xref: /llvm-project/llvm/lib/Target/PowerPC/PPCMIPeephole.cpp (revision 6262fd4b0a310d3afdb1457109f9aaad9d1aa928)
1 //===-------------- PPCMIPeephole.cpp - MI Peephole Cleanups -------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===---------------------------------------------------------------------===//
9 //
10 // This pass performs peephole optimizations to clean up ugly code
11 // sequences at the MachineInstruction layer.  It runs at the end of
12 // the SSA phases, following VSX swap removal.  A pass of dead code
13 // elimination follows this one for quick clean-up of any dead
14 // instructions introduced here.  Although we could do this as callbacks
15 // from the generic peephole pass, this would have a couple of bad
16 // effects:  it might remove optimization opportunities for VSX swap
17 // removal, and it would miss cleanups made possible following VSX
18 // swap removal.
19 //
20 //===---------------------------------------------------------------------===//
21 
22 #include "PPC.h"
23 #include "PPCInstrBuilder.h"
24 #include "PPCInstrInfo.h"
25 #include "PPCTargetMachine.h"
26 #include "llvm/ADT/Statistic.h"
27 #include "llvm/CodeGen/MachineDominators.h"
28 #include "llvm/CodeGen/MachineFunctionPass.h"
29 #include "llvm/CodeGen/MachineInstrBuilder.h"
30 #include "llvm/CodeGen/MachineRegisterInfo.h"
31 #include "llvm/Support/Debug.h"
32 #include "llvm/ADT/Statistic.h"
33 #include "MCTargetDesc/PPCPredicates.h"
34 
35 using namespace llvm;
36 
37 #define DEBUG_TYPE "ppc-mi-peepholes"
38 
39 STATISTIC(NumEliminatedSExt, "Number of eliminated sign-extensions");
40 STATISTIC(NumEliminatedZExt, "Number of eliminated zero-extensions");
41 STATISTIC(NumOptADDLIs, "Number of optimized ADD instruction fed by LI");
42 
43 static cl::opt<bool>
44     EnableSExtElimination("ppc-eliminate-signext",
45                           cl::desc("enable elimination of sign-extensions"),
46                           cl::init(false), cl::Hidden);
47 
48 static cl::opt<bool>
49     EnableZExtElimination("ppc-eliminate-zeroext",
50                           cl::desc("enable elimination of zero-extensions"),
51                           cl::init(false), cl::Hidden);
52 
53 namespace llvm {
54   void initializePPCMIPeepholePass(PassRegistry&);
55 }
56 
57 namespace {
58 
59 struct PPCMIPeephole : public MachineFunctionPass {
60 
61   static char ID;
62   const PPCInstrInfo *TII;
63   MachineFunction *MF;
64   MachineRegisterInfo *MRI;
65 
66   PPCMIPeephole() : MachineFunctionPass(ID) {
67     initializePPCMIPeepholePass(*PassRegistry::getPassRegistry());
68   }
69 
70 private:
71   MachineDominatorTree *MDT;
72 
73   // Initialize class variables.
74   void initialize(MachineFunction &MFParm);
75 
76   // Perform peepholes.
77   bool simplifyCode(void);
78 
79   // Perform peepholes.
80   bool eliminateRedundantCompare(void);
81 
82   // Find the "true" register represented by SrcReg (following chains
83   // of copies and subreg_to_reg operations).
84   unsigned lookThruCopyLike(unsigned SrcReg);
85 
86 public:
87 
88   void getAnalysisUsage(AnalysisUsage &AU) const override {
89     AU.addRequired<MachineDominatorTree>();
90     AU.addPreserved<MachineDominatorTree>();
91     MachineFunctionPass::getAnalysisUsage(AU);
92   }
93 
94   // Main entry point for this pass.
95   bool runOnMachineFunction(MachineFunction &MF) override {
96     if (skipFunction(*MF.getFunction()))
97       return false;
98     initialize(MF);
99     return simplifyCode();
100   }
101 };
102 
103 // Initialize class variables.
104 void PPCMIPeephole::initialize(MachineFunction &MFParm) {
105   MF = &MFParm;
106   MRI = &MF->getRegInfo();
107   MDT = &getAnalysis<MachineDominatorTree>();
108   TII = MF->getSubtarget<PPCSubtarget>().getInstrInfo();
109   DEBUG(dbgs() << "*** PowerPC MI peephole pass ***\n\n");
110   DEBUG(MF->dump());
111 }
112 
113 static MachineInstr *getVRegDefOrNull(MachineOperand *Op,
114                                       MachineRegisterInfo *MRI) {
115   assert(Op && "Invalid Operand!");
116   if (!Op->isReg())
117     return nullptr;
118 
119   unsigned Reg = Op->getReg();
120   if (!TargetRegisterInfo::isVirtualRegister(Reg))
121     return nullptr;
122 
123   return MRI->getVRegDef(Reg);
124 }
125 
126 // This function returns number of known zero bits in output of MI
127 // starting from the most significant bit.
128 static unsigned
129 getKnownLeadingZeroCount(MachineInstr *MI, const PPCInstrInfo *TII) {
130   unsigned Opcode = MI->getOpcode();
131   if (Opcode == PPC::RLDICL || Opcode == PPC::RLDICLo ||
132       Opcode == PPC::RLDCL  || Opcode == PPC::RLDCLo)
133     return MI->getOperand(3).getImm();
134 
135   if ((Opcode == PPC::RLDIC || Opcode == PPC::RLDICo) &&
136        MI->getOperand(3).getImm() <= 63 - MI->getOperand(2).getImm())
137     return MI->getOperand(3).getImm();
138 
139   if ((Opcode == PPC::RLWINM  || Opcode == PPC::RLWINMo ||
140        Opcode == PPC::RLWNM   || Opcode == PPC::RLWNMo  ||
141        Opcode == PPC::RLWINM8 || Opcode == PPC::RLWNM8) &&
142        MI->getOperand(3).getImm() <= MI->getOperand(4).getImm())
143     return 32 + MI->getOperand(3).getImm();
144 
145   if (Opcode == PPC::ANDIo) {
146     uint16_t Imm = MI->getOperand(2).getImm();
147     return 48 + countLeadingZeros(Imm);
148   }
149 
150   if (Opcode == PPC::CNTLZW  || Opcode == PPC::CNTLZWo ||
151       Opcode == PPC::CNTTZW  || Opcode == PPC::CNTTZWo ||
152       Opcode == PPC::CNTLZW8 || Opcode == PPC::CNTTZW8)
153     // The result ranges from 0 to 32.
154     return 58;
155 
156   if (Opcode == PPC::CNTLZD  || Opcode == PPC::CNTLZDo ||
157       Opcode == PPC::CNTTZD  || Opcode == PPC::CNTTZDo)
158     // The result ranges from 0 to 64.
