Merge llvm trunk r366426, resolve conflicts, and update FREEBSD-Xlist.

1 files changed, 760 insertions, 44 deletions

diff --git a/contrib/llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp b/contrib/llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp
index 69ddbfb53958..885239e2faed 100644
--- a/contrib/llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp
+++ b/contrib/llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp
@@ -1,9 +1,8 @@
 //===-- GCNHazardRecognizers.cpp - GCN Hazard Recognizer Impls ------------===//
 //
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
@@ -21,6 +20,7 @@
 #include "llvm/ADT/iterator_range.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineInstr.h"
+#include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineOperand.h"
 #include "llvm/CodeGen/ScheduleDAG.h"
 #include "llvm/MC/MCInstrDesc.h"
@@ -38,6 +38,7 @@ using namespace llvm;
 //===----------------------------------------------------------------------===//
 
 GCNHazardRecognizer::GCNHazardRecognizer(const MachineFunction &MF) :
+  IsHazardRecognizerMode(false),
   CurrCycleInstr(nullptr),
   MF(MF),
   ST(MF.getSubtarget<GCNSubtarget>()),
@@ -45,7 +46,8 @@ GCNHazardRecognizer::GCNHazardRecognizer(const MachineFunction &MF) :
   TRI(TII.getRegisterInfo()),
   ClauseUses(TRI.getNumRegUnits()),
   ClauseDefs(TRI.getNumRegUnits()) {
-  MaxLookAhead = 5;
+  MaxLookAhead = MF.getRegInfo().isPhysRegUsed(AMDGPU::AGPR0) ? 18 : 5;
+  TSchedModel.init(&ST);
 }
 
 void GCNHazardRecognizer::EmitInstruction(SUnit *SU) {
@@ -114,6 +116,12 @@ static bool isSendMsgTraceDataOrGDS(const SIInstrInfo &TII,
   }
 }
 
+static bool isPermlane(const MachineInstr &MI) {
+  unsigned Opcode = MI.getOpcode();
+  return Opcode == AMDGPU::V_PERMLANE16_B32 ||
+         Opcode == AMDGPU::V_PERMLANEX16_B32;
+}
+
 static unsigned getHWReg(const SIInstrInfo *TII, const MachineInstr &RegInstr) {
   const MachineOperand *RegOp = TII->getNamedOperand(RegInstr,
                                                      AMDGPU::OpName::simm16);
@@ -123,6 +131,8 @@ static unsigned getHWReg(const SIInstrInfo *TII, const MachineInstr &RegInstr) {
 ScheduleHazardRecognizer::HazardType
 GCNHazardRecognizer::getHazardType(SUnit *SU, int Stalls) {
   MachineInstr *MI = SU->getInstr();
+  if (MI->isBundle())
+   return NoHazard;
 
   if (SIInstrInfo::isSMRD(*MI) && checkSMRDHazards(MI) > 0)
     return NoopHazard;
@@ -133,6 +143,15 @@ GCNHazardRecognizer::getHazardType(SUnit *SU, int Stalls) {
       && checkVMEMHazards(MI) > 0)
     return NoopHazard;
 
+  if (ST.hasNSAtoVMEMBug() && checkNSAtoVMEMHazard(MI) > 0)
+    return NoopHazard;
+
+  if (checkFPAtomicToDenormModeHazard(MI) > 0)
+    return NoopHazard;
+
+  if (ST.hasNoDataDepHazard())
+    return NoHazard;
+
   if (SIInstrInfo::isVALU(*MI) && checkVALUHazards(MI) > 0)
     return NoopHazard;
 
@@ -163,6 +182,12 @@ GCNHazardRecognizer::getHazardType(SUnit *SU, int Stalls) {
       checkReadM0Hazards(MI) > 0)
     return NoopHazard;
 
+  if (SIInstrInfo::isMAI(*MI) && checkMAIHazards(MI) > 0)
+    return NoopHazard;
+
+  if ((MI->mayLoad() || MI->mayStore()) && checkMAILdStHazards(MI) > 0)
+    return NoopHazard;
+
   if (MI->isInlineAsm() && checkInlineAsmHazards(MI) > 0)
     return NoopHazard;
 
@@ -172,22 +197,74 @@ GCNHazardRecognizer::getHazardType(SUnit *SU, int Stalls) {
   return NoHazard;
 }
 
+static void insertNoopInBundle(MachineInstr *MI, const SIInstrInfo &TII) {
+  BuildMI(*MI->getParent(), MI, MI->getDebugLoc(), TII.get(AMDGPU::S_NOP))
+      .addImm(0);
+}
+
+void GCNHazardRecognizer::processBundle() {
+  MachineBasicBlock::instr_iterator MI = std::next(CurrCycleInstr->getIterator());
+  MachineBasicBlock::instr_iterator E = CurrCycleInstr->getParent()->instr_end();
+  // Check bundled MachineInstr's for hazards.
+  for (; MI != E && MI->isInsideBundle(); ++MI) {
+    CurrCycleInstr = &*MI;
+    unsigned WaitStates = PreEmitNoopsCommon(CurrCycleInstr);
+
+    if (IsHazardRecognizerMode)
+      fixHazards(CurrCycleInstr);
+
+    for (unsigned i = 0; i < WaitStates; ++i)
+      insertNoopInBundle(CurrCycleInstr, TII);
+
+    // It’s unnecessary to track more than MaxLookAhead instructions. Since we
+    // include the bundled MI directly after, only add a maximum of
+    // (MaxLookAhead - 1) noops to EmittedInstrs.
+    for (unsigned i = 0, e = std::min(WaitStates, MaxLookAhead - 1); i < e; ++i)
+      EmittedInstrs.push_front(nullptr);
+
+    EmittedInstrs.push_front(CurrCycleInstr);
+    EmittedInstrs.resize(MaxLookAhead);
+  }
+  CurrCycleInstr = nullptr;
+}
+
 unsigned GCNHazardRecognizer::PreEmitNoops(SUnit *SU) {
-  return PreEmitNoops(SU->getInstr());
+  IsHazardRecognizerMode = false;
+  return PreEmitNoopsCommon(SU->getInstr());
 }
 
 unsigned GCNHazardRecognizer::PreEmitNoops(MachineInstr *MI) {
+  IsHazardRecognizerMode = true;
+  CurrCycleInstr = MI;
+  unsigned W = PreEmitNoopsCommon(MI);
+  fixHazards(MI);
+  CurrCycleInstr = nullptr;
+  return W;
+}
+
+unsigned GCNHazardRecognizer::PreEmitNoopsCommon(MachineInstr *MI) {
+  if (MI->isBundle())
+    return 0;
+
   int WaitStates = std::max(0, checkAnyInstHazards(MI));
 
   if (SIInstrInfo::isSMRD(*MI))
     return std::max(WaitStates, checkSMRDHazards(MI));
 
