1 //===- HexagonConstPropagation.cpp ----------------------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 #define DEBUG_TYPE "hcp"
12 #include "HexagonInstrInfo.h"
13 #include "HexagonRegisterInfo.h"
14 #include "HexagonSubtarget.h"
15 #include "llvm/ADT/APFloat.h"
16 #include "llvm/ADT/APInt.h"
17 #include "llvm/ADT/PostOrderIterator.h"
18 #include "llvm/ADT/SetVector.h"
19 #include "llvm/ADT/SmallVector.h"
20 #include "llvm/ADT/StringRef.h"
21 #include "llvm/CodeGen/MachineBasicBlock.h"
22 #include "llvm/CodeGen/MachineFunction.h"
23 #include "llvm/CodeGen/MachineFunctionPass.h"
24 #include "llvm/CodeGen/MachineInstr.h"
25 #include "llvm/CodeGen/MachineInstrBuilder.h"
26 #include "llvm/CodeGen/MachineOperand.h"
27 #include "llvm/CodeGen/MachineRegisterInfo.h"
28 #include "llvm/CodeGen/TargetRegisterInfo.h"
29 #include "llvm/CodeGen/TargetSubtargetInfo.h"
30 #include "llvm/IR/Constants.h"
31 #include "llvm/IR/Type.h"
32 #include "llvm/Pass.h"
33 #include "llvm/Support/Casting.h"
34 #include "llvm/Support/Compiler.h"
35 #include "llvm/Support/Debug.h"
36 #include "llvm/Support/ErrorHandling.h"
37 #include "llvm/Support/MathExtras.h"
38 #include "llvm/Support/raw_ostream.h"
53 // Properties of a value that are tracked by the propagation.
54 // A property that is marked as present (i.e. bit is set) dentes that the
55 // value is known (proven) to have this property. Not all combinations
56 // of bits make sense, for example Zero and NonZero are mutually exclusive,
57 // but on the other hand, Zero implies Finite. In this case, whenever
58 // the Zero property is present, Finite should also be present.
59 class ConstantProperties {
69 NumericProperties = (Zero|NonZero|Finite|Infinity|NaN|SignedZero),
72 SignProperties = (PosOrZero|NegOrZero),
73 Everything = (NumericProperties|SignProperties)
76 // For a given constant, deduce the set of trackable properties that this
78 static uint32_t deduce(const Constant *C);
81 // A representation of a register as it can appear in a MachineOperand,
82 // i.e. a pair register:subregister.
86 explicit Register(unsigned R, unsigned SR = 0) : Reg(R), SubReg(SR) {}
87 explicit Register(const MachineOperand &MO)
88 : Reg(MO.getReg()), SubReg(MO.getSubReg()) {}
90 void print(const TargetRegisterInfo *TRI = nullptr) const {
91 dbgs() << printReg(Reg, TRI, SubReg);
94 bool operator== (const Register &R) const {
95 return (Reg == R.Reg) && (SubReg == R.SubReg);
99 // Lattice cell, based on that was described in the W-Z paper on constant
101 // Latice cell will be allowed to hold multiple constant values. While
102 // multiple values would normally indicate "bottom", we can still derive
103 // some useful information from them. For example, comparison X > 0
104 // could be folded if all the values in the cell associated with X are
108 enum { Normal, Top, Bottom };
110 static const unsigned MaxCellSize = 4;
114 unsigned IsSpecial:1;
120 const Constant *Value;
121 const Constant *Values[MaxCellSize];
124 LatticeCell() : Kind(Top), Size(0), IsSpecial(false) {
125 for (unsigned i = 0; i < MaxCellSize; ++i)
129 bool meet(const LatticeCell &L);
130 bool add(const Constant *C);
131 bool add(uint32_t Property);
132 uint32_t properties() const;
133 unsigned size() const { return Size; }
135 LatticeCell &operator= (const LatticeCell &L) {
137 // This memcpy also copies Properties (when L.Size == 0).
138 uint32_t N = L.IsSpecial ? sizeof L.Properties
139 : L.Size*sizeof(const Constant*);
140 memcpy(Values, L.Values, N);
143 IsSpecial = L.IsSpecial;
148 bool isSingle() const { return size() == 1; }
149 bool isProperty() const { return IsSpecial; }
150 bool isTop() const { return Kind == Top; }
151 bool isBottom() const { return Kind == Bottom; }
154 bool Changed = (Kind != Bottom);
161 void print(raw_ostream &os) const;
170 bool convertToProperty();
174 raw_ostream &operator<< (raw_ostream &os, const LatticeCell &L) {
180 class MachineConstEvaluator;
182 class MachineConstPropagator {
184 MachineConstPropagator(MachineConstEvaluator &E) : MCE(E) {
188 // Mapping: vreg -> cell
189 // The keys are registers _without_ subregisters. This won't allow
190 // definitions in the form of "vreg:subreg = ...". Such definitions
191 // would be questionable from the point of view of SSA, since the "vreg"
192 // could not be initialized in its entirety (specifically, an instruction
193 // defining the "other part" of "vreg" would also count as a definition
194 // of "vreg", which would violate the SSA).
195 // If a value of a pair vreg:subreg needs to be obtained, the cell for
196 // "vreg" needs to be looked up, and then the value of subregister "subreg"
197 // needs to be evaluated.
205 void clear() { Map.clear(); }
207 bool has(unsigned R) const {
208 // All non-virtual registers are considered "bottom".
209 if (!TargetRegisterInfo::isVirtualRegister(R))
211 MapType::const_iterator F = Map.find(R);
212 return F != Map.end();
215 const LatticeCell &get(unsigned R) const {
216 if (!TargetRegisterInfo::isVirtualRegister(R))
218 MapType::const_iterator F = Map.find(R);
224 // Invalidates any const references.
225 void update(unsigned R, const LatticeCell &L) {
229 void print(raw_ostream &os, const TargetRegisterInfo &TRI) const;
232 using MapType = std::map<unsigned, LatticeCell>;
235 // To avoid creating "top" entries, return a const reference to
236 // this cell in "get". Also, have a "Bottom" cell to return from
237 // get when a value of a physical register is requested.
238 LatticeCell Top, Bottom;
241 using const_iterator = MapType::const_iterator;
243 const_iterator begin() const { return Map.begin(); }
244 const_iterator end() const { return Map.end(); }
247 bool run(MachineFunction &MF);
250 void visitPHI(const MachineInstr &PN);
251 void visitNonBranch(const MachineInstr &MI);
252 void visitBranchesFrom(const MachineInstr &BrI);
253 void visitUsesOf(unsigned R);
254 bool computeBlockSuccessors(const MachineBasicBlock *MB,
255 SetVector<const MachineBasicBlock*> &Targets);
256 void removeCFGEdge(MachineBasicBlock *From, MachineBasicBlock *To);
258 void propagate(MachineFunction &MF);
259 bool rewrite(MachineFunction &MF);
261 MachineRegisterInfo *MRI;
262 MachineConstEvaluator &MCE;
264 using CFGEdge = std::pair<unsigned, unsigned>;
265 using SetOfCFGEdge = std::set<CFGEdge>;
266 using SetOfInstr = std::set<const MachineInstr *>;
267 using QueueOfCFGEdge = std::queue<CFGEdge>;
271 SetOfCFGEdge EdgeExec;
272 SetOfInstr InstrExec;
273 QueueOfCFGEdge FlowQ;
276 // The "evaluator/rewriter" of machine instructions. This is an abstract
277 // base class that provides the interface that the propagator will use,
278 // as well as some helper functions that are target-independent.
279 class MachineConstEvaluator {
281 MachineConstEvaluator(MachineFunction &Fn)
282 : TRI(*Fn.getSubtarget().getRegisterInfo()),
283 MF(Fn), CX(Fn.getFunction().getContext()) {}
284 virtual ~MachineConstEvaluator() = default;
286 // The required interface:
287 // - A set of three "evaluate" functions. Each returns "true" if the
288 // computation succeeded, "false" otherwise.
289 // (1) Given an instruction MI, and the map with input values "Inputs",
290 // compute the set of output values "Outputs". An example of when
291 // the computation can "fail" is if MI is not an instruction that
292 // is recognized by the evaluator.
293 // (2) Given a register R (as reg:subreg), compute the cell that
294 // corresponds to the "subreg" part of the given register.
295 // (3) Given a branch instruction BrI, compute the set of target blocks.
296 // If the branch can fall-through, add null (0) to the list of
298 // - A function "rewrite", that given the cell map after propagation,
299 // could rewrite instruction MI in a more beneficial form. Return
300 // "true" if a change has been made, "false" otherwise.
301 using CellMap = MachineConstPropagator::CellMap;
302 virtual bool evaluate(const MachineInstr &MI, const CellMap &Inputs,
303 CellMap &Outputs) = 0;
304 virtual bool evaluate(const Register &R, const LatticeCell &SrcC,
305 LatticeCell &Result) = 0;
306 virtual bool evaluate(const MachineInstr &BrI, const CellMap &Inputs,
307 SetVector<const MachineBasicBlock*> &Targets,
308 bool &CanFallThru) = 0;
309 virtual bool rewrite(MachineInstr &MI, const CellMap &Inputs) = 0;
311 const TargetRegisterInfo &TRI;
322 L = 0x04, // Less-than property.
323 G = 0x08, // Greater-than property.
324 U = 0x40, // Unsigned property.
335 static uint32_t negate(uint32_t Cmp) {
340 assert((Cmp & (L|G)) != (L|G));
347 bool getCell(const Register &R, const CellMap &Inputs, LatticeCell &RC);
348 bool constToInt(const Constant *C, APInt &Val) const;
349 bool constToFloat(const Constant *C, APFloat &Val) const;
350 const ConstantInt *intToConst(const APInt &Val) const;
353 bool evaluateCMPrr(uint32_t Cmp, const Register &R1, const Register &R2,
354 const CellMap &Inputs, bool &Result);
355 bool evaluateCMPri(uint32_t Cmp, const Register &R1, const APInt &A2,
356 const CellMap &Inputs, bool &Result);
357 bool evaluateCMPrp(uint32_t Cmp, const Register &R1, uint64_t Props2,
358 const CellMap &Inputs, bool &Result);
359 bool evaluateCMPii(uint32_t Cmp, const APInt &A1, const APInt &A2,
361 bool evaluateCMPpi(uint32_t Cmp, uint32_t Props, const APInt &A2,
363 bool evaluateCMPpp(uint32_t Cmp, uint32_t Props1, uint32_t Props2,
366 bool evaluateCOPY(const Register &R1, const CellMap &Inputs,
367 LatticeCell &Result);
369 // Logical operations.
370 bool evaluateANDrr(const Register &R1, const Register &R2,
371 const CellMap &Inputs, LatticeCell &Result);
372 bool evaluateANDri(const Register &R1, const APInt &A2,
373 const CellMap &Inputs, LatticeCell &Result);
374 bool evaluateANDii(const APInt &A1, const APInt &A2, APInt &Result);
375 bool evaluateORrr(const Register &R1, const Register &R2,
376 const CellMap &Inputs, LatticeCell &Result);
377 bool evaluateORri(const Register &R1, const APInt &A2,
378 const CellMap &Inputs, LatticeCell &Result);
379 bool evaluateORii(const APInt &A1, const APInt &A2, APInt &Result);
380 bool evaluateXORrr(const Register &R1, const Register &R2,
381 const CellMap &Inputs, LatticeCell &Result);
382 bool evaluateXORri(const Register &R1, const APInt &A2,
383 const CellMap &Inputs, LatticeCell &Result);
384 bool evaluateXORii(const APInt &A1, const APInt &A2, APInt &Result);
387 bool evaluateZEXTr(const Register &R1, unsigned Width, unsigned Bits,
388 const CellMap &Inputs, LatticeCell &Result);
389 bool evaluateZEXTi(const APInt &A1, unsigned Width, unsigned Bits,
391 bool evaluateSEXTr(const Register &R1, unsigned Width, unsigned Bits,
392 const CellMap &Inputs, LatticeCell &Result);
393 bool evaluateSEXTi(const APInt &A1, unsigned Width, unsigned Bits,
396 // Leading/trailing bits.
397 bool evaluateCLBr(const Register &R1, bool Zeros, bool Ones,
398 const CellMap &Inputs, LatticeCell &Result);
399 bool evaluateCLBi(const APInt &A1, bool Zeros, bool Ones, APInt &Result);
400 bool evaluateCTBr(const Register &R1, bool Zeros, bool Ones,
401 const CellMap &Inputs, LatticeCell &Result);
402 bool evaluateCTBi(const APInt &A1, bool Zeros, bool Ones, APInt &Result);
405 bool evaluateEXTRACTr(const Register &R1, unsigned Width, unsigned Bits,
406 unsigned Offset, bool Signed, const CellMap &Inputs,
407 LatticeCell &Result);
408 bool evaluateEXTRACTi(const APInt &A1, unsigned Bits, unsigned Offset,
409 bool Signed, APInt &Result);
410 // Vector operations.