159     return 57;
160 
161   if (Opcode == PPC::LHZ   || Opcode == PPC::LHZX  ||
162       Opcode == PPC::LHZ8  || Opcode == PPC::LHZX8 ||
163       Opcode == PPC::LHZU  || Opcode == PPC::LHZUX ||
164       Opcode == PPC::LHZU8 || Opcode == PPC::LHZUX8)
165     return 48;
166 
167   if (Opcode == PPC::LBZ   || Opcode == PPC::LBZX  ||
168       Opcode == PPC::LBZ8  || Opcode == PPC::LBZX8 ||
169       Opcode == PPC::LBZU  || Opcode == PPC::LBZUX ||
170       Opcode == PPC::LBZU8 || Opcode == PPC::LBZUX8)
171     return 56;
172 
173   if (TII->isZeroExtended(*MI))
174     return 32;
175 
176   return 0;
177 }
178 
179 // Perform peephole optimizations.
180 bool PPCMIPeephole::simplifyCode(void) {
181   bool Simplified = false;
182   MachineInstr* ToErase = nullptr;
183 
184   for (MachineBasicBlock &MBB : *MF) {
185     for (MachineInstr &MI : MBB) {
186 
187       // If the previous instruction was marked for elimination,
188       // remove it now.
189       if (ToErase) {
190         ToErase->eraseFromParent();
191         ToErase = nullptr;
192       }
193 
194       // Ignore debug instructions.
195       if (MI.isDebugValue())
196         continue;
197 
198       // Per-opcode peepholes.
199       switch (MI.getOpcode()) {
200 
201       default:
202         break;
203 
204       case PPC::XXPERMDI: {
205         // Perform simplifications of 2x64 vector swaps and splats.
206         // A swap is identified by an immediate value of 2, and a splat
207         // is identified by an immediate value of 0 or 3.
208         int Immed = MI.getOperand(3).getImm();
209 
210         if (Immed != 1) {
211 
212           // For each of these simplifications, we need the two source
213           // regs to match.  Unfortunately, MachineCSE ignores COPY and
214           // SUBREG_TO_REG, so for example we can see
215           //   XXPERMDI t, SUBREG_TO_REG(s), SUBREG_TO_REG(s), immed.
216           // We have to look through chains of COPY and SUBREG_TO_REG
217           // to find the real source values for comparison.
218           unsigned TrueReg1 = lookThruCopyLike(MI.getOperand(1).getReg());
219           unsigned TrueReg2 = lookThruCopyLike(MI.getOperand(2).getReg());
220 
221           if (TrueReg1 == TrueReg2
222               && TargetRegisterInfo::isVirtualRegister(TrueReg1)) {
223             MachineInstr *DefMI = MRI->getVRegDef(TrueReg1);
224             unsigned DefOpc = DefMI ? DefMI->getOpcode() : 0;
225 
226             // If this is a splat fed by a splatting load, the splat is
227             // redundant. Replace with a copy. This doesn't happen directly due
228             // to code in PPCDAGToDAGISel.cpp, but it can happen when converting
229             // a load of a double to a vector of 64-bit integers.
230             auto isConversionOfLoadAndSplat = [=]() -> bool {
231               if (DefOpc != PPC::XVCVDPSXDS && DefOpc != PPC::XVCVDPUXDS)
232                 return false;
233               unsigned DefReg = lookThruCopyLike(DefMI->getOperand(1).getReg());
234               if (TargetRegisterInfo::isVirtualRegister(DefReg)) {
235                 MachineInstr *LoadMI = MRI->getVRegDef(DefReg);
236                 if (LoadMI && LoadMI->getOpcode() == PPC::LXVDSX)
237                   return true;
238               }
239               return false;
240             };
241             if (DefMI && (Immed == 0 || Immed == 3)) {
242               if (DefOpc == PPC::LXVDSX || isConversionOfLoadAndSplat()) {
243                 DEBUG(dbgs()
244                       << "Optimizing load-and-splat/splat "
245                       "to load-and-splat/copy: ");
246                 DEBUG(MI.dump());
247                 BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::COPY),
248                         MI.getOperand(0).getReg())
249                     .add(MI.getOperand(1));
250                 ToErase = &MI;
251                 Simplified = true;
252               }
253             }
254 
255             // If this is a splat or a swap fed by another splat, we
256             // can replace it with a copy.
257             if (DefOpc == PPC::XXPERMDI) {
258               unsigned FeedImmed = DefMI->getOperand(3).getImm();
259               unsigned FeedReg1
260                 = lookThruCopyLike(DefMI->getOperand(1).getReg());
261               unsigned FeedReg2
262                 = lookThruCopyLike(DefMI->getOperand(2).getReg());
263 
264               if ((FeedImmed == 0 || FeedImmed == 3) && FeedReg1 == FeedReg2) {
265                 DEBUG(dbgs()
266                       << "Optimizing splat/swap or splat/splat "
267                       "to splat/copy: ");
268                 DEBUG(MI.dump());
269                 BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::COPY),
270                         MI.getOperand(0).getReg())
271                     .add(MI.getOperand(1));
272                 ToErase = &MI;
273                 Simplified = true;
274               }
275 
276               // If this is a splat fed by a swap, we can simplify modify
277               // the splat to splat the other value from the swap's input
278               // parameter.
279               else if ((Immed == 0 || Immed == 3)
280                        && FeedImmed == 2 && FeedReg1 == FeedReg2) {
281                 DEBUG(dbgs() << "Optimizing swap/splat => splat: ");
282                 DEBUG(MI.dump());
283                 MI.getOperand(1).setReg(DefMI->getOperand(1).getReg());
284                 MI.getOperand(2).setReg(DefMI->getOperand(2).getReg());
285                 MI.getOperand(3).setImm(3 - Immed);
286                 Simplified = true;
287               }
288 
289               // If this is a swap fed by a swap, we can replace it
290               // with a copy from the first swap's input.
291               else if (Immed == 2 && FeedImmed == 2 && FeedReg1 == FeedReg2) {
292                 DEBUG(dbgs() << "Optimizing swap/swap => copy: ");
293                 DEBUG(MI.dump());
294                 BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::COPY),
295                         MI.getOperand(0).getReg())
296                     .add(DefMI->getOperand(1));
297                 ToErase = &MI;
298                 Simplified = true;
299               }
300             } else if ((Immed == 0 || Immed == 3) && DefOpc == PPC::XXPERMDIs &&
301                        (DefMI->getOperand(2).getImm() == 0 ||
302                         DefMI->getOperand(2).getImm() == 3)) {
303               // Splat fed by another splat - switch the output of the first
304               // and remove the second.