-  if (SIInstrInfo::isVALU(*MI))
-    WaitStates = std::max(WaitStates, checkVALUHazards(MI));
-
   if (SIInstrInfo::isVMEM(*MI) || SIInstrInfo::isFLAT(*MI))
     WaitStates = std::max(WaitStates, checkVMEMHazards(MI));
 
+  if (ST.hasNSAtoVMEMBug())
+    WaitStates = std::max(WaitStates, checkNSAtoVMEMHazard(MI));
+
+  WaitStates = std::max(WaitStates, checkFPAtomicToDenormModeHazard(MI));
+
+  if (ST.hasNoDataDepHazard())
+    return WaitStates;
+
+  if (SIInstrInfo::isVALU(*MI))
+    WaitStates = std::max(WaitStates, checkVALUHazards(MI));
+
   if (SIInstrInfo::isDPP(*MI))
     WaitStates = std::max(WaitStates, checkDPPHazards(MI));
 
@@ -216,6 +293,12 @@ unsigned GCNHazardRecognizer::PreEmitNoops(MachineInstr *MI) {
   if (ST.hasReadM0SendMsgHazard() && isSendMsgTraceDataOrGDS(TII, *MI))
     return std::max(WaitStates, checkReadM0Hazards(MI));
 
+  if (SIInstrInfo::isMAI(*MI))
+    return std::max(WaitStates, checkMAIHazards(MI));
+
+  if (MI->mayLoad() || MI->mayStore())
+    return std::max(WaitStates, checkMAILdStHazards(MI));
+
   return WaitStates;
 }
 
@@ -232,10 +315,14 @@ void GCNHazardRecognizer::AdvanceCycle() {
   // Do not track non-instructions which do not affect the wait states.
   // If included, these instructions can lead to buffer overflow such that
   // detectable hazards are missed.
-  if (CurrCycleInstr->getOpcode() == AMDGPU::IMPLICIT_DEF)
+  if (CurrCycleInstr->isImplicitDef() || CurrCycleInstr->isDebugInstr() ||
+      CurrCycleInstr->isKill())
     return;
-  else if (CurrCycleInstr->isDebugInstr())
+
+  if (CurrCycleInstr->isBundle()) {
+    processBundle();
     return;
+  }
 
   unsigned NumWaitStates = TII.getNumWaitStates(*CurrCycleInstr);
 
@@ -266,41 +353,112 @@ void GCNHazardRecognizer::RecedeCycle() {
 // Helper Functions
 //===----------------------------------------------------------------------===//
 
-int GCNHazardRecognizer::getWaitStatesSince(
-    function_ref<bool(MachineInstr *)> IsHazard) {
+typedef function_ref<bool(MachineInstr *, int WaitStates)> IsExpiredFn;
+
+// Returns a minimum wait states since \p I walking all predecessors.
+// Only scans until \p IsExpired does not return true.
+// Can only be run in a hazard recognizer mode.
+static int getWaitStatesSince(GCNHazardRecognizer::IsHazardFn IsHazard,
+                              MachineBasicBlock *MBB,
+                              MachineBasicBlock::reverse_instr_iterator I,
+                              int WaitStates,
+                              IsExpiredFn IsExpired,
+                              DenseSet<const MachineBasicBlock *> &Visited) {
+  for (auto E = MBB->instr_rend(); I != E; ++I) {
+    // Don't add WaitStates for parent BUNDLE instructions.
+    if (I->isBundle())
+      continue;
+
+    if (IsHazard(&*I))
+      return WaitStates;
+
+    if (I->isInlineAsm() || I->isImplicitDef() || I->isDebugInstr())
+      continue;
+
+    WaitStates += SIInstrInfo::getNumWaitStates(*I);
+
+    if (IsExpired(&*I, WaitStates))
+      return std::numeric_limits<int>::max();
+  }
+
+  int MinWaitStates = WaitStates;
+  bool Found = false;
+  for (MachineBasicBlock *Pred : MBB->predecessors()) {
+    if (!Visited.insert(Pred).second)
+      continue;
+
+    int W = getWaitStatesSince(IsHazard, Pred, Pred->instr_rbegin(),
+                               WaitStates, IsExpired, Visited);
+
+    if (W == std::numeric_limits<int>::max())
+      continue;
+
+    MinWaitStates = Found ? std::min(MinWaitStates, W) : W;
+    if (IsExpired(nullptr, MinWaitStates))
+      return MinWaitStates;
+
+    Found = true;
+  }
+
+  if (Found)
+    return MinWaitStates;
+
+  return std::numeric_limits<int>::max();
+}
+
+static int getWaitStatesSince(GCNHazardRecognizer::IsHazardFn IsHazard,
+                              MachineInstr *MI,
+                              IsExpiredFn IsExpired) {
+  DenseSet<const MachineBasicBlock *> Visited;
+  return getWaitStatesSince(IsHazard, MI->getParent(),
+                            std::next(MI->getReverseIterator()),
+                            0, IsExpired, Visited);
+}
+
+int GCNHazardRecognizer::getWaitStatesSince(IsHazardFn IsHazard, int Limit) {
+  if (IsHazardRecognizerMode) {
+    auto IsExpiredFn = [Limit] (MachineInstr *, int WaitStates) {
+      return WaitStates >= Limit;
+    };
+    return ::getWaitStatesSince(IsHazard, CurrCycleInstr, IsExpiredFn);
+  }
+
   int WaitStates = 0;
   for (MachineInstr *MI : EmittedInstrs) {
     if (MI) {
       if (IsHazard(MI))
         return WaitStates;
 
