From 1d4d41411190dd9e62764e56713753d4155764dd Mon Sep 17 00:00:00 2001 From: Nate Begeman Date: Thu, 1 Sep 2005 00:19:25 +0000 Subject: [PATCH] First pass at the DAG Combiner. It isn't used anywhere yet, but it should be mostly functional. It currently has all folds from SelectionDAG.cpp that do not involve a condition code. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23184 91177308-0d34-0410-b5e6-96231b3b80d8 --- include/llvm/CodeGen/SelectionDAG.h | 6 + lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 1056 ++++++++++++++++++++++ 2 files changed, 1062 insertions(+) create mode 100644 lib/CodeGen/SelectionDAG/DAGCombiner.cpp diff --git a/include/llvm/CodeGen/SelectionDAG.h b/include/llvm/CodeGen/SelectionDAG.h index ed27fb11054..71bc2672e1a 100644 --- a/include/llvm/CodeGen/SelectionDAG.h +++ b/include/llvm/CodeGen/SelectionDAG.h @@ -79,6 +79,12 @@ public: /// const SDOperand &setRoot(SDOperand N) { return Root = N; } + /// Combine - This iterates over the nodes in the SelectionDAG, folding + /// certain types of nodes together, or eliminating superfluous nodes. When + /// the AfterLegalize argument is set to 'true', Combine takes care not to + /// generate any nodes that will be illegal on the target. + void Combine(bool AfterLegalize); + /// Legalize - This transforms the SelectionDAG into a SelectionDAG that is /// compatible with the target instruction selector, as indicated by the /// TargetLowering object. diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp new file mode 100644 index 00000000000..fc2c8f2e8ea --- /dev/null +++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp @@ -0,0 +1,1056 @@ +//===-- DAGCombiner.cpp - Implement a trivial DAG combiner ----------------===// +// +// The LLVM Compiler Infrastructure +// +// This file was developed by Nate Begeman and is distributed under the +// University of Illinois Open Source License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This pass combines dag nodes to form fewer, simpler DAG nodes. It can be run +// both before and after the DAG is legalized. +// +// FIXME: Missing folds +// sdiv, udiv, srem, urem (X, const) where X is an integer can be expanded into +// a sequence of multiplies, shifts, and adds. This should be controlled by +// some kind of hint from the target that int div is expensive. +// various folds of mulh[s,u] by constants such as -1, powers of 2, etc. +// +// FIXME: Should add a corresponding version of fold AND with +// ZERO_EXTEND/SIGN_EXTEND by converting them to an ANY_EXTEND node which +// we don't have yet. +// +// FIXME: mul (x, const) -> shifts + adds +// +// FIXME: undef values +// +// FIXME: zero extend when top bits are 0 -> drop it ? +// +//===----------------------------------------------------------------------===// + +#define DEBUG_TYPE "dagcombine" +#include "llvm/ADT/Statistic.h" +#include "llvm/CodeGen/SelectionDAG.h" +#include "llvm/Support/MathExtras.h" +#include "llvm/Target/TargetLowering.h" +#include +using namespace llvm; + +namespace { + Statistic<> NodesCombined ("dagcombiner", "Number of dag nodes combined"); + + class DAGCombiner { + SelectionDAG &DAG; + TargetLowering &TLI; + + // Worklist of all of the nodes that need to be simplified. + std::vector WorkList; + + /// AddUsersToWorkList - When an instruction is simplified, add all users of + /// the instruction to the work lists because they might get more simplified + /// now. + /// + void AddUsersToWorkList(SDNode *N) { + for (SDNode::use_iterator UI = N->use_begin(), UE = N->use_end(); + UI != UE; ++UI) { + SDNode *U = *UI; + for (unsigned i = 0, e = U->getNumOperands(); i != e; ++i) + WorkList.push_back(U->getOperand(i).Val); + } + } + + /// removeFromWorkList - remove all instances of N from the worklist. + void removeFromWorkList(SDNode *N) { + WorkList.erase(std::remove(WorkList.begin(), WorkList.end(), N), + WorkList.end()); + } + + /// visit - call the node-specific routine that knows how to fold each + /// particular type of node. + SDNode *visit(SDNode *N); + + // Visitation implementation - Implement dag node combining for different + // node types. The semantics are as follows: + // Return Value: + // null - No change was made + // otherwise - Node N should be replaced by the returned node. + // + SDNode *visitTokenFactor(SDNode *N); + SDNode *visitAdd(SDNode *N); + SDNode *visitSub(SDNode *N); + SDNode *visitMul(SDNode *N); + SDNode *visitSdiv(SDNode *N); + SDNode *visitUdiv(SDNode *N); + SDNode *visitSrem(SDNode *N); + SDNode *visitUrem(SDNode *N); + SDNode *visitMulHiU(SDNode *N); + SDNode *visitMulHiS(SDNode *N); + SDNode *visitAnd(SDNode *N); + SDNode *visitOr(SDNode *N); + SDNode *visitXor(SDNode *N); + SDNode *visitShl(SDNode *N); + SDNode *visitSra(SDNode *N); + SDNode *visitSrl(SDNode *N); + SDNode *visitCtlz(SDNode *N); + SDNode *visitCttz(SDNode *N); + SDNode *visitCtpop(SDNode *N); + // select + // select_cc + // setcc + SDNode *visitSignExtend(SDNode *N); + SDNode *visitZeroExtend(SDNode *N); + SDNode *visitSignExtendInReg(SDNode *N); + SDNode *visitTruncate(SDNode *N); + SDNode *visitSintToFP(SDNode *N); + SDNode *visitUintToFP(SDNode *N); + SDNode *visitFPToSint(SDNode *N); + SDNode *visitFPToUint(SDNode *N); + SDNode *visitFPRound(SDNode *N); + SDNode *visitFPRoundInReg(SDNode *N); + SDNode *visitFPExtend(SDNode *N); + SDNode *visitFneg(SDNode *N); + SDNode *visitFabs(SDNode *N); + SDNode *visitExtLoad(SDNode *N); + SDNode *visitSextLoad(SDNode *N); + SDNode *visitZextLoad(SDNode *N); + SDNode *visitTruncStore(SDNode *N); + // brcond + // brcondtwoway + // br_cc + // brtwoway_cc +public: + DAGCombiner(SelectionDAG &D) + : DAG(D), TLI(D.getTargetLoweringInfo()) { + // Add all the dag nodes to the worklist. + WorkList.insert(WorkList.end(), D.allnodes_begin(), D.allnodes_end()); + } + + /// Run - runs the dag combiner on all nodes in the work list + void Run(bool AfterLegalize); + }; +} + +/// MaskedValueIsZero - Return true if 'V & Mask' is known to be zero. We use +/// this predicate to simplify operations downstream. V and Mask are known to +/// be the same type. +static bool MaskedValueIsZero(const SDOperand &Op, uint64_t Mask, + const TargetLowering &TLI) { + unsigned SrcBits; + if (Mask == 0) return true; + + // If we know the result of a setcc has the top bits zero, use this info. + switch (Op.getOpcode()) { + case ISD::Constant: + return (cast(Op)->getValue() & Mask) == 0; + + case ISD::SETCC: + return ((Mask & 1) == 0) && + TLI.getSetCCResultContents() == TargetLowering::ZeroOrOneSetCCResult; + + case ISD::ZEXTLOAD: + SrcBits = MVT::getSizeInBits(cast(Op.getOperand(3))->getVT()); + return (Mask & ((1ULL << SrcBits)-1)) == 0; // Returning only the zext bits. + case ISD::ZERO_EXTEND: + case ISD::AssertZext: + SrcBits = MVT::getSizeInBits(Op.getOperand(0).getValueType()); + return MaskedValueIsZero(Op.getOperand(0),Mask & ((1ULL << SrcBits)-1),TLI); + + case ISD::AND: + // (X & C1) & C2 == 0 iff C1 & C2 == 0. + if (ConstantSDNode *AndRHS = dyn_cast(Op.getOperand(1))) + return MaskedValueIsZero(Op.getOperand(0),AndRHS->getValue() & Mask, TLI); + + // FALL THROUGH + case ISD::OR: + case ISD::XOR: + return MaskedValueIsZero(Op.getOperand(0), Mask, TLI) && + MaskedValueIsZero(Op.getOperand(1), Mask, TLI); + case ISD::SELECT: + return MaskedValueIsZero(Op.getOperand(1), Mask, TLI) && + MaskedValueIsZero(Op.getOperand(2), Mask, TLI); + case ISD::SELECT_CC: + return MaskedValueIsZero(Op.getOperand(2), Mask, TLI) && + MaskedValueIsZero(Op.getOperand(3), Mask, TLI); + case ISD::SRL: + // (ushr X, C1) & C2 == 0 iff X & (C2 << C1) == 0 + if (ConstantSDNode *ShAmt = dyn_cast(Op.getOperand(1))) { + uint64_t NewVal = Mask << ShAmt->getValue(); + SrcBits = MVT::getSizeInBits(Op.getValueType()); + if (SrcBits != 64) NewVal &= (1ULL << SrcBits)-1; + return MaskedValueIsZero(Op.getOperand(0), NewVal, TLI); + } + return false; + case ISD::SHL: + // (ushl X, C1) & C2 == 0 iff X & (C2 >> C1) == 0 + if (ConstantSDNode *ShAmt = dyn_cast(Op.getOperand(1))) { + uint64_t NewVal = Mask >> ShAmt->getValue(); + return MaskedValueIsZero(Op.getOperand(0), NewVal, TLI); + } + return false; + case ISD::CTTZ: + case ISD::CTLZ: + case ISD::CTPOP: + // Bit counting instructions can not set the high bits of the result + // register. The max number of bits sets depends on the input. + return (Mask & (MVT::getSizeInBits(Op.getValueType())*2-1)) == 0; + + // TODO we could handle some SRA cases here. + default: break; + } + + return false; +} + +// isInvertibleForFree - Return true if there is no cost to emitting the logical +// inverse of this