summaryrefslogtreecommitdiff
path: root/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
diff options
context:
space:
mode:
authorDimitry Andric <dim@FreeBSD.org>2017-05-02 18:30:13 +0000
committerDimitry Andric <dim@FreeBSD.org>2017-05-02 18:30:13 +0000
commita303c417bbdb53703c2c17398b08486bde78f1f6 (patch)
tree98366d6b93d863cefdc53f16c66c0c5ae7fb2261 /lib/CodeGen/SelectionDAG/DAGCombiner.cpp
parent12f3ca4cdb95b193af905a00e722a4dcb40b3de3 (diff)
Notes
Diffstat (limited to 'lib/CodeGen/SelectionDAG/DAGCombiner.cpp')
-rw-r--r--lib/CodeGen/SelectionDAG/DAGCombiner.cpp311
1 files changed, 227 insertions, 84 deletions
diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 1251ae6262b89..dc0276d576675 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -33,6 +33,7 @@
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/KnownBits.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetLowering.h"
@@ -236,10 +237,13 @@ namespace {
SDValue visitSUB(SDNode *N);
SDValue visitADDC(SDNode *N);
SDValue visitUADDO(SDNode *N);
+ SDValue visitUADDOLike(SDValue N0, SDValue N1, SDNode *N);
SDValue visitSUBC(SDNode *N);
SDValue visitUSUBO(SDNode *N);
SDValue visitADDE(SDNode *N);
+ SDValue visitADDCARRY(SDNode *N);
SDValue visitSUBE(SDNode *N);
+ SDValue visitSUBCARRY(SDNode *N);
SDValue visitMUL(SDNode *N);
SDValue useDivRem(SDNode *N);
SDValue visitSDIV(SDNode *N);
@@ -369,14 +373,14 @@ namespace {
SDValue BuildSDIVPow2(SDNode *N);
SDValue BuildUDIV(SDNode *N);
SDValue BuildLogBase2(SDValue Op, const SDLoc &DL);
- SDValue BuildReciprocalEstimate(SDValue Op, SDNodeFlags *Flags);
- SDValue buildRsqrtEstimate(SDValue Op, SDNodeFlags *Flags);
- SDValue buildSqrtEstimate(SDValue Op, SDNodeFlags *Flags);
- SDValue buildSqrtEstimateImpl(SDValue Op, SDNodeFlags *Flags, bool Recip);
+ SDValue BuildReciprocalEstimate(SDValue Op, SDNodeFlags Flags);
+ SDValue buildRsqrtEstimate(SDValue Op, SDNodeFlags Flags);
+ SDValue buildSqrtEstimate(SDValue Op, SDNodeFlags Flags);
+ SDValue buildSqrtEstimateImpl(SDValue Op, SDNodeFlags Flags, bool Recip);
SDValue buildSqrtNROneConst(SDValue Op, SDValue Est, unsigned Iterations,
- SDNodeFlags *Flags, bool Reciprocal);
+ SDNodeFlags Flags, bool Reciprocal);
SDValue buildSqrtNRTwoConst(SDValue Op, SDValue Est, unsigned Iterations,
- SDNodeFlags *Flags, bool Reciprocal);
+ SDNodeFlags Flags, bool Reciprocal);
SDValue MatchBSwapHWordLow(SDNode *N, SDValue N0, SDValue N1,
bool DemandHighBits = true);
SDValue MatchBSwapHWord(SDNode *N, SDValue N0, SDValue N1);
@@ -396,6 +400,7 @@ namespace {
SDValue createBuildVecShuffle(const SDLoc &DL, SDNode *N,
ArrayRef<int> VectorMask, SDValue VecIn1,
SDValue VecIn2, unsigned LeftIdx);
+ SDValue matchVSelectOpSizesWithSetCC(SDNode *N);
SDValue GetDemandedBits(SDValue V, const APInt &Mask);
@@ -644,7 +649,7 @@ static char isNegatibleForFree(SDValue Op, bool LegalOperations,
case ISD::FSUB:
// We can't turn -(A-B) into B-A when we honor signed zeros.
if (!Options->NoSignedZerosFPMath &&
- !Op.getNode()->getFlags()->hasNoSignedZeros())
+ !Op.getNode()->getFlags().hasNoSignedZeros())
return 0;
// fold (fneg (fsub A, B)) -> (fsub B, A)
@@ -682,7 +687,7 @@ static SDValue GetNegatedExpression(SDValue Op, SelectionDAG &DAG,
assert(Depth <= 6 && "GetNegatedExpression doesn't match isNegatibleForFree");
- const SDNodeFlags *Flags = Op.getNode()->getFlags();
+ const SDNodeFlags Flags = Op.getNode()->getFlags();
switch (Op.getOpcode()) {
default: llvm_unreachable("Unknown code");
@@ -965,8 +970,8 @@ CommitTargetLoweringOpt(const TargetLowering::TargetLoweringOpt &TLO) {
/// things it uses can be simplified by bit propagation. If so, return true.
bool DAGCombiner::SimplifyDemandedBits(SDValue Op, const APInt &Demanded) {
TargetLowering::TargetLoweringOpt TLO(DAG, LegalTypes, LegalOperations);
- APInt KnownZero, KnownOne;
- if (!TLI.SimplifyDemandedBits(Op, Demanded, KnownZero, KnownOne, TLO))
+ KnownBits Known;
+ if (!TLI.SimplifyDemandedBits(Op, Demanded, Known, TLO))
return false;
// Revisit the node.
