if (Subtarget->is64Bit()) {
setOperationAction(ISD::UINT_TO_FP , MVT::i32 , Promote);
setOperationAction(ISD::UINT_TO_FP , MVT::i64 , Custom);
- } else if (!TM.Options.UseSoftFloat) {
+ } else if (!Subtarget->useSoftFloat()) {
// We have an algorithm for SSE2->double, and we turn this into a
// 64-bit FILD followed by conditional FADD for other targets.
setOperationAction(ISD::UINT_TO_FP , MVT::i64 , Custom);
setOperationAction(ISD::SINT_TO_FP , MVT::i1 , Promote);
setOperationAction(ISD::SINT_TO_FP , MVT::i8 , Promote);
- if (!TM.Options.UseSoftFloat) {
+ if (!Subtarget->useSoftFloat()) {
// SSE has no i16 to fp conversion, only i32
if (X86ScalarSSEf32) {
setOperationAction(ISD::SINT_TO_FP , MVT::i16 , Promote);
if (Subtarget->is64Bit()) {
setOperationAction(ISD::FP_TO_UINT , MVT::i64 , Expand);
setOperationAction(ISD::FP_TO_UINT , MVT::i32 , Promote);
- } else if (!TM.Options.UseSoftFloat) {
+ } else if (!Subtarget->useSoftFloat()) {
// Since AVX is a superset of SSE3, only check for SSE here.
if (Subtarget->hasSSE1() && !Subtarget->hasSSE3())
// Expand FP_TO_UINT into a select.
// Special handling for half-precision floating point conversions.
// If we don't have F16C support, then lower half float conversions
// into library calls.
- if (TM.Options.UseSoftFloat || !Subtarget->hasF16C()) {
+ if (Subtarget->useSoftFloat() || !Subtarget->hasF16C()) {
setOperationAction(ISD::FP16_TO_FP, MVT::f32, Expand);
setOperationAction(ISD::FP_TO_FP16, MVT::f32, Expand);
}
setOperationAction(ISD::DYNAMIC_STACKALLOC, getPointerTy(), Custom);
- if (!TM.Options.UseSoftFloat && X86ScalarSSEf64) {
+ // GC_TRANSITION_START and GC_TRANSITION_END need custom lowering.
+ setOperationAction(ISD::GC_TRANSITION_START, MVT::Other, Custom);
+ setOperationAction(ISD::GC_TRANSITION_END, MVT::Other, Custom);
+
+ if (!Subtarget->useSoftFloat() && X86ScalarSSEf64) {
// f32 and f64 use SSE.
// Set up the FP register classes.
addRegisterClass(MVT::f32, &X86::FR32RegClass);
// cases we handle.
addLegalFPImmediate(APFloat(+0.0)); // xorpd
addLegalFPImmediate(APFloat(+0.0f)); // xorps
- } else if (!TM.Options.UseSoftFloat && X86ScalarSSEf32) {
+ } else if (!Subtarget->useSoftFloat() && X86ScalarSSEf32) {
// Use SSE for f32, x87 for f64.
// Set up the FP register classes.
addRegisterClass(MVT::f32, &X86::FR32RegClass);
setOperationAction(ISD::FCOS , MVT::f64, Expand);
setOperationAction(ISD::FSINCOS, MVT::f64, Expand);
}
- } else if (!TM.Options.UseSoftFloat) {
+ } else if (!Subtarget->useSoftFloat()) {
// f32 and f64 in x87.
// Set up the FP register classes.
addRegisterClass(MVT::f64, &X86::RFP64RegClass);
setOperationAction(ISD::FMA, MVT::f32, Expand);
// Long double always uses X87.
- if (!TM.Options.UseSoftFloat) {
+ if (!Subtarget->useSoftFloat()) {
addRegisterClass(MVT::f80, &X86::RFP80RegClass);
setOperationAction(ISD::UNDEF, MVT::f80, Expand);
setOperationAction(ISD::FCOPYSIGN, MVT::f80, Expand);
// them legal.
if (VT.getVectorElementType() == MVT::i1)
setLoadExtAction(ISD::EXTLOAD, InnerVT, VT, Expand);
+
+ // EXTLOAD for MVT::f16 vectors is not legal because f16 vectors are
+ // split/scalarized right now.
+ if (VT.getVectorElementType() == MVT::f16)
+ setLoadExtAction(ISD::EXTLOAD, InnerVT, VT, Expand);
}
}
// FIXME: In order to prevent SSE instructions being expanded to MMX ones
// with -msoft-float, disable use of MMX as well.