305               DefMI->getOperand(0).setReg(MI.getOperand(0).getReg());
306               ToErase = &MI;
307               Simplified = true;
308               DEBUG(dbgs() << "Removing redundant splat: ");
309               DEBUG(MI.dump());
310             }
311           }
312         }
313         break;
314       }
315       case PPC::VSPLTB:
316       case PPC::VSPLTH:
317       case PPC::XXSPLTW: {
318         unsigned MyOpcode = MI.getOpcode();
319         unsigned OpNo = MyOpcode == PPC::XXSPLTW ? 1 : 2;
320         unsigned TrueReg = lookThruCopyLike(MI.getOperand(OpNo).getReg());
321         if (!TargetRegisterInfo::isVirtualRegister(TrueReg))
322           break;
323         MachineInstr *DefMI = MRI->getVRegDef(TrueReg);
324         if (!DefMI)
325           break;
326         unsigned DefOpcode = DefMI->getOpcode();
327         auto isConvertOfSplat = [=]() -> bool {
328           if (DefOpcode != PPC::XVCVSPSXWS && DefOpcode != PPC::XVCVSPUXWS)
329             return false;
330           unsigned ConvReg = DefMI->getOperand(1).getReg();
331           if (!TargetRegisterInfo::isVirtualRegister(ConvReg))
332             return false;
333           MachineInstr *Splt = MRI->getVRegDef(ConvReg);
334           return Splt && (Splt->getOpcode() == PPC::LXVWSX ||
335             Splt->getOpcode() == PPC::XXSPLTW);
336         };
337         bool AlreadySplat = (MyOpcode == DefOpcode) ||
338           (MyOpcode == PPC::VSPLTB && DefOpcode == PPC::VSPLTBs) ||
339           (MyOpcode == PPC::VSPLTH && DefOpcode == PPC::VSPLTHs) ||
340           (MyOpcode == PPC::XXSPLTW && DefOpcode == PPC::XXSPLTWs) ||
341           (MyOpcode == PPC::XXSPLTW && DefOpcode == PPC::LXVWSX) ||
342           (MyOpcode == PPC::XXSPLTW && DefOpcode == PPC::MTVSRWS)||
343           (MyOpcode == PPC::XXSPLTW && isConvertOfSplat());
344         // If the instruction[s] that feed this splat have already splat
345         // the value, this splat is redundant.
346         if (AlreadySplat) {
347           DEBUG(dbgs() << "Changing redundant splat to a copy: ");
348           DEBUG(MI.dump());
349           BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::COPY),
350                   MI.getOperand(0).getReg())
351               .add(MI.getOperand(OpNo));
352           ToErase = &MI;
353           Simplified = true;
354         }
355         // Splat fed by a shift. Usually when we align value to splat into
356         // vector element zero.
357         if (DefOpcode == PPC::XXSLDWI) {
358           unsigned ShiftRes = DefMI->getOperand(0).getReg();
359           unsigned ShiftOp1 = DefMI->getOperand(1).getReg();
360           unsigned ShiftOp2 = DefMI->getOperand(2).getReg();
361           unsigned ShiftImm = DefMI->getOperand(3).getImm();
362           unsigned SplatImm = MI.getOperand(2).getImm();
363           if (ShiftOp1 == ShiftOp2) {
364             unsigned NewElem = (SplatImm + ShiftImm) & 0x3;
365             if (MRI->hasOneNonDBGUse(ShiftRes)) {
366               DEBUG(dbgs() << "Removing redundant shift: ");
367               DEBUG(DefMI->dump());
368               ToErase = DefMI;
369             }
370             Simplified = true;
371             DEBUG(dbgs() << "Changing splat immediate from " << SplatImm <<
372                   " to " << NewElem << " in instruction: ");
373             DEBUG(MI.dump());
374             MI.getOperand(1).setReg(ShiftOp1);
375             MI.getOperand(2).setImm(NewElem);
376           }
377         }
378         break;
379       }
380       case PPC::XVCVDPSP: {
381         // If this is a DP->SP conversion fed by an FRSP, the FRSP is redundant.
382         unsigned TrueReg = lookThruCopyLike(MI.getOperand(1).getReg());
383         if (!TargetRegisterInfo::isVirtualRegister(TrueReg))
384           break;
385         MachineInstr *DefMI = MRI->getVRegDef(TrueReg);
386 
387         // This can occur when building a vector of single precision or integer
388         // values.
389         if (DefMI && DefMI->getOpcode() == PPC::XXPERMDI) {
390           unsigned DefsReg1 = lookThruCopyLike(DefMI->getOperand(1).getReg());
391           unsigned DefsReg2 = lookThruCopyLike(DefMI->getOperand(2).getReg());
392           if (!TargetRegisterInfo::isVirtualRegister(DefsReg1) ||
393               !TargetRegisterInfo::isVirtualRegister(DefsReg2))
394             break;
395           MachineInstr *P1 = MRI->getVRegDef(DefsReg1);
396           MachineInstr *P2 = MRI->getVRegDef(DefsReg2);
397 
398           if (!P1 || !P2)
399             break;
400 
401           // Remove the passed FRSP instruction if it only feeds this MI and
402           // set any uses of that FRSP (in this MI) to the source of the FRSP.
403           auto removeFRSPIfPossible = [&](MachineInstr *RoundInstr) {
404             if (RoundInstr->getOpcode() == PPC::FRSP &&
405                 MRI->hasOneNonDBGUse(RoundInstr->getOperand(0).getReg())) {
406               Simplified = true;
407               unsigned ConvReg1 = RoundInstr->getOperand(1).getReg();
408               unsigned FRSPDefines = RoundInstr->getOperand(0).getReg();
409               MachineInstr &Use = *(MRI->use_instr_begin(FRSPDefines));
410               for (int i = 0, e = Use.getNumOperands(); i < e; ++i)
411                 if (Use.getOperand(i).isReg() &&
412                     Use.getOperand(i).getReg() == FRSPDefines)
413                   Use.getOperand(i).setReg(ConvReg1);
414               DEBUG(dbgs() << "Removing redundant FRSP:\n");
415               DEBUG(RoundInstr->dump());
416               DEBUG(dbgs() << "As it feeds instruction:\n");
417               DEBUG(MI.dump());
418               DEBUG(dbgs() << "Through instruction:\n");
419               DEBUG(DefMI->dump());
420               RoundInstr->eraseFromParent();
421             }
422           };
423 
424           // If the input to XVCVDPSP is a vector that was built (even
425           // partially) out of FRSP's, the FRSP(s) can safely be removed
426           // since this instruction performs the same operation.
427           if (P1 != P2) {
428             removeFRSPIfPossible(P1);
429             removeFRSPIfPossible(P2);
430             break;
431           }
432           removeFRSPIfPossible(P1);
433         }
434         break;
435       }
436       case PPC::EXTSH:
437       case PPC::EXTSH8:
438       case PPC::EXTSH8_32_64: {
439         if (!EnableSExtElimination) break;
440         unsigned NarrowReg = MI.getOperand(1).getReg();
441         if (!TargetRegisterInfo::isVirtualRegister(NarrowReg))
442           break;
443 
444         MachineInstr *SrcMI = MRI->getVRegDef(NarrowReg);
445         // If we've used a zero-extending load that we will sign-extend,
446         // just do a sign-extending load.
