// AArch64 does not have floating-point extending loads, i1 sign-extending
// load, floating-point truncating stores, or v2i32->v2i16 truncating store.
+ setLoadExtAction(ISD::EXTLOAD, MVT::f16, Expand);
setLoadExtAction(ISD::EXTLOAD, MVT::f32, Expand);
setLoadExtAction(ISD::EXTLOAD, MVT::f64, Expand);
setLoadExtAction(ISD::EXTLOAD, MVT::f80, Expand);
setTruncStoreAction(MVT::f128, MVT::f64, Expand);
setTruncStoreAction(MVT::f128, MVT::f32, Expand);
setTruncStoreAction(MVT::f128, MVT::f16, Expand);
+
+ setOperationAction(ISD::BITCAST, MVT::i16, Custom);
+ setOperationAction(ISD::BITCAST, MVT::f16, Custom);
+
// Indexed loads and stores are supported.
for (unsigned im = (unsigned)ISD::PRE_INC;
im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
unsigned AArch64TargetLowering::getMaximalGlobalOffset() const {
// FIXME: On AArch64, this depends on the type.
- // Basically, the addressable offsets are o to 4095 * Ty.getSizeInBytes().
+ // Basically, the addressable offsets are up to 4095 * Ty.getSizeInBytes().
// and the offset has to be a multiple of the related size in bytes.
return 4095;
}
StructType *RetTy = StructType::get(ArgTy, ArgTy, NULL);
TargetLowering::CallLoweringInfo CLI(DAG);
CLI.setDebugLoc(dl).setChain(DAG.getEntryNode())
- .setCallee(CallingConv::Fast, RetTy, Callee, &Args, 0);
+ .setCallee(CallingConv::Fast, RetTy, Callee, std::move(Args), 0);
std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
return CallResult.first;
}
+static SDValue LowerBITCAST(SDValue Op, SelectionDAG &DAG) {
+ if (Op.getValueType() != MVT::f16)
+ return SDValue();
+
+ assert(Op.getOperand(0).getValueType() == MVT::i16);
+ SDLoc DL(Op);
+
+ Op = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i32, Op.getOperand(0));
+ Op = DAG.getNode(ISD::BITCAST, DL, MVT::f32, Op);
+ return SDValue(
+ DAG.getMachineNode(TargetOpcode::EXTRACT_SUBREG, DL, MVT::f16, Op,
+ DAG.getTargetConstant(AArch64::hsub, MVT::i32)),
+ 0);
+}
+
+
SDValue AArch64TargetLowering::LowerOperation(SDValue Op,
SelectionDAG &DAG) const {
switch (Op.getOpcode()) {
default:
llvm_unreachable("unimplemented operand");
return SDValue();
+ case ISD::BITCAST:
+ return LowerBITCAST(Op, DAG);
case ISD::GlobalAddress:
return LowerGlobalAddress(Op, DAG);
case ISD::GlobalTLSAddress:
RC = &AArch64::GPR32RegClass;
else if (RegVT == MVT::i64)
RC = &AArch64::GPR64RegClass;
+ else if (RegVT == MVT::f16)
+ RC = &AArch64::FPR16RegClass;
else if (RegVT == MVT::f32)
RC = &AArch64::FPR32RegClass;
else if (RegVT == MVT::f64 || RegVT.is64BitVector())
return Op;
}
+// Normalize the operands of BUILD_VECTOR. The value of constant operands will
+// be truncated to fit element width.
+static SDValue NormalizeBuildVector(SDValue Op,
+ SelectionDAG &DAG) {
+ assert(Op.getOpcode() == ISD::BUILD_VECTOR && "Unknown opcode!");
+ SDLoc dl(Op);
+ EVT VT = Op.getValueType();
+ EVT EltTy= VT.getVectorElementType();
+
+ if (EltTy.isFloatingPoint() || EltTy.getSizeInBits() > 16)
+ return Op;
+
+ SmallVector<SDValue, 16> Ops;
+ for (unsigned I = 0, E = VT.getVectorNumElements(); I != E; ++I) {
+ SDValue Lane = Op.getOperand(I);
+ if (Lane.getOpcode() == ISD::Constant) {
+ APInt LowBits(EltTy.getSizeInBits(),
+ cast<ConstantSDNode>(Lane)->getZExtValue());
+ Lane = DAG.getConstant(LowBits.getZExtValue(), MVT::i32);
+ }
+ Ops.push_back(Lane);
+ }
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, Ops);
+}
+
SDValue AArch64TargetLowering::LowerBUILD_VECTOR(SDValue Op,
SelectionDAG &DAG) const {
- BuildVectorSDNode *BVN = cast<BuildVectorSDNode>(Op.getNode());
SDLoc dl(Op);
EVT VT = Op.getValueType();
+ Op = NormalizeBuildVector(Op, DAG);
+ BuildVectorSDNode *BVN = cast<BuildVectorSDNode>(Op.getNode());
APInt CnstBits(VT.getSizeInBits(), 0);
APInt UndefBits(VT.getSizeInBits(), 0);
SelectionDAG &DAG) const {
assert(Op.getOpcode() == ISD::INSERT_VECTOR_ELT && "Unknown opcode!");
- // Check for non-constant lane.
