NewMask << ShAmt);
+ // If the shift is exact, then it does demand the low bits (and knows that
+ // they are zero).
+ if (cast<BinaryWithFlagsSDNode>(Op)->Flags.hasExact())
+ InDemandedMask |= APInt::getLowBitsSet(BitWidth, ShAmt);
+
// If any of the demanded bits are produced by the sign extension, we also
// demand the input sign bit.
APInt HighBits = APInt::getHighBitsSet(BitWidth, ShAmt);
// If the input sign bit is known to be zero, or if none of the top bits
// are demanded, turn this into an unsigned shift right.
if (KnownZero.intersects(SignBit) || (HighBits & ~NewMask) == HighBits) {
+ SDNodeFlags Flags;
+ Flags.setExact(cast<BinaryWithFlagsSDNode>(Op)->Flags.hasExact());
+ return TLO.CombineTo(Op,
+ TLO.DAG.getNode(ISD::SRL, dl, VT, Op.getOperand(0),
+ Op.getOperand(1), &Flags));
+ }
+
+ int Log2 = NewMask.exactLogBase2();
+ if (Log2 >= 0) {
+ // The bit must come from the sign.
+ SDValue NewSA =
+ TLO.DAG.getConstant(BitWidth - 1 - Log2, dl,
+ Op.getOperand(1).getValueType());
return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SRL, dl, VT,
- Op.getOperand(0),
- Op.getOperand(1)));
- } else if (KnownOne.intersects(SignBit)) { // New bits are known one.
- KnownOne |= HighBits;
+ Op.getOperand(0), NewSA));
}
+
+ if (KnownOne.intersects(SignBit))
+ // New bits are known one.
+ KnownOne |= HighBits;
}
break;
case ISD::SIGN_EXTEND_INREG: {
APInt MsbMask = APInt::getHighBitsSet(BitWidth, 1);
// If we only care about the highest bit, don't bother shifting right.
- if (MsbMask == DemandedMask) {
+ if (MsbMask == NewMask) {
unsigned ShAmt = ExVT.getScalarType().getSizeInBits();
SDValue InOp = Op.getOperand(0);
-
- // Compute the correct shift amount type, which must be getShiftAmountTy
- // for scalar types after legalization.
- EVT ShiftAmtTy = Op.getValueType();
- if (TLO.LegalTypes() && !ShiftAmtTy.isVector())
- ShiftAmtTy = getShiftAmountTy(ShiftAmtTy);
-
- SDValue ShiftAmt = TLO.DAG.getConstant(BitWidth - ShAmt, ShiftAmtTy);
- return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SHL, dl,
- Op.getValueType(), InOp, ShiftAmt));
+ unsigned VTBits = Op->getValueType(0).getScalarType().getSizeInBits();
+ bool AlreadySignExtended =
+ TLO.DAG.ComputeNumSignBits(InOp) >= VTBits-ShAmt+1;
+ // However if the input is already sign extended we expect the sign
+ // extension to be dropped altogether later and do not simplify.
+ if (!AlreadySignExtended) {
+ // Compute the correct shift amount type, which must be getShiftAmountTy
+ // for scalar types after legalization.
+ EVT ShiftAmtTy = Op.getValueType();
+ if (TLO.LegalTypes() && !ShiftAmtTy.isVector())
+ ShiftAmtTy = getShiftAmountTy(ShiftAmtTy, DL);
+
+ SDValue ShiftAmt = TLO.DAG.getConstant(BitWidth - ShAmt, dl,
+ ShiftAmtTy);
+ return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SHL, dl,
+ Op.getValueType(), InOp,
+ ShiftAmt));
+ }
}
// Sign extension. Compute the demanded bits in the result that are not
}
break;
}
+ case ISD::BUILD_PAIR: {
+ EVT HalfVT = Op.getOperand(0).getValueType();
+ unsigned HalfBitWidth = HalfVT.getScalarSizeInBits();
+
+ APInt MaskLo = NewMask.getLoBits(HalfBitWidth).trunc(HalfBitWidth);
+ APInt MaskHi = NewMask.getHiBits(HalfBitWidth).trunc(HalfBitWidth);
+
+ APInt KnownZeroLo, KnownOneLo;
+ APInt KnownZeroHi, KnownOneHi;
+
+ if (SimplifyDemandedBits(Op.getOperand(0), MaskLo, KnownZeroLo,
+ KnownOneLo, TLO, Depth + 1))
+ return true;
+
+ if (SimplifyDemandedBits(Op.getOperand(1), MaskHi, KnownZeroHi,
+ KnownOneHi, TLO, Depth + 1))
+ return true;
+
+ KnownZero = KnownZeroLo.zext(BitWidth) |
+ KnownZeroHi.zext(BitWidth).shl(HalfBitWidth);
+
+ KnownOne = KnownOneLo.zext(BitWidth) |
+ KnownOneHi.zext(BitWidth).shl(HalfBitWidth);
+ break;
+ }
case ISD::ZERO_EXTEND: {
unsigned OperandBitWidth =
Op.getOperand(0).getValueType().getScalarType().getSizeInBits();
SDValue Shift = In.getOperand(1);
if (TLO.LegalTypes()) {
uint64_t ShVal = ShAmt->getZExtValue();
- Shift =
- TLO.DAG.getConstant(ShVal, getShiftAmountTy(Op.getValueType()));
+ Shift = TLO.DAG.getConstant(ShVal, dl,
+ getShiftAmountTy(Op.getValueType(), DL));
}
APInt HighBits = APInt::getHighBitsSet(OperandBitWidth,
Op.getOperand(0).getValueType().isFloatingPoint()) {
bool OpVTLegal = isOperationLegalOrCustom(ISD::FGETSIGN, Op.getValueType());
bool i32Legal = isOperationLegalOrCustom(ISD::FGETSIGN, MVT::i32);
- if ((OpVTLegal || i32Legal) && Op.getValueType().isSimple()) {
+ if ((OpVTLegal || i32Legal) && Op.getValueType().isSimple() &&
+ Op.getOperand(0).getValueType() != MVT::f128) {
+ // Cannot eliminate/lower SHL for f128 yet.
EVT Ty = OpVTLegal ? Op.getValueType() : MVT::i32;
// Make a FGETSIGN + SHL to move the sign bit into the appropriate
// place. We expect the SHL to be eliminated by other optimizations.
if (!OpVTLegal && OpVTSizeInBits > 32)
Sign = TLO.DAG.getNode(ISD::ZERO_EXTEND, dl, Op.getValueType(), Sign);
unsigned ShVal = Op.getValueType().getSizeInBits()-1;
- SDValue ShAmt = TLO.DAG.getConstant(ShVal, Op.getValueType());
+ SDValue ShAmt = TLO.DAG.getConstant(ShVal, dl, Op.getValueType());
return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SHL, dl,
Op.getValueType(),
Sign, ShAmt));
}
// FALL THROUGH
default:
- // Just use ComputeMaskedBits to compute output bits.
- TLO.DAG.ComputeMaskedBits(Op, KnownZero, KnownOne, Depth);
+ // Just use computeKnownBits to compute output bits.
+ TLO.DAG.computeKnownBits(Op, KnownZero, KnownOne, Depth);
break;
}
// If we know the value of all of the demanded bits, return this as a
// constant.
- if ((NewMask & (KnownZero|KnownOne)) == NewMask)
- return TLO.CombineTo(Op, TLO.DAG.getConstant(KnownOne, Op.getValueType()));
+ if ((NewMask & (KnownZero|KnownOne)) == NewMask) {
+ // Avoid folding to a constant if any OpaqueConstant is involved.
+ const SDNode *N = Op.getNode();
+ for (SDNodeIterator I = SDNodeIterator::begin(N),
+ E = SDNodeIterator::end(N); I != E; ++I) {
+ SDNode *Op = *I;
+ if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Op))
+ if (C->isOpaque())
+ return false;
+ }
+ return TLO.CombineTo(Op,
+ TLO.DAG.getConstant(KnownOne, dl, Op.getValueType()));
+ }
return false;
}
-/// computeMaskedBitsForTargetNode - Determine which of the bits specified
-/// in Mask are known to be either zero or one and return them in the
-/// KnownZero/KnownOne bitsets.
-void TargetLowering::computeMaskedBitsForTargetNode(const SDValue Op,
- APInt &KnownZero,
- APInt &KnownOne,
- const SelectionDAG &DAG,
- unsigned Depth) const {
+/// Determine which of the bits specified in Mask are known to be either zero or
+/// one and return them in the KnownZero/KnownOne bitsets.
+void TargetLowering::computeKnownBitsForTargetNode(const SDValue Op,
+ APInt &KnownZero,
+ APInt &KnownOne,
+ const SelectionDAG &DAG,
+ unsigned Depth) const {
assert((Op.getOpcode() >= ISD::BUILTIN_OP_END ||
Op.getOpcode() == ISD::INTRINSIC_WO_CHAIN ||
Op.getOpcode() == ISD::INTRINSIC_W_CHAIN ||
KnownZero = KnownOne = APInt(KnownOne.getBitWidth(), 0);
}
-/// ComputeNumSignBitsForTargetNode - This method can be implemented by
-/// targets that want to expose additional information about sign bits to the
-/// DAG Combiner.
+/// This method can be implemented by targets that want to expose additional
+/// information about sign bits to the DAG Combiner.
unsigned TargetLowering::ComputeNumSignBitsForTargetNode(SDValue Op,
+ const SelectionDAG &,
unsigned Depth) const {
assert((Op.getOpcode() >= ISD::BUILTIN_OP_END ||
Op.getOpcode() == ISD::INTRINSIC_WO_CHAIN ||
return 1;
}
-/// ValueHasExactlyOneBitSet - Test if the given value is known to have exactly
-/// one bit set. This differs from ComputeMaskedBits in that it doesn't need to
-/// determine which bit is set.
-///
+/// Test if the given value is known to have exactly one bit set. This differs
+/// from computeKnownBits in that it doesn't need to determine which bit is set.
static bool ValueHasExactlyOneBitSet(SDValue Val, const SelectionDAG &DAG) {
// A left-shift of a constant one will have exactly one bit set, because
// shifting the bit off the end is undefined.
// More could be done here, though the above checks are enough
// to handle some common cases.
- // Fall back to ComputeMaskedBits to catch other known cases.