447         if (SrcMI->getOpcode() == PPC::LHZ ||
448             SrcMI->getOpcode() == PPC::LHZX) {
449           if (!MRI->hasOneNonDBGUse(SrcMI->getOperand(0).getReg()))
450             break;
451           auto is64Bit = [] (unsigned Opcode) {
452             return Opcode == PPC::EXTSH8;
453           };
454           auto isXForm = [] (unsigned Opcode) {
455             return Opcode == PPC::LHZX;
456           };
457           auto getSextLoadOp = [] (bool is64Bit, bool isXForm) {
458             if (is64Bit)
459               if (isXForm) return PPC::LHAX8;
460               else         return PPC::LHA8;
461             else
462               if (isXForm) return PPC::LHAX;
463               else         return PPC::LHA;
464           };
465           unsigned Opc = getSextLoadOp(is64Bit(MI.getOpcode()),
466                                        isXForm(SrcMI->getOpcode()));
467           DEBUG(dbgs() << "Zero-extending load\n");
468           DEBUG(SrcMI->dump());
469           DEBUG(dbgs() << "and sign-extension\n");
470           DEBUG(MI.dump());
471           DEBUG(dbgs() << "are merged into sign-extending load\n");
472           SrcMI->setDesc(TII->get(Opc));
473           SrcMI->getOperand(0).setReg(MI.getOperand(0).getReg());
474           ToErase = &MI;
475           Simplified = true;
476           NumEliminatedSExt++;
477         }
478         break;
479       }
480       case PPC::EXTSW:
481       case PPC::EXTSW_32:
482       case PPC::EXTSW_32_64: {
483         if (!EnableSExtElimination) break;
484         unsigned NarrowReg = MI.getOperand(1).getReg();
485         if (!TargetRegisterInfo::isVirtualRegister(NarrowReg))
486           break;
487 
488         MachineInstr *SrcMI = MRI->getVRegDef(NarrowReg);
489         // If we've used a zero-extending load that we will sign-extend,
490         // just do a sign-extending load.
491         if (SrcMI->getOpcode() == PPC::LWZ ||
492             SrcMI->getOpcode() == PPC::LWZX) {
493           if (!MRI->hasOneNonDBGUse(SrcMI->getOperand(0).getReg()))
494             break;
495           auto is64Bit = [] (unsigned Opcode) {
496             return Opcode == PPC::EXTSW || Opcode == PPC::EXTSW_32_64;
497           };
498           auto isXForm = [] (unsigned Opcode) {
499             return Opcode == PPC::LWZX;
500           };
501           auto getSextLoadOp = [] (bool is64Bit, bool isXForm) {
502             if (is64Bit)
503               if (isXForm) return PPC::LWAX;
504               else         return PPC::LWA;
505             else
506               if (isXForm) return PPC::LWAX_32;
507               else         return PPC::LWA_32;
508           };
509           unsigned Opc = getSextLoadOp(is64Bit(MI.getOpcode()),
510                                        isXForm(SrcMI->getOpcode()));
511           DEBUG(dbgs() << "Zero-extending load\n");
512           DEBUG(SrcMI->dump());
513           DEBUG(dbgs() << "and sign-extension\n");
514           DEBUG(MI.dump());
515           DEBUG(dbgs() << "are merged into sign-extending load\n");
516           SrcMI->setDesc(TII->get(Opc));
517           SrcMI->getOperand(0).setReg(MI.getOperand(0).getReg());
518           ToErase = &MI;
519           Simplified = true;
520           NumEliminatedSExt++;
521         } else if (MI.getOpcode() == PPC::EXTSW_32_64 &&
522                    TII->isSignExtended(*SrcMI)) {
523           // We can eliminate EXTSW if the input is known to be already
524           // sign-extended.
525           DEBUG(dbgs() << "Removing redundant sign-extension\n");
526           unsigned TmpReg =
527             MF->getRegInfo().createVirtualRegister(&PPC::G8RCRegClass);
528           BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::IMPLICIT_DEF),
529                   TmpReg);
530           BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::INSERT_SUBREG),
531                   MI.getOperand(0).getReg())
532               .addReg(TmpReg)
533               .addReg(NarrowReg)
534               .addImm(PPC::sub_32);
535           ToErase = &MI;
536           Simplified = true;
537           NumEliminatedSExt++;
538         }
539         break;
540       }
541       case PPC::RLDICL: {
542         // We can eliminate RLDICL (e.g. for zero-extension)
543         // if all bits to clear are already zero in the input.
544         // This code assume following code sequence for zero-extension.
545         //   %vreg6<def> = COPY %vreg5:sub_32; (optional)
546         //   %vreg8<def> = IMPLICIT_DEF;
547         //   %vreg7<def,tied1> = INSERT_SUBREG %vreg8<tied0>, %vreg6, sub_32;
548         if (!EnableZExtElimination) break;
549 
550         if (MI.getOperand(2).getImm() != 0)
551           break;
552 
553         unsigned SrcReg = MI.getOperand(1).getReg();
554         if (!TargetRegisterInfo::isVirtualRegister(SrcReg))
555           break;
556 
557         MachineInstr *SrcMI = MRI->getVRegDef(SrcReg);
558         if (!(SrcMI && SrcMI->getOpcode() == PPC::INSERT_SUBREG &&
559               SrcMI->getOperand(0).isReg() && SrcMI->getOperand(1).isReg()))
560           break;
561 
562         MachineInstr *ImpDefMI, *SubRegMI;
563         ImpDefMI = MRI->getVRegDef(SrcMI->getOperand(1).getReg());
564         SubRegMI = MRI->getVRegDef(SrcMI->getOperand(2).getReg());
565         if (ImpDefMI->getOpcode() != PPC::IMPLICIT_DEF) break;
566 
567         SrcMI = SubRegMI;
568         if (SubRegMI->getOpcode() == PPC::COPY) {
569           unsigned CopyReg = SubRegMI->getOperand(1).getReg();
570           if (TargetRegisterInfo::isVirtualRegister(CopyReg))
571             SrcMI = MRI->getVRegDef(CopyReg);
572         }
573 
574         unsigned KnownZeroCount = getKnownLeadingZeroCount(SrcMI, TII);
575         if (MI.getOperand(3).getImm() <= KnownZeroCount) {
576           DEBUG(dbgs() << "Removing redundant zero-extension\n");
577           BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::COPY),
578                   MI.getOperand(0).getReg())
579               .addReg(SrcReg);
580           ToErase = &MI;
581           Simplified = true;
582           NumEliminatedZExt++;
583         }
584         break;
585       }
586 
587       // TODO: Any instruction that has an immediate form fed only by a PHI
588       // whose operands are all load immediate can be folded away. We currently
589       // do this for ADD instructions, but should expand it to arithmetic and
590       // binary instructions with immediate forms in the future.
591       case PPC::ADD4:
592       case PPC::ADD8: {
593         auto isSingleUsePHI = [&](MachineOperand *PhiOp) {
594           assert(PhiOp && "Invalid Operand!");
595           MachineInstr *DefPhiMI = getVRegDefOrNull(PhiOp, MRI);
596 
597           return DefPhiMI && (DefPhiMI->getOpcode() == PPC::PHI) &&
598                  MRI->hasOneNonDBGUse(DefPhiMI->getOperand(0).getReg());
599         };
600 
601         auto dominatesAllSingleUseLIs = [&](MachineOperand *DominatorOp,
602                                             MachineOperand *PhiOp) {
603           assert(PhiOp && "Invalid Operand!");
604           assert(DominatorOp && "Invalid Operand!");
605           MachineInstr *DefPhiMI = getVRegDefOrNull(PhiOp, MRI);
606           MachineInstr *DefDomMI = getVRegDefOrNull(DominatorOp, MRI);
607 
608           // Note: the vregs only show up at odd indices position of PHI Node,
609           // the even indices position save the BB info.
