fs-srv: GetFreeSpaceSize & GetTotalSpaceSize

Closes #3533

Turns out the functions were already implemented but just never added

externals/microprofile/microprofile.h

@@ -910,14 +910,14 @@ typedef void* (*MicroProfileThreadFunc)(void*);
 
 #ifndef _WIN32
 typedef pthread_t MicroProfileThread;
-void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func)
+inline void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func)
 {
     pthread_attr_t Attr;
     int r  = pthread_attr_init(&Attr);
     MP_ASSERT(r == 0);
     pthread_create(pThread, &Attr, Func, 0);
 }
-void MicroProfileThreadJoin(MicroProfileThread* pThread)
+inline void MicroProfileThreadJoin(MicroProfileThread* pThread)
 {
     int r = pthread_join(*pThread, 0);
     MP_ASSERT(r == 0);
@@ -930,11 +930,11 @@ DWORD _stdcall ThreadTrampoline(void* pFunc)
     return (uint32_t)F(0);
 }
 
-void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func)
+inline void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func)
 {
     *pThread = CreateThread(0, 0, ThreadTrampoline, Func, 0, 0);
 }
-void MicroProfileThreadJoin(MicroProfileThread* pThread)
+inline void MicroProfileThreadJoin(MicroProfileThread* pThread)
 {
     WaitForSingleObject(*pThread, INFINITE);
     CloseHandle(*pThread);
@@ -1131,7 +1131,7 @@ inline void MicroProfileSetThreadLog(MicroProfileThreadLog* pLog)
     pthread_setspecific(g_MicroProfileThreadLogKey, pLog);
 }
 #else
-MicroProfileThreadLog* MicroProfileGetThreadLog()
+inline MicroProfileThreadLog* MicroProfileGetThreadLog()
 {
     return g_MicroProfileThreadLog;
 }
@@ -1247,7 +1247,7 @@ MicroProfileToken MicroProfileFindToken(const char* pGroup, const char* pName)
     return MICROPROFILE_INVALID_TOKEN;
 }
 
-uint16_t MicroProfileGetGroup(const char* pGroup, MicroProfileTokenType Type)
+inline uint16_t MicroProfileGetGroup(const char* pGroup, MicroProfileTokenType Type)
 {
     for(uint32_t i = 0; i < S.nGroupCount; ++i)
     {
@@ -1276,7 +1276,7 @@ uint16_t MicroProfileGetGroup(const char* pGroup, MicroProfileTokenType Type)
     return nGroupIndex;
 }
 
-void MicroProfileRegisterGroup(const char* pGroup, const char* pCategory, uint32_t nColor)
+inline void MicroProfileRegisterGroup(const char* pGroup, const char* pCategory, uint32_t nColor)
 {
     int nCategoryIndex = -1;
     for(uint32_t i = 0; i < S.nCategoryCount; ++i)
@@ -1442,7 +1442,7 @@ void MicroProfileGpuLeave(MicroProfileToken nToken_, uint64_t nTickStart)
     }
 }
 
-void MicroProfileContextSwitchPut(MicroProfileContextSwitch* pContextSwitch)
+inline void MicroProfileContextSwitchPut(MicroProfileContextSwitch* pContextSwitch)
 {
     if(S.nRunning || pContextSwitch->nTicks <= S.nPauseTicks)
     {
@@ -1894,7 +1894,7 @@ void MicroProfileSetEnableAllGroups(bool bEnableAllGroups)
     S.nAllGroupsWanted = bEnableAllGroups ? 1 : 0;
 }
 
-void MicroProfileEnableCategory(const char* pCategory, bool bEnabled)
+inline void MicroProfileEnableCategory(const char* pCategory, bool bEnabled)
 {
     int nCategoryIndex = -1;
     for(uint32_t i = 0; i < S.nCategoryCount; ++i)
@@ -2004,7 +2004,7 @@ void MicroProfileForceDisableGroup(const char* pGroup, MicroProfileTokenType Typ
 }
 
 
-void MicroProfileCalcAllTimers(float* pTimers, float* pAverage, float* pMax, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, float* pTotal, uint32_t nSize)
+inline void MicroProfileCalcAllTimers(float* pTimers, float* pAverage, float* pMax, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, float* pTotal, uint32_t nSize)
 {
     for(uint32_t i = 0; i < S.nTotalTimers && i < nSize; ++i)
     {

externals/microprofile/microprofileui.h

@@ -417,19 +417,19 @@ void MicroProfileToggleDisplayMode()
 }
 
 
-void MicroProfileStringArrayClear(MicroProfileStringArray* pArray)
+inline void MicroProfileStringArrayClear(MicroProfileStringArray* pArray)
 {
     pArray->nNumStrings = 0;
     pArray->pBufferPos = &pArray->Buffer[0];
 }
 
-void MicroProfileStringArrayAddLiteral(MicroProfileStringArray* pArray, const char* pLiteral)
+inline void MicroProfileStringArrayAddLiteral(MicroProfileStringArray* pArray, const char* pLiteral)
 {
     MP_ASSERT(pArray->nNumStrings < MICROPROFILE_TOOLTIP_MAX_STRINGS);
     pArray->ppStrings[pArray->nNumStrings++] = pLiteral;
 }
 
-void MicroProfileStringArrayFormat(MicroProfileStringArray* pArray, const char* fmt, ...)
+inline void MicroProfileStringArrayFormat(MicroProfileStringArray* pArray, const char* fmt, ...)
 {
     MP_ASSERT(pArray->nNumStrings < MICROPROFILE_TOOLTIP_MAX_STRINGS);
     pArray->ppStrings[pArray->nNumStrings++] = pArray->pBufferPos;
@@ -439,7 +439,7 @@ void MicroProfileStringArrayFormat(MicroProfileStringArray* pArray, const char*
     va_end(args);
     MP_ASSERT(pArray->pBufferPos < pArray->Buffer + MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE);
 }
-void MicroProfileStringArrayCopy(MicroProfileStringArray* pDest, MicroProfileStringArray* pSrc)
+inline void MicroProfileStringArrayCopy(MicroProfileStringArray* pDest, MicroProfileStringArray* pSrc)
 {
     memcpy(&pDest->ppStrings[0], &pSrc->ppStrings[0], sizeof(pDest->ppStrings));
     memcpy(&pDest->Buffer[0], &pSrc->Buffer[0], sizeof(pDest->Buffer));
@@ -456,7 +456,7 @@ void MicroProfileStringArrayCopy(MicroProfileStringArray* pDest, MicroProfileStr
     pDest->nNumStrings = pSrc->nNumStrings;
 }
 
-void MicroProfileFloatWindowSize(const char** ppStrings, uint32_t nNumStrings, uint32_t* pColors, uint32_t& nWidth, uint32_t& nHeight, uint32_t* pStringLengths = 0)
+inline void MicroProfileFloatWindowSize(const char** ppStrings, uint32_t nNumStrings, uint32_t* pColors, uint32_t& nWidth, uint32_t& nHeight, uint32_t* pStringLengths = 0)
 {
     uint32_t* nStringLengths = pStringLengths ? pStringLengths : (uint32_t*)alloca(nNumStrings * sizeof(uint32_t));
     uint32_t nTextCount = nNumStrings/2;
@@ -474,7 +474,7 @@ void MicroProfileFloatWindowSize(const char** ppStrings, uint32_t nNumStrings, u
     nHeight = (MICROPROFILE_TEXT_HEIGHT+1) * nTextCount + 2 * MICROPROFILE_BORDER_SIZE;
 }
 
-void MicroProfileDrawFloatWindow(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0)
+inline void MicroProfileDrawFloatWindow(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0)
 {
     uint32_t nWidth = 0, nHeight = 0;
     uint32_t* nStringLengths = (uint32_t*)alloca(nNumStrings * sizeof(uint32_t));
@@ -503,7 +503,7 @@ void MicroProfileDrawFloatWindow(uint32_t nX, uint32_t nY, const char** ppString
         nY += (MICROPROFILE_TEXT_HEIGHT+1);
     }
 }
-void MicroProfileDrawTextBox(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0)
+inline void MicroProfileDrawTextBox(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0)
 {
     uint32_t nWidth = 0, nHeight = 0;
     uint32_t* nStringLengths = (uint32_t*)alloca(nNumStrings * sizeof(uint32_t));
@@ -529,7 +529,7 @@ void MicroProfileDrawTextBox(uint32_t nX, uint32_t nY, const char** ppStrings, u
 
 
 
-void MicroProfileToolTipMeta(MicroProfileStringArray* pToolTip)
+inline void MicroProfileToolTipMeta(MicroProfileStringArray* pToolTip)
 {
     MicroProfile& S = *MicroProfileGet();
     if(UI.nRangeBeginIndex != UI.nRangeEndIndex && UI.pRangeLog)
@@ -608,7 +608,7 @@ void MicroProfileToolTipMeta(MicroProfileStringArray* pToolTip)
     }
 }
 
-void MicroProfileDrawFloatTooltip(uint32_t nX, uint32_t nY, uint32_t nToken, uint64_t nTime)
+inline void MicroProfileDrawFloatTooltip(uint32_t nX, uint32_t nY, uint32_t nToken, uint64_t nTime)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -718,7 +718,7 @@ void MicroProfileDrawFloatTooltip(uint32_t nX, uint32_t nY, uint32_t nToken, uin
 }
 
 
-void MicroProfileZoomTo(int64_t nTickStart, int64_t nTickEnd)
+inline void MicroProfileZoomTo(int64_t nTickStart, int64_t nTickEnd)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -728,7 +728,7 @@ void MicroProfileZoomTo(int64_t nTickStart, int64_t nTickEnd)
     UI.fDetailedRangeTarget = MicroProfileLogTickDifference(nTickStart, nTickEnd) * fToMs;
 }
 
-void MicroProfileCenter(int64_t nTickCenter)
+inline void MicroProfileCenter(int64_t nTickCenter)
 {
     MicroProfile& S = *MicroProfileGet();
     int64_t nStart = S.Frames[S.nFrameCurrent].nFrameStartCpu;
@@ -739,7 +739,7 @@ void MicroProfileCenter(int64_t nTickCenter)
 #ifdef MICROPROFILE_DEBUG
 uint64_t* g_pMicroProfileDumpStart = 0;
 uint64_t* g_pMicroProfileDumpEnd = 0;
-void MicroProfileDebugDumpRange()
+inline void MicroProfileDebugDumpRange()
 {
     MicroProfile& S = *MicroProfileGet();
     if(g_pMicroProfileDumpStart != g_pMicroProfileDumpEnd)
@@ -777,7 +777,7 @@ void MicroProfileDebugDumpRange()
 
 #define MICROPROFILE_HOVER_DIST 0.5f
 
-void MicroProfileDrawDetailedContextSwitchBars(uint32_t nY, uint32_t nThreadId, uint32_t nContextSwitchStart, uint32_t nContextSwitchEnd, int64_t nBaseTicks, uint32_t nBaseY)
+inline void MicroProfileDrawDetailedContextSwitchBars(uint32_t nY, uint32_t nThreadId, uint32_t nContextSwitchStart, uint32_t nContextSwitchEnd, int64_t nBaseTicks, uint32_t nBaseY)
 {
     MicroProfile& S = *MicroProfileGet();
     int64_t nTickIn = -1;
@@ -841,7 +841,7 @@ void MicroProfileDrawDetailedContextSwitchBars(uint32_t nY, uint32_t nThreadId,
     }
 }
 
-void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int nBaseY, int nSelectedFrame)
+inline void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int nBaseY, int nSelectedFrame)
 {
     MicroProfile& S = *MicroProfileGet();
     MP_DEBUG_DUMP_RANGE();
@@ -1325,7 +1325,7 @@ void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int nBaseY,
 }
 
 
-void MicroProfileDrawDetailedFrameHistory(uint32_t nWidth, uint32_t nHeight, uint32_t nBaseY, uint32_t nSelectedFrame)
+inline void MicroProfileDrawDetailedFrameHistory(uint32_t nWidth, uint32_t nHeight, uint32_t nBaseY, uint32_t nSelectedFrame)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -1379,7 +1379,7 @@ void MicroProfileDrawDetailedFrameHistory(uint32_t nWidth, uint32_t nHeight, uin
     }
     MicroProfileDrawBox(fSelectionStart, nBaseY, fSelectionEnd, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, MICROPROFILE_FRAME_HISTORY_COLOR_HIGHTLIGHT, MicroProfileBoxTypeFlat);
 }
-void MicroProfileDrawDetailedView(uint32_t nWidth, uint32_t nHeight)
+inline void MicroProfileDrawDetailedView(uint32_t nWidth, uint32_t nHeight)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -1416,11 +1416,11 @@ void MicroProfileDrawDetailedView(uint32_t nWidth, uint32_t nHeight)
     MicroProfileDrawDetailedFrameHistory(nWidth, nHeight, nBaseY, nSelectedFrame);
 }
 
-void MicroProfileDrawTextRight(uint32_t nX, uint32_t nY, uint32_t nColor, const char* pStr, uint32_t nStrLen)
+inline void MicroProfileDrawTextRight(uint32_t nX, uint32_t nY, uint32_t nColor, const char* pStr, uint32_t nStrLen)
 {
     MicroProfileDrawText(nX - nStrLen * (MICROPROFILE_TEXT_WIDTH+1), nY, nColor, pStr, nStrLen);
 }
-void MicroProfileDrawHeader(int32_t nX, uint32_t nWidth, const char* pName)
+inline void MicroProfileDrawHeader(int32_t nX, uint32_t nWidth, const char* pName)
 {
     if(pName)
     {
@@ -1432,7 +1432,7 @@ void MicroProfileDrawHeader(int32_t nX, uint32_t nWidth, const char* pName)
 
 typedef void (*MicroProfileLoopGroupCallback)(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pData);
 
-void MicroProfileLoopActiveGroupsDraw(int32_t nX, int32_t nY, const char* pName, MicroProfileLoopGroupCallback CB, void* pData)
+inline void MicroProfileLoopActiveGroupsDraw(int32_t nX, int32_t nY, const char* pName, MicroProfileLoopGroupCallback CB, void* pData)
 {
     MicroProfile& S = *MicroProfileGet();
     nY += MICROPROFILE_TEXT_HEIGHT + 2;
@@ -1465,7 +1465,7 @@ void MicroProfileLoopActiveGroupsDraw(int32_t nX, int32_t nY, const char* pName,
 }
 
 
-void MicroProfileCalcTimers(float* pTimers, float* pAverage, float* pMax, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, uint64_t nGroup, uint32_t nSize)
+inline void MicroProfileCalcTimers(float* pTimers, float* pAverage, float* pMax, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, uint64_t nGroup, uint32_t nSize)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -1527,7 +1527,7 @@ void MicroProfileCalcTimers(float* pTimers, float* pAverage, float* pMax, float*
 
 #define SBUF_MAX 32
 
-void MicroProfileDrawBarArrayCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
+inline void MicroProfileDrawBarArrayCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
 {
     const uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT;
     const uint32_t nTextWidth = 6 * (1+MICROPROFILE_TEXT_WIDTH);
@@ -1547,7 +1547,7 @@ void MicroProfileDrawBarArrayCallback(uint32_t nTimer, uint32_t nIdx, uint64_t n
 }
 
 
-uint32_t MicroProfileDrawBarArray(int32_t nX, int32_t nY, float* pTimers, const char* pName, uint32_t nTotalHeight, float* pTimers2 = NULL)
+inline uint32_t MicroProfileDrawBarArray(int32_t nX, int32_t nY, float* pTimers, const char* pName, uint32_t nTotalHeight, float* pTimers2 = NULL)
 {
     const uint32_t nTextWidth = 6 * (1+MICROPROFILE_TEXT_WIDTH);
     const uint32_t nWidth = MICROPROFILE_BAR_WIDTH;
@@ -1559,7 +1559,7 @@ uint32_t MicroProfileDrawBarArray(int32_t nX, int32_t nY, float* pTimers, const
     return nWidth + 5 + nTextWidth;
 
 }
-void MicroProfileDrawBarCallCountCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
+inline void MicroProfileDrawBarCallCountCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
 {
     MicroProfile& S = *MicroProfileGet();
     char sBuffer[SBUF_MAX];
@@ -1567,7 +1567,7 @@ void MicroProfileDrawBarCallCountCallback(uint32_t nTimer, uint32_t nIdx, uint64
     MicroProfileDrawText(nX, nY, (uint32_t)-1, sBuffer, nLen);
 }
 
-uint32_t MicroProfileDrawBarCallCount(int32_t nX, int32_t nY, const char* pName)
+inline uint32_t MicroProfileDrawBarCallCount(int32_t nX, int32_t nY, const char* pName)
 {
     MicroProfileLoopActiveGroupsDraw(nX, nY, pName, MicroProfileDrawBarCallCountCallback, 0);
     const uint32_t nTextWidth = 6 * MICROPROFILE_TEXT_WIDTH;
@@ -1581,7 +1581,7 @@ struct MicroProfileMetaAverageArgs
     float fRcpFrames;
 };
 
-void MicroProfileDrawBarMetaAverageCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
+inline void MicroProfileDrawBarMetaAverageCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
 {
     MicroProfileMetaAverageArgs* pArgs = (MicroProfileMetaAverageArgs*)pExtra;
     uint64_t* pCounters = pArgs->pCounters;
@@ -1591,7 +1591,7 @@ void MicroProfileDrawBarMetaAverageCallback(uint32_t nTimer, uint32_t nIdx, uint
     MicroProfileDrawText(nX - nLen * (MICROPROFILE_TEXT_WIDTH+1), nY, (uint32_t)-1, sBuffer, nLen);
 }
 
-uint32_t MicroProfileDrawBarMetaAverage(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight)
+inline uint32_t MicroProfileDrawBarMetaAverage(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight)
 {
     if(!pName)
         return 0;
@@ -1605,7 +1605,7 @@ uint32_t MicroProfileDrawBarMetaAverage(int32_t nX, int32_t nY, uint64_t* pCount
 }
 
 
-void MicroProfileDrawBarMetaCountCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
+inline void MicroProfileDrawBarMetaCountCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
 {
     uint64_t* pCounters = (uint64_t*)pExtra;
     char sBuffer[SBUF_MAX];
@@ -1613,7 +1613,7 @@ void MicroProfileDrawBarMetaCountCallback(uint32_t nTimer, uint32_t nIdx, uint64
     MicroProfileDrawText(nX - nLen * (MICROPROFILE_TEXT_WIDTH+1), nY, (uint32_t)-1, sBuffer, nLen);
 }
 
-uint32_t MicroProfileDrawBarMetaCount(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight)
+inline uint32_t MicroProfileDrawBarMetaCount(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight)
 {
     if(!pName)
         return 0;
@@ -1625,7 +1625,7 @@ uint32_t MicroProfileDrawBarMetaCount(int32_t nX, int32_t nY, uint64_t* pCounter
     return 5 + nTextWidth;
 }
 
-void MicroProfileDrawBarLegendCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
+inline void MicroProfileDrawBarLegendCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra)
 {
     MicroProfile& S = *MicroProfileGet();
     if (S.TimerInfo[nTimer].bGraph)
@@ -1640,7 +1640,7 @@ void MicroProfileDrawBarLegendCallback(uint32_t nTimer, uint32_t nIdx, uint64_t
     }
 }
 
-uint32_t MicroProfileDrawBarLegend(int32_t nX, int32_t nY, uint32_t nTotalHeight, uint32_t nMaxWidth)
+inline uint32_t MicroProfileDrawBarLegend(int32_t nX, int32_t nY, uint32_t nTotalHeight, uint32_t nMaxWidth)
 {
     MicroProfileDrawLineVertical(nX-5, nY, nTotalHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);
     MicroProfileLoopActiveGroupsDraw(nMaxWidth, nY, 0, MicroProfileDrawBarLegendCallback, 0);
@@ -1807,7 +1807,7 @@ void MicroProfileDumpTimers()
     }
 }
 