-      unsigned Opcode = MI->getOpcode();
-      if (Opcode == AMDGPU::INLINEASM)
+      if (MI->isInlineAsm())
         continue;
     }
     ++WaitStates;
+
+    if (WaitStates >= Limit)
+      break;
   }
   return std::numeric_limits<int>::max();
 }
 
-int GCNHazardRecognizer::getWaitStatesSinceDef(
-    unsigned Reg, function_ref<bool(MachineInstr *)> IsHazardDef) {
+int GCNHazardRecognizer::getWaitStatesSinceDef(unsigned Reg,
+                                               IsHazardFn IsHazardDef,
+                                               int Limit) {
   const SIRegisterInfo *TRI = ST.getRegisterInfo();
 
   auto IsHazardFn = [IsHazardDef, TRI, Reg] (MachineInstr *MI) {
     return IsHazardDef(MI) && MI->modifiesRegister(Reg, TRI);
   };
 
-  return getWaitStatesSince(IsHazardFn);
+  return getWaitStatesSince(IsHazardFn, Limit);
 }
 
-int GCNHazardRecognizer::getWaitStatesSinceSetReg(
-    function_ref<bool(MachineInstr *)> IsHazard) {
+int GCNHazardRecognizer::getWaitStatesSinceSetReg(IsHazardFn IsHazard,
+                                                  int Limit) {
   auto IsHazardFn = [IsHazard] (MachineInstr *MI) {
     return isSSetReg(MI->getOpcode()) && IsHazard(MI);
   };
 
-  return getWaitStatesSince(IsHazardFn);
+  return getWaitStatesSince(IsHazardFn, Limit);
 }
 
 //===----------------------------------------------------------------------===//
@@ -342,9 +500,9 @@ int GCNHazardRecognizer::checkSoftClauseHazards(MachineInstr *MEM) {
   // instructions in this group may return out of order and/or may be
   // replayed (i.e. the same instruction issued more than once).
   //
-  // In order to handle these situations correctly we need to make sure
-  // that when a clause has more than one instruction, no instruction in the
-  // clause writes to a register that is read another instruction in the clause
+  // In order to handle these situations correctly we need to make sure that
+  // when a clause has more than one instruction, no instruction in the clause
+  // writes to a register that is read by another instruction in the clause
   // (including itself). If we encounter this situaion, we need to break the
   // clause by inserting a non SMEM instruction.
 
@@ -377,13 +535,12 @@ int GCNHazardRecognizer::checkSoftClauseHazards(MachineInstr *MEM) {
 }
 
 int GCNHazardRecognizer::checkSMRDHazards(MachineInstr *SMRD) {
-  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   int WaitStatesNeeded = 0;
 
   WaitStatesNeeded = checkSoftClauseHazards(SMRD);
 
   // This SMRD hazard only affects SI.
-  if (ST.getGeneration() != AMDGPUSubtarget::SOUTHERN_ISLANDS)
+  if (!ST.hasSMRDReadVALUDefHazard())
     return WaitStatesNeeded;
 
   // A read of an SGPR by SMRD instruction requires 4 wait states when the
@@ -398,7 +555,8 @@ int GCNHazardRecognizer::checkSMRDHazards(MachineInstr *SMRD) {
     if (!Use.isReg())
       continue;
     int WaitStatesNeededForUse =
-        SmrdSgprWaitStates - getWaitStatesSinceDef(Use.getReg(), IsHazardDefFn);
+        SmrdSgprWaitStates - getWaitStatesSinceDef(Use.getReg(), IsHazardDefFn,
+                                                   SmrdSgprWaitStates);
     WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
 
     // This fixes what appears to be undocumented hardware behavior in SI where
@@ -411,7 +569,8 @@ int GCNHazardRecognizer::checkSMRDHazards(MachineInstr *SMRD) {
     if (IsBufferSMRD) {
       int WaitStatesNeededForUse =
         SmrdSgprWaitStates - getWaitStatesSinceDef(Use.getReg(),
-                                                   IsBufferHazardDefFn);
+                                                   IsBufferHazardDefFn,
+                                                   SmrdSgprWaitStates);
       WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
     }
   }
@@ -420,7 +579,7 @@ int GCNHazardRecognizer::checkSMRDHazards(MachineInstr *SMRD) {
 }
 
 int GCNHazardRecognizer::checkVMEMHazards(MachineInstr* VMEM) {
-  if (ST.getGeneration() < AMDGPUSubtarget::VOLCANIC_ISLANDS)
+  if (!ST.hasVMEMReadSGPRVALUDefHazard())
     return 0;
 
   int WaitStatesNeeded = checkSoftClauseHazards(VMEM);
@@ -429,13 +588,13 @@ int GCNHazardRecognizer::checkVMEMHazards(MachineInstr* VMEM) {
   // SGPR was written by a VALU Instruction.
   const int VmemSgprWaitStates = 5;
   auto IsHazardDefFn = [this] (MachineInstr *MI) { return TII.isVALU(*MI); };
-
   for (const MachineOperand &Use : VMEM->uses()) {
     if (!Use.isReg() || TRI.isVGPR(MF.getRegInfo(), Use.getReg()))
       continue;
 
     int WaitStatesNeededForUse =
-        VmemSgprWaitStates - getWaitStatesSinceDef(Use.getReg(), IsHazardDefFn);
+        VmemSgprWaitStates - getWaitStatesSinceDef(Use.getReg(), IsHazardDefFn,
+                                                   VmemSgprWaitStates);
     WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
   }
   return WaitStatesNeeded;
@@ -455,13 +614,16 @@ int GCNHazardRecognizer::checkDPPHazards(MachineInstr *DPP) {
     if (!Use.isReg() || !TRI->isVGPR(MF.getRegInfo(), Use.getReg()))
       continue;
     int WaitStatesNeededForUse =
-        DppVgprWaitStates - getWaitStatesSinceDef(Use.getReg());
+        DppVgprWaitStates - getWaitStatesSinceDef(Use.getReg(),
+                              [](MachineInstr *) { return true; },
+                              DppVgprWaitStates);
     WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
   }
 