610           for (unsigned i = 1; i < DefPhiMI->getNumOperands(); i += 2) {
611             MachineInstr *LiMI =
612                 getVRegDefOrNull(&DefPhiMI->getOperand(i), MRI);
613             if (!LiMI ||
614                 (LiMI->getOpcode() != PPC::LI && LiMI->getOpcode() != PPC::LI8)
615                 || !MRI->hasOneNonDBGUse(LiMI->getOperand(0).getReg()) ||
616                 !MDT->dominates(DefDomMI, LiMI))
617               return false;
618           }
619 
620           return true;
621         };
622 
623         MachineOperand Op1 = MI.getOperand(1);
624         MachineOperand Op2 = MI.getOperand(2);
625         if (isSingleUsePHI(&Op2) && dominatesAllSingleUseLIs(&Op1, &Op2))
626           std::swap(Op1, Op2);
627         else if (!isSingleUsePHI(&Op1) || !dominatesAllSingleUseLIs(&Op2, &Op1))
628           break; // We don't have an ADD fed by LI's that can be transformed
629 
630         // Now we know that Op1 is the PHI node and Op2 is the dominator
631         unsigned DominatorReg = Op2.getReg();
632 
633         const TargetRegisterClass *TRC = MI.getOpcode() == PPC::ADD8
634                                              ? &PPC::G8RC_and_G8RC_NOX0RegClass
635                                              : &PPC::GPRC_and_GPRC_NOR0RegClass;
636         MRI->setRegClass(DominatorReg, TRC);
637 
638         // replace LIs with ADDIs
639         MachineInstr *DefPhiMI = getVRegDefOrNull(&Op1, MRI);
640         for (unsigned i = 1; i < DefPhiMI->getNumOperands(); i += 2) {
641           MachineInstr *LiMI = getVRegDefOrNull(&DefPhiMI->getOperand(i), MRI);
642           DEBUG(dbgs() << "Optimizing LI to ADDI: ");
643           DEBUG(LiMI->dump());
644 
645           // There could be repeated registers in the PHI, e.g: %vreg1<def> =
646           // PHI %vreg6, <BB#2>, %vreg8, <BB#3>, %vreg8, <BB#6>; So if we've
647           // already replaced the def instruction, skip.
648           if (LiMI->getOpcode() == PPC::ADDI || LiMI->getOpcode() == PPC::ADDI8)
649             continue;
650 
651           assert((LiMI->getOpcode() == PPC::LI ||
652                   LiMI->getOpcode() == PPC::LI8) &&
653                  "Invalid Opcode!");
654           auto LiImm = LiMI->getOperand(1).getImm(); // save the imm of LI
655           LiMI->RemoveOperand(1);                    // remove the imm of LI
656           LiMI->setDesc(TII->get(LiMI->getOpcode() == PPC::LI ? PPC::ADDI
657                                                               : PPC::ADDI8));
658           MachineInstrBuilder(*LiMI->getParent()->getParent(), *LiMI)
659               .addReg(DominatorReg)
660               .addImm(LiImm); // restore the imm of LI
661           DEBUG(LiMI->dump());
662         }
663 
664         // Replace ADD with COPY
665         DEBUG(dbgs() << "Optimizing ADD to COPY: ");
666         DEBUG(MI.dump());
667         BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(PPC::COPY),
668                 MI.getOperand(0).getReg())
669             .add(Op1);
670         ToErase = &MI;
671         Simplified = true;
672         NumOptADDLIs++;
673         break;
674       }
675       }
676     }
677 
678     // If the last instruction was marked for elimination,
679     // remove it now.
680     if (ToErase) {
681       ToErase->eraseFromParent();
682       ToErase = nullptr;
683     }
684   }
685 
686   // We try to eliminate redundant compare instruction.
687   Simplified |= eliminateRedundantCompare();
688 
689   return Simplified;
690 }
691 
692 // helper functions for eliminateRedundantCompare
693 static bool isEqOrNe(MachineInstr *BI) {
694   PPC::Predicate Pred = (PPC::Predicate)BI->getOperand(0).getImm();
695   unsigned PredCond = PPC::getPredicateCondition(Pred);
696   return (PredCond == PPC::PRED_EQ || PredCond == PPC::PRED_NE);
697 }
698 
699 static bool isSupportedCmpOp(unsigned opCode) {
700   return (opCode == PPC::CMPLD  || opCode == PPC::CMPD  ||
701           opCode == PPC::CMPLW  || opCode == PPC::CMPW  ||
702           opCode == PPC::CMPLDI || opCode == PPC::CMPDI ||
703           opCode == PPC::CMPLWI || opCode == PPC::CMPWI);
704 }
705 
706 static bool is64bitCmpOp(unsigned opCode) {
707   return (opCode == PPC::CMPLD  || opCode == PPC::CMPD ||
708           opCode == PPC::CMPLDI || opCode == PPC::CMPDI);
709 }
710 
711 static bool isSignedCmpOp(unsigned opCode) {
712   return (opCode == PPC::CMPD  || opCode == PPC::CMPW ||
713           opCode == PPC::CMPDI || opCode == PPC::CMPWI);
714 }
715 
716 static unsigned getSignedCmpOpCode(unsigned opCode) {
717   if (opCode == PPC::CMPLD)  return PPC::CMPD;
718   if (opCode == PPC::CMPLW)  return PPC::CMPW;
719   if (opCode == PPC::CMPLDI) return PPC::CMPDI;
720   if (opCode == PPC::CMPLWI) return PPC::CMPWI;
721   return opCode;
722 }
723 
724 // We can decrement immediate x in (GE x) by changing it to (GT x-1) or
725 // (LT x) to (LE x-1)
726 static unsigned getPredicateToDecImm(MachineInstr *BI, MachineInstr *CMPI) {
727   uint64_t Imm = CMPI->getOperand(2).getImm();
728   bool SignedCmp = isSignedCmpOp(CMPI->getOpcode());
729   if ((!SignedCmp && Imm == 0) || (SignedCmp && Imm == 0x8000))
730     return 0;
731 
732   PPC::Predicate Pred = (PPC::Predicate)BI->getOperand(0).getImm();
733   unsigned PredCond = PPC::getPredicateCondition(Pred);
734   unsigned PredHint = PPC::getPredicateHint(Pred);
735   if (PredCond == PPC::PRED_GE)
736     return PPC::getPredicate(PPC::PRED_GT, PredHint);
737   if (PredCond == PPC::PRED_LT)
738     return PPC::getPredicate(PPC::PRED_LE, PredHint);
739 
740   return 0;
741 }
742 
743 // We can increment immediate x in (GT x) by changing it to (GE x+1) or
744 // (LE x) to (LT x+1)
745 static unsigned getPredicateToIncImm(MachineInstr *BI, MachineInstr *CMPI) {
746   uint64_t Imm = CMPI->getOperand(2).getImm();
747   bool SignedCmp = isSignedCmpOp(CMPI->getOpcode());
748   if ((!SignedCmp && Imm == 0xFFFF) || (SignedCmp && Imm == 0x7FFF))
749     return 0;
750 
751   PPC::Predicate Pred = (PPC::Predicate)BI->getOperand(0).getImm();
752   unsigned PredCond = PPC::getPredicateCondition(Pred);
753   unsigned PredHint = PPC::getPredicateHint(Pred);
754   if (PredCond == PPC::PRED_GT)
755     return PPC::getPredicate(PPC::PRED_GE, PredHint);
756   if (PredCond == PPC::PRED_LE)
757     return PPC::getPredicate(PPC::PRED_LT, PredHint);
758 
759   return 0;
760 }
761 
762 // This takes a Phi node and returns a register value for the spefied BB.