-void MicroProfileDrawBarView(uint32_t nScreenWidth, uint32_t nScreenHeight)
+inline void MicroProfileDrawBarView(uint32_t nScreenWidth, uint32_t nScreenHeight)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -1951,7 +1951,7 @@ typedef const char* (*MicroProfileSubmenuCallback)(int, bool* bSelected);
 typedef void (*MicroProfileClickCallback)(int);
 
 
-const char* MicroProfileUIMenuMode(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuMode(int nIndex, bool* bSelected)
 {
     MicroProfile& S = *MicroProfileGet();
     switch(nIndex)
@@ -1979,7 +1979,7 @@ const char* MicroProfileUIMenuMode(int nIndex, bool* bSelected)
     }
 }
 
-const char* MicroProfileUIMenuGroups(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuGroups(int nIndex, bool* bSelected)
 {
     MicroProfile& S = *MicroProfileGet();
     *bSelected = false;
@@ -2012,7 +2012,7 @@ const char* MicroProfileUIMenuGroups(int nIndex, bool* bSelected)
     }
 }
 
-const char* MicroProfileUIMenuAggregate(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuAggregate(int nIndex, bool* bSelected)
 {
     MicroProfile& S = *MicroProfileGet();
     if(nIndex < sizeof(g_MicroProfileAggregatePresets)/sizeof(g_MicroProfileAggregatePresets[0]))
@@ -2032,7 +2032,7 @@ const char* MicroProfileUIMenuAggregate(int nIndex, bool* bSelected)
 
 }
 
-const char* MicroProfileUIMenuTimers(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuTimers(int nIndex, bool* bSelected)
 {
     MicroProfile& S = *MicroProfileGet();
     *bSelected = 0 != (S.nBars & (1 << nIndex));
@@ -2054,7 +2054,7 @@ const char* MicroProfileUIMenuTimers(int nIndex, bool* bSelected)
     return 0;
 }
 
-const char* MicroProfileUIMenuOptions(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuOptions(int nIndex, bool* bSelected)
 {
     MicroProfile& S = *MicroProfileGet();
     if(nIndex >= MICROPROFILE_OPTION_SIZE) return 0;
@@ -2094,7 +2094,7 @@ const char* MicroProfileUIMenuOptions(int nIndex, bool* bSelected)
     return UI.Options[nIndex].Text;
 }
 
-const char* MicroProfileUIMenuPreset(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuPreset(int nIndex, bool* bSelected)
 {
     static char buf[128];
     *bSelected = false;
@@ -2118,7 +2118,7 @@ const char* MicroProfileUIMenuPreset(int nIndex, bool* bSelected)
     }
 }
 
-const char* MicroProfileUIMenuCustom(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuCustom(int nIndex, bool* bSelected)
 {
     if((uint32_t)-1 == UI.nCustomActive)
     {
@@ -2145,13 +2145,13 @@ const char* MicroProfileUIMenuCustom(int nIndex, bool* bSelected)
     }
 }
 
-const char* MicroProfileUIMenuEmpty(int nIndex, bool* bSelected)
+inline const char* MicroProfileUIMenuEmpty(int nIndex, bool* bSelected)
 {
     return 0;
 }
 
 
-void MicroProfileUIClickMode(int nIndex)
+inline void MicroProfileUIClickMode(int nIndex)
 {
     MicroProfile& S = *MicroProfileGet();
     switch(nIndex)
@@ -2176,7 +2176,7 @@ void MicroProfileUIClickMode(int nIndex)
     }
 }
 
-void MicroProfileUIClickGroups(int nIndex)
+inline void MicroProfileUIClickGroups(int nIndex)
 {
     MicroProfile& S = *MicroProfileGet();
     if(nIndex == 0)
@@ -2208,7 +2208,7 @@ void MicroProfileUIClickGroups(int nIndex)
     }
 }
 
-void MicroProfileUIClickAggregate(int nIndex)
+inline void MicroProfileUIClickAggregate(int nIndex)
 {
     MicroProfile& S = *MicroProfileGet();
     S.nAggregateFlip = g_MicroProfileAggregatePresets[nIndex];
@@ -2218,13 +2218,13 @@ void MicroProfileUIClickAggregate(int nIndex)
     }
 }
 
-void MicroProfileUIClickTimers(int nIndex)
+inline void MicroProfileUIClickTimers(int nIndex)
 {
     MicroProfile& S = *MicroProfileGet();
     S.nBars ^= (1 << nIndex);
 }
 
-void MicroProfileUIClickOptions(int nIndex)
+inline void MicroProfileUIClickOptions(int nIndex)
 {
     MicroProfile& S = *MicroProfileGet();
     switch(UI.Options[nIndex].nSubType)
@@ -2271,7 +2271,7 @@ void MicroProfileUIClickOptions(int nIndex)
     }
 }
 
-void MicroProfileUIClickPreset(int nIndex)
+inline void MicroProfileUIClickPreset(int nIndex)
 {
     int nNumPresets = sizeof(g_MicroProfilePresetNames) / sizeof(g_MicroProfilePresetNames[0]);
     int nIndexSave = nIndex - nNumPresets - 1;
@@ -2285,7 +2285,7 @@ void MicroProfileUIClickPreset(int nIndex)
     }
 }
 
-void MicroProfileUIClickCustom(int nIndex)
+inline void MicroProfileUIClickCustom(int nIndex)
 {
     if(nIndex == 0)
     {
@@ -2298,13 +2298,13 @@ void MicroProfileUIClickCustom(int nIndex)
 
 }
 
-void MicroProfileUIClickEmpty(int nIndex)
+inline void MicroProfileUIClickEmpty(int nIndex)
 {
 
 }
 
 
-void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)
+inline void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)
 {
     MicroProfile& S = *MicroProfileGet();
 
@@ -2489,7 +2489,7 @@ void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)
 }
 
 
-void MicroProfileMoveGraph()
+inline void MicroProfileMoveGraph()
 {
 
     int nZoom = UI.nMouseWheelDelta;
@@ -2536,7 +2536,7 @@ void MicroProfileMoveGraph()
         UI.nOffsetY = 0;
 }
 
-void MicroProfileDrawCustom(uint32_t nWidth, uint32_t nHeight)
+inline void MicroProfileDrawCustom(uint32_t nWidth, uint32_t nHeight)
 {
     if((uint32_t)-1 != UI.nCustomActive)
     {
@@ -2633,7 +2633,7 @@ void MicroProfileDrawCustom(uint32_t nWidth, uint32_t nHeight)
         }
     }
 }
-void MicroProfileDraw(uint32_t nWidth, uint32_t nHeight)
+inline void MicroProfileDraw(uint32_t nWidth, uint32_t nHeight)
 {
     MICROPROFILE_SCOPE(g_MicroProfileDraw);
     MicroProfile& S = *MicroProfileGet();
@@ -3226,7 +3226,7 @@ void MicroProfileLoadPreset(const char* pSuffix)
     }
 }
 
-uint32_t MicroProfileCustomGroupFind(const char* pCustomName)
+inline uint32_t MicroProfileCustomGroupFind(const char* pCustomName)
 {
     for(uint32_t i = 0; i < UI.nCustomCount; ++i)
     {
@@ -3238,7 +3238,7 @@ uint32_t MicroProfileCustomGroupFind(const char* pCustomName)
     return (uint32_t)-1;
 }
 
-uint32_t MicroProfileCustomGroup(const char* pCustomName)
+inline uint32_t MicroProfileCustomGroup(const char* pCustomName)
 {
     for(uint32_t i = 0; i < UI.nCustomCount; ++i)
     {
@@ -3271,7 +3271,7 @@ void MicroProfileCustomGroup(const char* pCustomName, uint32_t nMaxTimers, uint3
     UI.Custom[nIndex].nAggregateFlip = nAggregateFlip;
 }
 
-void MicroProfileCustomGroupEnable(uint32_t nIndex)
+inline void MicroProfileCustomGroupEnable(uint32_t nIndex)
 {
     if(nIndex < UI.nCustomCount)
     {

src/CMakeLists.txt

@@ -54,8 +54,10 @@ else()
     add_compile_options(
         -Wall
         -Werror=implicit-fallthrough
+        -Werror=missing-declarations
         -Werror=reorder
         -Wextra
+        -Wmissing-declarations
         -Wno-attributes
         -Wno-unused-parameter
     )

src/common/bit_field.h

@@ -180,7 +180,7 @@ public:
     }
 
     constexpr void Assign(const T& value) {
-        storage = (static_cast<StorageType>(storage) & ~mask) | FormatValue(value);
+        storage = static_cast<StorageType>((storage & ~mask) | FormatValue(value));
     }
 
     constexpr T Value() const {

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -185,10 +185,9 @@ void ARM_Dynarmic_64::Step() {
 
 ARM_Dynarmic_64::ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor,
                                  std::size_t core_index)
-    : ARM_Interface{system},
-      cb(std::make_unique<DynarmicCallbacks64>(*this)), inner_unicorn{system},
-      core_index{core_index}, exclusive_monitor{
-                                  dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
+    : ARM_Interface{system}, cb(std::make_unique<DynarmicCallbacks64>(*this)),
+      inner_unicorn{system, ARM_Unicorn::Arch::AArch64}, core_index{core_index},
+      exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
 
 ARM_Dynarmic_64::~ARM_Dynarmic_64() = default;
 

src/core/arm/unicorn/arm_unicorn.cpp

@@ -62,8 +62,9 @@ static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int si
     return false;
 }
 
-ARM_Unicorn::ARM_Unicorn(System& system) : ARM_Interface{system} {
-    CHECKED(uc_open(UC_ARCH_ARM64, UC_MODE_ARM, &uc));
+ARM_Unicorn::ARM_Unicorn(System& system, Arch architecture) : ARM_Interface{system} {
+    const auto arch = architecture == Arch::AArch32 ? UC_ARCH_ARM : UC_ARCH_ARM64;
+    CHECKED(uc_open(arch, UC_MODE_ARM, &uc));
 
     auto fpv = 3 << 20;
     CHECKED(uc_reg_write(uc, UC_ARM64_REG_CPACR_EL1, &fpv));

src/core/arm/unicorn/arm_unicorn.h

@@ -15,7 +15,12 @@ class System;
 
 class ARM_Unicorn final : public ARM_Interface {
 public:
-    explicit ARM_Unicorn(System& system);
+    enum class Arch {
+        AArch32, // 32-bit ARM
+        AArch64, // 64-bit ARM
+    };
+
+    explicit ARM_Unicorn(System& system, Arch architecture);
     ~ARM_Unicorn() override;
 
     void SetPC(u64 pc) override;

src/core/crypto/partition_data_manager.cpp

@@ -202,8 +202,8 @@ static std::array<Key128, 0x20> FindEncryptedMasterKeyFromHex(const std::vector<
     return out;
 }
 
-FileSys::VirtualFile FindFileInDirWithNames(const FileSys::VirtualDir& dir,
-                                            const std::string& name) {
+static FileSys::VirtualFile FindFileInDirWithNames(const FileSys::VirtualDir& dir,
+                                                   const std::string& name) {
     const auto upper = Common::ToUpper(name);
 
     for (const auto& fname : {name, name + ".bin", upper, upper + ".BIN"}) {
@@ -345,8 +345,7 @@ FileSys::VirtualFile PartitionDataManager::GetPackage2Raw(Package2Type type) con
     return package2.at(static_cast<size_t>(type));
 }
 
-bool AttemptDecrypt(const std::array<u8, 16>& key, Package2Header& header) {
-
+static bool AttemptDecrypt(const std::array<u8, 16>& key, Package2Header& header) {
     const std::vector<u8> iv(header.header_ctr.begin(), header.header_ctr.end());
     Package2Header temp = header;
     AESCipher<Key128> cipher(key, Mode::CTR);

src/core/hle/kernel/memory/memory_block.h

@@ -17,7 +17,7 @@ namespace Kernel::Memory {
 
 enum class MemoryState : u32 {
     None = 0,
-    Mask = 0xFFFFFFFF, // TODO(bunnei): This should probable be 0xFF
+    Mask = 0xFF,
     All = ~None,
 
     FlagCanReprotect = (1 << 8),
@@ -253,6 +253,23 @@ public:
         };
     }
 
+    void ShareToDevice(MemoryPermission /*new_perm*/) {
+        ASSERT((attribute & MemoryAttribute::DeviceShared) == MemoryAttribute::DeviceShared ||
+               device_use_count == 0);
+        attribute |= MemoryAttribute::DeviceShared;
+        const u16 new_use_count{++device_use_count};
+        ASSERT(new_use_count > 0);
+    }
+
+    void UnshareToDevice(MemoryPermission /*new_perm*/) {
+        ASSERT((attribute & MemoryAttribute::DeviceShared) == MemoryAttribute::DeviceShared);
+        const u16 prev_use_count{device_use_count--};
+        ASSERT(prev_use_count > 0);
+        if (prev_use_count == 1) {
+            attribute &= ~MemoryAttribute::DeviceShared;
+        }
+    }
+
 private:
     constexpr bool HasProperties(MemoryState s, MemoryPermission p, MemoryAttribute a) const {
         constexpr MemoryAttribute AttributeIgnoreMask{MemoryAttribute::DontCareMask |
@@ -287,9 +304,9 @@ private:
         state = new_state;
         perm = new_perm;
 
-        // TODO(bunnei): Is this right?
         attribute = static_cast<MemoryAttribute>(
-            new_attribute /*| (attribute & (MemoryAttribute::IpcLocked | MemoryAttribute::DeviceShared))*/);
+            new_attribute |
+            (attribute & (MemoryAttribute::IpcLocked | MemoryAttribute::DeviceShared)));
     }
 
     constexpr MemoryBlock Split(VAddr split_addr) {

src/core/hle/kernel/memory/memory_block_manager.cpp

@@ -143,6 +143,42 @@ void MemoryBlockManager::Update(VAddr addr, std::size_t num_pages, MemoryState s
     }
 }
 
+void MemoryBlockManager::UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func,
+                                    MemoryPermission perm) {
+    const std::size_t prev_count{memory_block_tree.size()};
+    const VAddr end_addr{addr + num_pages * PageSize};
+    iterator node{memory_block_tree.begin()};
+
+    while (node != memory_block_tree.end()) {
+        MemoryBlock* block{&(*node)};
+        iterator next_node{std::next(node)};
+        const VAddr cur_addr{block->GetAddress()};
+        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
+
+        if (addr < cur_end_addr && cur_addr < end_addr) {
+            iterator new_node{node};
+
+            if (addr > cur_addr) {
+                memory_block_tree.insert(node, block->Split(addr));
+            }
+
+            if (end_addr < cur_end_addr) {
+                new_node = memory_block_tree.insert(node, block->Split(end_addr));
+            }
+
+            lock_func(new_node, perm);
+
+            MergeAdjacent(new_node, next_node);
+        }
+
+        if (cur_end_addr - 1 >= end_addr - 1) {
+            break;
+        }
+
+        node = next_node;
+    }
+}
+
 void MemoryBlockManager::IterateForRange(VAddr start, VAddr end, IterateFunc&& func) {
     const_iterator it{FindIterator(start)};
     MemoryInfo info{};

src/core/hle/kernel/memory/memory_block_manager.h

@@ -45,6 +45,9 @@ public:
                 MemoryPermission perm = MemoryPermission::None,
                 MemoryAttribute attribute = MemoryAttribute::None);
 
+    using LockFunc = std::function<void(iterator, MemoryPermission)>;
+    void UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func, MemoryPermission perm);
+
     using IterateFunc = std::function<void(const MemoryInfo&)>;
     void IterateForRange(VAddr start, VAddr end, IterateFunc&& func);
 

src/core/hle/kernel/memory/page_table.cpp

@@ -840,6 +840,50 @@ ResultVal<VAddr> PageTable::AllocateAndMapMemory(std::size_t needed_num_pages, s
     return MakeResult<VAddr>(addr);
 }
 
+ResultCode PageTable::LockForDeviceAddressSpace(VAddr addr, std::size_t size) {
+    std::lock_guard lock{page_table_lock};
+
+    MemoryPermission perm{};
+    if (const ResultCode result{CheckMemoryState(
+            nullptr, &perm, nullptr, addr, size, MemoryState::FlagCanChangeAttribute,
+            MemoryState::FlagCanChangeAttribute, MemoryPermission::None, MemoryPermission::None,
+            MemoryAttribute::LockedAndIpcLocked, MemoryAttribute::None,
+            MemoryAttribute::DeviceSharedAndUncached)};
+        result.IsError()) {
+        return result;
+    }
+
+    block_manager->UpdateLock(addr, size / PageSize,
+                              [](MemoryBlockManager::iterator block, MemoryPermission perm) {
+                                  block->ShareToDevice(perm);
+                              },
+                              perm);
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode PageTable::UnlockForDeviceAddressSpace(VAddr addr, std::size_t size) {
+    std::lock_guard lock{page_table_lock};
+
+    MemoryPermission perm{};
+    if (const ResultCode result{CheckMemoryState(
+            nullptr, &perm, nullptr, addr, size, MemoryState::FlagCanChangeAttribute,
+            MemoryState::FlagCanChangeAttribute, MemoryPermission::None, MemoryPermission::None,
+            MemoryAttribute::LockedAndIpcLocked, MemoryAttribute::None,
+            MemoryAttribute::DeviceSharedAndUncached)};
+        result.IsError()) {
+        return result;
+    }
+
+    block_manager->UpdateLock(addr, size / PageSize,
+                              [](MemoryBlockManager::iterator block, MemoryPermission perm) {
+                                  block->UnshareToDevice(perm);
+                              },
+                              perm);
+
+    return RESULT_SUCCESS;
+}
+
 ResultCode PageTable::InitializeMemoryLayout(VAddr start, VAddr end) {
     block_manager = std::make_unique<MemoryBlockManager>(start, end);
 

src/core/hle/kernel/memory/page_table.h

@@ -53,6 +53,8 @@ public:
                                           bool is_map_only, VAddr region_start,
                                           std::size_t region_num_pages, MemoryState state,
                                           MemoryPermission perm, PAddr map_addr = 0);
+    ResultCode LockForDeviceAddressSpace(VAddr addr, std::size_t size);
+    ResultCode UnlockForDeviceAddressSpace(VAddr addr, std::size_t size);
 
     Common::PageTable& PageTableImpl() {
         return page_table_impl;

src/core/hle/kernel/physical_core.cpp

@@ -27,7 +27,9 @@ PhysicalCore::PhysicalCore(Core::System& system, std::size_t id,
         std::make_unique<Core::ARM_Dynarmic_64>(system, exclusive_monitor, core_index);
 
 #else
-    arm_interface = std::make_shared<Core::ARM_Unicorn>(system);
+    using Core::ARM_Unicorn;
+    arm_interface_32 = std::make_unique<ARM_Unicorn>(system, ARM_Unicorn::Arch::AArch32);
+    arm_interface_64 = std::make_unique<ARM_Unicorn>(system, ARM_Unicorn::Arch::AArch64);
     LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
 

src/core/hle/kernel/shared_memory.cpp

@@ -36,22 +36,22 @@ std::shared_ptr<SharedMemory> SharedMemory::Create(
 }
 
 ResultCode SharedMemory::Map(Process& target_process, VAddr address, std::size_t size,
-                             Memory::MemoryPermission permission) {
+                             Memory::MemoryPermission permissions) {
     const u64 page_count{(size + Memory::PageSize - 1) / Memory::PageSize};
 
     if (page_list.GetNumPages() != page_count) {
         UNIMPLEMENTED_MSG("Page count does not match");
     }
 