+ // Fall back to computeKnownBits to catch other known cases.
EVT OpVT = Val.getValueType();
unsigned BitWidth = OpVT.getScalarType().getSizeInBits();
APInt KnownZero, KnownOne;
- DAG.ComputeMaskedBits(Val, KnownZero, KnownOne);
+ DAG.computeKnownBits(Val, KnownZero, KnownOne);
return (KnownZero.countPopulation() == BitWidth - 1) &&
(KnownOne.countPopulation() == 1);
}
-/// SimplifySetCC - Try to simplify a setcc built with the specified operands
-/// and cc. If it is unable to simplify it, return a null SDValue.
+bool TargetLowering::isConstTrueVal(const SDNode *N) const {
+ if (!N)
+ return false;
+
+ const ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N);
+ if (!CN) {
+ const BuildVectorSDNode *BV = dyn_cast<BuildVectorSDNode>(N);
+ if (!BV)
+ return false;
+
+ BitVector UndefElements;
+ CN = BV->getConstantSplatNode(&UndefElements);
+ // Only interested in constant splats, and we don't try to handle undef
+ // elements in identifying boolean constants.
+ if (!CN || UndefElements.none())
+ return false;
+ }
+
+ switch (getBooleanContents(N->getValueType(0))) {
+ case UndefinedBooleanContent:
+ return CN->getAPIntValue()[0];
+ case ZeroOrOneBooleanContent:
+ return CN->isOne();
+ case ZeroOrNegativeOneBooleanContent:
+ return CN->isAllOnesValue();
+ }
+
+ llvm_unreachable("Invalid boolean contents");
+}
+
+bool TargetLowering::isConstFalseVal(const SDNode *N) const {
+ if (!N)
+ return false;
+
+ const ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N);
+ if (!CN) {
+ const BuildVectorSDNode *BV = dyn_cast<BuildVectorSDNode>(N);
+ if (!BV)
+ return false;
+
+ BitVector UndefElements;
+ CN = BV->getConstantSplatNode(&UndefElements);
+ // Only interested in constant splats, and we don't try to handle undef
+ // elements in identifying boolean constants.
+ if (!CN || UndefElements.none())
+ return false;
+ }
+
+ if (getBooleanContents(N->getValueType(0)) == UndefinedBooleanContent)
+ return !CN->getAPIntValue()[0];
+
+ return CN->isNullValue();
+}
+
+/// Try to simplify a setcc built with the specified operands and cc. If it is
+/// unable to simplify it, return a null SDValue.
SDValue
TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
ISD::CondCode Cond, bool foldBooleans,
- DAGCombinerInfo &DCI, DebugLoc dl) const {
+ DAGCombinerInfo &DCI, SDLoc dl) const {
SelectionDAG &DAG = DCI.DAG;
// These setcc operations always fold.
switch (Cond) {
default: break;
case ISD::SETFALSE:
- case ISD::SETFALSE2: return DAG.getConstant(0, VT);
+ case ISD::SETFALSE2: return DAG.getConstant(0, dl, VT);
case ISD::SETTRUE:
- case ISD::SETTRUE2: return DAG.getConstant(1, VT);
+ case ISD::SETTRUE2: {
+ TargetLowering::BooleanContent Cnt =
+ getBooleanContents(N0->getValueType(0));
+ return DAG.getConstant(
+ Cnt == TargetLowering::ZeroOrNegativeOneBooleanContent ? -1ULL : 1, dl,
+ VT);
+ }
}
// Ensure that the constant occurs on the RHS, and fold constant
// comparisons.
- if (isa<ConstantSDNode>(N0.getNode()))
- return DAG.getSetCC(dl, VT, N1, N0, ISD::getSetCCSwappedOperands(Cond));
+ ISD::CondCode SwappedCC = ISD::getSetCCSwappedOperands(Cond);
+ if (isa<ConstantSDNode>(N0.getNode()) &&
+ (DCI.isBeforeLegalizeOps() ||
+ isCondCodeLegal(SwappedCC, N0.getSimpleValueType())))
+ return DAG.getSetCC(dl, VT, N1, N0, SwappedCC);
- if (ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1.getNode())) {
+ if (auto *N1C = dyn_cast<ConstantSDNode>(N1.getNode())) {
const APInt &C1 = N1C->getAPIntValue();
// If the LHS is '(srl (ctlz x), 5)', the RHS is 0/1, and this is an
// (srl (ctlz x), 5) == 1 -> X == 0
Cond = ISD::SETEQ;
}
- SDValue Zero = DAG.getConstant(0, N0.getValueType());
+ SDValue Zero = DAG.getConstant(0, dl, N0.getValueType());
return DAG.getSetCC(dl, VT, N0.getOperand(0).getOperand(0),
Zero, Cond);
}
// (ctpop x) u> 1 -> (x & x-1) != 0
if ((Cond == ISD::SETULT && C1 == 2) || (Cond == ISD::SETUGT && C1 == 1)){
SDValue Sub = DAG.getNode(ISD::SUB, dl, CTVT, CTOp,
- DAG.getConstant(1, CTVT));
+ DAG.getConstant(1, dl, CTVT));
SDValue And = DAG.getNode(ISD::AND, dl, CTVT, CTOp, Sub);
ISD::CondCode CC = Cond == ISD::SETULT ? ISD::SETEQ : ISD::SETNE;
- return DAG.getSetCC(dl, VT, And, DAG.getConstant(0, CTVT), CC);
+ return DAG.getSetCC(dl, VT, And, DAG.getConstant(0, dl, CTVT), CC);
}
// TODO: (ctpop x) == 1 -> x && (x & x-1) == 0 iff ctpop is illegal.
}
// (zext x) == C --> x == (trunc C)
- if (DCI.isBeforeLegalize() && N0->hasOneUse() &&
- (Cond == ISD::SETEQ || Cond == ISD::SETNE)) {
+ // (sext x) == C --> x == (trunc C)
+ if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
+ DCI.isBeforeLegalize() && N0->hasOneUse()) {
unsigned MinBits = N0.getValueSizeInBits();
- SDValue PreZExt;
+ SDValue PreExt;
+ bool Signed = false;
if (N0->getOpcode() == ISD::ZERO_EXTEND) {
// ZExt
MinBits = N0->getOperand(0).getValueSizeInBits();
- PreZExt = N0->getOperand(0);
+ PreExt = N0->getOperand(0);
} else if (N0->getOpcode() == ISD::AND) {
// DAGCombine turns costly ZExts into ANDs
- if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(N0->getOperand(1)))
+ if (auto *C = dyn_cast<ConstantSDNode>(N0->getOperand(1)))
if ((C->getAPIntValue()+1).isPowerOf2()) {
MinBits = C->getAPIntValue().countTrailingOnes();
- PreZExt = N0->getOperand(0);
+ PreExt = N0->getOperand(0);
}
- } else if (LoadSDNode *LN0 = dyn_cast<LoadSDNode>(N0)) {
- // ZEXTLOAD
+ } else if (N0->getOpcode() == ISD::SIGN_EXTEND) {
+ // SExt
+ MinBits = N0->getOperand(0).getValueSizeInBits();
+ PreExt = N0->getOperand(0);
+ Signed = true;
+ } else if (auto *LN0 = dyn_cast<LoadSDNode>(N0)) {
+ // ZEXTLOAD / SEXTLOAD
if (LN0->getExtensionType() == ISD::ZEXTLOAD) {
MinBits = LN0->getMemoryVT().getSizeInBits();
- PreZExt = N0;
+ PreExt = N0;
+ } else if (LN0->getExtensionType() == ISD::SEXTLOAD) {
+ Signed = true;
+ MinBits = LN0->getMemoryVT().getSizeInBits();
+ PreExt = N0;
}
}
+ // Figure out how many bits we need to preserve this constant.
+ unsigned ReqdBits = Signed ?
+ C1.getBitWidth() - C1.getNumSignBits() + 1 :
+ C1.getActiveBits();
+
// Make sure we're not losing bits from the constant.
- if (MinBits < C1.getBitWidth() && MinBits >= C1.getActiveBits()) {
+ if (MinBits > 0 &&
+ MinBits < C1.getBitWidth() &&
+ MinBits >= ReqdBits) {
EVT MinVT = EVT::getIntegerVT(*DAG.getContext(), MinBits);
if (isTypeDesirableForOp(ISD::SETCC, MinVT)) {
// Will get folded away.
- SDValue Trunc = DAG.getNode(ISD::TRUNCATE, dl, MinVT, PreZExt);
- SDValue C = DAG.getConstant(C1.trunc(MinBits), MinVT);
+ SDValue Trunc = DAG.getNode(ISD::TRUNCATE, dl, MinVT, PreExt);
+ SDValue C = DAG.getConstant(C1.trunc(MinBits), dl, MinVT);
return DAG.getSetCC(dl, VT, Trunc, C, Cond);
}
}
// the test is for equality or unsigned, and all 1 bits of the const are
// in the same partial word, see if we can shorten the load.
if (DCI.isBeforeLegalize() &&
+ !ISD::isSignedIntSetCC(Cond) &&
N0.getOpcode() == ISD::AND && C1 == 0 &&
N0.getNode()->hasOneUse() &&
isa<LoadSDNode>(N0.getOperand(0)) &&
APInt newMask = APInt::getLowBitsSet(maskWidth, width);
for (unsigned offset=0; offset<origWidth/width; offset++) {
if ((newMask & Mask) == Mask) {
- if (!getDataLayout()->isLittleEndian())
+ if (!DAG.getDataLayout().isLittleEndian())
bestOffset = (origWidth/width - offset - 1) * (width/8);
else
bestOffset = (uint64_t)offset * (width/8);
SDValue Ptr = Lod->getBasePtr();
if (bestOffset != 0)
Ptr = DAG.getNode(ISD::ADD, dl, PtrType, Lod->getBasePtr(),
- DAG.getConstant(bestOffset, PtrType));
+ DAG.getConstant(bestOffset, dl, PtrType));
unsigned NewAlign = MinAlign(Lod->getAlignment(), bestOffset);
SDValue NewLoad = DAG.getLoad(newVT, dl, Lod->getChain(), Ptr,
Lod->getPointerInfo().getWithOffset(bestOffset),
return DAG.getSetCC(dl, VT,
DAG.getNode(ISD::AND, dl, newVT, NewLoad,
DAG.getConstant(bestMask.trunc(bestWidth),
- newVT)),
- DAG.getConstant(0LL, newVT), Cond);
+ dl, newVT)),
+ DAG.getConstant(0LL, dl, newVT), Cond);
}
}
}
switch (Cond) {
case ISD::SETUGT:
case ISD::SETUGE:
- case ISD::SETEQ: return DAG.getConstant(0, VT);
+ case ISD::SETEQ: return DAG.getConstant(0, dl, VT);
case ISD::SETULT:
case ISD::SETULE:
- case ISD::SETNE: return DAG.getConstant(1, VT);
+ case ISD::SETNE: return DAG.getConstant(1, dl, VT);
case ISD::SETGT:
case ISD::SETGE:
// True if the sign bit of C1 is set.
