Cemu/src/Cafe/HW/Espresso/Recompiler/IML/IMLOptimizer.cpp

941 lines
37 KiB
C++

#include "Cafe/HW/Espresso/Interpreter/PPCInterpreterInternal.h"
#include "Cafe/HW/Espresso/Recompiler/IML/IML.h"
#include "Cafe/HW/Espresso/Recompiler/IML/IMLInstruction.h"
#include "../PPCRecompiler.h"
#include "../PPCRecompilerIml.h"
#include "../BackendX64/BackendX64.h"
//bool _RegExceedsFPRSpace(IMLReg r)
//{
// if (r.IsInvalid())
// return false;
// if (r.GetRegID() >= PPC_X64_FPR_USABLE_REGISTERS)
// return true;
// return false;
//}
IMLReg _FPRRegFromID(IMLRegID regId)
{
return IMLReg(IMLRegFormat::F64, IMLRegFormat::F64, 0, regId);
}
//bool PPCRecompiler_reduceNumberOfFPRRegisters(ppcImlGenContext_t* ppcImlGenContext)
//{
// // only xmm0 to xmm14 may be used, xmm15 is reserved
// // this method will reduce the number of fpr registers used
// // inefficient algorithm for optimizing away excess registers
// // we simply load, use and store excess registers into other unused registers when we need to
// // first we remove all name load and store instructions that involve out-of-bounds registers
// for (IMLSegment* segIt : ppcImlGenContext->segmentList2)
// {
// size_t imlIndex = 0;
// while( imlIndex < segIt->imlList.size() )
// {
// IMLInstruction& imlInstructionItr = segIt->imlList[imlIndex];
// if( imlInstructionItr.type == PPCREC_IML_TYPE_FPR_R_NAME || imlInstructionItr.type == PPCREC_IML_TYPE_FPR_NAME_R )
// {
// if(_RegExceedsFPRSpace(imlInstructionItr.op_r_name.regR))
// {
// imlInstructionItr.make_no_op();
// }
// }
// imlIndex++;
// }
// }
// // replace registers
// for (IMLSegment* segIt : ppcImlGenContext->segmentList2)
// {
// size_t imlIndex = 0;
// while( imlIndex < segIt->imlList.size() )
// {
// IMLUsedRegisters registersUsed;
// while( true )
// {
// segIt->imlList[imlIndex].CheckRegisterUsage(&registersUsed);
// if(_RegExceedsFPRSpace(registersUsed.readFPR1) || _RegExceedsFPRSpace(registersUsed.readFPR2) || _RegExceedsFPRSpace(registersUsed.readFPR3) || _RegExceedsFPRSpace(registersUsed.readFPR4) || _RegExceedsFPRSpace(registersUsed.writtenFPR1) )
// {
// // get index of register to replace
// sint32 fprToReplace = -1;
// if(_RegExceedsFPRSpace(registersUsed.readFPR1) )
// fprToReplace = registersUsed.readFPR1.GetRegID();
// else if(_RegExceedsFPRSpace(registersUsed.readFPR2) )
// fprToReplace = registersUsed.readFPR2.GetRegID();
// else if (_RegExceedsFPRSpace(registersUsed.readFPR3))
// fprToReplace = registersUsed.readFPR3.GetRegID();
// else if (_RegExceedsFPRSpace(registersUsed.readFPR4))
// fprToReplace = registersUsed.readFPR4.GetRegID();
// else if(_RegExceedsFPRSpace(registersUsed.writtenFPR1) )
// fprToReplace = registersUsed.writtenFPR1.GetRegID();
// if (fprToReplace >= 0)
// {
// // generate mask of useable registers
// uint8 useableRegisterMask = 0x7F; // lowest bit is fpr register 0
// if (registersUsed.readFPR1.IsValid())
// useableRegisterMask &= ~(1 << (registersUsed.readFPR1.GetRegID()));
// if (registersUsed.readFPR2.IsValid())
// useableRegisterMask &= ~(1 << (registersUsed.readFPR2.GetRegID()));
// if (registersUsed.readFPR3.IsValid())
// useableRegisterMask &= ~(1 << (registersUsed.readFPR3.GetRegID()));
// if (registersUsed.readFPR4.IsValid())
// useableRegisterMask &= ~(1 << (registersUsed.readFPR4.GetRegID()));
// if (registersUsed.writtenFPR1.IsValid())
// useableRegisterMask &= ~(1 << (registersUsed.writtenFPR1.GetRegID()));
// // get highest unused register index (0-6 range)
// sint32 unusedRegisterIndex = -1;
// for (sint32 f = 0; f < PPC_X64_FPR_USABLE_REGISTERS; f++)
// {
// if (useableRegisterMask & (1 << f))
// {
// unusedRegisterIndex = f;
// }
// }
// if (unusedRegisterIndex == -1)
// assert_dbg();
// // determine if the placeholder register is actually used (if not we must not load/store it)
// uint32 unusedRegisterName = ppcImlGenContext->mappedFPRRegister[unusedRegisterIndex];
// bool replacedRegisterIsUsed = true;
// if (unusedRegisterName >= PPCREC_NAME_FPR0 && unusedRegisterName < (PPCREC_NAME_FPR0 + 32))
// {
// replacedRegisterIsUsed = segIt->ppcFPRUsed[unusedRegisterName - PPCREC_NAME_FPR0];
// }
// // replace registers that are out of range
// segIt->imlList[imlIndex].ReplaceFPR(fprToReplace, unusedRegisterIndex);
// // add load/store name after instruction
// PPCRecompiler_pushBackIMLInstructions(segIt, imlIndex + 1, 2);
// // add load/store before current instruction
// PPCRecompiler_pushBackIMLInstructions(segIt, imlIndex, 2);