-    Memory::MemoryPermission expected =
+    const Memory::MemoryPermission expected =
         &target_process == owner_process ? owner_permission : user_permission;
 
-    if (permission != expected) {
+    if (permissions != expected) {
         UNIMPLEMENTED_MSG("Permission does not match");
     }
 
     return target_process.PageTable().MapPages(address, page_list, Memory::MemoryState::Shared,
-                                               permission);
+                                               permissions);
 }
 
 } // namespace Kernel

src/core/hle/kernel/shared_memory.h

@@ -51,7 +51,7 @@ public:
      * @param permissions Memory block map permissions (specified by SVC field)
      */
     ResultCode Map(Process& target_process, VAddr address, std::size_t size,
-                   Memory::MemoryPermission permission);
+                   Memory::MemoryPermission permissions);
 
     /**
      * Gets a pointer to the shared memory block

src/core/hle/kernel/svc.cpp

@@ -55,9 +55,6 @@ constexpr bool IsValidAddressRange(VAddr address, u64 size) {
     return address + size > address;
 }
 
-// 8 GiB
-constexpr u64 MAIN_MEMORY_SIZE = 0x200000000;
-
 // Helper function that performs the common sanity checks for svcMapMemory
 // and svcUnmapMemory. This is doable, as both functions perform their sanitizing
 // in the same order.
@@ -776,7 +773,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
             break;
         }
 
-        LOG_WARNING(Kernel_SVC, "(STUBBED) Unimplemented svcGetInfo id=0x{:016X}", info_id);
+        LOG_ERROR(Kernel_SVC, "Unimplemented svcGetInfo id=0x{:016X}", info_id);
         return ERR_INVALID_ENUM_VALUE;
     }
 
@@ -869,7 +866,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
     }
 
     default:
-        LOG_WARNING(Kernel_SVC, "(STUBBED) Unimplemented svcGetInfo id=0x{:016X}", info_id);
+        LOG_ERROR(Kernel_SVC, "Unimplemented svcGetInfo id=0x{:016X}", info_id);
         return ERR_INVALID_ENUM_VALUE;
     }
 }
@@ -1229,6 +1226,142 @@ static ResultCode QueryMemory32(Core::System& system, u32 memory_info_address,
     return QueryMemory(system, memory_info_address, page_info_address, query_address);
 }
 
+static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_handle, u64 dst_address,
+                                       u64 src_address, u64 size) {
+    LOG_DEBUG(Kernel_SVC,
+              "called. process_handle=0x{:08X}, dst_address=0x{:016X}, "
+              "src_address=0x{:016X}, size=0x{:016X}",
+              process_handle, dst_address, src_address, size);
+
+    if (!Common::Is4KBAligned(src_address)) {
+        LOG_ERROR(Kernel_SVC, "src_address is not page-aligned (src_address=0x{:016X}).",
+                  src_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (!Common::Is4KBAligned(dst_address)) {
+        LOG_ERROR(Kernel_SVC, "dst_address is not page-aligned (dst_address=0x{:016X}).",
+                  dst_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (size == 0 || !Common::Is4KBAligned(size)) {
+        LOG_ERROR(Kernel_SVC, "Size is zero or not page-aligned (size=0x{:016X})", size);
+        return ERR_INVALID_SIZE;
+    }
+
+    if (!IsValidAddressRange(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range overflows the address space (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsValidAddressRange(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range overflows the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
+    auto process = handle_table.Get<Process>(process_handle);
+    if (!process) {
+        LOG_ERROR(Kernel_SVC, "Invalid process handle specified (handle=0x{:08X}).",
+                  process_handle);
+        return ERR_INVALID_HANDLE;
+    }
+
+    auto& page_table = process->PageTable();
+    if (!page_table.IsInsideAddressSpace(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range is not within the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!page_table.IsInsideASLRRegion(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range is not within the ASLR region (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    return page_table.MapProcessCodeMemory(dst_address, src_address, size);
+}
+
+static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_handle,
+                                         u64 dst_address, u64 src_address, u64 size) {
+    LOG_DEBUG(Kernel_SVC,
+              "called. process_handle=0x{:08X}, dst_address=0x{:016X}, src_address=0x{:016X}, "
+              "size=0x{:016X}",
+              process_handle, dst_address, src_address, size);
+
+    if (!Common::Is4KBAligned(dst_address)) {
+        LOG_ERROR(Kernel_SVC, "dst_address is not page-aligned (dst_address=0x{:016X}).",
+                  dst_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (!Common::Is4KBAligned(src_address)) {
+        LOG_ERROR(Kernel_SVC, "src_address is not page-aligned (src_address=0x{:016X}).",
+                  src_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (size == 0 || Common::Is4KBAligned(size)) {
+        LOG_ERROR(Kernel_SVC, "Size is zero or not page-aligned (size=0x{:016X}).", size);
+        return ERR_INVALID_SIZE;
+    }
+
+    if (!IsValidAddressRange(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range overflows the address space (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsValidAddressRange(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range overflows the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
+    auto process = handle_table.Get<Process>(process_handle);
+    if (!process) {
+        LOG_ERROR(Kernel_SVC, "Invalid process handle specified (handle=0x{:08X}).",
+                  process_handle);
+        return ERR_INVALID_HANDLE;
+    }
+
+    auto& page_table = process->PageTable();
+    if (!page_table.IsInsideAddressSpace(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range is not within the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!page_table.IsInsideASLRRegion(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range is not within the ASLR region (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    return page_table.UnmapProcessCodeMemory(dst_address, src_address, size);
+}
+
 /// Exits the current process
 static void ExitProcess(Core::System& system) {
     auto* current_process = system.Kernel().CurrentProcess();
@@ -2256,8 +2389,8 @@ static const FunctionDef SVC_Table_64[] = {
     {0x74, nullptr, "MapProcessMemory"},
     {0x75, nullptr, "UnmapProcessMemory"},
     {0x76, SvcWrap64<QueryProcessMemory>, "QueryProcessMemory"},
-    {0x77, nullptr, "MapProcessCodeMemory"},
-    {0x78, nullptr, "UnmapProcessCodeMemory"},
+    {0x77, SvcWrap64<MapProcessCodeMemory>, "MapProcessCodeMemory"},
+    {0x78, SvcWrap64<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
     {0x79, nullptr, "CreateProcess"},
     {0x7A, nullptr, "StartProcess"},
     {0x7B, nullptr, "TerminateProcess"},

src/core/hle/service/acc/acc_su.cpp

@@ -33,8 +33,10 @@ ACC_SU::ACC_SU(std::shared_ptr<Module> module, std::shared_ptr<ProfileManager> p
         {111, nullptr, "ClearSaveDataThumbnail"},
         {112, nullptr, "LoadSaveDataThumbnail"},
         {113, nullptr, "GetSaveDataThumbnailExistence"},
+        {120, nullptr, "ListOpenUsersInApplication"},
         {130, nullptr, "ActivateOpenContextRetention"},
         {140, nullptr, "ListQualifiedUsers"},
+        {150, nullptr, "AuthenticateApplicationAsync"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
         {200, nullptr, "BeginUserRegistration"},

src/core/hle/service/acc/acc_u1.cpp

@@ -35,6 +35,7 @@ ACC_U1::ACC_U1(std::shared_ptr<Module> module, std::shared_ptr<ProfileManager> p
         {113, nullptr, "GetSaveDataThumbnailExistence"},
         {130, nullptr, "ActivateOpenContextRetention"},
         {140, nullptr, "ListQualifiedUsers"},
+        {150, nullptr, "AuthenticateApplicationAsync"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
         {997, nullptr, "DebugInvalidateTokenCacheForUser"},

src/core/hle/service/am/am.cpp

@@ -43,9 +43,9 @@
 
 namespace Service::AM {
 
-constexpr ResultCode ERR_NO_DATA_IN_CHANNEL{ErrorModule::AM, 0x2};
-constexpr ResultCode ERR_NO_MESSAGES{ErrorModule::AM, 0x3};
-constexpr ResultCode ERR_SIZE_OUT_OF_BOUNDS{ErrorModule::AM, 0x1F7};
+constexpr ResultCode ERR_NO_DATA_IN_CHANNEL{ErrorModule::AM, 2};
+constexpr ResultCode ERR_NO_MESSAGES{ErrorModule::AM, 3};
+constexpr ResultCode ERR_SIZE_OUT_OF_BOUNDS{ErrorModule::AM, 503};
 
 enum class LaunchParameterKind : u32 {
     ApplicationSpecific = 1,
@@ -235,6 +235,7 @@ IDebugFunctions::IDebugFunctions() : ServiceFramework{"IDebugFunctions"} {
         {30, nullptr, "RequestLaunchApplicationWithUserAndArgumentForDebug"},
         {40, nullptr, "GetAppletResourceUsageInfo"},
         {100, nullptr, "SetCpuBoostModeForApplet"},
+        {101, nullptr, "CancelCpuBoostModeForApplet"},
         {110, nullptr, "PushToAppletBoundChannelForDebug"},
         {111, nullptr, "TryPopFromAppletBoundChannelForDebug"},
         {120, nullptr, "AlarmSettingNotificationEnableAppEventReserve"},
@@ -277,6 +278,8 @@ ISelfController::ISelfController(Core::System& system,
         {41, nullptr, "IsSystemBufferSharingEnabled"},
         {42, nullptr, "GetSystemSharedLayerHandle"},
         {43, nullptr, "GetSystemSharedBufferHandle"},
+        {44, nullptr, "CreateManagedDisplaySeparableLayer"},
+        {45, nullptr, "SetManagedDisplayLayerSeparationMode"},
         {50, &ISelfController::SetHandlesRequestToDisplay, "SetHandlesRequestToDisplay"},
         {51, nullptr, "ApproveToDisplay"},
         {60, nullptr, "OverrideAutoSleepTimeAndDimmingTime"},
@@ -623,11 +626,15 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system,
         {64, nullptr, "SetTvPowerStateMatchingMode"},
         {65, nullptr, "GetApplicationIdByContentActionName"},
         {66, &ICommonStateGetter::SetCpuBoostMode, "SetCpuBoostMode"},
+        {67, nullptr, "CancelCpuBoostMode"},
         {80, nullptr, "PerformSystemButtonPressingIfInFocus"},
         {90, nullptr, "SetPerformanceConfigurationChangedNotification"},
         {91, nullptr, "GetCurrentPerformanceConfiguration"},
+        {100, nullptr, "SetHandlingHomeButtonShortPressedEnabled"},
         {200, nullptr, "GetOperationModeSystemInfo"},
         {300, nullptr, "GetSettingsPlatformRegion"},
+        {400, nullptr, "ActivateMigrationService"},
+        {401, nullptr, "DeactivateMigrationService"},
     };
     // clang-format on
 
@@ -835,6 +842,7 @@ public:
             {25, nullptr, "Terminate"},
             {30, &ILibraryAppletAccessor::GetResult, "GetResult"},
             {50, nullptr, "SetOutOfFocusApplicationSuspendingEnabled"},
+            {60, nullptr, "PresetLibraryAppletGpuTimeSliceZero"},
             {100, &ILibraryAppletAccessor::PushInData, "PushInData"},
             {101, &ILibraryAppletAccessor::PopOutData, "PopOutData"},
             {102, nullptr, "PushExtraStorage"},
@@ -1139,6 +1147,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {31, &IApplicationFunctions::EndBlockingHomeButtonShortAndLongPressed, "EndBlockingHomeButtonShortAndLongPressed"},
         {32, &IApplicationFunctions::BeginBlockingHomeButton, "BeginBlockingHomeButton"},
         {33, &IApplicationFunctions::EndBlockingHomeButton, "EndBlockingHomeButton"},
+        {34, nullptr, "SelectApplicationLicense"},
         {40, &IApplicationFunctions::NotifyRunning, "NotifyRunning"},
         {50, &IApplicationFunctions::GetPseudoDeviceId, "GetPseudoDeviceId"},
         {60, nullptr, "SetMediaPlaybackStateForApplication"},
@@ -1148,6 +1157,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {68, nullptr, "RequestFlushGamePlayingMovieForDebug"},
         {70, nullptr, "RequestToShutdown"},
         {71, nullptr, "RequestToReboot"},
+        {72, nullptr, "RequestToSleep"},
         {80, nullptr, "ExitAndRequestToShowThanksMessage"},
         {90, &IApplicationFunctions::EnableApplicationCrashReport, "EnableApplicationCrashReport"},
         {100, &IApplicationFunctions::InitializeApplicationCopyrightFrameBuffer, "InitializeApplicationCopyrightFrameBuffer"},

src/core/hle/service/audio/audctl.cpp

@@ -39,6 +39,8 @@ AudCtl::AudCtl() : ServiceFramework{"audctl"} {
         {25, nullptr, "GetAudioVolumeDataForPlayReport"},
         {26, nullptr, "UpdateHeadphoneSettings"},
         {27, nullptr, "SetVolumeMappingTableForDev"},
+        {28, nullptr, "GetAudioOutputChannelCountForPlayReport"},
+        {29, nullptr, "BindAudioOutputChannelCountUpdateEventForPlayReport"},
     };
     // clang-format on
 

src/core/hle/service/bcat/backend/boxcat.cpp

@@ -18,6 +18,7 @@
 #include "core/hle/service/bcat/backend/boxcat.h"
 #include "core/settings.h"
 
+namespace Service::BCAT {
 namespace {
 
 // Prevents conflicts with windows macro called CreateFile
@@ -30,10 +31,6 @@ bool VfsDeleteFileWrap(FileSys::VirtualDir dir, std::string_view name) {
     return dir->DeleteFile(name);
 }
 
-} // Anonymous namespace
-
-namespace Service::BCAT {
-
 constexpr ResultCode ERROR_GENERAL_BCAT_FAILURE{ErrorModule::BCAT, 1};
 
 constexpr char BOXCAT_HOSTNAME[] = "api.yuzu-emu.org";
@@ -90,8 +87,6 @@ constexpr u32 PORT = 443;
 constexpr u32 TIMEOUT_SECONDS = 30;
 [[maybe_unused]] constexpr u64 VFS_COPY_BLOCK_SIZE = 1ULL << 24; // 4MB
 
-namespace {
-
 std::string GetBINFilePath(u64 title_id) {
     return fmt::format("{}bcat/{:016X}/launchparam.bin",
                        FileUtil::GetUserPath(FileUtil::UserPath::CacheDir), title_id);

src/core/hle/service/bcat/module.cpp

@@ -141,6 +141,7 @@ public:
             {20301, nullptr, "RequestSuspendDeliveryTask"},
             {20400, nullptr, "RegisterSystemApplicationDeliveryTask"},
             {20401, nullptr, "UnregisterSystemApplicationDeliveryTask"},
+            {20410, nullptr, "SetSystemApplicationDeliveryTaskTimer"},
             {30100, &IBcatService::SetPassphrase, "SetPassphrase"},
             {30200, nullptr, "RegisterBackgroundDeliveryTask"},
             {30201, nullptr, "UnregisterBackgroundDeliveryTask"},

src/core/hle/service/es/es.cpp

@@ -4,6 +4,7 @@
 
 #include "core/crypto/key_manager.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/service/es/es.h"
 #include "core/hle/service/service.h"
 
 namespace Service::ES {

src/core/hle/service/filesystem/fsp_srv.cpp

@@ -316,8 +316,8 @@ public:
             {8, &IFileSystem::OpenFile, "OpenFile"},
             {9, &IFileSystem::OpenDirectory, "OpenDirectory"},
             {10, &IFileSystem::Commit, "Commit"},
-            {11, nullptr, "GetFreeSpaceSize"},
-            {12, nullptr, "GetTotalSpaceSize"},
+            {11, &IFileSystem::GetFreeSpaceSize, "GetFreeSpaceSize"},
+            {12, &IFileSystem::GetTotalSpaceSize, "GetTotalSpaceSize"},
             {13, &IFileSystem::CleanDirectoryRecursively, "CleanDirectoryRecursively"},
             {14, nullptr, "GetFileTimeStampRaw"},
             {15, nullptr, "QueryEntry"},
@@ -697,12 +697,14 @@ FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter)
         {68, nullptr, "OpenSaveDataInfoReaderBySaveDataFilter"},
         {69, nullptr, "ReadSaveDataFileSystemExtraDataBySaveDataAttribute"},
         {70, nullptr, "WriteSaveDataFileSystemExtraDataBySaveDataAttribute"},
+        {71, nullptr, "ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute"},
         {80, nullptr, "OpenSaveDataMetaFile"},
         {81, nullptr, "OpenSaveDataTransferManager"},
         {82, nullptr, "OpenSaveDataTransferManagerVersion2"},
         {83, nullptr, "OpenSaveDataTransferProhibiterForCloudBackUp"},
         {84, nullptr, "ListApplicationAccessibleSaveDataOwnerId"},
         {85, nullptr, "OpenSaveDataTransferManagerForSaveDataRepair"},
+        {86, nullptr, "OpenSaveDataMover"},
         {100, nullptr, "OpenImageDirectoryFileSystem"},
         {110, nullptr, "OpenContentStorageFileSystem"},
         {120, nullptr, "OpenCloudBackupWorkStorageFileSystem"},
@@ -762,9 +764,11 @@ FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter)
         {1011, &FSP_SRV::GetAccessLogVersionInfo, "GetAccessLogVersionInfo"},
         {1012, nullptr, "GetFsStackUsage"},
         {1013, nullptr, "UnsetSaveDataRootPath"},
+        {1014, nullptr, "OutputMultiProgramTagAccessLog"},
         {1100, nullptr, "OverrideSaveDataTransferTokenSignVerificationKey"},
         {1110, nullptr, "CorruptSaveDataFileSystemBySaveDataSpaceId2"},
         {1200, nullptr, "OpenMultiCommitManager"},
+        {1300, nullptr, "OpenBisWiper"},
     };
     // clang-format on
     RegisterHandlers(functions);

src/core/hle/service/friend/friend.cpp

@@ -96,6 +96,7 @@ public:
             {30830, nullptr, "ClearPlayLog"},
             {30900, nullptr, "SendFriendInvitation"},
             {30910, nullptr, "ReadFriendInvitation"},
+            {30911, nullptr, "ReadAllFriendInvitations"},
             {49900, nullptr, "DeleteNetworkServiceAccountCache"},
         };
         // clang-format on

src/core/hle/service/glue/errors.h

@@ -8,9 +8,9 @@
 
 namespace Service::Glue {
 
-constexpr ResultCode ERR_INVALID_RESOURCE{ErrorModule::ARP, 0x1E};
-constexpr ResultCode ERR_INVALID_PROCESS_ID{ErrorModule::ARP, 0x1F};
-constexpr ResultCode ERR_INVALID_ACCESS{ErrorModule::ARP, 0x2A};
-constexpr ResultCode ERR_NOT_REGISTERED{ErrorModule::ARP, 0x66};
+constexpr ResultCode ERR_INVALID_RESOURCE{ErrorModule::ARP, 30};
+constexpr ResultCode ERR_INVALID_PROCESS_ID{ErrorModule::ARP, 31};
+constexpr ResultCode ERR_INVALID_ACCESS{ErrorModule::ARP, 42};
+constexpr ResultCode ERR_NOT_REGISTERED{ErrorModule::ARP, 102};
 