- if (!TM.Options.UseSoftFloat && Subtarget->hasMMX()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasMMX()) {
addRegisterClass(MVT::x86mmx, &X86::VR64RegClass);
// No operations on x86mmx supported, everything uses intrinsics.
}
}
setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v1i64, Expand);
- if (!TM.Options.UseSoftFloat && Subtarget->hasSSE1()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasSSE1()) {
addRegisterClass(MVT::v4f32, &X86::VR128RegClass);
setOperationAction(ISD::FADD, MVT::v4f32, Legal);
setOperationAction(ISD::UINT_TO_FP, MVT::v4i32, Custom);
}
- if (!TM.Options.UseSoftFloat && Subtarget->hasSSE2()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasSSE2()) {
addRegisterClass(MVT::v2f64, &X86::VR128RegClass);
// FIXME: Unfortunately, -soft-float and -no-implicit-float mean XMM
setOperationAction(ISD::BITCAST, MVT::v8i8, Custom);
}
- if (!TM.Options.UseSoftFloat && Subtarget->hasSSE41()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasSSE41()) {
for (MVT RoundedTy : {MVT::f32, MVT::f64, MVT::v4f32, MVT::v2f64}) {
setOperationAction(ISD::FFLOOR, RoundedTy, Legal);
setOperationAction(ISD::FCEIL, RoundedTy, Legal);
setOperationAction(ISD::SRA, MVT::v4i32, Custom);
}
- if (!TM.Options.UseSoftFloat && Subtarget->hasFp256()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasFp256()) {
addRegisterClass(MVT::v32i8, &X86::VR256RegClass);
addRegisterClass(MVT::v16i16, &X86::VR256RegClass);
addRegisterClass(MVT::v8i32, &X86::VR256RegClass);
}
}
- if (!TM.Options.UseSoftFloat && Subtarget->hasAVX512()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasAVX512()) {
addRegisterClass(MVT::v16i32, &X86::VR512RegClass);
addRegisterClass(MVT::v16f32, &X86::VR512RegClass);
addRegisterClass(MVT::v8i64, &X86::VR512RegClass);
setOperationAction(ISD::UINT_TO_FP, MVT::v16i32, Legal);
setOperationAction(ISD::UINT_TO_FP, MVT::v8i32, Legal);
setOperationAction(ISD::UINT_TO_FP, MVT::v4i32, Legal);
+ setOperationAction(ISD::UINT_TO_FP, MVT::v16i8, Custom);
+ setOperationAction(ISD::UINT_TO_FP, MVT::v16i16, Custom);
setOperationAction(ISD::FP_ROUND, MVT::v8f32, Legal);
setOperationAction(ISD::FP_EXTEND, MVT::v8f32, Legal);
setOperationAction(ISD::SELECT, MVT::v8f64, Custom);
setOperationAction(ISD::SELECT, MVT::v8i64, Custom);
setOperationAction(ISD::SELECT, MVT::v16f32, Custom);
+ setOperationAction(ISD::SELECT, MVT::v16i1, Custom);
+ setOperationAction(ISD::SELECT, MVT::v8i1, Custom);
setOperationAction(ISD::ADD, MVT::v8i64, Legal);
setOperationAction(ISD::ADD, MVT::v16i32, Legal);
// Custom lower several nodes.
