//
//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "xcore-lower"
-
#include "XCoreISelLowering.h"
#include "XCore.h"
#include "XCoreMachineFunctionInfo.h"
#include "llvm/CodeGen/SelectionDAGISel.h"
#include "llvm/CodeGen/ValueTypes.h"
#include "llvm/IR/CallingConv.h"
+#include "llvm/IR/Constants.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalAlias.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
+#include <algorithm>
+
using namespace llvm;
+#define DEBUG_TYPE "xcore-lower"
+
const char *XCoreTargetLowering::
getTargetNodeName(unsigned Opcode) const
{
- switch (Opcode)
+ switch ((XCoreISD::NodeType)Opcode)
{
+ case XCoreISD::FIRST_NUMBER : break;
case XCoreISD::BL : return "XCoreISD::BL";
case XCoreISD::PCRelativeWrapper : return "XCoreISD::PCRelativeWrapper";
case XCoreISD::DPRelativeWrapper : return "XCoreISD::DPRelativeWrapper";
case XCoreISD::CPRelativeWrapper : return "XCoreISD::CPRelativeWrapper";
+ case XCoreISD::LDWSP : return "XCoreISD::LDWSP";
case XCoreISD::STWSP : return "XCoreISD::STWSP";
case XCoreISD::RETSP : return "XCoreISD::RETSP";
case XCoreISD::LADD : return "XCoreISD::LADD";
case XCoreISD::CRC8 : return "XCoreISD::CRC8";
case XCoreISD::BR_JT : return "XCoreISD::BR_JT";
case XCoreISD::BR_JT32 : return "XCoreISD::BR_JT32";
- default : return NULL;
+ case XCoreISD::FRAME_TO_ARGS_OFFSET : return "XCoreISD::FRAME_TO_ARGS_OFFSET";
+ case XCoreISD::EH_RETURN : return "XCoreISD::EH_RETURN";
+ case XCoreISD::MEMBARRIER : return "XCoreISD::MEMBARRIER";
}
+ return nullptr;
}
-XCoreTargetLowering::XCoreTargetLowering(XCoreTargetMachine &XTM)
- : TargetLowering(XTM, new XCoreTargetObjectFile()),
- TM(XTM),
- Subtarget(*XTM.getSubtargetImpl()) {
+XCoreTargetLowering::XCoreTargetLowering(const TargetMachine &TM,
+ const XCoreSubtarget &Subtarget)
+ : TargetLowering(TM), TM(TM), Subtarget(Subtarget) {
// Set up the register classes.
addRegisterClass(MVT::i32, &XCore::GRRegsRegClass);
// Compute derived properties from the register classes
- computeRegisterProperties();
-
- // Division is expensive
- setIntDivIsCheap(false);
+ computeRegisterProperties(Subtarget.getRegisterInfo());
setStackPointerRegisterToSaveRestore(XCore::SP);
- setSchedulingPreference(Sched::RegPressure);
+ setSchedulingPreference(Sched::Source);
// Use i32 for setcc operations results (slt, sgt, ...).
setBooleanContents(ZeroOrOneBooleanContent);
setBooleanVectorContents(ZeroOrOneBooleanContent); // FIXME: Is this correct?
// XCore does not have the NodeTypes below.
- setOperationAction(ISD::BR_CC, MVT::Other, Expand);
- setOperationAction(ISD::SELECT_CC, MVT::i32, Custom);
+ setOperationAction(ISD::BR_CC, MVT::i32, Expand);
+ setOperationAction(ISD::SELECT_CC, MVT::i32, Expand);
setOperationAction(ISD::ADDC, MVT::i32, Expand);
setOperationAction(ISD::ADDE, MVT::i32, Expand);
setOperationAction(ISD::SUBC, MVT::i32, Expand);
setOperationAction(ISD::SUBE, MVT::i32, Expand);
- // Stop the combiner recombining select and set_cc
- setOperationAction(ISD::SELECT_CC, MVT::Other, Expand);
-
// 64bit
setOperationAction(ISD::ADD, MVT::i64, Custom);
setOperationAction(ISD::SUB, MVT::i64, Custom);
setOperationAction(ISD::GlobalAddress, MVT::i32, Custom);
setOperationAction(ISD::BlockAddress, MVT::i32 , Custom);
- // Thread Local Storage
- setOperationAction(ISD::GlobalTLSAddress, MVT::i32, Custom);
-
// Conversion of i64 -> double produces constantpool nodes
setOperationAction(ISD::ConstantPool, MVT::i32, Custom);
// Loads
- setLoadExtAction(ISD::EXTLOAD, MVT::i1, Promote);
- setLoadExtAction(ISD::ZEXTLOAD, MVT::i1, Promote);
- setLoadExtAction(ISD::SEXTLOAD, MVT::i1, Promote);
+ for (MVT VT : MVT::integer_valuetypes()) {
+ setLoadExtAction(ISD::EXTLOAD, VT, MVT::i1, Promote);
+ setLoadExtAction(ISD::ZEXTLOAD, VT, MVT::i1, Promote);
+ setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote);
- setLoadExtAction(ISD::SEXTLOAD, MVT::i8, Expand);
- setLoadExtAction(ISD::ZEXTLOAD, MVT::i16, Expand);
+ setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i8, Expand);
+ setLoadExtAction(ISD::ZEXTLOAD, VT, MVT::i16, Expand);
+ }
// Custom expand misaligned loads / stores.
setOperationAction(ISD::LOAD, MVT::i32, Custom);
setOperationAction(ISD::STACKRESTORE, MVT::Other, Expand);
setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i32, Expand);
+ // Exception handling
+ setOperationAction(ISD::EH_RETURN, MVT::Other, Custom);
+ setExceptionPointerRegister(XCore::R0);
+ setExceptionSelectorRegister(XCore::R1);
+ setOperationAction(ISD::FRAME_TO_ARGS_OFFSET, MVT::i32, Custom);
+
+ // Atomic operations
+ // We request a fence for ATOMIC_* instructions, to reduce them to Monotonic.
+ // As we are always Sequential Consistent, an ATOMIC_FENCE becomes a no OP.
+ setInsertFencesForAtomic(true);
+ setOperationAction(ISD::ATOMIC_FENCE, MVT::Other, Custom);
+ setOperationAction(ISD::ATOMIC_LOAD, MVT::i32, Custom);
+ setOperationAction(ISD::ATOMIC_STORE, MVT::i32, Custom);
+
// TRAMPOLINE is custom lowered.
setOperationAction(ISD::INIT_TRAMPOLINE, MVT::Other, Custom);
setOperationAction(ISD::ADJUST_TRAMPOLINE, MVT::Other, Custom);
// We want to custom lower some of our intrinsics.
setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom);
- maxStoresPerMemset = maxStoresPerMemsetOptSize = 4;
- maxStoresPerMemmove = maxStoresPerMemmoveOptSize
- = maxStoresPerMemcpy = maxStoresPerMemcpyOptSize = 2;
+ MaxStoresPerMemset = MaxStoresPerMemsetOptSize = 4;
+ MaxStoresPerMemmove = MaxStoresPerMemmoveOptSize
+ = MaxStoresPerMemcpy = MaxStoresPerMemcpyOptSize = 2;
// We have target-specific dag combine patterns for the following nodes:
setTargetDAGCombine(ISD::STORE);
setTargetDAGCombine(ISD::ADD);
+ setTargetDAGCombine(ISD::INTRINSIC_VOID);
+ setTargetDAGCombine(ISD::INTRINSIC_W_CHAIN);
setMinFunctionAlignment(1);
+ setPrefFunctionAlignment(2);
+}
+
+bool XCoreTargetLowering::isZExtFree(SDValue Val, EVT VT2) const {
+ if (Val.getOpcode() != ISD::LOAD)
+ return false;
+
+ EVT VT1 = Val.getValueType();
+ if (!VT1.isSimple() || !VT1.isInteger() ||
+ !VT2.isSimple() || !VT2.isInteger())
+ return false;
+
+ switch (VT1.getSimpleVT().SimpleTy) {
+ default: break;
+ case MVT::i8:
+ return true;
+ }
+
+ return false;
}
SDValue XCoreTargetLowering::
LowerOperation(SDValue Op, SelectionDAG &DAG) const {
switch (Op.getOpcode())
{
+ case ISD::EH_RETURN: return LowerEH_RETURN(Op, DAG);
case ISD::GlobalAddress: return LowerGlobalAddress(Op, DAG);
- case ISD::GlobalTLSAddress: return LowerGlobalTLSAddress(Op, DAG);
case ISD::BlockAddress: return LowerBlockAddress(Op, DAG);
case ISD::ConstantPool: return LowerConstantPool(Op, DAG);
case ISD::BR_JT: return LowerBR_JT(Op, DAG);
case ISD::LOAD: return LowerLOAD(Op, DAG);
case ISD::STORE: return LowerSTORE(Op, DAG);
- case ISD::SELECT_CC: return LowerSELECT_CC(Op, DAG);
case ISD::VAARG: return LowerVAARG(Op, DAG);
case ISD::VASTART: return LowerVASTART(Op, DAG);
case ISD::SMUL_LOHI: return LowerSMUL_LOHI(Op, DAG);
case ISD::ADD:
case ISD::SUB: return ExpandADDSUB(Op.getNode(), DAG);
case ISD::FRAMEADDR: return LowerFRAMEADDR(Op, DAG);
+ case ISD::RETURNADDR: return LowerRETURNADDR(Op, DAG);
+ case ISD::FRAME_TO_ARGS_OFFSET: return LowerFRAME_TO_ARGS_OFFSET(Op, DAG);
case ISD::INIT_TRAMPOLINE: return LowerINIT_TRAMPOLINE(Op, DAG);
case ISD::ADJUST_TRAMPOLINE: return LowerADJUST_TRAMPOLINE(Op, DAG);
case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG);
+ case ISD::ATOMIC_FENCE: return LowerATOMIC_FENCE(Op, DAG);
+ case ISD::ATOMIC_LOAD: return LowerATOMIC_LOAD(Op, DAG);
+ case ISD::ATOMIC_STORE: return LowerATOMIC_STORE(Op, DAG);
default:
llvm_unreachable("unimplemented operand");
}
// Misc Lower Operation implementation
//===----------------------------------------------------------------------===//
-SDValue XCoreTargetLowering::
-LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const
-{
- DebugLoc dl = Op.getDebugLoc();
- SDValue Cond = DAG.getNode(ISD::SETCC, dl, MVT::i32, Op.getOperand(2),
- Op.getOperand(3), Op.getOperand(4));
- return DAG.getNode(ISD::SELECT, dl, MVT::i32, Cond, Op.getOperand(0),