 } // namespace Service::Glue

src/core/hle/service/hid/hid.cpp

@@ -233,7 +233,7 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {302, nullptr, "StopConsoleSixAxisSensor"},
         {303, nullptr, "ActivateSevenSixAxisSensor"},
         {304, nullptr, "StartSevenSixAxisSensor"},
-        {305, nullptr, "StopSevenSixAxisSensor"},
+        {305, &Hid::StopSevenSixAxisSensor, "StopSevenSixAxisSensor"},
         {306, &Hid::InitializeSevenSixAxisSensor, "InitializeSevenSixAxisSensor"},
         {307, nullptr, "FinalizeSevenSixAxisSensor"},
         {308, nullptr, "SetSevenSixAxisSensorFusionStrength"},
@@ -282,6 +282,7 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {1001, nullptr, "GetNpadCommunicationMode"},
         {1002, nullptr, "SetTouchScreenConfiguration"},
         {1003, nullptr, "IsFirmwareUpdateNeededForNotification"},
+        {2000, nullptr, "ActivateDigitizer"},
     };
     // clang-format on
 
@@ -852,6 +853,17 @@ void Hid::SetPalmaBoostMode(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
+void Hid::StopSevenSixAxisSensor(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID, "(STUBBED) called, applet_resource_user_id={}",
+                applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void Hid::InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_HID, "(STUBBED) called");
 
@@ -870,6 +882,7 @@ public:
             {10, nullptr, "DeactivateTouchScreen"},
             {11, nullptr, "SetTouchScreenAutoPilotState"},
             {12, nullptr, "UnsetTouchScreenAutoPilotState"},
+            {13, nullptr, "GetTouchScreenConfiguration"},
             {20, nullptr, "DeactivateMouse"},
             {21, nullptr, "SetMouseAutoPilotState"},
             {22, nullptr, "UnsetMouseAutoPilotState"},
@@ -879,7 +892,9 @@ public:
             {50, nullptr, "DeactivateXpad"},
             {51, nullptr, "SetXpadAutoPilotState"},
             {52, nullptr, "UnsetXpadAutoPilotState"},
-            {60, nullptr, "DeactivateJoyXpad"},
+            {60, nullptr, "ClearNpadSystemCommonPolicy"},
+            {61, nullptr, "DeactivateNpad"},
+            {62, nullptr, "ForceDisconnectNpad"},
             {91, nullptr, "DeactivateGesture"},
             {110, nullptr, "DeactivateHomeButton"},
             {111, nullptr, "SetHomeButtonAutoPilotState"},
@@ -899,6 +914,15 @@ public:
             {141, nullptr, "GetConsoleSixAxisSensorSamplingFrequency"},
             {142, nullptr, "DeactivateSevenSixAxisSensor"},
             {143, nullptr, "GetConsoleSixAxisSensorCountStates"},
+            {144, nullptr, "GetAccelerometerFsr"},
+            {145, nullptr, "SetAccelerometerFsr"},
+            {146, nullptr, "GetAccelerometerOdr"},
+            {147, nullptr, "SetAccelerometerOdr"},
+            {148, nullptr, "GetGyroscopeFsr"},
+            {149, nullptr, "SetGyroscopeFsr"},
+            {150, nullptr, "GetGyroscopeOdr"},
+            {151, nullptr, "SetGyroscopeOdr"},
+            {152, nullptr, "GetWhoAmI"},
             {201, nullptr, "ActivateFirmwareUpdate"},
             {202, nullptr, "DeactivateFirmwareUpdate"},
             {203, nullptr, "StartFirmwareUpdate"},
@@ -927,6 +951,17 @@ public:
             {233, nullptr, "ClearPairingInfo"},
             {234, nullptr, "GetUniquePadDeviceTypeSetInternal"},
             {235, nullptr, "EnableAnalogStickPower"},
+            {236, nullptr, "RequestKuinaUartClockCal"},
+            {237, nullptr, "GetKuinaUartClockCal"},
+            {238, nullptr, "SetKuinaUartClockTrim"},
+            {239, nullptr, "KuinaLoopbackTest"},
+            {240, nullptr, "RequestBatteryVoltage"},
+            {241, nullptr, "GetBatteryVoltage"},
+            {242, nullptr, "GetUniquePadPowerInfo"},
+            {243, nullptr, "RebootUniquePad"},
+            {244, nullptr, "RequestKuinaFirmwareVersion"},
+            {245, nullptr, "GetKuinaFirmwareVersion"},
+            {246, nullptr, "GetVidPid"},
             {301, nullptr, "GetAbstractedPadHandles"},
             {302, nullptr, "GetAbstractedPadState"},
             {303, nullptr, "GetAbstractedPadsState"},
@@ -945,6 +980,17 @@ public:
             {350, nullptr, "AddRegisteredDevice"},
             {400, nullptr, "DisableExternalMcuOnNxDevice"},
             {401, nullptr, "DisableRailDeviceFiltering"},
+            {402, nullptr, "EnableWiredPairing"},
+            {403, nullptr, "EnableShipmentModeAutoClear"},
+            {500, nullptr, "SetFactoryInt"},
+            {501, nullptr, "IsFactoryBootEnabled"},
+            {550, nullptr, "SetAnalogStickModelDataTemporarily"},
+            {551, nullptr, "GetAnalogStickModelData"},
+            {552, nullptr, "ResetAnalogStickModelData"},
+            {600, nullptr, "ConvertPadState"},
+            {2000, nullptr, "DeactivateDigitizer"},
+            {2001, nullptr, "SetDigitizerAutoPilotState"},
+            {2002, nullptr, "UnsetDigitizerAutoPilotState"},
         };
         // clang-format on
 

src/core/hle/service/hid/hid.h

@@ -128,6 +128,7 @@ private:
     void StopSixAxisSensor(Kernel::HLERequestContext& ctx);
     void SetIsPalmaAllConnectable(Kernel::HLERequestContext& ctx);
     void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
+    void StopSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
     void InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<IAppletResource> applet_resource;

src/core/hle/service/ldr/ldr.cpp

@@ -116,6 +116,7 @@ public:
             {1, nullptr, "GetProgramInfo"},
             {2, nullptr, "RegisterTitle"},
             {3, nullptr, "UnregisterTitle"},
+            {4, nullptr, "SetEnabledProgramVerification"},
         };
         // clang-format on
 

src/core/hle/service/ncm/ncm.cpp

@@ -122,6 +122,7 @@ public:
             {11, nullptr, "ActivateContentMetaDatabase"},
             {12, nullptr, "InactivateContentMetaDatabase"},
             {13, nullptr, "InvalidateRightsIdCache"},
+            {14, nullptr, "GetMemoryReport"},
         };
         // clang-format on
 

src/core/hle/service/npns/npns.cpp

@@ -48,6 +48,8 @@ public:
             {151, nullptr, "GetStateWithHandover"},
             {152, nullptr, "GetStateChangeEventWithHandover"},
             {153, nullptr, "GetDropEventWithHandover"},
+            {161, nullptr, "GetRequestChangeStateCancelEvent"},
+            {162, nullptr, "RequestChangeStateForceTimedWithCancelEvent"},
             {201, nullptr, "RequestChangeStateForceTimed"},
             {202, nullptr, "RequestChangeStateForceAsync"},
         };

src/core/hle/service/ns/ns.cpp

@@ -110,6 +110,10 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {100, nullptr, "ResetToFactorySettings"},
         {101, nullptr, "ResetToFactorySettingsWithoutUserSaveData"},
         {102, nullptr, "ResetToFactorySettingsForRefurbishment"},
+        {103, nullptr, "ResetToFactorySettingsWithPlatformRegion"},
+        {104, nullptr, "ResetToFactorySettingsWithPlatformRegionAuthentication"},
+        {105, nullptr, "RequestResetToFactorySettingsSecurely"},
+        {106, nullptr, "RequestResetToFactorySettingsWithPlatformRegionAuthenticationSecurely"},
         {200, nullptr, "CalculateUserSaveDataStatistics"},
         {201, nullptr, "DeleteUserSaveDataAll"},
         {210, nullptr, "DeleteUserSystemSaveData"},
@@ -191,6 +195,9 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {1307, nullptr, "TryDeleteRunningApplicationContentEntities"},
         {1308, nullptr, "DeleteApplicationCompletelyForDebug"},
         {1309, nullptr, "CleanupUnavailableAddOnContents"},
+        {1310, nullptr, "RequestMoveApplicationEntity"},
+        {1311, nullptr, "EstimateSizeToMove"},
+        {1312, nullptr, "HasMovableEntity"},
         {1400, nullptr, "PrepareShutdown"},
         {1500, nullptr, "FormatSdCard"},
         {1501, nullptr, "NeedsSystemUpdateToFormatSdCard"},
@@ -241,7 +248,7 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {2153, nullptr, "DeactivateRightsEnvironment"},
         {2154, nullptr, "ForceActivateRightsContextForExit"},
         {2155, nullptr, "UpdateRightsEnvironmentStatus"},
-        {2156, nullptr, "CreateRightsEnvironmentForPreomia"},
+        {2156, nullptr, "CreateRightsEnvironmentForMicroApplication"},
         {2160, nullptr, "AddTargetApplicationToRightsEnvironment"},
         {2161, nullptr, "SetUsersToRightsEnvironment"},
         {2170, nullptr, "GetRightsEnvironmentStatus"},
@@ -258,6 +265,7 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {2350, nullptr, "PerformAutoUpdateByApplicationId"},
         {2351, nullptr, "RequestNoDownloadRightsErrorResolution"},
         {2352, nullptr, "RequestResolveNoDownloadRightsError"},
+        {2353, nullptr, "GetApplicationDownloadTaskInfo"},
         {2400, nullptr, "GetPromotionInfo"},
         {2401, nullptr, "CountPromotionInfo"},
         {2402, nullptr, "ListPromotionInfo"},
@@ -266,9 +274,12 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {2500, nullptr, "ConfirmAvailableTime"},
         {2510, nullptr, "CreateApplicationResource"},
         {2511, nullptr, "GetApplicationResource"},
-        {2513, nullptr, "LaunchPreomia"},
+        {2513, nullptr, "LaunchMicroApplication"},
         {2514, nullptr, "ClearTaskOfAsyncTaskManager"},
+        {2515, nullptr, "CleanupAllPlaceHolderAndFragmentsIfNoTask"},
+        {2516, nullptr, "EnsureApplicationCertificate"},
         {2800, nullptr, "GetApplicationIdOfPreomia"},
+        {9999, nullptr, "GetApplicationCertificate"},
     };
     // clang-format on
 
@@ -360,10 +371,15 @@ ResultVal<u8> IApplicationManagerInterface::GetApplicationDesiredLanguage(
     // Convert to application language, get priority list
     const auto application_language = ConvertToApplicationLanguage(language_code);
     if (application_language == std::nullopt) {
+        LOG_ERROR(Service_NS, "Could not convert application language! language_code={}",
+                  language_code);
         return ERR_APPLICATION_LANGUAGE_NOT_FOUND;
     }
     const auto priority_list = GetApplicationLanguagePriorityList(*application_language);
     if (!priority_list) {
+        LOG_ERROR(Service_NS,
+                  "Could not find application language priorities! application_language={}",
+                  *application_language);
         return ERR_APPLICATION_LANGUAGE_NOT_FOUND;
     }
 
@@ -375,6 +391,8 @@ ResultVal<u8> IApplicationManagerInterface::GetApplicationDesiredLanguage(
         }
     }
 
+    LOG_ERROR(Service_NS, "Could not find a valid language! supported_languages={:08X}",
+              supported_languages);
     return ERR_APPLICATION_LANGUAGE_NOT_FOUND;
 }
 
@@ -399,6 +417,7 @@ ResultVal<u64> IApplicationManagerInterface::ConvertApplicationLanguageToLanguag
     const auto language_code =
         ConvertToLanguageCode(static_cast<ApplicationLanguage>(application_language));
     if (language_code == std::nullopt) {
+        LOG_ERROR(Service_NS, "Language not found! application_language={}", application_language);
         return ERR_APPLICATION_LANGUAGE_NOT_FOUND;
     }
 
@@ -505,6 +524,10 @@ IFactoryResetInterface::IFactoryResetInterface::IFactoryResetInterface()
             {100, nullptr, "ResetToFactorySettings"},
             {101, nullptr, "ResetToFactorySettingsWithoutUserSaveData"},
             {102, nullptr, "ResetToFactorySettingsForRefurbishment"},
+            {103, nullptr, "ResetToFactorySettingsWithPlatformRegion"},
+            {104, nullptr, "ResetToFactorySettingsWithPlatformRegionAuthentication"},
+            {105, nullptr, "RequestResetToFactorySettingsSecurely"},
+            {106, nullptr, "RequestResetToFactorySettingsWithPlatformRegionAuthenticationSecurely"},
         };
     // clang-format on
 
@@ -553,6 +576,9 @@ public:
             {10, nullptr, "TerminateApplication2"},
             {11, nullptr, "GetRunningApplicationProcessId"},
             {12, nullptr, "SetCurrentApplicationRightsEnvironmentCanBeActive"},
+            {13, nullptr, "CreateApplicationResourceForDevelop"},
+            {14, nullptr, "IsPreomiaForDevelop"},
+            {15, nullptr, "GetApplicationProgramIdFromHost"},
         };
         // clang-format on
 

src/core/hle/service/ns/pl_u.cpp

@@ -164,6 +164,7 @@ PL_U::PL_U(Core::System& system)
         {6, nullptr, "GetSharedFontInOrderOfPriorityForSystem"},
         {100, nullptr, "RequestApplicationFunctionAuthorization"},
         {101, nullptr, "RequestApplicationFunctionAuthorizationForSystem"},
+        {102, nullptr, "RequestApplicationFunctionAuthorizationByApplicationId"},
         {1000, nullptr, "LoadNgWordDataForPlatformRegionChina"},
         {1001, nullptr, "GetNgWordDataSizeForPlatformRegionChina"},
     };

src/core/hle/service/pctl/module.cpp

@@ -31,6 +31,8 @@ public:
             {1014, nullptr, "ConfirmPlayableApplicationVideoOld"},
             {1015, nullptr, "ConfirmPlayableApplicationVideo"},
             {1016, nullptr, "ConfirmShowNewsPermission"},
+            {1017, nullptr, "EndFreeCommunication"},
+            {1018, nullptr, "IsFreeCommunicationAvailable"},
             {1031, nullptr, "IsRestrictionEnabled"},
             {1032, nullptr, "GetSafetyLevel"},
             {1033, nullptr, "SetSafetyLevel"},

src/core/hle/service/prepo/prepo.cpp

@@ -21,8 +21,10 @@ public:
         static const FunctionInfo functions[] = {
             {10100, &PlayReport::SaveReport<Core::Reporter::PlayReportType::Old>, "SaveReportOld"},
             {10101, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::Old>, "SaveReportWithUserOld"},
-            {10102, &PlayReport::SaveReport<Core::Reporter::PlayReportType::New>, "SaveReport"},
-            {10103, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::New>, "SaveReportWithUser"},
+            {10102, &PlayReport::SaveReport<Core::Reporter::PlayReportType::Old2>, "SaveReportOld2"},
+            {10103, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::Old2>, "SaveReportWithUserOld2"},
+            {10104, nullptr, "SaveReport"},
+            {10105, nullptr, "SaveReportWithUser"},
             {10200, nullptr, "RequestImmediateTransmission"},
             {10300, nullptr, "GetTransmissionStatus"},
             {10400, nullptr, "GetSystemSessionId"},
@@ -35,8 +37,10 @@ public:
             {30400, nullptr, "GetStatistics"},
             {30401, nullptr, "GetThroughputHistory"},
             {30500, nullptr, "GetLastUploadError"},
+            {30600, nullptr, "GetApplicationUploadSummary"},
             {40100, nullptr, "IsUserAgreementCheckEnabled"},
             {40101, nullptr, "SetUserAgreementCheckEnabled"},
+            {50100, nullptr, "ReadAllApplicationReportFiles"},
             {90100, nullptr, "ReadAllReportFiles"},
         };
         // clang-format on
@@ -51,7 +55,7 @@ private:
         const auto process_id = rp.PopRaw<u64>();
 
         std::vector<std::vector<u8>> data{ctx.ReadBuffer(0)};
-        if (Type == Core::Reporter::PlayReportType::New) {
+        if constexpr (Type == Core::Reporter::PlayReportType::Old2) {
             data.emplace_back(ctx.ReadBuffer(1));
         }
 
@@ -71,7 +75,7 @@ private:
         const auto user_id = rp.PopRaw<u128>();
         const auto process_id = rp.PopRaw<u64>();
         std::vector<std::vector<u8>> data{ctx.ReadBuffer(0)};
-        if (Type == Core::Reporter::PlayReportType::New) {
+        if constexpr (Type == Core::Reporter::PlayReportType::Old2) {
             data.emplace_back(ctx.ReadBuffer(1));
         }
 

src/core/hle/service/set/set.cpp

@@ -67,6 +67,7 @@ void SET::MakeLanguageCode(Kernel::HLERequestContext& ctx) {
     const auto index = rp.Pop<u32>();
 
     if (index >= available_language_codes.size()) {
+        LOG_ERROR(Service_SET, "Invalid language code index! index={}", index);
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(ERR_INVALID_LANGUAGE);
         return;

src/core/hle/service/set/set_cal.cpp

@@ -50,6 +50,8 @@ SET_CAL::SET_CAL() : ServiceFramework("set:cal") {
         {39, nullptr, "GetConsoleSixAxisSensorModuleType"},
         {40, nullptr, "GetConsoleSixAxisSensorHorizontalOffset"},
         {41, nullptr, "GetBatteryVersion"},
+        {42, nullptr, "GetDeviceId"},
+        {43, nullptr, "GetConsoleSixAxisSensorMountType"},
     };
     // clang-format on
 

src/core/hle/service/set/set_sys.cpp

@@ -288,6 +288,18 @@ SET_SYS::SET_SYS() : ServiceFramework("set:sys") {
         {186, nullptr, "GetMemoryUsageRateFlag"},
         {187, nullptr, "GetTouchScreenMode"},
         {188, nullptr, "SetTouchScreenMode"},
+        {189, nullptr, "GetButtonConfigSettingsFull"},
+        {190, nullptr, "SetButtonConfigSettingsFull"},
+        {191, nullptr, "GetButtonConfigSettingsEmbedded"},
+        {192, nullptr, "SetButtonConfigSettingsEmbedded"},
+        {193, nullptr, "GetButtonConfigSettingsLeft"},
+        {194, nullptr, "SetButtonConfigSettingsLeft"},
+        {195, nullptr, "GetButtonConfigSettingsRight"},
+        {196, nullptr, "SetButtonConfigSettingsRight"},
+        {197, nullptr, "GetButtonConfigRegisteredSettingsEmbedded"},
+        {198, nullptr, "SetButtonConfigRegisteredSettingsEmbedded"},
+        {199, nullptr, "GetButtonConfigRegisteredSettings"},
+        {200, nullptr, "SetButtonConfigRegisteredSettings"},
     };
     // clang-format on
 

src/core/hle/service/sm/sm.cpp

@@ -28,9 +28,11 @@ void ServiceManager::InvokeControlRequest(Kernel::HLERequestContext& context) {
 
 static ResultCode ValidateServiceName(const std::string& name) {
     if (name.size() <= 0 || name.size() > 8) {
+        LOG_ERROR(Service_SM, "Invalid service name! service={}", name);
         return ERR_INVALID_NAME;
     }
     if (name.find('\0') != std::string::npos) {
+        LOG_ERROR(Service_SM, "A non null terminated service was passed");
         return ERR_INVALID_NAME;
     }
     return RESULT_SUCCESS;
@@ -51,8 +53,10 @@ ResultVal<std::shared_ptr<Kernel::ServerPort>> ServiceManager::RegisterService(
 
     CASCADE_CODE(ValidateServiceName(name));
 