// // name_unusedRegister = unusedRegister
// IMLInstruction* imlInstructionItr = segIt->imlList.data() + (imlIndex + 0);
// memset(imlInstructionItr, 0x00, sizeof(IMLInstruction));
// if (replacedRegisterIsUsed)
// {
// imlInstructionItr->type = PPCREC_IML_TYPE_FPR_NAME_R;
// imlInstructionItr->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionItr->op_r_name.regR = _FPRRegFromID(unusedRegisterIndex);
// imlInstructionItr->op_r_name.name = ppcImlGenContext->mappedFPRRegister[unusedRegisterIndex];
// }
// else
// imlInstructionItr->make_no_op();
// imlInstructionItr = segIt->imlList.data() + (imlIndex + 1);
// memset(imlInstructionItr, 0x00, sizeof(IMLInstruction));
// imlInstructionItr->type = PPCREC_IML_TYPE_FPR_R_NAME;
// imlInstructionItr->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionItr->op_r_name.regR = _FPRRegFromID(unusedRegisterIndex);
// imlInstructionItr->op_r_name.name = ppcImlGenContext->mappedFPRRegister[fprToReplace];
// // name_gprToReplace = unusedRegister
// imlInstructionItr = segIt->imlList.data() + (imlIndex + 3);
// memset(imlInstructionItr, 0x00, sizeof(IMLInstruction));
// imlInstructionItr->type = PPCREC_IML_TYPE_FPR_NAME_R;
// imlInstructionItr->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionItr->op_r_name.regR = _FPRRegFromID(unusedRegisterIndex);
// imlInstructionItr->op_r_name.name = ppcImlGenContext->mappedFPRRegister[fprToReplace];
// // unusedRegister = name_unusedRegister
// imlInstructionItr = segIt->imlList.data() + (imlIndex + 4);
// memset(imlInstructionItr, 0x00, sizeof(IMLInstruction));
// if (replacedRegisterIsUsed)
// {
// imlInstructionItr->type = PPCREC_IML_TYPE_FPR_R_NAME;
// imlInstructionItr->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionItr->op_r_name.regR = _FPRRegFromID(unusedRegisterIndex);
// imlInstructionItr->op_r_name.name = ppcImlGenContext->mappedFPRRegister[unusedRegisterIndex];
// }
// else
// imlInstructionItr->make_no_op();
// }
// }
// else
// break;
// }
// imlIndex++;
// }
// }
// return true;
//}
//
//typedef struct
//{
// bool isActive;
// uint32 virtualReg;
// sint32 lastUseIndex;
//}ppcRecRegisterMapping_t;
//
//typedef struct
//{
// ppcRecRegisterMapping_t currentMapping[PPC_X64_FPR_USABLE_REGISTERS];
// sint32 ppcRegToMapping[64];
// sint32 currentUseIndex;
//}ppcRecManageRegisters_t;
//
//ppcRecRegisterMapping_t* PPCRecompiler_findAvailableRegisterDepr(ppcRecManageRegisters_t* rCtx, IMLUsedRegisters* instructionUsedRegisters)
//{
// // find free register
// for (sint32 i = 0; i < PPC_X64_FPR_USABLE_REGISTERS; i++)
// {
// if (rCtx->currentMapping[i].isActive == false)
// {
// rCtx->currentMapping[i].isActive = true;
// rCtx->currentMapping[i].virtualReg = -1;
// rCtx->currentMapping[i].lastUseIndex = rCtx->currentUseIndex;
// return rCtx->currentMapping + i;
// }
// }
// // all registers are used
// return nullptr;
//}
//
//ppcRecRegisterMapping_t* PPCRecompiler_findUnloadableRegister(ppcRecManageRegisters_t* rCtx, IMLUsedRegisters* instructionUsedRegisters, uint32 unloadLockedMask)
//{
// // find unloadable register (with lowest lastUseIndex)
// sint32 unloadIndex = -1;
// sint32 unloadIndexLastUse = 0x7FFFFFFF;
// for (sint32 i = 0; i < PPC_X64_FPR_USABLE_REGISTERS; i++)
// {
// if (rCtx->currentMapping[i].isActive == false)
// continue;
// if( (unloadLockedMask&(1<<i)) != 0 )
// continue;
// IMLRegID virtualReg = rCtx->currentMapping[i].virtualReg;
// bool isReserved = instructionUsedRegisters->HasSameBaseFPRRegId(virtualReg);
// if (isReserved)
// continue;
// if (rCtx->currentMapping[i].lastUseIndex < unloadIndexLastUse)
// {
// unloadIndexLastUse = rCtx->currentMapping[i].lastUseIndex;
// unloadIndex = i;
// }
// }
// cemu_assert(unloadIndex != -1);
// return rCtx->currentMapping + unloadIndex;
//}
//
//bool PPCRecompiler_manageFPRRegistersForSegment(ppcImlGenContext_t* ppcImlGenContext, sint32 segmentIndex)
//{
// ppcRecManageRegisters_t rCtx = { 0 };
// for (sint32 i = 0; i < 64; i++)
// rCtx.ppcRegToMapping[i] = -1;
// IMLSegment* imlSegment = ppcImlGenContext->segmentList2[segmentIndex];
// size_t idx = 0;
// sint32 currentUseIndex = 0;
// IMLUsedRegisters registersUsed;
// while (idx < imlSegment->imlList.size())
// {
// IMLInstruction& idxInst = imlSegment->imlList[idx];
// if (idxInst.IsSuffixInstruction())
// break;
// idxInst.CheckRegisterUsage(&registersUsed);
// IMLReg fprMatch[4];
// IMLReg fprReplace[4];
// fprMatch[0] = IMLREG_INVALID;