- return DAG.getConstant(C1.isNegative(), VT);
+ return DAG.getConstant(C1.isNegative(), dl, VT);
case ISD::SETLT:
case ISD::SETLE:
// True if the sign bit of C1 isn't set.
- return DAG.getConstant(C1.isNonNegative(), VT);
+ return DAG.getConstant(C1.isNonNegative(), dl, VT);
default:
break;
}
EVT newVT = N0.getOperand(0).getValueType();
if (DCI.isBeforeLegalizeOps() ||
(isOperationLegal(ISD::SETCC, newVT) &&
- getCondCodeAction(Cond, newVT.getSimpleVT())==Legal))
- return DAG.getSetCC(dl, VT, N0.getOperand(0),
- DAG.getConstant(C1.trunc(InSize), newVT),
- Cond);
+ getCondCodeAction(Cond, newVT.getSimpleVT()) == Legal)) {
+ EVT NewSetCCVT =
+ getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), newVT);
+ SDValue NewConst = DAG.getConstant(C1.trunc(InSize), dl, newVT);
+
+ SDValue NewSetCC = DAG.getSetCC(dl, NewSetCCVT, N0.getOperand(0),
+ NewConst, Cond);
+ return DAG.getBoolExtOrTrunc(NewSetCC, dl, VT, N0.getValueType());
+ }
break;
}
default:
// If the constant doesn't fit into the number of bits for the source of
// the sign extension, it is impossible for both sides to be equal.
if (C1.getMinSignedBits() > ExtSrcTyBits)
- return DAG.getConstant(Cond == ISD::SETNE, VT);
+ return DAG.getConstant(Cond == ISD::SETNE, dl, VT);
SDValue ZextOp;
EVT Op0Ty = N0.getOperand(0).getValueType();
} else {
APInt Imm = APInt::getLowBitsSet(ExtDstTyBits, ExtSrcTyBits);
ZextOp = DAG.getNode(ISD::AND, dl, Op0Ty, N0.getOperand(0),
- DAG.getConstant(Imm, Op0Ty));
+ DAG.getConstant(Imm, dl, Op0Ty));
}
if (!DCI.isCalledByLegalizer())
DCI.AddToWorklist(ZextOp.getNode());
DAG.getConstant(C1 & APInt::getLowBitsSet(
ExtDstTyBits,
ExtSrcTyBits),
- ExtDstTy),
+ dl, ExtDstTy),
Cond);
} else if ((N1C->isNullValue() || N1C->getAPIntValue() == 1) &&
(Cond == ISD::SETEQ || Cond == ISD::SETNE)) {
ISD::CondCode CC = cast<CondCodeSDNode>(N0.getOperand(2))->get();
CC = ISD::getSetCCInverse(CC,
N0.getOperand(0).getValueType().isInteger());
- return DAG.getSetCC(dl, VT, N0.getOperand(0), N0.getOperand(1), CC);
+ if (DCI.isBeforeLegalizeOps() ||
+ isCondCodeLegal(CC, N0.getOperand(0).getSimpleValueType()))
+ return DAG.getSetCC(dl, VT, N0.getOperand(0), N0.getOperand(1), CC);
}
if ((N0.getOpcode() == ISD::XOR ||
}
} else if (N1C->getAPIntValue() == 1 &&
(VT == MVT::i1 ||
- getBooleanContents(false) == ZeroOrOneBooleanContent)) {
+ getBooleanContents(N0->getValueType(0)) ==
+ ZeroOrOneBooleanContent)) {
SDValue Op0 = N0;
if (Op0.getOpcode() == ISD::TRUNCATE)
Op0 = Op0.getOperand(0);
if (Op0.getValueType().bitsGT(VT))
Op0 = DAG.getNode(ISD::AND, dl, VT,
DAG.getNode(ISD::TRUNCATE, dl, VT, Op0.getOperand(0)),
- DAG.getConstant(1, VT));
+ DAG.getConstant(1, dl, VT));
else if (Op0.getValueType().bitsLT(VT))
Op0 = DAG.getNode(ISD::AND, dl, VT,
DAG.getNode(ISD::ANY_EXTEND, dl, VT, Op0.getOperand(0)),
- DAG.getConstant(1, VT));
+ DAG.getConstant(1, dl, VT));
return DAG.getSetCC(dl, VT, Op0,
- DAG.getConstant(0, Op0.getValueType()),
+ DAG.getConstant(0, dl, Op0.getValueType()),
Cond == ISD::SETEQ ? ISD::SETNE : ISD::SETEQ);
}
if (Op0.getOpcode() == ISD::AssertZext &&
cast<VTSDNode>(Op0.getOperand(1))->getVT() == MVT::i1)
return DAG.getSetCC(dl, VT, Op0,
- DAG.getConstant(0, Op0.getValueType()),
+ DAG.getConstant(0, dl, Op0.getValueType()),
Cond == ISD::SETEQ ? ISD::SETNE : ISD::SETEQ);
}
}
// Canonicalize GE/LE comparisons to use GT/LT comparisons.
if (Cond == ISD::SETGE || Cond == ISD::SETUGE) {
- if (C1 == MinVal) return DAG.getConstant(1, VT); // X >= MIN --> true
- // X >= C0 --> X > (C0-1)
- return DAG.getSetCC(dl, VT, N0,
- DAG.getConstant(C1-1, N1.getValueType()),
- (Cond == ISD::SETGE) ? ISD::SETGT : ISD::SETUGT);
+ if (C1 == MinVal) return DAG.getConstant(1, dl, VT); // X >= MIN --> true
+ // X >= C0 --> X > (C0 - 1)
+ APInt C = C1 - 1;
+ ISD::CondCode NewCC = (Cond == ISD::SETGE) ? ISD::SETGT : ISD::SETUGT;
+ if ((DCI.isBeforeLegalizeOps() ||
+ isCondCodeLegal(NewCC, VT.getSimpleVT())) &&
+ (!N1C->isOpaque() || (N1C->isOpaque() && C.getBitWidth() <= 64 &&
+ isLegalICmpImmediate(C.getSExtValue())))) {
+ return DAG.getSetCC(dl, VT, N0,
+ DAG.getConstant(C, dl, N1.getValueType()),
+ NewCC);
+ }
}
if (Cond == ISD::SETLE || Cond == ISD::SETULE) {
- if (C1 == MaxVal) return DAG.getConstant(1, VT); // X <= MAX --> true
- // X <= C0 --> X < (C0+1)
- return DAG.getSetCC(dl, VT, N0,
- DAG.getConstant(C1+1, N1.getValueType()),
- (Cond == ISD::SETLE) ? ISD::SETLT : ISD::SETULT);
+ if (C1 == MaxVal) return DAG.getConstant(1, dl, VT); // X <= MAX --> true
+ // X <= C0 --> X < (C0 + 1)
+ APInt C = C1 + 1;
+ ISD::CondCode NewCC = (Cond == ISD::SETLE) ? ISD::SETLT : ISD::SETULT;
+ if ((DCI.isBeforeLegalizeOps() ||
+ isCondCodeLegal(NewCC, VT.getSimpleVT())) &&
+ (!N1C->isOpaque() || (N1C->isOpaque() && C.getBitWidth() <= 64 &&
+ isLegalICmpImmediate(C.getSExtValue())))) {
+ return DAG.getSetCC(dl, VT, N0,
+ DAG.getConstant(C, dl, N1.getValueType()),
+ NewCC);
+ }
}
if ((Cond == ISD::SETLT || Cond == ISD::SETULT) && C1 == MinVal)
- return DAG.getConstant(0, VT); // X < MIN --> false
+ return DAG.getConstant(0, dl, VT); // X < MIN --> false
if ((Cond == ISD::SETGE || Cond == ISD::SETUGE) && C1 == MinVal)
- return DAG.getConstant(1, VT); // X >= MIN --> true
+ return DAG.getConstant(1, dl, VT); // X >= MIN --> true
if ((Cond == ISD::SETGT || Cond == ISD::SETUGT) && C1 == MaxVal)
- return DAG.getConstant(0, VT); // X > MAX --> false
+ return DAG.getConstant(0, dl, VT); // X > MAX --> false
if ((Cond == ISD::SETLE || Cond == ISD::SETULE) && C1 == MaxVal)
- return DAG.getConstant(1, VT); // X <= MAX --> true
+ return DAG.getConstant(1, dl, VT); // X <= MAX --> true
// Canonicalize setgt X, Min --> setne X, Min
if ((Cond == ISD::SETGT || Cond == ISD::SETUGT) && C1 == MinVal)
// If we have setult X, 1, turn it into seteq X, 0
if ((Cond == ISD::SETLT || Cond == ISD::SETULT) && C1 == MinVal+1)
return DAG.getSetCC(dl, VT, N0,
- DAG.getConstant(MinVal, N0.getValueType()),
+ DAG.getConstant(MinVal, dl, N0.getValueType()),
ISD::SETEQ);
// If we have setugt X, Max-1, turn it into seteq X, Max
if ((Cond == ISD::SETGT || Cond == ISD::SETUGT) && C1 == MaxVal-1)
return DAG.getSetCC(dl, VT, N0,
- DAG.getConstant(MaxVal, N0.getValueType()),
+ DAG.getConstant(MaxVal, dl, N0.getValueType()),
ISD::SETEQ);
// If we have "setcc X, C0", check to see if we can shrink the immediate
if (Cond == ISD::SETUGT &&
C1 == APInt::getSignedMaxValue(OperandBitSize))
return DAG.getSetCC(dl, VT, N0,
- DAG.getConstant(0, N1.getValueType()),
+ DAG.getConstant(0, dl, N1.getValueType()),
ISD::SETLT);
// SETULT X, SINTMIN -> SETGT X, -1
if (Cond == ISD::SETULT &&
C1 == APInt::getSignedMinValue(OperandBitSize)) {
SDValue ConstMinusOne =
- DAG.getConstant(APInt::getAllOnesValue(OperandBitSize),
+ DAG.getConstant(APInt::getAllOnesValue(OperandBitSize), dl,
N1.getValueType());
return DAG.getSetCC(dl, VT, N0, ConstMinusOne, ISD::SETGT);
}
if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
(VT == N0.getValueType() ||
(isTypeLegal(VT) && VT.bitsLE(N0.getValueType()))) &&
- N0.getOpcode() == ISD::AND)
- if (ConstantSDNode *AndRHS =
- dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
- EVT ShiftTy = DCI.isBeforeLegalizeOps() ?