- Op.getOperand(1));
-}
-
-SDValue XCoreTargetLowering::
-getGlobalAddressWrapper(SDValue GA, const GlobalValue *GV,
- SelectionDAG &DAG) const
-{
+SDValue XCoreTargetLowering::getGlobalAddressWrapper(SDValue GA,
+ const GlobalValue *GV,
+ SelectionDAG &DAG) const {
// FIXME there is no actual debug info here
- DebugLoc dl = GA.getDebugLoc();
- const GlobalValue *UnderlyingGV = GV;
- // If GV is an alias then use the aliasee to determine the wrapper type
- if (const GlobalAlias *GA = dyn_cast<GlobalAlias>(GV))
- UnderlyingGV = GA->resolveAliasedGlobal();
- if (const GlobalVariable *GVar = dyn_cast<GlobalVariable>(UnderlyingGV)) {
- if (GVar->isConstant())
- return DAG.getNode(XCoreISD::CPRelativeWrapper, dl, MVT::i32, GA);
- return DAG.getNode(XCoreISD::DPRelativeWrapper, dl, MVT::i32, GA);
- }
- return DAG.getNode(XCoreISD::PCRelativeWrapper, dl, MVT::i32, GA);
-}
+ SDLoc dl(GA);
-SDValue XCoreTargetLowering::
-LowerGlobalAddress(SDValue Op, SelectionDAG &DAG) const
-{
- const GlobalValue *GV = cast<GlobalAddressSDNode>(Op)->getGlobal();
- SDValue GA = DAG.getTargetGlobalAddress(GV, Op.getDebugLoc(), MVT::i32);
- return getGlobalAddressWrapper(GA, GV, DAG);
-}
+ if (GV->getType()->getElementType()->isFunctionTy())
+ return DAG.getNode(XCoreISD::PCRelativeWrapper, dl, MVT::i32, GA);
-static inline SDValue BuildGetId(SelectionDAG &DAG, DebugLoc dl) {
- return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, MVT::i32,
- DAG.getConstant(Intrinsic::xcore_getid, MVT::i32));
+ const auto *GVar = dyn_cast<GlobalVariable>(GV);
+ if ((GV->hasSection() && StringRef(GV->getSection()).startswith(".cp.")) ||
+ (GVar && GVar->isConstant() && GV->hasLocalLinkage()))
+ return DAG.getNode(XCoreISD::CPRelativeWrapper, dl, MVT::i32, GA);
+
+ return DAG.getNode(XCoreISD::DPRelativeWrapper, dl, MVT::i32, GA);
}
-static inline bool isZeroLengthArray(Type *Ty) {
- ArrayType *AT = dyn_cast_or_null<ArrayType>(Ty);
- return AT && (AT->getNumElements() == 0);
+static bool IsSmallObject(const GlobalValue *GV, const XCoreTargetLowering &XTL) {
+ if (XTL.getTargetMachine().getCodeModel() == CodeModel::Small)
+ return true;
+
+ Type *ObjType = GV->getType()->getPointerElementType();
+ if (!ObjType->isSized())
+ return false;
+
+ auto &DL = GV->getParent()->getDataLayout();
+ unsigned ObjSize = DL.getTypeAllocSize(ObjType);
+ return ObjSize < CodeModelLargeSize && ObjSize != 0;
}
SDValue XCoreTargetLowering::
-LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const
+LowerGlobalAddress(SDValue Op, SelectionDAG &DAG) const
{
- // FIXME there isn't really debug info here
- DebugLoc dl = Op.getDebugLoc();
- // transform to label + getid() * size
- const GlobalValue *GV = cast<GlobalAddressSDNode>(Op)->getGlobal();
- SDValue GA = DAG.getTargetGlobalAddress(GV, dl, MVT::i32);
- const GlobalVariable *GVar = dyn_cast<GlobalVariable>(GV);
- if (!GVar) {
- // If GV is an alias then use the aliasee to determine size
- if (const GlobalAlias *GA = dyn_cast<GlobalAlias>(GV))
- GVar = dyn_cast_or_null<GlobalVariable>(GA->resolveAliasedGlobal());
- }
- if (!GVar) {
- llvm_unreachable("Thread local object not a GlobalVariable?");
- }
- Type *Ty = cast<PointerType>(GV->getType())->getElementType();
- if (!Ty->isSized() || isZeroLengthArray(Ty)) {
-#ifndef NDEBUG
- errs() << "Size of thread local object " << GVar->getName()
- << " is unknown\n";
-#endif
- llvm_unreachable(0);
+ const GlobalAddressSDNode *GN = cast<GlobalAddressSDNode>(Op);
+ const GlobalValue *GV = GN->getGlobal();
+ SDLoc DL(GN);
+ int64_t Offset = GN->getOffset();
+ if (IsSmallObject(GV, *this)) {
+ // We can only fold positive offsets that are a multiple of the word size.
+ int64_t FoldedOffset = std::max(Offset & ~3, (int64_t)0);
+ SDValue GA = DAG.getTargetGlobalAddress(GV, DL, MVT::i32, FoldedOffset);
+ GA = getGlobalAddressWrapper(GA, GV, DAG);
+ // Handle the rest of the offset.
+ if (Offset != FoldedOffset) {
+ SDValue Remaining = DAG.getConstant(Offset - FoldedOffset, DL, MVT::i32);
+ GA = DAG.getNode(ISD::ADD, DL, MVT::i32, GA, Remaining);
+ }
+ return GA;
+ } else {
+ // Ideally we would not fold in offset with an index <= 11.
+ Type *Ty = Type::getInt8PtrTy(*DAG.getContext());
+ Constant *GA = ConstantExpr::getBitCast(const_cast<GlobalValue*>(GV), Ty);
+ Ty = Type::getInt32Ty(*DAG.getContext());
+ Constant *Idx = ConstantInt::get(Ty, Offset);
+ Constant *GAI = ConstantExpr::getGetElementPtr(
+ Type::getInt8Ty(*DAG.getContext()), GA, Idx);
+ SDValue CP = DAG.getConstantPool(GAI, MVT::i32);
+ return DAG.getLoad(getPointerTy(DAG.getDataLayout()), DL,
+ DAG.getEntryNode(), CP, MachinePointerInfo(), false,
+ false, false, 0);
}
- SDValue base = getGlobalAddressWrapper(GA, GV, DAG);
- const DataLayout *TD = TM.getDataLayout();
- unsigned Size = TD->getTypeAllocSize(Ty);
- SDValue offset = DAG.getNode(ISD::MUL, dl, MVT::i32, BuildGetId(DAG, dl),
- DAG.getConstant(Size, MVT::i32));
- return DAG.getNode(ISD::ADD, dl, MVT::i32, base, offset);
}
SDValue XCoreTargetLowering::
LowerBlockAddress(SDValue Op, SelectionDAG &DAG) const
{
- DebugLoc DL = Op.getDebugLoc();
-
+ SDLoc DL(Op);
+ auto PtrVT = getPointerTy(DAG.getDataLayout());
const BlockAddress *BA = cast<BlockAddressSDNode>(Op)->getBlockAddress();
- SDValue Result = DAG.getTargetBlockAddress(BA, getPointerTy());
+ SDValue Result = DAG.getTargetBlockAddress(BA, PtrVT);
- return DAG.getNode(XCoreISD::PCRelativeWrapper, DL, getPointerTy(), Result);
+ return DAG.getNode(XCoreISD::PCRelativeWrapper, DL, PtrVT, Result);
}
SDValue XCoreTargetLowering::
{
ConstantPoolSDNode *CP = cast<ConstantPoolSDNode>(Op);
// FIXME there isn't really debug info here
- DebugLoc dl = CP->getDebugLoc();
+ SDLoc dl(CP);
EVT PtrVT = Op.getValueType();
SDValue Res;
if (CP->isMachineConstantPoolEntry()) {
Res = DAG.getTargetConstantPool(CP->getMachineCPVal(), PtrVT,
- CP->getAlignment());
+ CP->getAlignment(), CP->getOffset());
} else {
Res = DAG.getTargetConstantPool(CP->getConstVal(), PtrVT,
- CP->getAlignment());
+ CP->getAlignment(), CP->getOffset());
}
return DAG.getNode(XCoreISD::CPRelativeWrapper, dl, MVT::i32, Res);
}
SDValue Chain = Op.getOperand(0);
SDValue Table = Op.getOperand(1);
SDValue Index = Op.getOperand(2);
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
JumpTableSDNode *JT = cast<JumpTableSDNode>(Table);
unsigned JTI = JT->getIndex();
MachineFunction &MF = DAG.getMachineFunction();
}
assert((NumEntries >> 31) == 0);
SDValue ScaledIndex = DAG.getNode(ISD::SHL, dl, MVT::i32, Index,
- DAG.getConstant(1, MVT::i32));
+ DAG.getConstant(1, dl, MVT::i32));
return DAG.getNode(XCoreISD::BR_JT32, dl, MVT::Other, Chain, TargetJT,
ScaledIndex);
}
-static bool
-IsWordAlignedBasePlusConstantOffset(SDValue Addr, SDValue &AlignedBase,
- int64_t &Offset)
+SDValue XCoreTargetLowering::
+lowerLoadWordFromAlignedBasePlusOffset(SDLoc DL, SDValue Chain, SDValue Base,
+ int64_t Offset, SelectionDAG &DAG) const
{
- if (Addr.getOpcode() != ISD::ADD) {
- return false;
- }
- ConstantSDNode *CN = 0;
- if (!(CN = dyn_cast<ConstantSDNode>(Addr.getOperand(1)))) {
- return false;
- }
- int64_t off = CN->getSExtValue();
- const SDValue &Base = Addr.getOperand(0);
- const SDValue *Root = &Base;
- if (Base.getOpcode() == ISD::ADD &&
- Base.getOperand(1).getOpcode() == ISD::SHL) {
- ConstantSDNode *CN = dyn_cast<ConstantSDNode>(Base.getOperand(1)
- .getOperand(1));
- if (CN && (CN->getSExtValue() >= 2)) {
- Root = &Base.getOperand(0);
- }
- }
- if (isa<FrameIndexSDNode>(*Root)) {
- // All frame indicies are word aligned
- AlignedBase = Base;
- Offset = off;
- return true;
- }
- if (Root->getOpcode() == XCoreISD::DPRelativeWrapper ||
- Root->getOpcode() == XCoreISD::CPRelativeWrapper) {
- // All dp / cp relative addresses are word aligned
- AlignedBase = Base;
- Offset = off;
- return true;
+ auto PtrVT = getPointerTy(DAG.getDataLayout());
+ if ((Offset & 0x3) == 0) {
+ return DAG.getLoad(PtrVT, DL, Chain, Base, MachinePointerInfo(), false,
+ false, false, 0);
}
- // Check for an aligned global variable.