node. +static bool isInvertibleForFree(SDOperand N) { + if (isa(N.Val)) return true; + if (N.Val->getOpcode() == ISD::SETCC && N.Val->hasOneUse()) + return true; + return false; +} + +// isSetCCEquivalent - Return true if this node is a select_cc that selects +// between the values 1 and 0, making it equivalent to a setcc. +static bool isSetCCEquivalent(SDOperand N) { + if (N.getOpcode() == ISD::SELECT_CC && + N.getOperand(2).getOpcode() == ISD::Constant && + N.getOperand(3).getOpcode() == ISD::Constant && + cast(N.getOperand(2))->getValue() == 1 && + cast(N.getOperand(3))->isNullValue()) + return true; + return false; +} + +void DAGCombiner::Run(bool AfterLegalize) { + // while the worklist isn't empty, inspect the node on the end of it and + // try and combine it. + while (!WorkList.empty()) { + SDNode *N = WorkList.back(); + WorkList.pop_back(); + + // If N has no uses, it is dead. Make sure to revisit all N's operands once + // N is deleted from the DAG, since they too may now be dead. + if (N->use_empty()) { + for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) + WorkList.push_back(N->getOperand(i).Val); + + DAG.DeleteNode(N); + removeFromWorkList(N); + continue; + } + + if (SDNode *Result = visit(N)) { + ++NodesCombined; + assert(Result != N && "Modifying DAG nodes in place is illegal!"); + + std::cerr << "DC: Old = "; N->dump(); + std::cerr << " New = "; Result->dump(); + std::cerr << '\n'; + DAG.ReplaceAllUsesWith(N, Result); + + // Push the new node and any users onto the worklist + WorkList.push_back(Result); + AddUsersToWorkList(Result); + + // Nodes can end up on the worklist more than once. Make sure we do + // not process a node that has been replaced. + removeFromWorkList(N); + } + } +} + +SDNode *DAGCombiner::visit(SDNode *N) { + switch(N->getOpcode()) { + default: break; + case ISD::TokenFactor: + return visitTokenFactor(N); + case ISD::ADD: + return visitAdd(N); + case ISD::SUB: + return visitSub(N); + case ISD::MUL: + return visitMul(N); + case ISD::SDIV: + return visitSdiv(N); + case ISD::UDIV: + return visitUdiv(N); + case ISD::SREM: + return visitSrem(N); + case ISD::UREM: + return visitUrem(N); + case ISD::SIGN_EXTEND: + return visitSignExtend(N); + case ISD::ZERO_EXTEND: + return visitZeroExtend(N); + case ISD::FNEG: + return visitFneg(N); + } + return 0; +} + +SDNode *DAGCombiner::visitTokenFactor(SDNode *N) { + // If the token factor only has one operand, fold TF(x) -> x + if (N->getNumOperands() == 1) + return N->getOperand(0).Val; + + // If the token factor has two operands and one is the entry token, replace + // the token factor with the other operand. + if (N->getNumOperands() == 2) { + if (N->getOperand(0).getOpcode() == ISD::EntryToken) + return N->getOperand(1).Val; + if (N->getOperand(1).getOpcode() == ISD::EntryToken) + return N->getOperand(0).Val; + } + return 0; +} + +SDNode *DAGCombiner::visitAdd(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + ConstantFPSDNode *N1CFP = dyn_cast(N0.Val); + ConstantFPSDNode *N2CFP = dyn_cast(N1.Val); + + // fold (add c1, c2) -> c1+c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() + N2C->getValue(), + N->getValueType(0)).Val; + // fold (add x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // fold floating point (add c1, c2) -> c1+c2 + if (N1CFP && N2CFP) + return DAG.getConstantFP(N1CFP->getValue() + N2CFP->getValue(), + N->getValueType(0)).Val; + // fold (A + (-B)) -> A-B + if (N1.getOpcode() == ISD::FNEG) + return DAG.getNode(ISD::SUB, N->getValueType(0), N0, N1.getOperand(0)).Val; + // fold ((-A) + B) -> B-A + if (N0.getOpcode() == ISD::FNEG) + return DAG.getNode(ISD::SUB, N->getValueType(0), N1, N0.getOperand(0)).Val; + // fold ((0-A) + B) -> B-A + if (N0.getOpcode() == ISD::SUB && isa(N0.getOperand(0)) && + cast(N0.getOperand(0))->isNullValue()) + return DAG.getNode(ISD::SUB, N->getValueType(0), N1, N0.getOperand(1)).Val; + // fold (A + (0-B)) -> A-B + if (N1.getOpcode() == ISD::SUB && isa(N1.getOperand(0)) && + cast(N1.getOperand(0))->isNullValue()) + return DAG.getNode(ISD::SUB, N->getValueType(0), N0, N1.getOperand(1)).Val; + // fold (A+(B-A)) -> B for non-fp types + if (N1.getOpcode() == ISD::SUB && N0 == N1.getOperand(1) && + !MVT::isFloatingPoint(N1.getValueType())) + return N1.getOperand(0).Val; + return 