- getPointerTy() : getShiftAmountTy(N0.getValueType());
+ N0.getOpcode() == ISD::AND) {
+ auto &DL = DAG.getDataLayout();
+ if (auto *AndRHS = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
+ EVT ShiftTy = DCI.isBeforeLegalize()
+ ? getPointerTy(DL)
+ : getShiftAmountTy(N0.getValueType(), DL);
if (Cond == ISD::SETNE && C1 == 0) {// (X & 8) != 0 --> (X & 8) >> 3
// Perform the xform if the AND RHS is a single bit.
if (AndRHS->getAPIntValue().isPowerOf2()) {
return DAG.getNode(ISD::TRUNCATE, dl, VT,
DAG.getNode(ISD::SRL, dl, N0.getValueType(), N0,
- DAG.getConstant(AndRHS->getAPIntValue().logBase2(), ShiftTy)));
+ DAG.getConstant(AndRHS->getAPIntValue().logBase2(), dl,
+ ShiftTy)));
}
} else if (Cond == ISD::SETEQ && C1 == AndRHS->getAPIntValue()) {
// (X & 8) == 8 --> (X & 8) >> 3
if (C1.isPowerOf2()) {
return DAG.getNode(ISD::TRUNCATE, dl, VT,
DAG.getNode(ISD::SRL, dl, N0.getValueType(), N0,
- DAG.getConstant(C1.logBase2(), ShiftTy)));
+ DAG.getConstant(C1.logBase2(), dl,
+ ShiftTy)));
}
}
}
+ }
if (C1.getMinSignedBits() <= 64 &&
!isLegalICmpImmediate(C1.getSExtValue())) {
// (X & -256) == 256 -> (X >> 8) == 1
if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
N0.getOpcode() == ISD::AND && N0.hasOneUse()) {
- if (ConstantSDNode *AndRHS =
- dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
+ if (auto *AndRHS = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
const APInt &AndRHSC = AndRHS->getAPIntValue();
if ((-AndRHSC).isPowerOf2() && (AndRHSC & C1) == C1) {
unsigned ShiftBits = AndRHSC.countTrailingZeros();
- EVT ShiftTy = DCI.isBeforeLegalizeOps() ?
- getPointerTy() : getShiftAmountTy(N0.getValueType());
+ auto &DL = DAG.getDataLayout();
+ EVT ShiftTy = DCI.isBeforeLegalize()
+ ? getPointerTy(DL)
+ : getShiftAmountTy(N0.getValueType(), DL);
EVT CmpTy = N0.getValueType();
SDValue Shift = DAG.getNode(ISD::SRL, dl, CmpTy, N0.getOperand(0),
- DAG.getConstant(ShiftBits, ShiftTy));
- SDValue CmpRHS = DAG.getConstant(C1.lshr(ShiftBits), CmpTy);
+ DAG.getConstant(ShiftBits, dl,
+ ShiftTy));
+ SDValue CmpRHS = DAG.getConstant(C1.lshr(ShiftBits), dl, CmpTy);
return DAG.getSetCC(dl, VT, Shift, CmpRHS, Cond);
}
}
ShiftBits = C1.countTrailingZeros();
}
NewC = NewC.lshr(ShiftBits);
- if (ShiftBits && isLegalICmpImmediate(NewC.getSExtValue())) {
- EVT ShiftTy = DCI.isBeforeLegalizeOps() ?
- getPointerTy() : getShiftAmountTy(N0.getValueType());
+ if (ShiftBits && NewC.getMinSignedBits() <= 64 &&
+ isLegalICmpImmediate(NewC.getSExtValue())) {
+ auto &DL = DAG.getDataLayout();
+ EVT ShiftTy = DCI.isBeforeLegalize()
+ ? getPointerTy(DL)
+ : getShiftAmountTy(N0.getValueType(), DL);
EVT CmpTy = N0.getValueType();
SDValue Shift = DAG.getNode(ISD::SRL, dl, CmpTy, N0,
- DAG.getConstant(ShiftBits, ShiftTy));
- SDValue CmpRHS = DAG.getConstant(NewC, CmpTy);
+ DAG.getConstant(ShiftBits, dl, ShiftTy));
+ SDValue CmpRHS = DAG.getConstant(NewC, dl, CmpTy);
return DAG.getSetCC(dl, VT, Shift, CmpRHS, NewCond);
}
}
// Constant fold or commute setcc.
SDValue O = DAG.FoldSetCC(VT, N0, N1, Cond, dl);
if (O.getNode()) return O;
- } else if (ConstantFPSDNode *CFP = dyn_cast<ConstantFPSDNode>(N1.getNode())) {
+ } else if (auto *CFP = dyn_cast<ConstantFPSDNode>(N1.getNode())) {
// If the RHS of an FP comparison is a constant, simplify it away in
// some cases.
if (CFP->getValueAPF().isNaN()) {
switch (ISD::getUnorderedFlavor(Cond)) {
default: llvm_unreachable("Unknown flavor!");
case 0: // Known false.
- return DAG.getConstant(0, VT);
+ return DAG.getConstant(0, dl, VT);
case 1: // Known true.
- return DAG.getConstant(1, VT);
+ return DAG.getConstant(1, dl, VT);
case 2: // Undefined.
return DAG.getUNDEF(VT);
}
// The sext(setcc()) => setcc() optimization relies on the appropriate
// constant being emitted.
uint64_t EqVal = 0;
- switch (getBooleanContents(N0.getValueType().isVector())) {
+ switch (getBooleanContents(N0.getValueType())) {
case UndefinedBooleanContent:
case ZeroOrOneBooleanContent:
EqVal = ISD::isTrueWhenEqual(Cond);
// We can always fold X == X for integer setcc's.
if (N0.getValueType().isInteger()) {
- return DAG.getConstant(EqVal, VT);
+ return DAG.getConstant(EqVal, dl, VT);
}
unsigned UOF = ISD::getUnorderedFlavor(Cond);
if (UOF == 2) // FP operators that are undefined on NaNs.
- return DAG.getConstant(EqVal, VT);
+ return DAG.getConstant(EqVal, dl, VT);
if (UOF == unsigned(ISD::isTrueWhenEqual(Cond)))
- return DAG.getConstant(EqVal, VT);
+ return DAG.getConstant(EqVal, dl, VT);
// Otherwise, we can't fold it. However, we can simplify it to SETUO/SETO
// if it is not already.
ISD::CondCode NewCond = UOF == 0 ? ISD::SETO : ISD::SETUO;
// to be careful about increasing register pressure needlessly.
bool LegalRHSImm = false;
- if (ConstantSDNode *RHSC = dyn_cast<ConstantSDNode>(N1)) {
- if (ConstantSDNode *LHSR = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
+ if (auto *RHSC = dyn_cast<ConstantSDNode>(N1)) {
+ if (auto *LHSR = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
// Turn (X+C1) == C2 --> X == C2-C1
if (N0.getOpcode() == ISD::ADD && N0.getNode()->hasOneUse()) {
return DAG.getSetCC(dl, VT, N0.getOperand(0),
DAG.getConstant(RHSC->getAPIntValue()-
LHSR->getAPIntValue(),
- N0.getValueType()), Cond);
+ dl, N0.getValueType()), Cond);
}
// Turn (X^C1) == C2 into X == C1^C2 iff X&~C1 = 0.