// fprMatch[1] = IMLREG_INVALID;
// fprMatch[2] = IMLREG_INVALID;
// fprMatch[3] = IMLREG_INVALID;
// fprReplace[0] = IMLREG_INVALID;
// fprReplace[1] = IMLREG_INVALID;
// fprReplace[2] = IMLREG_INVALID;
// fprReplace[3] = IMLREG_INVALID;
// // generate a mask of registers that we may not free
// sint32 numReplacedOperands = 0;
// uint32 unloadLockedMask = 0;
// for (sint32 f = 0; f < 5; f++)
// {
// IMLReg virtualFpr;
// if (f == 0)
// virtualFpr = registersUsed.readFPR1;
// else if (f == 1)
// virtualFpr = registersUsed.readFPR2;
// else if (f == 2)
// virtualFpr = registersUsed.readFPR3;
// else if (f == 3)
// virtualFpr = registersUsed.readFPR4;
// else if (f == 4)
// virtualFpr = registersUsed.writtenFPR1;
// if(virtualFpr.IsInvalid())
// continue;
// cemu_assert_debug(virtualFpr.GetBaseFormat() == IMLRegFormat::F64);
// cemu_assert_debug(virtualFpr.GetRegFormat() == IMLRegFormat::F64);
// cemu_assert_debug(virtualFpr.GetRegID() < 64);
// // check if this virtual FPR is already loaded in any real register
// ppcRecRegisterMapping_t* regMapping;
// if (rCtx.ppcRegToMapping[virtualFpr.GetRegID()] == -1)
// {
// // not loaded
// // find available register
// while (true)
// {
// regMapping = PPCRecompiler_findAvailableRegisterDepr(&rCtx, &registersUsed);
// if (regMapping == NULL)
// {
// // unload least recently used register and try again
// ppcRecRegisterMapping_t* unloadRegMapping = PPCRecompiler_findUnloadableRegister(&rCtx, &registersUsed, unloadLockedMask);
// // mark as locked
// unloadLockedMask |= (1<<(unloadRegMapping- rCtx.currentMapping));
// // create unload instruction
// PPCRecompiler_pushBackIMLInstructions(imlSegment, idx, 1);
// IMLInstruction* imlInstructionTemp = imlSegment->imlList.data() + idx;
// memset(imlInstructionTemp, 0x00, sizeof(IMLInstruction));
// imlInstructionTemp->type = PPCREC_IML_TYPE_FPR_NAME_R;
// imlInstructionTemp->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionTemp->op_r_name.regR = _FPRRegFromID((uint8)(unloadRegMapping - rCtx.currentMapping));
// imlInstructionTemp->op_r_name.name = ppcImlGenContext->mappedFPRRegister[unloadRegMapping->virtualReg];
// idx++;
// // update mapping
// unloadRegMapping->isActive = false;
// rCtx.ppcRegToMapping[unloadRegMapping->virtualReg] = -1;
// }
// else
// break;
// }
// // create load instruction
// PPCRecompiler_pushBackIMLInstructions(imlSegment, idx, 1);
// IMLInstruction* imlInstructionTemp = imlSegment->imlList.data() + idx;
// memset(imlInstructionTemp, 0x00, sizeof(IMLInstruction));
// imlInstructionTemp->type = PPCREC_IML_TYPE_FPR_R_NAME;
// imlInstructionTemp->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionTemp->op_r_name.regR = _FPRRegFromID((uint8)(regMapping-rCtx.currentMapping));
// imlInstructionTemp->op_r_name.name = ppcImlGenContext->mappedFPRRegister[virtualFpr.GetRegID()];
// idx++;
// // update mapping
// regMapping->virtualReg = virtualFpr.GetRegID();
// rCtx.ppcRegToMapping[virtualFpr.GetRegID()] = (sint32)(regMapping - rCtx.currentMapping);
// regMapping->lastUseIndex = rCtx.currentUseIndex;
// rCtx.currentUseIndex++;
// }
// else
// {
// regMapping = rCtx.currentMapping + rCtx.ppcRegToMapping[virtualFpr.GetRegID()];
// regMapping->lastUseIndex = rCtx.currentUseIndex;
// rCtx.currentUseIndex++;
// }
// // replace FPR
// bool entryFound = false;
// for (sint32 t = 0; t < numReplacedOperands; t++)
// {
// if (fprMatch[t].IsValid() && fprMatch[t].GetRegID() == virtualFpr.GetRegID())
// {
// cemu_assert_debug(fprReplace[t] == _FPRRegFromID(regMapping - rCtx.currentMapping));
// entryFound = true;
// break;
// }
// }
// if (entryFound == false)
// {
// cemu_assert_debug(numReplacedOperands != 4);
// fprMatch[numReplacedOperands] = virtualFpr;
// fprReplace[numReplacedOperands] = _FPRRegFromID(regMapping - rCtx.currentMapping);
// numReplacedOperands++;
// }
// }
// if (numReplacedOperands > 0)
// {
// imlSegment->imlList[idx].ReplaceFPRs(fprMatch, fprReplace);
// }
// // next
// idx++;
// }
// // count loaded registers
// sint32 numLoadedRegisters = 0;
// for (sint32 i = 0; i < PPC_X64_FPR_USABLE_REGISTERS; i++)
// {
// if (rCtx.currentMapping[i].isActive)
// numLoadedRegisters++;
// }
// // store all loaded registers
// if (numLoadedRegisters > 0)
// {
// PPCRecompiler_pushBackIMLInstructions(imlSegment, idx, numLoadedRegisters);
// for (sint32 i = 0; i < PPC_X64_FPR_USABLE_REGISTERS; i++)
// {
// if (rCtx.currentMapping[i].isActive == false)
// continue;