- if (GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(*Root)) {
- const GlobalValue *GV = GA->getGlobal();
- if (GA->getOffset() == 0 && GV->getAlignment() >= 4) {
- AlignedBase = Base;
- Offset = off;
- return true;
- }
+ // Lower to pair of consecutive word aligned loads plus some bit shifting.
+ int32_t HighOffset = RoundUpToAlignment(Offset, 4);
+ int32_t LowOffset = HighOffset - 4;
+ SDValue LowAddr, HighAddr;
+ if (GlobalAddressSDNode *GASD =
+ dyn_cast<GlobalAddressSDNode>(Base.getNode())) {
+ LowAddr = DAG.getGlobalAddress(GASD->getGlobal(), DL, Base.getValueType(),
+ LowOffset);
+ HighAddr = DAG.getGlobalAddress(GASD->getGlobal(), DL, Base.getValueType(),
+ HighOffset);
+ } else {
+ LowAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base,
+ DAG.getConstant(LowOffset, DL, MVT::i32));
+ HighAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base,
+ DAG.getConstant(HighOffset, DL, MVT::i32));
}
- return false;
+ SDValue LowShift = DAG.getConstant((Offset - LowOffset) * 8, DL, MVT::i32);
+ SDValue HighShift = DAG.getConstant((HighOffset - Offset) * 8, DL, MVT::i32);
+
+ SDValue Low = DAG.getLoad(PtrVT, DL, Chain, LowAddr, MachinePointerInfo(),
+ false, false, false, 0);
+ SDValue High = DAG.getLoad(PtrVT, DL, Chain, HighAddr, MachinePointerInfo(),
+ false, false, false, 0);
+ SDValue LowShifted = DAG.getNode(ISD::SRL, DL, MVT::i32, Low, LowShift);
+ SDValue HighShifted = DAG.getNode(ISD::SHL, DL, MVT::i32, High, HighShift);
+ SDValue Result = DAG.getNode(ISD::OR, DL, MVT::i32, LowShifted, HighShifted);
+ Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Low.getValue(1),
+ High.getValue(1));
+ SDValue Ops[] = { Result, Chain };
+ return DAG.getMergeValues(Ops, DL);
+}
+
+static bool isWordAligned(SDValue Value, SelectionDAG &DAG)
+{
+ APInt KnownZero, KnownOne;
+ DAG.computeKnownBits(Value, KnownZero, KnownOne);
+ return KnownZero.countTrailingOnes() >= 2;
}
SDValue XCoreTargetLowering::
LowerLOAD(SDValue Op, SelectionDAG &DAG) const {
+ const TargetLowering &TLI = DAG.getTargetLoweringInfo();
LoadSDNode *LD = cast<LoadSDNode>(Op);
assert(LD->getExtensionType() == ISD::NON_EXTLOAD &&
"Unexpected extension type");
assert(LD->getMemoryVT() == MVT::i32 && "Unexpected load EVT");
- if (allowsUnalignedMemoryAccesses(LD->getMemoryVT()))
+ if (allowsMisalignedMemoryAccesses(LD->getMemoryVT(),
+ LD->getAddressSpace(),
+ LD->getAlignment()))
return SDValue();
- unsigned ABIAlignment = getDataLayout()->
- getABITypeAlignment(LD->getMemoryVT().getTypeForEVT(*DAG.getContext()));
+ auto &TD = DAG.getDataLayout();
+ unsigned ABIAlignment = TD.getABITypeAlignment(
+ LD->getMemoryVT().getTypeForEVT(*DAG.getContext()));
// Leave aligned load alone.
if (LD->getAlignment() >= ABIAlignment)
return SDValue();
SDValue Chain = LD->getChain();
SDValue BasePtr = LD->getBasePtr();
- DebugLoc DL = Op.getDebugLoc();
-
- SDValue Base;
- int64_t Offset;
- if (!LD->isVolatile() &&
- IsWordAlignedBasePlusConstantOffset(BasePtr, Base, Offset)) {
- if (Offset % 4 == 0) {
- // We've managed to infer better alignment information than the load
- // already has. Use an aligned load.
- //
- return DAG.getLoad(getPointerTy(), DL, Chain, BasePtr,
- MachinePointerInfo(),
- false, false, false, 0);
+ SDLoc DL(Op);
+
+ if (!LD->isVolatile()) {
+ const GlobalValue *GV;
+ int64_t Offset = 0;
+ if (DAG.isBaseWithConstantOffset(BasePtr) &&
+ isWordAligned(BasePtr->getOperand(0), DAG)) {
+ SDValue NewBasePtr = BasePtr->getOperand(0);
+ Offset = cast<ConstantSDNode>(BasePtr->getOperand(1))->getSExtValue();
+ return lowerLoadWordFromAlignedBasePlusOffset(DL, Chain, NewBasePtr,
+ Offset, DAG);
+ }
+ if (TLI.isGAPlusOffset(BasePtr.getNode(), GV, Offset) &&
+ MinAlign(GV->getAlignment(), 4) == 4) {
+ SDValue NewBasePtr = DAG.getGlobalAddress(GV, DL,
+ BasePtr->getValueType(0));
+ return lowerLoadWordFromAlignedBasePlusOffset(DL, Chain, NewBasePtr,
+ Offset, DAG);
}
- // Lower to
- // ldw low, base[offset >> 2]
- // ldw high, base[(offset >> 2) + 1]
- // shr low_shifted, low, (offset & 0x3) * 8
- // shl high_shifted, high, 32 - (offset & 0x3) * 8
- // or result, low_shifted, high_shifted
- SDValue LowOffset = DAG.getConstant(Offset & ~0x3, MVT::i32);
- SDValue HighOffset = DAG.getConstant((Offset & ~0x3) + 4, MVT::i32);
- SDValue LowShift = DAG.getConstant((Offset & 0x3) * 8, MVT::i32);
- SDValue HighShift = DAG.getConstant(32 - (Offset & 0x3) * 8, MVT::i32);
-
- SDValue LowAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, LowOffset);
- SDValue HighAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, HighOffset);
-
- SDValue Low = DAG.getLoad(getPointerTy(), DL, Chain,
- LowAddr, MachinePointerInfo(),
- false, false, false, 0);
- SDValue High = DAG.getLoad(getPointerTy(), DL, Chain,
- HighAddr, MachinePointerInfo(),
- false, false, false, 0);
- SDValue LowShifted = DAG.getNode(ISD::SRL, DL, MVT::i32, Low, LowShift);
- SDValue HighShifted = DAG.getNode(ISD::SHL, DL, MVT::i32, High, HighShift);
- SDValue Result = DAG.getNode(ISD::OR, DL, MVT::i32, LowShifted, HighShifted);
- Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Low.getValue(1),
- High.getValue(1));
- SDValue Ops[] = { Result, Chain };
- return DAG.getMergeValues(Ops, 2, DL);
}
if (LD->getAlignment() == 2) {
SDValue Low = DAG.getExtLoad(ISD::ZEXTLOAD, DL, MVT::i32, Chain,
BasePtr, LD->getPointerInfo(), MVT::i16,
- LD->isVolatile(), LD->isNonTemporal(), 2);
+ LD->isVolatile(), LD->isNonTemporal(),
+ LD->isInvariant(), 2);
SDValue HighAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, BasePtr,
- DAG.getConstant(2, MVT::i32));
+ DAG.getConstant(2, DL, MVT::i32));
SDValue High = DAG.getExtLoad(ISD::EXTLOAD, DL, MVT::i32, Chain,
HighAddr,
LD->getPointerInfo().getWithOffset(2),
MVT::i16, LD->isVolatile(),
- LD->isNonTemporal(), 2);
+ LD->isNonTemporal(), LD->isInvariant(), 2);
SDValue HighShifted = DAG.getNode(ISD::SHL, DL, MVT::i32, High,
- DAG.getConstant(16, MVT::i32));
+ DAG.getConstant(16, DL, MVT::i32));
SDValue Result = DAG.getNode(ISD::OR, DL, MVT::i32, Low, HighShifted);
Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Low.getValue(1),
High.getValue(1));
SDValue Ops[] = { Result, Chain };
- return DAG.getMergeValues(Ops, 2, DL);
+ return DAG.getMergeValues(Ops, DL);
}
// Lower to a call to __misaligned_load(BasePtr).
