llvm.org GIT mirror llvm / 25e2616 lib / Target / AMDGPU / SIWholeQuadMode.cpp
25e2616

Tree @25e2616 (Download .tar.gz)

SIWholeQuadMode.cpp @25e2616raw · history · blame

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
//===-- SIWholeQuadMode.cpp - enter and suspend whole quad mode -----------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
/// \file
/// \brief This pass adds instructions to enable whole quad mode for pixel
/// shaders.
///
/// Whole quad mode is required for derivative computations, but it interferes
/// with shader side effects (stores and atomics). This pass is run on the
/// scheduled machine IR but before register coalescing, so that machine SSA is
/// available for analysis. It ensures that WQM is enabled when necessary, but
/// disabled around stores and atomics.
///
/// When necessary, this pass creates a function prolog
///
///   S_MOV_B64 LiveMask, EXEC
///   S_WQM_B64 EXEC, EXEC
///
/// to enter WQM at the top of the function and surrounds blocks of Exact
/// instructions by
///
///   S_AND_SAVEEXEC_B64 Tmp, LiveMask
///   ...
///   S_MOV_B64 EXEC, Tmp
///
/// In order to avoid excessive switching during sequences of Exact
/// instructions, the pass first analyzes which instructions must be run in WQM
/// (aka which instructions produce values that lead to derivative
/// computations).
///
/// Basic blocks are always exited in WQM as long as some successor needs WQM.
///
/// There is room for improvement given better control flow analysis:
///
///  (1) at the top level (outside of control flow statements, and as long as
///      kill hasn't been used), one SGPR can be saved by recovering WQM from
///      the LiveMask (this is implemented for the entry block).
///
///  (2) when entire regions (e.g. if-else blocks or entire loops) only
///      consist of exact and don't-care instructions, the switch only has to
///      be done at the entry and exit points rather than potentially in each
///      block of the region.
///
//===----------------------------------------------------------------------===//

#include "AMDGPU.h"
#include "AMDGPUSubtarget.h"
#include "SIInstrInfo.h"
#include "SIMachineFunctionInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"

using namespace llvm;

#define DEBUG_TYPE "si-wqm"

namespace {

enum {
  StateWQM = 0x1,
  StateExact = 0x2,
};

struct InstrInfo {
  char Needs = 0;
  char OutNeeds = 0;
};

struct BlockInfo {
  char Needs = 0;
  char InNeeds = 0;
  char OutNeeds = 0;
};

struct WorkItem {
  MachineBasicBlock *MBB = nullptr;
  MachineInstr *MI = nullptr;

  WorkItem() {}
  WorkItem(MachineBasicBlock *MBB) : MBB(MBB) {}
  WorkItem(MachineInstr *MI) : MI(MI) {}
};

class SIWholeQuadMode : public MachineFunctionPass {
private:
  const SIInstrInfo *TII;
  const SIRegisterInfo *TRI;
  MachineRegisterInfo *MRI;
  LiveIntervals *LIS;

  DenseMap<const MachineInstr *, InstrInfo> Instructions;
  DenseMap<MachineBasicBlock *, BlockInfo> Blocks;
  SmallVector<const MachineInstr *, 2> ExecExports;
  SmallVector<MachineInstr *, 1> LiveMaskQueries;

  void markInstruction(MachineInstr &MI, char Flag,
                       std::vector<WorkItem> &Worklist);
  char scanInstructions(MachineFunction &MF, std::vector<WorkItem> &Worklist);
  void propagateInstruction(MachineInstr &MI, std::vector<WorkItem> &Worklist);
  void propagateBlock(MachineBasicBlock &MBB, std::vector<WorkItem> &Worklist);
  char analyzeFunction(MachineFunction &MF);

  void toExact(MachineBasicBlock &MBB, MachineBasicBlock::iterator Before,
               unsigned SaveWQM, unsigned LiveMaskReg);
  void toWQM(MachineBasicBlock &MBB, MachineBasicBlock::iterator Before,
             unsigned SavedWQM);
  void processBlock(MachineBasicBlock &MBB, unsigned LiveMaskReg, bool isEntry);

  void lowerLiveMaskQueries(unsigned LiveMaskReg);

public:
  static char ID;

  SIWholeQuadMode() :
    MachineFunctionPass(ID) { }

  bool runOnMachineFunction(MachineFunction &MF) override;

  const char *getPassName() const override {
    return "SI Whole Quad Mode";
  }

  void getAnalysisUsage(AnalysisUsage &AU) const override {
    AU.addRequired<LiveIntervals>();
    AU.setPreservesCFG();
    MachineFunctionPass::getAnalysisUsage(AU);
  }
};

} // End anonymous namespace

char SIWholeQuadMode::ID = 0;