// IMLInstruction* imlInstructionTemp = imlSegment->imlList.data() + idx;
// memset(imlInstructionTemp, 0x00, sizeof(IMLInstruction));
// imlInstructionTemp->type = PPCREC_IML_TYPE_FPR_NAME_R;
// imlInstructionTemp->operation = PPCREC_IML_OP_ASSIGN;
// imlInstructionTemp->op_r_name.regR = _FPRRegFromID(i);
// imlInstructionTemp->op_r_name.name = ppcImlGenContext->mappedFPRRegister[rCtx.currentMapping[i].virtualReg];
// idx++;
// }
// }
// return true;
//}
//
//bool PPCRecompiler_manageFPRRegisters(ppcImlGenContext_t* ppcImlGenContext)
//{
// for (sint32 s = 0; s < ppcImlGenContext->segmentList2.size(); s++)
// {
// if (PPCRecompiler_manageFPRRegistersForSegment(ppcImlGenContext, s) == false)
// return false;
// }
// return true;
//}
/*
* Returns true if the loaded value is guaranteed to be overwritten
*/
bool PPCRecompiler_trackRedundantNameLoadInstruction(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 startIndex, IMLInstruction* nameStoreInstruction, sint32 scanDepth)
{
IMLReg registerIndex = nameStoreInstruction->op_r_name.regR;
for(size_t i=startIndex; i<imlSegment->imlList.size(); i++)
{
IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
IMLUsedRegisters registersUsed;
imlInstruction->CheckRegisterUsage(&registersUsed);
if( registersUsed.readGPR1 == registerIndex || registersUsed.readGPR2 == registerIndex || registersUsed.readGPR3 == registerIndex )
return false;
if (registersUsed.IsBaseGPRWritten(registerIndex))
return true;
}
// todo: Scan next segment(s)
return false;
}
/*
* Returns true if the loaded value is guaranteed to be overwritten
*/
bool PPCRecompiler_trackRedundantFPRNameLoadInstruction(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 startIndex, IMLInstruction* nameStoreInstruction, sint32 scanDepth)
{
IMLRegID regId = nameStoreInstruction->op_r_name.regR.GetRegID();
for(size_t i=startIndex; i<imlSegment->imlList.size(); i++)
{
IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
IMLUsedRegisters registersUsed;
imlInstruction->CheckRegisterUsage(&registersUsed);
if( registersUsed.readFPR1.IsValidAndSameRegID(regId) || registersUsed.readFPR2.IsValidAndSameRegID(regId) || registersUsed.readFPR3.IsValidAndSameRegID(regId) || registersUsed.readFPR4.IsValidAndSameRegID(regId))
return false;
if( registersUsed.writtenFPR1.IsValidAndSameRegID(regId) )
return true;
}
// todo: Scan next segment(s)
return false;
}
/*
* Returns true if the loaded name is never changed
*/
bool PPCRecompiler_trackRedundantNameStoreInstruction(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 startIndex, IMLInstruction* nameStoreInstruction, sint32 scanDepth)
{
IMLReg regR = nameStoreInstruction->op_r_name.regR;
for(sint32 i=startIndex; i>=0; i--)
{
IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
IMLUsedRegisters registersUsed;
imlInstruction->CheckRegisterUsage(&registersUsed);
if( registersUsed.IsBaseGPRWritten(regR) )
{
if( imlSegment->imlList[i].type == PPCREC_IML_TYPE_R_NAME )
return true;
return false;
}
}
return false;
}
sint32 debugCallCounter1 = 0;
/*
* Returns true if the name is overwritten in the current or any following segments
*/
bool PPCRecompiler_trackOverwrittenNameStoreInstruction(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 startIndex, IMLInstruction* nameStoreInstruction, sint32 scanDepth)
{
uint32 name = nameStoreInstruction->op_r_name.name;
for(size_t i=startIndex; i<imlSegment->imlList.size(); i++)
{
const IMLInstruction& imlInstruction = imlSegment->imlList[i];
if(imlInstruction.type == PPCREC_IML_TYPE_R_NAME )
{
// name is loaded before being written
if (imlInstruction.op_r_name.name == name)
return false;
}
else if(imlInstruction.type == PPCREC_IML_TYPE_NAME_R )
{
// name is written before being loaded
if (imlInstruction.op_r_name.name == name)
return true;
}
}
if( scanDepth >= 2 )
return false;
if( imlSegment->nextSegmentIsUncertain )
return false;
if( imlSegment->nextSegmentBranchTaken && PPCRecompiler_trackOverwrittenNameStoreInstruction(ppcImlGenContext, imlSegment->nextSegmentBranchTaken, 0, nameStoreInstruction, scanDepth+1) == false )
return false;
if( imlSegment->nextSegmentBranchNotTaken && PPCRecompiler_trackOverwrittenNameStoreInstruction(ppcImlGenContext, imlSegment->nextSegmentBranchNotTaken, 0, nameStoreInstruction, scanDepth+1) == false )
return false;
if( imlSegment->nextSegmentBranchTaken == nullptr && imlSegment->nextSegmentBranchNotTaken == nullptr)
return false;
return true;
}
/*
* Returns true if the loaded FPR name is never changed