411 bool evaluateSplatr(const Register &R1, unsigned Bits, unsigned Count,
412 const CellMap &Inputs, LatticeCell &Result);
413 bool evaluateSplati(const APInt &A1, unsigned Bits, unsigned Count,
417 } // end anonymous namespace
419 uint32_t ConstantProperties::deduce(const Constant *C) {
420 if (isa<ConstantInt>(C)) {
421 const ConstantInt *CI = cast<ConstantInt>(C);
423 return Zero | PosOrZero | NegOrZero | Finite;
424 uint32_t Props = (NonZero | Finite);
425 if (CI->isNegative())
426 return Props | NegOrZero;
427 return Props | PosOrZero;
430 if (isa<ConstantFP>(C)) {
431 const ConstantFP *CF = cast<ConstantFP>(C);
432 uint32_t Props = CF->isNegative() ? (NegOrZero|NonZero)
435 return (Props & ~NumericProperties) | (Zero|Finite);
436 Props = (Props & ~NumericProperties) | NonZero;
438 return (Props & ~NumericProperties) | NaN;
439 const APFloat &Val = CF->getValueAPF();
440 if (Val.isInfinity())
441 return (Props & ~NumericProperties) | Infinity;
449 // Convert a cell from a set of specific values to a cell that tracks
451 bool LatticeCell::convertToProperty() {
454 // Corner case: converting a fresh (top) cell to "special".
455 // This can happen, when adding a property to a top cell.
456 uint32_t Everything = ConstantProperties::Everything;
457 uint32_t Ps = !isTop() ? properties()
459 if (Ps != ConstantProperties::Unknown) {
469 void LatticeCell::print(raw_ostream &os) const {
472 uint32_t Ps = properties();
473 if (Ps & ConstantProperties::Zero)
475 if (Ps & ConstantProperties::NonZero)
477 if (Ps & ConstantProperties::Finite)
479 if (Ps & ConstantProperties::Infinity)
481 if (Ps & ConstantProperties::NaN)
483 if (Ps & ConstantProperties::PosOrZero)
485 if (Ps & ConstantProperties::NegOrZero)
494 } else if (isTop()) {
497 for (unsigned i = 0; i < size(); ++i) {
498 const Constant *C = Values[i];
508 // "Meet" operation on two cells. This is the key of the propagation
510 bool LatticeCell::meet(const LatticeCell &L) {
511 bool Changed = false;
513 Changed = setBottom();
514 if (isBottom() || L.isTop())
518 // L can be neither Top nor Bottom, so *this must have changed.
522 // Top/bottom cases covered. Need to integrate L's set into ours.
524 return add(L.properties());
525 for (unsigned i = 0; i < L.size(); ++i) {
526 const Constant *LC = L.Values[i];
532 // Add a new constant to the cell. This is actually where the cell update
533 // happens. If a cell has room for more constants, the new constant is added.
534 // Otherwise, the cell is converted to a "property" cell (i.e. a cell that
535 // will track properties of the associated values, and not the values
536 // themselves. Care is taken to handle special cases, like "bottom", etc.
537 bool LatticeCell::add(const Constant *LC) {
543 // Cell is not special. Try to add the constant here first,
546 while (Index < Size) {
547 const Constant *C = Values[Index];
548 // If the constant is already here, no change is needed.
553 if (Index < MaxCellSize) {
561 bool Changed = false;
563 // This cell is special, or is not special, but is full. After this
564 // it will be special.
565 Changed = convertToProperty();
566 uint32_t Ps = properties();
567 uint32_t NewPs = Ps & ConstantProperties::deduce(LC);
568 if (NewPs == ConstantProperties::Unknown) {
579 // Add a property to the cell. This will force the cell to become a property-
581 bool LatticeCell::add(uint32_t Property) {
582 bool Changed = convertToProperty();
583 uint32_t Ps = properties();
584 if (Ps == (Ps & Property))
586 Properties = Property & Ps;
590 // Return the properties of the values in the cell. This is valid for any
591 // cell, and does not alter the cell itself.
592 uint32_t LatticeCell::properties() const {
595 assert(!isTop() && "Should not call this for a top cell");
597 return ConstantProperties::Unknown;
599 assert(size() > 0 && "Empty cell");
600 uint32_t Ps = ConstantProperties::deduce(Values[0]);
601 for (unsigned i = 1; i < size(); ++i) {
602 if (Ps == ConstantProperties::Unknown)
604 Ps &= ConstantProperties::deduce(Values[i]);
610 void MachineConstPropagator::CellMap::print(raw_ostream &os,
611 const TargetRegisterInfo &TRI) const {
613 dbgs() << " " << printReg(I.first, &TRI) << " -> " << I.second << '\n';
617 void MachineConstPropagator::visitPHI(const MachineInstr &PN) {
618 const MachineBasicBlock *MB = PN.getParent();
619 unsigned MBN = MB->getNumber();
620 DEBUG(dbgs() << "Visiting FI(" << printMBBReference(*MB) << "): " << PN);
622 const MachineOperand &MD = PN.getOperand(0);
624 assert(TargetRegisterInfo::isVirtualRegister(DefR.Reg));
626 bool Changed = false;
628 // If the def has a sub-register, set the corresponding cell to "bottom".
631 const LatticeCell &T = Cells.get(DefR.Reg);
632 Changed = !T.isBottom();
633 Cells.update(DefR.Reg, Bottom);
635 visitUsesOf(DefR.Reg);
639 LatticeCell DefC = Cells.get(DefR.Reg);
641 for (unsigned i = 1, n = PN.getNumOperands(); i < n; i += 2) {
642 const MachineBasicBlock *PB = PN.getOperand(i+1).getMBB();
643 unsigned PBN = PB->getNumber();
644 if (!EdgeExec.count(CFGEdge(PBN, MBN))) {
645 DEBUG(dbgs() << " edge " << printMBBReference(*PB) << "->"
646 << printMBBReference(*MB) << " not executable\n");
649 const MachineOperand &SO = PN.getOperand(i);
651 // If the input is not a virtual register, we don't really know what
653 if (!TargetRegisterInfo::isVirtualRegister(UseR.Reg))
655 // If there is no cell for an input register, it means top.
656 if (!Cells.has(UseR.Reg))
660 bool Eval = MCE.evaluate(UseR, Cells.get(UseR.Reg), SrcC);
661 DEBUG(dbgs() << " edge from " << printMBBReference(*PB) << ": "
662 << printReg(UseR.Reg, &MCE.TRI, UseR.SubReg) << SrcC << '\n');
663 Changed |= Eval ? DefC.meet(SrcC)
665 Cells.update(DefR.Reg, DefC);
670 visitUsesOf(DefR.Reg);
673 void MachineConstPropagator::visitNonBranch(const MachineInstr &MI) {
674 DEBUG(dbgs() << "Visiting MI(" << printMBBReference(*MI.getParent())
677 bool Eval = MCE.evaluate(MI, Cells, Outputs);
680 dbgs() << " outputs:";
681 for (auto &I : Outputs)
682 dbgs() << ' ' << I.second;
687 // Update outputs. If the value was not computed, set all the
688 // def cells to bottom.
689 for (const MachineOperand &MO : MI.operands()) {
690 if (!MO.isReg() || !MO.isDef())
693 // Only track virtual registers.
694 if (!TargetRegisterInfo::isVirtualRegister(DefR.Reg))
696 bool Changed = false;
697 // If the evaluation failed, set cells for all output registers to bottom.
699 const LatticeCell &T = Cells.get(DefR.Reg);
700 Changed = !T.isBottom();
701 Cells.update(DefR.Reg, Bottom);
703 // Find the corresponding cell in the computed outputs.
704 // If it's not there, go on to the next def.
705 if (!Outputs.has(DefR.Reg))
707 LatticeCell RC = Cells.get(DefR.Reg);
708 Changed = RC.meet(Outputs.get(DefR.Reg));
709 Cells.update(DefR.Reg, RC);
712 visitUsesOf(DefR.Reg);
716 // \brief Starting at a given branch, visit remaining branches in the block.
717 // Traverse over the subsequent branches for as long as the preceding one
718 // can fall through. Add all the possible targets to the flow work queue,
719 // including the potential fall-through to the layout-successor block.
720 void MachineConstPropagator::visitBranchesFrom(const MachineInstr &BrI) {
721 const MachineBasicBlock &B = *BrI.getParent();
722 unsigned MBN = B.getNumber();
723 MachineBasicBlock::const_iterator It = BrI.getIterator();
724 MachineBasicBlock::const_iterator End = B.end();
726 SetVector<const MachineBasicBlock*> Targets;
727 bool EvalOk = true, FallsThru = true;
729 const MachineInstr &MI = *It;
730 InstrExec.insert(&MI);
731 DEBUG(dbgs() << "Visiting " << (EvalOk ? "BR" : "br") << "("
732 << printMBBReference(B) << "): " << MI);
733 // Do not evaluate subsequent branches if the evaluation of any of the
734 // previous branches failed. Keep iterating over the branches only
735 // to mark them as executable.
736 EvalOk = EvalOk && MCE.evaluate(MI, Cells, Targets, FallsThru);
745 // Need to add all CFG successors that lead to EH landing pads.
746 // There won't be explicit branches to these blocks, but they must
748 for (const MachineBasicBlock *SB : B.successors()) {
753 const MachineFunction &MF = *B.getParent();
754 MachineFunction::const_iterator BI = B.getIterator();
755 MachineFunction::const_iterator Next = std::next(BI);
756 if (Next != MF.end())
757 Targets.insert(&*Next);
760 // If the evaluation of the branches failed, make "Targets" to be the
761 // set of all successors of the block from the CFG.
762 // If the evaluation succeeded for all visited branches, then if the
763 // last one set "FallsThru", then add an edge to the layout successor
766 DEBUG(dbgs() << " failed to evaluate a branch...adding all CFG "
768 for (const MachineBasicBlock *SB : B.successors())
772 for (const MachineBasicBlock *TB : Targets) {
773 unsigned TBN = TB->getNumber();
774 DEBUG(dbgs() << " pushing edge " << printMBBReference(B) << " -> "
775 << printMBBReference(*TB) << "\n");
776 FlowQ.push(CFGEdge(MBN, TBN));
780 void MachineConstPropagator::visitUsesOf(unsigned Reg) {
781 DEBUG(dbgs() << "Visiting uses of " << printReg(Reg, &MCE.TRI)
782 << Cells.get(Reg) << '\n');
783 for (MachineInstr &MI : MRI->use_nodbg_instructions(Reg)) {
784 // Do not process non-executable instructions. They can become exceutable
785 // later (via a flow-edge in the work queue). In such case, the instruc-
786 // tion will be visited at that time.
787 if (!InstrExec.count(&MI))
791 else if (!MI.isBranch())
794 visitBranchesFrom(MI);
798 bool MachineConstPropagator::computeBlockSuccessors(const MachineBasicBlock *MB,
799 SetVector<const MachineBasicBlock*> &Targets) {
800 MachineBasicBlock::const_iterator FirstBr = MB->end();
801 for (const MachineInstr &MI : *MB) {
802 if (MI.isDebugValue())
805 FirstBr = MI.getIterator();
811 MachineBasicBlock::const_iterator End = MB->end();
814 for (MachineBasicBlock::const_iterator I = FirstBr; I != End; ++I) {
815 const MachineInstr &MI = *I;
816 // Can there be debug instructions between branches?
817 if (MI.isDebugValue())
819 if (!InstrExec.count(&MI))
821 bool Eval = MCE.evaluate(MI, Cells, Targets, DoNext);
827 // If the last branch could fall-through, add block's layout successor.
829 MachineFunction::const_iterator BI = MB->getIterator();
830 MachineFunction::const_iterator NextI = std::next(BI);
831 if (NextI != MB->getParent()->end())
832 Targets.insert(&*NextI);
835 // Add all the EH landing pads.
836 for (const MachineBasicBlock *SB : MB->successors())
843 void MachineConstPropagator::removeCFGEdge(MachineBasicBlock *From,
844 MachineBasicBlock *To) {
845 // First, remove the CFG successor/predecessor information.
846 From->removeSuccessor(To);
847 // Remove all corresponding PHI operands in the To block.
848 for (auto I = To->begin(), E = To->getFirstNonPHI(); I != E; ++I) {
849 MachineInstr *PN = &*I;
850 // reg0 = PHI reg1, bb2, reg3, bb4, ...
851 int N = PN->getNumOperands()-2;
853 if (PN->getOperand(N+1).getMBB() == From) {
854 PN->RemoveOperand(N+1);
855 PN->RemoveOperand(N);
862 void MachineConstPropagator::propagate(MachineFunction &MF) {
863 MachineBasicBlock *Entry = GraphTraits<MachineFunction*>::getEntryNode(&MF);
864 unsigned EntryNum = Entry->getNumber();
866 // Start with a fake edge, just to process the entry node.