763 static unsigned getIncomingRegForBlock(MachineInstr *Phi,
764                                        MachineBasicBlock *MBB) {
765   for (unsigned I = 2, E = Phi->getNumOperands() + 1; I != E; I += 2) {
766     MachineOperand &MO = Phi->getOperand(I);
767     if (MO.getMBB() == MBB)
768       return Phi->getOperand(I-1).getReg();
769   }
770   llvm_unreachable("invalid src basic block for this Phi node\n");
771   return 0;
772 }
773 
774 // This function tracks the source of the register through register copy.
775 // If BB1 and BB2 are non-NULL, we also track PHI instruction in BB2
776 // assuming that the control comes from BB1 into BB2.
777 static unsigned getSrcVReg(unsigned Reg, MachineBasicBlock *BB1,
778                            MachineBasicBlock *BB2, MachineRegisterInfo *MRI) {
779   unsigned SrcReg = Reg;
780   while (1) {
781     unsigned NextReg = SrcReg;
782     MachineInstr *Inst = MRI->getVRegDef(SrcReg);
783     if (BB1 && Inst->getOpcode() == PPC::PHI && Inst->getParent() == BB2) {
784       NextReg = getIncomingRegForBlock(Inst, BB1);
785       // We track through PHI only once to avoid infinite loop.
786       BB1 = nullptr;
787     }
788     else if (Inst->isFullCopy())
789       NextReg = Inst->getOperand(1).getReg();
790     if (NextReg == SrcReg || !TargetRegisterInfo::isVirtualRegister(NextReg))
791       break;
792     SrcReg = NextReg;
793   }
794   return SrcReg;
795 }
796 
797 static bool eligibleForCompareElimination(MachineBasicBlock &MBB,
798                                           MachineBasicBlock *&PredMBB,
799                                           MachineBasicBlock *&MBBtoMoveCmp,
800                                           MachineRegisterInfo *MRI) {
801 
802   auto isEligibleBB = [&](MachineBasicBlock &BB) {
803     auto BII = BB.getFirstInstrTerminator();
804     // We optimize BBs ending with a conditional branch.
805     // We check only for BCC here, not BCCLR, because BCCLR
806     // will be formed only later in the pipeline.
807     if (BB.succ_size() == 2 &&
808         BII != BB.instr_end() &&
809         (*BII).getOpcode() == PPC::BCC &&
810         (*BII).getOperand(1).isReg()) {
811       // We optimize only if the condition code is used only by one BCC.
812       unsigned CndReg = (*BII).getOperand(1).getReg();
813       if (!TargetRegisterInfo::isVirtualRegister(CndReg) ||
814           !MRI->hasOneNonDBGUse(CndReg))
815         return false;
816 
817       // We skip this BB if a physical register is used in comparison.
818       MachineInstr *CMPI = MRI->getVRegDef(CndReg);
819       for (MachineOperand &MO : CMPI->operands())
820         if (MO.isReg() && !TargetRegisterInfo::isVirtualRegister(MO.getReg()))
821           return false;
822 
823       return true;
824     }
825     return false;
826   };
827 
828   // If this BB has more than one successor, we can create a new BB and
829   // move the compare instruction in the new BB.
830   // So far, we do not move compare instruction to a BB having multiple
831   // successors to avoid potentially increasing code size.
832   auto isEligibleForMoveCmp = [](MachineBasicBlock &BB) {
833     return BB.succ_size() == 1;
834   };
835 
836   if (!isEligibleBB(MBB))
837     return false;
838 
839   unsigned NumPredBBs = MBB.pred_size();
840   if (NumPredBBs == 1) {
841     MachineBasicBlock *TmpMBB = *MBB.pred_begin();
842     if (isEligibleBB(*TmpMBB)) {
843       PredMBB = TmpMBB;
844       MBBtoMoveCmp = nullptr;
845       return true;
846     }
847   }
848   else if (NumPredBBs == 2) {
849     // We check for partially redundant case.
850     // So far, we support cases with only two predecessors
851     // to avoid increasing the number of instructions.
852     MachineBasicBlock::pred_iterator PI = MBB.pred_begin();
853     MachineBasicBlock *Pred1MBB = *PI;
854     MachineBasicBlock *Pred2MBB = *(PI+1);
855 
856     if (isEligibleBB(*Pred1MBB) && isEligibleForMoveCmp(*Pred2MBB)) {
857       // We assume Pred1MBB is the BB containing the compare to be merged and
858       // Pred2MBB is the BB to which we will append a compare instruction.
859       // Hence we can proceed as is.
860     }
861     else if (isEligibleBB(*Pred2MBB) && isEligibleForMoveCmp(*Pred1MBB)) {
862       // We need to swap Pred1MBB and Pred2MBB to canonicalize.
863       std::swap(Pred1MBB, Pred2MBB);
864     }
865     else return false;
866 
867     // Here, Pred2MBB is the BB to which we need to append a compare inst.
868     // We cannot move the compare instruction if operands are not available
869     // in Pred2MBB (i.e. defined in MBB by an instruction other than PHI).
870     MachineInstr *BI = &*MBB.getFirstInstrTerminator();
871     MachineInstr *CMPI = MRI->getVRegDef(BI->getOperand(1).getReg());
872     for (int I = 1; I <= 2; I++)
873       if (CMPI->getOperand(I).isReg()) {
874         MachineInstr *Inst = MRI->getVRegDef(CMPI->getOperand(I).getReg());
875         if (Inst->getParent() == &MBB && Inst->getOpcode() != PPC::PHI)
876           return false;
877       }
878 
879     PredMBB = Pred1MBB;
880     MBBtoMoveCmp = Pred2MBB;
881     return true;
882   }
883 
884   return false;
885 }
886 
887 // If multiple conditional branches are executed based on the (essentially)
888 // same comparison, we merge compare instructions into one and make multiple
889 // conditional branches on this comparison.
890 // For example,
891 //   if (a == 0) { ... }
892 //   else if (a < 0) { ... }
893 // can be executed by one compare and two conditional branches instead of
894 // two pairs of a compare and a conditional branch.
895 //
896 // This method merges two compare instructions in two MBBs and modifies the
897 // compare and conditional branch instructions if needed.