*/
bool PPCRecompiler_trackRedundantFPRNameStoreInstruction(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 startIndex, IMLInstruction* nameStoreInstruction, sint32 scanDepth)
{
IMLRegID regId = nameStoreInstruction->op_r_name.regR.GetRegID();
for(sint32 i=startIndex; i>=0; i--)
{
IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
IMLUsedRegisters registersUsed;
imlInstruction->CheckRegisterUsage(&registersUsed);
if( registersUsed.writtenFPR1.IsValidAndSameRegID(regId))
{
if(imlInstruction->type == PPCREC_IML_TYPE_FPR_R_NAME )
return true;
return false;
}
}
// todo: Scan next segment(s)
return false;
}
void PPCRecompiler_optimizeDirectFloatCopiesScanForward(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 imlIndexLoad, IMLReg fprReg)
{
IMLRegID fprIndex = fprReg.GetRegID();
IMLInstruction* imlInstructionLoad = imlSegment->imlList.data() + imlIndexLoad;
if (imlInstructionLoad->op_storeLoad.flags2.notExpanded)
return;
IMLUsedRegisters registersUsed;
sint32 scanRangeEnd = std::min<sint32>(imlIndexLoad + 25, imlSegment->imlList.size()); // don't scan too far (saves performance and also the chances we can merge the load+store become low at high distances)
bool foundMatch = false;
sint32 lastStore = -1;
for (sint32 i = imlIndexLoad + 1; i < scanRangeEnd; i++)
{
IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
if (imlInstruction->IsSuffixInstruction())
break;
// check if FPR is stored
if ((imlInstruction->type == PPCREC_IML_TYPE_FPR_STORE && imlInstruction->op_storeLoad.mode == PPCREC_FPR_ST_MODE_SINGLE_FROM_PS0) ||
(imlInstruction->type == PPCREC_IML_TYPE_FPR_STORE_INDEXED && imlInstruction->op_storeLoad.mode == PPCREC_FPR_ST_MODE_SINGLE_FROM_PS0))
{
if (imlInstruction->op_storeLoad.registerData.GetRegID() == fprIndex)
{
if (foundMatch == false)
{
// flag the load-single instruction as "don't expand" (leave single value as-is)
imlInstructionLoad->op_storeLoad.flags2.notExpanded = true;
}
// also set the flag for the store instruction
IMLInstruction* imlInstructionStore = imlInstruction;
imlInstructionStore->op_storeLoad.flags2.notExpanded = true;
foundMatch = true;
lastStore = i + 1;
continue;
}
}
// check if FPR is overwritten (we can actually ignore read operations?)
imlInstruction->CheckRegisterUsage(&registersUsed);
if (registersUsed.writtenFPR1.IsValidAndSameRegID(fprIndex))
break;
if (registersUsed.readFPR1.IsValidAndSameRegID(fprIndex))
break;
if (registersUsed.readFPR2.IsValidAndSameRegID(fprIndex))
break;
if (registersUsed.readFPR3.IsValidAndSameRegID(fprIndex))
break;
if (registersUsed.readFPR4.IsValidAndSameRegID(fprIndex))
break;
}
if (foundMatch)
{
// insert expand instruction after store
IMLInstruction* newExpand = PPCRecompiler_insertInstruction(imlSegment, lastStore);
PPCRecompilerImlGen_generateNewInstruction_fpr_r(ppcImlGenContext, newExpand, PPCREC_IML_OP_FPR_EXPAND_BOTTOM32_TO_BOTTOM64_AND_TOP64, _FPRRegFromID(fprIndex));
}
}
/*
* Scans for patterns:
* <Load sp float into register f>
* <Random unrelated instructions>
* <Store sp float from register f>
* For these patterns the store and load is modified to work with un-extended values (float remains as float, no double conversion)
* The float->double extension is then executed later
* Advantages:
* Keeps denormals and other special float values intact
* Slightly improves performance
*/
void PPCRecompiler_optimizeDirectFloatCopies(ppcImlGenContext_t* ppcImlGenContext)
{
for (IMLSegment* segIt : ppcImlGenContext->segmentList2)
{
for (sint32 i = 0; i < segIt->imlList.size(); i++)
{
IMLInstruction* imlInstruction = segIt->imlList.data() + i;
if (imlInstruction->type == PPCREC_IML_TYPE_FPR_LOAD && imlInstruction->op_storeLoad.mode == PPCREC_FPR_LD_MODE_SINGLE_INTO_PS0_PS1)
{
PPCRecompiler_optimizeDirectFloatCopiesScanForward(ppcImlGenContext, segIt, i, imlInstruction->op_storeLoad.registerData);
}
else if (imlInstruction->type == PPCREC_IML_TYPE_FPR_LOAD_INDEXED && imlInstruction->op_storeLoad.mode == PPCREC_FPR_LD_MODE_SINGLE_INTO_PS0_PS1)
{
PPCRecompiler_optimizeDirectFloatCopiesScanForward(ppcImlGenContext, segIt, i, imlInstruction->op_storeLoad.registerData);
}
}
}
}
void PPCRecompiler_optimizeDirectIntegerCopiesScanForward(ppcImlGenContext_t* ppcImlGenContext, IMLSegment* imlSegment, sint32 imlIndexLoad, IMLReg gprReg)
{
cemu_assert_debug(gprReg.GetBaseFormat() == IMLRegFormat::I64); // todo - proper handling required for non-standard sizes