INITIALIZE_PASS_BEGIN(SIWholeQuadMode, DEBUG_TYPE, "SI Whole Quad Mode", false,
                      false)
INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
INITIALIZE_PASS_END(SIWholeQuadMode, DEBUG_TYPE, "SI Whole Quad Mode", false,
                    false)

char &llvm::SIWholeQuadModeID = SIWholeQuadMode::ID;

FunctionPass *llvm::createSIWholeQuadModePass() {
  return new SIWholeQuadMode;
}

void SIWholeQuadMode::markInstruction(MachineInstr &MI, char Flag,
                                      std::vector<WorkItem> &Worklist) {
  InstrInfo &II = Instructions[&MI];

  assert(Flag == StateWQM || Flag == StateExact);

  // Ignore if the instruction is already marked. The typical case is that we
  // mark an instruction WQM multiple times, but for atomics it can happen that
  // Flag is StateWQM, but Needs is already set to StateExact. In this case,
  // letting the atomic run in StateExact is correct as per the relevant specs.
  if (II.Needs)
    return;

  II.Needs = Flag;
  Worklist.push_back(&MI);
}

// Scan instructions to determine which ones require an Exact execmask and
// which ones seed WQM requirements.
char SIWholeQuadMode::scanInstructions(MachineFunction &MF,
                                       std::vector<WorkItem> &Worklist) {
  char GlobalFlags = 0;
  bool WQMOutputs = MF.getFunction()->hasFnAttribute("amdgpu-ps-wqm-outputs");

  for (auto BI = MF.begin(), BE = MF.end(); BI != BE; ++BI) {
    MachineBasicBlock &MBB = *BI;

    for (auto II = MBB.begin(), IE = MBB.end(); II != IE; ++II) {
      MachineInstr &MI = *II;
      unsigned Opcode = MI.getOpcode();
      char Flags = 0;

      if (TII->isWQM(Opcode) || TII->isDS(Opcode)) {
        Flags = StateWQM;
      } else if (TII->isDisableWQM(MI)) {
        Flags = StateExact;
      } else {
        // Handle export instructions with the exec mask valid flag set
        if (Opcode == AMDGPU::EXP) {
          if (MI.getOperand(4).getImm() != 0)
            ExecExports.push_back(&MI);
        } else if (Opcode == AMDGPU::SI_PS_LIVE) {
          LiveMaskQueries.push_back(&MI);
        } else if (WQMOutputs) {
          // The function is in machine SSA form, which means that physical
          // VGPRs correspond to shader inputs and outputs. Inputs are
          // only used, outputs are only defined.
          for (const MachineOperand &MO : MI.defs()) {
            if (!MO.isReg())
              continue;

            unsigned Reg = MO.getReg();

            if (!TRI->isVirtualRegister(Reg) &&
                TRI->hasVGPRs(TRI->getPhysRegClass(Reg))) {
              Flags = StateWQM;
              break;
            }
          }
        }

        if (!Flags)
          continue;
      }

      markInstruction(MI, Flags, Worklist);
      GlobalFlags |= Flags;
    }
  }

  return GlobalFlags;
}

void SIWholeQuadMode::propagateInstruction(MachineInstr &MI,
                                           std::vector<WorkItem>& Worklist) {
  MachineBasicBlock *MBB = MI.getParent();
  InstrInfo II = Instructions[&MI]; // take a copy to prevent dangling references
  BlockInfo &BI = Blocks[MBB];

  // Control flow-type instructions and stores to temporary memory that are
  // followed by WQM computations must themselves be in WQM.
  if ((II.OutNeeds & StateWQM) && !II.Needs &&
      (MI.isTerminator() || (TII->usesVM_CNT(MI) && MI.mayStore()))) {
    Instructions[&MI].Needs = StateWQM;
    II.Needs = StateWQM;
  }

  // Propagate to block level
  BI.Needs |= II.Needs;
  if ((BI.InNeeds | II.Needs) != BI.InNeeds) {
    BI.InNeeds |= II.Needs;
    Worklist.push_back(MBB);
  }

  // Propagate backwards within block
  if (MachineInstr *PrevMI = MI.getPrevNode()) {
    char InNeeds = II.Needs | II.OutNeeds;
    if (!PrevMI->isPHI()) {
      InstrInfo &PrevII = Instructions[PrevMI];
      if ((PrevII.OutNeeds | InNeeds) != PrevII.OutNeeds) {
        PrevII.OutNeeds |= InNeeds;
        Worklist.push_back(PrevMI);
      }
    }
  }

  // Propagate WQM flag to instruction inputs
  assert(II.Needs != (StateWQM | StateExact));
  if (II.Needs != StateWQM)
    return;

  for (const MachineOperand &Use : MI.uses()) {
    if (!Use.isReg() || !Use.isUse())
      continue;

    unsigned Reg = Use.getReg();