898 // For the above example, the input for this pass looks like:
899 //   cmplwi r3, 0
900 //   beq    0, .LBB0_3
901 //   cmpwi  r3, -1
902 //   bgt    0, .LBB0_4
903 // So, before merging two compares, we need to modify these instructions as
904 //   cmpwi  r3, 0       ; cmplwi and cmpwi yield same result for beq
905 //   beq    0, .LBB0_3
906 //   cmpwi  r3, 0       ; greather than -1 means greater or equal to 0
907 //   bge    0, .LBB0_4
908 
909 bool PPCMIPeephole::eliminateRedundantCompare(void) {
910   bool Simplified = false;
911 
912   for (MachineBasicBlock &MBB2 : *MF) {
913     MachineBasicBlock *MBB1 = nullptr, *MBBtoMoveCmp = nullptr;
914 
915     // For fully redundant case, we select two basic blocks MBB1 and MBB2
916     // as an optimization target if
917     // - both MBBs end with a conditional branch,
918     // - MBB1 is the only predecessor of MBB2, and
919     // - compare does not take a physical register as a operand in both MBBs.
920     // In this case, eligibleForCompareElimination sets MBBtoMoveCmp nullptr.
921     //
922     // As partially redundant case, we additionally handle if MBB2 has one
923     // additional predecessor, which has only one successor (MBB2).
924     // In this case, we move the compre instruction originally in MBB2 into
925     // MBBtoMoveCmp. This partially redundant case is typically appear by
926     // compiling a while loop; here, MBBtoMoveCmp is the loop preheader.
927     //
928     // Overview of CFG of related basic blocks
929     // Fully redundant case        Partially redundant case
930     //   --------                   ----------------  --------
931     //   | MBB1 | (w/ 2 succ)       | MBBtoMoveCmp |  | MBB1 | (w/ 2 succ)
932     //   --------                   ----------------  --------
933     //      |    \                     (w/ 1 succ) \     |    \
934     //      |     \                                 \    |     \
935     //      |                                        \   |
936     //   --------                                     --------
937     //   | MBB2 | (w/ 1 pred                          | MBB2 | (w/ 2 pred
938     //   -------- and 2 succ)                         -------- and 2 succ)
939     //      |    \                                       |    \
940     //      |     \                                      |     \
941     //
942     if (!eligibleForCompareElimination(MBB2, MBB1, MBBtoMoveCmp, MRI))
943       continue;
944 
945     MachineInstr *BI1   = &*MBB1->getFirstInstrTerminator();
946     MachineInstr *CMPI1 = MRI->getVRegDef(BI1->getOperand(1).getReg());
947 
948     MachineInstr *BI2   = &*MBB2.getFirstInstrTerminator();
949     MachineInstr *CMPI2 = MRI->getVRegDef(BI2->getOperand(1).getReg());
950     bool IsPartiallyRedundant = (MBBtoMoveCmp != nullptr);
951 
952     // We cannot optimize an unsupported compare opcode or
953     // a mix of 32-bit and 64-bit comaprisons
954     if (!isSupportedCmpOp(CMPI1->getOpcode()) ||
955         !isSupportedCmpOp(CMPI2->getOpcode()) ||
956         is64bitCmpOp(CMPI1->getOpcode()) != is64bitCmpOp(CMPI2->getOpcode()))
957       continue;
958 
959     unsigned NewOpCode = 0;
960     unsigned NewPredicate1 = 0, NewPredicate2 = 0;
961     int16_t Imm1 = 0, NewImm1 = 0, Imm2 = 0, NewImm2 = 0;
962     bool SwapOperands = false;
963 
964     if (CMPI1->getOpcode() != CMPI2->getOpcode()) {
965       // Typically, unsigned comparison is used for equality check, but
966       // we replace it with a signed comparison if the comparison
967       // to be merged is a signed comparison.
968       // In other cases of opcode mismatch, we cannot optimize this.
969       if (isEqOrNe(BI2) &&
970           CMPI1->getOpcode() == getSignedCmpOpCode(CMPI2->getOpcode()))
971         NewOpCode = CMPI1->getOpcode();
972       else if (isEqOrNe(BI1) &&
973                getSignedCmpOpCode(CMPI1->getOpcode()) == CMPI2->getOpcode())
974         NewOpCode = CMPI2->getOpcode();
975       else continue;
976     }
977 
978     if (CMPI1->getOperand(2).isReg() && CMPI2->getOperand(2).isReg()) {
979       // In case of comparisons between two registers, these two registers
980       // must be same to merge two comparisons.
981       unsigned Cmp1Operand1 = getSrcVReg(CMPI1->getOperand(1).getReg(),
982                                          nullptr, nullptr, MRI);
983       unsigned Cmp1Operand2 = getSrcVReg(CMPI1->getOperand(2).getReg(),
984                                          nullptr, nullptr, MRI);
985       unsigned Cmp2Operand1 = getSrcVReg(CMPI2->getOperand(1).getReg(),
986                                          MBB1, &MBB2, MRI);
987       unsigned Cmp2Operand2 = getSrcVReg(CMPI2->getOperand(2).getReg(),
988                                          MBB1, &MBB2, MRI);
989 
990       if (Cmp1Operand1 == Cmp2Operand1 && Cmp1Operand2 == Cmp2Operand2) {
991         // Same pair of registers in the same order; ready to merge as is.
992       }
993       else if (Cmp1Operand1 == Cmp2Operand2 && Cmp1Operand2 == Cmp2Operand1) {
994         // Same pair of registers in different order.
995         // We reverse the predicate to merge compare instructions.
996         PPC::Predicate Pred = (PPC::Predicate)BI2->getOperand(0).getImm();
997         NewPredicate2 = (unsigned)PPC::getSwappedPredicate(Pred);
998         // In case of partial redundancy, we need to swap operands
999         // in another compare instruction.
1000         SwapOperands = true;
1001       }
1002       else continue;
1003     }
1004     else if (CMPI1->getOperand(2).isImm() && CMPI2->getOperand(2).isImm()) {
1005       // In case of comparisons between a register and an immediate,
1006       // the operand register must be same for two compare instructions.
1007       unsigned Cmp1Operand1 = getSrcVReg(CMPI1->getOperand(1).getReg(),
1008                                          nullptr, nullptr, MRI);
1009       unsigned Cmp2Operand1 = getSrcVReg(CMPI2->getOperand(1).getReg(),
1010                                          MBB1, &MBB2, MRI);
1011       if (Cmp1Operand1 != Cmp2Operand1)
1012         continue;
1013 
1014       NewImm1 = Imm1 = (int16_t)CMPI1->getOperand(2).getImm();
1015       NewImm2 = Imm2 = (int16_t)CMPI2->getOperand(2).getImm();
1016 
1017       // If immediate are not same, we try to adjust by changing predicate;
1018       // e.g. GT imm means GE (imm+1).