cemu_assert_debug(gprReg.GetRegFormat() == IMLRegFormat::I32);
IMLRegID gprIndex = gprReg.GetRegID();
IMLInstruction* imlInstructionLoad = imlSegment->imlList.data() + imlIndexLoad;
if ( imlInstructionLoad->op_storeLoad.flags2.swapEndian == false )
return;
bool foundMatch = false;
IMLUsedRegisters registersUsed;
sint32 scanRangeEnd = std::min<sint32>(imlIndexLoad + 25, imlSegment->imlList.size()); // don't scan too far (saves performance and also the chances we can merge the load+store become low at high distances)
sint32 i = imlIndexLoad + 1;
for (; i < scanRangeEnd; i++)
{
IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
if (imlInstruction->IsSuffixInstruction())
break;
// check if GPR is stored
if ((imlInstruction->type == PPCREC_IML_TYPE_STORE && imlInstruction->op_storeLoad.copyWidth == 32 ) )
{
if (imlInstruction->op_storeLoad.registerMem.GetRegID() == gprIndex)
break;
if (imlInstruction->op_storeLoad.registerData.GetRegID() == gprIndex)
{
IMLInstruction* imlInstructionStore = imlInstruction;
if (foundMatch == false)
{
// switch the endian swap flag for the load instruction
imlInstructionLoad->op_storeLoad.flags2.swapEndian = !imlInstructionLoad->op_storeLoad.flags2.swapEndian;
foundMatch = true;
}
// switch the endian swap flag for the store instruction
imlInstructionStore->op_storeLoad.flags2.swapEndian = !imlInstructionStore->op_storeLoad.flags2.swapEndian;
// keep scanning
continue;
}
}
// check if GPR is accessed
imlInstruction->CheckRegisterUsage(&registersUsed);
if (registersUsed.readGPR1.IsValidAndSameRegID(gprIndex) ||
registersUsed.readGPR2.IsValidAndSameRegID(gprIndex) ||
registersUsed.readGPR3.IsValidAndSameRegID(gprIndex))
{
break;
}
if (registersUsed.IsBaseGPRWritten(gprReg))
return; // GPR overwritten, we don't need to byte swap anymore
}
if (foundMatch)
{
PPCRecompiler_insertInstruction(imlSegment, i)->make_r_r(PPCREC_IML_OP_ENDIAN_SWAP, gprReg, gprReg);
}
}
/*
* Scans for patterns:
* <Load sp integer into register r>
* <Random unrelated instructions>
* <Store sp integer from register r>
* For these patterns the store and load is modified to work with non-swapped values
* The big_endian->little_endian conversion is then executed later
* Advantages:
* Slightly improves performance
*/
void PPCRecompiler_optimizeDirectIntegerCopies(ppcImlGenContext_t* ppcImlGenContext)
{
for (IMLSegment* segIt : ppcImlGenContext->segmentList2)
{
for (sint32 i = 0; i < segIt->imlList.size(); i++)
{
IMLInstruction* imlInstruction = segIt->imlList.data() + i;
if (imlInstruction->type == PPCREC_IML_TYPE_LOAD && imlInstruction->op_storeLoad.copyWidth == 32 && imlInstruction->op_storeLoad.flags2.swapEndian )
{
PPCRecompiler_optimizeDirectIntegerCopiesScanForward(ppcImlGenContext, segIt, i, imlInstruction->op_storeLoad.registerData);
}
}
}
}
IMLName PPCRecompilerImlGen_GetRegName(ppcImlGenContext_t* ppcImlGenContext, IMLReg reg);
sint32 _getGQRIndexFromRegister(ppcImlGenContext_t* ppcImlGenContext, IMLReg gqrReg)
{
if (gqrReg.IsInvalid())
return -1;
sint32 namedReg = PPCRecompilerImlGen_GetRegName(ppcImlGenContext, gqrReg);
if (namedReg >= (PPCREC_NAME_SPR0 + SPR_UGQR0) && namedReg <= (PPCREC_NAME_SPR0 + SPR_UGQR7))
{
return namedReg - (PPCREC_NAME_SPR0 + SPR_UGQR0);
}
else
{
cemu_assert_suspicious();
}
return -1;
}
bool PPCRecompiler_isUGQRValueKnown(ppcImlGenContext_t* ppcImlGenContext, sint32 gqrIndex, uint32& gqrValue)
{
// UGQR 2 to 7 are initialized by the OS and we assume that games won't ever permanently touch those
// todo - hack - replace with more accurate solution
if (gqrIndex == 2)
gqrValue = 0x00040004;
else if (gqrIndex == 3)
gqrValue = 0x00050005;
else if (gqrIndex == 4)
gqrValue = 0x00060006;
else if (gqrIndex == 5)
gqrValue = 0x00070007;
else
return false;
return true;
}
/*
* If value of GQR can be predicted for a given PSQ load or store instruction then replace it with an optimized version
*/
void PPCRecompiler_optimizePSQLoadAndStore(ppcImlGenContext_t* ppcImlGenContext)
{
for (IMLSegment* segIt : ppcImlGenContext->segmentList2)
{
for(IMLInstruction& instIt : segIt->imlList)
{
if (instIt.type == PPCREC_IML_TYPE_FPR_LOAD || instIt.type == PPCREC_IML_TYPE_FPR_LOAD_INDEXED)
{
if(instIt.op_storeLoad.mode != PPCREC_FPR_LD_MODE_PSQ_GENERIC_PS0 &&
instIt.op_storeLoad.mode != PPCREC_FPR_LD_MODE_PSQ_GENERIC_PS0_PS1 )
continue;
// get GQR value
cemu_assert_debug(instIt.op_storeLoad.registerGQR.IsValid());
sint32 gqrIndex = _getGQRIndexFromRegister(ppcImlGenContext, instIt.op_storeLoad.registerGQR);