   WaitStatesNeeded = std::max(
       WaitStatesNeeded,
-      DppExecWaitStates - getWaitStatesSinceDef(AMDGPU::EXEC, IsHazardDefFn));
+      DppExecWaitStates - getWaitStatesSinceDef(AMDGPU::EXEC, IsHazardDefFn,
+                                                DppExecWaitStates));
 
   return WaitStatesNeeded;
 }
@@ -473,7 +635,8 @@ int GCNHazardRecognizer::checkDivFMasHazards(MachineInstr *DivFMas) {
   // instruction.
   const int DivFMasWaitStates = 4;
   auto IsHazardDefFn = [TII] (MachineInstr *MI) { return TII->isVALU(*MI); };
-  int WaitStatesNeeded = getWaitStatesSinceDef(AMDGPU::VCC, IsHazardDefFn);
+  int WaitStatesNeeded = getWaitStatesSinceDef(AMDGPU::VCC, IsHazardDefFn,
+                                               DivFMasWaitStates);
 
   return DivFMasWaitStates - WaitStatesNeeded;
 }
@@ -486,7 +649,7 @@ int GCNHazardRecognizer::checkGetRegHazards(MachineInstr *GetRegInstr) {
   auto IsHazardFn = [TII, GetRegHWReg] (MachineInstr *MI) {
     return GetRegHWReg == getHWReg(TII, *MI);
   };
-  int WaitStatesNeeded = getWaitStatesSinceSetReg(IsHazardFn);
+  int WaitStatesNeeded = getWaitStatesSinceSetReg(IsHazardFn, GetRegWaitStates);
 
   return GetRegWaitStates - WaitStatesNeeded;
 }
@@ -495,12 +658,11 @@ int GCNHazardRecognizer::checkSetRegHazards(MachineInstr *SetRegInstr) {
   const SIInstrInfo *TII = ST.getInstrInfo();
   unsigned HWReg = getHWReg(TII, *SetRegInstr);
 
-  const int SetRegWaitStates =
-      ST.getGeneration() <= AMDGPUSubtarget::SEA_ISLANDS ? 1 : 2;
+  const int SetRegWaitStates = ST.getSetRegWaitStates();
   auto IsHazardFn = [TII, HWReg] (MachineInstr *MI) {
     return HWReg == getHWReg(TII, *MI);
   };
-  int WaitStatesNeeded = getWaitStatesSinceSetReg(IsHazardFn);
+  int WaitStatesNeeded = getWaitStatesSinceSetReg(IsHazardFn, SetRegWaitStates);
   return SetRegWaitStates - WaitStatesNeeded;
 }
 
@@ -571,7 +733,7 @@ int GCNHazardRecognizer::checkVALUHazardsHelper(const MachineOperand &Def,
     TRI->regsOverlap(MI->getOperand(DataIdx).getReg(), Reg);
   };
   int WaitStatesNeededForDef =
-    VALUWaitStates - getWaitStatesSince(IsHazardFn);
+    VALUWaitStates - getWaitStatesSince(IsHazardFn, VALUWaitStates);
   WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForDef);
 
   return WaitStatesNeeded;
@@ -636,12 +798,13 @@ int GCNHazardRecognizer::checkRWLaneHazards(MachineInstr *RWLane) {
   };
 
   const int RWLaneWaitStates = 4;
-  int WaitStatesSince = getWaitStatesSinceDef(LaneSelectReg, IsHazardFn);
+  int WaitStatesSince = getWaitStatesSinceDef(LaneSelectReg, IsHazardFn,
+                                              RWLaneWaitStates);
   return RWLaneWaitStates - WaitStatesSince;
 }
 
 int GCNHazardRecognizer::checkRFEHazards(MachineInstr *RFE) {
-  if (ST.getGeneration() < AMDGPUSubtarget::VOLCANIC_ISLANDS)
+  if (!ST.hasRFEHazards())
     return 0;
 
   const SIInstrInfo *TII = ST.getInstrInfo();
@@ -651,7 +814,7 @@ int GCNHazardRecognizer::checkRFEHazards(MachineInstr *RFE) {
   auto IsHazardFn = [TII] (MachineInstr *MI) {
     return getHWReg(TII, *MI) == AMDGPU::Hwreg::ID_TRAPSTS;
   };
-  int WaitStatesNeeded = getWaitStatesSinceSetReg(IsHazardFn);
+  int WaitStatesNeeded = getWaitStatesSinceSetReg(IsHazardFn, RFEWaitStates);
   return RFEWaitStates - WaitStatesNeeded;
 }
 
@@ -675,7 +838,8 @@ int GCNHazardRecognizer::checkAnyInstHazards(MachineInstr *MI) {
       return MI->getOpcode() == AMDGPU::S_MOV_FED_B32;
     };
     int WaitStatesNeededForUse =
-        MovFedWaitStates - getWaitStatesSinceDef(Use.getReg(), IsHazardFn);
+        MovFedWaitStates - getWaitStatesSinceDef(Use.getReg(), IsHazardFn,
+                                                 MovFedWaitStates);
     WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
   }
 
