//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/SelectionDAG.h"
-#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/Triple.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineJumpTableInfo.h"
SDNode *Node, bool isSigned);
SDValue ExpandFPLibCall(SDNode *Node, RTLIB::Libcall Call_F32,
RTLIB::Libcall Call_F64, RTLIB::Libcall Call_F80,
- RTLIB::Libcall Call_F128, RTLIB::Libcall Call_PPCF128);
+ RTLIB::Libcall Call_F128,
+ RTLIB::Libcall Call_PPCF128);
SDValue ExpandIntLibCall(SDNode *Node, bool isSigned,
RTLIB::Libcall Call_I8,
RTLIB::Libcall Call_I16,
RTLIB::Libcall Call_I64,
RTLIB::Libcall Call_I128);
void ExpandDivRemLibCall(SDNode *Node, SmallVectorImpl<SDValue> &Results);
+ void ExpandSinCosLibCall(SDNode *Node, SmallVectorImpl<SDValue> &Results);
SDValue EmitStackConvert(SDValue SrcOp, EVT SlotVT, EVT DestVT, DebugLoc dl);
SDValue ExpandBUILD_VECTOR(SDNode *Node);
Results.push_back(Rem);
}
+/// isSinCosLibcallAvailable - Return true if sincos libcall is available.
+static bool isSinCosLibcallAvailable(SDNode *Node, const TargetLowering &TLI) {
+ RTLIB::Libcall LC;
+ switch (Node->getValueType(0).getSimpleVT().SimpleTy) {
+ default: llvm_unreachable("Unexpected request for libcall!");
+ case MVT::f32: LC = RTLIB::SINCOS_F32; break;
+ case MVT::f64: LC = RTLIB::SINCOS_F64; break;
+ case MVT::f80: LC = RTLIB::SINCOS_F80; break;
+ case MVT::f128: LC = RTLIB::SINCOS_F128; break;
+ case MVT::ppcf128: LC = RTLIB::SINCOS_PPCF128; break;
+ }
+ return TLI.getLibcallName(LC) != 0;
+}
+
+/// canCombineSinCosLibcall - Return true if sincos libcall is available and
+/// can be used to combine sin and cos.
+static bool canCombineSinCosLibcall(SDNode *Node, const TargetLowering &TLI,
+ const TargetMachine &TM) {
+ if (!isSinCosLibcallAvailable(Node, TLI))
+ return false;
+ // GNU sin/cos functions set errno while sincos does not. Therefore
+ // combining sin and cos is only safe if unsafe-fpmath is enabled.
+ bool isGNU = Triple(TM.getTargetTriple()).getEnvironment() == Triple::GNU;
+ if (isGNU && !TM.Options.UnsafeFPMath)
+ return false;
+ return true;
+}
+
+/// useSinCos - Only issue sincos libcall if both sin and cos are
+/// needed.
+static bool useSinCos(SDNode *Node) {
+ unsigned OtherOpcode = Node->getOpcode() == ISD::FSIN
+ ? ISD::FCOS : ISD::FSIN;
+
+ SDValue Op0 = Node->getOperand(0);
+ for (SDNode::use_iterator UI = Op0.getNode()->use_begin(),
+ UE = Op0.getNode()->use_end(); UI != UE; ++UI) {
+ SDNode *User = *UI;
+ if (User == Node)
+ continue;
+ // The other user might have been turned into sincos already.
+ if (User->getOpcode() == OtherOpcode || User->getOpcode() == ISD::FSINCOS)
+ return true;
+ }
+ return false;
+}
+
+/// ExpandSinCosLibCall - Issue libcalls to sincos to compute sin / cos
+/// pairs.
+void
+SelectionDAGLegalize::ExpandSinCosLibCall(SDNode *Node,
+ SmallVectorImpl<SDValue> &Results) {
+ RTLIB::Libcall LC;
+ switch (Node->getValueType(0).getSimpleVT().SimpleTy) {
+ default: llvm_unreachable("Unexpected request for libcall!");
+ case MVT::f32: LC = RTLIB::SINCOS_F32; break;
+ case MVT::f64: LC = RTLIB::SINCOS_F64; break;
+ case MVT::f80: LC = RTLIB::SINCOS_F80; break;
+ case MVT::f128: LC = RTLIB::SINCOS_F128; break;
+ case MVT::ppcf128: LC = RTLIB::SINCOS_PPCF128; break;
+ }
+
+ // The input chain to this libcall is the entry node of the function.
+ // Legalizing the call will automatically add the previous call to the
+ // dependence.