0; +} + +SDNode *DAGCombiner::visitSub(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + ConstantFPSDNode *N1CFP = dyn_cast(N0.Val); + ConstantFPSDNode *N2CFP = dyn_cast(N1.Val); + + // fold (sub c1, c2) -> c1-c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() - N2C->getValue(), + N->getValueType(0)).Val; + // fold (sub x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // fold floating point (sub c1, c2) -> c1-c2 + if (N1CFP && N2CFP) + return DAG.getConstantFP(N1CFP->getValue() - N2CFP->getValue(), + N->getValueType(0)).Val; + // fold (A+B)-A -> B + if (N0.getOpcode() == ISD::ADD && N0.getOperand(0) == N1 && + !MVT::isFloatingPoint(N1.getValueType())) + return N0.getOperand(1).Val; + // fold (A+B)-B -> A + if (N0.getOpcode() == ISD::ADD && N0.getOperand(1) == N1 && + !MVT::isFloatingPoint(N1.getValueType())) + return N0.getOperand(0).Val; + // fold (A-(-B)) -> A+B + if (N1.getOpcode() == ISD::FNEG) + return DAG.getNode(ISD::ADD, N0.getValueType(), N0, N1.getOperand(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitMul(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + ConstantFPSDNode *N1CFP = dyn_cast(N0.Val); + ConstantFPSDNode *N2CFP = dyn_cast(N1.Val); + + // fold (mul c1, c2) -> c1*c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() * N2C->getValue(), + N->getValueType(0)).Val; + // fold (mul x, 0) -> 0 + if (N2C && N2C->isNullValue()) + return N1.Val; + // fold (mul x, -1) -> 0-x + if (N2C && N2C->isAllOnesValue()) + return DAG.getNode(ISD::SUB, N->getValueType(0), + DAG.getConstant(0, N->getValueType(0)), N0).Val; + // fold (mul x, (1 << c)) -> x << c + if (N2C && isPowerOf2_64(N2C->getValue())) + return DAG.getNode(ISD::SHL, N->getValueType(0), N0, + DAG.getConstant(Log2_64(N2C->getValue()), + TLI.getShiftAmountTy())).Val; + // fold floating point (mul c1, c2) -> c1*c2 + if (N1CFP && N2CFP) + return DAG.getConstantFP(N1CFP->getValue() * N2CFP->getValue(), + N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitSdiv(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + ConstantFPSDNode *N1CFP = dyn_cast(N0.Val); + ConstantFPSDNode *N2CFP = dyn_cast(N1.Val); + + // fold (sdiv c1, c2) -> c1/c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getSignExtended() / N2C->getSignExtended(), + N->getValueType(0)).Val; + // fold floating point (sdiv c1, c2) -> c1/c2 + if (N1CFP && N2CFP) + return DAG.getConstantFP(N1CFP->getValue() / N2CFP->getValue(), + N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitUdiv(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + + // fold (udiv c1, c2) -> c1/c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() / N2C->getValue(), + N->getValueType(0)).Val; + // fold (udiv x, (1 << c)) -> x >>u c + if (N2C && isPowerOf2_64(N2C->getValue())) + return DAG.getNode(ISD::SRL, N->getValueType(0), N0, + DAG.getConstant(Log2_64(N2C->getValue()), + TLI.getShiftAmountTy())).Val; + return 0; +} + +SDNode *DAGCombiner::visitSrem(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + ConstantFPSDNode *N1CFP = dyn_cast(N0.Val); + ConstantFPSDNode *N2CFP = dyn_cast(N1.Val); + + // fold (srem c1, c2) -> c1%c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getSignExtended() % N2C->getSignExtended(), + N->getValueType(0)).Val; + // fold floating point (srem c1, c2) -> fmod(c1, c2) + if (N1CFP && N2CFP) + return DAG.getConstantFP(fmod(N1CFP->getValue(),N2CFP->getValue()), + N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitUrem(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + + // fold (urem c1, c2) -> c1%c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() % N2C->getValue(), + N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitMulHiS(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N2C = dyn_cast(N1.Val); + + // fold (mulhs x, 0) -> 0 + if (N2C && N2C->isNullValue()) + return N1.Val; + + // fold (mulhs x, 1) -> (sra x, size(x)-1) + if (N2C && N2C->getValue() == 1) + return DAG.getNode(ISD::SRA, N0.getValueType(), N0, + DAG.getConstant(MVT::getSizeInBits(N0.getValueType())-1, + TLI.getShiftAmountTy())).Val; + return 0; +} + +SDNode *DAGCombiner::visitMulHiU(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N2C = dyn_cast(N1.Val); + + // fold (mulhu x, 0) -> 0 + if (N2C && N2C->isNullValue()) + return N1.Val; + + // fold (mulhu