867 FlowQ.push(CFGEdge(EntryNum, EntryNum));
869 while (!FlowQ.empty()) {
870 CFGEdge Edge = FlowQ.front();
873 DEBUG(dbgs() << "Picked edge "
874 << printMBBReference(*MF.getBlockNumbered(Edge.first)) << "->"
875 << printMBBReference(*MF.getBlockNumbered(Edge.second))
877 if (Edge.first != EntryNum)
878 if (EdgeExec.count(Edge))
880 EdgeExec.insert(Edge);
881 MachineBasicBlock *SB = MF.getBlockNumbered(Edge.second);
883 // Process the block in three stages:
884 // - visit all PHI nodes,
885 // - visit all non-branch instructions,
886 // - visit block branches.
887 MachineBasicBlock::const_iterator It = SB->begin(), End = SB->end();
889 // Visit PHI nodes in the successor block.
890 while (It != End && It->isPHI()) {
891 InstrExec.insert(&*It);
896 // If the successor block just became executable, visit all instructions.
897 // To see if this is the first time we're visiting it, check the first
898 // non-debug instruction to see if it is executable.
899 while (It != End && It->isDebugValue())
901 assert(It == End || !It->isPHI());
902 // If this block has been visited, go on to the next one.
903 if (It != End && InstrExec.count(&*It))
905 // For now, scan all non-branch instructions. Branches require different
907 while (It != End && !It->isBranch()) {
908 if (!It->isDebugValue()) {
909 InstrExec.insert(&*It);
915 // Time to process the end of the block. This is different from
916 // processing regular (non-branch) instructions, because there can
917 // be multiple branches in a block, and they can cause the block to
920 visitBranchesFrom(*It);
922 // If the block didn't have a branch, add all successor edges to the
923 // work queue. (There should really be only one successor in such case.)
924 unsigned SBN = SB->getNumber();
925 for (const MachineBasicBlock *SSB : SB->successors())
926 FlowQ.push(CFGEdge(SBN, SSB->getNumber()));
931 dbgs() << "Cells after propagation:\n";
932 Cells.print(dbgs(), MCE.TRI);
933 dbgs() << "Dead CFG edges:\n";
934 for (const MachineBasicBlock &B : MF) {
935 unsigned BN = B.getNumber();
936 for (const MachineBasicBlock *SB : B.successors()) {
937 unsigned SN = SB->getNumber();
938 if (!EdgeExec.count(CFGEdge(BN, SN)))
939 dbgs() << " " << printMBBReference(B) << " -> "
940 << printMBBReference(*SB) << '\n';
946 bool MachineConstPropagator::rewrite(MachineFunction &MF) {
947 bool Changed = false;
948 // Rewrite all instructions based on the collected cell information.
950 // Traverse the instructions in a post-order, so that rewriting an
951 // instruction can make changes "downstream" in terms of control-flow
952 // without affecting the rewriting process. (We should not change
953 // instructions that have not yet been visited by the rewriter.)
954 // The reason for this is that the rewriter can introduce new vregs,
955 // and replace uses of old vregs (which had corresponding cells
956 // computed during propagation) with these new vregs (which at this
957 // point would not have any cells, and would appear to be "top").
958 // If an attempt was made to evaluate an instruction with a fresh
959 // "top" vreg, it would cause an error (abend) in the evaluator.
961 // Collect the post-order-traversal block ordering. The subsequent
962 // traversal/rewrite will update block successors, so it's safer
963 // if the visiting order it computed ahead of time.
964 std::vector<MachineBasicBlock*> POT;
965 for (MachineBasicBlock *B : post_order(&MF))
969 for (MachineBasicBlock *B : POT) {
970 // Walk the block backwards (which usually begin with the branches).
971 // If any branch is rewritten, we may need to update the successor
972 // information for this block. Unless the block's successors can be
973 // precisely determined (which may not be the case for indirect
974 // branches), we cannot modify any branch.
976 // Compute the successor information.
977 SetVector<const MachineBasicBlock*> Targets;
978 bool HaveTargets = computeBlockSuccessors(B, Targets);
979 // Rewrite the executable instructions. Skip branches if we don't
980 // have block successor information.
981 for (auto I = B->rbegin(), E = B->rend(); I != E; ++I) {
982 MachineInstr &MI = *I;
983 if (InstrExec.count(&MI)) {
984 if (MI.isBranch() && !HaveTargets)
986 Changed |= MCE.rewrite(MI, Cells);
989 // The rewriting could rewrite PHI nodes to non-PHI nodes, causing
990 // regular instructions to appear in between PHI nodes. Bring all
991 // the PHI nodes to the beginning of the block.
992 for (auto I = B->begin(), E = B->end(); I != E; ++I) {
995 // I is not PHI. Find the next PHI node P.
1003 // Splice P right before I.
1005 // Reset I to point at the just spliced PHI node.
1008 // Update the block successor information: remove unnecessary successors.
1010 SmallVector<MachineBasicBlock*,2> ToRemove;
1011 for (MachineBasicBlock *SB : B->successors()) {
1012 if (!Targets.count(SB))
1013 ToRemove.push_back(const_cast<MachineBasicBlock*>(SB));
1016 for (unsigned i = 0, n = ToRemove.size(); i < n; ++i)
1017 removeCFGEdge(B, ToRemove[i]);
1018 // If there are any blocks left in the computed targets, it means that
1019 // we think that the block could go somewhere, but the CFG does not.
1020 // This could legitimately happen in blocks that have non-returning
1021 // calls---we would think that the execution can continue, but the
1022 // CFG will not have a successor edge.
1025 // Need to do some final post-processing.
1026 // If a branch was not executable, it will not get rewritten, but should
1027 // be removed (or replaced with something equivalent to a A2_nop). We can't
1028 // erase instructions during rewriting, so this needs to be delayed until
1030 for (MachineBasicBlock &B : MF) {
1031 MachineBasicBlock::iterator I = B.begin(), E = B.end();
1033 auto Next = std::next(I);
1034 if (I->isBranch() && !InstrExec.count(&*I))
1042 // This is the constant propagation algorithm as described by Wegman-Zadeck.
1043 // Most of the terminology comes from there.
1044 bool MachineConstPropagator::run(MachineFunction &MF) {
1045 DEBUG(MF.print(dbgs() << "Starting MachineConstPropagator\n", nullptr));
1047 MRI = &MF.getRegInfo();
1052 assert(FlowQ.empty());
1055 bool Changed = rewrite(MF);
1058 dbgs() << "End of MachineConstPropagator (Changed=" << Changed << ")\n";
1060 MF.print(dbgs(), nullptr);
1065 // --------------------------------------------------------------------
1066 // Machine const evaluator.
1068 bool MachineConstEvaluator::getCell(const Register &R, const CellMap &Inputs,
1070 if (!TargetRegisterInfo::isVirtualRegister(R.Reg))
1072 const LatticeCell &L = Inputs.get(R.Reg);
1075 return !RC.isBottom();
1077 bool Eval = evaluate(R, L, RC);
1078 return Eval && !RC.isBottom();
1081 bool MachineConstEvaluator::constToInt(const Constant *C,
1083 const ConstantInt *CI = dyn_cast<ConstantInt>(C);
1086 Val = CI->getValue();
1090 const ConstantInt *MachineConstEvaluator::intToConst(const APInt &Val) const {
1091 return ConstantInt::get(CX, Val);
1094 bool MachineConstEvaluator::evaluateCMPrr(uint32_t Cmp, const Register &R1,
1095 const Register &R2, const CellMap &Inputs, bool &Result) {
1096 assert(Inputs.has(R1.Reg) && Inputs.has(R2.Reg));
1097 LatticeCell LS1, LS2;
1098 if (!getCell(R1, Inputs, LS1) || !getCell(R2, Inputs, LS2))
1101 bool IsProp1 = LS1.isProperty();
1102 bool IsProp2 = LS2.isProperty();
1104 uint32_t Prop1 = LS1.properties();
1106 return evaluateCMPpp(Cmp, Prop1, LS2.properties(), Result);
1107 uint32_t NegCmp = Comparison::negate(Cmp);
1108 return evaluateCMPrp(NegCmp, R2, Prop1, Inputs, Result);
1111 uint32_t Prop2 = LS2.properties();
1112 return evaluateCMPrp(Cmp, R1, Prop2, Inputs, Result);
1116 bool IsTrue = true, IsFalse = true;
1117 for (unsigned i = 0; i < LS2.size(); ++i) {
1119 bool Computed = constToInt(LS2.Values[i], A) &&
1120 evaluateCMPri(Cmp, R1, A, Inputs, Res);
1126 assert(!IsTrue || !IsFalse);
1127 // The actual logical value of the comparison is same as IsTrue.
1129 // Return true if the result was proven to be true or proven to be false.
1130 return IsTrue || IsFalse;
1133 bool MachineConstEvaluator::evaluateCMPri(uint32_t Cmp, const Register &R1,
1134 const APInt &A2, const CellMap &Inputs, bool &Result) {
1135 assert(Inputs.has(R1.Reg));
1137 if (!getCell(R1, Inputs, LS))
1139 if (LS.isProperty())
1140 return evaluateCMPpi(Cmp, LS.properties(), A2, Result);
1143 bool IsTrue = true, IsFalse = true;
1144 for (unsigned i = 0; i < LS.size(); ++i) {
1146 bool Computed = constToInt(LS.Values[i], A) &&
1147 evaluateCMPii(Cmp, A, A2, Res);
1153 assert(!IsTrue || !IsFalse);
1154 // The actual logical value of the comparison is same as IsTrue.
1156 // Return true if the result was proven to be true or proven to be false.
1157 return IsTrue || IsFalse;
1160 bool MachineConstEvaluator::evaluateCMPrp(uint32_t Cmp, const Register &R1,
1161 uint64_t Props2, const CellMap &Inputs, bool &Result) {
1162 assert(Inputs.has(R1.Reg));
1164 if (!getCell(R1, Inputs, LS))
1166 if (LS.isProperty())
1167 return evaluateCMPpp(Cmp, LS.properties(), Props2, Result);
1170 uint32_t NegCmp = Comparison::negate(Cmp);
1171 bool IsTrue = true, IsFalse = true;
1172 for (unsigned i = 0; i < LS.size(); ++i) {
1174 bool Computed = constToInt(LS.Values[i], A) &&
1175 evaluateCMPpi(NegCmp, Props2, A, Res);
1181 assert(!IsTrue || !IsFalse);
1183 return IsTrue || IsFalse;
1186 bool MachineConstEvaluator::evaluateCMPii(uint32_t Cmp, const APInt &A1,
1187 const APInt &A2, bool &Result) {
1188 // NE is a special kind of comparison (not composed of smaller properties).
1189 if (Cmp == Comparison::NE) {
1190 Result = !APInt::isSameValue(A1, A2);
1193 if (Cmp == Comparison::EQ) {
1194 Result = APInt::isSameValue(A1, A2);
1197 if (Cmp & Comparison::EQ) {
1198 if (APInt::isSameValue(A1, A2))
1199 return (Result = true);
1201 assert((Cmp & (Comparison::L | Comparison::G)) && "Malformed comparison");
1204 unsigned W1 = A1.getBitWidth();
1205 unsigned W2 = A2.getBitWidth();
1206 unsigned MaxW = (W1 >= W2) ? W1 : W2;
1207 if (Cmp & Comparison::U) {
1208 const APInt Zx1 = A1.zextOrSelf(MaxW);
1209 const APInt Zx2 = A2.zextOrSelf(MaxW);
1210 if (Cmp & Comparison::L)
1211 Result = Zx1.ult(Zx2);
1212 else if (Cmp & Comparison::G)
1213 Result = Zx2.ult(Zx1);
1217 // Signed comparison.
1218 const APInt Sx1 = A1.sextOrSelf(MaxW);
1219 const APInt Sx2 = A2.sextOrSelf(MaxW);
1220 if (Cmp & Comparison::L)
1221 Result = Sx1.slt(Sx2);
1222 else if (Cmp & Comparison::G)
1223 Result = Sx2.slt(Sx1);
1227 bool MachineConstEvaluator::evaluateCMPpi(uint32_t Cmp, uint32_t Props,
1228 const APInt &A2, bool &Result) {
1229 if (Props == ConstantProperties::Unknown)
1232 // Should never see NaN here, but check for it for completeness.
1233 if (Props & ConstantProperties::NaN)
1235 // Infinity could theoretically be compared to a number, but the
1236 // presence of infinity here would be very suspicious. If we don't
1237 // know for sure that the number is finite, bail out.
1238 if (!(Props & ConstantProperties::Finite))
1241 // Let X be a number that has properties Props.
1243 if (Cmp & Comparison::U) {
1244 // In case of unsigned comparisons, we can only compare against 0.