@@ -1412,7 +1417,9 @@ SDValue DAGCombiner::visit(SDNode *N) {
case ISD::SUBC: return visitSUBC(N);
case ISD::USUBO: return visitUSUBO(N);
case ISD::ADDE: return visitADDE(N);
+ case ISD::ADDCARRY: return visitADDCARRY(N);
case ISD::SUBE: return visitSUBE(N);
+ case ISD::SUBCARRY: return visitSUBCARRY(N);
case ISD::MUL: return visitMUL(N);
case ISD::SDIV: return visitSDIV(N);
case ISD::UDIV: return visitUDIV(N);
@@ -1866,14 +1873,31 @@ SDValue DAGCombiner::visitADD(SDNode *N) {
if (isNullConstant(N1))
return N0;
- // fold ((c1-A)+c2) -> (c1+c2)-A
if (isConstantOrConstantVector(N1, /* NoOpaque */ true)) {
- if (N0.getOpcode() == ISD::SUB)
- if (isConstantOrConstantVector(N0.getOperand(0), /* NoOpaque */ true)) {
- return DAG.getNode(ISD::SUB, DL, VT,
- DAG.getNode(ISD::ADD, DL, VT, N1, N0.getOperand(0)),
- N0.getOperand(1));
+ // fold ((c1-A)+c2) -> (c1+c2)-A
+ if (N0.getOpcode() == ISD::SUB &&
+ isConstantOrConstantVector(N0.getOperand(0), /* NoOpaque */ true)) {
+ // FIXME: Adding 2 constants should be handled by FoldConstantArithmetic.
+ return DAG.getNode(ISD::SUB, DL, VT,
+ DAG.getNode(ISD::ADD, DL, VT, N1, N0.getOperand(0)),
+ N0.getOperand(1));
+ }
+
+ // add (sext i1 X), 1 -> zext (not i1 X)
+ // We don't transform this pattern:
+ // add (zext i1 X), -1 -> sext (not i1 X)
+ // because most (?) targets generate better code for the zext form.
+ if (N0.getOpcode() == ISD::SIGN_EXTEND && N0.hasOneUse() &&
+ isOneConstantOrOneSplatConstant(N1)) {
+ SDValue X = N0.getOperand(0);
+ if ((!LegalOperations ||
+ (TLI.isOperationLegal(ISD::XOR, X.getValueType()) &&
+ TLI.isOperationLegal(ISD::ZERO_EXTEND, VT))) &&
+ X.getScalarValueSizeInBits() == 1) {
+ SDValue Not = DAG.getNOT(DL, X, X.getValueType());
+ return DAG.getNode(ISD::ZERO_EXTEND, DL, VT, Not);
}
+ }
}
if (SDValue NewSel = foldBinOpIntoSelect(N))
@@ -1992,6 +2016,11 @@ SDValue DAGCombiner::visitADDLike(SDValue N0, SDValue N1, SDNode *LocReference)
}
}
+ // (add X, (addcarry Y, 0, Carry)) -> (addcarry X, Y, Carry)
+ if (N1.getOpcode() == ISD::ADDCARRY && isNullConstant(N1.getOperand(1)))
+ return DAG.getNode(ISD::ADDCARRY, DL, N1->getVTList(),
+ N0, N1.getOperand(0), N1.getOperand(2));
+
return SDValue();
}
@@ -2055,6 +2084,26 @@ SDValue DAGCombiner::visitUADDO(SDNode *N) {
return CombineTo(N, DAG.getNode(ISD::ADD, DL, VT, N0, N1),
DAG.getConstant(0, DL, CarryVT));
+ if (SDValue Combined = visitUADDOLike(N0, N1, N))
+ return Combined;
+
+ if (SDValue Combined = visitUADDOLike(N1, N0, N))
+ return Combined;
+
+ return SDValue();
+}
+
+SDValue DAGCombiner::visitUADDOLike(SDValue N0, SDValue N1, SDNode *N) {
+ // (uaddo X, (addcarry Y, 0, Carry)) -> (addcarry X, Y, Carry)
+ // If Y + 1 cannot overflow.