- Type *IntPtrTy = getDataLayout()->getIntPtrType(*DAG.getContext());
+ Type *IntPtrTy = TD.getIntPtrType(*DAG.getContext());
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Node = BasePtr;
Args.push_back(Entry);
- TargetLowering::CallLoweringInfo CLI(Chain, IntPtrTy, false, false,
- false, false, 0, CallingConv::C, /*isTailCall=*/false,
- /*doesNotRet=*/false, /*isReturnValueUsed=*/true,
- DAG.getExternalSymbol("__misaligned_load", getPointerTy()),
- Args, DAG, DL);
- std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
-
- SDValue Ops[] =
- { CallResult.first, CallResult.second };
+ TargetLowering::CallLoweringInfo CLI(DAG);
+ CLI.setDebugLoc(DL).setChain(Chain).setCallee(
+ CallingConv::C, IntPtrTy,
+ DAG.getExternalSymbol("__misaligned_load",
+ getPointerTy(DAG.getDataLayout())),
+ std::move(Args), 0);
- return DAG.getMergeValues(Ops, 2, DL);
+ std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
+ SDValue Ops[] = { CallResult.first, CallResult.second };
+ return DAG.getMergeValues(Ops, DL);
}
SDValue XCoreTargetLowering::
StoreSDNode *ST = cast<StoreSDNode>(Op);
assert(!ST->isTruncatingStore() && "Unexpected store type");
assert(ST->getMemoryVT() == MVT::i32 && "Unexpected store EVT");
- if (allowsUnalignedMemoryAccesses(ST->getMemoryVT())) {
+ if (allowsMisalignedMemoryAccesses(ST->getMemoryVT(),
+ ST->getAddressSpace(),
+ ST->getAlignment())) {
return SDValue();
}
- unsigned ABIAlignment = getDataLayout()->
- getABITypeAlignment(ST->getMemoryVT().getTypeForEVT(*DAG.getContext()));
+ unsigned ABIAlignment = DAG.getDataLayout().getABITypeAlignment(
+ ST->getMemoryVT().getTypeForEVT(*DAG.getContext()));
// Leave aligned store alone.
if (ST->getAlignment() >= ABIAlignment) {
return SDValue();
SDValue Chain = ST->getChain();
SDValue BasePtr = ST->getBasePtr();
SDValue Value = ST->getValue();
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
if (ST->getAlignment() == 2) {
SDValue Low = Value;
SDValue High = DAG.getNode(ISD::SRL, dl, MVT::i32, Value,
- DAG.getConstant(16, MVT::i32));
+ DAG.getConstant(16, dl, MVT::i32));
SDValue StoreLow = DAG.getTruncStore(Chain, dl, Low, BasePtr,
ST->getPointerInfo(), MVT::i16,
ST->isVolatile(), ST->isNonTemporal(),
2);
SDValue HighAddr = DAG.getNode(ISD::ADD, dl, MVT::i32, BasePtr,
- DAG.getConstant(2, MVT::i32));
+ DAG.getConstant(2, dl, MVT::i32));
SDValue StoreHigh = DAG.getTruncStore(Chain, dl, High, HighAddr,
ST->getPointerInfo().getWithOffset(2),
MVT::i16, ST->isVolatile(),
}
// Lower to a call to __misaligned_store(BasePtr, Value).
- Type *IntPtrTy = getDataLayout()->getIntPtrType(*DAG.getContext());
+ Type *IntPtrTy = DAG.getDataLayout().getIntPtrType(*DAG.getContext());
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Node = Value;
Args.push_back(Entry);
- TargetLowering::CallLoweringInfo CLI(Chain,
- Type::getVoidTy(*DAG.getContext()), false, false,
- false, false, 0, CallingConv::C, /*isTailCall=*/false,
- /*doesNotRet=*/false, /*isReturnValueUsed=*/true,
- DAG.getExternalSymbol("__misaligned_store", getPointerTy()),
- Args, DAG, dl);
- std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
+ TargetLowering::CallLoweringInfo CLI(DAG);
+ CLI.setDebugLoc(dl).setChain(Chain).setCallee(
+ CallingConv::C, Type::getVoidTy(*DAG.getContext()),
+ DAG.getExternalSymbol("__misaligned_store",
+ getPointerTy(DAG.getDataLayout())),
+ std::move(Args), 0);
+ std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
return CallResult.second;
}
{
assert(Op.getValueType() == MVT::i32 && Op.getOpcode() == ISD::SMUL_LOHI &&
"Unexpected operand to lower!");
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
SDValue LHS = Op.getOperand(0);
SDValue RHS = Op.getOperand(1);
- SDValue Zero = DAG.getConstant(0, MVT::i32);
+ SDValue Zero = DAG.getConstant(0, dl, MVT::i32);
SDValue Hi = DAG.getNode(XCoreISD::MACCS, dl,
DAG.getVTList(MVT::i32, MVT::i32), Zero, Zero,
LHS, RHS);
SDValue Lo(Hi.getNode(), 1);
SDValue Ops[] = { Lo, Hi };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
SDValue XCoreTargetLowering::
{
assert(Op.getValueType() == MVT::i32 && Op.getOpcode() == ISD::UMUL_LOHI &&
"Unexpected operand to lower!");
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
SDValue LHS = Op.getOperand(0);
SDValue RHS = Op.getOperand(1);
- SDValue Zero = DAG.getConstant(0, MVT::i32);
+ SDValue Zero = DAG.getConstant(0, dl, MVT::i32);
SDValue Hi = DAG.getNode(XCoreISD::LMUL, dl,
DAG.getVTList(MVT::i32, MVT::i32), LHS, RHS,
Zero, Zero);
SDValue Lo(Hi.getNode(), 1);
SDValue Ops[] = { Lo, Hi };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
/// isADDADDMUL - Return whether Op is in a form that is equivalent to
} else {
return SDValue();
}
- DebugLoc dl = N->getDebugLoc();
+ SDLoc dl(N);
SDValue LL, RL, AddendL, AddendH;
LL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Mul.getOperand(0), DAG.getConstant(0, MVT::i32));
+ Mul.getOperand(0), DAG.getConstant(0, dl, MVT::i32));
RL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Mul.getOperand(1), DAG.getConstant(0, MVT::i32));
+ Mul.getOperand(1), DAG.getConstant(0, dl, MVT::i32));
AddendL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Other, DAG.getConstant(0, MVT::i32));
+ Other, DAG.getConstant(0, dl, MVT::i32));
AddendH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Other, DAG.getConstant(1, MVT::i32));
+ Other, DAG.getConstant(1, dl, MVT::i32));
APInt HighMask = APInt::getHighBitsSet(64, 32);
unsigned LHSSB = DAG.ComputeNumSignBits(Mul.getOperand(0));
unsigned RHSSB = DAG.ComputeNumSignBits(Mul.getOperand(1));
}
SDValue LH, RH;
LH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Mul.getOperand(0), DAG.getConstant(1, MVT::i32));
+ Mul.getOperand(0), DAG.getConstant(1, dl, MVT::i32));
RH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Mul.getOperand(1), DAG.getConstant(1, MVT::i32));
+ Mul.getOperand(1), DAG.getConstant(1, dl, MVT::i32));
SDValue Hi = DAG.getNode(XCoreISD::MACCU, dl,
DAG.getVTList(MVT::i32, MVT::i32), AddendH,
AddendL, LL, RL);
if (N->getOpcode() == ISD::ADD) {
SDValue Result = TryExpandADDWithMul(N, DAG);
- if (Result.getNode() != 0)
+ if (Result.getNode())
return Result;
}
- DebugLoc dl = N->getDebugLoc();
+ SDLoc dl(N);
// Extract components
SDValue LHSL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- N->getOperand(0), DAG.getConstant(0, MVT::i32));
+ N->getOperand(0),
+ DAG.getConstant(0, dl, MVT::i32));
SDValue LHSH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- N->getOperand(0), DAG.getConstant(1, MVT::i32));
+ N->getOperand(0),
+ DAG.getConstant(1, dl, MVT::i32));
SDValue RHSL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- N->getOperand(1), DAG.getConstant(0, MVT::i32));
+ N->getOperand(1),
+ DAG.getConstant(0, dl, MVT::i32));
SDValue RHSH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- N->getOperand(1), DAG.getConstant(1, MVT::i32));
+ N->getOperand(1),
+ DAG.getConstant(1, dl, MVT::i32));
// Expand
unsigned Opcode = (N->getOpcode() == ISD::ADD) ? XCoreISD::LADD :
XCoreISD::LSUB;
- SDValue Zero = DAG.getConstant(0, MVT::i32);
+ SDValue Zero = DAG.getConstant(0, dl, MVT::i32);
SDValue Lo = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32),
LHSL, RHSL, Zero);
SDValue Carry(Lo.getNode(), 1);
SDValue XCoreTargetLowering::
LowerVAARG(SDValue Op, SelectionDAG &DAG) const
{
- llvm_unreachable("unimplemented");
- // FIXME Arguments passed by reference need a extra dereference.
+ // Whist llvm does not support aggregate varargs we can ignore
+ // the possibility of the ValueType being an implicit byVal vararg.
SDNode *Node = Op.getNode();
- DebugLoc dl = Node->getDebugLoc();
- const Value *V = cast<SrcValueSDNode>(Node->getOperand(2))->getValue();
- EVT VT = Node->getValueType(0);
- SDValue VAList = DAG.getLoad(getPointerTy(), dl, Node->getOperand(0),
- Node->getOperand(1), MachinePointerInfo(V),
+ EVT VT = Node->getValueType(0); // not an aggregate
+ SDValue InChain = Node->getOperand(0);
+ SDValue VAListPtr = Node->getOperand(1);
+ EVT PtrVT = VAListPtr.getValueType();
+ const Value *SV = cast<SrcValueSDNode>(Node->getOperand(2))->getValue();
+ SDLoc dl(Node);
+ SDValue VAList = DAG.getLoad(PtrVT, dl, InChain,
+ VAListPtr, MachinePointerInfo(SV),
false, false, false, 0);
// Increment the pointer, VAList, to the next vararg
- SDValue Tmp3 = DAG.getNode(ISD::ADD, dl, getPointerTy(), VAList,
- DAG.getConstant(VT.getSizeInBits(),
- getPointerTy()));
+ SDValue nextPtr = DAG.getNode(ISD::ADD, dl, PtrVT, VAList,
+ DAG.getIntPtrConstant(VT.getSizeInBits() / 8,
+ dl));
// Store the incremented VAList to the legalized pointer
- Tmp3 = DAG.getStore(VAList.getValue(1), dl, Tmp3, Node->getOperand(1),
- MachinePointerInfo(V), false, false, 0);
+ InChain = DAG.getStore(VAList.getValue(1), dl, nextPtr, VAListPtr,
+ MachinePointerInfo(SV), false, false, 0);
// Load the actual argument out of the pointer VAList
- return DAG.getLoad(VT, dl, Tmp3, VAList, MachinePointerInfo(),
+ return DAG.getLoad(VT, dl, InChain, VAList, MachinePointerInfo(),
false, false, false, 0);
}
SDValue XCoreTargetLowering::
LowerVASTART(SDValue Op, SelectionDAG &DAG) const
{
- DebugLoc dl = Op.getDebugLoc();
+ SDLoc dl(Op);
// vastart stores the address of the VarArgsFrameIndex slot into the
// memory location argument
MachineFunction &MF = DAG.getMachineFunction();
SDValue XCoreTargetLowering::LowerFRAMEADDR(SDValue Op,
SelectionDAG &DAG) const {
- DebugLoc dl = Op.getDebugLoc();
+ // This nodes represent llvm.frameaddress on the DAG.
+ // It takes one operand, the index of the frame address to return.