1246 // Any x!=0 will be considered >0 in an unsigned comparison.
1247 if (Props & ConstantProperties::Zero)
1248 Result = (Cmp & Comparison::EQ);
1249 else if (Props & ConstantProperties::NonZero)
1250 Result = (Cmp & Comparison::G) || (Cmp == Comparison::NE);
1255 // A2 is not zero. The only handled case is if X = 0.
1256 if (Props & ConstantProperties::Zero) {
1257 Result = (Cmp & Comparison::L) || (Cmp == Comparison::NE);
1263 // Signed comparisons are different.
1264 if (Props & ConstantProperties::Zero) {
1266 Result = (Cmp & Comparison::EQ);
1268 Result = (Cmp == Comparison::NE) ||
1269 ((Cmp & Comparison::L) && !A2.isNegative()) ||
1270 ((Cmp & Comparison::G) && A2.isNegative());
1273 if (Props & ConstantProperties::PosOrZero) {
1274 // X >= 0 and !(A2 < 0) => cannot compare
1275 if (!A2.isNegative())
1277 // X >= 0 and A2 < 0
1278 Result = (Cmp & Comparison::G) || (Cmp == Comparison::NE);
1281 if (Props & ConstantProperties::NegOrZero) {
1282 // X <= 0 and Src1 < 0 => cannot compare
1283 if (A2 == 0 || A2.isNegative())
1285 // X <= 0 and A2 > 0
1286 Result = (Cmp & Comparison::L) || (Cmp == Comparison::NE);
1293 bool MachineConstEvaluator::evaluateCMPpp(uint32_t Cmp, uint32_t Props1,
1294 uint32_t Props2, bool &Result) {
1295 using P = ConstantProperties;
1297 if ((Props1 & P::NaN) && (Props2 & P::NaN))
1299 if (!(Props1 & P::Finite) || !(Props2 & P::Finite))
1302 bool Zero1 = (Props1 & P::Zero), Zero2 = (Props2 & P::Zero);
1303 bool NonZero1 = (Props1 & P::NonZero), NonZero2 = (Props2 & P::NonZero);
1304 if (Zero1 && Zero2) {
1305 Result = (Cmp & Comparison::EQ);
1308 if (Cmp == Comparison::NE) {
1309 if ((Zero1 && NonZero2) || (NonZero1 && Zero2))
1310 return (Result = true);
1314 if (Cmp & Comparison::U) {
1315 // In unsigned comparisons, we can only compare against a known zero,
1316 // or a known non-zero.
1317 if (Zero1 && NonZero2) {
1318 Result = (Cmp & Comparison::L);
1321 if (NonZero1 && Zero2) {
1322 Result = (Cmp & Comparison::G);
1328 // Signed comparison. The comparison is not NE.
1329 bool Poz1 = (Props1 & P::PosOrZero), Poz2 = (Props2 & P::PosOrZero);
1330 bool Nez1 = (Props1 & P::NegOrZero), Nez2 = (Props2 & P::NegOrZero);
1332 if (NonZero1 || NonZero2) {
1333 Result = (Cmp & Comparison::L);
1336 // Either (or both) could be zero. Can only say that X <= Y.
1337 if ((Cmp & Comparison::EQ) && (Cmp & Comparison::L))
1338 return (Result = true);
1341 if (NonZero1 || NonZero2) {
1342 Result = (Cmp & Comparison::G);
1345 // Either (or both) could be zero. Can only say that X >= Y.
1346 if ((Cmp & Comparison::EQ) && (Cmp & Comparison::G))
1347 return (Result = true);
1353 bool MachineConstEvaluator::evaluateCOPY(const Register &R1,
1354 const CellMap &Inputs, LatticeCell &Result) {
1355 return getCell(R1, Inputs, Result);
1358 bool MachineConstEvaluator::evaluateANDrr(const Register &R1,
1359 const Register &R2, const CellMap &Inputs, LatticeCell &Result) {
1360 assert(Inputs.has(R1.Reg) && Inputs.has(R2.Reg));
1361 const LatticeCell &L1 = Inputs.get(R2.Reg);
1362 const LatticeCell &L2 = Inputs.get(R2.Reg);
1363 // If both sources are bottom, exit. Otherwise try to evaluate ANDri
1364 // with the non-bottom argument passed as the immediate. This is to
1365 // catch cases of ANDing with 0.
1366 if (L2.isBottom()) {
1369 return evaluateANDrr(R2, R1, Inputs, Result);
1372 if (!evaluate(R2, L2, LS2))
1374 if (LS2.isBottom() || LS2.isProperty())
1378 for (unsigned i = 0; i < LS2.size(); ++i) {
1380 bool Eval = constToInt(LS2.Values[i], A) &&
1381 evaluateANDri(R1, A, Inputs, RC);
1386 return !Result.isBottom();
1389 bool MachineConstEvaluator::evaluateANDri(const Register &R1,
1390 const APInt &A2, const CellMap &Inputs, LatticeCell &Result) {
1391 assert(Inputs.has(R1.Reg));
1393 return getCell(R1, Inputs, Result);
1396 RC.add(intToConst(A2));
1397 // Overwrite Result.
1402 if (!getCell(R1, Inputs, LS1))
1404 if (LS1.isBottom() || LS1.isProperty())
1408 for (unsigned i = 0; i < LS1.size(); ++i) {
1409 bool Eval = constToInt(LS1.Values[i], A) &&
1410 evaluateANDii(A, A2, ResA);
1413 const Constant *C = intToConst(ResA);
1416 return !Result.isBottom();
1419 bool MachineConstEvaluator::evaluateANDii(const APInt &A1,
1420 const APInt &A2, APInt &Result) {
1425 bool MachineConstEvaluator::evaluateORrr(const Register &R1,
1426 const Register &R2, const CellMap &Inputs, LatticeCell &Result) {
1427 assert(Inputs.has(R1.Reg) && Inputs.has(R2.Reg));
1428 const LatticeCell &L1 = Inputs.get(R2.Reg);
1429 const LatticeCell &L2 = Inputs.get(R2.Reg);
1430 // If both sources are bottom, exit. Otherwise try to evaluate ORri
1431 // with the non-bottom argument passed as the immediate. This is to
1432 // catch cases of ORing with -1.
1433 if (L2.isBottom()) {
1436 return evaluateORrr(R2, R1, Inputs, Result);
1439 if (!evaluate(R2, L2, LS2))
1441 if (LS2.isBottom() || LS2.isProperty())
1445 for (unsigned i = 0; i < LS2.size(); ++i) {
1447 bool Eval = constToInt(LS2.Values[i], A) &&
1448 evaluateORri(R1, A, Inputs, RC);
1453 return !Result.isBottom();
1456 bool MachineConstEvaluator::evaluateORri(const Register &R1,
1457 const APInt &A2, const CellMap &Inputs, LatticeCell &Result) {
1458 assert(Inputs.has(R1.Reg));
1460 return getCell(R1, Inputs, Result);
1463 RC.add(intToConst(A2));
1464 // Overwrite Result.
1469 if (!getCell(R1, Inputs, LS1))
1471 if (LS1.isBottom() || LS1.isProperty())
1475 for (unsigned i = 0; i < LS1.size(); ++i) {
1476 bool Eval = constToInt(LS1.Values[i], A) &&
1477 evaluateORii(A, A2, ResA);
1480 const Constant *C = intToConst(ResA);
1483 return !Result.isBottom();
1486 bool MachineConstEvaluator::evaluateORii(const APInt &A1,
1487 const APInt &A2, APInt &Result) {
1492 bool MachineConstEvaluator::evaluateXORrr(const Register &R1,
1493 const Register &R2, const CellMap &Inputs, LatticeCell &Result) {
1494 assert(Inputs.has(R1.Reg) && Inputs.has(R2.Reg));
1495 LatticeCell LS1, LS2;
1496 if (!getCell(R1, Inputs, LS1) || !getCell(R2, Inputs, LS2))
1498 if (LS1.isProperty()) {
1499 if (LS1.properties() & ConstantProperties::Zero)
1500 return !(Result = LS2).isBottom();
1503 if (LS2.isProperty()) {
1504 if (LS2.properties() & ConstantProperties::Zero)
1505 return !(Result = LS1).isBottom();
1510 for (unsigned i = 0; i < LS2.size(); ++i) {
1512 bool Eval = constToInt(LS2.Values[i], A) &&
1513 evaluateXORri(R1, A, Inputs, RC);
1518 return !Result.isBottom();
1521 bool MachineConstEvaluator::evaluateXORri(const Register &R1,
1522 const APInt &A2, const CellMap &Inputs, LatticeCell &Result) {
1523 assert(Inputs.has(R1.Reg));
1525 if (!getCell(R1, Inputs, LS1))
1527 if (LS1.isProperty()) {
1528 if (LS1.properties() & ConstantProperties::Zero) {
1529 const Constant *C = intToConst(A2);
1531 return !Result.isBottom();
1537 for (unsigned i = 0; i < LS1.size(); ++i) {
1538 bool Eval = constToInt(LS1.Values[i], A) &&
1539 evaluateXORii(A, A2, XA);
1542 const Constant *C = intToConst(XA);
1545 return !Result.isBottom();
1548 bool MachineConstEvaluator::evaluateXORii(const APInt &A1,
1549 const APInt &A2, APInt &Result) {
1554 bool MachineConstEvaluator::evaluateZEXTr(const Register &R1, unsigned Width,
1555 unsigned Bits, const CellMap &Inputs, LatticeCell &Result) {
1556 assert(Inputs.has(R1.Reg));
1558 if (!getCell(R1, Inputs, LS1))
1560 if (LS1.isProperty())
1564 for (unsigned i = 0; i < LS1.size(); ++i) {
1565 bool Eval = constToInt(LS1.Values[i], A) &&
1566 evaluateZEXTi(A, Width, Bits, XA);
1569 const Constant *C = intToConst(XA);
1575 bool MachineConstEvaluator::evaluateZEXTi(const APInt &A1, unsigned Width,
1576 unsigned Bits, APInt &Result) {
1577 unsigned BW = A1.getBitWidth();
1579 assert(Width >= Bits && BW >= Bits);
1580 APInt Mask = APInt::getLowBitsSet(Width, Bits);
1581 Result = A1.zextOrTrunc(Width) & Mask;
1585 bool MachineConstEvaluator::evaluateSEXTr(const Register &R1, unsigned Width,
1586 unsigned Bits, const CellMap &Inputs, LatticeCell &Result) {
1587 assert(Inputs.has(R1.Reg));
1589 if (!getCell(R1, Inputs, LS1))
1591 if (LS1.isBottom() || LS1.isProperty())
1595 for (unsigned i = 0; i < LS1.size(); ++i) {
1596 bool Eval = constToInt(LS1.Values[i], A) &&
1597 evaluateSEXTi(A, Width, Bits, XA);
1600 const Constant *C = intToConst(XA);
1606 bool MachineConstEvaluator::evaluateSEXTi(const APInt &A1, unsigned Width,
1607 unsigned Bits, APInt &Result) {
1608 unsigned BW = A1.getBitWidth();
1609 assert(Width >= Bits && BW >= Bits);
1610 // Special case to make things faster for smaller source widths.
1611 // Sign extension of 0 bits generates 0 as a result. This is consistent
1612 // with what the HW does.
1614 Result = APInt(Width, 0);
1617 // In C, shifts by 64 invoke undefined behavior: handle that case in APInt.
1618 if (BW <= 64 && Bits != 0) {
1619 int64_t V = A1.getSExtValue();
1622 V = static_cast<int8_t>(V);
1625 V = static_cast<int16_t>(V);
1628 V = static_cast<int32_t>(V);
1631 // Shift left to lose all bits except lower "Bits" bits, then shift
1632 // the value back, replicating what was a sign bit after the first
1634 V = (V << (64-Bits)) >> (64-Bits);
1637 // V is a 64-bit sign-extended value. Convert it to APInt of desired
1639 Result = APInt(Width, V, true);
1642 // Slow case: the value doesn't fit in int64_t.