for (MVT VT : MVT::vector_valuetypes()) {
unsigned EltSize = VT.getVectorElementType().getSizeInBits();
+ if (EltSize == 1) {
+ setOperationAction(ISD::AND, VT, Legal);
+ setOperationAction(ISD::OR, VT, Legal);
+ setOperationAction(ISD::XOR, VT, Legal);
+ }
if (EltSize >= 32 && VT.getSizeInBits() <= 512) {
setOperationAction(ISD::MGATHER, VT, Custom);
setOperationAction(ISD::MSCATTER, VT, Custom);
}
}// has AVX-512
- if (!TM.Options.UseSoftFloat && Subtarget->hasBWI()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasBWI()) {
addRegisterClass(MVT::v32i16, &X86::VR512RegClass);
addRegisterClass(MVT::v64i8, &X86::VR512RegClass);
setOperationAction(ISD::CONCAT_VECTORS, MVT::v64i1, Custom);
setOperationAction(ISD::INSERT_SUBVECTOR, MVT::v32i1, Custom);
setOperationAction(ISD::INSERT_SUBVECTOR, MVT::v64i1, Custom);
+ setOperationAction(ISD::SELECT, MVT::v32i1, Custom);
+ setOperationAction(ISD::SELECT, MVT::v64i1, Custom);
for (int i = MVT::v32i8; i != MVT::v8i64; ++i) {
const MVT VT = (MVT::SimpleValueType)i;
}
}
- if (!TM.Options.UseSoftFloat && Subtarget->hasVLX()) {
+ if (!Subtarget->useSoftFloat() && Subtarget->hasVLX()) {
addRegisterClass(MVT::v4i1, &X86::VK4RegClass);
addRegisterClass(MVT::v2i1, &X86::VK2RegClass);
setOperationAction(ISD::CONCAT_VECTORS, MVT::v8i1, Custom);
setOperationAction(ISD::INSERT_SUBVECTOR, MVT::v8i1, Custom);
setOperationAction(ISD::INSERT_SUBVECTOR, MVT::v4i1, Custom);
+ setOperationAction(ISD::SELECT, MVT::v4i1, Custom);
+ setOperationAction(ISD::SELECT, MVT::v2i1, Custom);
setOperationAction(ISD::AND, MVT::v8i32, Legal);
setOperationAction(ISD::OR, MVT::v8i32, Legal);
setTargetDAGCombine(ISD::ANY_EXTEND);
setTargetDAGCombine(ISD::SIGN_EXTEND);
setTargetDAGCombine(ISD::SIGN_EXTEND_INREG);
- setTargetDAGCombine(ISD::TRUNCATE);
setTargetDAGCombine(ISD::SINT_TO_FP);
setTargetDAGCombine(ISD::SETCC);
setTargetDAGCombine(ISD::INTRINSIC_WO_CHAIN);
return TargetLowering::getJumpTableEncoding();
}
+bool X86TargetLowering::useSoftFloat() const {
+ return Subtarget->useSoftFloat();
+}
+
const MCExpr *
X86TargetLowering::LowerCustomJumpTableEntry(const MachineJumpTableInfo *MJTI,
const MachineBasicBlock *MBB,
const Function *Fn = MF.getFunction();
bool NoImplicitFloatOps = Fn->hasFnAttribute(Attribute::NoImplicitFloat);
- bool isSoftFloat = MF.getTarget().Options.UseSoftFloat;
+ bool isSoftFloat = Subtarget->useSoftFloat();
assert(!(isSoftFloat && NoImplicitFloatOps) &&
"SSE register cannot be used when SSE is disabled!");
if (isSoftFloat || NoImplicitFloatOps || !Subtarget->hasSSE1())
bool IsWinEHParent = WinEHParent && WinEHParent == Fn;
// Figure out if XMM registers are in use.
- assert(!(MF.getTarget().Options.UseSoftFloat &&
+ assert(!(Subtarget->useSoftFloat() &&
Fn->hasFnAttribute(Attribute::NoImplicitFloat)) &&
"SSE register cannot be used when SSE is disabled!");
// This isn't right, although it's probably harmless on x86; liveouts
// should be computed from returns not tail calls. Consider a void
// function making a tail call to a function returning int.
+ MF.getFrameInfo()->setHasTailCall();
return DAG.getNode(X86ISD::TC_RETURN, dl, NodeTys, Ops);
}
return false;
SDValue Ptr = MaskLoad->getBasePtr();
- if (Ptr->getOpcode() == X86ISD::Wrapper)
+ if (Ptr->getOpcode() == X86ISD::Wrapper ||
+ Ptr->getOpcode() == X86ISD::WrapperRIP)
Ptr = Ptr->getOperand(0);
auto *MaskCP = dyn_cast<ConstantPoolSDNode>(Ptr);
case MVT::v4i32:
case MVT::v8i32:
return lowerUINT_TO_FP_vXi32(Op, DAG, *Subtarget);
+ case MVT::v16i8:
+ case MVT::v16i16:
+ if (Subtarget->hasAVX512())
+ return DAG.getNode(ISD::UINT_TO_FP, dl, Op.getValueType(),
+ DAG.getNode(ISD::ZERO_EXTEND, dl, MVT::v16i32, N0));
}
llvm_unreachable(nullptr);
}
}
}
+ if (VT == MVT::v4i1 || VT == MVT::v2i1) {
+ SDValue zeroConst = DAG.getIntPtrConstant(0, DL);
+ Op1 = DAG.getNode(ISD::INSERT_SUBVECTOR, DL, MVT::v8i1,
+ DAG.getUNDEF(MVT::v8i1), Op1, zeroConst);
+ Op2 = DAG.getNode(ISD::INSERT_SUBVECTOR, DL, MVT::v8i1,
+ DAG.getUNDEF(MVT::v8i1), Op2, zeroConst);
+ SDValue newSelect = DAG.getNode(ISD::SELECT, DL, MVT::v8i1,
+ Cond, Op1, Op2);
+ return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, newSelect, zeroConst);
+ }
+
if (Cond.getOpcode() == ISD::SETCC) {
SDValue NewCond = LowerSETCC(Cond, DAG);
if (NewCond.getNode())
if (ArgMode == 2) {
// Sanity Check: Make sure using fp_offset makes sense.
