[AMDGPU][Waitcnt] NFC. Cleanup some code

[msearles]

Summary

Naming consistency:

• s/SWaitcnt/Waitcnt s/WaitCnt/Waitcnt

[arsenm]

LGTM

[msearles]

https://reviews.llvm.org/rL330730

lib/Target/AMDGPU/SIInsertWaitcnts.cpp

public:
   void addKillWaitBracket(BlockWaitcntBrackets *Bracket) {
     // The waitcnt information is copied because it changes as the block is
     // traversed.
     KillWaitBrackets.push_back(
         llvm::make_unique<BlockWaitcntBrackets>(*Bracket));
   }
 
   bool mayAccessLDSThroughFlat(const MachineInstr &MI) const;
-  void generateSWaitCntInstBefore(MachineInstr &MI,
+  void generateWaitcntInstBefore(MachineInstr &MI,
                                   BlockWaitcntBrackets *ScoreBrackets);
-  void updateEventWaitCntAfter(MachineInstr &Inst,
+  void updateEventWaitcntAfter(MachineInstr &Inst,
                                BlockWaitcntBrackets *ScoreBrackets);
   void mergeInputScoreBrackets(MachineBasicBlock &Block);
   bool isLoopBottom(const MachineLoop *Loop, const MachineBasicBlock *Block);
   unsigned countNumBottomBlocks(const MachineLoop *Loop);
   void insertWaitcntInBlock(MachineFunction &MF, MachineBasicBlock &Block);
   void insertWaitcntBeforeCF(MachineBasicBlock &Block, MachineInstr *Inst);
   bool isWaitcntStronger(unsigned LHS, unsigned RHS);
   unsigned combineWaitcnt(unsigned LHS, unsigned RHS);

 ///  but instructions of different types can complete out of order.
 ///  We rely on this in-order completion
 ///  and simply assign a score to the memory access instructions.
 ///  We keep track of the active "score bracket" to determine
 ///  if an access of a memory read requires an s_waitcnt
 ///  and if so what the value of each counter is.
 ///  The "score bracket" is bound by the lower bound and upper bound
 ///  scores (*_score_LB and *_score_ub respectively).
-void SIInsertWaitcnts::generateSWaitCntInstBefore(
+void SIInsertWaitcnts::generateWaitcntInstBefore(
     MachineInstr &MI, BlockWaitcntBrackets *ScoreBrackets) {
   // To emit, or not to emit - that's the question!
   // Start with an assumption that there is no need to emit.
-  unsigned int EmitSwaitcnt = 0;
+  unsigned int EmitWaitcnt = 0;
   // No need to wait before phi. If a phi-move exists, then the wait should
   // has been inserted before the move. If a phi-move does not exist, then
   // wait should be inserted before the real use. The same is true for
   // sc-merge. It is not a coincident that all these cases correspond to the
   // instructions that are skipped in the assembling loop.
   bool NeedLineMapping = false; // TODO: Check on this.
   if (MI.isDebugValue() &&
       // TODO: any other opcode?
       !NeedLineMapping) {
     return;
   }
 
   // See if an s_waitcnt is forced at block entry, or is needed at
   // program end.
   if (ScoreBrackets->getWaitAtBeginning()) {
     // Note that we have already cleared the state, so we don't need to update
     // it.
     ScoreBrackets->clearWaitAtBeginning();
     for (enum InstCounterType T = VM_CNT; T < NUM_INST_CNTS;
          T = (enum InstCounterType)(T + 1)) {
-      EmitSwaitcnt |= CNT_MASK(T);
+      EmitWaitcnt |= CNT_MASK(T);
       ScoreBrackets->setScoreLB(T, ScoreBrackets->getScoreUB(T));
     }
   }
 
   // See if this instruction has a forced S_WAITCNT VM.
   // TODO: Handle other cases of NeedsWaitcntVmBefore()
   else if (MI.getOpcode() == AMDGPU::BUFFER_WBINVL1 ||
            MI.getOpcode() == AMDGPU::BUFFER_WBINVL1_SC ||
            MI.getOpcode() == AMDGPU::BUFFER_WBINVL1_VOL) {
-    EmitSwaitcnt |=
+    EmitWaitcnt |=
         ScoreBrackets->updateByWait(VM_CNT, ScoreBrackets->getScoreUB(VM_CNT));
   }
 