@@ -688,5 +852,557 @@ int GCNHazardRecognizer::checkReadM0Hazards(MachineInstr *MI) {
   auto IsHazardFn = [TII] (MachineInstr *MI) {
     return TII->isSALU(*MI);
   };
-  return SMovRelWaitStates - getWaitStatesSinceDef(AMDGPU::M0, IsHazardFn);
+  return SMovRelWaitStates - getWaitStatesSinceDef(AMDGPU::M0, IsHazardFn,
+                                                   SMovRelWaitStates);
+}
+
+void GCNHazardRecognizer::fixHazards(MachineInstr *MI) {
+  fixVMEMtoScalarWriteHazards(MI);
+  fixVcmpxPermlaneHazards(MI);
+  fixSMEMtoVectorWriteHazards(MI);
+  fixVcmpxExecWARHazard(MI);
+  fixLdsBranchVmemWARHazard(MI);
+}
+
+bool GCNHazardRecognizer::fixVcmpxPermlaneHazards(MachineInstr *MI) {
+  if (!ST.hasVcmpxPermlaneHazard() || !isPermlane(*MI))
+    return false;
+
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  auto IsHazardFn = [TII] (MachineInstr *MI) {
+    return TII->isVOPC(*MI);
+  };
+
+  auto IsExpiredFn = [] (MachineInstr *MI, int) {
+    if (!MI)
+      return false;
+    unsigned Opc = MI->getOpcode();
+    return SIInstrInfo::isVALU(*MI) &&
+           Opc != AMDGPU::V_NOP_e32 &&
+           Opc != AMDGPU::V_NOP_e64 &&
+           Opc != AMDGPU::V_NOP_sdwa;
+  };
+
+  if (::getWaitStatesSince(IsHazardFn, MI, IsExpiredFn) ==
+      std::numeric_limits<int>::max())
+    return false;
+
+  // V_NOP will be discarded by SQ.
+  // Use V_MOB_B32 v?, v?. Register must be alive so use src0 of V_PERMLANE*
+  // which is always a VGPR and available.
+  auto *Src0 = TII->getNamedOperand(*MI, AMDGPU::OpName::src0);
+  unsigned Reg = Src0->getReg();
+  bool IsUndef = Src0->isUndef();
+  BuildMI(*MI->getParent(), MI, MI->getDebugLoc(),
+          TII->get(AMDGPU::V_MOV_B32_e32))
+    .addReg(Reg, RegState::Define | (IsUndef ? RegState::Dead : 0))
+    .addReg(Reg, IsUndef ? RegState::Undef : RegState::Kill);
+
+  return true;
+}
+
+bool GCNHazardRecognizer::fixVMEMtoScalarWriteHazards(MachineInstr *MI) {
+  if (!ST.hasVMEMtoScalarWriteHazard())
+    return false;
+
+  if (!SIInstrInfo::isSALU(*MI) && !SIInstrInfo::isSMRD(*MI))
+    return false;
+
+  if (MI->getNumDefs() == 0)
+    return false;
+
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
+
+  auto IsHazardFn = [TRI, MI] (MachineInstr *I) {
+    if (!SIInstrInfo::isVMEM(*I) && !SIInstrInfo::isDS(*I) &&
+        !SIInstrInfo::isFLAT(*I))
+      return false;
+
+    for (const MachineOperand &Def : MI->defs()) {
+      MachineOperand *Op = I->findRegisterUseOperand(Def.getReg(), false, TRI);
+      if (!Op)
+        continue;
+      return true;
+    }
+    return false;
+  };
+
+  auto IsExpiredFn = [] (MachineInstr *MI, int) {
+    return MI && (SIInstrInfo::isVALU(*MI) ||
+                  (MI->getOpcode() == AMDGPU::S_WAITCNT &&
+                   !MI->getOperand(0).getImm()));
+  };
+
+  if (::getWaitStatesSince(IsHazardFn, MI, IsExpiredFn) ==
+      std::numeric_limits<int>::max())
+    return false;
+
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  BuildMI(*MI->getParent(), MI, MI->getDebugLoc(), TII->get(AMDGPU::V_NOP_e32));
+  return true;
+}
+
+bool GCNHazardRecognizer::fixSMEMtoVectorWriteHazards(MachineInstr *MI) {
+  if (!ST.hasSMEMtoVectorWriteHazard())
+    return false;
+
+  if (!SIInstrInfo::isVALU(*MI))
+    return false;
+
+  unsigned SDSTName;
+  switch (MI->getOpcode()) {
+  case AMDGPU::V_READLANE_B32:
+  case AMDGPU::V_READFIRSTLANE_B32:
+    SDSTName = AMDGPU::OpName::vdst;
+    break;
+  default:
+    SDSTName = AMDGPU::OpName::sdst;
+    break;
+  }
+
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
+  const AMDGPU::IsaVersion IV = AMDGPU::getIsaVersion(ST.getCPU());
+  const MachineOperand *SDST = TII->getNamedOperand(*MI, SDSTName);
+  if (!SDST) {
+    for (const auto &MO : MI->implicit_operands()) {
+      if (MO.isDef() && TRI->isSGPRClass(TRI->getPhysRegClass(MO.getReg()))) {
+        SDST = &MO;
+        break;
+      }
+    }
+  }
+
+  if (!SDST)
+    return false;
+
+  const unsigned SDSTReg = SDST->getReg();
+  auto IsHazardFn = [SDSTReg, TRI] (MachineInstr *I) {
+    return SIInstrInfo::isSMRD(*I) && I->readsRegister(SDSTReg, TRI);
+  };
+
+  auto IsExpiredFn = [TII, IV] (MachineInstr *MI, int) {
+    if (MI) {
+      if (TII->isSALU(*MI)) {
+        switch (MI->getOpcode()) {
+        case AMDGPU::S_SETVSKIP:
+        case AMDGPU::S_VERSION:
+        case AMDGPU::S_WAITCNT_VSCNT:
+        case AMDGPU::S_WAITCNT_VMCNT:
+        case AMDGPU::S_WAITCNT_EXPCNT:
+          // These instructions cannot not mitigate the hazard.
+          return false;
+        case AMDGPU::S_WAITCNT_LGKMCNT:
+          // Reducing lgkmcnt count to 0 always mitigates the hazard.
+          return (MI->getOperand(1).getImm() == 0) &&