-    if (registered_services.find(name) != registered_services.end())
+    if (registered_services.find(name) != registered_services.end()) {
+        LOG_ERROR(Service_SM, "Service is already registered! service={}", name);
         return ERR_ALREADY_REGISTERED;
+    }
 
     auto& kernel = Core::System::GetInstance().Kernel();
     auto [server_port, client_port] =
@@ -66,9 +70,10 @@ ResultCode ServiceManager::UnregisterService(const std::string& name) {
     CASCADE_CODE(ValidateServiceName(name));
 
     const auto iter = registered_services.find(name);
-    if (iter == registered_services.end())
+    if (iter == registered_services.end()) {
+        LOG_ERROR(Service_SM, "Server is not registered! service={}", name);
         return ERR_SERVICE_NOT_REGISTERED;
-
+    }
     registered_services.erase(iter);
     return RESULT_SUCCESS;
 }
@@ -79,6 +84,7 @@ ResultVal<std::shared_ptr<Kernel::ClientPort>> ServiceManager::GetServicePort(
     CASCADE_CODE(ValidateServiceName(name));
     auto it = registered_services.find(name);
     if (it == registered_services.end()) {
+        LOG_ERROR(Service_SM, "Server is not registered! service={}", name);
         return ERR_SERVICE_NOT_REGISTERED;
     }
 

src/core/hle/service/sockets/bsd.cpp

@@ -148,6 +148,7 @@ BSD::BSD(const char* name) : ServiceFramework(name) {
         {30, nullptr, "SendMMsg"},
         {31, nullptr, "EventFd"},
         {32, nullptr, "RegisterResourceStatisticsName"},
+        {33, nullptr, "Initialize2"},
     };
     // clang-format on
 

src/core/hle/service/time/time.cpp

@@ -20,8 +20,8 @@ namespace Service::Time {
 
 class ISystemClock final : public ServiceFramework<ISystemClock> {
 public:
-    ISystemClock(Clock::SystemClockCore& clock_core)
-        : ServiceFramework("ISystemClock"), clock_core{clock_core} {
+    explicit ISystemClock(Clock::SystemClockCore& clock_core, Core::System& system)
+        : ServiceFramework("ISystemClock"), clock_core{clock_core}, system{system} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &ISystemClock::GetCurrentTime, "GetCurrentTime"},
@@ -46,9 +46,8 @@ private:
         }
 
         s64 posix_time{};
-        if (const ResultCode result{
-                clock_core.GetCurrentTime(Core::System::GetInstance(), posix_time)};
-            result != RESULT_SUCCESS) {
+        if (const ResultCode result{clock_core.GetCurrentTime(system, posix_time)};
+            result.IsError()) {
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(result);
             return;
@@ -69,9 +68,8 @@ private:
         }
 
         Clock::SystemClockContext system_clock_context{};
-        if (const ResultCode result{
-                clock_core.GetClockContext(Core::System::GetInstance(), system_clock_context)};
-            result != RESULT_SUCCESS) {
+        if (const ResultCode result{clock_core.GetClockContext(system, system_clock_context)};
+            result.IsError()) {
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(result);
             return;
@@ -83,12 +81,13 @@ private:
     }
 
     Clock::SystemClockCore& clock_core;
+    Core::System& system;
 };
 
 class ISteadyClock final : public ServiceFramework<ISteadyClock> {
 public:
-    ISteadyClock(Clock::SteadyClockCore& clock_core)
-        : ServiceFramework("ISteadyClock"), clock_core{clock_core} {
+    explicit ISteadyClock(Clock::SteadyClockCore& clock_core, Core::System& system)
+        : ServiceFramework("ISteadyClock"), clock_core{clock_core}, system{system} {
         static const FunctionInfo functions[] = {
             {0, &ISteadyClock::GetCurrentTimePoint, "GetCurrentTimePoint"},
         };
@@ -105,14 +104,14 @@ private:
             return;
         }
 
-        const Clock::SteadyClockTimePoint time_point{
-            clock_core.GetCurrentTimePoint(Core::System::GetInstance())};
+        const Clock::SteadyClockTimePoint time_point{clock_core.GetCurrentTimePoint(system)};
         IPC::ResponseBuilder rb{ctx, (sizeof(Clock::SteadyClockTimePoint) / 4) + 2};
         rb.Push(RESULT_SUCCESS);
         rb.PushRaw(time_point);
     }
 
     Clock::SteadyClockCore& clock_core;
+    Core::System& system;
 };
 
 ResultCode Module::Interface::GetClockSnapshotFromSystemClockContextInternal(
@@ -134,7 +133,7 @@ ResultCode Module::Interface::GetClockSnapshotFromSystemClockContextInternal(
     }
 
     const auto current_time_point{
-        time_manager.GetStandardSteadyClockCore().GetCurrentTimePoint(Core::System::GetInstance())};
+        time_manager.GetStandardSteadyClockCore().GetCurrentTimePoint(system)};
     if (const ResultCode result{Clock::ClockSnapshot::GetCurrentTime(
             clock_snapshot.user_time, current_time_point, clock_snapshot.user_context)};
         result != RESULT_SUCCESS) {
@@ -176,21 +175,24 @@ void Module::Interface::GetStandardUserSystemClock(Kernel::HLERequestContext& ct
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardUserSystemClockCore());
+    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardUserSystemClockCore(),
+                                      system);
 }
 
 void Module::Interface::GetStandardNetworkSystemClock(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardNetworkSystemClockCore());
+    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardNetworkSystemClockCore(),
+                                      system);
 }
 
 void Module::Interface::GetStandardSteadyClock(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISteadyClock>(module->GetTimeManager().GetStandardSteadyClockCore());
+    rb.PushIpcInterface<ISteadyClock>(module->GetTimeManager().GetStandardSteadyClockCore(),
+                                      system);
 }
 
 void Module::Interface::GetTimeZoneService(Kernel::HLERequestContext& ctx) {
@@ -204,7 +206,8 @@ void Module::Interface::GetStandardLocalSystemClock(Kernel::HLERequestContext& c
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardLocalSystemClockCore());
+    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardLocalSystemClockCore(),
+                                      system);
 }
 
 void Module::Interface::IsStandardNetworkSystemClockAccuracySufficient(
@@ -228,8 +231,7 @@ void Module::Interface::CalculateMonotonicSystemClockBaseTimePoint(Kernel::HLERe
 
     IPC::RequestParser rp{ctx};
     const auto context{rp.PopRaw<Clock::SystemClockContext>()};
-    const auto current_time_point{
-        steady_clock_core.GetCurrentTimePoint(Core::System::GetInstance())};
+    const auto current_time_point{steady_clock_core.GetCurrentTimePoint(system)};
 
     if (current_time_point.clock_source_id == context.steady_time_point.clock_source_id) {
         const auto ticks{Clock::TimeSpanType::FromTicks(
@@ -255,8 +257,8 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     Clock::SystemClockContext user_context{};
     if (const ResultCode result{
             module->GetTimeManager().GetStandardUserSystemClockCore().GetClockContext(
-                Core::System::GetInstance(), user_context)};
-        result != RESULT_SUCCESS) {
+                system, user_context)};
+        result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
         return;
@@ -264,8 +266,8 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     Clock::SystemClockContext network_context{};
     if (const ResultCode result{
             module->GetTimeManager().GetStandardNetworkSystemClockCore().GetClockContext(
-                Core::System::GetInstance(), network_context)};
-        result != RESULT_SUCCESS) {
+                system, network_context)};
+        result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
         return;
@@ -274,7 +276,7 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     Clock::ClockSnapshot clock_snapshot{};
     if (const ResultCode result{GetClockSnapshotFromSystemClockContextInternal(
             &ctx.GetThread(), user_context, network_context, type, clock_snapshot)};
-        result != RESULT_SUCCESS) {
+        result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
         return;

src/core/hle/service/vi/vi.cpp

@@ -267,7 +267,7 @@ protected:
 
 private:
     struct Data {
-        u32_le unk_0;
+        u32_le unk_0{};
     };
 
     Data data{};
@@ -614,6 +614,14 @@ private:
             ctx.WriteBuffer(response.Serialize());
             break;
         }
+        case TransactionId::SetBufferCount: {
+            LOG_WARNING(Service_VI, "(STUBBED) called, transaction=SetBufferCount");
+            [[maybe_unused]] const auto buffer = ctx.ReadBuffer();
+
+            IGBPEmptyResponseParcel response{};
+            ctx.WriteBuffer(response.Serialize());
+            break;
+        }
         default:
             ASSERT_MSG(false, "Unimplemented");
         }
@@ -859,6 +867,7 @@ private:
 
         const auto layer_id = nv_flinger->CreateLayer(display);
         if (!layer_id) {
+            LOG_ERROR(Service_VI, "Layer not found! display=0x{:016X}", display);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -975,6 +984,7 @@ private:
 
         const auto display_id = nv_flinger->OpenDisplay(name);
         if (!display_id) {
+            LOG_ERROR(Service_VI, "Display not found! display_name={}", name);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -1074,6 +1084,7 @@ private:
 
         const auto display_id = nv_flinger->OpenDisplay(display_name);
         if (!display_id) {
+            LOG_ERROR(Service_VI, "Layer not found! layer_id={}", layer_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -1081,6 +1092,7 @@ private:
 
         const auto buffer_queue_id = nv_flinger->FindBufferQueueId(*display_id, layer_id);
         if (!buffer_queue_id) {
+            LOG_ERROR(Service_VI, "Buffer queue id not found! display_id={}", *display_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -1116,6 +1128,7 @@ private:
 
         const auto layer_id = nv_flinger->CreateLayer(display_id);
         if (!layer_id) {
+            LOG_ERROR(Service_VI, "Layer not found! layer_id={}", *layer_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -1123,6 +1136,7 @@ private:
 
         const auto buffer_queue_id = nv_flinger->FindBufferQueueId(display_id, *layer_id);
         if (!buffer_queue_id) {
+            LOG_ERROR(Service_VI, "Buffer queue id not found! display_id={}", display_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -1153,6 +1167,7 @@ private:
 
         const auto vsync_event = nv_flinger->FindVsyncEvent(display_id);
         if (!vsync_event) {
+            LOG_ERROR(Service_VI, "Vsync event was not found for display_id={}", display_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;
@@ -1193,6 +1208,7 @@ private:
         case NintendoScaleMode::PreserveAspectRatio:
             return MakeResult(ConvertedScaleMode::PreserveAspectRatio);
         default:
+            LOG_ERROR(Service_VI, "Invalid scaling mode specified, mode={}", mode);
             return ERR_OPERATION_FAILED;
         }
     }
@@ -1249,6 +1265,7 @@ void detail::GetDisplayServiceImpl(Kernel::HLERequestContext& ctx,
     const auto policy = rp.PopEnum<Policy>();
 
     if (!IsValidServiceAccess(permission, policy)) {
+        LOG_ERROR(Service_VI, "Permission denied for policy {}", static_cast<u32>(policy));
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(ERR_PERMISSION_DENIED);
         return;

src/core/reporter.h

@@ -56,6 +56,7 @@ public:
 
     enum class PlayReportType {
         Old,
+        Old2,
         New,
         System,
     };

src/core/settings.cpp

@@ -92,7 +92,7 @@ void LogSettings() {
     LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
     LogSetting("Renderer_FrameLimit", Settings::values.frame_limit);
     LogSetting("Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
-    LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
+    LogSetting("Renderer_GPUAccuracyLevel", Settings::values.gpu_accuracy);
     LogSetting("Renderer_UseAsynchronousGpuEmulation",
                Settings::values.use_asynchronous_gpu_emulation);
     LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
@@ -109,4 +109,12 @@ void LogSettings() {
     LogSetting("Services_BCATBoxcatLocal", Settings::values.bcat_boxcat_local);
 }
 
+bool IsGPULevelExtreme() {
+    return values.gpu_accuracy == GPUAccuracy::Extreme;
+}
+
+bool IsGPULevelHigh() {
+    return values.gpu_accuracy == GPUAccuracy::Extreme || values.gpu_accuracy == GPUAccuracy::High;
+}
+
 } // namespace Settings

src/core/settings.h

@@ -376,6 +376,12 @@ enum class RendererBackend {
     Vulkan = 1,
 };
 
+enum class GPUAccuracy : u32 {
+    Normal = 0,
+    High = 1,
+    Extreme = 2,
+};
+
 struct Values {
     // System
     bool use_docked_mode;
@@ -436,10 +442,11 @@ struct Values {
     bool use_frame_limit;
     u16 frame_limit;
     bool use_disk_shader_cache;
-    bool use_accurate_gpu_emulation;
+    GPUAccuracy gpu_accuracy;
     bool use_asynchronous_gpu_emulation;
     bool use_vsync;
     bool force_30fps_mode;
+    bool use_fast_gpu_time;
 
     float bg_red;
     float bg_green;
@@ -480,6 +487,9 @@ struct Values {
     std::map<u64, std::vector<std::string>> disabled_addons;
 } extern values;
 
+bool IsGPULevelExtreme();
+bool IsGPULevelHigh();
+
 void Apply();
 void LogSettings();
 } // namespace Settings

src/core/telemetry_session.cpp

@@ -56,6 +56,18 @@ static const char* TranslateRenderer(Settings::RendererBackend backend) {
     return "Unknown";
 }
 
+static const char* TranslateGPUAccuracyLevel(Settings::GPUAccuracy backend) {
+    switch (backend) {
+    case Settings::GPUAccuracy::Normal:
+        return "Normal";
+    case Settings::GPUAccuracy::High:
+        return "High";
+    case Settings::GPUAccuracy::Extreme:
+        return "Extreme";
+    }
+    return "Unknown";
+}
+
 u64 GetTelemetryId() {
     u64 telemetry_id{};
     const std::string filename{FileUtil::GetUserPath(FileUtil::UserPath::ConfigDir) +
@@ -184,8 +196,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
     AddField(field_type, "Renderer_UseFrameLimit", Settings::values.use_frame_limit);
     AddField(field_type, "Renderer_FrameLimit", Settings::values.frame_limit);
     AddField(field_type, "Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
-    AddField(field_type, "Renderer_UseAccurateGpuEmulation",
-             Settings::values.use_accurate_gpu_emulation);
+    AddField(field_type, "Renderer_GPUAccuracyLevel",
+             TranslateGPUAccuracyLevel(Settings::values.gpu_accuracy));
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation);
     AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);

src/tests/core/core_timing.cpp

@@ -14,13 +14,14 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 
+namespace {
 // Numbers are chosen randomly to make sure the correct one is given.
-static constexpr std::array<u64, 5> CB_IDS{{42, 144, 93, 1026, UINT64_C(0xFFFF7FFFF7FFFF)}};
-static constexpr int MAX_SLICE_LENGTH = 10000; // Copied from CoreTiming internals
+constexpr std::array<u64, 5> CB_IDS{{42, 144, 93, 1026, UINT64_C(0xFFFF7FFFF7FFFF)}};
+constexpr int MAX_SLICE_LENGTH = 10000; // Copied from CoreTiming internals
 
-static std::bitset<CB_IDS.size()> callbacks_ran_flags;
-static u64 expected_callback = 0;
-static s64 lateness = 0;
+std::bitset<CB_IDS.size()> callbacks_ran_flags;
+u64 expected_callback = 0;
+s64 lateness = 0;
 
 template <unsigned int IDX>
 void CallbackTemplate(u64 userdata, s64 cycles_late) {
@@ -31,7 +32,7 @@ void CallbackTemplate(u64 userdata, s64 cycles_late) {
     REQUIRE(lateness == cycles_late);
 }
 
-static u64 callbacks_done = 0;
+u64 callbacks_done = 0;
 
 void EmptyCallback(u64 userdata, s64 cycles_late) {
     ++callbacks_done;
@@ -48,8 +49,8 @@ struct ScopeInit final {
     Core::Timing::CoreTiming core_timing;
 };
 
-static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32 context = 0,
-                            int expected_lateness = 0, int cpu_downcount = 0) {
+void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32 context = 0,
+                     int expected_lateness = 0, int cpu_downcount = 0) {
     callbacks_ran_flags = 0;
     expected_callback = CB_IDS[idx];
     lateness = expected_lateness;
@@ -62,6 +63,7 @@ static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32
 
     REQUIRE(decltype(callbacks_ran_flags)().set(idx) == callbacks_ran_flags);
 }
+} // Anonymous namespace
 
 TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
     ScopeInit guard;

src/video_core/CMakeLists.txt

@@ -23,6 +23,7 @@ add_library(video_core STATIC
     engines/shader_bytecode.h
     engines/shader_header.h
     engines/shader_type.h
+    fence_manager.h
     gpu.cpp
     gpu.h
     gpu_asynch.cpp
@@ -51,6 +52,8 @@ add_library(video_core STATIC
     renderer_opengl/gl_buffer_cache.h
     renderer_opengl/gl_device.cpp
     renderer_opengl/gl_device.h
+    renderer_opengl/gl_fence_manager.cpp
+    renderer_opengl/gl_fence_manager.h
     renderer_opengl/gl_framebuffer_cache.cpp
     renderer_opengl/gl_framebuffer_cache.h
     renderer_opengl/gl_rasterizer.cpp
@@ -121,6 +124,8 @@ add_library(video_core STATIC
     shader/decode.cpp
     shader/expr.cpp
     shader/expr.h
+    shader/memory_util.cpp
+    shader/memory_util.h
     shader/node_helper.cpp
     shader/node_helper.h
     shader/node.h
@@ -176,6 +181,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_descriptor_pool.h
         renderer_vulkan/vk_device.cpp
         renderer_vulkan/vk_device.h
+        renderer_vulkan/vk_fence_manager.cpp
+        renderer_vulkan/vk_fence_manager.h
         renderer_vulkan/vk_graphics_pipeline.cpp
         renderer_vulkan/vk_graphics_pipeline.h
         renderer_vulkan/vk_image.cpp

src/video_core/buffer_cache/buffer_cache.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <list>
 #include <memory>
 #include <mutex>
 #include <unordered_map>
@@ -18,8 +19,10 @@
 
 #include "common/alignment.h"
 #include "common/common_types.h"
+#include "common/logging/log.h"
 #include "core/core.h"
 #include "core/memory.h"
+#include "core/settings.h"
 #include "video_core/buffer_cache/buffer_block.h"
 #include "video_core/buffer_cache/map_interval.h"
 #include "video_core/memory_manager.h"
@@ -79,6 +82,9 @@ public:
         auto map = MapAddress(block, gpu_addr, cpu_addr, size);
         if (is_written) {
             map->MarkAsModified(true, GetModifiedTicks());
+            if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
+                MarkForAsyncFlush(map);
+            }
             if (!map->IsWritten()) {
                 map->MarkAsWritten(true);
                 MarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
@@ -137,11 +143,22 @@ public:
         });
         for (auto& object : objects) {
             if (object->IsModified() && object->IsRegistered()) {
+                mutex.unlock();
                 FlushMap(object);
+                mutex.lock();
             }
         }
     }
 
+    bool MustFlushRegion(VAddr addr, std::size_t size) {
+        std::lock_guard lock{mutex};
+
+        const std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+        return std::any_of(objects.cbegin(), objects.cend(), [](const MapInterval& map) {
+            return map->IsModified() && map->IsRegistered();
+        });
+    }
+
     /// Mark the specified region as being invalidated
     void InvalidateRegion(VAddr addr, u64 size) {
         std::lock_guard lock{mutex};
@@ -154,6 +171,77 @@ public:
         }
     }
 