+ SDValue InChain = DAG.getEntryNode();
+
+ EVT RetVT = Node->getValueType(0);
+ Type *RetTy = RetVT.getTypeForEVT(*DAG.getContext());
+
+ TargetLowering::ArgListTy Args;
+ TargetLowering::ArgListEntry Entry;
+
+ // Pass the argument.
+ Entry.Node = Node->getOperand(0);
+ Entry.Ty = RetTy;
+ Entry.isSExt = false;
+ Entry.isZExt = false;
+ Args.push_back(Entry);
+
+ // Pass the return address of sin.
+ SDValue SinPtr = DAG.CreateStackTemporary(RetVT);
+ Entry.Node = SinPtr;
+ Entry.Ty = RetTy->getPointerTo();
+ Entry.isSExt = false;
+ Entry.isZExt = false;
+ Args.push_back(Entry);
+
+ // Also pass the return address of the cos.
+ SDValue CosPtr = DAG.CreateStackTemporary(RetVT);
+ Entry.Node = CosPtr;
+ Entry.Ty = RetTy->getPointerTo();
+ Entry.isSExt = false;
+ Entry.isZExt = false;
+ Args.push_back(Entry);
+
+ SDValue Callee = DAG.getExternalSymbol(TLI.getLibcallName(LC),
+ TLI.getPointerTy());
+
+ DebugLoc dl = Node->getDebugLoc();
+ TargetLowering::
+ CallLoweringInfo CLI(InChain, Type::getVoidTy(*DAG.getContext()),
+ false, false, false, false,
+ 0, TLI.getLibcallCallingConv(LC), /*isTailCall=*/false,
+ /*doesNotReturn=*/false, /*isReturnValueUsed=*/true,
+ Callee, Args, DAG, dl);
+ std::pair<SDValue, SDValue> CallInfo = TLI.LowerCallTo(CLI);
+
+ Results.push_back(DAG.getLoad(RetVT, dl, CallInfo.second, SinPtr,
+ MachinePointerInfo(), false, false, false, 0));
+ Results.push_back(DAG.getLoad(RetVT, dl, CallInfo.second, CosPtr,
+ MachinePointerInfo(), false, false, false, 0));
+}
+
/// ExpandLegalINT_TO_FP - This function is responsible for legalizing a
/// INT_TO_FP operation of the specified operand when the target requests that
/// we expand it. At this point, we know that the result and operand types are
}
}
-/// SplatByte - Distribute ByteVal over NumBits bits.
-// FIXME: Move this helper to a common place.
-static APInt SplatByte(unsigned NumBits, uint8_t ByteVal) {
- APInt Val = APInt(NumBits, ByteVal);
- unsigned Shift = 8;
- for (unsigned i = NumBits; i > 8; i >>= 1) {
- Val = (Val << Shift) | Val;
- Shift <<= 1;
- }
- return Val;
-}
-
/// ExpandBitCount - Expand the specified bitcount instruction into operations.
///
SDValue SelectionDAGLegalize::ExpandBitCount(unsigned Opc, SDValue Op,
// This is the "best" algorithm from
// http://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
- SDValue Mask55 = DAG.getConstant(SplatByte(Len, 0x55), VT);
- SDValue Mask33 = DAG.getConstant(SplatByte(Len, 0x33), VT);
- SDValue Mask0F = DAG.getConstant(SplatByte(Len, 0x0F), VT);
- SDValue Mask01 = DAG.getConstant(SplatByte(Len, 0x01), VT);
+ SDValue Mask55 = DAG.getConstant(APInt::getSplat(Len, APInt(8, 0x55)), VT);
+ SDValue Mask33 = DAG.getConstant(APInt::getSplat(Len, APInt(8, 0x33)), VT);
+ SDValue Mask0F = DAG.getConstant(APInt::getSplat(Len, APInt(8, 0x0F)), VT);
+ SDValue Mask01 = DAG.getConstant(APInt::getSplat(Len, APInt(8, 0x01)), VT);
// v = v - ((v >> 1) & 0x55555555...)
Op = DAG.getNode(ISD::SUB, dl, VT, Op,
Results.push_back(DAG.getConstant(0, MVT::i32));
Results.push_back(Node->getOperand(0));
break;
- case ISD::ATOMIC_FENCE:
- case ISD::MEMBARRIER: {
+ case ISD::ATOMIC_FENCE: {
// If the target didn't lower this, lower it to '__sync_synchronize()' call
// FIXME: handle "fence singlethread" more efficiently.