    // Handle physical registers that we need to track; this is mostly relevant
    // for VCC, which can appear as the (implicit) input of a uniform branch,
    // e.g. when a loop counter is stored in a VGPR.
    if (!TargetRegisterInfo::isVirtualRegister(Reg)) {
      if (Reg == AMDGPU::EXEC)
        continue;

      for (MCRegUnitIterator RegUnit(Reg, TRI); RegUnit.isValid(); ++RegUnit) {
        LiveRange &LR = LIS->getRegUnit(*RegUnit);
        const VNInfo *Value = LR.Query(LIS->getInstructionIndex(MI)).valueIn();
        if (!Value)
          continue;

        // Since we're in machine SSA, we do not need to track physical
        // registers across basic blocks.
        if (Value->isPHIDef())
          continue;

        markInstruction(*LIS->getInstructionFromIndex(Value->def), StateWQM,
                        Worklist);
      }

      continue;
    }

    for (MachineInstr &DefMI : MRI->def_instructions(Use.getReg()))
      markInstruction(DefMI, StateWQM, Worklist);
  }
}

void SIWholeQuadMode::propagateBlock(MachineBasicBlock &MBB,
                                     std::vector<WorkItem>& Worklist) {
  BlockInfo BI = Blocks[&MBB]; // Make a copy to prevent dangling references.

  // Propagate through instructions
  if (!MBB.empty()) {
    MachineInstr *LastMI = &*MBB.rbegin();
    InstrInfo &LastII = Instructions[LastMI];
    if ((LastII.OutNeeds | BI.OutNeeds) != LastII.OutNeeds) {
      LastII.OutNeeds |= BI.OutNeeds;
      Worklist.push_back(LastMI);
    }
  }

  // Predecessor blocks must provide for our WQM/Exact needs.
  for (MachineBasicBlock *Pred : MBB.predecessors()) {
    BlockInfo &PredBI = Blocks[Pred];
    if ((PredBI.OutNeeds | BI.InNeeds) == PredBI.OutNeeds)
      continue;

    PredBI.OutNeeds |= BI.InNeeds;
    PredBI.InNeeds |= BI.InNeeds;
    Worklist.push_back(Pred);
  }

  // All successors must be prepared to accept the same set of WQM/Exact data.
  for (MachineBasicBlock *Succ : MBB.successors()) {
    BlockInfo &SuccBI = Blocks[Succ];
    if ((SuccBI.InNeeds | BI.OutNeeds) == SuccBI.InNeeds)
      continue;

    SuccBI.InNeeds |= BI.OutNeeds;
    Worklist.push_back(Succ);
  }
}

char SIWholeQuadMode::analyzeFunction(MachineFunction &MF) {
  std::vector<WorkItem> Worklist;
  char GlobalFlags = scanInstructions(MF, Worklist);

  while (!Worklist.empty()) {
    WorkItem WI = Worklist.back();
    Worklist.pop_back();

    if (WI.MI)
      propagateInstruction(*WI.MI, Worklist);
    else
      propagateBlock(*WI.MBB, Worklist);
  }

  return GlobalFlags;
}

void SIWholeQuadMode::toExact(MachineBasicBlock &MBB,
                              MachineBasicBlock::iterator Before,
                              unsigned SaveWQM, unsigned LiveMaskReg) {
  if (SaveWQM) {
    BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::S_AND_SAVEEXEC_B64),
            SaveWQM)
        .addReg(LiveMaskReg);
  } else {
    BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::S_AND_B64),
            AMDGPU::EXEC)
        .addReg(AMDGPU::EXEC)
        .addReg(LiveMaskReg);
  }
}

void SIWholeQuadMode::toWQM(MachineBasicBlock &MBB,
                            MachineBasicBlock::iterator Before,
                            unsigned SavedWQM) {
  if (SavedWQM) {
    BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::COPY), AMDGPU::EXEC)
        .addReg(SavedWQM);
  } else {
    BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::S_WQM_B64),
            AMDGPU::EXEC)
        .addReg(AMDGPU::EXEC);
  }
}

void SIWholeQuadMode::processBlock(MachineBasicBlock &MBB, unsigned LiveMaskReg,
                                   bool isEntry) {
  auto BII = Blocks.find(&MBB);
  if (BII == Blocks.end())
    return;

  const BlockInfo &BI = BII->second;

  if (!(BI.InNeeds & StateWQM))
    return;

  // This is a non-entry block that is WQM throughout, so no need to do
  // anything.
  if (!isEntry && !(BI.Needs & StateExact) && BI.OutNeeds != StateExact)
    return;

  unsigned SavedWQMReg = 0;
  bool WQMFromExec = isEntry;
  char State = isEntry ? StateExact : StateWQM;

  auto II = MBB.getFirstNonPHI(), IE = MBB.end();
  while (II != IE) {
    MachineInstr &MI = *II;
    ++II;