1644 Result = A1.trunc(Bits).sext(Width);
1646 Result = A1.sext(Width);
1650 bool MachineConstEvaluator::evaluateCLBr(const Register &R1, bool Zeros,
1651 bool Ones, const CellMap &Inputs, LatticeCell &Result) {
1652 assert(Inputs.has(R1.Reg));
1654 if (!getCell(R1, Inputs, LS1))
1656 if (LS1.isBottom() || LS1.isProperty())
1660 for (unsigned i = 0; i < LS1.size(); ++i) {
1661 bool Eval = constToInt(LS1.Values[i], A) &&
1662 evaluateCLBi(A, Zeros, Ones, CA);
1665 const Constant *C = intToConst(CA);
1671 bool MachineConstEvaluator::evaluateCLBi(const APInt &A1, bool Zeros,
1672 bool Ones, APInt &Result) {
1673 unsigned BW = A1.getBitWidth();
1674 if (!Zeros && !Ones)
1677 if (Zeros && (Count == 0))
1678 Count = A1.countLeadingZeros();
1679 if (Ones && (Count == 0))
1680 Count = A1.countLeadingOnes();
1681 Result = APInt(BW, static_cast<uint64_t>(Count), false);
1685 bool MachineConstEvaluator::evaluateCTBr(const Register &R1, bool Zeros,
1686 bool Ones, const CellMap &Inputs, LatticeCell &Result) {
1687 assert(Inputs.has(R1.Reg));
1689 if (!getCell(R1, Inputs, LS1))
1691 if (LS1.isBottom() || LS1.isProperty())
1695 for (unsigned i = 0; i < LS1.size(); ++i) {
1696 bool Eval = constToInt(LS1.Values[i], A) &&
1697 evaluateCTBi(A, Zeros, Ones, CA);
1700 const Constant *C = intToConst(CA);
1706 bool MachineConstEvaluator::evaluateCTBi(const APInt &A1, bool Zeros,
1707 bool Ones, APInt &Result) {
1708 unsigned BW = A1.getBitWidth();
1709 if (!Zeros && !Ones)
1712 if (Zeros && (Count == 0))
1713 Count = A1.countTrailingZeros();
1714 if (Ones && (Count == 0))
1715 Count = A1.countTrailingOnes();
1716 Result = APInt(BW, static_cast<uint64_t>(Count), false);
1720 bool MachineConstEvaluator::evaluateEXTRACTr(const Register &R1,
1721 unsigned Width, unsigned Bits, unsigned Offset, bool Signed,
1722 const CellMap &Inputs, LatticeCell &Result) {
1723 assert(Inputs.has(R1.Reg));
1724 assert(Bits+Offset <= Width);
1726 if (!getCell(R1, Inputs, LS1))
1730 if (LS1.isProperty()) {
1731 uint32_t Ps = LS1.properties();
1732 if (Ps & ConstantProperties::Zero) {
1733 const Constant *C = intToConst(APInt(Width, 0, false));
1741 for (unsigned i = 0; i < LS1.size(); ++i) {
1742 bool Eval = constToInt(LS1.Values[i], A) &&
1743 evaluateEXTRACTi(A, Bits, Offset, Signed, CA);
1746 const Constant *C = intToConst(CA);
1752 bool MachineConstEvaluator::evaluateEXTRACTi(const APInt &A1, unsigned Bits,
1753 unsigned Offset, bool Signed, APInt &Result) {
1754 unsigned BW = A1.getBitWidth();
1755 assert(Bits+Offset <= BW);
1756 // Extracting 0 bits generates 0 as a result (as indicated by the HW people).
1758 Result = APInt(BW, 0);
1762 int64_t V = A1.getZExtValue();
1763 V <<= (64-Bits-Offset);
1767 V = static_cast<uint64_t>(V) >> (64-Bits);
1768 Result = APInt(BW, V, Signed);
1772 Result = A1.shl(BW-Bits-Offset).ashr(BW-Bits);
1774 Result = A1.shl(BW-Bits-Offset).lshr(BW-Bits);
1778 bool MachineConstEvaluator::evaluateSplatr(const Register &R1,
1779 unsigned Bits, unsigned Count, const CellMap &Inputs,
1780 LatticeCell &Result) {
1781 assert(Inputs.has(R1.Reg));
1783 if (!getCell(R1, Inputs, LS1))
1785 if (LS1.isBottom() || LS1.isProperty())
1789 for (unsigned i = 0; i < LS1.size(); ++i) {
1790 bool Eval = constToInt(LS1.Values[i], A) &&
1791 evaluateSplati(A, Bits, Count, SA);
1794 const Constant *C = intToConst(SA);
1800 bool MachineConstEvaluator::evaluateSplati(const APInt &A1, unsigned Bits,
1801 unsigned Count, APInt &Result) {
1803 unsigned BW = A1.getBitWidth(), SW = Count*Bits;
1804 APInt LoBits = (Bits < BW) ? A1.trunc(Bits) : A1.zextOrSelf(Bits);
1806 LoBits = LoBits.zext(SW);
1808 APInt Res(SW, 0, false);
1809 for (unsigned i = 0; i < Count; ++i) {
1817 // ----------------------------------------------------------------------
1818 // Hexagon-specific code.
1822 FunctionPass *createHexagonConstPropagationPass();
1823 void initializeHexagonConstPropagationPass(PassRegistry &Registry);
1825 } // end namespace llvm
1829 class HexagonConstEvaluator : public MachineConstEvaluator {
1831 HexagonConstEvaluator(MachineFunction &Fn);
1833 bool evaluate(const MachineInstr &MI, const CellMap &Inputs,
1834 CellMap &Outputs) override;
1835 bool evaluate(const Register &R, const LatticeCell &SrcC,
1836 LatticeCell &Result) override;
1837 bool evaluate(const MachineInstr &BrI, const CellMap &Inputs,
1838 SetVector<const MachineBasicBlock*> &Targets, bool &FallsThru)
1840 bool rewrite(MachineInstr &MI, const CellMap &Inputs) override;
1843 unsigned getRegBitWidth(unsigned Reg) const;
1845 static uint32_t getCmp(unsigned Opc);
1846 static APInt getCmpImm(unsigned Opc, unsigned OpX,
1847 const MachineOperand &MO);
1848 void replaceWithNop(MachineInstr &MI);
1850 bool evaluateHexRSEQ32(Register RL, Register RH, const CellMap &Inputs,
1851 LatticeCell &Result);
1852 bool evaluateHexCompare(const MachineInstr &MI, const CellMap &Inputs,
1854 // This is suitable to be called for compare-and-jump instructions.
1855 bool evaluateHexCompare2(uint32_t Cmp, const MachineOperand &Src1,
1856 const MachineOperand &Src2, const CellMap &Inputs, bool &Result);
1857 bool evaluateHexLogical(const MachineInstr &MI, const CellMap &Inputs,
1859 bool evaluateHexCondMove(const MachineInstr &MI, const CellMap &Inputs,
1861 bool evaluateHexExt(const MachineInstr &MI, const CellMap &Inputs,
1863 bool evaluateHexVector1(const MachineInstr &MI, const CellMap &Inputs,
1865 bool evaluateHexVector2(const MachineInstr &MI, const CellMap &Inputs,
1868 void replaceAllRegUsesWith(unsigned FromReg, unsigned ToReg);
1869 bool rewriteHexBranch(MachineInstr &BrI, const CellMap &Inputs);
1870 bool rewriteHexConstDefs(MachineInstr &MI, const CellMap &Inputs,
1872 bool rewriteHexConstUses(MachineInstr &MI, const CellMap &Inputs);
1874 MachineRegisterInfo *MRI;
1875 const HexagonInstrInfo &HII;
1876 const HexagonRegisterInfo &HRI;
1879 class HexagonConstPropagation : public MachineFunctionPass {
1883 HexagonConstPropagation() : MachineFunctionPass(ID) {
1884 PassRegistry &Registry = *PassRegistry::getPassRegistry();
1885 initializeHexagonConstPropagationPass(Registry);
1888 StringRef getPassName() const override {
1889 return "Hexagon Constant Propagation";
1892 bool runOnMachineFunction(MachineFunction &MF) override {
1893 const Function &F = MF.getFunction();
1894 if (skipFunction(F))
1897 HexagonConstEvaluator HCE(MF);
1898 return MachineConstPropagator(HCE).run(MF);
1902 } // end anonymous namespace
1904 char HexagonConstPropagation::ID = 0;
1906 INITIALIZE_PASS(HexagonConstPropagation, "hcp", "Hexagon Constant Propagation",
1909 HexagonConstEvaluator::HexagonConstEvaluator(MachineFunction &Fn)
1910 : MachineConstEvaluator(Fn),
1911 HII(*Fn.getSubtarget<HexagonSubtarget>().getInstrInfo()),
1912 HRI(*Fn.getSubtarget<HexagonSubtarget>().getRegisterInfo()) {
1913 MRI = &Fn.getRegInfo();
1916 bool HexagonConstEvaluator::evaluate(const MachineInstr &MI,
1917 const CellMap &Inputs, CellMap &Outputs) {
1920 if (MI.getNumOperands() == 0 || !MI.getOperand(0).isReg())
1922 const MachineOperand &MD = MI.getOperand(0);
1926 unsigned Opc = MI.getOpcode();
1928 assert(!DefR.SubReg);
1929 if (!TargetRegisterInfo::isVirtualRegister(DefR.Reg))
1934 Register SrcR(MI.getOperand(1));
1935 bool Eval = evaluateCOPY(SrcR, Inputs, RC);
1938 Outputs.update(DefR.Reg, RC);
1941 if (MI.isRegSequence()) {
1942 unsigned Sub1 = MI.getOperand(2).getImm();
1943 unsigned Sub2 = MI.getOperand(4).getImm();
1944 const TargetRegisterClass &DefRC = *MRI->getRegClass(DefR.Reg);
1945 unsigned SubLo = HRI.getHexagonSubRegIndex(DefRC, Hexagon::ps_sub_lo);
1946 unsigned SubHi = HRI.getHexagonSubRegIndex(DefRC, Hexagon::ps_sub_hi);
1947 if (Sub1 != SubLo && Sub1 != SubHi)
1949 if (Sub2 != SubLo && Sub2 != SubHi)
1951 assert(Sub1 != Sub2);
1952 bool LoIs1 = (Sub1 == SubLo);
1953 const MachineOperand &OpLo = LoIs1 ? MI.getOperand(1) : MI.getOperand(3);
1954 const MachineOperand &OpHi = LoIs1 ? MI.getOperand(3) : MI.getOperand(1);
1956 Register SrcRL(OpLo), SrcRH(OpHi);
1957 bool Eval = evaluateHexRSEQ32(SrcRL, SrcRH, Inputs, RC);
1960 Outputs.update(DefR.Reg, RC);
1963 if (MI.isCompare()) {
1964 bool Eval = evaluateHexCompare(MI, Inputs, Outputs);
1971 case Hexagon::A2_tfrsi:
1972 case Hexagon::A2_tfrpi:
1973 case Hexagon::CONST32:
1974 case Hexagon::CONST64:
1976 const MachineOperand &VO = MI.getOperand(1);
1977 // The operand of CONST32 can be a blockaddress, e.g.
1978 // %0 = CONST32 <blockaddress(@eat, %l)>
1979 // Do this check for all instructions for safety.
1982 int64_t V = MI.getOperand(1).getImm();
1983 unsigned W = getRegBitWidth(DefR.Reg);
1984 if (W != 32 && W != 64)
1986 IntegerType *Ty = (W == 32) ? Type::getInt32Ty(CX)
1987 : Type::getInt64Ty(CX);
1988 const ConstantInt *CI = ConstantInt::get(Ty, V, true);
1989 LatticeCell RC = Outputs.get(DefR.Reg);
1991 Outputs.update(DefR.Reg, RC);
1995 case Hexagon::PS_true:
1996 case Hexagon::PS_false:
1998 LatticeCell RC = Outputs.get(DefR.Reg);
1999 bool NonZero = (Opc == Hexagon::PS_true);
2000 uint32_t P = NonZero ? ConstantProperties::NonZero
2001 : ConstantProperties::Zero;
2003 Outputs.update(DefR.Reg, RC);
2007 case Hexagon::A2_and:
2008 case Hexagon::A2_andir:
2009 case Hexagon::A2_andp:
2010 case Hexagon::A2_or:
2011 case Hexagon::A2_orir:
2012 case Hexagon::A2_orp:
2013 case Hexagon::A2_xor:
2014 case Hexagon::A2_xorp:
2016 bool Eval = evaluateHexLogical(MI, Inputs, Outputs);
2022 case Hexagon::A2_combineii: // combine(#s8Ext, #s8)
2023 case Hexagon::A4_combineii: // combine(#s8, #u6Ext)
2025 uint64_t Hi = MI.getOperand(1).getImm();
2026 uint64_t Lo = MI.getOperand(2).getImm();
2027 uint64_t Res = (Hi << 32) | (Lo & 0xFFFFFFFF);
2028 IntegerType *Ty = Type::getInt64Ty(CX);
2029 const ConstantInt *CI = ConstantInt::get(Ty, Res, false);
2030 LatticeCell RC = Outputs.get(DefR.Reg);
2032 Outputs.update(DefR.Reg, RC);
2036 case Hexagon::S2_setbit_i:
2038 int64_t B = MI.getOperand(2).getImm();
2039 assert(B >=0 && B < 32);
2040 APInt A(32, (1ull << B), false);
2041 Register R(MI.getOperand(1));
2042 LatticeCell RC = Outputs.get(DefR.Reg);
2043 bool Eval = evaluateORri(R, A, Inputs, RC);
2046 Outputs.update(DefR.Reg, RC);
2050 case Hexagon::C2_mux:
2051 case Hexagon::C2_muxir:
2052 case Hexagon::C2_muxri:
2053 case Hexagon::C2_muxii:
2055 bool Eval = evaluateHexCondMove(MI, Inputs, Outputs);
2061 case Hexagon::A2_sxtb:
2062 case Hexagon::A2_sxth:
2063 case Hexagon::A2_sxtw:
2064 case Hexagon::A2_zxtb:
2065 case Hexagon::A2_zxth:
2067 bool Eval = evaluateHexExt(MI, Inputs, Outputs);
2073 case Hexagon::S2_ct0:
2074 case Hexagon::S2_ct0p:
2075 case Hexagon::S2_ct1:
2076 case Hexagon::S2_ct1p:
2078 using namespace Hexagon;
2080 bool Ones = (Opc == S2_ct1) || (Opc == S2_ct1p);
2081 Register R1(MI.getOperand(1));
2082 assert(Inputs.has(R1.Reg));
2084 bool Eval = evaluateCTBr(R1, !Ones, Ones, Inputs, T);
2087 // All of these instructions return a 32-bit value. The evaluate
2088 // will generate the same type as the operand, so truncate the
2089 // result if necessary.