- if (!isa<ConstantSDNode>(Op.getOperand(2)))
+ // Check for non-constant or out of range lane.
+ EVT VT = Op.getOperand(0).getValueType();
+ ConstantSDNode *CI = dyn_cast<ConstantSDNode>(Op.getOperand(2));
+ if (!CI || CI->getZExtValue() >= VT.getVectorNumElements())
return SDValue();
- EVT VT = Op.getOperand(0).getValueType();
// Insertion/extraction are legal for V128 types.
if (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32 ||
SelectionDAG &DAG) const {
assert(Op.getOpcode() == ISD::EXTRACT_VECTOR_ELT && "Unknown opcode!");
- // Check for non-constant lane.
- if (!isa<ConstantSDNode>(Op.getOperand(1)))
+ // Check for non-constant or out of range lane.
+ EVT VT = Op.getOperand(0).getValueType();
+ ConstantSDNode *CI = dyn_cast<ConstantSDNode>(Op.getOperand(1));
+ if (!CI || CI->getZExtValue() >= VT.getVectorNumElements())
return SDValue();
- EVT VT = Op.getOperand(0).getValueType();
// Insertion/extraction are legal for V128 types.
if (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32 ||
if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(N->getOperand(1))) {
APInt Value = C->getAPIntValue();
EVT VT = N->getValueType(0);
- APInt VM1 = Value - 1;
- if (VM1.isPowerOf2()) {
- // Multiplying by one more than a power of two, replace with a shift
- // and an add.
- SDValue ShiftedVal =
- DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
- DAG.getConstant(VM1.logBase2(), MVT::i64));
- return DAG.getNode(ISD::ADD, SDLoc(N), VT, ShiftedVal, N->getOperand(0));
- }
- APInt VP1 = Value + 1;
- if (VP1.isPowerOf2()) {
- // Multiplying by one less than a power of two, replace with a shift
- // and a subtract.
- SDValue ShiftedVal =
- DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
- DAG.getConstant(VP1.logBase2(), MVT::i64));
- return DAG.getNode(ISD::SUB, SDLoc(N), VT, ShiftedVal, N->getOperand(0));
+ if (Value.isNonNegative()) {
+ // (mul x, 2^N + 1) => (add (shl x, N), x)
+ APInt VM1 = Value - 1;
+ if (VM1.isPowerOf2()) {
+ SDValue ShiftedVal =
+ DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+ DAG.getConstant(VM1.logBase2(), MVT::i64));
+ return DAG.getNode(ISD::ADD, SDLoc(N), VT, ShiftedVal,
+ N->getOperand(0));
+ }
+ // (mul x, 2^N - 1) => (sub (shl x, N), x)
+ APInt VP1 = Value + 1;
+ if (VP1.isPowerOf2()) {
+ SDValue ShiftedVal =
+ DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+ DAG.getConstant(VP1.logBase2(), MVT::i64));
+ return DAG.getNode(ISD::SUB, SDLoc(N), VT, ShiftedVal,
+ N->getOperand(0));
+ }
+ } else {
+ // (mul x, -(2^N + 1)) => - (add (shl x, N), x)
+ APInt VNM1 = -Value - 1;
+ if (VNM1.isPowerOf2()) {
+ SDValue ShiftedVal =
+ DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+ DAG.getConstant(VNM1.logBase2(), MVT::i64));
+ SDValue Add =
+ DAG.getNode(ISD::ADD, SDLoc(N), VT, ShiftedVal, N->getOperand(0));
+ return DAG.getNode(ISD::SUB, SDLoc(N), VT, DAG.getConstant(0, VT), Add);
+ }
+ // (mul x, -(2^N - 1)) => (sub x, (shl x, N))
+ APInt VNP1 = -Value + 1;
+ if (VNP1.isPowerOf2()) {
+ SDValue ShiftedVal =
+ DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+ DAG.getConstant(VNP1.logBase2(), MVT::i64));
+ return DAG.getNode(ISD::SUB, SDLoc(N), VT, N->getOperand(0),
+ ShiftedVal);
+ }
}
}
return SDValue();
}
+static SDValue performVectorCompareAndMaskUnaryOpCombine(SDNode *N,
+ SelectionDAG &DAG) {
+ // Take advantage of vector comparisons producing 0 or -1 in each lane to
+ // optimize away operation when it's from a constant.