   // All waits must be resolved at call return.
   // NOTE: this could be improved with knowledge of all call sites or
   //   with knowledge of the called routines.
   if (MI.getOpcode() == AMDGPU::RETURN ||
       MI.getOpcode() == AMDGPU::SI_RETURN_TO_EPILOG ||
       MI.getOpcode() == AMDGPU::S_SETPC_B64_return) {
     for (enum InstCounterType T = VM_CNT; T < NUM_INST_CNTS;
          T = (enum InstCounterType)(T + 1)) {
       if (ScoreBrackets->getScoreUB(T) > ScoreBrackets->getScoreLB(T)) {
         ScoreBrackets->setScoreLB(T, ScoreBrackets->getScoreUB(T));
-        EmitSwaitcnt |= CNT_MASK(T);
+        EmitWaitcnt |= CNT_MASK(T);
       }
     }
   }
   // Resolve vm waits before gs-done.
   else if ((MI.getOpcode() == AMDGPU::S_SENDMSG ||
             MI.getOpcode() == AMDGPU::S_SENDMSGHALT) &&
            ((MI.getOperand(0).getImm() & AMDGPU::SendMsg::ID_MASK_) ==
             AMDGPU::SendMsg::ID_GS_DONE)) {
     if (ScoreBrackets->getScoreUB(VM_CNT) > ScoreBrackets->getScoreLB(VM_CNT)) {
       ScoreBrackets->setScoreLB(VM_CNT, ScoreBrackets->getScoreUB(VM_CNT));
-      EmitSwaitcnt |= CNT_MASK(VM_CNT);
+      EmitWaitcnt |= CNT_MASK(VM_CNT);
     }
   }
 #if 0 // TODO: the following blocks of logic when we have fence.
   else if (MI.getOpcode() == SC_FENCE) {
     const unsigned int group_size =
       context->shader_info->GetMaxThreadGroupSize();
     // group_size == 0 means thread group size is unknown at compile time
     const bool group_is_multi_wave =
       (group_size == 0 || group_size > target_info->GetWaveFrontSize());
     const bool fence_is_global = !((SCInstInternalMisc*)Inst)->IsGroupFence();
 
     for (unsigned int i = 0; i < Inst->NumSrcOperands(); i++) {
       SCRegType src_type = Inst->GetSrcType(i);
       switch (src_type) {
         case SCMEM_LDS:
           if (group_is_multi_wave ||
             context->OptFlagIsOn(OPT_R1100_LDSMEM_FENCE_CHICKEN_BIT)) {
-            EmitSwaitcnt |= ScoreBrackets->updateByWait(LGKM_CNT,
+            EmitWaitcnt |= ScoreBrackets->updateByWait(LGKM_CNT,
                                ScoreBrackets->getScoreUB(LGKM_CNT));
             // LDS may have to wait for VM_CNT after buffer load to LDS
             if (target_info->HasBufferLoadToLDS()) {
-              EmitSwaitcnt |= ScoreBrackets->updateByWait(VM_CNT,
+              EmitWaitcnt |= ScoreBrackets->updateByWait(VM_CNT,
                                  ScoreBrackets->getScoreUB(VM_CNT));
             }
           }
           break;
 
         case SCMEM_GDS:
           if (group_is_multi_wave || fence_is_global) {
-            EmitSwaitcnt |= ScoreBrackets->updateByWait(EXP_CNT,
+            EmitWaitcnt |= ScoreBrackets->updateByWait(EXP_CNT,
               ScoreBrackets->getScoreUB(EXP_CNT));
-            EmitSwaitcnt |= ScoreBrackets->updateByWait(LGKM_CNT,
+            EmitWaitcnt |= ScoreBrackets->updateByWait(LGKM_CNT,
               ScoreBrackets->getScoreUB(LGKM_CNT));
           }
           break;
 
         case SCMEM_UAV:
         case SCMEM_TFBUF:
         case SCMEM_RING:
         case SCMEM_SCATTER:
           if (group_is_multi_wave || fence_is_global) {
-            EmitSwaitcnt |= ScoreBrackets->updateByWait(EXP_CNT,
+            EmitWaitcnt |= ScoreBrackets->updateByWait(EXP_CNT,
               ScoreBrackets->getScoreUB(EXP_CNT));
-            EmitSwaitcnt |= ScoreBrackets->updateByWait(VM_CNT,
+            EmitWaitcnt |= ScoreBrackets->updateByWait(VM_CNT,
               ScoreBrackets->getScoreUB(VM_CNT));
           }
           break;
 
         case SCMEM_SCRATCH:
         default:
           break;
       }
     }
   }
 #endif
 