DAG.getSetCC(dl, VT, N0.getOperand(0),
DAG.getConstant(LHSR->getAPIntValue() ^
RHSC->getAPIntValue(),
- N0.getValueType()),
+ dl, N0.getValueType()),
Cond);
}
// Turn (C1-X) == C2 --> X == C1-C2
- if (ConstantSDNode *SUBC = dyn_cast<ConstantSDNode>(N0.getOperand(0))) {
+ if (auto *SUBC = dyn_cast<ConstantSDNode>(N0.getOperand(0))) {
if (N0.getOpcode() == ISD::SUB && N0.getNode()->hasOneUse()) {
return
DAG.getSetCC(dl, VT, N0.getOperand(1),
DAG.getConstant(SUBC->getAPIntValue() -
RHSC->getAPIntValue(),
- N0.getValueType()),
+ dl, N0.getValueType()),
Cond);
}
}
if (!LegalRHSImm || N0.getNode()->hasOneUse()) {
if (N0.getOperand(0) == N1)
return DAG.getSetCC(dl, VT, N0.getOperand(1),
- DAG.getConstant(0, N0.getValueType()), Cond);
+ DAG.getConstant(0, dl, N0.getValueType()), Cond);
if (N0.getOperand(1) == N1) {
if (DAG.isCommutativeBinOp(N0.getOpcode()))
return DAG.getSetCC(dl, VT, N0.getOperand(0),
- DAG.getConstant(0, N0.getValueType()), Cond);
+ DAG.getConstant(0, dl, N0.getValueType()),
+ Cond);
if (N0.getNode()->hasOneUse()) {
assert(N0.getOpcode() == ISD::SUB && "Unexpected operation!");
+ auto &DL = DAG.getDataLayout();
// (Z-X) == X --> Z == X<<1
- SDValue SH = DAG.getNode(ISD::SHL, dl, N1.getValueType(), N1,
- DAG.getConstant(1, getShiftAmountTy(N1.getValueType())));
+ SDValue SH = DAG.getNode(
+ ISD::SHL, dl, N1.getValueType(), N1,
+ DAG.getConstant(1, dl,
+ getShiftAmountTy(N1.getValueType(), DL)));
if (!DCI.isCalledByLegalizer())
DCI.AddToWorklist(SH.getNode());
return DAG.getSetCC(dl, VT, N0.getOperand(0), SH, Cond);
// Simplify X == (X+Z) --> Z == 0
if (N1.getOperand(0) == N0)
return DAG.getSetCC(dl, VT, N1.getOperand(1),
- DAG.getConstant(0, N1.getValueType()), Cond);
+ DAG.getConstant(0, dl, N1.getValueType()), Cond);
if (N1.getOperand(1) == N0) {
if (DAG.isCommutativeBinOp(N1.getOpcode()))
return DAG.getSetCC(dl, VT, N1.getOperand(0),
- DAG.getConstant(0, N1.getValueType()), Cond);
+ DAG.getConstant(0, dl, N1.getValueType()), Cond);
if (N1.getNode()->hasOneUse()) {
assert(N1.getOpcode() == ISD::SUB && "Unexpected operation!");
+ auto &DL = DAG.getDataLayout();
// X == (Z-X) --> X<<1 == Z
- SDValue SH = DAG.getNode(ISD::SHL, dl, N1.getValueType(), N0,
- DAG.getConstant(1, getShiftAmountTy(N0.getValueType())));
+ SDValue SH = DAG.getNode(
+ ISD::SHL, dl, N1.getValueType(), N0,
+ DAG.getConstant(1, dl, getShiftAmountTy(N0.getValueType(), DL)));
if (!DCI.isCalledByLegalizer())
DCI.AddToWorklist(SH.getNode());
return DAG.getSetCC(dl, VT, SH, N1.getOperand(0), Cond);
if (N0.getOperand(0) == N1 || N0.getOperand(1) == N1) {
if (ValueHasExactlyOneBitSet(N1, DAG)) {
Cond = ISD::getSetCCInverse(Cond, /*isInteger=*/true);
- SDValue Zero = DAG.getConstant(0, N1.getValueType());
- return DAG.getSetCC(dl, VT, N0, Zero, Cond);
+ if (DCI.isBeforeLegalizeOps() ||
+ isCondCodeLegal(Cond, N0.getSimpleValueType())) {
+ SDValue Zero = DAG.getConstant(0, dl, N1.getValueType());
+ return DAG.getSetCC(dl, VT, N0, Zero, Cond);
+ }
}
}
if (N1.getOpcode() == ISD::AND)
if (N1.getOperand(0) == N0 || N1.getOperand(1) == N0) {
if (ValueHasExactlyOneBitSet(N0, DAG)) {
Cond = ISD::getSetCCInverse(Cond, /*isInteger=*/true);
- SDValue Zero = DAG.getConstant(0, N0.getValueType());
- return DAG.getSetCC(dl, VT, N1, Zero, Cond);
+ if (DCI.isBeforeLegalizeOps() ||
+ isCondCodeLegal(Cond, N1.getSimpleValueType())) {
+ SDValue Zero = DAG.getConstant(0, dl, N0.getValueType());
+ return DAG.getSetCC(dl, VT, N1, Zero, Cond);
+ }
}
}
}
return SDValue();
}
-/// isGAPlusOffset - Returns true (and the GlobalValue and the offset) if the
-/// node is a GlobalAddress + offset.
+/// Returns true (and the GlobalValue and the offset) if the node is a
+/// GlobalAddress + offset.
bool TargetLowering::isGAPlusOffset(SDNode *N, const GlobalValue *&GA,
int64_t &Offset) const {
- if (isa<GlobalAddressSDNode>(N)) {
- GlobalAddressSDNode *GASD = cast<GlobalAddressSDNode>(N);
+ if (auto *GASD = dyn_cast<GlobalAddressSDNode>(N)) {
GA = GASD->getGlobal();
Offset += GASD->getOffset();
return true;
SDValue N1 = N->getOperand(0);
SDValue N2 = N->getOperand(1);
if (isGAPlusOffset(N1.getNode(), GA, Offset)) {
- ConstantSDNode *V = dyn_cast<ConstantSDNode>(N2);
- if (V) {
+ if (auto *V = dyn_cast<ConstantSDNode>(N2)) {
Offset += V->getSExtValue();
return true;
}
} else if (isGAPlusOffset(N2.getNode(), GA, Offset)) {
- ConstantSDNode *V = dyn_cast<ConstantSDNode>(N1);
- if (V) {
+ if (auto *V = dyn_cast<ConstantSDNode>(N1)) {
Offset += V->getSExtValue();
return true;
}
return false;
}
-
-SDValue TargetLowering::
-PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const {
+SDValue TargetLowering::PerformDAGCombine(SDNode *N,
+ DAGCombinerInfo &DCI) const {
// Default implementation: no optimization.
return SDValue();
}
// Inline Assembler Implementation Methods
//===----------------------------------------------------------------------===//
-
TargetLowering::ConstraintType
-TargetLowering::getConstraintType(const std::string &Constraint) const {
+TargetLowering::getConstraintType(StringRef Constraint) const {
unsigned S = Constraint.size();
if (S == 1) {
}
if (S > 1 && Constraint[0] == '{' && Constraint[S-1] == '}') {
- if (S == 8 && !Constraint.compare(1, 6, "memory", 6)) // "{memory}"
+ if (S == 8 && Constraint.substr(1, 6) == "memory") // "{memory}"
return C_Memory;
return C_Register;
}
return C_Unknown;
}
-/// LowerXConstraint - try to replace an X constraint, which matches anything,
-/// with another that has more specific requirements based on the type of the
-/// corresponding operand.
+/// Try to replace an X constraint, which matches anything, with another that
+/// has more specific requirements based on the type of the corresponding
+/// operand.
const char *TargetLowering::LowerXConstraint(EVT ConstraintVT) const{
if (ConstraintVT.isInteger())
return "r";
if (ConstraintVT.isFloatingPoint())
return "f"; // works for many targets
- return 0;
+ return nullptr;
}
-/// LowerAsmOperandForConstraint - Lower the specified operand into the Ops
-/// vector. If it is invalid, don't add anything to Ops.
+/// Lower the specified operand into the Ops vector.
+/// If it is invalid, don't add anything to Ops.
void TargetLowering::LowerAsmOperandForConstraint(SDValue Op,
std::string &Constraint,
std::vector<SDValue> &Ops,
if (Op.getOpcode() == ISD::ADD) {
C = dyn_cast<ConstantSDNode>(Op.getOperand(1));
GA = dyn_cast<GlobalAddressSDNode>(Op.getOperand(0));
- if (C == 0 || GA == 0) {
+ if (!C || !GA) {
C = dyn_cast<ConstantSDNode>(Op.getOperand(0));
GA = dyn_cast<GlobalAddressSDNode>(Op.getOperand(1));
}
- if (C == 0 || GA == 0)
- C = 0, GA = 0;
+ if (!C || !GA)
+ C = nullptr, GA = nullptr;
}
// If we find a valid operand, map to the TargetXXX version so that the
int64_t Offs = GA->getOffset();
if (C) Offs += C->getZExtValue();
Ops.push_back(DAG.getTargetGlobalAddress(GA->getGlobal(),
- C ? C->getDebugLoc() : DebugLoc(),
+ C ? SDLoc(C) : SDLoc(),
Op.getValueType(), Offs));
- return;
}
+ return;
}
if (C) { // just C, no GV.
// Simple constants are not allowed for 's'.
// now; without this it would get ZExt'd later in
// ScheduleDAGSDNodes::EmitNode, which is very generic.
Ops.push_back(DAG.getTargetConstant(C->getAPIntValue().getSExtValue(),
- MVT::i64));
- return;
+ SDLoc(C), MVT::i64));
}
+ return;
}
break;
}
}
}
-std::pair<unsigned, const TargetRegisterClass*> TargetLowering::
-getRegForInlineAsmConstraint(const std::string &Constraint,
- EVT VT) const {
- if (Constraint[0] != '{')
- return std::make_pair(0u, static_cast<TargetRegisterClass*>(0));
+std::pair<unsigned, const TargetRegisterClass *>
+TargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *RI,
+ StringRef Constraint,
+ MVT VT) const {
+ if (Constraint.empty() || Constraint[0] != '{')
+ return std::make_pair(0u, static_cast<TargetRegisterClass*>(nullptr));
assert(*(Constraint.end()-1) == '}' && "Not a brace enclosed constraint?");
// Remove the braces from around the name.
StringRef RegName(Constraint.data()+1, Constraint.size()-2);
std::pair<unsigned, const TargetRegisterClass*> R =
- std::make_pair(0u, static_cast<const TargetRegisterClass*>(0));
+ std::make_pair(0u, static_cast<const TargetRegisterClass*>(nullptr));
// Figure out which register class contains this reg.
- const TargetRegisterInfo *RI = getTargetMachine().getRegisterInfo();
for (TargetRegisterInfo::regclass_iterator RCI = RI->regclass_begin(),
E = RI->regclass_end(); RCI != E; ++RCI) {
const TargetRegisterClass *RC = *RCI;
//===----------------------------------------------------------------------===//
// Constraint Selection.
-/// isMatchingInputConstraint - Return true of this is an input operand that is
-/// a matching constraint like "4".
+/// Return true of this is an input operand that is a matching constraint like
+/// "4".
bool TargetLowering::AsmOperandInfo::isMatchingInputConstraint() const {
assert(!ConstraintCode.empty() && "No known constraint!");
return isdigit(static_cast<unsigned char>(ConstraintCode[0]));
}
-/// getMatchedOperand - If this is an input matching constraint, this method
-/// returns the output operand it matches.
+/// If this is an input matching constraint, this method returns the output
+/// operand it matches.
unsigned TargetLowering::AsmOperandInfo::getMatchedOperand() const {
assert(!ConstraintCode.empty() && "No known constraint!");
return atoi(ConstraintCode.c_str());
}
-
-/// ParseConstraints - Split up the constraint string from the inline
-/// assembly value into the specific constraints and their prefixes,
-/// and also tie in the associated operand values.
+/// Split up the constraint string from the inline assembly value into the
+/// specific constraints and their prefixes, and also tie in the associated
+/// operand values.
/// If this returns an empty vector, and if the constraint string itself
/// isn't empty, there was an error parsing.
