SDValue ARM64SelectionDAGInfo::EmitTargetCodeForMemset(
SelectionDAG &DAG, SDLoc dl, SDValue Chain, SDValue Dst, SDValue Src,
SDValue Size, unsigned Align, bool isVolatile,
MachinePointerInfo DstPtrInfo) const {
// Check to see if there is a specialized entry-point for memory zeroing.
ConstantSDNode *V = dyn_cast<ConstantSDNode>(Src);
ConstantSDNode *SizeValue = dyn_cast<ConstantSDNode>(Size);
const char *bzeroEntry =
(V && V->isNullValue()) ? Subtarget->getBZeroEntry() : 0;
// For small size (< 256), it is not beneficial to use bzero
// instead of memset.
if (bzeroEntry && (!SizeValue || SizeValue->getZExtValue() > 256)) {
const ARM64TargetLowering &TLI = *static_cast<const ARM64TargetLowering *>(
DAG.getTarget().getTargetLowering());
EVT IntPtr = TLI.getPointerTy();
Type *IntPtrTy = getDataLayout()->getIntPtrType(*DAG.getContext());
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Node = Dst;
Entry.Ty = IntPtrTy;
Args.push_back(Entry);
Entry.Node = Size;
Args.push_back(Entry);
TargetLowering::CallLoweringInfo CLI(
Chain, Type::getVoidTy(*DAG.getContext()), false, false, false, false,
0, CallingConv::C, /*isTailCall=*/false,
/*doesNotRet=*/false, /*isReturnValueUsed=*/false,
DAG.getExternalSymbol(bzeroEntry, IntPtr), Args, DAG, dl);
std::pair<SDValue, SDValue> CallResult = TLI.LowerCallTo(CLI);
return CallResult.second;
}
return SDValue();
}
SDValue
X86SelectionDAGInfo::EmitTargetCodeForMemset(SelectionDAG &DAG, SDLoc dl,
SDValue Chain,
SDValue Dst, SDValue Src,
SDValue Size, unsigned Align,
bool isVolatile,
MachinePointerInfo DstPtrInfo) const {
ConstantSDNode *ConstantSize = dyn_cast<ConstantSDNode>(Size);
// If to a segment-relative address space, use the default lowering.
if (DstPtrInfo.getAddrSpace() >= 256)
return SDValue();
// If not DWORD aligned or size is more than the threshold, call the library.
// The libc version is likely to be faster for these cases. It can use the
// address value and run time information about the CPU.
if ((Align & 3) != 0 ||
!ConstantSize ||
ConstantSize->getZExtValue() >
Subtarget->getMaxInlineSizeThreshold()) {
// Check to see if there is a specialized entry-point for memory zeroing.
ConstantSDNode *V = dyn_cast<ConstantSDNode>(Src);
if (const char *bzeroEntry = V &&
V->isNullValue() ? Subtarget->getBZeroEntry() : nullptr) {
EVT IntPtr = TLI.getPointerTy();
Type *IntPtrTy = getDataLayout()->getIntPtrType(*DAG.getContext());
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Node = Dst;
Entry.Ty = IntPtrTy;
Args.push_back(Entry);
Entry.Node = Size;
Args.push_back(Entry);
TargetLowering::CallLoweringInfo CLI(DAG);
CLI.setDebugLoc(dl).setChain(Chain)
.setCallee(CallingConv::C, Type::getVoidTy(*DAG.getContext()),
DAG.getExternalSymbol(bzeroEntry, IntPtr), &Args, 0)
.setDiscardResult();
std::pair<SDValue,SDValue> CallResult = TLI.LowerCallTo(CLI);
return CallResult.second;
}
// Otherwise have the target-independent code call memset.
return SDValue();
}
uint64_t SizeVal = ConstantSize->getZExtValue();
SDValue InFlag;
EVT AVT;
SDValue Count;
ConstantSDNode *ValC = dyn_cast<ConstantSDNode>(Src);
unsigned BytesLeft = 0;
bool TwoRepStos = false;
if (ValC) {
unsigned ValReg;
uint64_t Val = ValC->getZExtValue() & 255;
// If the value is a constant, then we can potentially use larger sets.