   // Export & GDS instructions do not read the EXEC mask until after the export
   // is granted (which can occur well after the instruction is issued).
   // The shader program must flush all EXP operations on the export-count
   // before overwriting the EXEC mask.
   else {
     if (MI.modifiesRegister(AMDGPU::EXEC, TRI)) {
       // Export and GDS are tracked individually, either may trigger a waitcnt
       // for EXEC.
-      EmitSwaitcnt |= ScoreBrackets->updateByWait(
+      EmitWaitcnt |= ScoreBrackets->updateByWait(
           EXP_CNT, ScoreBrackets->getEventUB(EXP_GPR_LOCK));
-      EmitSwaitcnt |= ScoreBrackets->updateByWait(
+      EmitWaitcnt |= ScoreBrackets->updateByWait(
           EXP_CNT, ScoreBrackets->getEventUB(EXP_PARAM_ACCESS));
-      EmitSwaitcnt |= ScoreBrackets->updateByWait(
+      EmitWaitcnt |= ScoreBrackets->updateByWait(
           EXP_CNT, ScoreBrackets->getEventUB(EXP_POS_ACCESS));
-      EmitSwaitcnt |= ScoreBrackets->updateByWait(
+      EmitWaitcnt |= ScoreBrackets->updateByWait(
           EXP_CNT, ScoreBrackets->getEventUB(GDS_GPR_LOCK));
     }
 
 #if 0 // TODO: the following code to handle CALL.
     // The argument passing for CALLs should suffice for VM_CNT and LGKM_CNT.
     // However, there is a problem with EXP_CNT, because the call cannot
     // easily tell if a register is used in the function, and if it did, then
     // the referring instruction would have to have an S_WAITCNT, which is
     // dependent on all call sites. So Instead, force S_WAITCNT for EXP_CNTs
     // before the call.
     if (MI.getOpcode() == SC_CALL) {
       if (ScoreBrackets->getScoreUB(EXP_CNT) >
         ScoreBrackets->getScoreLB(EXP_CNT)) {
         ScoreBrackets->setScoreLB(EXP_CNT, ScoreBrackets->getScoreUB(EXP_CNT));
-        EmitSwaitcnt |= CNT_MASK(EXP_CNT);
+        EmitWaitcnt |= CNT_MASK(EXP_CNT);
       }
     }
 #endif
 
     // FIXME: Should not be relying on memoperands.
     // Look at the source operands of every instruction to see if
     // any of them results from a previous memory operation that affects
     // its current usage. If so, an s_waitcnt instruction needs to be
     // emitted.
     // If the source operand was defined by a load, add the s_waitcnt
     // instruction.
     for (const MachineMemOperand *Memop : MI.memoperands()) {
       unsigned AS = Memop->getAddrSpace();
       if (AS != AMDGPUASI.LOCAL_ADDRESS)
         continue;
       unsigned RegNo = SQ_MAX_PGM_VGPRS + EXTRA_VGPR_LDS;
       // VM_CNT is only relevant to vgpr or LDS.
-      EmitSwaitcnt |= ScoreBrackets->updateByWait(
+      EmitWaitcnt |= ScoreBrackets->updateByWait(
           VM_CNT, ScoreBrackets->getRegScore(RegNo, VM_CNT));
     }
 
     for (unsigned I = 0, E = MI.getNumOperands(); I != E; ++I) {
       const MachineOperand &Op = MI.getOperand(I);
       const MachineRegisterInfo &MRIA = *MRI;
       RegInterval Interval =
           ScoreBrackets->getRegInterval(&MI, TII, MRI, TRI, I, false);
       for (signed RegNo = Interval.first; RegNo < Interval.second; ++RegNo) {
         if (TRI->isVGPR(MRIA, Op.getReg())) {
           // VM_CNT is only relevant to vgpr or LDS.
-          EmitSwaitcnt |= ScoreBrackets->updateByWait(
+          EmitWaitcnt |= ScoreBrackets->updateByWait(
               VM_CNT, ScoreBrackets->getRegScore(RegNo, VM_CNT));
         }
-        EmitSwaitcnt |= ScoreBrackets->updateByWait(
+        EmitWaitcnt |= ScoreBrackets->updateByWait(
             LGKM_CNT, ScoreBrackets->getRegScore(RegNo, LGKM_CNT));
       }
     }
     // End of for loop that looks at all source operands to decide vm_wait_cnt
     // and lgk_wait_cnt.
 