-TargetLowering::AsmOperandInfoVector TargetLowering::ParseConstraints(
- ImmutableCallSite CS) const {
- /// ConstraintOperands - Information about all of the constraints.
+TargetLowering::AsmOperandInfoVector
+TargetLowering::ParseConstraints(const DataLayout &DL,
+ const TargetRegisterInfo *TRI,
+ ImmutableCallSite CS) const {
+ /// Information about all of the constraints.
AsmOperandInfoVector ConstraintOperands;
const InlineAsm *IA = cast<InlineAsm>(CS.getCalledValue());
unsigned maCount = 0; // Largest number of multiple alternative constraints.
// Do a prepass over the constraints, canonicalizing them, and building up the
// ConstraintOperands list.
- InlineAsm::ConstraintInfoVector
- ConstraintInfos = IA->ParseConstraints();
-
unsigned ArgNo = 0; // ArgNo - The argument of the CallInst.
unsigned ResNo = 0; // ResNo - The result number of the next output.
- for (unsigned i = 0, e = ConstraintInfos.size(); i != e; ++i) {
- ConstraintOperands.push_back(AsmOperandInfo(ConstraintInfos[i]));
+ for (InlineAsm::ConstraintInfo &CI : IA->ParseConstraints()) {
+ ConstraintOperands.emplace_back(std::move(CI));
AsmOperandInfo &OpInfo = ConstraintOperands.back();
// Update multiple alternative constraint count.
assert(!CS.getType()->isVoidTy() &&
"Bad inline asm!");
if (StructType *STy = dyn_cast<StructType>(CS.getType())) {
- OpInfo.ConstraintVT = getSimpleValueType(STy->getElementType(ResNo));
+ OpInfo.ConstraintVT =
+ getSimpleValueType(DL, STy->getElementType(ResNo));
} else {
assert(ResNo == 0 && "Asm only has one result!");
- OpInfo.ConstraintVT = getSimpleValueType(CS.getType());
+ OpInfo.ConstraintVT = getSimpleValueType(DL, CS.getType());
}
++ResNo;
break;
// If OpTy is not a single value, it may be a struct/union that we
// can tile with integers.
if (!OpTy->isSingleValueType() && OpTy->isSized()) {
- unsigned BitSize = getDataLayout()->getTypeSizeInBits(OpTy);
+ unsigned BitSize = DL.getTypeSizeInBits(OpTy);
switch (BitSize) {
default: break;
case 1:
break;
}
} else if (PointerType *PT = dyn_cast<PointerType>(OpTy)) {
- OpInfo.ConstraintVT = MVT::getIntegerVT(
- 8*getDataLayout()->getPointerSize(PT->getAddressSpace()));
+ unsigned PtrSize = DL.getPointerSizeInBits(PT->getAddressSpace());
+ OpInfo.ConstraintVT = MVT::getIntegerVT(PtrSize);
} else {
OpInfo.ConstraintVT = MVT::getVT(OpTy, true);
}
}
// If we have multiple alternative constraints, select the best alternative.
- if (ConstraintInfos.size()) {
+ if (!ConstraintOperands.empty()) {
if (maCount) {
unsigned bestMAIndex = 0;
int bestWeight = -1;
AsmOperandInfo &Input = ConstraintOperands[OpInfo.MatchingInput];
if (OpInfo.ConstraintVT != Input.ConstraintVT) {
- std::pair<unsigned, const TargetRegisterClass*> MatchRC =
- getRegForInlineAsmConstraint(OpInfo.ConstraintCode,
- OpInfo.ConstraintVT);
- std::pair<unsigned, const TargetRegisterClass*> InputRC =
- getRegForInlineAsmConstraint(Input.ConstraintCode,
- Input.ConstraintVT);
+ std::pair<unsigned, const TargetRegisterClass *> MatchRC =
+ getRegForInlineAsmConstraint(TRI, OpInfo.ConstraintCode,
+ OpInfo.ConstraintVT);
+ std::pair<unsigned, const TargetRegisterClass *> InputRC =
+ getRegForInlineAsmConstraint(TRI, Input.ConstraintCode,
+ Input.ConstraintVT);
if ((OpInfo.ConstraintVT.isInteger() !=
Input.ConstraintVT.isInteger()) ||
(MatchRC.second != InputRC.second)) {
" incompatible type!");
}
}
-
}
}
return ConstraintOperands;
}
-
-/// getConstraintGenerality - Return an integer indicating how general CT
-/// is.
+/// Return an integer indicating how general CT is.
static unsigned getConstraintGenerality(TargetLowering::ConstraintType CT) {
switch (CT) {
case TargetLowering::C_Other:
Value *CallOperandVal = info.CallOperandVal;
// If we don't have a value, we can't do a match,
// but allow it at the lowest weight.
- if (CallOperandVal == NULL)
+ if (!CallOperandVal)
return CW_Default;
// Look at the constraint type.
switch (*constraint) {
return weight;
}
-/// ChooseConstraint - If there are multiple different constraints that we
-/// could pick for this operand (e.g. "imr") try to pick the 'best' one.
+/// If there are multiple different constraints that we could pick for this
+/// operand (e.g. "imr") try to pick the 'best' one.
/// This is somewhat tricky: constraints fall into four classes:
/// Other -> immediates and magic values
/// Register -> one specific register
OpInfo.ConstraintType = BestType;
}
-/// ComputeConstraintToUse - Determines the constraint code and constraint
-/// type to use for the specific AsmOperandInfo, setting
-/// OpInfo.ConstraintCode and OpInfo.ConstraintType.
+/// Determines the constraint code and constraint type to use for the specific
+/// AsmOperandInfo, setting OpInfo.ConstraintCode and OpInfo.ConstraintType.
void TargetLowering::ComputeConstraintToUse(AsmOperandInfo &OpInfo,
SDValue Op,
SelectionDAG *DAG) const {
}
}
-/// BuildExactDiv - Given an exact SDIV by a constant, create a multiplication
+/// \brief Given an exact SDIV by a constant, create a multiplication
/// with the multiplicative inverse of the constant.
-SDValue TargetLowering::BuildExactSDIV(SDValue Op1, SDValue Op2, DebugLoc dl,
- SelectionDAG &DAG) const {
- ConstantSDNode *C = cast<ConstantSDNode>(Op2);
- APInt d = C->getAPIntValue();
+static SDValue BuildExactSDIV(const TargetLowering &TLI, SDValue Op1, APInt d,
+ SDLoc dl, SelectionDAG &DAG,
+ std::vector<SDNode *> &Created) {
assert(d != 0 && "Division by zero!");
// Shift the value upfront if it is even, so the LSB is one.
unsigned ShAmt = d.countTrailingZeros();
if (ShAmt) {
// TODO: For UDIV use SRL instead of SRA.
- SDValue Amt = DAG.getConstant(ShAmt, getShiftAmountTy(Op1.getValueType()));
- Op1 = DAG.getNode(ISD::SRA, dl, Op1.getValueType(), Op1, Amt);
+ SDValue Amt =
+ DAG.getConstant(ShAmt, dl, TLI.getShiftAmountTy(Op1.getValueType(),
+ DAG.getDataLayout()));
+ SDNodeFlags Flags;
+ Flags.setExact(true);
+ Op1 = DAG.getNode(ISD::SRA, dl, Op1.getValueType(), Op1, Amt, &Flags);
+ Created.push_back(Op1.getNode());
d = d.ashr(ShAmt);
}
while ((t = d*xn) != 1)
xn *= APInt(d.getBitWidth(), 2) - t;
- Op2 = DAG.getConstant(xn, Op1.getValueType());
- return DAG.getNode(ISD::MUL, dl, Op1.getValueType(), Op1, Op2);
+ SDValue Op2 = DAG.getConstant(xn, dl, Op1.getValueType());
+ SDValue Mul = DAG.getNode(ISD::MUL, dl, Op1.getValueType(), Op1, Op2);
+ Created.push_back(Mul.getNode());
+ return Mul;
+}
+
+SDValue TargetLowering::BuildSDIVPow2(SDNode *N, const APInt &Divisor,
+ SelectionDAG &DAG,
+ std::vector<SDNode *> *Created) const {
+ AttributeSet Attr = DAG.getMachineFunction().getFunction()->getAttributes();
+ const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+ if (TLI.isIntDivCheap(N->getValueType(0), Attr))
+ return SDValue(N,0); // Lower SDIV as SDIV
+ return SDValue();
}
-/// BuildSDIVSequence - Given an ISD::SDIV node expressing a divide by constant,
+/// \brief Given an ISD::SDIV node expressing a divide by constant,
/// return a DAG expression to select that will generate the same value by
-/// multiplying by a magic number. See:
-/// <http://the.wall.riscom.net/books/proc/ppc/cwg/code2.html>
-SDValue TargetLowering::
-BuildSDIV(SDNode *N, SelectionDAG &DAG, bool IsAfterLegalization,
- std::vector<SDNode*> *Created) const {
+/// multiplying by a magic number.
+/// Ref: "Hacker's Delight" or "The PowerPC Compiler Writer's Guide".
+SDValue TargetLowering::BuildSDIV(SDNode *N, const APInt &Divisor,
+ SelectionDAG &DAG, bool IsAfterLegalization,
+ std::vector<SDNode *> *Created) const {
+ assert(Created && "No vector to hold sdiv ops.");
+
EVT VT = N->getValueType(0);
- DebugLoc dl= N->getDebugLoc();
+ SDLoc dl(N);
// Check to see if we can do this.
// FIXME: We should be more aggressive here.
if (!isTypeLegal(VT))
return SDValue();
- APInt d = cast<ConstantSDNode>(N->getOperand(1))->getAPIntValue();
- APInt::ms magics = d.magic();
+ // If the sdiv has an 'exact' bit we can use a simpler lowering.