switch (Align & 3) {
case 2: // WORD aligned
AVT = MVT::i16;
ValReg = X86::AX;
Val = (Val << 8) | Val;
break;
case 0: // DWORD aligned
AVT = MVT::i32;
ValReg = X86::EAX;
Val = (Val << 8) | Val;
Val = (Val << 16) | Val;
if (Subtarget->is64Bit() && ((Align & 0x7) == 0)) { // QWORD aligned
AVT = MVT::i64;
ValReg = X86::RAX;
Val = (Val << 32) | Val;
}
break;
default: // Byte aligned
AVT = MVT::i8;
ValReg = X86::AL;
Count = DAG.getIntPtrConstant(SizeVal);
break;
}
if (AVT.bitsGT(MVT::i8)) {
unsigned UBytes = AVT.getSizeInBits() / 8;
Count = DAG.getIntPtrConstant(SizeVal / UBytes);
BytesLeft = SizeVal % UBytes;
}
Chain = DAG.getCopyToReg(Chain, dl, ValReg, DAG.getConstant(Val, AVT),
InFlag);
InFlag = Chain.getValue(1);
} else {
AVT = MVT::i8;
Count = DAG.getIntPtrConstant(SizeVal);
Chain = DAG.getCopyToReg(Chain, dl, X86::AL, Src, InFlag);
InFlag = Chain.getValue(1);
}
Chain = DAG.getCopyToReg(Chain, dl, Subtarget->is64Bit() ? X86::RCX :
X86::ECX,
Count, InFlag);
InFlag = Chain.getValue(1);
Chain = DAG.getCopyToReg(Chain, dl, Subtarget->is64Bit() ? X86::RDI :
X86::EDI,
Dst, InFlag);
InFlag = Chain.getValue(1);
SDVTList Tys = DAG.getVTList(MVT::Other, MVT::Glue);
SDValue Ops[] = { Chain, DAG.getValueType(AVT), InFlag };
Chain = DAG.getNode(X86ISD::REP_STOS, dl, Tys, Ops);
if (TwoRepStos) {
InFlag = Chain.getValue(1);
Count = Size;
EVT CVT = Count.getValueType();
SDValue Left = DAG.getNode(ISD::AND, dl, CVT, Count,
DAG.getConstant((AVT == MVT::i64) ? 7 : 3, CVT));
Chain = DAG.getCopyToReg(Chain, dl, (CVT == MVT::i64) ? X86::RCX :
X86::ECX,
Left, InFlag);
InFlag = Chain.getValue(1);
Tys = DAG.getVTList(MVT::Other, MVT::Glue);
SDValue Ops[] = { Chain, DAG.getValueType(MVT::i8), InFlag };
Chain = DAG.getNode(X86ISD::REP_STOS, dl, Tys, Ops);
} else if (BytesLeft) {
// Handle the last 1 - 7 bytes.
unsigned Offset = SizeVal - BytesLeft;
EVT AddrVT = Dst.getValueType();
EVT SizeVT = Size.getValueType();
Chain = DAG.getMemset(Chain, dl,
DAG.getNode(ISD::ADD, dl, AddrVT, Dst,
DAG.getConstant(Offset, AddrVT)),
Src,
DAG.getConstant(BytesLeft, SizeVT),
Align, isVolatile, DstPtrInfo.getWithOffset(Offset));
}
// TODO: Use a Tokenfactor, as in memcpy, instead of a single chain.
return Chain;
}
static bool isZero(SDValue V)
{
ConstantSDNode *C = dyn_cast<ConstantSDNode>(V);
return C && C->isNullValue();
}
SDValue X86SelectionDAGInfo::EmitTargetCodeForMemset(
SelectionDAG &DAG, const SDLoc &dl, SDValue Chain, SDValue Dst, SDValue Val,
SDValue Size, unsigned Align, bool isVolatile,
MachinePointerInfo DstPtrInfo) const {
ConstantSDNode *ConstantSize = dyn_cast<ConstantSDNode>(Size);
const X86Subtarget &Subtarget =
DAG.getMachineFunction().getSubtarget<X86Subtarget>();
#ifndef NDEBUG
// If the base register might conflict with our physical registers, bail out.
const MCPhysReg ClobberSet[] = {X86::RCX, X86::RAX, X86::RDI,
X86::ECX, X86::EAX, X86::EDI};
assert(!isBaseRegConflictPossible(DAG, ClobberSet));
#endif
// If to a segment-relative address space, use the default lowering.
if (DstPtrInfo.getAddrSpace() >= 256)
return SDValue();
// If not DWORD aligned or size is more than the threshold, call the library.
// The libc version is likely to be faster for these cases. It can use the
// address value and run time information about the CPU.
if ((Align & 3) != 0 || !ConstantSize ||
ConstantSize->getZExtValue() > Subtarget.getMaxInlineSizeThreshold()) {
// Check to see if there is a specialized entry-point for memory zeroing.