+ if (N1.getOpcode() == ISD::ADDCARRY && isNullConstant(N1.getOperand(1))) {
+ SDValue Y = N1.getOperand(0);
+ SDValue One = DAG.getConstant(1, SDLoc(N), Y.getValueType());
+ if (DAG.computeOverflowKind(Y, One) == SelectionDAG::OFK_Never)
+ return DAG.getNode(ISD::ADDCARRY, SDLoc(N), N->getVTList(), N0, Y,
+ N1.getOperand(2));
+ }
+
return SDValue();
}
@@ -2077,6 +2126,25 @@ SDValue DAGCombiner::visitADDE(SDNode *N) {
return SDValue();
}
+SDValue DAGCombiner::visitADDCARRY(SDNode *N) {
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ SDValue CarryIn = N->getOperand(2);
+
+ // canonicalize constant to RHS
+ ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
+ ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
+ if (N0C && !N1C)
+ return DAG.getNode(ISD::ADDCARRY, SDLoc(N), N->getVTList(),
+ N1, N0, CarryIn);
+
+ // fold (addcarry x, y, false) -> (uaddo x, y)
+ if (isNullConstant(CarryIn))
+ return DAG.getNode(ISD::UADDO, SDLoc(N), N->getVTList(), N0, N1);
+
+ return SDValue();
+}
+
// Since it may not be valid to emit a fold to zero for vector initializers
// check if we can before folding.
static SDValue tryFoldToZero(const SDLoc &DL, const TargetLowering &TLI, EVT VT,
@@ -2143,13 +2211,13 @@ SDValue DAGCombiner::visitSUB(SDNode *N) {
}
// 0 - X --> 0 if the sub is NUW.
- if (N->getFlags()->hasNoUnsignedWrap())
+ if (N->getFlags().hasNoUnsignedWrap())
return N0;
if (DAG.MaskedValueIsZero(N1, ~APInt::getSignMask(BitWidth))) {
// N1 is either 0 or the minimum signed value. If the sub is NSW, then
// N1 must be 0 because negating the minimum signed value is undefined.
- if (N->getFlags()->hasNoSignedWrap())
+ if (N->getFlags().hasNoSignedWrap())
return N0;
// 0 - X --> X if X is 0 or the minimum signed value.
@@ -2309,6 +2377,18 @@ SDValue DAGCombiner::visitSUBE(SDNode *N) {
return SDValue();
}
+SDValue DAGCombiner::visitSUBCARRY(SDNode *N) {
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ SDValue CarryIn = N->getOperand(2);
+
+ // fold (subcarry x, y, false) -> (usubo x, y)
+ if (isNullConstant(CarryIn))
+ return DAG.getNode(ISD::USUBO, SDLoc(N), N->getVTList(), N0, N1);
+
+ return SDValue();
+}
+
SDValue DAGCombiner::visitMUL(SDNode *N) {
SDValue N0 = N->getOperand(0);
SDValue N1 = N->getOperand(1);
@@ -2589,9 +2669,8 @@ SDValue DAGCombiner::visitSDIV(SDNode *N) {
// better results in that case. The target-specific lowering should learn how
// to handle exact sdivs efficiently.
if (N1C && !N1C->isNullValue() && !N1C->isOpaque() &&
- !cast<BinaryWithFlagsSDNode>(N)->Flags.hasExact() &&
- (N1C->getAPIntValue().isPowerOf2() ||
- (-N1C->getAPIntValue()).isPowerOf2())) {
+ !N->getFlags().hasExact() && (N1C->getAPIntValue().isPowerOf2() ||
+ (-N1C->getAPIntValue()).isPowerOf2())) {
// Target-specific implementation of sdiv x, pow2.