+ if (cast<BinaryWithFlagsSDNode>(N)->Flags.hasExact())
+ return BuildExactSDIV(*this, N->getOperand(0), Divisor, dl, DAG, *Created);
+
+ APInt::ms magics = Divisor.magic();
// Multiply the numerator (operand 0) by the magic value
// FIXME: We should support doing a MUL in a wider type
if (IsAfterLegalization ? isOperationLegal(ISD::MULHS, VT) :
isOperationLegalOrCustom(ISD::MULHS, VT))
Q = DAG.getNode(ISD::MULHS, dl, VT, N->getOperand(0),
- DAG.getConstant(magics.m, VT));
+ DAG.getConstant(magics.m, dl, VT));
else if (IsAfterLegalization ? isOperationLegal(ISD::SMUL_LOHI, VT) :
isOperationLegalOrCustom(ISD::SMUL_LOHI, VT))
Q = SDValue(DAG.getNode(ISD::SMUL_LOHI, dl, DAG.getVTList(VT, VT),
N->getOperand(0),
- DAG.getConstant(magics.m, VT)).getNode(), 1);
+ DAG.getConstant(magics.m, dl, VT)).getNode(), 1);
else
return SDValue(); // No mulhs or equvialent
// If d > 0 and m < 0, add the numerator
- if (d.isStrictlyPositive() && magics.m.isNegative()) {
+ if (Divisor.isStrictlyPositive() && magics.m.isNegative()) {
Q = DAG.getNode(ISD::ADD, dl, VT, Q, N->getOperand(0));
- if (Created)
- Created->push_back(Q.getNode());
+ Created->push_back(Q.getNode());
}
// If d < 0 and m > 0, subtract the numerator.
- if (d.isNegative() && magics.m.isStrictlyPositive()) {
+ if (Divisor.isNegative() && magics.m.isStrictlyPositive()) {
Q = DAG.getNode(ISD::SUB, dl, VT, Q, N->getOperand(0));
- if (Created)
- Created->push_back(Q.getNode());
+ Created->push_back(Q.getNode());
}
+ auto &DL = DAG.getDataLayout();
// Shift right algebraic if shift value is nonzero
if (magics.s > 0) {
- Q = DAG.getNode(ISD::SRA, dl, VT, Q,
- DAG.getConstant(magics.s, getShiftAmountTy(Q.getValueType())));
- if (Created)
- Created->push_back(Q.getNode());
+ Q = DAG.getNode(
+ ISD::SRA, dl, VT, Q,
+ DAG.getConstant(magics.s, dl, getShiftAmountTy(Q.getValueType(), DL)));
+ Created->push_back(Q.getNode());
}
// Extract the sign bit and add it to the quotient
SDValue T =
- DAG.getNode(ISD::SRL, dl, VT, Q, DAG.getConstant(VT.getSizeInBits()-1,
- getShiftAmountTy(Q.getValueType())));
- if (Created)
- Created->push_back(T.getNode());
+ DAG.getNode(ISD::SRL, dl, VT, Q,
+ DAG.getConstant(VT.getScalarSizeInBits() - 1, dl,
+ getShiftAmountTy(Q.getValueType(), DL)));
+ Created->push_back(T.getNode());
return DAG.getNode(ISD::ADD, dl, VT, Q, T);
}
-/// BuildUDIVSequence - Given an ISD::UDIV node expressing a divide by constant,
+/// \brief Given an ISD::UDIV node expressing a divide by constant,
/// return a DAG expression to select that will generate the same value by
-/// multiplying by a magic number. See:
-/// <http://the.wall.riscom.net/books/proc/ppc/cwg/code2.html>
-SDValue TargetLowering::
-BuildUDIV(SDNode *N, SelectionDAG &DAG, bool IsAfterLegalization,
- std::vector<SDNode*> *Created) const {
+/// multiplying by a magic number.
+/// Ref: "Hacker's Delight" or "The PowerPC Compiler Writer's Guide".
+SDValue TargetLowering::BuildUDIV(SDNode *N, const APInt &Divisor,
+ SelectionDAG &DAG, bool IsAfterLegalization,
+ std::vector<SDNode *> *Created) const {
+ assert(Created && "No vector to hold udiv ops.");
+
EVT VT = N->getValueType(0);
- DebugLoc dl = N->getDebugLoc();
+ SDLoc dl(N);
+ auto &DL = DAG.getDataLayout();
// Check to see if we can do this.
// FIXME: We should be more aggressive here.
// FIXME: We should use a narrower constant when the upper
// bits are known to be zero.
- const APInt &N1C = cast<ConstantSDNode>(N->getOperand(1))->getAPIntValue();
- APInt::mu magics = N1C.magicu();
+ APInt::mu magics = Divisor.magicu();
SDValue Q = N->getOperand(0);
// If the divisor is even, we can avoid using the expensive fixup by shifting
// the divided value upfront.
- if (magics.a != 0 && !N1C[0]) {
- unsigned Shift = N1C.countTrailingZeros();
- Q = DAG.getNode(ISD::SRL, dl, VT, Q,
- DAG.getConstant(Shift, getShiftAmountTy(Q.getValueType())));
- if (Created)
- Created->push_back(Q.getNode());
+ if (magics.a != 0 && !Divisor[0]) {
+ unsigned Shift = Divisor.countTrailingZeros();
+ Q = DAG.getNode(
+ ISD::SRL, dl, VT, Q,
+ DAG.getConstant(Shift, dl, getShiftAmountTy(Q.getValueType(), DL)));
+ Created->push_back(Q.getNode());
// Get magic number for the shifted divisor.
- magics = N1C.lshr(Shift).magicu(Shift);
+ magics = Divisor.lshr(Shift).magicu(Shift);
assert(magics.a == 0 && "Should use cheap fixup now");
}
// FIXME: We should support doing a MUL in a wider type
if (IsAfterLegalization ? isOperationLegal(ISD::MULHU, VT) :
isOperationLegalOrCustom(ISD::MULHU, VT))
- Q = DAG.getNode(ISD::MULHU, dl, VT, Q, DAG.getConstant(magics.m, VT));
+ Q = DAG.getNode(ISD::MULHU, dl, VT, Q, DAG.getConstant(magics.m, dl, VT));
else if (IsAfterLegalization ? isOperationLegal(ISD::UMUL_LOHI, VT) :
isOperationLegalOrCustom(ISD::UMUL_LOHI, VT))
Q = SDValue(DAG.getNode(ISD::UMUL_LOHI, dl, DAG.getVTList(VT, VT), Q,
- DAG.getConstant(magics.m, VT)).getNode(), 1);
+ DAG.getConstant(magics.m, dl, VT)).getNode(), 1);
else
return SDValue(); // No mulhu or equvialent
- if (Created)
- Created->push_back(Q.getNode());
+
+ Created->push_back(Q.getNode());
if (magics.a == 0) {
- assert(magics.s < N1C.getBitWidth() &&
+ assert(magics.s < Divisor.getBitWidth() &&
"We shouldn't generate an undefined shift!");
- return DAG.getNode(ISD::SRL, dl, VT, Q,
- DAG.getConstant(magics.s, getShiftAmountTy(Q.getValueType())));
+ return DAG.getNode(
+ ISD::SRL, dl, VT, Q,
+ DAG.getConstant(magics.s, dl, getShiftAmountTy(Q.getValueType(), DL)));
} else {
SDValue NPQ = DAG.getNode(ISD::SUB, dl, VT, N->getOperand(0), Q);
- if (Created)
- Created->push_back(NPQ.getNode());
- NPQ = DAG.getNode(ISD::SRL, dl, VT, NPQ,
- DAG.getConstant(1, getShiftAmountTy(NPQ.getValueType())));
- if (Created)
- Created->push_back(NPQ.getNode());
+ Created->push_back(NPQ.getNode());
+ NPQ = DAG.getNode(
+ ISD::SRL, dl, VT, NPQ,
+ DAG.getConstant(1, dl, getShiftAmountTy(NPQ.getValueType(), DL)));
+ Created->push_back(NPQ.getNode());
NPQ = DAG.getNode(ISD::ADD, dl, VT, NPQ, Q);
- if (Created)
- Created->push_back(NPQ.getNode());
- return DAG.getNode(ISD::SRL, dl, VT, NPQ,
- DAG.getConstant(magics.s-1, getShiftAmountTy(NPQ.getValueType())));
+ Created->push_back(NPQ.getNode());
+ return DAG.getNode(
+ ISD::SRL, dl, VT, NPQ,
+ DAG.getConstant(magics.s - 1, dl,
+ getShiftAmountTy(NPQ.getValueType(), DL)));
+ }
+}
+
+bool TargetLowering::
+verifyReturnAddressArgumentIsConstant(SDValue Op, SelectionDAG &DAG) const {
+ if (!isa<ConstantSDNode>(Op.getOperand(0))) {
+ DAG.getContext()->emitError("argument to '__builtin_return_address' must "
+ "be a constant integer");
+ return true;
+ }
+
+ return false;
+}
+
+//===----------------------------------------------------------------------===//
+// Legalization Utilities
+//===----------------------------------------------------------------------===//
+
+bool TargetLowering::expandMUL(SDNode *N, SDValue &Lo, SDValue &Hi, EVT HiLoVT,
+ SelectionDAG &DAG, SDValue LL, SDValue LH,
+ SDValue RL, SDValue RH) const {
+ EVT VT = N->getValueType(0);
+ SDLoc dl(N);
+
+ bool HasMULHS = isOperationLegalOrCustom(ISD::MULHS, HiLoVT);
+ bool HasMULHU = isOperationLegalOrCustom(ISD::MULHU, HiLoVT);
+ bool HasSMUL_LOHI = isOperationLegalOrCustom(ISD::SMUL_LOHI, HiLoVT);
+ bool HasUMUL_LOHI = isOperationLegalOrCustom(ISD::UMUL_LOHI, HiLoVT);
+ if (HasMULHU || HasMULHS || HasUMUL_LOHI || HasSMUL_LOHI) {
+ unsigned OuterBitSize = VT.getSizeInBits();
+ unsigned InnerBitSize = HiLoVT.getSizeInBits();
+ unsigned LHSSB = DAG.ComputeNumSignBits(N->getOperand(0));
+ unsigned RHSSB = DAG.ComputeNumSignBits(N->getOperand(1));
+
+ // LL, LH, RL, and RH must be either all NULL or all set to a value.
+ assert((LL.getNode() && LH.getNode() && RL.getNode() && RH.getNode()) ||
+ (!LL.getNode() && !LH.getNode() && !RL.getNode() && !RH.getNode()));
+
+ if (!LL.getNode() && !RL.getNode() &&
+ isOperationLegalOrCustom(ISD::TRUNCATE, HiLoVT)) {
+ LL = DAG.getNode(ISD::TRUNCATE, dl, HiLoVT, N->getOperand(0));
+ RL = DAG.getNode(ISD::TRUNCATE, dl, HiLoVT, N->getOperand(1));
+ }
+
+ if (!LL.getNode())
+ return false;
+
+ APInt HighMask = APInt::getHighBitsSet(OuterBitSize, InnerBitSize);
+ if (DAG.MaskedValueIsZero(N->getOperand(0), HighMask) &&
+ DAG.MaskedValueIsZero(N->getOperand(1), HighMask)) {
+ // The inputs are both zero-extended.