+    void OnCPUWrite(VAddr addr, std::size_t size) {
+        std::lock_guard lock{mutex};
+
+        for (const auto& object : GetMapsInRange(addr, size)) {
+            if (object->IsMemoryMarked() && object->IsRegistered()) {
+                UnmarkMemory(object);
+                object->SetSyncPending(true);
+                marked_for_unregister.emplace_back(object);
+            }
+        }
+    }
+
+    void SyncGuestHost() {
+        std::lock_guard lock{mutex};
+
+        for (const auto& object : marked_for_unregister) {
+            if (object->IsRegistered()) {
+                object->SetSyncPending(false);
+                Unregister(object);
+            }
+        }
+        marked_for_unregister.clear();
+    }
+
+    void CommitAsyncFlushes() {
+        if (uncommitted_flushes) {
+            auto commit_list = std::make_shared<std::list<MapInterval>>();
+            for (auto& map : *uncommitted_flushes) {
+                if (map->IsRegistered() && map->IsModified()) {
+                    // TODO(Blinkhawk): Implement backend asynchronous flushing
+                    // AsyncFlushMap(map)
+                    commit_list->push_back(map);
+                }
+            }
+            if (!commit_list->empty()) {
+                committed_flushes.push_back(commit_list);
+            } else {
+                committed_flushes.emplace_back();
+            }
+        } else {
+            committed_flushes.emplace_back();
+        }
+        uncommitted_flushes.reset();
+    }
+
+    bool ShouldWaitAsyncFlushes() const {
+        return !committed_flushes.empty() && committed_flushes.front() != nullptr;
+    }
+
+    bool HasUncommittedFlushes() const {
+        return uncommitted_flushes != nullptr;
+    }
+
+    void PopAsyncFlushes() {
+        if (committed_flushes.empty()) {
+            return;
+        }
+        auto& flush_list = committed_flushes.front();
+        if (!flush_list) {
+            committed_flushes.pop_front();
+            return;
+        }
+        for (MapInterval& map : *flush_list) {
+            if (map->IsRegistered()) {
+                // TODO(Blinkhawk): Replace this for reading the asynchronous flush
+                FlushMap(map);
+            }
+        }
+        committed_flushes.pop_front();
+    }
+
     virtual BufferType GetEmptyBuffer(std::size_t size) = 0;
 
 protected:
@@ -196,17 +284,30 @@ protected:
         const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
         mapped_addresses.insert({interval, new_map});
         rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
+        new_map->SetMemoryMarked(true);
         if (inherit_written) {
             MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
             new_map->MarkAsWritten(true);
         }
     }
 
-    /// Unregisters an object from the cache
-    void Unregister(MapInterval& map) {
+    void UnmarkMemory(const MapInterval& map) {
+        if (!map->IsMemoryMarked()) {
+            return;
+        }
         const std::size_t size = map->GetEnd() - map->GetStart();
         rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1);
+        map->SetMemoryMarked(false);
+    }
+
+    /// Unregisters an object from the cache
+    void Unregister(const MapInterval& map) {
+        UnmarkMemory(map);
         map->MarkAsRegistered(false);
+        if (map->IsSyncPending()) {
+            marked_for_unregister.remove(map);
+            map->SetSyncPending(false);
+        }
         if (map->IsWritten()) {
             UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
         }
@@ -264,6 +365,9 @@ private:
         MapInterval new_map = CreateMap(new_start, new_end, new_gpu_addr);
         if (modified_inheritance) {
             new_map->MarkAsModified(true, GetModifiedTicks());
+            if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
+                MarkForAsyncFlush(new_map);
+            }
         }
         Register(new_map, write_inheritance);
         return new_map;
@@ -450,6 +554,13 @@ private:
         return false;
     }
 
+    void MarkForAsyncFlush(MapInterval& map) {
+        if (!uncommitted_flushes) {
+            uncommitted_flushes = std::make_shared<std::unordered_set<MapInterval>>();
+        }
+        uncommitted_flushes->insert(map);
+    }
+
     VideoCore::RasterizerInterface& rasterizer;
     Core::System& system;
 
@@ -479,6 +590,10 @@ private:
     u64 modified_ticks = 0;
 
     std::vector<u8> staging_buffer;
+    std::list<MapInterval> marked_for_unregister;
+
+    std::shared_ptr<std::unordered_set<MapInterval>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::list<MapInterval>>> committed_flushes;
 
     std::recursive_mutex mutex;
 };

src/video_core/buffer_cache/map_interval.h

@@ -46,6 +46,22 @@ public:
         return is_registered;
     }
 
+    void SetMemoryMarked(bool is_memory_marked_) {
+        is_memory_marked = is_memory_marked_;
+    }
+
+    bool IsMemoryMarked() const {
+        return is_memory_marked;
+    }
+
+    void SetSyncPending(bool is_sync_pending_) {
+        is_sync_pending = is_sync_pending_;
+    }
+
+    bool IsSyncPending() const {
+        return is_sync_pending;
+    }
+
     VAddr GetStart() const {
         return start;
     }
@@ -83,6 +99,8 @@ private:
     bool is_written{};
     bool is_modified{};
     bool is_registered{};
+    bool is_memory_marked{};
+    bool is_sync_pending{};
     u64 ticks{};
 };
 

src/video_core/dma_pusher.cpp

@@ -21,6 +21,7 @@ MICROPROFILE_DEFINE(DispatchCalls, "GPU", "Execute command buffer", MP_RGB(128,
 void DmaPusher::DispatchCalls() {
     MICROPROFILE_SCOPE(DispatchCalls);
 
+    gpu.SyncGuestHost();
     // On entering GPU code, assume all memory may be touched by the ARM core.
     gpu.Maxwell3D().OnMemoryWrite();
 
@@ -32,6 +33,8 @@ void DmaPusher::DispatchCalls() {
         }
     }
     gpu.FlushCommands();
+    gpu.SyncGuestHost();
+    gpu.OnCommandListEnd();
 }
 
 bool DmaPusher::Step() {
@@ -68,16 +71,22 @@ bool DmaPusher::Step() {
     gpu.MemoryManager().ReadBlockUnsafe(dma_get, command_headers.data(),
                                         command_list_header.size * sizeof(u32));
 
-    for (const CommandHeader& command_header : command_headers) {
+    for (std::size_t index = 0; index < command_headers.size();) {
+        const CommandHeader& command_header = command_headers[index];
 
-        // now, see if we're in the middle of a command
-        if (dma_state.length_pending) {
-            // Second word of long non-inc methods command - method count
-            dma_state.length_pending = 0;
-            dma_state.method_count = command_header.method_count_;
-        } else if (dma_state.method_count) {
+        if (dma_state.method_count) {
             // Data word of methods command
-            CallMethod(command_header.argument);
+            if (dma_state.non_incrementing) {
+                const u32 max_write = static_cast<u32>(
+                    std::min<std::size_t>(index + dma_state.method_count, command_headers.size()) -
+                    index);
+                CallMultiMethod(&command_header.argument, max_write);
+                dma_state.method_count -= max_write;
+                index += max_write;
+                continue;
+            } else {
+                CallMethod(command_header.argument);
+            }
 
             if (!dma_state.non_incrementing) {
                 dma_state.method++;
@@ -117,6 +126,7 @@ bool DmaPusher::Step() {
                 break;
             }
         }
+        index++;
     }
 
     if (!non_main) {
@@ -137,4 +147,9 @@ void DmaPusher::CallMethod(u32 argument) const {
     gpu.CallMethod({dma_state.method, argument, dma_state.subchannel, dma_state.method_count});
 }
 
+void DmaPusher::CallMultiMethod(const u32* base_start, u32 num_methods) const {
+    gpu.CallMultiMethod(dma_state.method, dma_state.subchannel, base_start, num_methods,
+                        dma_state.method_count);
+}
+
 } // namespace Tegra

src/video_core/dma_pusher.h

@@ -75,6 +75,7 @@ private:
     void SetState(const CommandHeader& command_header);
 
     void CallMethod(u32 argument) const;
+    void CallMultiMethod(const u32* base_start, u32 num_methods) const;
 
     std::vector<CommandHeader> command_headers; ///< Buffer for list of commands fetched at once
 

src/video_core/engines/fermi_2d.cpp

@@ -28,7 +28,13 @@ void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
-std::pair<u32, u32> DelimitLine(u32 src_1, u32 src_2, u32 dst_1, u32 dst_2, u32 src_line) {
+void Fermi2D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
+static std::pair<u32, u32> DelimitLine(u32 src_1, u32 src_2, u32 dst_1, u32 dst_2, u32 src_line) {
     const u32 line_a = src_2 - src_1;
     const u32 line_b = dst_2 - dst_1;
     const u32 excess = std::max<s32>(0, line_a - src_line + src_1);

src/video_core/engines/fermi_2d.h

@@ -39,6 +39,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     enum class Origin : u32 {
         Center = 0,
         Corner = 1,

src/video_core/engines/kepler_compute.cpp

@@ -51,6 +51,13 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void KeplerCompute::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                    u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
 Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) const {
     const std::bitset<8> cbuf_mask = launch_description.const_buffer_enable_mask.Value();
     ASSERT(cbuf_mask[regs.tex_cb_index]);

src/video_core/engines/kepler_compute.h

@@ -202,6 +202,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     Texture::FullTextureInfo GetTexture(std::size_t offset) const;
 
     /// Given a texture handle, returns the TSC and TIC entries.

src/video_core/engines/kepler_memory.cpp

@@ -41,4 +41,11 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void KeplerMemory::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                   u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
 } // namespace Tegra::Engines

src/video_core/engines/kepler_memory.h

@@ -40,6 +40,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     struct Regs {
         static constexpr size_t NUM_REGS = 0x7F;
 

src/video_core/engines/maxwell_3d.cpp

@@ -280,6 +280,58 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                u32 methods_pending) {
+    // Methods after 0xE00 are special, they're actually triggers for some microcode that was
+    // uploaded to the GPU during initialization.
+    if (method >= MacroRegistersStart) {
+        // We're trying to execute a macro
+        if (executing_macro == 0) {
+            // A macro call must begin by writing the macro method's register, not its argument.
+            ASSERT_MSG((method % 2) == 0,
+                       "Can't start macro execution by writing to the ARGS register");
+            executing_macro = method;
+        }
+
+        for (std::size_t i = 0; i < amount; i++) {
+            macro_params.push_back(base_start[i]);
+        }
+
+        // Call the macro when there are no more parameters in the command buffer
+        if (amount == methods_pending) {
+            CallMacroMethod(executing_macro, macro_params.size(), macro_params.data());
+            macro_params.clear();
+        }
+        return;
+    }
+    switch (method) {
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[3]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[4]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[5]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[6]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[7]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[8]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[9]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[10]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[11]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[12]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[13]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[14]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[15]): {
+        ProcessCBMultiData(method, base_start, amount);
+        break;
+    }
+    default: {
+        for (std::size_t i = 0; i < amount; i++) {
+            CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+        }
+    }
+    }
+}
+
 void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
     if (mme_draw.current_mode == MMEDrawMode::Undefined) {
         if (mme_draw.gl_begin_consume) {
@@ -404,7 +456,11 @@ void Maxwell3D::ProcessQueryGet() {
 
     switch (regs.query.query_get.operation) {
     case Regs::QueryOperation::Release:
-        StampQueryResult(regs.query.query_sequence, regs.query.query_get.short_query == 0);
+        if (regs.query.query_get.fence == 1) {
+            rasterizer.SignalSemaphore(regs.query.QueryAddress(), regs.query.query_sequence);
+        } else {
+            StampQueryResult(regs.query.query_sequence, regs.query.query_get.short_query == 0);
+        }
         break;
     case Regs::QueryOperation::Acquire:
         // TODO(Blinkhawk): Under this operation, the GPU waits for the CPU to write a value that
@@ -483,7 +539,7 @@ void Maxwell3D::ProcessSyncPoint() {
     const u32 increment = regs.sync_info.increment.Value();
     [[maybe_unused]] const u32 cache_flush = regs.sync_info.unknown.Value();
     if (increment) {
-        system.GPU().IncrementSyncPoint(sync_point);
+        rasterizer.SignalSyncPoint(sync_point);
     }
 }
 
@@ -566,6 +622,28 @@ void Maxwell3D::StartCBData(u32 method) {
     ProcessCBData(regs.const_buffer.cb_data[cb_data_state.id]);
 }
 
+void Maxwell3D::ProcessCBMultiData(u32 method, const u32* start_base, u32 amount) {
+    if (cb_data_state.current != method) {
+        if (cb_data_state.current != null_cb_data) {
+            FinishCBData();
+        }
+        constexpr u32 first_cb_data = MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]);
+        cb_data_state.start_pos = regs.const_buffer.cb_pos;
+        cb_data_state.id = method - first_cb_data;
+        cb_data_state.current = method;
+        cb_data_state.counter = 0;
+    }
+    const std::size_t id = cb_data_state.id;
+    const std::size_t size = amount;
+    std::size_t i = 0;
+    for (; i < size; i++) {
+        cb_data_state.buffer[id][cb_data_state.counter] = start_base[i];
+        cb_data_state.counter++;
+    }
+    // Increment the current buffer position.
+    regs.const_buffer.cb_pos = regs.const_buffer.cb_pos + 4 * amount;
+}
+
 void Maxwell3D::FinishCBData() {
     // Write the input value to the current const buffer at the current position.
     const GPUVAddr buffer_address = regs.const_buffer.BufferAddress();

src/video_core/engines/maxwell_3d.h

@@ -1259,7 +1259,8 @@ public:
 
                     GPUVAddr LimitAddress() const {
                         return static_cast<GPUVAddr>((static_cast<GPUVAddr>(limit_high) << 32) |
-                                                     limit_low);
+                                                     limit_low) +
+                               1;
                     }
                 } vertex_array_limit[NumVertexArrays];
 
@@ -1358,6 +1359,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     /// Write the value to the register identified by method.
     void CallMethodFromMME(const GPU::MethodCall& method_call);
 
@@ -1511,6 +1515,7 @@ private:
     /// Handles a write to the CB_DATA[i] register.
     void StartCBData(u32 method);
     void ProcessCBData(u32 value);
+    void ProcessCBMultiData(u32 method, const u32* start_base, u32 amount);
     void FinishCBData();
 
     /// Handles a write to the CB_BIND register.

src/video_core/engines/maxwell_dma.cpp

@@ -36,6 +36,13 @@ void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
 #undef MAXWELLDMA_REG_INDEX
 }
 
+void MaxwellDMA::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                 u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
 void MaxwellDMA::HandleCopy() {
     LOG_TRACE(HW_GPU, "Requested a DMA copy");
 
@@ -104,8 +111,13 @@ void MaxwellDMA::HandleCopy() {
             write_buffer.resize(dst_size);
         }
 
-        memory_manager.ReadBlock(source, read_buffer.data(), src_size);
-        memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
+        if (Settings::IsGPULevelExtreme()) {
+            memory_manager.ReadBlock(source, read_buffer.data(), src_size);
+            memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
+        } else {
+            memory_manager.ReadBlockUnsafe(source, read_buffer.data(), src_size);
+            memory_manager.ReadBlockUnsafe(dest, write_buffer.data(), dst_size);
+        }
 
         Texture::UnswizzleSubrect(
             regs.x_count, regs.y_count, regs.dst_pitch, regs.src_params.size_x, bytes_per_pixel,
@@ -136,7 +148,7 @@ void MaxwellDMA::HandleCopy() {
             write_buffer.resize(dst_size);
         }
 
-        if (Settings::values.use_accurate_gpu_emulation) {
+        if (Settings::IsGPULevelExtreme()) {
             memory_manager.ReadBlock(source, read_buffer.data(), src_size);
             memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
         } else {

src/video_core/engines/maxwell_dma.h

@@ -35,6 +35,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     struct Regs {
         static constexpr std::size_t NUM_REGS = 0x1D6;
 

src/video_core/engines/shader_bytecode.h

@@ -655,6 +655,7 @@ union Instruction {
     }
 
     constexpr Instruction(u64 value) : value{value} {}
+    constexpr Instruction(const Instruction& instr) : value(instr.value) {}
 
     BitField<0, 8, Register> gpr0;
     BitField<8, 8, Register> gpr8;
@@ -817,11 +818,9 @@ union Instruction {
         BitField<32, 1, u64> saturate;
         BitField<49, 2, HalfMerge> merge;
 
-        BitField<43, 1, u64> negate_a;
         BitField<44, 1, u64> abs_a;
         BitField<47, 2, HalfType> type_a;
 
-        BitField<31, 1, u64> negate_b;
         BitField<30, 1, u64> abs_b;
         BitField<28, 2, HalfType> type_b;
 

src/video_core/fence_manager.h

@@ -0,0 +1,170 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <array>
+#include <memory>
+#include <queue>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/core.h"
+#include "core/memory.h"
+#include "core/settings.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
+
+namespace VideoCommon {
+
+class FenceBase {
+public:
+    FenceBase(u32 payload, bool is_stubbed)
+        : address{}, payload{payload}, is_semaphore{false}, is_stubbed{is_stubbed} {}
+
+    FenceBase(GPUVAddr address, u32 payload, bool is_stubbed)
+        : address{address}, payload{payload}, is_semaphore{true}, is_stubbed{is_stubbed} {}
+
+    GPUVAddr GetAddress() const {
+        return address;
+    }
+
+    u32 GetPayload() const {
+        return payload;
+    }
+
+    bool IsSemaphore() const {
+        return is_semaphore;
+    }
+
+private:
+    GPUVAddr address;
+    u32 payload;
+    bool is_semaphore;
+
+protected:
+    bool is_stubbed;
+};
+
+template <typename TFence, typename TTextureCache, typename TTBufferCache, typename TQueryCache>
+class FenceManager {
+public:
+    void SignalSemaphore(GPUVAddr addr, u32 value) {
+        TryReleasePendingFences();
+        const bool should_flush = ShouldFlush();
+        CommitAsyncFlushes();
+        TFence new_fence = CreateFence(addr, value, !should_flush);
+        fences.push(new_fence);
+        QueueFence(new_fence);
+        if (should_flush) {
+            rasterizer.FlushCommands();
+        }
+        rasterizer.SyncGuestHost();
+    }
+
+    void SignalSyncPoint(u32 value) {
+        TryReleasePendingFences();
+        const bool should_flush = ShouldFlush();
+        CommitAsyncFlushes();
+        TFence new_fence = CreateFence(value, !should_flush);
+        fences.push(new_fence);
+        QueueFence(new_fence);
+        if (should_flush) {
+            rasterizer.FlushCommands();
+        }
+        rasterizer.SyncGuestHost();
+    }
+
+    void WaitPendingFences() {
+        auto& gpu{system.GPU()};
+        auto& memory_manager{gpu.MemoryManager()};
+        while (!fences.empty()) {
+            TFence& current_fence = fences.front();
+            if (ShouldWait()) {
+                WaitFence(current_fence);
+            }
+            PopAsyncFlushes();
+            if (current_fence->IsSemaphore()) {
+                memory_manager.Write<u32>(current_fence->GetAddress(), current_fence->GetPayload());
+            } else {
+                gpu.IncrementSyncPoint(current_fence->GetPayload());
+            }
+            fences.pop();
+        }
+    }
+
+protected:
+    FenceManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                 TTextureCache& texture_cache, TTBufferCache& buffer_cache,
+                 TQueryCache& query_cache)
+        : system{system}, rasterizer{rasterizer}, texture_cache{texture_cache},
+          buffer_cache{buffer_cache}, query_cache{query_cache} {}
+
+    virtual ~FenceManager() {}
+
+    /// Creates a Sync Point Fence Interface, does not create a backend fence if 'is_stubbed' is
+    /// true
+    virtual TFence CreateFence(u32 value, bool is_stubbed) = 0;
+    /// Creates a Semaphore Fence Interface, does not create a backend fence if 'is_stubbed' is true
+    virtual TFence CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) = 0;
+    /// Queues a fence into the backend if the fence isn't stubbed.
+    virtual void QueueFence(TFence& fence) = 0;
+    /// Notifies that the backend fence has been signaled/reached in host GPU.
+    virtual bool IsFenceSignaled(TFence& fence) const = 0;
+    /// Waits until a fence has been signalled by the host GPU.
+    virtual void WaitFence(TFence& fence) = 0;
+
+    Core::System& system;
+    VideoCore::RasterizerInterface& rasterizer;
+    TTextureCache& texture_cache;
+    TTBufferCache& buffer_cache;
+    TQueryCache& query_cache;
+
+private:
+    void TryReleasePendingFences() {
+        auto& gpu{system.GPU()};
+        auto& memory_manager{gpu.MemoryManager()};
+        while (!fences.empty()) {
+            TFence& current_fence = fences.front();
+            if (ShouldWait() && !IsFenceSignaled(current_fence)) {
+                return;
+            }
+            PopAsyncFlushes();
+            if (current_fence->IsSemaphore()) {
+                memory_manager.Write<u32>(current_fence->GetAddress(), current_fence->GetPayload());
+            } else {
+                gpu.IncrementSyncPoint(current_fence->GetPayload());
+            }
+            fences.pop();
+        }
+    }
+
+    bool ShouldWait() const {
+        return texture_cache.ShouldWaitAsyncFlushes() || buffer_cache.ShouldWaitAsyncFlushes() ||
+               query_cache.ShouldWaitAsyncFlushes();
+    }
+
+    bool ShouldFlush() const {
+        return texture_cache.HasUncommittedFlushes() || buffer_cache.HasUncommittedFlushes() ||
+               query_cache.HasUncommittedFlushes();
+    }
+
+    void PopAsyncFlushes() {
+        texture_cache.PopAsyncFlushes();
+        buffer_cache.PopAsyncFlushes();
+        query_cache.PopAsyncFlushes();
+    }
+
+    void CommitAsyncFlushes() {
+        texture_cache.CommitAsyncFlushes();
+        buffer_cache.CommitAsyncFlushes();
+        query_cache.CommitAsyncFlushes();
+    }
+
+    std::queue<TFence> fences;
+};
+
+} // namespace VideoCommon

src/video_core/gpu.cpp

@@ -9,6 +9,7 @@
 #include "core/core_timing_util.h"
 #include "core/frontend/emu_window.h"
 #include "core/memory.h"
+#include "core/settings.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/kepler_memory.h"
@@ -125,6 +126,28 @@ bool GPU::CancelSyncptInterrupt(const u32 syncpoint_id, const u32 value) {
     return true;
 }
 