if (SDValue Res = BuildSDIVPow2(N))
return Res;
@@ -3766,7 +3845,7 @@ SDValue DAGCombiner::MatchBSwapHWordLow(SDNode *N, SDValue N0, SDValue N1,
EVT VT = N->getValueType(0);
if (VT != MVT::i64 && VT != MVT::i32 && VT != MVT::i16)
return SDValue();
- if (!TLI.isOperationLegal(ISD::BSWAP, VT))
+ if (!TLI.isOperationLegalOrCustom(ISD::BSWAP, VT))
return SDValue();
// Recognize (and (shl a, 8), 0xff), (and (srl a, 8), 0xff00)
@@ -3880,8 +3959,15 @@ static bool isBSwapHWordElement(SDValue N, MutableArrayRef<SDNode *> Parts) {
SDValue N0 = N.getOperand(0);
unsigned Opc0 = N0.getOpcode();
+ if (Opc0 != ISD::AND && Opc0 != ISD::SHL && Opc0 != ISD::SRL)
+ return false;
- ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N.getOperand(1));
+ ConstantSDNode *N1C = nullptr;
+ // SHL or SRL: look upstream for AND mask operand
+ if (Opc == ISD::AND)
+ N1C = dyn_cast<ConstantSDNode>(N.getOperand(1));
+ else if (Opc0 == ISD::AND)
+ N1C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
if (!N1C)
return false;
@@ -3952,7 +4038,7 @@ SDValue DAGCombiner::MatchBSwapHWord(SDNode *N, SDValue N0, SDValue N1) {
EVT VT = N->getValueType(0);
if (VT != MVT::i32)
return SDValue();
- if (!TLI.isOperationLegal(ISD::BSWAP, VT))
+ if (!TLI.isOperationLegalOrCustom(ISD::BSWAP, VT))
return SDValue();
// Look for either
@@ -3967,18 +4053,16 @@ SDValue DAGCombiner::MatchBSwapHWord(SDNode *N, SDValue N0, SDValue N1) {
if (N1.getOpcode() == ISD::OR &&
N00.getNumOperands() == 2 && N01.getNumOperands() == 2) {
// (or (or (and), (and)), (or (and), (and)))
- SDValue N000 = N00.getOperand(0);
- if (!isBSwapHWordElement(N000, Parts))
+ if (!isBSwapHWordElement(N00, Parts))
return SDValue();
- SDValue N001 = N00.getOperand(1);
- if (!isBSwapHWordElement(N001, Parts))
+ if (!isBSwapHWordElement(N01, Parts))
return SDValue();
- SDValue N010 = N01.getOperand(0);
- if (!isBSwapHWordElement(N010, Parts))
+ SDValue N10 = N1.getOperand(0);
+ if (!isBSwapHWordElement(N10, Parts))
return SDValue();
- SDValue N011 = N01.getOperand(1);
- if (!isBSwapHWordElement(N011, Parts))
+ SDValue N11 = N1.getOperand(1);
+ if (!isBSwapHWordElement(N11, Parts))
return SDValue();
} else {
// (or (or (or (and), (and)), (and)), (and))
@@ -5322,7 +5406,7 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
// fold (shl (sr[la] exact X, C1), C2) -> (shl X, (C2-C1)) if C1 <= C2
// fold (shl (sr[la] exact X, C1), C2) -> (sr[la] X, (C2-C1)) if C1 > C2
if (N1C && (N0.getOpcode() == ISD::SRL || N0.getOpcode() == ISD::SRA) &&
- cast<BinaryWithFlagsSDNode>(N0)->Flags.hasExact()) {
+ N0->getFlags().hasExact()) {
if (ConstantSDNode *N0C1 = isConstOrConstSplat(N0.getOperand(1))) {
uint64_t C1 = N0C1->getZExtValue();
uint64_t C2 = N1C->getZExtValue();
@@ -5347,7 +5431,7 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
APInt Mask = APInt::getHighBitsSet(OpSizeInBits, OpSizeInBits - c1);
SDValue Shift;
if (c2 > c1) {
- Mask = Mask.shl(c2 - c1);
+ Mask <<= c2 - c1;
SDLoc DL(N);
Shift = DAG.getNode(ISD::SHL, DL, VT, N0.getOperand(0),
DAG.getConstant(c2 - c1, DL, N1.getValueType()));
@@ -5680,20 +5764,20 @@ SDValue DAGCombiner::visitSRL(SDNode *N) {
// fold (srl (ctlz x), "5") -> x iff x has one bit set (the low bit).
if (N1C && N0.getOpcode() == ISD::CTLZ &&
N1C->getAPIntValue() == Log2_32(OpSizeInBits)) {
- APInt KnownZero, KnownOne;
- DAG.computeKnownBits(N0.getOperand(0), KnownZero, KnownOne);
+ KnownBits Known;
+ DAG.computeKnownBits(N0.getOperand(0), Known);
// If any of the input bits are KnownOne, then the input couldn't be all
// zeros, thus the result of the srl will always be zero.
- if (KnownOne.getBoolValue()) return DAG.getConstant(0, SDLoc(N0), VT);
+ if (Known.One.getBoolValue()) return DAG.getConstant(0, SDLoc(N0), VT);
// If all of the bits input the to ctlz node are known to be zero, then
// the result of the ctlz is "32" and the result of the shift is one.
- APInt UnknownBits = ~KnownZero;
+ APInt UnknownBits = ~Known.Zero;
if (UnknownBits == 0) return DAG.getConstant(1, SDLoc(N0), VT);
// Otherwise, check to see if there is exactly one bit input to the ctlz.
- if ((UnknownBits & (UnknownBits - 1)) == 0) {
+ if (UnknownBits.isPowerOf2()) {
// Okay, we know that only that the single bit specified by UnknownBits
// could be set on input to the CTLZ node. If this bit is set, the SRL
// will return 0, if it is clear, it returns 1. Change the CTLZ/SRL pair
@@ -6889,6 +6973,51 @@ SDValue DAGCombiner::CombineExtLoad(SDNode *N) {
return SDValue(N, 0); // Return N so it doesn't get rechecked!