2091 LatticeCell RC = Outputs.get(DefR.Reg);
2092 for (unsigned i = 0; i < T.size(); ++i) {
2093 const Constant *CI = T.Values[i];
2094 if (constToInt(CI, C) && C.getBitWidth() > 32)
2095 CI = intToConst(C.trunc(32));
2098 Outputs.update(DefR.Reg, RC);
2102 case Hexagon::S2_cl0:
2103 case Hexagon::S2_cl0p:
2104 case Hexagon::S2_cl1:
2105 case Hexagon::S2_cl1p:
2106 case Hexagon::S2_clb:
2107 case Hexagon::S2_clbp:
2109 using namespace Hexagon;
2111 bool OnlyZeros = (Opc == S2_cl0) || (Opc == S2_cl0p);
2112 bool OnlyOnes = (Opc == S2_cl1) || (Opc == S2_cl1p);
2113 Register R1(MI.getOperand(1));
2114 assert(Inputs.has(R1.Reg));
2116 bool Eval = evaluateCLBr(R1, !OnlyOnes, !OnlyZeros, Inputs, T);
2119 // All of these instructions return a 32-bit value. The evaluate
2120 // will generate the same type as the operand, so truncate the
2121 // result if necessary.
2123 LatticeCell RC = Outputs.get(DefR.Reg);
2124 for (unsigned i = 0; i < T.size(); ++i) {
2125 const Constant *CI = T.Values[i];
2126 if (constToInt(CI, C) && C.getBitWidth() > 32)
2127 CI = intToConst(C.trunc(32));
2130 Outputs.update(DefR.Reg, RC);
2134 case Hexagon::S4_extract:
2135 case Hexagon::S4_extractp:
2136 case Hexagon::S2_extractu:
2137 case Hexagon::S2_extractup:
2139 bool Signed = (Opc == Hexagon::S4_extract) ||
2140 (Opc == Hexagon::S4_extractp);
2141 Register R1(MI.getOperand(1));
2142 unsigned BW = getRegBitWidth(R1.Reg);
2143 unsigned Bits = MI.getOperand(2).getImm();
2144 unsigned Offset = MI.getOperand(3).getImm();
2145 LatticeCell RC = Outputs.get(DefR.Reg);
2147 APInt Zero(BW, 0, false);
2148 RC.add(intToConst(Zero));
2151 if (Offset+Bits > BW) {
2152 // If the requested bitfield extends beyond the most significant bit,
2153 // the extra bits are treated as 0s. To emulate this behavior, reduce
2154 // the number of requested bits, and make the extract unsigned.
2158 bool Eval = evaluateEXTRACTr(R1, BW, Bits, Offset, Signed, Inputs, RC);
2161 Outputs.update(DefR.Reg, RC);
2165 case Hexagon::S2_vsplatrb:
2166 case Hexagon::S2_vsplatrh:
2170 // vrndwh, vrndwh:sat
2171 // vsathb, vsathub, vsatwuh
2173 // vtrunehb, vtrunohb
2176 bool Eval = evaluateHexVector1(MI, Inputs, Outputs);
2192 bool HexagonConstEvaluator::evaluate(const Register &R,
2193 const LatticeCell &Input, LatticeCell &Result) {
2198 const TargetRegisterClass *RC = MRI->getRegClass(R.Reg);
2199 if (RC != &Hexagon::DoubleRegsRegClass)
2201 if (R.SubReg != Hexagon::isub_lo && R.SubReg != Hexagon::isub_hi)
2204 assert(!Input.isTop());
2205 if (Input.isBottom())
2208 using P = ConstantProperties;
2210 if (Input.isProperty()) {
2211 uint32_t Ps = Input.properties();
2212 if (Ps & (P::Zero|P::NaN)) {
2213 uint32_t Ns = (Ps & (P::Zero|P::NaN|P::SignProperties));
2217 if (R.SubReg == Hexagon::isub_hi) {
2218 uint32_t Ns = (Ps & P::SignProperties);
2225 // The Input cell contains some known values. Pick the word corresponding
2226 // to the subregister.
2228 for (unsigned i = 0; i < Input.size(); ++i) {
2229 const Constant *C = Input.Values[i];
2230 if (!constToInt(C, A))
2234 uint64_t U = A.getZExtValue();
2235 if (R.SubReg == Hexagon::isub_hi)
2238 uint32_t U32 = Lo_32(U);
2240 memcpy(&V32, &U32, sizeof V32);
2241 IntegerType *Ty = Type::getInt32Ty(CX);
2242 const ConstantInt *C32 = ConstantInt::get(Ty, static_cast<int64_t>(V32));
2248 bool HexagonConstEvaluator::evaluate(const MachineInstr &BrI,
2249 const CellMap &Inputs, SetVector<const MachineBasicBlock*> &Targets,
2251 // We need to evaluate one branch at a time. TII::analyzeBranch checks
2252 // all the branches in a basic block at once, so we cannot use it.
2253 unsigned Opc = BrI.getOpcode();
2254 bool SimpleBranch = false;
2255 bool Negated = false;
2257 case Hexagon::J2_jumpf:
2258 case Hexagon::J2_jumpfnew:
2259 case Hexagon::J2_jumpfnewpt:
2262 case Hexagon::J2_jumpt:
2263 case Hexagon::J2_jumptnew:
2264 case Hexagon::J2_jumptnewpt:
2265 // Simple branch: if([!]Pn) jump ...
2266 // i.e. Op0 = predicate, Op1 = branch target.
2267 SimpleBranch = true;
2269 case Hexagon::J2_jump:
2270 Targets.insert(BrI.getOperand(0).getMBB());
2275 // If the branch is of unknown type, assume that all successors are
2277 FallsThru = !BrI.isUnconditionalBranch();
2282 const MachineOperand &MD = BrI.getOperand(0);
2284 // If the condition operand has a subregister, this is not something
2285 // we currently recognize.
2288 assert(Inputs.has(PR.Reg));
2289 const LatticeCell &PredC = Inputs.get(PR.Reg);
2290 if (PredC.isBottom())
2293 uint32_t Props = PredC.properties();
2294 bool CTrue = false, CFalse = false;
2295 if (Props & ConstantProperties::Zero)
2297 else if (Props & ConstantProperties::NonZero)
2299 // If the condition is not known to be either, bail out.
2300 if (!CTrue && !CFalse)
2303 const MachineBasicBlock *BranchTarget = BrI.getOperand(1).getMBB();
2306 if ((!Negated && CTrue) || (Negated && CFalse))
2307 Targets.insert(BranchTarget);
2308 else if ((!Negated && CFalse) || (Negated && CTrue))
2317 bool HexagonConstEvaluator::rewrite(MachineInstr &MI, const CellMap &Inputs) {
2319 return rewriteHexBranch(MI, Inputs);
2321 unsigned Opc = MI.getOpcode();
2325 case Hexagon::A2_tfrsi:
2326 case Hexagon::A2_tfrpi:
2327 case Hexagon::CONST32:
2328 case Hexagon::CONST64:
2329 case Hexagon::PS_true:
2330 case Hexagon::PS_false:
2334 unsigned NumOp = MI.getNumOperands();
2338 bool AllDefs, Changed;
2339 Changed = rewriteHexConstDefs(MI, Inputs, AllDefs);
2340 // If not all defs have been rewritten (i.e. the instruction defines
2341 // a register that is not compile-time constant), then try to rewrite
2342 // register operands that are known to be constant with immediates.
2344 Changed |= rewriteHexConstUses(MI, Inputs);
2349 unsigned HexagonConstEvaluator::getRegBitWidth(unsigned Reg) const {
2350 const TargetRegisterClass *RC = MRI->getRegClass(Reg);
2351 if (Hexagon::IntRegsRegClass.hasSubClassEq(RC))
2353 if (Hexagon::DoubleRegsRegClass.hasSubClassEq(RC))
2355 if (Hexagon::PredRegsRegClass.hasSubClassEq(RC))
2357 llvm_unreachable("Invalid register");
2361 uint32_t HexagonConstEvaluator::getCmp(unsigned Opc) {
2363 case Hexagon::C2_cmpeq:
2364 case Hexagon::C2_cmpeqp:
2365 case Hexagon::A4_cmpbeq:
2366 case Hexagon::A4_cmpheq:
2367 case Hexagon::A4_cmpbeqi:
2368 case Hexagon::A4_cmpheqi:
2369 case Hexagon::C2_cmpeqi:
2370 case Hexagon::J4_cmpeqn1_t_jumpnv_nt:
2371 case Hexagon::J4_cmpeqn1_t_jumpnv_t:
2372 case Hexagon::J4_cmpeqi_t_jumpnv_nt:
2373 case Hexagon::J4_cmpeqi_t_jumpnv_t:
2374 case Hexagon::J4_cmpeq_t_jumpnv_nt:
2375 case Hexagon::J4_cmpeq_t_jumpnv_t:
2376 return Comparison::EQ;
2378 case Hexagon::C4_cmpneq:
2379 case Hexagon::C4_cmpneqi:
2380 case Hexagon::J4_cmpeqn1_f_jumpnv_nt:
2381 case Hexagon::J4_cmpeqn1_f_jumpnv_t:
2382 case Hexagon::J4_cmpeqi_f_jumpnv_nt:
2383 case Hexagon::J4_cmpeqi_f_jumpnv_t:
2384 case Hexagon::J4_cmpeq_f_jumpnv_nt:
2385 case Hexagon::J4_cmpeq_f_jumpnv_t:
2386 return Comparison::NE;
2388 case Hexagon::C2_cmpgt:
2389 case Hexagon::C2_cmpgtp:
2390 case Hexagon::A4_cmpbgt:
2391 case Hexagon::A4_cmphgt:
2392 case Hexagon::A4_cmpbgti:
2393 case Hexagon::A4_cmphgti:
2394 case Hexagon::C2_cmpgti:
2395 case Hexagon::J4_cmpgtn1_t_jumpnv_nt:
2396 case Hexagon::J4_cmpgtn1_t_jumpnv_t:
2397 case Hexagon::J4_cmpgti_t_jumpnv_nt:
2398 case Hexagon::J4_cmpgti_t_jumpnv_t:
2399 case Hexagon::J4_cmpgt_t_jumpnv_nt:
2400 case Hexagon::J4_cmpgt_t_jumpnv_t:
2401 return Comparison::GTs;
2403 case Hexagon::C4_cmplte:
2404 case Hexagon::C4_cmpltei:
2405 case Hexagon::J4_cmpgtn1_f_jumpnv_nt:
2406 case Hexagon::J4_cmpgtn1_f_jumpnv_t:
2407 case Hexagon::J4_cmpgti_f_jumpnv_nt:
2408 case Hexagon::J4_cmpgti_f_jumpnv_t:
2409 case Hexagon::J4_cmpgt_f_jumpnv_nt:
2410 case Hexagon::J4_cmpgt_f_jumpnv_t:
2411 return Comparison::LEs;
2413 case Hexagon::C2_cmpgtu:
2414 case Hexagon::C2_cmpgtup:
2415 case Hexagon::A4_cmpbgtu:
2416 case Hexagon::A4_cmpbgtui:
2417 case Hexagon::A4_cmphgtu:
2418 case Hexagon::A4_cmphgtui:
2419 case Hexagon::C2_cmpgtui:
2420 case Hexagon::J4_cmpgtui_t_jumpnv_nt:
2421 case Hexagon::J4_cmpgtui_t_jumpnv_t:
2422 case Hexagon::J4_cmpgtu_t_jumpnv_nt:
2423 case Hexagon::J4_cmpgtu_t_jumpnv_t:
2424 return Comparison::GTu;
2426 case Hexagon::J4_cmpltu_f_jumpnv_nt:
2427 case Hexagon::J4_cmpltu_f_jumpnv_t:
2428 return Comparison::GEu;
2430 case Hexagon::J4_cmpltu_t_jumpnv_nt:
2431 case Hexagon::J4_cmpltu_t_jumpnv_t:
2432 return Comparison::LTu;
2434 case Hexagon::J4_cmplt_f_jumpnv_nt:
2435 case Hexagon::J4_cmplt_f_jumpnv_t:
2436 return Comparison::GEs;
2438 case Hexagon::C4_cmplteu:
2439 case Hexagon::C4_cmplteui:
2440 case Hexagon::J4_cmpgtui_f_jumpnv_nt:
2441 case Hexagon::J4_cmpgtui_f_jumpnv_t:
2442 case Hexagon::J4_cmpgtu_f_jumpnv_nt:
2443 case Hexagon::J4_cmpgtu_f_jumpnv_t:
2444 return Comparison::LEu;
2446 case Hexagon::J4_cmplt_t_jumpnv_nt:
2447 case Hexagon::J4_cmplt_t_jumpnv_t:
2448 return Comparison::LTs;
2453 return Comparison::Unk;
2456 APInt HexagonConstEvaluator::getCmpImm(unsigned Opc, unsigned OpX,
2457 const MachineOperand &MO) {
2458 bool Signed = false;
2460 case Hexagon::A4_cmpbgtui: // u7
2461 case Hexagon::A4_cmphgtui: // u7
2463 case Hexagon::A4_cmpheqi: // s8
2464 case Hexagon::C4_cmpneqi: // s8
2466 case Hexagon::A4_cmpbeqi: // u8
2468 case Hexagon::C2_cmpgtui: // u9
2469 case Hexagon::C4_cmplteui: // u9
2471 case Hexagon::C2_cmpeqi: // s10
2472 case Hexagon::C2_cmpgti: // s10
2473 case Hexagon::C4_cmpltei: // s10
2476 case Hexagon::J4_cmpeqi_f_jumpnv_nt: // u5
2477 case Hexagon::J4_cmpeqi_f_jumpnv_t: // u5
2478 case Hexagon::J4_cmpeqi_t_jumpnv_nt: // u5
2479 case Hexagon::J4_cmpeqi_t_jumpnv_t: // u5
2480 case Hexagon::J4_cmpgti_f_jumpnv_nt: // u5
2481 case Hexagon::J4_cmpgti_f_jumpnv_t: // u5
2482 case Hexagon::J4_cmpgti_t_jumpnv_nt: // u5
2483 case Hexagon::J4_cmpgti_t_jumpnv_t: // u5
2484 case Hexagon::J4_cmpgtui_f_jumpnv_nt: // u5
2485 case Hexagon::J4_cmpgtui_f_jumpnv_t: // u5
2486 case Hexagon::J4_cmpgtui_t_jumpnv_nt: // u5
2487 case Hexagon::J4_cmpgtui_t_jumpnv_t: // u5
2490 llvm_unreachable("Unhandled instruction");
2494 uint64_t Val = MO.getImm();
2495 return APInt(32, Val, Signed);
2498 void HexagonConstEvaluator::replaceWithNop(MachineInstr &MI) {
2499 MI.setDesc(HII.get(Hexagon::A2_nop));
2500 while (MI.getNumOperands() > 0)
2501 MI.RemoveOperand(0);
2504 bool HexagonConstEvaluator::evaluateHexRSEQ32(Register RL, Register RH,
2505 const CellMap &Inputs, LatticeCell &Result) {
2506 assert(Inputs.has(RL.Reg) && Inputs.has(RH.Reg));
2507 LatticeCell LSL, LSH;
2508 if (!getCell(RL, Inputs, LSL) || !getCell(RH, Inputs, LSH))
2510 if (LSL.isProperty() || LSH.isProperty())
2513 unsigned LN = LSL.size(), HN = LSH.size();
2514 SmallVector<APInt,4> LoVs(LN), HiVs(HN);
2515 for (unsigned i = 0; i < LN; ++i) {
2516 bool Eval = constToInt(LSL.Values[i], LoVs[i]);
2519 assert(LoVs[i].getBitWidth() == 32);
2521 for (unsigned i = 0; i < HN; ++i) {
2522 bool Eval = constToInt(LSH.Values[i], HiVs[i]);
2525 assert(HiVs[i].getBitWidth() == 32);
2528 for (unsigned i = 0; i < HiVs.size(); ++i) {
2529 APInt HV = HiVs[i].zextOrSelf(64) << 32;
2530 for (unsigned j = 0; j < LoVs.size(); ++j) {
2531 APInt LV = LoVs[j].zextOrSelf(64);
2532 const Constant *C = intToConst(HV | LV);
2534 if (Result.isBottom())
2538 return !Result.isBottom();
2541 bool HexagonConstEvaluator::evaluateHexCompare(const MachineInstr &MI,
2542 const CellMap &Inputs, CellMap &Outputs) {
2543 unsigned Opc = MI.getOpcode();
2544 bool Classic = false;
2546 case Hexagon::C2_cmpeq:
2547 case Hexagon::C2_cmpeqp:
2548 case Hexagon::C2_cmpgt:
2549 case Hexagon::C2_cmpgtp:
2550 case Hexagon::C2_cmpgtu:
2551 case Hexagon::C2_cmpgtup:
2552 case Hexagon::C2_cmpeqi:
2553 case Hexagon::C2_cmpgti:
2554 case Hexagon::C2_cmpgtui:
2555 // Classic compare: Dst0 = CMP Src1, Src2
2559 // Not handling other compare instructions now.
2564 const MachineOperand &Src1 = MI.getOperand(1);
2565 const MachineOperand &Src2 = MI.getOperand(2);
2568 unsigned Opc = MI.getOpcode();
2569 bool Computed = evaluateHexCompare2(Opc, Src1, Src2, Inputs, Result);
2571 // Only create a zero/non-zero cell. At this time there isn't really
2572 // much need for specific values.
2573 Register DefR(MI.getOperand(0));
2574 LatticeCell L = Outputs.get(DefR.Reg);
2575 uint32_t P = Result ? ConstantProperties::NonZero
2576 : ConstantProperties::Zero;
2578 Outputs.update(DefR.Reg, L);
2586 bool HexagonConstEvaluator::evaluateHexCompare2(unsigned Opc,
2587 const MachineOperand &Src1, const MachineOperand &Src2,
2588 const CellMap &Inputs, bool &Result) {
2589 uint32_t Cmp = getCmp(Opc);
2590 bool Reg1 = Src1.isReg(), Reg2 = Src2.isReg();
2591 bool Imm1 = Src1.isImm(), Imm2 = Src2.isImm();
2596 return evaluateCMPrr(Cmp, R1, R2, Inputs, Result);
2598 APInt A2 = getCmpImm(Opc, 2, Src2);
2599 return evaluateCMPri(Cmp, R1, A2, Inputs, Result);
2602 APInt A1 = getCmpImm(Opc, 1, Src1);
2605 uint32_t NegCmp = Comparison::negate(Cmp);
2606 return evaluateCMPri(NegCmp, R2, A1, Inputs, Result);
2608 APInt A2 = getCmpImm(Opc, 2, Src2);
2609 return evaluateCMPii(Cmp, A1, A2, Result);
2612 // Unknown kind of comparison.
2616 bool HexagonConstEvaluator::evaluateHexLogical(const MachineInstr &MI,
2617 const CellMap &Inputs, CellMap &Outputs) {
2618 unsigned Opc = MI.getOpcode();
2619 if (MI.getNumOperands() != 3)
2621 const MachineOperand &Src1 = MI.getOperand(1);
2622 const MachineOperand &Src2 = MI.getOperand(2);
2629 case Hexagon::A2_and:
2630 case Hexagon::A2_andp:
2631 Eval = evaluateANDrr(R1, Register(Src2), Inputs, RC);
2633 case Hexagon::A2_andir: {
2634 APInt A(32, Src2.getImm(), true);
2635 Eval = evaluateANDri(R1, A, Inputs, RC);
2638 case Hexagon::A2_or:
2639 case Hexagon::A2_orp:
2640 Eval = evaluateORrr(R1, Register(Src2), Inputs, RC);
2642 case Hexagon::A2_orir: {
2643 APInt A(32, Src2.getImm(), true);
2644 Eval = evaluateORri(R1, A, Inputs, RC);
2647 case Hexagon::A2_xor:
2648 case Hexagon::A2_xorp:
2649 Eval = evaluateXORrr(R1, Register(Src2), Inputs, RC);
2653 Register DefR(MI.getOperand(0));
2654 Outputs.update(DefR.Reg, RC);
2659 bool HexagonConstEvaluator::evaluateHexCondMove(const MachineInstr &MI,
2660 const CellMap &Inputs, CellMap &Outputs) {
2661 // Dst0 = Cond1 ? Src2 : Src3
2662 Register CR(MI.getOperand(1));
2663 assert(Inputs.has(CR.Reg));
2665 if (!getCell(CR, Inputs, LS))
2667 uint32_t Ps = LS.properties();
2669 if (Ps & ConstantProperties::Zero)
2671 else if (Ps & ConstantProperties::NonZero)
2676 const MachineOperand &ValOp = MI.getOperand(TakeOp);
2677 Register DefR(MI.getOperand(0));
2678 LatticeCell RC = Outputs.get(DefR.Reg);
2680 if (ValOp.isImm()) {
2681 int64_t V = ValOp.getImm();
2682 unsigned W = getRegBitWidth(DefR.Reg);
2683 APInt A(W, V, true);
2684 const Constant *C = intToConst(A);
2686 Outputs.update(DefR.Reg, RC);
2689 if (ValOp.isReg()) {
2691 const LatticeCell &LR = Inputs.get(R.Reg);
2693 if (!evaluate(R, LR, LSR))
2696 Outputs.update(DefR.Reg, RC);
2702 bool HexagonConstEvaluator::evaluateHexExt(const MachineInstr &MI,
2703 const CellMap &Inputs, CellMap &Outputs) {
2705 Register R1(MI.getOperand(1));
2706 assert(Inputs.has(R1.Reg));
2708 unsigned Opc = MI.getOpcode();
2711 case Hexagon::A2_sxtb:
2712 case Hexagon::A2_zxtb:
2715 case Hexagon::A2_sxth:
2716 case Hexagon::A2_zxth:
2719 case Hexagon::A2_sxtw:
2724 bool Signed = false;
2726 case Hexagon::A2_sxtb:
2727 case Hexagon::A2_sxth:
2728 case Hexagon::A2_sxtw:
2733 Register DefR(MI.getOperand(0));
2734 unsigned BW = getRegBitWidth(DefR.Reg);
2735 LatticeCell RC = Outputs.get(DefR.Reg);
2736 bool Eval = Signed ? evaluateSEXTr(R1, BW, Bits, Inputs, RC)
2737 : evaluateZEXTr(R1, BW, Bits, Inputs, RC);
2740 Outputs.update(DefR.Reg, RC);
2744 bool HexagonConstEvaluator::evaluateHexVector1(const MachineInstr &MI,
2745 const CellMap &Inputs, CellMap &Outputs) {
2747 Register DefR(MI.getOperand(0));
2748 Register R1(MI.getOperand(1));
2749 assert(Inputs.has(R1.Reg));
2750 LatticeCell RC = Outputs.get(DefR.Reg);
2753 unsigned Opc = MI.getOpcode();
2755 case Hexagon::S2_vsplatrb:
2756 // Rd = 4 times Rs:0..7
2757 Eval = evaluateSplatr(R1, 8, 4, Inputs, RC);
2759 case Hexagon::S2_vsplatrh:
2760 // Rdd = 4 times Rs:0..15
2761 Eval = evaluateSplatr(R1, 16, 4, Inputs, RC);
2769 Outputs.update(DefR.Reg, RC);
2773 bool HexagonConstEvaluator::rewriteHexConstDefs(MachineInstr &MI,
2774 const CellMap &Inputs, bool &AllDefs) {
2777 // Some diagnostics.
2778 // DEBUG({...}) gets confused with all this code as an argument.
2780 bool Debugging = DebugFlag && isCurrentDebugType(DEBUG_TYPE);
2782 bool Const = true, HasUse = false;
2783 for (const MachineOperand &MO : MI.operands()) {
2784 if (!MO.isReg() || !MO.isUse() || MO.isImplicit())
2787 if (!TargetRegisterInfo::isVirtualRegister(R.Reg))
2790 // PHIs can legitimately have "top" cells after propagation.
2791 if (!MI.isPHI() && !Inputs.has(R.Reg)) {
2792 dbgs() << "Top " << printReg(R.Reg, &HRI, R.SubReg)
2793 << " in MI: " << MI;
2796 const LatticeCell &L = Inputs.get(R.Reg);
2797 Const &= L.isSingle();
2801 if (HasUse && Const) {
2803 dbgs() << "CONST: " << MI;
2804 for (const MachineOperand &MO : MI.operands()) {
2805 if (!MO.isReg() || !MO.isUse() || MO.isImplicit())
2807 unsigned R = MO.getReg();
2808 dbgs() << printReg(R, &TRI) << ": " << Inputs.get(R) << "\n";
2815 // Avoid generating TFRIs for register transfers---this will keep the
2816 // coalescing opportunities.