    // Skip instructions that are not affected by EXEC
    if (TII->isScalarUnit(MI) && !MI.isTerminator())
      continue;

    // Generic instructions such as COPY will either disappear by register
    // coalescing or be lowered to SALU or VALU instructions.
    if (TargetInstrInfo::isGenericOpcode(MI.getOpcode())) {
      if (MI.getNumExplicitOperands() >= 1) {
        const MachineOperand &Op = MI.getOperand(0);
        if (Op.isReg()) {
          if (TRI->isSGPRReg(*MRI, Op.getReg())) {
            // SGPR instructions are not affected by EXEC
            continue;
          }
        }
      }
    }

    char Needs = 0;
    char OutNeeds = 0;
    auto InstrInfoIt = Instructions.find(&MI);
    if (InstrInfoIt != Instructions.end()) {
      Needs = InstrInfoIt->second.Needs;
      OutNeeds = InstrInfoIt->second.OutNeeds;

      // Make sure to switch to Exact mode before the end of the block when
      // Exact and only Exact is needed further downstream.
      if (OutNeeds == StateExact && MI.isTerminator()) {
        assert(Needs == 0);
        Needs = StateExact;
      }
    }

    // State switching
    if (Needs && State != Needs) {
      if (Needs == StateExact) {
        assert(!SavedWQMReg);

        if (!WQMFromExec && (OutNeeds & StateWQM))
          SavedWQMReg = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);

        toExact(MBB, &MI, SavedWQMReg, LiveMaskReg);
      } else {
        assert(WQMFromExec == (SavedWQMReg == 0));
        toWQM(MBB, &MI, SavedWQMReg);
        SavedWQMReg = 0;
      }

      State = Needs;
    }
  }

  if ((BI.OutNeeds & StateWQM) && State != StateWQM) {
    assert(WQMFromExec == (SavedWQMReg == 0));
    toWQM(MBB, MBB.end(), SavedWQMReg);
  } else if (BI.OutNeeds == StateExact && State != StateExact) {
    toExact(MBB, MBB.end(), 0, LiveMaskReg);
  }
}

void SIWholeQuadMode::lowerLiveMaskQueries(unsigned LiveMaskReg) {
  for (MachineInstr *MI : LiveMaskQueries) {
    const DebugLoc &DL = MI->getDebugLoc();
    unsigned Dest = MI->getOperand(0).getReg();
    BuildMI(*MI->getParent(), MI, DL, TII->get(AMDGPU::COPY), Dest)
        .addReg(LiveMaskReg);
    MI->eraseFromParent();
  }
}

bool SIWholeQuadMode::runOnMachineFunction(MachineFunction &MF) {
  if (MF.getFunction()->getCallingConv() != CallingConv::AMDGPU_PS)
    return false;

  Instructions.clear();
  Blocks.clear();
  ExecExports.clear();
  LiveMaskQueries.clear();

  const SISubtarget &ST = MF.getSubtarget<SISubtarget>();

  TII = ST.getInstrInfo();
  TRI = &TII->getRegisterInfo();
  MRI = &MF.getRegInfo();
  LIS = &getAnalysis<LiveIntervals>();

  char GlobalFlags = analyzeFunction(MF);
  if (!(GlobalFlags & StateWQM)) {
    lowerLiveMaskQueries(AMDGPU::EXEC);
    return !LiveMaskQueries.empty();
  }

  // Store a copy of the original live mask when required
  unsigned LiveMaskReg = 0;
  {
    MachineBasicBlock &Entry = MF.front();
    MachineBasicBlock::iterator EntryMI = Entry.getFirstNonPHI();

    if (GlobalFlags & StateExact || !LiveMaskQueries.empty()) {
      LiveMaskReg = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
      BuildMI(Entry, EntryMI, DebugLoc(), TII->get(AMDGPU::COPY), LiveMaskReg)
          .addReg(AMDGPU::EXEC);
    }

    if (GlobalFlags == StateWQM) {
      // For a shader that needs only WQM, we can just set it once.
      BuildMI(Entry, EntryMI, DebugLoc(), TII->get(AMDGPU::S_WQM_B64),
              AMDGPU::EXEC)
          .addReg(AMDGPU::EXEC);

      lowerLiveMaskQueries(LiveMaskReg);
      // EntryMI may become invalid here
      return true;
    }
  }

  lowerLiveMaskQueries(LiveMaskReg);

  // Handle the general case
  for (auto BII : Blocks)
    processBlock(*BII.first, LiveMaskReg, BII.first == &*MF.begin());

  return true;
}