}
+/// If we're narrowing or widening the result of a vector select and the final
+/// size is the same size as a setcc (compare) feeding the select, then try to
+/// apply the cast operation to the select's operands because matching vector
+/// sizes for a select condition and other operands should be more efficient.
+SDValue DAGCombiner::matchVSelectOpSizesWithSetCC(SDNode *Cast) {
+ unsigned CastOpcode = Cast->getOpcode();
+ assert((CastOpcode == ISD::SIGN_EXTEND || CastOpcode == ISD::ZERO_EXTEND ||
+ CastOpcode == ISD::TRUNCATE || CastOpcode == ISD::FP_EXTEND ||
+ CastOpcode == ISD::FP_ROUND) &&
+ "Unexpected opcode for vector select narrowing/widening");
+
+ // We only do this transform before legal ops because the pattern may be
+ // obfuscated by target-specific operations after legalization. Do not create
+ // an illegal select op, however, because that may be difficult to lower.
+ EVT VT = Cast->getValueType(0);
+ if (LegalOperations || !TLI.isOperationLegalOrCustom(ISD::VSELECT, VT))
+ return SDValue();
+
+ SDValue VSel = Cast->getOperand(0);
+ if (VSel.getOpcode() != ISD::VSELECT || !VSel.hasOneUse() ||
+ VSel.getOperand(0).getOpcode() != ISD::SETCC)
+ return SDValue();
+
+ // Does the setcc have the same vector size as the casted select?
+ SDValue SetCC = VSel.getOperand(0);
+ EVT SetCCVT = getSetCCResultType(SetCC.getOperand(0).getValueType());
+ if (SetCCVT.getSizeInBits() != VT.getSizeInBits())
+ return SDValue();
+
+ // cast (vsel (setcc X), A, B) --> vsel (setcc X), (cast A), (cast B)
+ SDValue A = VSel.getOperand(1);
+ SDValue B = VSel.getOperand(2);
+ SDValue CastA, CastB;
+ SDLoc DL(Cast);
+ if (CastOpcode == ISD::FP_ROUND) {
+ // FP_ROUND (fptrunc) has an extra flag operand to pass along.
+ CastA = DAG.getNode(CastOpcode, DL, VT, A, Cast->getOperand(1));
+ CastB = DAG.getNode(CastOpcode, DL, VT, B, Cast->getOperand(1));
+ } else {
+ CastA = DAG.getNode(CastOpcode, DL, VT, A);
+ CastB = DAG.getNode(CastOpcode, DL, VT, B);
+ }
+ return DAG.getNode(ISD::VSELECT, DL, VT, SetCC, CastA, CastB);
+}
+
SDValue DAGCombiner::visitSIGN_EXTEND(SDNode *N) {
SDValue N0 = N->getOperand(0);
EVT VT = N->getValueType(0);
@@ -7112,19 +7241,21 @@ SDValue DAGCombiner::visitSIGN_EXTEND(SDNode *N) {
DAG.SignBitIsZero(N0))
return DAG.getNode(ISD::ZERO_EXTEND, DL, VT, N0);
+ if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
+ return NewVSel;
+
return SDValue();
}
// isTruncateOf - If N is a truncate of some other value, return true, record
-// the value being truncated in Op and which of Op's bits are zero in KnownZero.
-// This function computes KnownZero to avoid a duplicated call to
+// the value being truncated in Op and which of Op's bits are zero/one in Known.
+// This function computes KnownBits to avoid a duplicated call to
// computeKnownBits in the caller.
static bool isTruncateOf(SelectionDAG &DAG, SDValue N, SDValue &Op,
- APInt &KnownZero) {
- APInt KnownOne;
+ KnownBits &Known) {
if (N->getOpcode() == ISD::TRUNCATE) {
Op = N->getOperand(0);
- DAG.computeKnownBits(Op, KnownZero, KnownOne);
+ DAG.computeKnownBits(Op, Known);
return true;
}
@@ -7143,9 +7274,9 @@ static bool isTruncateOf(SelectionDAG &DAG, SDValue N, SDValue &Op,
else
return false;
- DAG.computeKnownBits(Op, KnownZero, KnownOne);
+ DAG.computeKnownBits(Op, Known);
- if (!(KnownZero | APInt(Op.getValueSizeInBits(), 1)).isAllOnesValue())
+ if (!(Known.Zero | 1).isAllOnesValue())
return false;
return true;
@@ -7170,8 +7301,8 @@ SDValue DAGCombiner::visitZERO_EXTEND(SDNode *N) {
// This is valid when the truncated bits of x are already zero.
// FIXME: We should extend this to work for vectors too.
SDValue Op;
- APInt KnownZero;
- if (!VT.isVector() && isTruncateOf(DAG, N0, Op, KnownZero)) {
+ KnownBits Known;
+ if (!VT.isVector() && isTruncateOf(DAG, N0, Op, Known)) {
APInt TruncatedBits =
(Op.getValueSizeInBits() == N0.getValueSizeInBits()) ?