+                 (MI->getOperand(0).getReg() == AMDGPU::SGPR_NULL);
+        case AMDGPU::S_WAITCNT: {
+          const int64_t Imm = MI->getOperand(0).getImm();
+          AMDGPU::Waitcnt Decoded = AMDGPU::decodeWaitcnt(IV, Imm);
+          return (Decoded.LgkmCnt == 0);
+        }
+        default:
+          // SOPP instructions cannot mitigate the hazard.
+          if (TII->isSOPP(*MI))
+            return false;
+          // At this point the SALU can be assumed to mitigate the hazard
+          // because either:
+          // (a) it is independent of the at risk SMEM (breaking chain),
+          // or
+          // (b) it is dependent on the SMEM, in which case an appropriate
+          //     s_waitcnt lgkmcnt _must_ exist between it and the at risk
+          //     SMEM instruction.
+          return true;
+        }
+      }
+    }
+    return false;
+  };
+
+  if (::getWaitStatesSince(IsHazardFn, MI, IsExpiredFn) ==
+      std::numeric_limits<int>::max())
+    return false;
+
+  BuildMI(*MI->getParent(), MI, MI->getDebugLoc(),
+          TII->get(AMDGPU::S_MOV_B32), AMDGPU::SGPR_NULL)
+      .addImm(0);
+  return true;
+}
+
+bool GCNHazardRecognizer::fixVcmpxExecWARHazard(MachineInstr *MI) {
+  if (!ST.hasVcmpxExecWARHazard() || !SIInstrInfo::isVALU(*MI))
+    return false;
+
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
+  if (!MI->modifiesRegister(AMDGPU::EXEC, TRI))
+    return false;
+
+  auto IsHazardFn = [TRI] (MachineInstr *I) {
+    if (SIInstrInfo::isVALU(*I))
+      return false;
+    return I->readsRegister(AMDGPU::EXEC, TRI);
+  };
+
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  auto IsExpiredFn = [TII, TRI] (MachineInstr *MI, int) {
+    if (!MI)
+      return false;
+    if (SIInstrInfo::isVALU(*MI)) {
+      if (TII->getNamedOperand(*MI, AMDGPU::OpName::sdst))
+        return true;
+      for (auto MO : MI->implicit_operands())
+        if (MO.isDef() && TRI->isSGPRClass(TRI->getPhysRegClass(MO.getReg())))
+          return true;
+    }
+    if (MI->getOpcode() == AMDGPU::S_WAITCNT_DEPCTR &&
+        (MI->getOperand(0).getImm() & 0xfffe) == 0xfffe)
+      return true;
+    return false;
+  };
+
+  if (::getWaitStatesSince(IsHazardFn, MI, IsExpiredFn) ==
+      std::numeric_limits<int>::max())
+    return false;
+
+  BuildMI(*MI->getParent(), MI, MI->getDebugLoc(),
+          TII->get(AMDGPU::S_WAITCNT_DEPCTR))
+    .addImm(0xfffe);
+  return true;
+}
+
+bool GCNHazardRecognizer::fixLdsBranchVmemWARHazard(MachineInstr *MI) {
+  if (!ST.hasLdsBranchVmemWARHazard())
+    return false;
+
+  auto IsHazardInst = [] (const MachineInstr *MI) {
+    if (SIInstrInfo::isDS(*MI))
+      return 1;
+    if (SIInstrInfo::isVMEM(*MI) || SIInstrInfo::isSegmentSpecificFLAT(*MI))
+      return 2;
+    return 0;
+  };
+
+  auto InstType = IsHazardInst(MI);
+  if (!InstType)
+    return false;
+
+  auto IsExpiredFn = [&IsHazardInst] (MachineInstr *I, int) {
+    return I && (IsHazardInst(I) ||
+                 (I->getOpcode() == AMDGPU::S_WAITCNT_VSCNT &&
+                  I->getOperand(0).getReg() == AMDGPU::SGPR_NULL &&
+                  !I->getOperand(1).getImm()));
+  };
+
+  auto IsHazardFn = [InstType, &IsHazardInst] (MachineInstr *I) {
+    if (!I->isBranch())
+      return false;
+
+    auto IsHazardFn = [InstType, IsHazardInst] (MachineInstr *I) {
+      auto InstType2 = IsHazardInst(I);
+      return InstType2 && InstType != InstType2;
+    };
+
+    auto IsExpiredFn = [InstType, &IsHazardInst] (MachineInstr *I, int) {
+      if (!I)
+        return false;
+
+      auto InstType2 = IsHazardInst(I);
+      if (InstType == InstType2)
+        return true;
+
+      return I->getOpcode() == AMDGPU::S_WAITCNT_VSCNT &&
+             I->getOperand(0).getReg() == AMDGPU::SGPR_NULL &&
+             !I->getOperand(1).getImm();
+    };
+
+    return ::getWaitStatesSince(IsHazardFn, I, IsExpiredFn) !=
+           std::numeric_limits<int>::max();
+  };
+
+  if (::getWaitStatesSince(IsHazardFn, MI, IsExpiredFn) ==
+      std::numeric_limits<int>::max())
+    return false;
+
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  BuildMI(*MI->getParent(), MI, MI->getDebugLoc(),
+          TII->get(AMDGPU::S_WAITCNT_VSCNT))
+    .addReg(AMDGPU::SGPR_NULL, RegState::Undef)
+    .addImm(0);
+
+  return true;
+}
+
+int GCNHazardRecognizer::checkNSAtoVMEMHazard(MachineInstr *MI) {
+  int NSAtoVMEMWaitStates = 1;
+
+  if (!ST.hasNSAtoVMEMBug())
+    return 0;
+
+  if (!SIInstrInfo::isMUBUF(*MI) && !SIInstrInfo::isMTBUF(*MI))
+    return 0;
+
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  const auto *Offset = TII->getNamedOperand(*MI, AMDGPU::OpName::offset);