+ // An index of zero corresponds to the current function's frame address.
+ // An index of one to the parent's frame address, and so on.
+ // Depths > 0 not supported yet!
+ if (cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue() > 0)
+ return SDValue();
+
+ MachineFunction &MF = DAG.getMachineFunction();
+ const TargetRegisterInfo *RegInfo = Subtarget.getRegisterInfo();
+ return DAG.getCopyFromReg(DAG.getEntryNode(), SDLoc(Op),
+ RegInfo->getFrameRegister(MF), MVT::i32);
+}
+
+SDValue XCoreTargetLowering::
+LowerRETURNADDR(SDValue Op, SelectionDAG &DAG) const {
+ // This nodes represent llvm.returnaddress on the DAG.
+ // It takes one operand, the index of the return address to return.
+ // An index of zero corresponds to the current function's return address.
+ // An index of one to the parent's return address, and so on.
// Depths > 0 not supported yet!
if (cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue() > 0)
return SDValue();
MachineFunction &MF = DAG.getMachineFunction();
- const TargetRegisterInfo *RegInfo = getTargetMachine().getRegisterInfo();
- return DAG.getCopyFromReg(DAG.getEntryNode(), dl,
+ XCoreFunctionInfo *XFI = MF.getInfo<XCoreFunctionInfo>();
+ int FI = XFI->createLRSpillSlot(MF);
+ SDValue FIN = DAG.getFrameIndex(FI, MVT::i32);
+ return DAG.getLoad(
+ getPointerTy(DAG.getDataLayout()), SDLoc(Op), DAG.getEntryNode(), FIN,
+ MachinePointerInfo::getFixedStack(MF, FI), false, false, false, 0);
+}
+
+SDValue XCoreTargetLowering::
+LowerFRAME_TO_ARGS_OFFSET(SDValue Op, SelectionDAG &DAG) const {
+ // This node represents offset from frame pointer to first on-stack argument.
+ // This is needed for correct stack adjustment during unwind.
+ // However, we don't know the offset until after the frame has be finalised.
+ // This is done during the XCoreFTAOElim pass.
+ return DAG.getNode(XCoreISD::FRAME_TO_ARGS_OFFSET, SDLoc(Op), MVT::i32);
+}
+
+SDValue XCoreTargetLowering::
+LowerEH_RETURN(SDValue Op, SelectionDAG &DAG) const {
+ // OUTCHAIN = EH_RETURN(INCHAIN, OFFSET, HANDLER)
+ // This node represents 'eh_return' gcc dwarf builtin, which is used to
+ // return from exception. The general meaning is: adjust stack by OFFSET and
+ // pass execution to HANDLER.
+ MachineFunction &MF = DAG.getMachineFunction();
+ SDValue Chain = Op.getOperand(0);
+ SDValue Offset = Op.getOperand(1);
+ SDValue Handler = Op.getOperand(2);
+ SDLoc dl(Op);
+
+ // Absolute SP = (FP + FrameToArgs) + Offset
+ const TargetRegisterInfo *RegInfo = Subtarget.getRegisterInfo();
+ SDValue Stack = DAG.getCopyFromReg(DAG.getEntryNode(), dl,
RegInfo->getFrameRegister(MF), MVT::i32);
+ SDValue FrameToArgs = DAG.getNode(XCoreISD::FRAME_TO_ARGS_OFFSET, dl,
+ MVT::i32);
+ Stack = DAG.getNode(ISD::ADD, dl, MVT::i32, Stack, FrameToArgs);
+ Stack = DAG.getNode(ISD::ADD, dl, MVT::i32, Stack, Offset);
+
+ // R0=ExceptionPointerRegister R1=ExceptionSelectorRegister
+ // which leaves 2 caller saved registers, R2 & R3 for us to use.
+ unsigned StackReg = XCore::R2;
+ unsigned HandlerReg = XCore::R3;
+
+ SDValue OutChains[] = {
+ DAG.getCopyToReg(Chain, dl, StackReg, Stack),
+ DAG.getCopyToReg(Chain, dl, HandlerReg, Handler)
+ };
+
+ Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains);
+
+ return DAG.getNode(XCoreISD::EH_RETURN, dl, MVT::Other, Chain,
+ DAG.getRegister(StackReg, MVT::i32),
+ DAG.getRegister(HandlerReg, MVT::i32));
+
}
SDValue XCoreTargetLowering::
SDValue Addr = Trmp;
- DebugLoc dl = Op.getDebugLoc();
- OutChains[0] = DAG.getStore(Chain, dl, DAG.getConstant(0x0a3cd805, MVT::i32),
- Addr, MachinePointerInfo(TrmpAddr), false, false,
- 0);
+ SDLoc dl(Op);
+ OutChains[0] = DAG.getStore(Chain, dl,
+ DAG.getConstant(0x0a3cd805, dl, MVT::i32), Addr,
+ MachinePointerInfo(TrmpAddr), false, false, 0);
Addr = DAG.getNode(ISD::ADD, dl, MVT::i32, Trmp,
- DAG.getConstant(4, MVT::i32));
- OutChains[1] = DAG.getStore(Chain, dl, DAG.getConstant(0xd80456c0, MVT::i32),
- Addr, MachinePointerInfo(TrmpAddr, 4), false,
- false, 0);
+ DAG.getConstant(4, dl, MVT::i32));
+ OutChains[1] = DAG.getStore(Chain, dl,
+ DAG.getConstant(0xd80456c0, dl, MVT::i32), Addr,
+ MachinePointerInfo(TrmpAddr, 4), false, false, 0);
Addr = DAG.getNode(ISD::ADD, dl, MVT::i32, Trmp,
- DAG.getConstant(8, MVT::i32));
- OutChains[2] = DAG.getStore(Chain, dl, DAG.getConstant(0x27fb0a3c, MVT::i32),
- Addr, MachinePointerInfo(TrmpAddr, 8), false,
- false, 0);
+ DAG.getConstant(8, dl, MVT::i32));
+ OutChains[2] = DAG.getStore(Chain, dl,
+ DAG.getConstant(0x27fb0a3c, dl, MVT::i32), Addr,
+ MachinePointerInfo(TrmpAddr, 8), false, false, 0);
Addr = DAG.getNode(ISD::ADD, dl, MVT::i32, Trmp,
- DAG.getConstant(12, MVT::i32));
+ DAG.getConstant(12, dl, MVT::i32));
OutChains[3] = DAG.getStore(Chain, dl, Nest, Addr,
MachinePointerInfo(TrmpAddr, 12), false, false,
0);
Addr = DAG.getNode(ISD::ADD, dl, MVT::i32, Trmp,
- DAG.getConstant(16, MVT::i32));
+ DAG.getConstant(16, dl, MVT::i32));
OutChains[4] = DAG.getStore(Chain, dl, FPtr, Addr,
MachinePointerInfo(TrmpAddr, 16), false, false,
0);
- return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains, 5);
+ return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains);
}
SDValue XCoreTargetLowering::
LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const {
- DebugLoc DL = Op.getDebugLoc();
+ SDLoc DL(Op);
unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
switch (IntNo) {
case Intrinsic::xcore_crc8:
Op.getOperand(1), Op.getOperand(2) , Op.getOperand(3));
SDValue Crc(Data.getNode(), 1);
SDValue Results[] = { Crc, Data };
- return DAG.getMergeValues(Results, 2, DL);
+ return DAG.getMergeValues(Results, DL);
}
return SDValue();
}
+SDValue XCoreTargetLowering::
+LowerATOMIC_FENCE(SDValue Op, SelectionDAG &DAG) const {
+ SDLoc DL(Op);
+ return DAG.getNode(XCoreISD::MEMBARRIER, DL, MVT::Other, Op.getOperand(0));
+}
+
+SDValue XCoreTargetLowering::
+LowerATOMIC_LOAD(SDValue Op, SelectionDAG &DAG) const {
+ AtomicSDNode *N = cast<AtomicSDNode>(Op);
+ assert(N->getOpcode() == ISD::ATOMIC_LOAD && "Bad Atomic OP");
+ assert(N->getOrdering() <= Monotonic &&
+ "setInsertFencesForAtomic(true) and yet greater than Monotonic");
+ if (N->getMemoryVT() == MVT::i32) {
+ if (N->getAlignment() < 4)
+ report_fatal_error("atomic load must be aligned");
+ return DAG.getLoad(getPointerTy(DAG.getDataLayout()), SDLoc(Op),
+ N->getChain(), N->getBasePtr(), N->getPointerInfo(),
+ N->isVolatile(), N->isNonTemporal(), N->isInvariant(),
+ N->getAlignment(), N->getAAInfo(), N->getRanges());
+ }
+ if (N->getMemoryVT() == MVT::i16) {
+ if (N->getAlignment() < 2)
+ report_fatal_error("atomic load must be aligned");
+ return DAG.getExtLoad(ISD::EXTLOAD, SDLoc(Op), MVT::i32, N->getChain(),
+ N->getBasePtr(), N->getPointerInfo(), MVT::i16,
+ N->isVolatile(), N->isNonTemporal(),
+ N->isInvariant(), N->getAlignment(), N->getAAInfo());
+ }
+ if (N->getMemoryVT() == MVT::i8)
+ return DAG.getExtLoad(ISD::EXTLOAD, SDLoc(Op), MVT::i32, N->getChain(),
+ N->getBasePtr(), N->getPointerInfo(), MVT::i8,
+ N->isVolatile(), N->isNonTemporal(),
+ N->isInvariant(), N->getAlignment(), N->getAAInfo());
+ return SDValue();
+}
+
+SDValue XCoreTargetLowering::
+LowerATOMIC_STORE(SDValue Op, SelectionDAG &DAG) const {
+ AtomicSDNode *N = cast<AtomicSDNode>(Op);
+ assert(N->getOpcode() == ISD::ATOMIC_STORE && "Bad Atomic OP");
+ assert(N->getOrdering() <= Monotonic &&
+ "setInsertFencesForAtomic(true) and yet greater than Monotonic");
+ if (N->getMemoryVT() == MVT::i32) {
+ if (N->getAlignment() < 4)
+ report_fatal_error("atomic store must be aligned");
+ return DAG.getStore(N->getChain(), SDLoc(Op), N->getVal(),
+ N->getBasePtr(), N->getPointerInfo(),
+ N->isVolatile(), N->isNonTemporal(),
+ N->getAlignment(), N->getAAInfo());
+ }
+ if (N->getMemoryVT() == MVT::i16) {
+ if (N->getAlignment() < 2)
+ report_fatal_error("atomic store must be aligned");
+ return DAG.getTruncStore(N->getChain(), SDLoc(Op), N->getVal(),
+ N->getBasePtr(), N->getPointerInfo(), MVT::i16,
+ N->isVolatile(), N->isNonTemporal(),
+ N->getAlignment(), N->getAAInfo());
+ }
+ if (N->getMemoryVT() == MVT::i8)
+ return DAG.getTruncStore(N->getChain(), SDLoc(Op), N->getVal(),
+ N->getBasePtr(), N->getPointerInfo(), MVT::i8,
+ N->isVolatile(), N->isNonTemporal(),
+ N->getAlignment(), N->getAAInfo());
+ return SDValue();
+}
+
//===----------------------------------------------------------------------===//
// Calling Convention Implementation
//===----------------------------------------------------------------------===//
XCoreTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
SmallVectorImpl<SDValue> &InVals) const {
SelectionDAG &DAG = CLI.DAG;
- DebugLoc &dl = CLI.DL;
- SmallVector<ISD::OutputArg, 32> &Outs = CLI.Outs;
- SmallVector<SDValue, 32> &OutVals = CLI.OutVals;
- SmallVector<ISD::InputArg, 32> &Ins = CLI.Ins;
+ SDLoc &dl = CLI.DL;
+ SmallVectorImpl<ISD::OutputArg> &Outs = CLI.Outs;
+ SmallVectorImpl<SDValue> &OutVals = CLI.OutVals;
+ SmallVectorImpl<ISD::InputArg> &Ins = CLI.Ins;
SDValue Chain = CLI.Chain;
SDValue Callee = CLI.Callee;
bool &isTailCall = CLI.IsTailCall;
}
}
+/// LowerCallResult - Lower the result values of a call into the
+/// appropriate copies out of appropriate physical registers / memory locations.