+u64 GPU::RequestFlush(VAddr addr, std::size_t size) {
+    std::unique_lock lck{flush_request_mutex};
+    const u64 fence = ++last_flush_fence;
+    flush_requests.emplace_back(fence, addr, size);
+    return fence;
+}
+
+void GPU::TickWork() {
+    std::unique_lock lck{flush_request_mutex};
+    while (!flush_requests.empty()) {
+        auto& request = flush_requests.front();
+        const u64 fence = request.fence;
+        const VAddr addr = request.addr;
+        const std::size_t size = request.size;
+        flush_requests.pop_front();
+        flush_request_mutex.unlock();
+        renderer->Rasterizer().FlushRegion(addr, size);
+        current_flush_fence.store(fence);
+        flush_request_mutex.lock();
+    }
+}
+
 u64 GPU::GetTicks() const {
     // This values were reversed engineered by fincs from NVN
     // The gpu clock is reported in units of 385/625 nanoseconds
@@ -132,7 +155,10 @@ u64 GPU::GetTicks() const {
     constexpr u64 gpu_ticks_den = 625;
 
     const u64 cpu_ticks = system.CoreTiming().GetTicks();
-    const u64 nanoseconds = Core::Timing::CyclesToNs(cpu_ticks).count();
+    u64 nanoseconds = Core::Timing::CyclesToNs(cpu_ticks).count();
+    if (Settings::values.use_fast_gpu_time) {
+        nanoseconds /= 256;
+    }
     const u64 nanoseconds_num = nanoseconds / gpu_ticks_den;
     const u64 nanoseconds_rem = nanoseconds % gpu_ticks_den;
     return nanoseconds_num * gpu_ticks_num + (nanoseconds_rem * gpu_ticks_num) / gpu_ticks_den;
@@ -142,6 +168,13 @@ void GPU::FlushCommands() {
     renderer->Rasterizer().FlushCommands();
 }
 
+void GPU::SyncGuestHost() {
+    renderer->Rasterizer().SyncGuestHost();
+}
+
+void GPU::OnCommandListEnd() {
+    renderer->Rasterizer().ReleaseFences();
+}
 // Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
 // their numbers are written down multiplied by 4 in Docs. Here we are not multiply by 4.
 // So the values you see in docs might be multiplied by 4.
@@ -180,16 +213,32 @@ void GPU::CallMethod(const MethodCall& method_call) {
 
     ASSERT(method_call.subchannel < bound_engines.size());
 
-    if (ExecuteMethodOnEngine(method_call)) {
+    if (ExecuteMethodOnEngine(method_call.method)) {
         CallEngineMethod(method_call);
     } else {
         CallPullerMethod(method_call);
     }
 }
 
-bool GPU::ExecuteMethodOnEngine(const MethodCall& method_call) {
-    const auto method = static_cast<BufferMethods>(method_call.method);
-    return method >= BufferMethods::NonPullerMethods;
+void GPU::CallMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                          u32 methods_pending) {
+    LOG_TRACE(HW_GPU, "Processing method {:08X} on subchannel {}", method, subchannel);
+
+    ASSERT(subchannel < bound_engines.size());
+
+    if (ExecuteMethodOnEngine(method)) {
+        CallEngineMultiMethod(method, subchannel, base_start, amount, methods_pending);
+    } else {
+        for (std::size_t i = 0; i < amount; i++) {
+            CallPullerMethod(
+                {method, base_start[i], subchannel, methods_pending - static_cast<u32>(i)});
+        }
+    }
+}
+
+bool GPU::ExecuteMethodOnEngine(u32 method) {
+    const auto buffer_method = static_cast<BufferMethods>(method);
+    return buffer_method >= BufferMethods::NonPullerMethods;
 }
 
 void GPU::CallPullerMethod(const MethodCall& method_call) {
@@ -269,6 +318,31 @@ void GPU::CallEngineMethod(const MethodCall& method_call) {
     }
 }
 
+void GPU::CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                                u32 methods_pending) {
+    const EngineID engine = bound_engines[subchannel];
+
+    switch (engine) {
+    case EngineID::FERMI_TWOD_A:
+        fermi_2d->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::MAXWELL_B:
+        maxwell_3d->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::KEPLER_COMPUTE_B:
+        kepler_compute->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::MAXWELL_DMA_COPY_A:
+        maxwell_dma->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::KEPLER_INLINE_TO_MEMORY_B:
+        kepler_memory->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented engine");
+    }
+}
+
 void GPU::ProcessBindMethod(const MethodCall& method_call) {
     // Bind the current subchannel to the desired engine id.
     LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel,

src/video_core/gpu.h

@@ -155,7 +155,27 @@ public:
     /// Calls a GPU method.
     void CallMethod(const MethodCall& method_call);
 
+    /// Calls a GPU multivalue method.
+    void CallMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                         u32 methods_pending);
+
+    /// Flush all current written commands into the host GPU for execution.
     void FlushCommands();
+    /// Synchronizes CPU writes with Host GPU memory.
+    void SyncGuestHost();
+    /// Signal the ending of command list.
+    virtual void OnCommandListEnd();
+
+    /// Request a host GPU memory flush from the CPU.
+    u64 RequestFlush(VAddr addr, std::size_t size);
+
+    /// Obtains current flush request fence id.
+    u64 CurrentFlushRequestFence() const {
+        return current_flush_fence.load(std::memory_order_relaxed);
+    }
+
+    /// Tick pending requests within the GPU.
+    void TickWork();
 
     /// Returns a reference to the Maxwell3D GPU engine.
     Engines::Maxwell3D& Maxwell3D();
@@ -293,8 +313,12 @@ private:
     /// Calls a GPU engine method.
     void CallEngineMethod(const MethodCall& method_call);
 
+    /// Calls a GPU engine multivalue method.
+    void CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                               u32 methods_pending);
+
     /// Determines where the method should be executed.
-    bool ExecuteMethodOnEngine(const MethodCall& method_call);
+    bool ExecuteMethodOnEngine(u32 method);
 
 protected:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
@@ -325,6 +349,19 @@ private:
 
     std::condition_variable sync_cv;
 
+    struct FlushRequest {
+        FlushRequest(u64 fence, VAddr addr, std::size_t size)
+            : fence{fence}, addr{addr}, size{size} {}
+        u64 fence;
+        VAddr addr;
+        std::size_t size;
+    };
+
+    std::list<FlushRequest> flush_requests;
+    std::atomic<u64> current_flush_fence{};
+    u64 last_flush_fence{};
+    std::mutex flush_request_mutex;
+
     const bool is_async;
 };
 

src/video_core/gpu_asynch.cpp

@@ -52,4 +52,8 @@ void GPUAsynch::WaitIdle() const {
     gpu_thread.WaitIdle();
 }
 
+void GPUAsynch::OnCommandListEnd() {
+    gpu_thread.OnCommandListEnd();
+}
+
 } // namespace VideoCommon

src/video_core/gpu_asynch.h

@@ -32,6 +32,8 @@ public:
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void WaitIdle() const override;
 
+    void OnCommandListEnd() override;
+
 protected:
     void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const override;
 

src/video_core/gpu_thread.cpp

@@ -6,6 +6,7 @@
 #include "common/microprofile.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
+#include "core/settings.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
@@ -14,8 +15,9 @@
 namespace VideoCommon::GPUThread {
 
 /// Runs the GPU thread
-static void RunThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
-                      Tegra::DmaPusher& dma_pusher, SynchState& state) {
+static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
+                      Core::Frontend::GraphicsContext& context, Tegra::DmaPusher& dma_pusher,
+                      SynchState& state) {
     MicroProfileOnThreadCreate("GpuThread");
 
     // Wait for first GPU command before acquiring the window context
@@ -37,10 +39,14 @@ static void RunThread(VideoCore::RendererBase& renderer, Core::Frontend::Graphic
             dma_pusher.DispatchCalls();
         } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) {
             renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
+        } else if (const auto data = std::get_if<OnCommandListEndCommand>(&next.data)) {
+            renderer.Rasterizer().ReleaseFences();
+        } else if (const auto data = std::get_if<GPUTickCommand>(&next.data)) {
+            system.GPU().TickWork();
         } else if (const auto data = std::get_if<FlushRegionCommand>(&next.data)) {
             renderer.Rasterizer().FlushRegion(data->addr, data->size);
         } else if (const auto data = std::get_if<InvalidateRegionCommand>(&next.data)) {
-            renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
+            renderer.Rasterizer().OnCPUWrite(data->addr, data->size);
         } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
             return;
         } else {
@@ -65,8 +71,8 @@ ThreadManager::~ThreadManager() {
 void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
                                 Core::Frontend::GraphicsContext& context,
                                 Tegra::DmaPusher& dma_pusher) {
-    thread = std::thread{RunThread, std::ref(renderer), std::ref(context), std::ref(dma_pusher),
-                         std::ref(state)};
+    thread = std::thread{RunThread,         std::ref(system),     std::ref(renderer),
+                         std::ref(context), std::ref(dma_pusher), std::ref(state)};
 }
 
 void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
@@ -78,16 +84,29 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
 }
 
 void ThreadManager::FlushRegion(VAddr addr, u64 size) {
-    PushCommand(FlushRegionCommand(addr, size));
+    if (!Settings::IsGPULevelHigh()) {
+        PushCommand(FlushRegionCommand(addr, size));
+        return;
+    }
+    if (!Settings::IsGPULevelExtreme()) {
+        return;
+    }
+    if (system.Renderer().Rasterizer().MustFlushRegion(addr, size)) {
+        auto& gpu = system.GPU();
+        u64 fence = gpu.RequestFlush(addr, size);
+        PushCommand(GPUTickCommand());
+        while (fence > gpu.CurrentFlushRequestFence()) {
+        }
+    }
 }
 
 void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
-    system.Renderer().Rasterizer().InvalidateRegion(addr, size);
+    system.Renderer().Rasterizer().OnCPUWrite(addr, size);
 }
 
 void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
-    InvalidateRegion(addr, size);
+    system.Renderer().Rasterizer().OnCPUWrite(addr, size);
 }
 
 void ThreadManager::WaitIdle() const {
@@ -95,6 +114,10 @@ void ThreadManager::WaitIdle() const {
     }
 }
 
+void ThreadManager::OnCommandListEnd() {
+    PushCommand(OnCommandListEndCommand());
+}
+
 u64 ThreadManager::PushCommand(CommandData&& command_data) {
     const u64 fence{++state.last_fence};
     state.queue.Push(CommandDataContainer(std::move(command_data), fence));

src/video_core/gpu_thread.h

@@ -70,9 +70,16 @@ struct FlushAndInvalidateRegionCommand final {
     u64 size;
 };
 
+/// Command called within the gpu, to schedule actions after a command list end
+struct OnCommandListEndCommand final {};
+
+/// Command to make the gpu look into pending requests
+struct GPUTickCommand final {};
+
 using CommandData =
     std::variant<EndProcessingCommand, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
-                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand>;
+                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand,
+                 GPUTickCommand>;
 
 struct CommandDataContainer {
     CommandDataContainer() = default;
@@ -122,6 +129,8 @@ public:
     // Wait until the gpu thread is idle.
     void WaitIdle() const;
 
+    void OnCommandListEnd();
+
 private:
     /// Pushes a command to be executed by the GPU thread
     u64 PushCommand(CommandData&& command_data);

src/video_core/memory_manager.cpp

@@ -51,11 +51,8 @@ GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) {
     const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)};
 
     MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr);
-    ASSERT(system.CurrentProcess()
-               ->PageTable()
-               .SetMemoryAttribute(cpu_addr, size, Kernel::Memory::MemoryAttribute::DeviceShared,
-                                   Kernel::Memory::MemoryAttribute::DeviceShared)
-               .IsSuccess());
+    ASSERT(
+        system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess());
 
     return gpu_addr;
 }
@@ -66,11 +63,8 @@ GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size)
     const u64 aligned_size{Common::AlignUp(size, page_size)};
 
     MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr);
-    ASSERT(system.CurrentProcess()
-               ->PageTable()
-               .SetMemoryAttribute(cpu_addr, size, Kernel::Memory::MemoryAttribute::DeviceShared,
-                                   Kernel::Memory::MemoryAttribute::DeviceShared)
-               .IsSuccess());
+    ASSERT(
+        system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess());
     return gpu_addr;
 }
 
@@ -87,9 +81,7 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
     UnmapRange(gpu_addr, aligned_size);
     ASSERT(system.CurrentProcess()
                ->PageTable()
-               .SetMemoryAttribute(cpu_addr.value(), size,
-                                   Kernel::Memory::MemoryAttribute::DeviceShared,
-                                   Kernel::Memory::MemoryAttribute::None)
+               .UnlockForDeviceAddressSpace(cpu_addr.value(), size)
                .IsSuccess());
 
     return gpu_addr;

src/video_core/query_cache.h

@@ -12,10 +12,12 @@
 #include <mutex>
 #include <optional>
 #include <unordered_map>
+#include <unordered_set>
 #include <vector>
 
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/settings.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
@@ -130,6 +132,9 @@ public:
         }
 
         query->BindCounter(Stream(type).Current(), timestamp);
+        if (Settings::values.use_asynchronous_gpu_emulation) {
+            AsyncFlushQuery(cpu_addr);
+        }
     }
 
     /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
@@ -170,6 +175,37 @@ public:
         return streams[static_cast<std::size_t>(type)];
     }
 
+    void CommitAsyncFlushes() {
+        committed_flushes.push_back(uncommitted_flushes);
+        uncommitted_flushes.reset();
+    }
+
+    bool HasUncommittedFlushes() const {
+        return uncommitted_flushes != nullptr;
+    }
+
+    bool ShouldWaitAsyncFlushes() const {
+        if (committed_flushes.empty()) {
+            return false;
+        }
+        return committed_flushes.front() != nullptr;
+    }
+
+    void PopAsyncFlushes() {
+        if (committed_flushes.empty()) {
+            return;
+        }
+        auto& flush_list = committed_flushes.front();
+        if (!flush_list) {
+            committed_flushes.pop_front();
+            return;
+        }
+        for (VAddr query_address : *flush_list) {
+            FlushAndRemoveRegion(query_address, 4);
+        }
+        committed_flushes.pop_front();
+    }
+
 protected:
     std::array<QueryPool, VideoCore::NumQueryTypes> query_pools;
 
@@ -224,6 +260,13 @@ private:
         return found != std::end(contents) ? &*found : nullptr;
     }
 
+    void AsyncFlushQuery(VAddr addr) {
+        if (!uncommitted_flushes) {
+            uncommitted_flushes = std::make_shared<std::unordered_set<VAddr>>();
+        }
+        uncommitted_flushes->insert(addr);
+    }
+
     static constexpr std::uintptr_t PAGE_SIZE = 4096;
     static constexpr unsigned PAGE_SHIFT = 12;
 
@@ -235,6 +278,9 @@ private:
     std::unordered_map<u64, std::vector<CachedQuery>> cached_queries;
 
     std::array<CounterStream, VideoCore::NumQueryTypes> streams;
+
+    std::shared_ptr<std::unordered_set<VAddr>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::unordered_set<VAddr>>> committed_flushes;
 };
 
 template <class QueryCache, class HostCounter>

src/video_core/rasterizer_interface.h

@@ -49,15 +49,33 @@ public:
     /// Records a GPU query and caches it
     virtual void Query(GPUVAddr gpu_addr, QueryType type, std::optional<u64> timestamp) = 0;
 
+    /// Signal a GPU based semaphore as a fence
+    virtual void SignalSemaphore(GPUVAddr addr, u32 value) = 0;
+
+    /// Signal a GPU based syncpoint as a fence
+    virtual void SignalSyncPoint(u32 value) = 0;
+
+    /// Release all pending fences.
+    virtual void ReleaseFences() = 0;
+
     /// Notify rasterizer that all caches should be flushed to Switch memory
     virtual void FlushAll() = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     virtual void FlushRegion(VAddr addr, u64 size) = 0;
 
+    /// Check if the the specified memory area requires flushing to CPU Memory.
+    virtual bool MustFlushRegion(VAddr addr, u64 size) = 0;
+
     /// Notify rasterizer that any caches of the specified region should be invalidated
     virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
 
+    /// Notify rasterizer that any caches of the specified region are desync with guest
+    virtual void OnCPUWrite(VAddr addr, u64 size) = 0;
+
+    /// Sync memory between guest and host.
+    virtual void SyncGuestHost() = 0;
+
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     /// and invalidated
     virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -52,7 +52,7 @@ Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
 }
 
 void OGLBufferCache::WriteBarrier() {
-    glMemoryBarrier(GL_ALL_BARRIER_BITS);
+    glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);
 }
 