cemu_assert(gqrIndex >= 0);
if (ppcImlGenContext->tracking.modifiesGQR[gqrIndex])
continue;
uint32 gqrValue;
if (!PPCRecompiler_isUGQRValueKnown(ppcImlGenContext, gqrIndex, gqrValue))
continue;
uint32 formatType = (gqrValue >> 16) & 7;
uint32 scale = (gqrValue >> 24) & 0x3F;
if (scale != 0)
continue; // only generic handler supports scale
if (instIt.op_storeLoad.mode == PPCREC_FPR_LD_MODE_PSQ_GENERIC_PS0)
{
if (formatType == 0)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_FLOAT_PS0;
else if (formatType == 4)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_U8_PS0;
else if (formatType == 5)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_U16_PS0;
else if (formatType == 6)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_S8_PS0;
else if (formatType == 7)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_S16_PS0;
if (instIt.op_storeLoad.mode != PPCREC_FPR_LD_MODE_PSQ_GENERIC_PS0)
instIt.op_storeLoad.registerGQR = IMLREG_INVALID;
}
else if (instIt.op_storeLoad.mode == PPCREC_FPR_LD_MODE_PSQ_GENERIC_PS0_PS1)
{
if (formatType == 0)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_FLOAT_PS0_PS1;
else if (formatType == 4)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_U8_PS0_PS1;
else if (formatType == 5)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_U16_PS0_PS1;
else if (formatType == 6)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_S8_PS0_PS1;
else if (formatType == 7)
instIt.op_storeLoad.mode = PPCREC_FPR_LD_MODE_PSQ_S16_PS0_PS1;
if (instIt.op_storeLoad.mode != PPCREC_FPR_LD_MODE_PSQ_GENERIC_PS0_PS1)
instIt.op_storeLoad.registerGQR = IMLREG_INVALID;
}
}
else if (instIt.type == PPCREC_IML_TYPE_FPR_STORE || instIt.type == PPCREC_IML_TYPE_FPR_STORE_INDEXED)
{
if(instIt.op_storeLoad.mode != PPCREC_FPR_ST_MODE_PSQ_GENERIC_PS0 &&
instIt.op_storeLoad.mode != PPCREC_FPR_ST_MODE_PSQ_GENERIC_PS0_PS1)
continue;
// get GQR value
cemu_assert_debug(instIt.op_storeLoad.registerGQR.IsValid());
sint32 gqrIndex = _getGQRIndexFromRegister(ppcImlGenContext, instIt.op_storeLoad.registerGQR);
cemu_assert(gqrIndex >= 0 && gqrIndex < 8);
if (ppcImlGenContext->tracking.modifiesGQR[gqrIndex])
continue;
uint32 gqrValue;
if(!PPCRecompiler_isUGQRValueKnown(ppcImlGenContext, gqrIndex, gqrValue))
continue;
uint32 formatType = (gqrValue >> 16) & 7;
uint32 scale = (gqrValue >> 24) & 0x3F;
if (scale != 0)
continue; // only generic handler supports scale
if (instIt.op_storeLoad.mode == PPCREC_FPR_ST_MODE_PSQ_GENERIC_PS0)
{
if (formatType == 0)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_FLOAT_PS0;
else if (formatType == 4)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_U8_PS0;
else if (formatType == 5)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_U16_PS0;
else if (formatType == 6)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_S8_PS0;
else if (formatType == 7)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_S16_PS0;
if (instIt.op_storeLoad.mode != PPCREC_FPR_ST_MODE_PSQ_GENERIC_PS0)
instIt.op_storeLoad.registerGQR = IMLREG_INVALID;
}
else if (instIt.op_storeLoad.mode == PPCREC_FPR_ST_MODE_PSQ_GENERIC_PS0_PS1)
{
if (formatType == 0)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_FLOAT_PS0_PS1;
else if (formatType == 4)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_U8_PS0_PS1;
else if (formatType == 5)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_U16_PS0_PS1;
else if (formatType == 6)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_S8_PS0_PS1;
else if (formatType == 7)
instIt.op_storeLoad.mode = PPCREC_FPR_ST_MODE_PSQ_S16_PS0_PS1;
if (instIt.op_storeLoad.mode != PPCREC_FPR_ST_MODE_PSQ_GENERIC_PS0_PS1)
instIt.op_storeLoad.registerGQR = IMLREG_INVALID;
}
}
}
}
}
///*
// * Returns true if registerWrite overwrites any of the registers read by registerRead
// */
//bool PPCRecompilerAnalyzer_checkForGPROverwrite(IMLUsedRegisters* registerRead, IMLUsedRegisters* registerWrite)
//{
// if (registerWrite->writtenNamedReg1 < 0)
// return false;
//
// if (registerWrite->writtenNamedReg1 == registerRead->readNamedReg1)
// return true;
// if (registerWrite->writtenNamedReg1 == registerRead->readNamedReg2)
// return true;
// if (registerWrite->writtenNamedReg1 == registerRead->readNamedReg3)
// return true;
// return false;
//}
void _reorderConditionModifyInstructions(IMLSegment* imlSegment)
{
// IMLInstruction* lastInstruction = imlSegment->GetLastInstruction();
// // last instruction is a conditional branch?