+  if (!Offset || (Offset->getImm() & 6) == 0)
+    return 0;
+
+  auto IsHazardFn = [TII] (MachineInstr *I) {
+    if (!SIInstrInfo::isMIMG(*I))
+      return false;
+    const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(I->getOpcode());
+    return Info->MIMGEncoding == AMDGPU::MIMGEncGfx10NSA &&
+           TII->getInstSizeInBytes(*I) >= 16;
+  };
+
+  return NSAtoVMEMWaitStates - getWaitStatesSince(IsHazardFn, 1);
+}
+
+int GCNHazardRecognizer::checkFPAtomicToDenormModeHazard(MachineInstr *MI) {
+  int FPAtomicToDenormModeWaitStates = 3;
+
+  if (MI->getOpcode() != AMDGPU::S_DENORM_MODE)
+    return 0;
+
+  auto IsHazardFn = [] (MachineInstr *I) {
+    if (!SIInstrInfo::isVMEM(*I) && !SIInstrInfo::isFLAT(*I))
+      return false;
+    return SIInstrInfo::isFPAtomic(*I);
+  };
+
+  auto IsExpiredFn = [] (MachineInstr *MI, int WaitStates) {
+    if (WaitStates >= 3 || SIInstrInfo::isVALU(*MI))
+      return true;
+
+    switch (MI->getOpcode()) {
+    case AMDGPU::S_WAITCNT:
+    case AMDGPU::S_WAITCNT_VSCNT:
+    case AMDGPU::S_WAITCNT_VMCNT:
+    case AMDGPU::S_WAITCNT_EXPCNT:
+    case AMDGPU::S_WAITCNT_LGKMCNT:
+    case AMDGPU::S_WAITCNT_IDLE:
+      return true;
+    default:
+      break;
+    }
+
+    return false;
+  };
+
+
+  return FPAtomicToDenormModeWaitStates -
+         ::getWaitStatesSince(IsHazardFn, MI, IsExpiredFn);
+}
+
+int GCNHazardRecognizer::checkMAIHazards(MachineInstr *MI) {
+  assert(SIInstrInfo::isMAI(*MI));
+
+  int WaitStatesNeeded = 0;
+  unsigned Opc = MI->getOpcode();
+
+  auto IsVALUFn = [] (MachineInstr *MI) {
+    return SIInstrInfo::isVALU(*MI);
+  };
+
+  if (Opc != AMDGPU::V_ACCVGPR_READ_B32) { // MFMA or v_accvgpr_write
+    const int LegacyVALUWritesVGPRWaitStates = 2;
+    const int VALUWritesExecWaitStates = 4;
+    const int MaxWaitStates = 4;
+
+    int WaitStatesNeededForUse = VALUWritesExecWaitStates -
+      getWaitStatesSinceDef(AMDGPU::EXEC, IsVALUFn, MaxWaitStates);
+    WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+
+    if (WaitStatesNeeded < MaxWaitStates) {
+      for (const MachineOperand &Use : MI->explicit_uses()) {
+        const int MaxWaitStates = 2;
+
+        if (!Use.isReg() || !TRI.isVGPR(MF.getRegInfo(), Use.getReg()))
+          continue;
+
+        int WaitStatesNeededForUse = LegacyVALUWritesVGPRWaitStates -
+          getWaitStatesSinceDef(Use.getReg(), IsVALUFn, MaxWaitStates);
+        WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+
+        if (WaitStatesNeeded == MaxWaitStates)
+          break;
+      }
+    }
+  }
+
+  auto IsMFMAFn = [] (MachineInstr *MI) {
+    return SIInstrInfo::isMAI(*MI) &&
+           MI->getOpcode() != AMDGPU::V_ACCVGPR_WRITE_B32 &&
+           MI->getOpcode() != AMDGPU::V_ACCVGPR_READ_B32;
+  };
+
+  for (const MachineOperand &Op : MI->explicit_operands()) {
+    if (!Op.isReg() || !TRI.isAGPR(MF.getRegInfo(), Op.getReg()))
+      continue;
+
+    if (Op.isDef() && Opc != AMDGPU::V_ACCVGPR_WRITE_B32)
+      continue;
+
+    const int MFMAWritesAGPROverlappedSrcABWaitStates = 4;
+    const int MFMAWritesAGPROverlappedSrcCWaitStates = 2;
+    const int MFMA4x4WritesAGPRAccVgprReadWaitStates = 4;
+    const int MFMA16x16WritesAGPRAccVgprReadWaitStates = 10;
+    const int MFMA32x32WritesAGPRAccVgprReadWaitStates = 18;
+    const int MFMA4x4WritesAGPRAccVgprWriteWaitStates = 1;
+    const int MFMA16x16WritesAGPRAccVgprWriteWaitStates = 7;
+    const int MFMA32x32WritesAGPRAccVgprWriteWaitStates = 15;
+    const int MaxWaitStates = 18;
+    unsigned Reg = Op.getReg();
+    unsigned HazardDefLatency = 0;
+
+    auto IsOverlappedMFMAFn = [Reg, &IsMFMAFn, &HazardDefLatency, this]
+                              (MachineInstr *MI) {
+      if (!IsMFMAFn(MI))
+        return false;
+      unsigned DstReg = MI->getOperand(0).getReg();
+      if (DstReg == Reg)
+        return false;
+      HazardDefLatency = std::max(HazardDefLatency,
+                                  TSchedModel.computeInstrLatency(MI));
+      return TRI.regsOverlap(DstReg, Reg);
+    };
+
+    int WaitStatesSinceDef = getWaitStatesSinceDef(Reg, IsOverlappedMFMAFn,
+                                                   MaxWaitStates);
+    int NeedWaitStates = MFMAWritesAGPROverlappedSrcABWaitStates;
+    int SrcCIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::src2);
+    int OpNo = MI->getOperandNo(&Op);
+    if (OpNo == SrcCIdx) {
+      NeedWaitStates = MFMAWritesAGPROverlappedSrcCWaitStates;
+    } else if (Opc == AMDGPU::V_ACCVGPR_READ_B32) {
+      switch (HazardDefLatency) {
+      case 2:  NeedWaitStates = MFMA4x4WritesAGPRAccVgprReadWaitStates;
+               break;
+      case 8:  NeedWaitStates = MFMA16x16WritesAGPRAccVgprReadWaitStates;