+static SDValue
+LowerCallResult(SDValue Chain, SDValue InFlag,
+ const SmallVectorImpl<CCValAssign> &RVLocs,
+ SDLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals) {
+ SmallVector<std::pair<int, unsigned>, 4> ResultMemLocs;
+ // Copy results out of physical registers.
+ for (unsigned i = 0, e = RVLocs.size(); i != e; ++i) {
+ const CCValAssign &VA = RVLocs[i];
+ if (VA.isRegLoc()) {
+ Chain = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), VA.getValVT(),
+ InFlag).getValue(1);
+ InFlag = Chain.getValue(2);
+ InVals.push_back(Chain.getValue(0));
+ } else {
+ assert(VA.isMemLoc());
+ ResultMemLocs.push_back(std::make_pair(VA.getLocMemOffset(),
+ InVals.size()));
+ // Reserve space for this result.
+ InVals.push_back(SDValue());
+ }
+ }
+
+ // Copy results out of memory.
+ SmallVector<SDValue, 4> MemOpChains;
+ for (unsigned i = 0, e = ResultMemLocs.size(); i != e; ++i) {
+ int offset = ResultMemLocs[i].first;
+ unsigned index = ResultMemLocs[i].second;
+ SDVTList VTs = DAG.getVTList(MVT::i32, MVT::Other);
+ SDValue Ops[] = { Chain, DAG.getConstant(offset / 4, dl, MVT::i32) };
+ SDValue load = DAG.getNode(XCoreISD::LDWSP, dl, VTs, Ops);
+ InVals[index] = load;
+ MemOpChains.push_back(load.getValue(1));
+ }
+
+ // Transform all loads nodes into one single node because
+ // all load nodes are independent of each other.
+ if (!MemOpChains.empty())
+ Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains);
+
+ return Chain;
+}
+
/// LowerCCCCallTo - functions arguments are copied from virtual
/// regs to (physical regs)/(stack frame), CALLSEQ_START and
/// CALLSEQ_END are emitted.
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
const SmallVectorImpl<ISD::InputArg> &Ins,
- DebugLoc dl, SelectionDAG &DAG,
+ SDLoc dl, SelectionDAG &DAG,
SmallVectorImpl<SDValue> &InVals) const {
// Analyze operands of the call, assigning locations to each operand.
SmallVector<CCValAssign, 16> ArgLocs;
- CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(),
- getTargetMachine(), ArgLocs, *DAG.getContext());
+ CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
+ *DAG.getContext());
// The ABI dictates there should be one stack slot available to the callee
// on function entry (for saving lr).
CCInfo.AnalyzeCallOperands(Outs, CC_XCore);
+ SmallVector<CCValAssign, 16> RVLocs;
+ // Analyze return values to determine the number of bytes of stack required.
+ CCState RetCCInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs,
+ *DAG.getContext());
+ RetCCInfo.AllocateStack(CCInfo.getNextStackOffset(), 4);
+ RetCCInfo.AnalyzeCallResult(Ins, RetCC_XCore);
+
// Get a count of how many bytes are to be pushed on the stack.
- unsigned NumBytes = CCInfo.getNextStackOffset();
+ unsigned NumBytes = RetCCInfo.getNextStackOffset();
+ auto PtrVT = getPointerTy(DAG.getDataLayout());
- Chain = DAG.getCALLSEQ_START(Chain,DAG.getConstant(NumBytes,
- getPointerTy(), true));
+ Chain = DAG.getCALLSEQ_START(Chain,
+ DAG.getConstant(NumBytes, dl, PtrVT, true), dl);
SmallVector<std::pair<unsigned, SDValue>, 4> RegsToPass;
SmallVector<SDValue, 12> MemOpChains;
MemOpChains.push_back(DAG.getNode(XCoreISD::STWSP, dl, MVT::Other,
Chain, Arg,
- DAG.getConstant(Offset/4, MVT::i32)));
+ DAG.getConstant(Offset/4, dl,
+ MVT::i32)));
}
}
// Transform all store nodes into one single node because
// all store nodes are independent of each other.
if (!MemOpChains.empty())
- Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
- &MemOpChains[0], MemOpChains.size());
+ Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains);
// Build a sequence of copy-to-reg nodes chained together with token
// chain and flag operands which copy the outgoing args into registers.
if (InFlag.getNode())
Ops.push_back(InFlag);
- Chain = DAG.getNode(XCoreISD::BL, dl, NodeTys, &Ops[0], Ops.size());
+ Chain = DAG.getNode(XCoreISD::BL, dl, NodeTys, Ops);
InFlag = Chain.getValue(1);
// Create the CALLSEQ_END node.
- Chain = DAG.getCALLSEQ_END(Chain,
- DAG.getConstant(NumBytes, getPointerTy(), true),
- DAG.getConstant(0, getPointerTy(), true),
- InFlag);
+ Chain = DAG.getCALLSEQ_END(Chain, DAG.getConstant(NumBytes, dl, PtrVT, true),
+ DAG.getConstant(0, dl, PtrVT, true), InFlag, dl);
InFlag = Chain.getValue(1);
// Handle result values, copying them out of physregs into vregs that we
// return.
- return LowerCallResult(Chain, InFlag, CallConv, isVarArg,
- Ins, dl, DAG, InVals);
-}
-
-/// LowerCallResult - Lower the result values of a call into the
-/// appropriate copies out of appropriate physical registers.
-SDValue
-XCoreTargetLowering::LowerCallResult(SDValue Chain, SDValue InFlag,
- CallingConv::ID CallConv, bool isVarArg,
- const SmallVectorImpl<ISD::InputArg> &Ins,
- DebugLoc dl, SelectionDAG &DAG,
- SmallVectorImpl<SDValue> &InVals) const {
-
- // Assign locations to each value returned by this call.
- SmallVector<CCValAssign, 16> RVLocs;
- CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(),
- getTargetMachine(), RVLocs, *DAG.getContext());
-
- CCInfo.AnalyzeCallResult(Ins, RetCC_XCore);
-
- // Copy all of the result registers out of their specified physreg.
- for (unsigned i = 0; i != RVLocs.size(); ++i) {
- Chain = DAG.getCopyFromReg(Chain, dl, RVLocs[i].getLocReg(),
- RVLocs[i].getValVT(), InFlag).getValue(1);
- InFlag = Chain.getValue(2);
- InVals.push_back(Chain.getValue(0));
- }
-
- return Chain;
+ return LowerCallResult(Chain, InFlag, RVLocs, dl, DAG, InVals);
}
//===----------------------------------------------------------------------===//
// Formal Arguments Calling Convention Implementation
//===----------------------------------------------------------------------===//
+namespace {
+ struct ArgDataPair { SDValue SDV; ISD::ArgFlagsTy Flags; };
+}
+
/// XCore formal arguments implementation
SDValue
XCoreTargetLowering::LowerFormalArguments(SDValue Chain,
CallingConv::ID CallConv,
bool isVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins,
- DebugLoc dl,
+ SDLoc dl,
SelectionDAG &DAG,
SmallVectorImpl<SDValue> &InVals)
const {
bool isVarArg,
const SmallVectorImpl<ISD::InputArg>
&Ins,
- DebugLoc dl,
+ SDLoc dl,
SelectionDAG &DAG,
SmallVectorImpl<SDValue> &InVals) const {
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo *MFI = MF.getFrameInfo();
MachineRegisterInfo &RegInfo = MF.getRegInfo();
+ XCoreFunctionInfo *XFI = MF.getInfo<XCoreFunctionInfo>();
// Assign locations to all of the incoming arguments.
SmallVector<CCValAssign, 16> ArgLocs;
- CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(),
- getTargetMachine(), ArgLocs, *DAG.getContext());
+ CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
+ *DAG.getContext());
CCInfo.AnalyzeFormalArguments(Ins, CC_XCore);
unsigned LRSaveSize = StackSlotSize;
+ if (!isVarArg)
+ XFI->setReturnStackOffset(CCInfo.getNextStackOffset() + LRSaveSize);
+
+ // All getCopyFromReg ops must precede any getMemcpys to prevent the
+ // scheduler clobbering a register before it has been copied.
+ // The stages are:
+ // 1. CopyFromReg (and load) arg & vararg registers.
+ // 2. Chain CopyFromReg nodes into a TokenFactor.
+ // 3. Memcpy 'byVal' args & push final InVals.
+ // 4. Chain mem ops nodes into a TokenFactor.