TargetLowering::ArgListTy Args;
SDValue True, False;
EVT VT = Node->getOperand(0).getValueType();
EVT NVT = Node->getValueType(0);
- APFloat apf(APInt::getNullValue(VT.getSizeInBits()));
+ APFloat apf(DAG.EVTToAPFloatSemantics(VT),
+ APInt::getNullValue(VT.getSizeInBits()));
APInt x = APInt::getSignBit(NVT.getSizeInBits());
(void)apf.convertFromAPInt(x, false, APFloat::rmNearestTiesToEven);
Tmp1 = DAG.getConstantFP(apf, VT);
RTLIB::SQRT_PPCF128));
break;
case ISD::FSIN:
- Results.push_back(ExpandFPLibCall(Node, RTLIB::SIN_F32, RTLIB::SIN_F64,
- RTLIB::SIN_F80, RTLIB::SIN_F128,
- RTLIB::SIN_PPCF128));
+ case ISD::FCOS: {
+ EVT VT = Node->getValueType(0);
+ bool isSIN = Node->getOpcode() == ISD::FSIN;
+ // Turn fsin / fcos into ISD::FSINCOS node if there are a pair of fsin /
+ // fcos which share the same operand and both are used.
+ if ((TLI.isOperationLegalOrCustom(ISD::FSINCOS, VT) ||
+ canCombineSinCosLibcall(Node, TLI, TM))
+ && useSinCos(Node)) {
+ SDVTList VTs = DAG.getVTList(VT, VT);
+ Tmp1 = DAG.getNode(ISD::FSINCOS, dl, VTs, Node->getOperand(0));
+ if (!isSIN)
+ Tmp1 = Tmp1.getValue(1);
+ Results.push_back(Tmp1);
+ } else if (isSIN) {
+ Results.push_back(ExpandFPLibCall(Node, RTLIB::SIN_F32, RTLIB::SIN_F64,
+ RTLIB::SIN_F80, RTLIB::SIN_F128,
+ RTLIB::SIN_PPCF128));
+ } else {
+ Results.push_back(ExpandFPLibCall(Node, RTLIB::COS_F32, RTLIB::COS_F64,
+ RTLIB::COS_F80, RTLIB::COS_F128,
+ RTLIB::COS_PPCF128));
+ }
break;
- case ISD::FCOS:
- Results.push_back(ExpandFPLibCall(Node, RTLIB::COS_F32, RTLIB::COS_F64,
- RTLIB::COS_F80, RTLIB::COS_F128,
- RTLIB::COS_PPCF128));
+ }
+ case ISD::FSINCOS:
+ // Expand into sincos libcall.
+ ExpandSinCosLibCall(Node, Results);
break;
case ISD::FLOG:
Results.push_back(ExpandFPLibCall(Node, RTLIB::LOG_F32, RTLIB::LOG_F64,
case ISD::UREM:
case ISD::SREM: {
EVT VT = Node->getValueType(0);
- SDVTList VTs = DAG.getVTList(VT, VT);
bool isSigned = Node->getOpcode() == ISD::SREM;
unsigned DivOpc = isSigned ? ISD::SDIV : ISD::UDIV;
unsigned DivRemOpc = isSigned ? ISD::SDIVREM : ISD::UDIVREM;
// If div is legal, it's better to do the normal expansion
!TLI.isOperationLegalOrCustom(DivOpc, Node->getValueType(0)) &&
useDivRem(Node, isSigned, false))) {
+ SDVTList VTs = DAG.getVTList(VT, VT);
Tmp1 = DAG.getNode(DivRemOpc, dl, VTs, Tmp2, Tmp3).getValue(1);
} else if (TLI.isOperationLegalOrCustom(DivOpc, VT)) {
// X % Y -> X-X/Y*Y
// Otherwise, SETCC for the given comparison type must be completely
// illegal; expand it into a SELECT_CC.
EVT VT = Node->getValueType(0);
+ int TrueValue;
+ switch (TLI.getBooleanContents(VT.isVector())) {
+ case TargetLowering::ZeroOrOneBooleanContent:
+ case TargetLowering::UndefinedBooleanContent:
+ TrueValue = 1;
+ break;
+ case TargetLowering::ZeroOrNegativeOneBooleanContent:
+ TrueValue = -1;
+ break;
+ }
Tmp1 = DAG.getNode(ISD::SELECT_CC, dl, VT, Tmp1, Tmp2,
- DAG.getConstant(1, VT), DAG.getConstant(0, VT), Tmp3);
+ DAG.getConstant(TrueValue, VT), DAG.getConstant(0, VT),
+ Tmp3);
Results.push_back(Tmp1);
break;
}
projects / oota-llvm.git / blobdiff