2820 // Collect all virtual register-def operands.
2821 SmallVector<unsigned,2> DefRegs;
2822 for (const MachineOperand &MO : MI.operands()) {
2823 if (!MO.isReg() || !MO.isDef())
2825 unsigned R = MO.getReg();
2826 if (!TargetRegisterInfo::isVirtualRegister(R))
2828 assert(!MO.getSubReg());
2829 assert(Inputs.has(R));
2830 DefRegs.push_back(R);
2833 MachineBasicBlock &B = *MI.getParent();
2834 const DebugLoc &DL = MI.getDebugLoc();
2835 unsigned ChangedNum = 0;
2837 SmallVector<const MachineInstr*,4> NewInstrs;
2840 // For each defined register, if it is a constant, create an instruction
2842 // and replace all uses of the defined register with NewR.
2843 for (unsigned i = 0, n = DefRegs.size(); i < n; ++i) {
2844 unsigned R = DefRegs[i];
2845 const LatticeCell &L = Inputs.get(R);
2848 const TargetRegisterClass *RC = MRI->getRegClass(R);
2849 MachineBasicBlock::iterator At = MI.getIterator();
2851 if (!L.isSingle()) {
2852 // If this a zero/non-zero cell, we can fold a definition
2853 // of a predicate register.
2854 using P = ConstantProperties;
2856 uint64_t Ps = L.properties();
2857 if (!(Ps & (P::Zero|P::NonZero)))
2859 const TargetRegisterClass *PredRC = &Hexagon::PredRegsRegClass;
2862 const MCInstrDesc *NewD = (Ps & P::Zero) ?
2863 &HII.get(Hexagon::PS_false) :
2864 &HII.get(Hexagon::PS_true);
2865 unsigned NewR = MRI->createVirtualRegister(PredRC);
2866 const MachineInstrBuilder &MIB = BuildMI(B, At, DL, *NewD, NewR);
2869 NewInstrs.push_back(&*MIB);
2871 replaceAllRegUsesWith(R, NewR);
2873 // This cell has a single value.
2875 if (!constToInt(L.Value, A) || !A.isSignedIntN(64))
2877 const TargetRegisterClass *NewRC;
2878 const MCInstrDesc *NewD;
2880 unsigned W = getRegBitWidth(R);
2881 int64_t V = A.getSExtValue();
2882 assert(W == 32 || W == 64);
2884 NewRC = &Hexagon::IntRegsRegClass;
2886 NewRC = &Hexagon::DoubleRegsRegClass;
2887 unsigned NewR = MRI->createVirtualRegister(NewRC);
2888 const MachineInstr *NewMI;
2891 NewD = &HII.get(Hexagon::A2_tfrsi);
2892 NewMI = BuildMI(B, At, DL, *NewD, NewR)
2895 if (A.isSignedIntN(8)) {
2896 NewD = &HII.get(Hexagon::A2_tfrpi);
2897 NewMI = BuildMI(B, At, DL, *NewD, NewR)
2900 int32_t Hi = V >> 32;
2901 int32_t Lo = V & 0xFFFFFFFFLL;
2902 if (isInt<8>(Hi) && isInt<8>(Lo)) {
2903 NewD = &HII.get(Hexagon::A2_combineii);
2904 NewMI = BuildMI(B, At, DL, *NewD, NewR)
2908 NewD = &HII.get(Hexagon::CONST64);
2909 NewMI = BuildMI(B, At, DL, *NewD, NewR)
2916 NewInstrs.push_back(NewMI);
2918 replaceAllRegUsesWith(R, NewR);
2924 if (!NewInstrs.empty()) {
2925 MachineFunction &MF = *MI.getParent()->getParent();
2926 dbgs() << "In function: " << MF.getName() << "\n";
2927 dbgs() << "Rewrite: for " << MI << " created " << *NewInstrs[0];
2928 for (unsigned i = 1; i < NewInstrs.size(); ++i)
2929 dbgs() << " " << *NewInstrs[i];
2933 AllDefs = (ChangedNum == DefRegs.size());
2934 return ChangedNum > 0;
2937 bool HexagonConstEvaluator::rewriteHexConstUses(MachineInstr &MI,
2938 const CellMap &Inputs) {
2939 bool Changed = false;
2940 unsigned Opc = MI.getOpcode();
2941 MachineBasicBlock &B = *MI.getParent();
2942 const DebugLoc &DL = MI.getDebugLoc();
2943 MachineBasicBlock::iterator At = MI.getIterator();
2944 MachineInstr *NewMI = nullptr;
2947 case Hexagon::M2_maci:
2948 // Convert DefR += mpyi(R2, R3)
2949 // to DefR += mpyi(R, #imm),
2950 // or DefR -= mpyi(R, #imm).
2952 Register DefR(MI.getOperand(0));
2953 assert(!DefR.SubReg);
2954 Register R2(MI.getOperand(2));
2955 Register R3(MI.getOperand(3));
2956 assert(Inputs.has(R2.Reg) && Inputs.has(R3.Reg));
2957 LatticeCell LS2, LS3;
2958 // It is enough to get one of the input cells, since we will only try
2959 // to replace one argument---whichever happens to be a single constant.
2960 bool HasC2 = getCell(R2, Inputs, LS2), HasC3 = getCell(R3, Inputs, LS3);
2961 if (!HasC2 && !HasC3)
2963 bool Zero = ((HasC2 && (LS2.properties() & ConstantProperties::Zero)) ||
2964 (HasC3 && (LS3.properties() & ConstantProperties::Zero)));
2965 // If one of the operands is zero, eliminate the multiplication.
2967 // DefR == R1 (tied operands).
2968 MachineOperand &Acc = MI.getOperand(1);
2970 unsigned NewR = R1.Reg;
2972 // Generate COPY. FIXME: Replace with the register:subregister.
2973 const TargetRegisterClass *RC = MRI->getRegClass(DefR.Reg);
2974 NewR = MRI->createVirtualRegister(RC);
2975 NewMI = BuildMI(B, At, DL, HII.get(TargetOpcode::COPY), NewR)
2976 .addReg(R1.Reg, getRegState(Acc), R1.SubReg);
2978 replaceAllRegUsesWith(DefR.Reg, NewR);
2979 MRI->clearKillFlags(NewR);
2985 if (!LS3.isSingle()) {
2986 if (!LS2.isSingle())
2990 const LatticeCell &LI = Swap ? LS2 : LS3;
2991 const MachineOperand &OpR2 = Swap ? MI.getOperand(3)
2993 // LI is single here.
2995 if (!constToInt(LI.Value, A) || !A.isSignedIntN(8))
2997 int64_t V = A.getSExtValue();
2998 const MCInstrDesc &D = (V >= 0) ? HII.get(Hexagon::M2_macsip)
2999 : HII.get(Hexagon::M2_macsin);
3002 const TargetRegisterClass *RC = MRI->getRegClass(DefR.Reg);
3003 unsigned NewR = MRI->createVirtualRegister(RC);
3004 const MachineOperand &Src1 = MI.getOperand(1);
3005 NewMI = BuildMI(B, At, DL, D, NewR)
3006 .addReg(Src1.getReg(), getRegState(Src1), Src1.getSubReg())
3007 .addReg(OpR2.getReg(), getRegState(OpR2), OpR2.getSubReg())
3009 replaceAllRegUsesWith(DefR.Reg, NewR);
3014 case Hexagon::A2_and:
3016 Register R1(MI.getOperand(1));
3017 Register R2(MI.getOperand(2));
3018 assert(Inputs.has(R1.Reg) && Inputs.has(R2.Reg));
3019 LatticeCell LS1, LS2;
3020 unsigned CopyOf = 0;
3021 // Check if any of the operands is -1 (i.e. all bits set).
3022 if (getCell(R1, Inputs, LS1) && LS1.isSingle()) {
3024 if (constToInt(LS1.Value, M1) && !~M1)
3027 else if (getCell(R2, Inputs, LS2) && LS2.isSingle()) {
3029 if (constToInt(LS2.Value, M1) && !~M1)
3034 MachineOperand &SO = MI.getOperand(CopyOf);
3036 Register DefR(MI.getOperand(0));
3037 unsigned NewR = SR.Reg;
3039 const TargetRegisterClass *RC = MRI->getRegClass(DefR.Reg);
3040 NewR = MRI->createVirtualRegister(RC);
3041 NewMI = BuildMI(B, At, DL, HII.get(TargetOpcode::COPY), NewR)
3042 .addReg(SR.Reg, getRegState(SO), SR.SubReg);
3044 replaceAllRegUsesWith(DefR.Reg, NewR);
3045 MRI->clearKillFlags(NewR);
3050 case Hexagon::A2_or:
3052 Register R1(MI.getOperand(1));
3053 Register R2(MI.getOperand(2));
3054 assert(Inputs.has(R1.Reg) && Inputs.has(R2.Reg));
3055 LatticeCell LS1, LS2;
3056 unsigned CopyOf = 0;
3058 using P = ConstantProperties;
3060 if (getCell(R1, Inputs, LS1) && (LS1.properties() & P::Zero))
3062 else if (getCell(R2, Inputs, LS2) && (LS2.properties() & P::Zero))
3066 MachineOperand &SO = MI.getOperand(CopyOf);
3068 Register DefR(MI.getOperand(0));
3069 unsigned NewR = SR.Reg;
3071 const TargetRegisterClass *RC = MRI->getRegClass(DefR.Reg);
3072 NewR = MRI->createVirtualRegister(RC);
3073 NewMI = BuildMI(B, At, DL, HII.get(TargetOpcode::COPY), NewR)
3074 .addReg(SR.Reg, getRegState(SO), SR.SubReg);
3076 replaceAllRegUsesWith(DefR.Reg, NewR);
3077 MRI->clearKillFlags(NewR);
3084 // clear all the kill flags of this new instruction.
3085 for (MachineOperand &MO : NewMI->operands())
3086 if (MO.isReg() && MO.isUse())
3087 MO.setIsKill(false);
3092 dbgs() << "Rewrite: for " << MI;
3094 dbgs() << " created " << *NewMI;
3096 dbgs() << " modified the instruction itself and created:" << *NewMI;
3103 void HexagonConstEvaluator::replaceAllRegUsesWith(unsigned FromReg,
3105 assert(TargetRegisterInfo::isVirtualRegister(FromReg));
3106 assert(TargetRegisterInfo::isVirtualRegister(ToReg));
3107 for (auto I = MRI->use_begin(FromReg), E = MRI->use_end(); I != E;) {
3108 MachineOperand &O = *I;
3114 bool HexagonConstEvaluator::rewriteHexBranch(MachineInstr &BrI,
3115 const CellMap &Inputs) {
3116 MachineBasicBlock &B = *BrI.getParent();
3117 unsigned NumOp = BrI.getNumOperands();
3122 SetVector<const MachineBasicBlock*> Targets;
3123 bool Eval = evaluate(BrI, Inputs, Targets, FallsThru);
3124 unsigned NumTargets = Targets.size();
3125 if (!Eval || NumTargets > 1 || (NumTargets == 1 && FallsThru))
3127 if (BrI.getOpcode() == Hexagon::J2_jump)
3130 DEBUG(dbgs() << "Rewrite(" << printMBBReference(B) << "):" << BrI);
3131 bool Rewritten = false;
3132 if (NumTargets > 0) {
3133 assert(!FallsThru && "This should have been checked before");
3134 // MIB.addMBB needs non-const pointer.
3135 MachineBasicBlock *TargetB = const_cast<MachineBasicBlock*>(Targets[0]);
3136 bool Moot = B.isLayoutSuccessor(TargetB);
3138 // If we build a branch here, we must make sure that it won't be
3139 // erased as "non-executable". We can't mark any new instructions
3140 // as executable here, so we need to overwrite the BrI, which we
3141 // know is executable.
3142 const MCInstrDesc &JD = HII.get(Hexagon::J2_jump);
3143 auto NI = BuildMI(B, BrI.getIterator(), BrI.getDebugLoc(), JD)
3146 while (BrI.getNumOperands() > 0)
3147 BrI.RemoveOperand(0);
3148 // This ensures that all implicit operands (e.g. implicit-def %r31, etc)
3149 // are present in the rewritten branch.
3150 for (auto &Op : NI->operands())
3152 NI->eraseFromParent();
3157 // Do not erase instructions. A newly created instruction could get
3158 // the same address as an instruction marked as executable during the
3161 replaceWithNop(BrI);
3165 FunctionPass *llvm::createHexagonConstPropagationPass() {
3166 return new HexagonConstPropagation();