+ SmallVector<SDValue, 4> CFRegNode;
+ SmallVector<ArgDataPair, 4> ArgData;
+ SmallVector<SDValue, 4> MemOps;
+
+ // 1a. CopyFromReg (and load) arg registers.
for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
CCValAssign &VA = ArgLocs[i];
+ SDValue ArgIn;
if (VA.isRegLoc()) {
// Arguments passed in registers
errs() << "LowerFormalArguments Unhandled argument type: "
<< RegVT.getSimpleVT().SimpleTy << "\n";
#endif
- llvm_unreachable(0);
+ llvm_unreachable(nullptr);
}
case MVT::i32:
unsigned VReg = RegInfo.createVirtualRegister(&XCore::GRRegsRegClass);
RegInfo.addLiveIn(VA.getLocReg(), VReg);
- InVals.push_back(DAG.getCopyFromReg(Chain, dl, VReg, RegVT));
+ ArgIn = DAG.getCopyFromReg(Chain, dl, VReg, RegVT);
+ CFRegNode.push_back(ArgIn.getValue(ArgIn->getNumValues() - 1));
}
} else {
// sanity check
// Create the SelectionDAG nodes corresponding to a load
//from this parameter
SDValue FIN = DAG.getFrameIndex(FI, MVT::i32);
- InVals.push_back(DAG.getLoad(VA.getLocVT(), dl, Chain, FIN,
- MachinePointerInfo::getFixedStack(FI),
- false, false, false, 0));
+ ArgIn = DAG.getLoad(VA.getLocVT(), dl, Chain, FIN,
+ MachinePointerInfo::getFixedStack(MF, FI), false,
+ false, false, 0);
}
+ const ArgDataPair ADP = { ArgIn, Ins[i].Flags };
+ ArgData.push_back(ADP);
}
+ // 1b. CopyFromReg vararg registers.
if (isVarArg) {
- /* Argument registers */
- static const uint16_t ArgRegs[] = {
+ // Argument registers
+ static const MCPhysReg ArgRegs[] = {
XCore::R0, XCore::R1, XCore::R2, XCore::R3
};
XCoreFunctionInfo *XFI = MF.getInfo<XCoreFunctionInfo>();
- unsigned FirstVAReg = CCInfo.getFirstUnallocated(ArgRegs,
- array_lengthof(ArgRegs));
+ unsigned FirstVAReg = CCInfo.getFirstUnallocated(ArgRegs);
if (FirstVAReg < array_lengthof(ArgRegs)) {
- SmallVector<SDValue, 4> MemOps;
int offset = 0;
// Save remaining registers, storing higher register numbers at a higher
// address
unsigned VReg = RegInfo.createVirtualRegister(&XCore::GRRegsRegClass);
RegInfo.addLiveIn(ArgRegs[i], VReg);
SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i32);
+ CFRegNode.push_back(Val.getValue(Val->getNumValues() - 1));
// Move argument from virt reg -> stack
SDValue Store = DAG.getStore(Val.getValue(1), dl, Val, FIN,
MachinePointerInfo(), false, false, 0);
MemOps.push_back(Store);
}
- if (!MemOps.empty())
- Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
- &MemOps[0], MemOps.size());
} else {
// This will point to the next argument passed via stack.
XFI->setVarArgsFrameIndex(
}
}
+ // 2. chain CopyFromReg nodes into a TokenFactor.
+ if (!CFRegNode.empty())
+ Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, CFRegNode);
+
+ // 3. Memcpy 'byVal' args & push final InVals.
+ // Aggregates passed "byVal" need to be copied by the callee.
+ // The callee will use a pointer to this copy, rather than the original
+ // pointer.
+ for (SmallVectorImpl<ArgDataPair>::const_iterator ArgDI = ArgData.begin(),
+ ArgDE = ArgData.end();
+ ArgDI != ArgDE; ++ArgDI) {
+ if (ArgDI->Flags.isByVal() && ArgDI->Flags.getByValSize()) {
+ unsigned Size = ArgDI->Flags.getByValSize();
+ unsigned Align = std::max(StackSlotSize, ArgDI->Flags.getByValAlign());
+ // Create a new object on the stack and copy the pointee into it.
+ int FI = MFI->CreateStackObject(Size, Align, false);
+ SDValue FIN = DAG.getFrameIndex(FI, MVT::i32);
+ InVals.push_back(FIN);
+ MemOps.push_back(DAG.getMemcpy(Chain, dl, FIN, ArgDI->SDV,
+ DAG.getConstant(Size, dl, MVT::i32),
+ Align, false, false, false,
+ MachinePointerInfo(),
+ MachinePointerInfo()));
+ } else {
+ InVals.push_back(ArgDI->SDV);
+ }
+ }
+
+ // 4, chain mem ops nodes into a TokenFactor.
+ if (!MemOps.empty()) {
+ MemOps.push_back(Chain);
+ Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOps);
+ }
+
return Chain;
}
const SmallVectorImpl<ISD::OutputArg> &Outs,
LLVMContext &Context) const {
SmallVector<CCValAssign, 16> RVLocs;
- CCState CCInfo(CallConv, isVarArg, MF, getTargetMachine(), RVLocs, Context);
- return CCInfo.CheckReturn(Outs, RetCC_XCore);
+ CCState CCInfo(CallConv, isVarArg, MF, RVLocs, Context);
+ if (!CCInfo.CheckReturn(Outs, RetCC_XCore))
+ return false;
+ if (CCInfo.getNextStackOffset() != 0 && isVarArg)
+ return false;
+ return true;
}
SDValue
CallingConv::ID CallConv, bool isVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
- DebugLoc dl, SelectionDAG &DAG) const {
+ SDLoc dl, SelectionDAG &DAG) const {
+
+ XCoreFunctionInfo *XFI =
+ DAG.getMachineFunction().getInfo<XCoreFunctionInfo>();
+ MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
// CCValAssign - represent the assignment of
// the return value to a location
SmallVector<CCValAssign, 16> RVLocs;
// CCState - Info about the registers and stack slot.
- CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(),
- getTargetMachine(), RVLocs, *DAG.getContext());
+ CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs,
+ *DAG.getContext());
// Analyze return values.
+ if (!isVarArg)
+ CCInfo.AllocateStack(XFI->getReturnStackOffset(), 4);
+
CCInfo.AnalyzeReturn(Outs, RetCC_XCore);
SDValue Flag;
SmallVector<SDValue, 4> RetOps(1, Chain);
// Return on XCore is always a "retsp 0"
- RetOps.push_back(DAG.getConstant(0, MVT::i32));
+ RetOps.push_back(DAG.getConstant(0, dl, MVT::i32));
- // Copy the result values into the output registers.
- for (unsigned i = 0; i != RVLocs.size(); ++i) {
+ SmallVector<SDValue, 4> MemOpChains;
+ // Handle return values that must be copied to memory.
+ for (unsigned i = 0, e = RVLocs.size(); i != e; ++i) {
CCValAssign &VA = RVLocs[i];
- assert(VA.isRegLoc() && "Can only return in registers!");
+ if (VA.isRegLoc())
+ continue;
+ assert(VA.isMemLoc());
+ if (isVarArg) {
+ report_fatal_error("Can't return value from vararg function in memory");
+ }
+
+ int Offset = VA.getLocMemOffset();
+ unsigned ObjSize = VA.getLocVT().getSizeInBits() / 8;
+ // Create the frame index object for the memory location.
+ int FI = MFI->CreateFixedObject(ObjSize, Offset, false);
+
+ // Create a SelectionDAG node corresponding to a store
+ // to this memory location.
+ SDValue FIN = DAG.getFrameIndex(FI, MVT::i32);
+ MemOpChains.push_back(DAG.getStore(
+ Chain, dl, OutVals[i], FIN,
+ MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI), false,
+ false, 0));
+ }
+
+ // Transform all store nodes into one single node because
+ // all stores are independent of each other.
+ if (!MemOpChains.empty())
+ Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains);
- Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
- OutVals[i], Flag);
+ // Now handle return values copied to registers.
+ for (unsigned i = 0, e = RVLocs.size(); i != e; ++i) {
+ CCValAssign &VA = RVLocs[i];
+ if (!VA.isRegLoc())
+ continue;
+ // Copy the result values into the output registers.
+ Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), OutVals[i], Flag);
// guarantee that all emitted copies are
// stuck together, avoiding something bad
if (Flag.getNode())
RetOps.push_back(Flag);
- return DAG.getNode(XCoreISD::RETSP, dl, MVT::Other,
- &RetOps[0], RetOps.size());
+ return DAG.getNode(XCoreISD::RETSP, dl, MVT::Other, RetOps);
}
//===----------------------------------------------------------------------===//
MachineBasicBlock *
XCoreTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
MachineBasicBlock *BB) const {
- const TargetInstrInfo &TII = *getTargetMachine().getInstrInfo();
+ const TargetInstrInfo &TII = *Subtarget.getInstrInfo();
DebugLoc dl = MI->getDebugLoc();
assert((MI->getOpcode() == XCore::SELECT_CC) &&
"Unexpected instr type to insert");
// Transfer the remainder of BB and its successor edges to sinkMBB.
sinkMBB->splice(sinkMBB->begin(), BB,
- llvm::next(MachineBasicBlock::iterator(MI)),
- BB->end());
+ std::next(MachineBasicBlock::iterator(MI)), BB->end());
sinkMBB->transferSuccessorsAndUpdatePHIs(BB);
// Next, add the true and fallthrough blocks as its successors.
SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N,
DAGCombinerInfo &DCI) const {
SelectionDAG &DAG = DCI.DAG;
- DebugLoc dl = N->getDebugLoc();
+ SDLoc dl(N);
switch (N->getOpcode()) {
default: break;
+ case ISD::INTRINSIC_VOID:
+ switch (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue()) {
+ case Intrinsic::xcore_outt:
+ case Intrinsic::xcore_outct:
+ case Intrinsic::xcore_chkct: {
+ SDValue OutVal = N->getOperand(3);
+ // These instructions ignore the high bits.