+ //
+ // The general transformation is:
+ // UNARYOP(AND(VECTOR_CMP(x,y), constant)) -->
+ // AND(VECTOR_CMP(x,y), constant2)
+ // constant2 = UNARYOP(constant)
+
+ // Early exit if this isn't a vector operation or if the operand of the
+ // unary operation isn't a bitwise AND.
+ EVT VT = N->getValueType(0);
+ if (!VT.isVector() || N->getOperand(0)->getOpcode() != ISD::AND ||
+ N->getOperand(0)->getOperand(0)->getOpcode() != ISD::SETCC)
+ return SDValue();
+
+ // Now check that the other operand of the AND is a constant splat. We could
+ // make the transformation for non-constant splats as well, but it's unclear
+ // that would be a benefit as it would not eliminate any operations, just
+ // perform one more step in scalar code before moving to the vector unit.
+ if (BuildVectorSDNode *BV =
+ dyn_cast<BuildVectorSDNode>(N->getOperand(0)->getOperand(1))) {
+ // Bail out if the vector isn't a constant splat.
+ if (!BV->getConstantSplatNode())
+ return SDValue();
+
+ // Everything checks out. Build up the new and improved node.
+ SDLoc DL(N);
+ EVT IntVT = BV->getValueType(0);
+ // Create a new constant of the appropriate type for the transformed
+ // DAG.
+ SDValue SourceConst = DAG.getNode(N->getOpcode(), DL, VT, SDValue(BV, 0));
+ // The AND node needs bitcasts to/from an integer vector type around it.
+ SDValue MaskConst = DAG.getNode(ISD::BITCAST, DL, IntVT, SourceConst);
+ SDValue NewAnd = DAG.getNode(ISD::AND, DL, IntVT,
+ N->getOperand(0)->getOperand(0), MaskConst);
+ SDValue Res = DAG.getNode(ISD::BITCAST, DL, VT, NewAnd);
+ return Res;
+ }
+
+ return SDValue();
+}
+
static SDValue performIntToFpCombine(SDNode *N, SelectionDAG &DAG) {
+ // First try to optimize away the conversion when it's conditionally from
+ // a constant. Vectors only.
+ SDValue Res = performVectorCompareAndMaskUnaryOpCombine(N, DAG);
+ if (Res != SDValue())
+ return Res;
+
EVT VT = N->getValueType(0);
if (VT != MVT::f32 && VT != MVT::f64)
return SDValue();
+
// Only optimize when the source and destination types have the same width.
if (VT.getSizeInBits() != N->getOperand(0).getValueType().getSizeInBits())
return SDValue();
return true;
}
+static void ReplaceBITCASTResults(SDNode *N, SmallVectorImpl<SDValue> &Results,
+ SelectionDAG &DAG) {
+ if (N->getValueType(0) != MVT::i16)
+ return;
+
+ SDLoc DL(N);
+ SDValue Op = N->getOperand(0);
+ assert(Op.getValueType() == MVT::f16 &&
+ "Inconsistent bitcast? Only 16-bit types should be i16 or f16");
+ Op = SDValue(
+ DAG.getMachineNode(TargetOpcode::INSERT_SUBREG, DL, MVT::f32,
+ DAG.getUNDEF(MVT::i32), Op,
+ DAG.getTargetConstant(AArch64::hsub, MVT::i32)),
+ 0);
+ Op = DAG.getNode(ISD::BITCAST, DL, MVT::i32, Op);
+ Results.push_back(DAG.getNode(ISD::TRUNCATE, DL, MVT::i16, Op));
+}
+
void AArch64TargetLowering::ReplaceNodeResults(
SDNode *N, SmallVectorImpl<SDValue> &Results, SelectionDAG &DAG) const {
switch (N->getOpcode()) {
default:
llvm_unreachable("Don't know how to custom expand this");
+ case ISD::BITCAST:
+ ReplaceBITCASTResults(N, Results, DAG);
+ return;
case ISD::FP_TO_UINT:
case ISD::FP_TO_SINT:
assert(N->getValueType(0) == MVT::i128 && "unexpected illegal conversion");
return Inst->getType()->getPrimitiveSizeInBits() <= 128;
}
+TargetLoweringBase::LegalizeTypeAction
+AArch64TargetLowering::getPreferredVectorAction(EVT VT) const {
+ MVT SVT = VT.getSimpleVT();
+ // During type legalization, we prefer to widen v1i8, v1i16, v1i32 to v8i8,
+ // v4i16, v2i32 instead of to promote.
+ if (SVT == MVT::v1i8 || SVT == MVT::v1i16 || SVT == MVT::v1i32
+ || SVT == MVT::v1f32)
+ return TypeWidenVector;
+
+ return TargetLoweringBase::getPreferredVectorAction(VT);
+}
+
Value *AArch64TargetLowering::emitLoadLinked(IRBuilder<> &Builder, Value *Addr,
AtomicOrdering Ord) const {
Module *M = Builder.GetInsertBlock()->getParent()->getParent();