1019       if (Imm1 != Imm2 && (!isEqOrNe(BI2) || !isEqOrNe(BI1))) {
1020         int Diff = Imm1 - Imm2;
1021         if (Diff < -2 || Diff > 2)
1022           continue;
1023 
1024         unsigned PredToInc1 = getPredicateToIncImm(BI1, CMPI1);
1025         unsigned PredToDec1 = getPredicateToDecImm(BI1, CMPI1);
1026         unsigned PredToInc2 = getPredicateToIncImm(BI2, CMPI2);
1027         unsigned PredToDec2 = getPredicateToDecImm(BI2, CMPI2);
1028         if (Diff == 2) {
1029           if (PredToInc2 && PredToDec1) {
1030             NewPredicate2 = PredToInc2;
1031             NewPredicate1 = PredToDec1;
1032             NewImm2++;
1033             NewImm1--;
1034           }
1035         }
1036         else if (Diff == 1) {
1037           if (PredToInc2) {
1038             NewImm2++;
1039             NewPredicate2 = PredToInc2;
1040           }
1041           else if (PredToDec1) {
1042             NewImm1--;
1043             NewPredicate1 = PredToDec1;
1044           }
1045         }
1046         else if (Diff == -1) {
1047           if (PredToDec2) {
1048             NewImm2--;
1049             NewPredicate2 = PredToDec2;
1050           }
1051           else if (PredToInc1) {
1052             NewImm1++;
1053             NewPredicate1 = PredToInc1;
1054           }
1055         }
1056         else if (Diff == -2) {
1057           if (PredToDec2 && PredToInc1) {
1058             NewPredicate2 = PredToDec2;
1059             NewPredicate1 = PredToInc1;
1060             NewImm2--;
1061             NewImm1++;
1062           }
1063         }
1064       }
1065 
1066       // We cannnot merge two compares if the immediates are not same.
1067       if (NewImm2 != NewImm1)
1068         continue;
1069     }
1070 
1071     DEBUG(dbgs() << "Optimize two pairs of compare and branch:\n");
1072     DEBUG(CMPI1->dump());
1073     DEBUG(BI1->dump());
1074     DEBUG(CMPI2->dump());
1075     DEBUG(BI2->dump());
1076 
1077     // We adjust opcode, predicates and immediate as we determined above.
1078     if (NewOpCode != 0 && NewOpCode != CMPI1->getOpcode()) {
1079       CMPI1->setDesc(TII->get(NewOpCode));
1080     }
1081     if (NewPredicate1) {
1082       BI1->getOperand(0).setImm(NewPredicate1);
1083     }
1084     if (NewPredicate2) {
1085       BI2->getOperand(0).setImm(NewPredicate2);
1086     }
1087     if (NewImm1 != Imm1) {
1088       CMPI1->getOperand(2).setImm(NewImm1);
1089     }
1090 
1091     if (IsPartiallyRedundant) {
1092       // We touch up the compare instruction in MBB2 and move it to
1093       // a previous BB to handle partially redundant case.
1094       if (SwapOperands) {
1095         unsigned Op1 = CMPI2->getOperand(1).getReg();
1096         unsigned Op2 = CMPI2->getOperand(2).getReg();
1097         CMPI2->getOperand(1).setReg(Op2);
1098         CMPI2->getOperand(2).setReg(Op1);
1099       }
1100       if (NewImm2 != Imm2)
1101         CMPI2->getOperand(2).setImm(NewImm2);
1102 
1103       for (int I = 1; I <= 2; I++) {
1104         if (CMPI2->getOperand(I).isReg()) {
1105           MachineInstr *Inst = MRI->getVRegDef(CMPI2->getOperand(I).getReg());
1106           if (Inst->getParent() != &MBB2)
1107             continue;
1108 
1109           assert(Inst->getOpcode() == PPC::PHI &&
1110                  "We cannot support if an operand comes from this BB.");
1111           unsigned SrcReg = getIncomingRegForBlock(Inst, MBBtoMoveCmp);
1112           CMPI2->getOperand(I).setReg(SrcReg);
1113         }
1114       }
1115       auto I = MachineBasicBlock::iterator(MBBtoMoveCmp->getFirstTerminator());
1116       MBBtoMoveCmp->splice(I, &MBB2, MachineBasicBlock::iterator(CMPI2));
1117 
1118       DebugLoc DL = CMPI2->getDebugLoc();
1119       unsigned NewVReg = MRI->createVirtualRegister(&PPC::CRRCRegClass);
1120       BuildMI(MBB2, MBB2.begin(), DL,
1121               TII->get(PPC::PHI), NewVReg)
1122         .addReg(BI1->getOperand(1).getReg()).addMBB(MBB1)
1123         .addReg(BI2->getOperand(1).getReg()).addMBB(MBBtoMoveCmp);
1124       BI2->getOperand(1).setReg(NewVReg);
1125     }
1126     else {
1127       // We finally eliminate compare instruction in MBB2.
1128       BI2->getOperand(1).setReg(BI1->getOperand(1).getReg());
1129       CMPI2->eraseFromParent();
1130     }
1131     BI2->getOperand(1).setIsKill(true);
1132     BI1->getOperand(1).setIsKill(false);
1133 
1134     DEBUG(dbgs() << "into a compare and two branches:\n");
1135     DEBUG(CMPI1->dump());
1136     DEBUG(BI1->dump());
1137     DEBUG(BI2->dump());
1138     if (IsPartiallyRedundant) {
1139       DEBUG(dbgs() << "The following compare is moved into BB#" <<
1140             MBBtoMoveCmp->getNumber() << " to handle partial redundancy.\n");
1141       DEBUG(CMPI2->dump());
1142     }
1143 
1144     Simplified = true;
1145   }
1146 
1147   return Simplified;
1148 }
1149 
1150 // This is used to find the "true" source register for an
1151 // XXPERMDI instruction, since MachineCSE does not handle the
1152 // "copy-like" operations (Copy and SubregToReg).  Returns
1153 // the original SrcReg unless it is the target of a copy-like
1154 // operation, in which case we chain backwards through all
1155 // such operations to the ultimate source register.  If a
1156 // physical register is encountered, we stop the search.
1157 unsigned PPCMIPeephole::lookThruCopyLike(unsigned SrcReg) {
1158 
1159   while (true) {
1160 
1161     MachineInstr *MI = MRI->getVRegDef(SrcReg);
1162     if (!MI->isCopyLike())
1163       return SrcReg;
1164 
1165     unsigned CopySrcReg;
1166     if (MI->isCopy())
1167       CopySrcReg = MI->getOperand(1).getReg();
1168     else {
1169       assert(MI->isSubregToReg() && "bad opcode for lookThruCopyLike");
1170       CopySrcReg = MI->getOperand(2).getReg();
1171     }
1172 
1173     if (!TargetRegisterInfo::isVirtualRegister(CopySrcReg))
1174       return CopySrcReg;
1175 
1176     SrcReg = CopySrcReg;
1177   }
1178 }
1179 
1180 } // end default namespace
1181 
1182 INITIALIZE_PASS_BEGIN(PPCMIPeephole, DEBUG_TYPE,
1183                       "PowerPC MI Peephole Optimization", false, false)
1184 INITIALIZE_PASS_END(PPCMIPeephole, DEBUG_TYPE,
1185                     "PowerPC MI Peephole Optimization", false, false)
1186 
1187 char PPCMIPeephole::ID = 0;
1188 FunctionPass*
1189 llvm::createPPCMIPeepholePass() { return new PPCMIPeephole(); }
1190 
1191