+ if (OutVal.hasOneUse()) {
+ unsigned BitWidth = OutVal.getValueSizeInBits();
+ APInt DemandedMask = APInt::getLowBitsSet(BitWidth, 8);
+ APInt KnownZero, KnownOne;
+ TargetLowering::TargetLoweringOpt TLO(DAG, !DCI.isBeforeLegalize(),
+ !DCI.isBeforeLegalizeOps());
+ const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+ if (TLO.ShrinkDemandedConstant(OutVal, DemandedMask) ||
+ TLI.SimplifyDemandedBits(OutVal, DemandedMask, KnownZero, KnownOne,
+ TLO))
+ DCI.CommitTargetLoweringOpt(TLO);
+ }
+ break;
+ }
+ case Intrinsic::xcore_setpt: {
+ SDValue Time = N->getOperand(3);
+ // This instruction ignores the high bits.
+ if (Time.hasOneUse()) {
+ unsigned BitWidth = Time.getValueSizeInBits();
+ APInt DemandedMask = APInt::getLowBitsSet(BitWidth, 16);
+ APInt KnownZero, KnownOne;
+ TargetLowering::TargetLoweringOpt TLO(DAG, !DCI.isBeforeLegalize(),
+ !DCI.isBeforeLegalizeOps());
+ const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+ if (TLO.ShrinkDemandedConstant(Time, DemandedMask) ||
+ TLI.SimplifyDemandedBits(Time, DemandedMask, KnownZero, KnownOne,
+ TLO))
+ DCI.CommitTargetLoweringOpt(TLO);
+ }
+ break;
+ }
+ }
+ break;
case XCoreISD::LADD: {
SDValue N0 = N->getOperand(0);
SDValue N1 = N->getOperand(1);
// fold (ladd 0, 0, x) -> 0, x & 1
if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) {
- SDValue Carry = DAG.getConstant(0, VT);
+ SDValue Carry = DAG.getConstant(0, dl, VT);
SDValue Result = DAG.getNode(ISD::AND, dl, VT, N2,
- DAG.getConstant(1, VT));
+ DAG.getConstant(1, dl, VT));
SDValue Ops[] = { Result, Carry };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
// fold (ladd x, 0, y) -> 0, add x, y iff carry is unused and y has only the
APInt KnownZero, KnownOne;
APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
VT.getSizeInBits() - 1);
- DAG.ComputeMaskedBits(N2, KnownZero, KnownOne);
+ DAG.computeKnownBits(N2, KnownZero, KnownOne);
if ((KnownZero & Mask) == Mask) {
- SDValue Carry = DAG.getConstant(0, VT);
+ SDValue Carry = DAG.getConstant(0, dl, VT);
SDValue Result = DAG.getNode(ISD::ADD, dl, VT, N0, N2);
SDValue Ops[] = { Result, Carry };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
}
}
APInt KnownZero, KnownOne;
APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
VT.getSizeInBits() - 1);
- DAG.ComputeMaskedBits(N2, KnownZero, KnownOne);
+ DAG.computeKnownBits(N2, KnownZero, KnownOne);
if ((KnownZero & Mask) == Mask) {
SDValue Borrow = N2;
SDValue Result = DAG.getNode(ISD::SUB, dl, VT,
- DAG.getConstant(0, VT), N2);
+ DAG.getConstant(0, dl, VT), N2);
SDValue Ops[] = { Result, Borrow };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
}
APInt KnownZero, KnownOne;
APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
VT.getSizeInBits() - 1);
- DAG.ComputeMaskedBits(N2, KnownZero, KnownOne);
+ DAG.computeKnownBits(N2, KnownZero, KnownOne);
if ((KnownZero & Mask) == Mask) {
- SDValue Borrow = DAG.getConstant(0, VT);
+ SDValue Borrow = DAG.getConstant(0, dl, VT);
SDValue Result = DAG.getNode(ISD::SUB, dl, VT, N0, N2);
SDValue Ops[] = { Result, Borrow };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
}
}
if (N->hasNUsesOfValue(0, 0)) {
SDValue Lo = DAG.getNode(ISD::ADD, dl, VT, N2, N3);
SDValue Ops[] = { Lo, Lo };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
// Otherwise fold to ladd(a, b, 0)
SDValue Result =
DAG.getNode(XCoreISD::LADD, dl, DAG.getVTList(VT, VT), N2, N3, N1);
SDValue Carry(Result.getNode(), 1);
SDValue Ops[] = { Carry, Result };
- return DAG.getMergeValues(Ops, 2, dl);
+ return DAG.getMergeValues(Ops, dl);
}
}
break;
DAG.MaskedValueIsZero(Addend0, HighMask) &&
DAG.MaskedValueIsZero(Addend1, HighMask)) {
SDValue Mul0L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Mul0, DAG.getConstant(0, MVT::i32));
+ Mul0, DAG.getConstant(0, dl, MVT::i32));
SDValue Mul1L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Mul1, DAG.getConstant(0, MVT::i32));
+ Mul1, DAG.getConstant(0, dl, MVT::i32));
SDValue Addend0L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Addend0, DAG.getConstant(0, MVT::i32));
+ Addend0, DAG.getConstant(0, dl, MVT::i32));
SDValue Addend1L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
- Addend1, DAG.getConstant(0, MVT::i32));
+ Addend1, DAG.getConstant(0, dl, MVT::i32));
SDValue Hi = DAG.getNode(XCoreISD::LMUL, dl,
DAG.getVTList(MVT::i32, MVT::i32), Mul0L, Mul1L,
Addend0L, Addend1L);
// Replace unaligned store of unaligned load with memmove.
StoreSDNode *ST = cast<StoreSDNode>(N);
if (!DCI.isBeforeLegalize() ||
- allowsUnalignedMemoryAccesses(ST->getMemoryVT()) ||
+ allowsMisalignedMemoryAccesses(ST->getMemoryVT(),
+ ST->getAddressSpace(),
+ ST->getAlignment()) ||
ST->isVolatile() || ST->isIndexed()) {
break;
}
SDValue Chain = ST->getChain();
unsigned StoreBits = ST->getMemoryVT().getStoreSizeInBits();
- if (StoreBits % 8) {
- break;
- }
- unsigned ABIAlignment = getDataLayout()->getABITypeAlignment(
+ assert((StoreBits % 8) == 0 &&
+ "Store size in bits must be a multiple of 8");
+ unsigned ABIAlignment = DAG.getDataLayout().getABITypeAlignment(
ST->getMemoryVT().getTypeForEVT(*DCI.DAG.getContext()));
unsigned Alignment = ST->getAlignment();
if (Alignment >= ABIAlignment) {
LD->getAlignment() == Alignment &&
!LD->isVolatile() && !LD->isIndexed() &&
Chain.reachesChainWithoutSideEffects(SDValue(LD, 1))) {
+ bool isTail = isInTailCallPosition(DAG, ST, Chain);
return DAG.getMemmove(Chain, dl, ST->getBasePtr(),
LD->getBasePtr(),
- DAG.getConstant(StoreBits/8, MVT::i32),
- Alignment, false, ST->getPointerInfo(),
+ DAG.getConstant(StoreBits/8, dl, MVT::i32),
+ Alignment, false, isTail, ST->getPointerInfo(),
LD->getPointerInfo());
}
}
return SDValue();
}
-void XCoreTargetLowering::computeMaskedBitsForTargetNode(const SDValue Op,
- APInt &KnownZero,
- APInt &KnownOne,
- const SelectionDAG &DAG,
- unsigned Depth) const {
+void XCoreTargetLowering::computeKnownBitsForTargetNode(const SDValue Op,
+ APInt &KnownZero,
+ APInt &KnownOne,
+ const SelectionDAG &DAG,
+ unsigned Depth) const {
KnownZero = KnownOne = APInt(KnownZero.getBitWidth(), 0);
switch (Op.getOpcode()) {
default: break;
KnownZero.getBitWidth() - 1);
}
break;
+ case ISD::INTRINSIC_W_CHAIN:
+ {
+ unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
+ switch (IntNo) {
+ case Intrinsic::xcore_getts:
+ // High bits are known to be zero.
+ KnownZero = APInt::getHighBitsSet(KnownZero.getBitWidth(),
+ KnownZero.getBitWidth() - 16);
+ break;
+ case Intrinsic::xcore_int:
+ case Intrinsic::xcore_inct:
+ // High bits are known to be zero.
+ KnownZero = APInt::getHighBitsSet(KnownZero.getBitWidth(),
+ KnownZero.getBitWidth() - 8);
+ break;
+ case Intrinsic::xcore_testct:
+ // Result is either 0 or 1.
+ KnownZero = APInt::getHighBitsSet(KnownZero.getBitWidth(),
+ KnownZero.getBitWidth() - 1);
+ break;
+ case Intrinsic::xcore_testwct:
+ // Result is in the range 0 - 4.
+ KnownZero = APInt::getHighBitsSet(KnownZero.getBitWidth(),
+ KnownZero.getBitWidth() - 3);
+ break;
+ }
+ }
+ break;
}
}
/// isLegalAddressingMode - Return true if the addressing mode represented
/// by AM is legal for this target, for a load/store of the specified type.
-bool
-XCoreTargetLowering::isLegalAddressingMode(const AddrMode &AM,
- Type *Ty) const {
+bool XCoreTargetLowering::isLegalAddressingMode(const DataLayout &DL,
+ const AddrMode &AM, Type *Ty,
+ unsigned AS) const {
if (Ty->getTypeID() == Type::VoidTyID)
return AM.Scale == 0 && isImmUs(AM.BaseOffs) && isImmUs4(AM.BaseOffs);
- const DataLayout *TD = TM.getDataLayout();
- unsigned Size = TD->getTypeAllocSize(Ty);
+ unsigned Size = DL.getTypeAllocSize(Ty);
if (AM.BaseGV) {
return Size >= 4 && !AM.HasBaseReg && AM.Scale == 0 &&
AM.BaseOffs%4 == 0;
// XCore Inline Assembly Support
//===----------------------------------------------------------------------===//
-std::pair<unsigned, const TargetRegisterClass*>
-XCoreTargetLowering::
-getRegForInlineAsmConstraint(const std::string &Constraint,
- EVT VT) const {
+std::pair<unsigned, const TargetRegisterClass *>
+XCoreTargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
+ StringRef Constraint,
+ MVT VT) const {
if (Constraint.size() == 1) {
switch (Constraint[0]) {
default : break;
}
// Use the default implementation in TargetLowering to convert the register
// constraint into a member of a register class.
- return TargetLowering::getRegForInlineAsmConstraint(Constraint, VT);
+ return TargetLowering::getRegForInlineAsmConstraint(TRI, Constraint, VT);
}