+ if (HasUMUL_LOHI) {
+ // We can emit a umul_lohi.
+ Lo = DAG.getNode(ISD::UMUL_LOHI, dl, DAG.getVTList(HiLoVT, HiLoVT), LL,
+ RL);
+ Hi = SDValue(Lo.getNode(), 1);
+ return true;
+ }
+ if (HasMULHU) {
+ // We can emit a mulhu+mul.
+ Lo = DAG.getNode(ISD::MUL, dl, HiLoVT, LL, RL);
+ Hi = DAG.getNode(ISD::MULHU, dl, HiLoVT, LL, RL);
+ return true;
+ }
+ }
+ if (LHSSB > InnerBitSize && RHSSB > InnerBitSize) {
+ // The input values are both sign-extended.
+ if (HasSMUL_LOHI) {
+ // We can emit a smul_lohi.
+ Lo = DAG.getNode(ISD::SMUL_LOHI, dl, DAG.getVTList(HiLoVT, HiLoVT), LL,
+ RL);
+ Hi = SDValue(Lo.getNode(), 1);
+ return true;
+ }
+ if (HasMULHS) {
+ // We can emit a mulhs+mul.
+ Lo = DAG.getNode(ISD::MUL, dl, HiLoVT, LL, RL);
+ Hi = DAG.getNode(ISD::MULHS, dl, HiLoVT, LL, RL);
+ return true;
+ }
+ }
+
+ if (!LH.getNode() && !RH.getNode() &&
+ isOperationLegalOrCustom(ISD::SRL, VT) &&
+ isOperationLegalOrCustom(ISD::TRUNCATE, HiLoVT)) {
+ auto &DL = DAG.getDataLayout();
+ unsigned ShiftAmt = VT.getSizeInBits() - HiLoVT.getSizeInBits();
+ SDValue Shift = DAG.getConstant(ShiftAmt, dl, getShiftAmountTy(VT, DL));
+ LH = DAG.getNode(ISD::SRL, dl, VT, N->getOperand(0), Shift);
+ LH = DAG.getNode(ISD::TRUNCATE, dl, HiLoVT, LH);
+ RH = DAG.getNode(ISD::SRL, dl, VT, N->getOperand(1), Shift);
+ RH = DAG.getNode(ISD::TRUNCATE, dl, HiLoVT, RH);
+ }
+
+ if (!LH.getNode())
+ return false;
+
+ if (HasUMUL_LOHI) {
+ // Lo,Hi = umul LHS, RHS.
+ SDValue UMulLOHI = DAG.getNode(ISD::UMUL_LOHI, dl,
+ DAG.getVTList(HiLoVT, HiLoVT), LL, RL);
+ Lo = UMulLOHI;
+ Hi = UMulLOHI.getValue(1);
+ RH = DAG.getNode(ISD::MUL, dl, HiLoVT, LL, RH);
+ LH = DAG.getNode(ISD::MUL, dl, HiLoVT, LH, RL);
+ Hi = DAG.getNode(ISD::ADD, dl, HiLoVT, Hi, RH);
+ Hi = DAG.getNode(ISD::ADD, dl, HiLoVT, Hi, LH);
+ return true;
+ }
+ if (HasMULHU) {
+ Lo = DAG.getNode(ISD::MUL, dl, HiLoVT, LL, RL);
+ Hi = DAG.getNode(ISD::MULHU, dl, HiLoVT, LL, RL);
+ RH = DAG.getNode(ISD::MUL, dl, HiLoVT, LL, RH);
+ LH = DAG.getNode(ISD::MUL, dl, HiLoVT, LH, RL);
+ Hi = DAG.getNode(ISD::ADD, dl, HiLoVT, Hi, RH);
+ Hi = DAG.getNode(ISD::ADD, dl, HiLoVT, Hi, LH);
+ return true;
+ }
}
+ return false;
+}
+
+bool TargetLowering::expandFP_TO_SINT(SDNode *Node, SDValue &Result,
+ SelectionDAG &DAG) const {
+ EVT VT = Node->getOperand(0).getValueType();
+ EVT NVT = Node->getValueType(0);
+ SDLoc dl(SDValue(Node, 0));
+
+ // FIXME: Only f32 to i64 conversions are supported.
+ if (VT != MVT::f32 || NVT != MVT::i64)
+ return false;
+
+ // Expand f32 -> i64 conversion
+ // This algorithm comes from compiler-rt's implementation of fixsfdi:
+ // https://github.com/llvm-mirror/compiler-rt/blob/master/lib/builtins/fixsfdi.c
+ EVT IntVT = EVT::getIntegerVT(*DAG.getContext(),
+ VT.getSizeInBits());
+ SDValue ExponentMask = DAG.getConstant(0x7F800000, dl, IntVT);
+ SDValue ExponentLoBit = DAG.getConstant(23, dl, IntVT);
+ SDValue Bias = DAG.getConstant(127, dl, IntVT);
+ SDValue SignMask = DAG.getConstant(APInt::getSignBit(VT.getSizeInBits()), dl,
+ IntVT);
+ SDValue SignLowBit = DAG.getConstant(VT.getSizeInBits() - 1, dl, IntVT);
+ SDValue MantissaMask = DAG.getConstant(0x007FFFFF, dl, IntVT);
+
+ SDValue Bits = DAG.getNode(ISD::BITCAST, dl, IntVT, Node->getOperand(0));
+
+ auto &DL = DAG.getDataLayout();
+ SDValue ExponentBits = DAG.getNode(
+ ISD::SRL, dl, IntVT, DAG.getNode(ISD::AND, dl, IntVT, Bits, ExponentMask),
+ DAG.getZExtOrTrunc(ExponentLoBit, dl, getShiftAmountTy(IntVT, DL)));
+ SDValue Exponent = DAG.getNode(ISD::SUB, dl, IntVT, ExponentBits, Bias);
+
+ SDValue Sign = DAG.getNode(
+ ISD::SRA, dl, IntVT, DAG.getNode(ISD::AND, dl, IntVT, Bits, SignMask),
+ DAG.getZExtOrTrunc(SignLowBit, dl, getShiftAmountTy(IntVT, DL)));
+ Sign = DAG.getSExtOrTrunc(Sign, dl, NVT);
+
+ SDValue R = DAG.getNode(ISD::OR, dl, IntVT,
+ DAG.getNode(ISD::AND, dl, IntVT, Bits, MantissaMask),
+ DAG.getConstant(0x00800000, dl, IntVT));
+
+ R = DAG.getZExtOrTrunc(R, dl, NVT);
+
+ R = DAG.getSelectCC(
+ dl, Exponent, ExponentLoBit,
+ DAG.getNode(ISD::SHL, dl, NVT, R,
+ DAG.getZExtOrTrunc(
+ DAG.getNode(ISD::SUB, dl, IntVT, Exponent, ExponentLoBit),
+ dl, getShiftAmountTy(IntVT, DL))),
+ DAG.getNode(ISD::SRL, dl, NVT, R,
+ DAG.getZExtOrTrunc(
+ DAG.getNode(ISD::SUB, dl, IntVT, ExponentLoBit, Exponent),
+ dl, getShiftAmountTy(IntVT, DL))),
+ ISD::SETGT);
+
+ SDValue Ret = DAG.getNode(ISD::SUB, dl, NVT,
+ DAG.getNode(ISD::XOR, dl, NVT, R, Sign),
+ Sign);
+
+ Result = DAG.getSelectCC(dl, Exponent, DAG.getConstant(0, dl, IntVT),
+ DAG.getConstant(0, dl, NVT), Ret, ISD::SETLT);
+ return true;
+}
+
+//===----------------------------------------------------------------------===//
+// Implementation of Emulated TLS Model
+//===----------------------------------------------------------------------===//
+
+SDValue TargetLowering::LowerToTLSEmulatedModel(const GlobalAddressSDNode *GA,
+ SelectionDAG &DAG) const {
+ // Access to address of TLS varialbe xyz is lowered to a function call:
+ // __emutls_get_address( address of global variable named "__emutls_v.xyz" )
+ EVT PtrVT = getPointerTy(DAG.getDataLayout());
+ PointerType *VoidPtrType = Type::getInt8PtrTy(*DAG.getContext());
+ SDLoc dl(GA);
+
+ ArgListTy Args;
+ ArgListEntry Entry;
+ std::string NameString = ("__emutls_v." + GA->getGlobal()->getName()).str();
+ Module *VariableModule = const_cast<Module*>(GA->getGlobal()->getParent());
+ StringRef EmuTlsVarName(NameString);
+ GlobalVariable *EmuTlsVar = VariableModule->getNamedGlobal(EmuTlsVarName);
+ if (!EmuTlsVar)
+ EmuTlsVar = dyn_cast_or_null<GlobalVariable>(
+ VariableModule->getOrInsertGlobal(EmuTlsVarName, VoidPtrType));
+ Entry.Node = DAG.getGlobalAddress(EmuTlsVar, dl, PtrVT);
+ Entry.Ty = VoidPtrType;
+ Args.push_back(Entry);
+
+ SDValue EmuTlsGetAddr = DAG.getExternalSymbol("__emutls_get_address", PtrVT);
+
+ TargetLowering::CallLoweringInfo CLI(DAG);
+ CLI.setDebugLoc(dl).setChain(DAG.getEntryNode());
+ CLI.setCallee(CallingConv::C, VoidPtrType, EmuTlsGetAddr, std::move(Args), 0);
+ std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
+
+ // TLSADDR will be codegen'ed as call. Inform MFI that function has calls.
+ // At last for X86 targets, maybe good for other targets too?
+ MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
+ MFI->setAdjustsStack(true); // Is this only for X86 target?
+ MFI->setHasCalls(true);
+
+ assert((GA->getOffset() == 0) &&
+ "Emulated TLS must have zero offset in GlobalAddressSDNode");
+ return CallResult.first;
}