x, 1) -> 0 + if (N2C && N2C->getValue() == 1) + return DAG.getConstant(0, N0.getValueType()).Val; + return 0; +} + +SDNode *DAGCombiner::visitAnd(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + MVT::ValueType VT = N1.getValueType(); + + // fold (and c1, c2) -> c1&c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() & N2C->getValue(), VT).Val; + // fold (and x, 0) -> 0 + if (N2C && N2C->isNullValue()) + return N1.Val; + // fold (and x, -1) -> x + if (N2C && N2C->isAllOnesValue()) + return N0.Val; + // fold (and x, 0) -> 0 + if (MaskedValueIsZero(N0, N2C->getValue(), TLI)) + return DAG.getConstant(0, VT).Val; + // fold (and x, mask containing x) -> x + uint64_t NotC2 = ~N2C->getValue(); + if (MVT::i64 != VT) NotC2 &= (1ULL << MVT::getSizeInBits(VT))-1; + if (MaskedValueIsZero(N0, NotC2, TLI)) + return N0.Val; + // fold (and (sign_extend_inreg x, i16 to i32), 1) -> (and x, 1) + if (N0.getOpcode() == ISD::SIGN_EXTEND_INREG) { + unsigned ExtendBits = + MVT::getSizeInBits(cast(N0.getOperand(1))->getVT()); + if ((N2C->getValue() & (~0ULL << ExtendBits)) == 0) + return DAG.getNode(ISD::AND, VT, N0.getOperand(0), N1).Val; + } + // fold (and (or x, 0xFFFF), 0xFF) -> 0xFF + if (N0.getOpcode() == ISD::OR) + if (ConstantSDNode *ORI = dyn_cast(N0.getOperand(1))) + if ((ORI->getValue() & N2C->getValue()) == N2C->getValue()) + return N1.Val; + // fold (and (assert_zext x, i16), 0xFFFF) -> (assert_zext x, i16) + if (N0.getOpcode() == ISD::AssertZext) { + unsigned ExtendBits = + MVT::getSizeInBits(cast(N0.getOperand(1))->getVT()); + if (N2C->getValue() == (1ULL << ExtendBits)-1) + return N0.Val; + } + return 0; +} + +SDNode *DAGCombiner::visitOr(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + + // fold (or c1, c2) -> c1|c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() | N2C->getValue(), + N->getValueType(0)).Val; + // fold (or x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // fold (or x, -1) -> -1 + if (N2C && N2C->isAllOnesValue()) + return N1.Val; + return 0; +} + +SDNode *DAGCombiner::visitXor(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + MVT::ValueType VT = N0.getValueType(); + + // fold (xor c1, c2) -> c1^c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() ^ N2C->getValue(), VT).Val; + // fold (xor x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // fold !(x cc y) -> (x !cc y) + if (N2C && N2C->isAllOnesValue() && N0.getOpcode() == ISD::SETCC) { + bool isInt = MVT::isInteger(N0.getOperand(0).getValueType()); + ISD::CondCode CC = cast(N0.getOperand(2))->get(); + return DAG.getSetCC(VT, N0.getOperand(0), N0.getOperand(1), + ISD::getSetCCInverse(CC, isInt)).Val; + } + // fold !(x cc y) -> (x !cc y) + if (N2C && N2C->isAllOnesValue() && isSetCCEquivalent(N0)) { + bool isInt = MVT::isInteger(N0.getOperand(0).getValueType()); + ISD::CondCode CC = cast(N0.getOperand(4))->get(); + return DAG.getSelectCC(N0.getOperand(0), N0.getOperand(1), + N0.getOperand(2), N0.getOperand(3), + ISD::getSetCCInverse(CC, isInt)).Val; + } + // fold !(x or y) -> (!x and !y) iff x or y are freely invertible + if (N2C && N2C->isAllOnesValue() && N0.getOpcode() == ISD::OR) { + SDOperand LHS = N0.getOperand(0), RHS = N0.getOperand(1); + if (isInvertibleForFree(RHS) || isInvertibleForFree(LHS)) { + LHS = DAG.getNode(ISD::XOR, VT, LHS, N1); // RHS = ~LHS + RHS = DAG.getNode(ISD::XOR, VT, RHS, N1); // RHS = ~RHS + return DAG.getNode(ISD::AND, VT, LHS, RHS).Val; + } + } + // fold !(x and y) -> (!x or !y) iff x or y are freely invertible + if (N2C && N2C->isAllOnesValue() && N0.getOpcode() == ISD::AND) { + SDOperand LHS = N0.getOperand(0), RHS = N0.getOperand(1); + if (isInvertibleForFree(RHS) || isInvertibleForFree(LHS)) { + LHS = DAG.getNode(ISD::XOR, VT, LHS, N1); // RHS = ~LHS + RHS = DAG.getNode(ISD::XOR, VT, RHS, N1); // RHS = ~RHS + return DAG.getNode(ISD::OR, VT, LHS, RHS).Val; + } + } + return 0; +} + +SDNode *DAGCombiner::visitShl(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + MVT::ValueType VT = N0.getValueType(); + unsigned OpSizeInBits = MVT::getSizeInBits(VT); + + // fold (shl c1, c2) -> c1<getValue() << N2C->getValue(), VT).Val; + // fold (shl 0, x) -> 0 + if (N1C && N1C->isNullValue()) + return N0.Val; + // fold (shl x, c >= size(x)) -> undef + if (N2C && N2C->getValue() >= OpSizeInBits) + return