APInt(Op.getValueSizeInBits(), 0) :
@@ -7179,7 +7310,7 @@ SDValue DAGCombiner::visitZERO_EXTEND(SDNode *N) {
N0.getValueSizeInBits(),
std::min(Op.getValueSizeInBits(),
VT.getSizeInBits()));
- if (TruncatedBits == (KnownZero & TruncatedBits)) {
+ if (TruncatedBits.isSubsetOf(Known.Zero)) {
if (VT.bitsGT(Op.getValueType()))
return DAG.getNode(ISD::ZERO_EXTEND, SDLoc(N), VT, Op);
if (VT.bitsLT(Op.getValueType()))
@@ -7446,6 +7577,9 @@ SDValue DAGCombiner::visitZERO_EXTEND(SDNode *N) {
ShAmt);
}
+ if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
+ return NewVSel;
+
return SDValue();
}
@@ -7802,7 +7936,7 @@ SDValue DAGCombiner::ReduceLoadWidth(SDNode *N) {
SDValue NewPtr = DAG.getNode(ISD::ADD, DL,
PtrType, LN0->getBasePtr(),
DAG.getConstant(PtrOff, DL, PtrType),
- &Flags);
+ Flags);
AddToWorklist(NewPtr.getNode());
SDValue Load;
@@ -8228,17 +8362,21 @@ SDValue DAGCombiner::visitTRUNCATE(SDNode *N) {
return SDValue(N, 0);
// (trunc adde(X, Y, Carry)) -> (adde trunc(X), trunc(Y), Carry)
+ // (trunc addcarry(X, Y, Carry)) -> (addcarry trunc(X), trunc(Y), Carry)
// When the adde's carry is not used.
- if (N0.getOpcode() == ISD::ADDE && N0.hasOneUse() &&
- !N0.getNode()->hasAnyUseOfValue(1) &&
- (!LegalOperations || TLI.isOperationLegal(ISD::ADDE, VT))) {
+ if ((N0.getOpcode() == ISD::ADDE || N0.getOpcode() == ISD::ADDCARRY) &&
+ N0.hasOneUse() && !N0.getNode()->hasAnyUseOfValue(1) &&
+ (!LegalOperations || TLI.isOperationLegal(N0.getOpcode(), VT))) {
SDLoc SL(N);
auto X = DAG.getNode(ISD::TRUNCATE, SL, VT, N0.getOperand(0));
auto Y = DAG.getNode(ISD::TRUNCATE, SL, VT, N0.getOperand(1));
- return DAG.getNode(ISD::ADDE, SL, DAG.getVTList(VT, MVT::Glue),
- X, Y, N0.getOperand(2));
+ auto VTs = DAG.getVTList(VT, N0->getValueType(1));
+ return DAG.getNode(N0.getOpcode(), SL, VTs, X, Y, N0.getOperand(2));
}
+ if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
+ return NewVSel;
+
return SDValue();
}
@@ -8701,7 +8839,7 @@ ConstantFoldBITCASTofBUILD_VECTOR(SDNode *BV, EVT DstEltVT) {
}
static bool isContractable(SDNode *N) {
- SDNodeFlags F = cast<BinaryWithFlagsSDNode>(N)->Flags;
+ SDNodeFlags F = N->getFlags();
return F.hasAllowContract() || F.hasUnsafeAlgebra();
}
@@ -9287,7 +9425,7 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
EVT VT = N->getValueType(0);
SDLoc DL(N);
const TargetOptions &Options = DAG.getTarget().Options;
- const SDNodeFlags *Flags = &cast<BinaryWithFlagsSDNode>(N)->Flags;
+ const SDNodeFlags Flags = N->getFlags();
// fold vector ops
if (VT.isVector())
@@ -9318,7 +9456,7 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
GetNegatedExpression(N0, DAG, LegalOperations), Flags);
// FIXME: Auto-upgrade the target/function-level option.
- if (Options.NoSignedZerosFPMath || N->getFlags()->hasNoSignedZeros()) {
+ if (Options.NoSignedZerosFPMath || N->getFlags().hasNoSignedZeros()) {
// fold (fadd A, 0) -> A
if (ConstantFPSDNode *N1C = isConstOrConstSplatFP(N1))
if (N1C->isZero())
@@ -9441,7 +9579,7 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
EVT VT = N->getValueType(0);
SDLoc DL(N);
const TargetOptions &Options = DAG.getTarget().Options;
- const SDNodeFlags *Flags = &cast<BinaryWithFlagsSDNode>(N)->Flags;
+ const SDNodeFlags Flags = N->getFlags();
// fold vector ops
if (VT.isVector())
@@ -9461,7 +9599,7 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
GetNegatedExpression(N1, DAG, LegalOperations), Flags);
// FIXME: Auto-upgrade the target/function-level option.