- assert(!DAG.getTarget().Options.UseSoftFloat &&
+ assert(!Subtarget->useSoftFloat() &&
!(DAG.getMachineFunction().getFunction()->hasFnAttribute(
Attribute::NoImplicitFloat)) &&
Subtarget->hasSSE1());
return Op;
}
+SDValue X86TargetLowering::LowerGC_TRANSITION_START(SDValue Op,
+ SelectionDAG &DAG) const {
+ // TODO: Eventually, the lowering of these nodes should be informed by or
+ // deferred to the GC strategy for the function in which they appear. For
+ // now, however, they must be lowered to something. Since they are logically
+ // no-ops in the case of a null GC strategy (or a GC strategy which does not
+ // require special handling for these nodes), lower them as literal NOOPs for
+ // the time being.
+ SmallVector<SDValue, 2> Ops;
+
+ Ops.push_back(Op.getOperand(0));
+ if (Op->getGluedNode())
+ Ops.push_back(Op->getOperand(Op->getNumOperands() - 1));
+
+ SDLoc OpDL(Op);
+ SDVTList VTs = DAG.getVTList(MVT::Other, MVT::Glue);
+ SDValue NOOP(DAG.getMachineNode(X86::NOOP, SDLoc(Op), VTs, Ops), 0);
+
+ return NOOP;
+}
+
+SDValue X86TargetLowering::LowerGC_TRANSITION_END(SDValue Op,
+ SelectionDAG &DAG) const {
+ // TODO: Eventually, the lowering of these nodes should be informed by or
+ // deferred to the GC strategy for the function in which they appear. For
+ // now, however, they must be lowered to something. Since they are logically
+ // no-ops in the case of a null GC strategy (or a GC strategy which does not
+ // require special handling for these nodes), lower them as literal NOOPs for
+ // the time being.
+ SmallVector<SDValue, 2> Ops;
+
+ Ops.push_back(Op.getOperand(0));
+ if (Op->getGluedNode())
+ Ops.push_back(Op->getOperand(Op->getNumOperands() - 1));
+
+ SDLoc OpDL(Op);
+ SDVTList VTs = DAG.getVTList(MVT::Other, MVT::Glue);
+ SDValue NOOP(DAG.getMachineNode(X86::NOOP, SDLoc(Op), VTs, Ops), 0);
+
+ return NOOP;
+}
+
/// LowerOperation - Provide custom lowering hooks for some operations.
///
SDValue X86TargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const {
case ISD::FSINCOS: return LowerFSINCOS(Op, Subtarget, DAG);
case ISD::MGATHER: return LowerMGATHER(Op, Subtarget, DAG);
case ISD::MSCATTER: return LowerMSCATTER(Op, Subtarget, DAG);
+ case ISD::GC_TRANSITION_START:
+ return LowerGC_TRANSITION_START(Op, DAG);
+ case ISD::GC_TRANSITION_END: return LowerGC_TRANSITION_END(Op, DAG);
}
}
return;
}
case ISD::FP_TO_SINT:
+ // FP_TO_INT*_IN_MEM is not legal for f16 inputs. Do not convert
+ // (FP_TO_SINT (load f16)) to FP_TO_INT*.
+ if (N->getOperand(0).getValueType() == MVT::f16)
+ break;
+ // fallthrough
case ISD::FP_TO_UINT: {
bool IsSigned = N->getOpcode() == ISD::FP_TO_SINT;
Results.push_back(V);
return;
}
+ case ISD::FP_EXTEND: {
+ // Right now, only MVT::v2f32 has OperationAction for FP_EXTEND.
+ // No other ValueType for FP_EXTEND should reach this point.