+               break;
+      case 16: LLVM_FALLTHROUGH;
+      default: NeedWaitStates = MFMA32x32WritesAGPRAccVgprReadWaitStates;
+               break;
+      }
+    } else if (Opc == AMDGPU::V_ACCVGPR_WRITE_B32) {
+      switch (HazardDefLatency) {
+      case 2:  NeedWaitStates = MFMA4x4WritesAGPRAccVgprWriteWaitStates;
+               break;
+      case 8:  NeedWaitStates = MFMA16x16WritesAGPRAccVgprWriteWaitStates;
+               break;
+      case 16: LLVM_FALLTHROUGH;
+      default: NeedWaitStates = MFMA32x32WritesAGPRAccVgprWriteWaitStates;
+               break;
+      }
+    }
+
+    int WaitStatesNeededForUse = NeedWaitStates - WaitStatesSinceDef;
+    WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+
+    if (WaitStatesNeeded == MaxWaitStates)
+      return WaitStatesNeeded; // Early exit.
+
+    auto IsAccVgprWriteFn = [Reg, this] (MachineInstr *MI) {
+      if (MI->getOpcode() != AMDGPU::V_ACCVGPR_WRITE_B32)
+        return false;
+      unsigned DstReg = MI->getOperand(0).getReg();
+      return TRI.regsOverlap(Reg, DstReg);
+    };
+
+    const int AccVGPRWriteMFMAReadSrcCWaitStates = 1;
+    const int AccVGPRWriteMFMAReadSrcABWaitStates = 3;
+    const int AccVGPRWriteAccVgprReadWaitStates = 3;
+    NeedWaitStates = AccVGPRWriteMFMAReadSrcABWaitStates;
+    if (OpNo == SrcCIdx)
+      NeedWaitStates = AccVGPRWriteMFMAReadSrcCWaitStates;
+    else if (Opc == AMDGPU::V_ACCVGPR_READ_B32)
+      NeedWaitStates = AccVGPRWriteAccVgprReadWaitStates;
+
+    WaitStatesNeededForUse = NeedWaitStates -
+      getWaitStatesSinceDef(Reg, IsAccVgprWriteFn, MaxWaitStates);
+    WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+
+    if (WaitStatesNeeded == MaxWaitStates)
+      return WaitStatesNeeded; // Early exit.
+  }
+
+  if (Opc == AMDGPU::V_ACCVGPR_WRITE_B32) {
+    const int MFMA4x4ReadSrcCAccVgprWriteWaitStates = 0;
+    const int MFMA16x16ReadSrcCAccVgprWriteWaitStates = 5;
+    const int MFMA32x32ReadSrcCAccVgprWriteWaitStates = 13;
+    const int MaxWaitStates = 13;
+    unsigned DstReg = MI->getOperand(0).getReg();
+    unsigned HazardDefLatency = 0;
+
+    auto IsSrcCMFMAFn = [DstReg, &IsMFMAFn, &HazardDefLatency, this]
+                         (MachineInstr *MI) {
+      if (!IsMFMAFn(MI))
+        return false;
+      unsigned Reg = TII.getNamedOperand(*MI, AMDGPU::OpName::src2)->getReg();
+      HazardDefLatency = std::max(HazardDefLatency,
+                                  TSchedModel.computeInstrLatency(MI));
+      return TRI.regsOverlap(Reg, DstReg);
+    };
+
+    int WaitStatesSince = getWaitStatesSince(IsSrcCMFMAFn, MaxWaitStates);
+    int NeedWaitStates;
+    switch (HazardDefLatency) {
+    case 2:  NeedWaitStates = MFMA4x4ReadSrcCAccVgprWriteWaitStates;
+             break;
+    case 8:  NeedWaitStates = MFMA16x16ReadSrcCAccVgprWriteWaitStates;
+             break;
+    case 16: LLVM_FALLTHROUGH;
+    default: NeedWaitStates = MFMA32x32ReadSrcCAccVgprWriteWaitStates;
+             break;
+    }
+
+    int WaitStatesNeededForUse = NeedWaitStates - WaitStatesSince;
+    WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+  }
+
+  return WaitStatesNeeded;
+}
+
+int GCNHazardRecognizer::checkMAILdStHazards(MachineInstr *MI) {
+  if (!ST.hasMAIInsts())
+    return 0;
+
+  int WaitStatesNeeded = 0;
+
+  auto IsAccVgprReadFn = [] (MachineInstr *MI) {
+    return MI->getOpcode() == AMDGPU::V_ACCVGPR_READ_B32;
+  };
+
+  for (const MachineOperand &Op : MI->explicit_uses()) {
+    if (!Op.isReg() || !TRI.isVGPR(MF.getRegInfo(), Op.getReg()))
+      continue;
+
+    unsigned Reg = Op.getReg();
+
+    const int AccVgprReadLdStWaitStates = 2;
+    const int VALUWriteAccVgprReadLdStDepVALUWaitStates = 1;
+    const int MaxWaitStates = 2;
+
+    int WaitStatesNeededForUse = AccVgprReadLdStWaitStates -
+      getWaitStatesSinceDef(Reg, IsAccVgprReadFn, MaxWaitStates);
+    WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+
+    if (WaitStatesNeeded == MaxWaitStates)
+      return WaitStatesNeeded; // Early exit.
+
+    auto IsVALUAccVgprReadCheckFn = [Reg, this] (MachineInstr *MI) {
+      if (MI->getOpcode() != AMDGPU::V_ACCVGPR_READ_B32)
+        return false;
+      auto IsVALUFn = [] (MachineInstr *MI) {
+        return SIInstrInfo::isVALU(*MI) && !SIInstrInfo::isMAI(*MI);
+      };
+      return getWaitStatesSinceDef(Reg, IsVALUFn, 2 /*MaxWaitStates*/) <
+             std::numeric_limits<int>::max();
+    };
+
+    WaitStatesNeededForUse = VALUWriteAccVgprReadLdStDepVALUWaitStates -
+      getWaitStatesSince(IsVALUAccVgprReadCheckFn, MaxWaitStates);
+    WaitStatesNeeded = std::max(WaitStatesNeeded, WaitStatesNeededForUse);
+  }
+
+  return WaitStatesNeeded;
 }