     // Two cases are handled for destination operands:
     // 1) If the destination operand was defined by a load, add the s_waitcnt
     // instruction to guarantee the right WAW order.
     // 2) If a destination operand that was used by a recent export/store ins,
     // add s_waitcnt on exp_cnt to guarantee the WAR order.
     if (MI.mayStore()) {
       // FIXME: Should not be relying on memoperands.
       for (const MachineMemOperand *Memop : MI.memoperands()) {
         unsigned AS = Memop->getAddrSpace();
         if (AS != AMDGPUASI.LOCAL_ADDRESS)
           continue;
         unsigned RegNo = SQ_MAX_PGM_VGPRS + EXTRA_VGPR_LDS;
-        EmitSwaitcnt |= ScoreBrackets->updateByWait(
+        EmitWaitcnt |= ScoreBrackets->updateByWait(
             VM_CNT, ScoreBrackets->getRegScore(RegNo, VM_CNT));
-        EmitSwaitcnt |= ScoreBrackets->updateByWait(
+        EmitWaitcnt |= ScoreBrackets->updateByWait(
             EXP_CNT, ScoreBrackets->getRegScore(RegNo, EXP_CNT));
       }
     }
     for (unsigned I = 0, E = MI.getNumOperands(); I != E; ++I) {
       MachineOperand &Def = MI.getOperand(I);
       const MachineRegisterInfo &MRIA = *MRI;
       RegInterval Interval =
           ScoreBrackets->getRegInterval(&MI, TII, MRI, TRI, I, true);
       for (signed RegNo = Interval.first; RegNo < Interval.second; ++RegNo) {
         if (TRI->isVGPR(MRIA, Def.getReg())) {
-          EmitSwaitcnt |= ScoreBrackets->updateByWait(
+          EmitWaitcnt |= ScoreBrackets->updateByWait(
               VM_CNT, ScoreBrackets->getRegScore(RegNo, VM_CNT));
-          EmitSwaitcnt |= ScoreBrackets->updateByWait(
+          EmitWaitcnt |= ScoreBrackets->updateByWait(
               EXP_CNT, ScoreBrackets->getRegScore(RegNo, EXP_CNT));
         }
-        EmitSwaitcnt |= ScoreBrackets->updateByWait(
+        EmitWaitcnt |= ScoreBrackets->updateByWait(
             LGKM_CNT, ScoreBrackets->getRegScore(RegNo, LGKM_CNT));
       }
     } // End of for loop that looks at all dest operands.
   }
 
   // TODO: Tie force zero to a compiler triage option.
   bool ForceZero = false;
 
   // Check to see if this is an S_BARRIER, and if an implicit S_WAITCNT 0
   // occurs before the instruction. Doing it here prevents any additional
   // S_WAITCNTs from being emitted if the instruction was marked as
   // requiring a WAITCNT beforehand.
   if (MI.getOpcode() == AMDGPU::S_BARRIER &&
       !ST->hasAutoWaitcntBeforeBarrier()) {
-    EmitSwaitcnt |=
+    EmitWaitcnt |=
         ScoreBrackets->updateByWait(VM_CNT, ScoreBrackets->getScoreUB(VM_CNT));
-    EmitSwaitcnt |= ScoreBrackets->updateByWait(
+    EmitWaitcnt |= ScoreBrackets->updateByWait(
         EXP_CNT, ScoreBrackets->getScoreUB(EXP_CNT));
-    EmitSwaitcnt |= ScoreBrackets->updateByWait(
+    EmitWaitcnt |= ScoreBrackets->updateByWait(
         LGKM_CNT, ScoreBrackets->getScoreUB(LGKM_CNT));
   }
 
   // TODO: Remove this work-around, enable the assert for Bug 457939
   //       after fixing the scheduler. Also, the Shader Compiler code is
   //       independent of target.
   if (readsVCCZ(MI) && ST->getGeneration() <= SISubtarget::SEA_ISLANDS) {
     if (ScoreBrackets->getScoreLB(LGKM_CNT) <
             ScoreBrackets->getScoreUB(LGKM_CNT) &&
         ScoreBrackets->hasPendingSMEM()) {
       // Wait on everything, not just LGKM.  vccz reads usually come from
       // terminators, and we always wait on everything at the end of the
       // block, so if we only wait on LGKM here, we might end up with
       // another s_waitcnt inserted right after this if there are non-LGKM
       // instructions still outstanding.
       ForceZero = true;
-      EmitSwaitcnt = true;
+      EmitWaitcnt = true;
     }
   }
 