- if (Options.NoSignedZerosFPMath || N->getFlags()->hasNoSignedZeros()) {
+ if (Options.NoSignedZerosFPMath || N->getFlags().hasNoSignedZeros()) {
// (fsub 0, B) -> -B
if (N0CFP && N0CFP->isZero()) {
if (isNegatibleForFree(N1, LegalOperations, TLI, &Options))
@@ -9512,7 +9650,7 @@ SDValue DAGCombiner::visitFMUL(SDNode *N) {
EVT VT = N->getValueType(0);
SDLoc DL(N);
const TargetOptions &Options = DAG.getTarget().Options;
- const SDNodeFlags *Flags = &cast<BinaryWithFlagsSDNode>(N)->Flags;
+ const SDNodeFlags Flags = N->getFlags();
// fold vector ops
if (VT.isVector()) {
@@ -9656,7 +9794,7 @@ SDValue DAGCombiner::visitFMA(SDNode *N) {
isConstantFPBuildVectorOrConstantFP(N2.getOperand(1))) {
return DAG.getNode(ISD::FMUL, DL, VT, N0,
DAG.getNode(ISD::FADD, DL, VT, N1, N2.getOperand(1),
- &Flags), &Flags);
+ Flags), Flags);
}
// (fma (fmul x, c1), c2, y) -> (fma x, c1*c2, y)
@@ -9666,7 +9804,7 @@ SDValue DAGCombiner::visitFMA(SDNode *N) {
return DAG.getNode(ISD::FMA, DL, VT,
N0.getOperand(0),
DAG.getNode(ISD::FMUL, DL, VT, N1, N0.getOperand(1),
- &Flags),
+ Flags),
N2);
}
}
@@ -9692,16 +9830,16 @@ SDValue DAGCombiner::visitFMA(SDNode *N) {
if (N1CFP && N0 == N2) {
return DAG.getNode(ISD::FMUL, DL, VT, N0,
DAG.getNode(ISD::FADD, DL, VT, N1,
- DAG.getConstantFP(1.0, DL, VT), &Flags),
- &Flags);
+ DAG.getConstantFP(1.0, DL, VT), Flags),
+ Flags);
}
// (fma x, c, (fneg x)) -> (fmul x, (c-1))
if (N1CFP && N2.getOpcode() == ISD::FNEG && N2.getOperand(0) == N0) {
return DAG.getNode(ISD::FMUL, DL, VT, N0,
DAG.getNode(ISD::FADD, DL, VT, N1,
- DAG.getConstantFP(-1.0, DL, VT), &Flags),
- &Flags);
+ DAG.getConstantFP(-1.0, DL, VT), Flags),
+ Flags);
}
}
@@ -9717,8 +9855,8 @@ SDValue DAGCombiner::visitFMA(SDNode *N) {
// is the critical path is increased from "one FDIV" to "one FDIV + one FMUL".
SDValue DAGCombiner::combineRepeatedFPDivisors(SDNode *N) {
bool UnsafeMath = DAG.getTarget().Options.UnsafeFPMath;
- const SDNodeFlags *Flags = N->getFlags();
- if (!UnsafeMath && !Flags->hasAllowReciprocal())
+ const SDNodeFlags Flags = N->getFlags();
+ if (!UnsafeMath && !Flags.hasAllowReciprocal())
return SDValue();
// Skip if current node is a reciprocal.
@@ -9741,7 +9879,7 @@ SDValue DAGCombiner::combineRepeatedFPDivisors(SDNode *N) {
if (U->getOpcode() == ISD::FDIV && U->getOperand(1) == N1) {
// This division is eligible for optimization only if global unsafe math
// is enabled or if this division allows reciprocal formation.
- if (UnsafeMath || U->getFlags()->hasAllowReciprocal())
+ if (UnsafeMath || U->getFlags().hasAllowReciprocal())
Users.insert(U);
}
}
@@ -9780,7 +9918,7 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
EVT VT = N->getValueType(0);
SDLoc DL(N);
const TargetOptions &Options = DAG.getTarget().Options;
- SDNodeFlags *Flags = &cast<BinaryWithFlagsSDNode>(N)->Flags;
+ SDNodeFlags Flags = N->getFlags();
// fold vector ops
if (VT.isVector())
@@ -9894,8 +10032,7 @@ SDValue DAGCombiner::visitFREM(SDNode *N) {
// fold (frem c1, c2) -> fmod(c1,c2)
if (N0CFP && N1CFP)
- return DAG.getNode(ISD::FREM, SDLoc(N), VT, N0, N1,
- &cast<BinaryWithFlagsSDNode>(N)->Flags);
+ return DAG.getNode(ISD::FREM, SDLoc(N), VT, N0, N1, N->getFlags());
if (SDValue NewSel = foldBinOpIntoSelect(N))
return NewSel;
@@ -9915,7 +10052,7 @@ SDValue DAGCombiner::visitFSQRT(SDNode *N) {
// For now, create a Flags object for use with all unsafe math transforms.