ConstantSDNode *ValC = dyn_cast<ConstantSDNode>(Val);
if (const char *bzeroName = (ValC && ValC->isNullValue())
? DAG.getTargetLoweringInfo().getLibcallName(RTLIB::BZERO)
: nullptr) {
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
EVT IntPtr = TLI.getPointerTy(DAG.getDataLayout());
Type *IntPtrTy = DAG.getDataLayout().getIntPtrType(*DAG.getContext());
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Node = Dst;
Entry.Ty = IntPtrTy;
Args.push_back(Entry);
Entry.Node = Size;
Args.push_back(Entry);
TargetLowering::CallLoweringInfo CLI(DAG);
CLI.setDebugLoc(dl)
.setChain(Chain)
.setLibCallee(CallingConv::C, Type::getVoidTy(*DAG.getContext()),
DAG.getExternalSymbol(bzeroName, IntPtr),
std::move(Args))
.setDiscardResult();
std::pair<SDValue,SDValue> CallResult = TLI.LowerCallTo(CLI);
return CallResult.second;
}
// Otherwise have the target-independent code call memset.
return SDValue();
}
uint64_t SizeVal = ConstantSize->getZExtValue();
SDValue InFlag;
EVT AVT;
SDValue Count;
ConstantSDNode *ValC = dyn_cast<ConstantSDNode>(Val);
unsigned BytesLeft = 0;
if (ValC) {
unsigned ValReg;
uint64_t Val = ValC->getZExtValue() & 255;
// If the value is a constant, then we can potentially use larger sets.
switch (Align & 3) {
case 2: // WORD aligned
AVT = MVT::i16;
ValReg = X86::AX;
Val = (Val << 8) | Val;
break;
case 0: // DWORD aligned
AVT = MVT::i32;
ValReg = X86::EAX;
Val = (Val << 8) | Val;
Val = (Val << 16) | Val;
if (Subtarget.is64Bit() && ((Align & 0x7) == 0)) { // QWORD aligned
AVT = MVT::i64;
ValReg = X86::RAX;
Val = (Val << 32) | Val;
}
break;
default: // Byte aligned
AVT = MVT::i8;
ValReg = X86::AL;
Count = DAG.getIntPtrConstant(SizeVal, dl);
break;
}
if (AVT.bitsGT(MVT::i8)) {
unsigned UBytes = AVT.getSizeInBits() / 8;
Count = DAG.getIntPtrConstant(SizeVal / UBytes, dl);
BytesLeft = SizeVal % UBytes;
}
Chain = DAG.getCopyToReg(Chain, dl, ValReg, DAG.getConstant(Val, dl, AVT),
InFlag);
InFlag = Chain.getValue(1);
} else {
AVT = MVT::i8;
Count = DAG.getIntPtrConstant(SizeVal, dl);
Chain = DAG.getCopyToReg(Chain, dl, X86::AL, Val, InFlag);
InFlag = Chain.getValue(1);
}
bool Use64BitRegs = Subtarget.isTarget64BitLP64();
Chain = DAG.getCopyToReg(Chain, dl, Use64BitRegs ? X86::RCX : X86::ECX,
Count, InFlag);
InFlag = Chain.getValue(1);
Chain = DAG.getCopyToReg(Chain, dl, Use64BitRegs ? X86::RDI : X86::EDI,
Dst, InFlag);
InFlag = Chain.getValue(1);
SDVTList Tys = DAG.getVTList(MVT::Other, MVT::Glue);
SDValue Ops[] = { Chain, DAG.getValueType(AVT), InFlag };
Chain = DAG.getNode(X86ISD::REP_STOS, dl, Tys, Ops);
if (BytesLeft) {
// Handle the last 1 - 7 bytes.
unsigned Offset = SizeVal - BytesLeft;
EVT AddrVT = Dst.getValueType();
EVT SizeVT = Size.getValueType();
Chain = DAG.getMemset(Chain, dl,
DAG.getNode(ISD::ADD, dl, AddrVT, Dst,
DAG.getConstant(Offset, dl, AddrVT)),
Val,
DAG.getConstant(BytesLeft, dl, SizeVT),
Align, isVolatile, false,
DstPtrInfo.getWithOffset(Offset));
}
// TODO: Use a Tokenfactor, as in memcpy, instead of a single chain.
return Chain;
}