   // Does this operand processing indicate s_wait counter update?
-  if (EmitSwaitcnt) {
+  if (EmitWaitcnt) {
     int CntVal[NUM_INST_CNTS];
 
     bool UseDefaultWaitcntStrategy = true;
     if (ForceZero) {
       // Force all waitcnts to 0.
       for (enum InstCounterType T = VM_CNT; T < NUM_INST_CNTS;
            T = (enum InstCounterType)(T + 1)) {
         ScoreBrackets->setScoreLB(T, ScoreBrackets->getScoreUB(T));
       }
       CntVal[VM_CNT] = 0;
       CntVal[EXP_CNT] = 0;
       CntVal[LGKM_CNT] = 0;
       UseDefaultWaitcntStrategy = false;
     }
 
     if (UseDefaultWaitcntStrategy) {
       for (enum InstCounterType T = VM_CNT; T < NUM_INST_CNTS;
            T = (enum InstCounterType)(T + 1)) {
-        if (EmitSwaitcnt & CNT_MASK(T)) {
+        if (EmitWaitcnt & CNT_MASK(T)) {
           int Delta =
               ScoreBrackets->getScoreUB(T) - ScoreBrackets->getScoreLB(T);
           int MaxDelta = ScoreBrackets->getWaitCountMax(T);
           if (Delta >= MaxDelta) {
             Delta = -1;
             if (T != EXP_CNT) {
               ScoreBrackets->setScoreLB(
                   T, ScoreBrackets->getScoreUB(T) - MaxDelta);
             }
-            EmitSwaitcnt &= ~CNT_MASK(T);
+            EmitWaitcnt &= ~CNT_MASK(T);
           }
           CntVal[T] = Delta;
         } else {
           // If we are not waiting for a particular counter then encode
           // it as -1 which means "don't care."
           CntVal[T] = -1;
         }
       }
     }
 
     // If we are not waiting on any counter we can skip the wait altogether.
-    if (EmitSwaitcnt != 0) {
+    if (EmitWaitcnt != 0) {
       MachineInstr *OldWaitcnt = ScoreBrackets->getWaitcnt();
       int Imm = (!OldWaitcnt) ? 0 : OldWaitcnt->getOperand(0).getImm();
       if (!OldWaitcnt ||
           (AMDGPU::decodeVmcnt(IV, Imm) !=
                           (CntVal[VM_CNT] & AMDGPU::getVmcntBitMask(IV))) ||
           (AMDGPU::decodeExpcnt(IV, Imm) !=
            (CntVal[EXP_CNT] & AMDGPU::getExpcntBitMask(IV))) ||
           (AMDGPU::decodeLgkmcnt(IV, Imm) !=
for (const MachineMemOperand *Memop : MI.memoperands()) {
     unsigned AS = Memop->getAddrSpace();
     if (AS == AMDGPUASI.LOCAL_ADDRESS || AS == AMDGPUASI.FLAT_ADDRESS)
       return true;
   }
 
   return false;
 }
 
-void SIInsertWaitcnts::updateEventWaitCntAfter(
+void SIInsertWaitcnts::updateEventWaitcntAfter(
     MachineInstr &Inst, BlockWaitcntBrackets *ScoreBrackets) {
   // Now look at the instruction opcode. If it is a memory access
   // instruction, update the upper-bound of the appropriate counter's
   // bracket and the destination operand scores.
   // TODO: Use the (TSFlags & SIInstrFlags::LGKM_CNT) property everywhere.
   if (TII->isDS(Inst) && TII->usesLGKM_CNT(Inst)) {
     if (TII->hasModifiersSet(Inst, AMDGPU::OpName::gds)) {
       ScoreBrackets->updateByEvent(TII, TRI, MRI, GDS_ACCESS, Inst);
if (readsVCCZ(Inst) &&
           ScoreBrackets->hasPendingSMEM()) {
         if (ST->getGeneration() <= SISubtarget::SEA_ISLANDS)
           VCCZBugWorkAround = true;
       }
     }
 
     // Generate an s_waitcnt instruction to be placed before
     // cur_Inst, if needed.
-    generateSWaitCntInstBefore(Inst, ScoreBrackets);
+    generateWaitcntInstBefore(Inst, ScoreBrackets);
 
-    updateEventWaitCntAfter(Inst, ScoreBrackets);
+    updateEventWaitcntAfter(Inst, ScoreBrackets);
 
 #if 0 // TODO: implement resource type check controlled by options with ub = LB.
     // If this instruction generates a S_SETVSKIP because it is an
     // indexed resource, and we are on Tahiti, then it will also force
     // an S_WAITCNT vmcnt(0)
     if (RequireCheckResourceType(Inst, context)) {
       // Force the score to as if an S_WAITCNT vmcnt(0) is emitted.
       ScoreBrackets->setScoreLB(VM_CNT,