SDNodeFlags Flags;
Flags.setUnsafeAlgebra(true);
- return buildSqrtEstimate(N0, &Flags);
+ return buildSqrtEstimate(N0, Flags);
}
/// copysign(x, fp_extend(y)) -> copysign(x, y)
@@ -10190,6 +10327,9 @@ SDValue DAGCombiner::visitFP_ROUND(SDNode *N) {
Tmp, N0.getOperand(1));
}
+ if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
+ return NewVSel;
+
return SDValue();
}
@@ -10256,6 +10396,9 @@ SDValue DAGCombiner::visitFP_EXTEND(SDNode *N) {
return SDValue(N, 0); // Return N so it doesn't get rechecked!
}
+ if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
+ return NewVSel;
+
return SDValue();
}
@@ -10341,10 +10484,10 @@ SDValue DAGCombiner::visitFNEG(SDNode *N) {
if (Level >= AfterLegalizeDAG &&
(TLI.isFPImmLegal(CVal, VT) ||
TLI.isOperationLegal(ISD::ConstantFP, VT)))
- return DAG.getNode(ISD::FMUL, SDLoc(N), VT, N0.getOperand(0),
- DAG.getNode(ISD::FNEG, SDLoc(N), VT,
- N0.getOperand(1)),
- &cast<BinaryWithFlagsSDNode>(N0)->Flags);
+ return DAG.getNode(
+ ISD::FMUL, SDLoc(N), VT, N0.getOperand(0),
+ DAG.getNode(ISD::FNEG, SDLoc(N), VT, N0.getOperand(1)),
+ N0->getFlags());
}
}
@@ -15832,7 +15975,7 @@ SDValue DAGCombiner::BuildLogBase2(SDValue V, const SDLoc &DL) {
/// =>
/// X_{i+1} = X_i (2 - A X_i) = X_i + X_i (1 - A X_i) [this second form
/// does not require additional intermediate precision]
-SDValue DAGCombiner::BuildReciprocalEstimate(SDValue Op, SDNodeFlags *Flags) {
+SDValue DAGCombiner::BuildReciprocalEstimate(SDValue Op, SDNodeFlags Flags) {
if (Level >= AfterLegalizeDAG)
return SDValue();
@@ -15887,7 +16030,7 @@ SDValue DAGCombiner::BuildReciprocalEstimate(SDValue Op, SDNodeFlags *Flags) {
/// As a result, we precompute A/2 prior to the iteration loop.
SDValue DAGCombiner::buildSqrtNROneConst(SDValue Arg, SDValue Est,
unsigned Iterations,
- SDNodeFlags *Flags, bool Reciprocal) {
+ SDNodeFlags Flags, bool Reciprocal) {
EVT VT = Arg.getValueType();
SDLoc DL(Arg);
SDValue ThreeHalves = DAG.getConstantFP(1.5, DL, VT);
@@ -15931,7 +16074,7 @@ SDValue DAGCombiner::buildSqrtNROneConst(SDValue Arg, SDValue Est,
/// X_{i+1} = (-0.5 * X_i) * (A * X_i * X_i + (-3.0))
SDValue DAGCombiner::buildSqrtNRTwoConst(SDValue Arg, SDValue Est,
unsigned Iterations,
- SDNodeFlags *Flags, bool Reciprocal) {
+ SDNodeFlags Flags, bool Reciprocal) {
EVT VT = Arg.getValueType();
SDLoc DL(Arg);
SDValue MinusThree = DAG.getConstantFP(-3.0, DL, VT);
@@ -15976,7 +16119,7 @@ SDValue DAGCombiner::buildSqrtNRTwoConst(SDValue Arg, SDValue Est,
/// Build code to calculate either rsqrt(Op) or sqrt(Op). In the latter case
/// Op*rsqrt(Op) is actually computed, so additional postprocessing is needed if
/// Op can be zero.
-SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, SDNodeFlags *Flags,
+SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, SDNodeFlags Flags,
bool Reciprocal) {
if (Level >= AfterLegalizeDAG)
return SDValue();
@@ -16029,11 +16172,11 @@ SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, SDNodeFlags *Flags,
return SDValue();
}
-SDValue DAGCombiner::buildRsqrtEstimate(SDValue Op, SDNodeFlags *Flags) {
+SDValue DAGCombiner::buildRsqrtEstimate(SDValue Op, SDNodeFlags Flags) {
return buildSqrtEstimateImpl(Op, Flags, true);
}
-SDValue DAGCombiner::buildSqrtEstimate(SDValue Op, SDNodeFlags *Flags) {
+SDValue DAGCombiner::buildSqrtEstimate(SDValue Op, SDNodeFlags Flags) {
return buildSqrtEstimateImpl(Op, Flags, false);
}