DAG.getNode(ISD::UNDEF, VT).Val; + // fold (shl x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // if (shl x, c) is known to be zero, return 0 + if (N2C && MaskedValueIsZero(N0,(~0ULL >> (64-OpSizeInBits))>>N2C->getValue(), + TLI)) + return DAG.getConstant(0, VT).Val; + // fold (shl (shl x, c1), c2) -> 0 or (shl x, c1+c2) + if (N2C && N0.getOpcode() == ISD::SHL && + N0.getOperand(1).getOpcode() == ISD::Constant) { + uint64_t c1 = cast(N0.getOperand(1))->getValue(); + uint64_t c2 = N2C->getValue(); + if (c1 + c2 > OpSizeInBits) + return DAG.getConstant(0, VT).Val; + return DAG.getNode(ISD::SHL, VT, N0.getOperand(0), + DAG.getConstant(c1 + c2, N1.getValueType())).Val; + } + // fold (shl (srl x, c1), c2) -> (shl (and x, -1 << c1), c2-c1) or + // (srl (and x, -1 << c1), c1-c2) + if (N2C && N0.getOpcode() == ISD::SRL && + N0.getOperand(1).getOpcode() == ISD::Constant) { + uint64_t c1 = cast(N0.getOperand(1))->getValue(); + uint64_t c2 = N2C->getValue(); + SDOperand Mask = DAG.getNode(ISD::AND, VT, N0.getOperand(0), + DAG.getConstant(~0ULL << c1, VT)); + if (c2 > c1) + return DAG.getNode(ISD::SHL, VT, Mask, + DAG.getConstant(c2-c1, N1.getValueType())).Val; + else + return DAG.getNode(ISD::SRL, VT, Mask, + DAG.getConstant(c1-c2, N1.getValueType())).Val; + } + // fold (shl (sra x, c1), c1) -> (and x, -1 << c1) + if (N2C && N0.getOpcode() == ISD::SRA && + N0.getOperand(1).getOpcode() == ISD::Constant) { + uint64_t c1 = cast(N0.getOperand(1))->getValue(); + uint64_t c2 = N2C->getValue(); + if (c1 == c2) + return DAG.getNode(ISD::AND, VT, N0.getOperand(0), + DAG.getConstant(~0ULL << c1, VT)).Val; + } + return 0; +} + +SDNode *DAGCombiner::visitSra(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + MVT::ValueType VT = N0.getValueType(); + unsigned OpSizeInBits = MVT::getSizeInBits(VT); + + // fold (sra c1, c2) -> c1>>c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getSignExtended() >> N2C->getValue(), VT).Val; + // fold (sra 0, x) -> 0 + if (N1C && N1C->isNullValue()) + return N0.Val; + // fold (sra -1, x) -> -1 + if (N1C && N1C->isAllOnesValue()) + return N0.Val; + // fold (sra x, c >= size(x)) -> undef + if (N2C && N2C->getValue() >= OpSizeInBits) + return DAG.getNode(ISD::UNDEF, VT).Val; + // fold (sra x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // If the sign bit is known to be zero, switch this to a SRL. + if (N2C && MaskedValueIsZero(N0, (1ULL << (OpSizeInBits-1)), TLI)) + return DAG.getNode(ISD::SRL, VT, N0, N1).Val; + return 0; +} + +SDNode *DAGCombiner::visitSrl(SDNode *N) { + SDOperand N0 = N->getOperand(0); + SDOperand N1 = N->getOperand(1); + ConstantSDNode *N1C = dyn_cast(N0.Val); + ConstantSDNode *N2C = dyn_cast(N1.Val); + MVT::ValueType VT = N0.getValueType(); + unsigned OpSizeInBits = MVT::getSizeInBits(VT); + + // fold (srl c1, c2) -> c1 >>u c2 + if (N1C && N2C) + return DAG.getConstant(N1C->getValue() >> N2C->getValue(), VT).Val; + // fold (srl 0, x) -> 0 + if (N1C && N1C->isNullValue()) + return N0.Val; + // fold (srl x, c >= size(x)) -> undef + if (N2C && N2C->getValue() >= OpSizeInBits) + return DAG.getNode(ISD::UNDEF, VT).Val; + // fold (srl x, 0) -> x + if (N2C && N2C->isNullValue()) + return N0.Val; + // if (srl x, c) is known to be zero, return 0 + if (N2C && MaskedValueIsZero(N0,(~0ULL >> (64-OpSizeInBits))<getValue(), + TLI)) + return DAG.getConstant(0, VT).Val; + // fold (srl (srl x, c1), c2) -> 0 or (srl x, c1+c2) + if (N2C && N0.getOpcode() == ISD::SRL && + N0.getOperand(1).getOpcode() == ISD::Constant) { + uint64_t c1 = cast(N0.getOperand(1))->getValue(); + uint64_t c2 = N2C->getValue(); + if (c1 + c2 > OpSizeInBits) + return DAG.getConstant(0, VT).Val; + return DAG.getNode(ISD::SRL, VT, N0.getOperand(0), + DAG.getConstant(c1 + c2, N1.getValueType())).Val; + } + return 0; +} + +SDNode *DAGCombiner::visitCtlz(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + + // fold (ctlz c1) -> c2 + if (N1C) + return DAG.getConstant(CountLeadingZeros_64(N1C->getValue()), + N0.getValueType()).Val; + return 0; +} + +SDNode *DAGCombiner::visitCttz(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + + // fold (cttz c1) -> c2 + if (N1C) + return DAG.getConstant(CountTrailingZeros_64(N1C->getValue()), + N0.getValueType()).Val; + return 0; +} + +SDNode *DAGCombiner::visitCtpop(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + + // fold (ctpop c1) -> c2 + if (N1C) + return DAG.getConstant(CountPopulation_64(N1C->getValue()), + N0.getValueType()).Val; + return 0; +} + +SDNode *DAGCombiner::visitSignExtend(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + MVT::ValueType VT = N->getValueType(0); + + // noop sext + if (N0.getValueType() == N->getValueType(0)) + return N0.Val; + // fold (sext c1) -> c1 + if (N1C) + return DAG.getConstant(N1C->getSignExtended(), VT).Val; + // fold (sext (sext x)) -> (sext x) + if (N0.getOpcode() == ISD::SIGN_EXTEND) + return DAG.getNode(ISD::SIGN_EXTEND, VT, N0.getOperand(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitZeroExtend(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + MVT::ValueType VT = N->getValueType(0); + + // noop zext + if (N0.getValueType() == N->getValueType(0)) + return N0.Val; + // fold (zext c1) -> c1 + if (N1C) + return DAG.getConstant(N1C->getValue(), VT).Val; + // fold (zext (zext x)) -> (zext x) + if (N0.getOpcode() == ISD::ZERO_EXTEND) + return DAG.getNode(ISD::ZERO_EXTEND, VT, N0.getOperand(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitSignExtendInReg(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + MVT::ValueType VT = N->getValueType(0); + MVT::ValueType EVT = cast(N->getOperand(1))->getVT(); + + // noop sext_in_reg + if (EVT == VT) + return N0.Val; + // fold (sext_in_reg c1) -> c1 + if (N1C) { + SDOperand Truncate = DAG.getConstant(N1C->getValue(), EVT); + return DAG.getNode(ISD::SIGN_EXTEND, VT, Truncate).Val; + } + // fold (sext_in_reg (sext_in_reg x)) -> (sext_in_reg x) + if (N0.getOpcode() == ISD::SIGN_EXTEND_INREG && + cast(N0.getOperand(1))->getVT() <= EVT) { + return N0.Val; + } + // fold (sext_in_reg (assert_sext x)) -> (assert_sext x) + if (N0.getOpcode() == ISD::AssertSext && + cast(N0.getOperand(1))->getVT() <= EVT) { + return N0.Val; + } + // fold (sext_in_reg (sextload x)) -> (sextload x) + if (N0.getOpcode() == ISD::SEXTLOAD && + cast(N0.getOperand(3))->getVT() <= EVT) { + return N0.Val; + } + // fold (sext_in_reg (setcc x)) -> setcc x iff (setcc x) == 0 or 1 + if (N0.getOpcode() == ISD::SETCC && + TLI.getSetCCResultContents() == + TargetLowering::ZeroOrNegativeOneSetCCResult) + return N0.Val; + // FIXME: this code is currently just ported over from SelectionDAG.cpp + // we probably actually want to handle this in two pieces. Rather than + // checking all the top bits for zero, just check the sign bit here and turn + // it into a zero extend inreg (AND with constant). + // then, let the code for AND figure out if the mask is superfluous rather + // than doing so here. + if (N0.getOpcode() == ISD::AND && + N0.getOperand(1).getOpcode() == ISD::Constant) { + uint64_t Mask = cast(N0.getOperand(1))->getValue(); + unsigned NumBits = MVT::getSizeInBits(EVT); + if ((Mask & (~0ULL << (NumBits-1))) == 0) + return N0.Val; + } + return 0; +} + +SDNode *DAGCombiner::visitTruncate(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + MVT::ValueType VT = N->getValueType(0); + + // noop truncate + if (N0.getValueType() == N->getValueType(0)) + return N0.Val; + // fold (truncate c1) -> c1 + if (N1C) + return DAG.getConstant(N1C->getValue(), VT).Val; + // fold (truncate (truncate x)) -> (truncate x) + if (N0.getOpcode() == ISD::TRUNCATE) + return DAG.getNode(ISD::TRUNCATE, VT, N0.getOperand(0)).Val; + // fold (truncate (ext x)) -> (ext x) or (truncate x) or x + if (N0.getOpcode() == ISD::ZERO_EXTEND || N0.getOpcode() == ISD::SIGN_EXTEND){ + if (N0.getValueType() < VT) + // if the source is smaller than the dest, we still need an extend + return DAG.getNode(N0.getOpcode(), VT, N0.getOperand(0)).Val; + else if (N0.getValueType() > VT) + // if the source is larger than the dest, than we just need the truncate + return DAG.getNode(ISD::TRUNCATE, VT, N0.getOperand(0)).Val; + else + // if the source and dest are the same type, we can drop both the extend + // and the truncate + return N0.getOperand(0).Val; + } + return 0; +} + +SDNode *DAGCombiner::visitSintToFP(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + MVT::ValueType VT = N->getValueType(0); + + // fold (sint_to_fp c1) -> c1fp + if (N1C) + return DAG.getConstantFP(N1C->getSignExtended(), VT).Val; + return 0; +} + +SDNode *DAGCombiner::visitUintToFP(SDNode *N) { + SDOperand N0 = N->getOperand(0); + ConstantSDNode *N1C = dyn_cast(N0.Val); + MVT::ValueType VT = N->getValueType(0); + + // fold (uint_to_fp c1) -> c1fp + if (N1C) + return DAG.getConstantFP(N1C->getValue(), VT).Val; + return 0; +} + +SDNode *DAGCombiner::visitFPToSint(SDNode *N) { + ConstantFPSDNode *N1CFP = dyn_cast(N->getOperand(0)); + + // fold (fp_to_sint c1fp) -> c1 + if (N1CFP) + return DAG.getConstant((int64_t)N1CFP->getValue(), N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitFPToUint(SDNode *N) { + ConstantFPSDNode *N1CFP = dyn_cast(N->getOperand(0)); + + // fold (fp_to_uint c1fp) -> c1 + if (N1CFP) + return DAG.getConstant((uint64_t)N1CFP->getValue(), N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitFPRound(SDNode *N) { + ConstantFPSDNode *N1CFP = dyn_cast(N->getOperand(0)); + + // fold (fp_round c1fp) -> c1fp + if (N1CFP) + return DAG.getConstantFP(N1CFP->getValue(), N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitFPRoundInReg(SDNode *N) { + SDOperand N0 = N->getOperand(0); + MVT::ValueType VT = N->getValueType(0); + MVT::ValueType EVT = cast(N->getOperand(1))->getVT(); + ConstantFPSDNode *N1CFP = dyn_cast(N0); + + // noop fp_round_inreg + if (EVT == VT) + return N0.Val; + // fold (fp_round_inreg c1fp) -> c1fp + if (N1CFP) { + SDOperand Round = DAG.getConstantFP(N1CFP->getValue(), EVT); + return DAG.getNode(ISD::FP_EXTEND, VT, Round).Val; + } + return 0; +} + +SDNode *DAGCombiner::visitFPExtend(SDNode *N) { + ConstantFPSDNode *N1CFP = dyn_cast(N->getOperand(0)); + + // fold (fp_extend c1fp) -> c1fp + if (N1CFP) + return DAG.getConstantFP(N1CFP->getValue(), N->getValueType(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitFneg(SDNode *N) { + ConstantFPSDNode *N1CFP = dyn_cast(N->getOperand(0)); + // fold (neg c1) -> -c1 + if (N1CFP) + return DAG.getConstantFP(-N1CFP->getValue(), N->getValueType(0)).Val; + // fold (neg (sub x, y)) -> (sub y, x) + if (N->getOperand(0).getOpcode() == ISD::SUB) + return DAG.getNode(ISD::SUB, N->getValueType(0), N->getOperand(1), + N->getOperand(0)).Val; + // fold (neg (neg x)) -> x + if (N->getOperand(0).getOpcode() == ISD::FNEG) + return N->getOperand(0).getOperand(0).Val; + return 0; +} + +SDNode *DAGCombiner::visitFabs(SDNode *N) { + ConstantFPSDNode *N1CFP = dyn_cast(N->getOperand(0)); + // fold (fabs c1) -> fabs(c1) + if (N1CFP) + return DAG.getConstantFP(fabs(N1CFP->getValue()), N->getValueType(0)).Val; + // fold (fabs (fabs x)) -> (fabs x) + if (N->getOperand(0).getOpcode() == ISD::FABS) + return N->getOperand(0).Val; + // fold (fabs (fneg x)) -> (fabs x) + if (N->getOperand(0).getOpcode() == ISD::FNEG) + return DAG.getNode(ISD::FABS, N->getValueType(0), + N->getOperand(0).getOperand(0)).Val; + return 0; +} + +SDNode *DAGCombiner::visitExtLoad(SDNode *N) { + MVT::ValueType VT = N->getValueType(0); + MVT::ValueType EVT = cast(N->getOperand(3))->getVT(); + + // fold (extload vt, x) -> (load x) + if (EVT == VT) + return DAG.getLoad(VT, N->getOperand(0), N->getOperand(1), + N->getOperand(2)).Val; + return 0; +} + +SDNode *DAGCombiner::visitSextLoad(SDNode *N) { + MVT::ValueType VT = N->getValueType(0); + MVT::ValueType EVT = cast(N->getOperand(3))->getVT(); + + // fold (sextload vt, x) -> (load x) + if (EVT == VT) + return DAG.getLoad(VT, N->getOperand(0), N->getOperand(1), + N->getOperand(2)).Val; + return 0; +} + +SDNode *DAGCombiner::visitZextLoad(SDNode *N) { + MVT::ValueType VT = N->getValueType(0); + MVT::ValueType EVT = cast(N->getOperand(3))->getVT(); + + // fold (zextload vt, x) -> (load x) + if (EVT == VT) + return DAG.getLoad(VT, N->getOperand(0), N->getOperand(1), + N->getOperand(2)).Val; + return 0; +} + +SDNode *DAGCombiner::visitTruncStore(SDNode *N) { + MVT::ValueType VT = N->getValueType(0); + MVT::ValueType EVT = cast(N->getOperand(4))->getVT(); + + // fold (truncstore x, vt) -> (store x) + if (N->getOperand(0).getValueType() == EVT) + return DAG.getNode(ISD::STORE, VT, N->getOperand(0), N->getOperand(1), + N->getOperand(2), N->getOperand(3)).Val; + return 0; +} + +// SelectionDAG::Combine - This is the entry point for the file. +// +void SelectionDAG::Combine(bool AfterLegalize) { + /// run - This is the main entry point to this class. + /// + DAGCombiner(*this).Run(AfterLegalize); +} -- 2.34.1