// if (lastInstruction == nullptr || lastInstruction->type != PPCREC_IML_TYPE_CJUMP)
// return;
// if (lastInstruction->op_conditionalJump.crRegisterIndex >= 8)
// return;
// // get CR bitmask of bit required for conditional jump
// PPCRecCRTracking_t crTracking;
// IMLAnalyzer_GetCRTracking(lastInstruction, &crTracking);
// uint32 requiredCRBits = crTracking.readCRBits;
//
// // scan backwards until we find the instruction that sets the CR
// sint32 crSetterInstructionIndex = -1;
// sint32 unsafeInstructionIndex = -1;
// for (sint32 i = imlSegment->imlList.size() - 2; i >= 0; i--)
// {
// IMLInstruction* imlInstruction = imlSegment->imlList.data() + i;
// IMLAnalyzer_GetCRTracking(imlInstruction, &crTracking);
// if (crTracking.readCRBits != 0)
// return; // dont handle complex cases for now
// if (crTracking.writtenCRBits != 0)
// {
// if ((crTracking.writtenCRBits&requiredCRBits) != 0)
// {
// crSetterInstructionIndex = i;
// break;
// }
// else
// {
// return; // other CR bits overwritten (dont handle complex cases)
// }
// }
// // is safe? (no risk of overwriting x64 eflags)
// if ((imlInstruction->type == PPCREC_IML_TYPE_NAME_R || imlInstruction->type == PPCREC_IML_TYPE_R_NAME || imlInstruction->type == PPCREC_IML_TYPE_NO_OP) ||
// (imlInstruction->type == PPCREC_IML_TYPE_FPR_NAME_R || imlInstruction->type == PPCREC_IML_TYPE_FPR_R_NAME) ||
// (imlInstruction->type == PPCREC_IML_TYPE_R_S32 && (imlInstruction->operation == PPCREC_IML_OP_ASSIGN)) ||
// (imlInstruction->type == PPCREC_IML_TYPE_R_R && (imlInstruction->operation == PPCREC_IML_OP_ASSIGN)) )
// continue;
// // not safe
// if (unsafeInstructionIndex == -1)
// unsafeInstructionIndex = i;
// }
// if (crSetterInstructionIndex < 0)
// return;
// if (unsafeInstructionIndex < 0)
// return; // no danger of overwriting eflags, don't reorder
// // check if we can move the CR setter instruction to after unsafeInstructionIndex
// PPCRecCRTracking_t crTrackingSetter = crTracking;
// IMLUsedRegisters regTrackingCRSetter;
// imlSegment->imlList[crSetterInstructionIndex].CheckRegisterUsage(&regTrackingCRSetter);
// if (regTrackingCRSetter.writtenFPR1 >= 0 || regTrackingCRSetter.readFPR1 >= 0 || regTrackingCRSetter.readFPR2 >= 0 || regTrackingCRSetter.readFPR3 >= 0 || regTrackingCRSetter.readFPR4 >= 0)
// return; // we don't handle FPR dependency yet so just ignore FPR instructions
// IMLUsedRegisters registerTracking;
// if (regTrackingCRSetter.writtenNamedReg1 >= 0)
// {
// // CR setter does write GPR
// for (sint32 i = crSetterInstructionIndex + 1; i <= unsafeInstructionIndex; i++)
// {
// imlSegment->imlList[i].CheckRegisterUsage(&registerTracking);
// // reads register written by CR setter?
// if (PPCRecompilerAnalyzer_checkForGPROverwrite(&registerTracking, &regTrackingCRSetter))
// {
// return; // cant move CR setter because of dependency
// }
// // writes register read by CR setter?
// if (PPCRecompilerAnalyzer_checkForGPROverwrite(&regTrackingCRSetter, &registerTracking))
// {
// return; // cant move CR setter because of dependency
// }
// // overwrites register written by CR setter?
// if (regTrackingCRSetter.writtenNamedReg1 == registerTracking.writtenNamedReg1)
// return;
// }
// }
// else
// {
// // CR setter does not write GPR
// for (sint32 i = crSetterInstructionIndex + 1; i <= unsafeInstructionIndex; i++)
// {
// imlSegment->imlList[i].CheckRegisterUsage(&registerTracking);
// // writes register read by CR setter?
// if (PPCRecompilerAnalyzer_checkForGPROverwrite(&regTrackingCRSetter, &registerTracking))
// {
// return; // cant move CR setter because of dependency
// }
// }
// }
//
// // move CR setter instruction
//#ifdef CEMU_DEBUG_ASSERT
// if ((unsafeInstructionIndex + 1) <= crSetterInstructionIndex)
// assert_dbg();
//#endif
// IMLInstruction* newCRSetterInstruction = PPCRecompiler_insertInstruction(imlSegment, unsafeInstructionIndex+1);
// memcpy(newCRSetterInstruction, imlSegment->imlList.data() + crSetterInstructionIndex, sizeof(IMLInstruction));
// imlSegment->imlList[crSetterInstructionIndex].make_no_op();
}
/*
* Move instructions which update the condition flags closer to the instruction that consumes them
* On x64 this improves performance since we often can avoid storing CR in memory
*/
void PPCRecompiler_reorderConditionModifyInstructions(ppcImlGenContext_t* ppcImlGenContext)
{
// check if this segment has a conditional branch
for (IMLSegment* segIt : ppcImlGenContext->segmentList2)
{
_reorderConditionModifyInstructions(segIt);
}
}