+ assert(N->getValueType(0) == MVT::v2f32 &&
+ "Do not know how to legalize this Node");
+ return;
+ }
case ISD::INTRINSIC_W_CHAIN: {
unsigned IntNo = cast<ConstantSDNode>(N->getOperand(1))->getZExtValue();
switch (IntNo) {
case X86ISD::SMAX: return "X86ISD::SMAX";
case X86ISD::SMIN: return "X86ISD::SMIN";
case X86ISD::FMAX: return "X86ISD::FMAX";
+ case X86ISD::FMAX_RND: return "X86ISD::FMAX_RND";
case X86ISD::FMIN: return "X86ISD::FMIN";
+ case X86ISD::FMIN_RND: return "X86ISD::FMIN_RND";
case X86ISD::FMAXC: return "X86ISD::FMAXC";
case X86ISD::FMINC: return "X86ISD::FMINC";
case X86ISD::FRSQRT: return "X86ISD::FRSQRT";
if (!VT.isSimple())
return false;
+ // Not for i1 vectors
+ if (VT.getScalarType() == MVT::i1)
+ return false;
+
// Very little shuffling can be done for 64-bit vectors right now.
if (VT.getSizeInBits() == 64)
return false;
case X86::CMOV_RFP32:
case X86::CMOV_RFP64:
case X86::CMOV_RFP80:
+ case X86::CMOV_V8I1:
+ case X86::CMOV_V16I1:
+ case X86::CMOV_V32I1:
+ case X86::CMOV_V64I1:
return EmitLoweredSelect(MI, BB);
case X86::FP32_TO_INT16_IN_MEM:
return SDValue();
}
-/// PerformTruncateCombine - Converts truncate operation to
-/// a sequence of vector shuffle operations.
-/// It is possible when we truncate 256-bit vector to 128-bit vector
-static SDValue PerformTruncateCombine(SDNode *N, SelectionDAG &DAG,
- TargetLowering::DAGCombinerInfo &DCI,
- const X86Subtarget *Subtarget) {
- return SDValue();
-}
-
/// XFormVExtractWithShuffleIntoLoad - Check if a vector extract from a target
/// specific shuffle of a load can be folded into a single element load.
/// Similar handling for VECTOR_SHUFFLE is performed by DAGCombiner, but
const Function *F = DAG.getMachineFunction().getFunction();
bool NoImplicitFloatOps = F->hasFnAttribute(Attribute::NoImplicitFloat);
- bool F64IsLegal = !DAG.getTarget().Options.UseSoftFloat && !NoImplicitFloatOps
- && Subtarget->hasSSE2();
+ bool F64IsLegal =
+ !Subtarget->useSoftFloat() && !NoImplicitFloatOps && Subtarget->hasSSE2();
if ((VT.isVector() ||
(VT == MVT::i64 && F64IsLegal && !Subtarget->is64Bit())) &&
isa<LoadSDNode>(St->getValue()) &&
SDValue OldExtract = St->getOperand(1);
SDValue ExtOp0 = OldExtract.getOperand(0);
unsigned VecSize = ExtOp0.getValueSizeInBits();
- MVT VecVT = MVT::getVectorVT(MVT::f64, VecSize / 64);
+ EVT VecVT = EVT::getVectorVT(*DAG.getContext(), MVT::f64, VecSize / 64);
SDValue BitCast = DAG.getNode(ISD::BITCAST, dl, VecVT, ExtOp0);
SDValue NewExtract = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::f64,
BitCast, OldExtract.getOperand(1));
if (Op0.getOpcode() == ISD::LOAD) {
LoadSDNode *Ld = cast<LoadSDNode>(Op0.getNode());
EVT VT = Ld->getValueType(0);
+
+ // This transformation is not supported if the result type is f16
+ if (N->getValueType(0) == MVT::f16)
+ return SDValue();
+
if (!Ld->isVolatile() && !N->getValueType(0).isVector() &&
ISD::isNON_EXTLoad(Op0.getNode()) && Op0.hasOneUse() &&
!Subtarget->is64Bit() && VT == MVT::i64) {
case ISD::SIGN_EXTEND: return PerformSExtCombine(N, DAG, DCI, Subtarget);
case ISD::SIGN_EXTEND_INREG:
return PerformSIGN_EXTEND_INREGCombine(N, DAG, Subtarget);
- case ISD::TRUNCATE: return PerformTruncateCombine(N, DAG,DCI,Subtarget);
case ISD::SETCC: return PerformISDSETCCCombine(N, DAG, Subtarget);
case X86ISD::SETCC: return PerformSETCCCombine(N, DAG, DCI, Subtarget);
case X86ISD::BRCOND: return PerformBrCondCombine(N, DAG, DCI, Subtarget);
projects / oota-llvm.git / blobdiff