 GLuint OGLBufferCache::ToHandle(const Buffer& buffer) {
@@ -72,6 +72,7 @@ void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, s
 void OGLBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                                        u8* data) {
     MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
+    glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
     glGetNamedBufferSubData(buffer->GetHandle(), static_cast<GLintptr>(offset),
                             static_cast<GLsizeiptr>(size), data);
 }

src/video_core/renderer_opengl/gl_fence_manager.cpp

@@ -0,0 +1,72 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+
+#include "video_core/renderer_opengl/gl_fence_manager.h"
+
+namespace OpenGL {
+
+GLInnerFence::GLInnerFence(u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(payload, is_stubbed), sync_object{} {}
+
+GLInnerFence::GLInnerFence(GPUVAddr address, u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(address, payload, is_stubbed), sync_object{} {}
+
+GLInnerFence::~GLInnerFence() = default;
+
+void GLInnerFence::Queue() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(sync_object.handle == 0);
+    sync_object.Create();
+}
+
+bool GLInnerFence::IsSignaled() const {
+    if (is_stubbed) {
+        return true;
+    }
+    ASSERT(sync_object.handle != 0);
+    GLsizei length;
+    GLint sync_status;
+    glGetSynciv(sync_object.handle, GL_SYNC_STATUS, sizeof(GLint), &length, &sync_status);
+    return sync_status == GL_SIGNALED;
+}
+
+void GLInnerFence::Wait() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(sync_object.handle != 0);
+    glClientWaitSync(sync_object.handle, 0, GL_TIMEOUT_IGNORED);
+}
+
+FenceManagerOpenGL::FenceManagerOpenGL(Core::System& system,
+                                       VideoCore::RasterizerInterface& rasterizer,
+                                       TextureCacheOpenGL& texture_cache,
+                                       OGLBufferCache& buffer_cache, QueryCache& query_cache)
+    : GenericFenceManager(system, rasterizer, texture_cache, buffer_cache, query_cache) {}
+
+Fence FenceManagerOpenGL::CreateFence(u32 value, bool is_stubbed) {
+    return std::make_shared<GLInnerFence>(value, is_stubbed);
+}
+
+Fence FenceManagerOpenGL::CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) {
+    return std::make_shared<GLInnerFence>(addr, value, is_stubbed);
+}
+
+void FenceManagerOpenGL::QueueFence(Fence& fence) {
+    fence->Queue();
+}
+
+bool FenceManagerOpenGL::IsFenceSignaled(Fence& fence) const {
+    return fence->IsSignaled();
+}
+
+void FenceManagerOpenGL::WaitFence(Fence& fence) {
+    fence->Wait();
+}
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_fence_manager.h

@@ -0,0 +1,53 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <glad/glad.h>
+
+#include "common/common_types.h"
+#include "video_core/fence_manager.h"
+#include "video_core/renderer_opengl/gl_buffer_cache.h"
+#include "video_core/renderer_opengl/gl_query_cache.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_texture_cache.h"
+
+namespace OpenGL {
+
+class GLInnerFence : public VideoCommon::FenceBase {
+public:
+    GLInnerFence(u32 payload, bool is_stubbed);
+    GLInnerFence(GPUVAddr address, u32 payload, bool is_stubbed);
+    ~GLInnerFence();
+
+    void Queue();
+
+    bool IsSignaled() const;
+
+    void Wait();
+
+private:
+    OGLSync sync_object;
+};
+
+using Fence = std::shared_ptr<GLInnerFence>;
+using GenericFenceManager =
+    VideoCommon::FenceManager<Fence, TextureCacheOpenGL, OGLBufferCache, QueryCache>;
+
+class FenceManagerOpenGL final : public GenericFenceManager {
+public:
+    FenceManagerOpenGL(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                       TextureCacheOpenGL& texture_cache, OGLBufferCache& buffer_cache,
+                       QueryCache& query_cache);
+
+protected:
+    Fence CreateFence(u32 value, bool is_stubbed) override;
+    Fence CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) override;
+    void QueueFence(Fence& fence) override;
+    bool IsFenceSignaled(Fence& fence) const override;
+    void WaitFence(Fence& fence) override;
+};
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -99,9 +99,10 @@ RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWind
                                    ScreenInfo& info, GLShader::ProgramManager& program_manager,
                                    StateTracker& state_tracker)
     : RasterizerAccelerated{system.Memory()}, texture_cache{system, *this, device, state_tracker},
-      shader_cache{*this, system, emu_window, device}, query_cache{system, *this}, system{system},
-      screen_info{info}, program_manager{program_manager}, state_tracker{state_tracker},
-      buffer_cache{*this, system, device, STREAM_BUFFER_SIZE} {
+      shader_cache{*this, system, emu_window, device}, query_cache{system, *this},
+      buffer_cache{*this, system, device, STREAM_BUFFER_SIZE},
+      fence_manager{system, *this, texture_cache, buffer_cache, query_cache}, system{system},
+      screen_info{info}, program_manager{program_manager}, state_tracker{state_tracker} {
     CheckExtensions();
 }
 
@@ -185,8 +186,12 @@ void RasterizerOpenGL::SetupVertexBuffer() {
         const GPUVAddr start = vertex_array.StartAddress();
         const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
 
-        ASSERT(end > start);
-        const u64 size = end - start + 1;
+        ASSERT(end >= start);
+        const u64 size = end - start;
+        if (size == 0) {
+            glBindVertexBuffer(static_cast<GLuint>(index), 0, 0, vertex_array.stride);
+            continue;
+        }
         const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
         glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
                            vertex_array.stride);
@@ -310,8 +315,8 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
         const GPUVAddr start = regs.vertex_array[index].StartAddress();
         const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
 
-        ASSERT(end > start);
-        size += end - start + 1;
+        size += end - start;
+        ASSERT(end >= start);
     }
 
     return size;
@@ -599,6 +604,8 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     EndTransformFeedback();
 
     ++num_queued_commands;
+
+    system.GPU().TickWork();
 }
 
 void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
@@ -649,6 +656,13 @@ void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
     query_cache.FlushRegion(addr, size);
 }
 
+bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size) {
+    if (!Settings::IsGPULevelHigh()) {
+        return buffer_cache.MustFlushRegion(addr, size);
+    }
+    return texture_cache.MustFlushRegion(addr, size) || buffer_cache.MustFlushRegion(addr, size);
+}
+
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     if (addr == 0 || size == 0) {
@@ -660,8 +674,52 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     query_cache.InvalidateRegion(addr, size);
 }
 
+void RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    if (addr == 0 || size == 0) {
+        return;
+    }
+    texture_cache.OnCPUWrite(addr, size);
+    shader_cache.InvalidateRegion(addr, size);
+    buffer_cache.OnCPUWrite(addr, size);
+    query_cache.InvalidateRegion(addr, size);
+}
+
+void RasterizerOpenGL::SyncGuestHost() {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    texture_cache.SyncGuestHost();
+    buffer_cache.SyncGuestHost();
+}
+
+void RasterizerOpenGL::SignalSemaphore(GPUVAddr addr, u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        auto& memory_manager{gpu.MemoryManager()};
+        memory_manager.Write<u32>(addr, value);
+        return;
+    }
+    fence_manager.SignalSemaphore(addr, value);
+}
+
+void RasterizerOpenGL::SignalSyncPoint(u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        gpu.IncrementSyncPoint(value);
+        return;
+    }
+    fence_manager.SignalSyncPoint(value);
+}
+
+void RasterizerOpenGL::ReleaseFences() {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        return;
+    }
+    fence_manager.WaitPendingFences();
+}
+
 void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
-    if (Settings::values.use_accurate_gpu_emulation) {
+    if (Settings::IsGPULevelExtreme()) {
         FlushRegion(addr, size);
     }
     InvalidateRegion(addr, size);

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -23,6 +23,7 @@
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_device.h"
+#include "video_core/renderer_opengl/gl_fence_manager.h"
 #include "video_core/renderer_opengl/gl_framebuffer_cache.h"
 #include "video_core/renderer_opengl/gl_query_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
@@ -66,7 +67,13 @@ public:
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
     void FlushRegion(VAddr addr, u64 size) override;
+    bool MustFlushRegion(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size) override;
+    void OnCPUWrite(VAddr addr, u64 size) override;
+    void SyncGuestHost() override;
+    void SignalSemaphore(GPUVAddr addr, u32 value) override;
+    void SignalSyncPoint(u32 value) override;
+    void ReleaseFences() override;
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void FlushCommands() override;
     void TickFrame() override;
@@ -222,6 +229,8 @@ private:
     SamplerCacheOpenGL sampler_cache;
     FramebufferCacheOpenGL framebuffer_cache;
     QueryCache query_cache;
+    OGLBufferCache buffer_cache;
+    FenceManagerOpenGL fence_manager;
 
     Core::System& system;
     ScreenInfo& screen_info;
@@ -229,7 +238,6 @@ private:
     StateTracker& state_tracker;
 
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
-    OGLBufferCache buffer_cache;
 
     GLint vertex_binding = 0;
 

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -10,8 +10,6 @@
 #include <thread>
 #include <unordered_set>
 
-#include <boost/functional/hash.hpp>
-
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
@@ -28,76 +26,26 @@
 #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
 #include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/utils.h"
+#include "video_core/shader/memory_util.h"
 #include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace OpenGL {
 
 using Tegra::Engines::ShaderType;
+using VideoCommon::Shader::GetShaderAddress;
+using VideoCommon::Shader::GetShaderCode;
+using VideoCommon::Shader::GetUniqueIdentifier;
+using VideoCommon::Shader::KERNEL_MAIN_OFFSET;
 using VideoCommon::Shader::ProgramCode;
 using VideoCommon::Shader::Registry;
 using VideoCommon::Shader::ShaderIR;
+using VideoCommon::Shader::STAGE_MAIN_OFFSET;
 
 namespace {
 
-constexpr u32 STAGE_MAIN_OFFSET = 10;
-constexpr u32 KERNEL_MAIN_OFFSET = 0;
-
 constexpr VideoCommon::Shader::CompilerSettings COMPILER_SETTINGS{};
 
-/// Gets the address for the specified shader stage program
-GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) {
-    const auto& gpu{system.GPU().Maxwell3D()};
-    const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
-    return gpu.regs.code_address.CodeAddress() + shader_config.offset;
-}
-
-/// Gets if the current instruction offset is a scheduler instruction
-constexpr bool IsSchedInstruction(std::size_t offset, std::size_t main_offset) {
-    // Sched instructions appear once every 4 instructions.
-    constexpr std::size_t SchedPeriod = 4;
-    const std::size_t absolute_offset = offset - main_offset;
-    return (absolute_offset % SchedPeriod) == 0;
-}
-
-/// Calculates the size of a program stream
-std::size_t CalculateProgramSize(const ProgramCode& program) {
-    constexpr std::size_t start_offset = 10;
-    // This is the encoded version of BRA that jumps to itself. All Nvidia
-    // shaders end with one.
-    constexpr u64 self_jumping_branch = 0xE2400FFFFF07000FULL;
-    constexpr u64 mask = 0xFFFFFFFFFF7FFFFFULL;
-    std::size_t offset = start_offset;
-    while (offset < program.size()) {
-        const u64 instruction = program[offset];
-        if (!IsSchedInstruction(offset, start_offset)) {
-            if ((instruction & mask) == self_jumping_branch) {
-                // End on Maxwell's "nop" instruction
-                break;
-            }
-            if (instruction == 0) {
-                break;
-            }
-        }
-        offset++;
-    }
-    // The last instruction is included in the program size
-    return std::min(offset + 1, program.size());
-}
-
-/// Gets the shader program code from memory for the specified address
-ProgramCode GetShaderCode(Tegra::MemoryManager& memory_manager, const GPUVAddr gpu_addr,
-                          const u8* host_ptr) {
-    ProgramCode code(VideoCommon::Shader::MAX_PROGRAM_LENGTH);
-    ASSERT_OR_EXECUTE(host_ptr != nullptr, {
-        std::fill(code.begin(), code.end(), 0);
-        return code;
-    });
-    memory_manager.ReadBlockUnsafe(gpu_addr, code.data(), code.size() * sizeof(u64));
-    code.resize(CalculateProgramSize(code));
-    return code;
-}
-
 /// Gets the shader type from a Maxwell program type
 constexpr GLenum GetGLShaderType(ShaderType shader_type) {
     switch (shader_type) {
@@ -114,17 +62,6 @@ constexpr GLenum GetGLShaderType(ShaderType shader_type) {
     }
 }
 
-/// Hashes one (or two) program streams
-u64 GetUniqueIdentifier(ShaderType shader_type, bool is_a, const ProgramCode& code,
-                        const ProgramCode& code_b = {}) {
-    u64 unique_identifier = boost::hash_value(code);
-    if (is_a) {
-        // VertexA programs include two programs
-        boost::hash_combine(unique_identifier, boost::hash_value(code_b));
-    }
-    return unique_identifier;
-}
-
 constexpr const char* GetShaderTypeName(ShaderType shader_type) {
     switch (shader_type) {
     case ShaderType::Vertex:
@@ -448,7 +385,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
 
     // Look up shader in the cache based on address
     const auto cpu_addr{memory_manager.GpuToCpuAddress(address)};
-    Shader shader{cpu_addr ? TryGet(*cpu_addr) : nullptr};
+    Shader shader{cpu_addr ? TryGet(*cpu_addr) : null_shader};
     if (shader) {
         return last_shaders[static_cast<std::size_t>(program)] = shader;
     }
@@ -456,11 +393,12 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
     const auto host_ptr{memory_manager.GetPointer(address)};
 
     // No shader found - create a new one
-    ProgramCode code{GetShaderCode(memory_manager, address, host_ptr)};
+    ProgramCode code{GetShaderCode(memory_manager, address, host_ptr, false)};
     ProgramCode code_b;
     if (program == Maxwell::ShaderProgram::VertexA) {
         const GPUVAddr address_b{GetShaderAddress(system, Maxwell::ShaderProgram::VertexB)};
-        code_b = GetShaderCode(memory_manager, address_b, memory_manager.GetPointer(address_b));
+        const u8* host_ptr_b = memory_manager.GetPointer(address_b);
+        code_b = GetShaderCode(memory_manager, address_b, host_ptr_b, false);
     }
 
     const auto unique_identifier = GetUniqueIdentifier(
@@ -477,7 +415,12 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
         const std::size_t size_in_bytes = code.size() * sizeof(u64);
         shader = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
     }
-    Register(shader);
+
+    if (cpu_addr) {
+        Register(shader);
+    } else {
+        null_shader = shader;
+    }
 
     return last_shaders[static_cast<std::size_t>(program)] = shader;
 }
@@ -486,14 +429,14 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
     auto& memory_manager{system.GPU().MemoryManager()};
     const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)};
 
-    auto kernel = cpu_addr ? TryGet(*cpu_addr) : nullptr;
+    auto kernel = cpu_addr ? TryGet(*cpu_addr) : null_kernel;
     if (kernel) {
         return kernel;
     }
 
     const auto host_ptr{memory_manager.GetPointer(code_addr)};
     // No kernel found, create a new one
-    auto code{GetShaderCode(memory_manager, code_addr, host_ptr)};
+    auto code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
     const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
 
     const ShaderParameters params{system,    disk_cache, device,
@@ -507,7 +450,11 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
         kernel = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
     }
 
-    Register(kernel);
+    if (cpu_addr) {
+        Register(kernel);
+    } else {
+        null_kernel = kernel;
+    }
     return kernel;
 }
 

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -125,6 +125,9 @@ private:
     ShaderDiskCacheOpenGL disk_cache;
     std::unordered_map<u64, PrecompiledShader> runtime_cache;
 
+    Shader null_shader{};
+    Shader null_kernel{};
+
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 };
 

src/video_core/renderer_vulkan/fixed_pipeline_state.cpp

@@ -140,6 +140,12 @@ void FixedPipelineState::BlendingAttachment::Fill(const Maxwell& regs, std::size
     enable.Assign(1);
 }
 
+void FixedPipelineState::Fill(const Maxwell& regs) {
+    rasterizer.Fill(regs);
+    depth_stencil.Fill(regs);
+    color_blending.Fill(regs);
+}
+
 std::size_t FixedPipelineState::Hash() const noexcept {
     const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), sizeof *this);
     return static_cast<std::size_t>(hash);
@@ -149,15 +155,6 @@ bool FixedPipelineState::operator==(const FixedPipelineState& rhs) const noexcep
     return std::memcmp(this, &rhs, sizeof *this) == 0;
 }
 
-FixedPipelineState GetFixedPipelineState(const Maxwell& regs) {
-    FixedPipelineState fixed_state;
-    fixed_state.rasterizer.Fill(regs);
-    fixed_state.depth_stencil.Fill(regs);
-    fixed_state.color_blending.Fill(regs);
-    fixed_state.padding = {};
-    return fixed_state;
-}
-
 u32 FixedPipelineState::PackComparisonOp(Maxwell::ComparisonOp op) noexcept {
     // OpenGL enums go from 0x200 to 0x207 and the others from 1 to 8
     // If we substract 0x200 to OpenGL enums and 1 to the others we get a 0-7 range.

src/video_core/renderer_vulkan/fixed_pipeline_state.h

@@ -17,7 +17,7 @@ namespace Vulkan {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
-struct alignas(32) FixedPipelineState {
+struct FixedPipelineState {
     static u32 PackComparisonOp(Maxwell::ComparisonOp op) noexcept;
     static Maxwell::ComparisonOp UnpackComparisonOp(u32 packed) noexcept;
 
@@ -129,7 +129,7 @@ struct alignas(32) FixedPipelineState {
             auto& binding = bindings[index];
             binding.raw = 0;
             binding.enabled.Assign(enabled ? 1 : 0);
-            binding.stride.Assign(stride);
+            binding.stride.Assign(static_cast<u16>(stride));
             binding_divisors[index] = divisor;
         }
 
@@ -237,7 +237,8 @@ struct alignas(32) FixedPipelineState {
     Rasterizer rasterizer;
     DepthStencil depth_stencil;
     ColorBlending color_blending;
-    std::array<u8, 20> padding;
+
+    void Fill(const Maxwell& regs);
 
     std::size_t Hash() const noexcept;
 
@@ -250,9 +251,6 @@ struct alignas(32) FixedPipelineState {
 static_assert(std::has_unique_object_representations_v<FixedPipelineState>);
 static_assert(std::is_trivially_copyable_v<FixedPipelineState>);
 static_assert(std::is_trivially_constructible_v<FixedPipelineState>);
-static_assert(sizeof(FixedPipelineState) % 32 == 0, "Size is not aligned");
-
-FixedPipelineState GetFixedPipelineState(const Maxwell& regs);
 
 } // namespace Vulkan
 

src/video_core/renderer_vulkan/vk_device.h

@@ -82,11 +82,6 @@ public:
         return present_family;
     }
 
-    /// Returns true if the device is integrated with the host CPU.
-    bool IsIntegrated() const {
-        return properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU;
-    }
-
     /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers.
     u32 GetApiVersion() const {
         return properties.apiVersion;

src/video_core/renderer_vulkan/vk_fence_manager.cpp

@@ -0,0 +1,101 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <thread>
+
+#include "video_core/renderer_vulkan/vk_buffer_cache.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_fence_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_texture_cache.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Vulkan {
+
+InnerFence::InnerFence(const VKDevice& device, VKScheduler& scheduler, u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(payload, is_stubbed), device{device}, scheduler{scheduler} {}
+
+InnerFence::InnerFence(const VKDevice& device, VKScheduler& scheduler, GPUVAddr address,
+                       u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(address, payload, is_stubbed), device{device}, scheduler{scheduler} {}
+
+InnerFence::~InnerFence() = default;
+
+void InnerFence::Queue() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(!event);
+
+    event = device.GetLogical().CreateEvent();
+    ticks = scheduler.Ticks();
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([event = *event](vk::CommandBuffer cmdbuf) {
+        cmdbuf.SetEvent(event, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT);
+    });
+}
+
+bool InnerFence::IsSignaled() const {
+    if (is_stubbed) {
+        return true;
+    }
+    ASSERT(event);
+    return IsEventSignalled();
+}
+
+void InnerFence::Wait() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(event);
+
+    if (ticks >= scheduler.Ticks()) {
+        scheduler.Flush();
+    }
+    while (!IsEventSignalled()) {
+        std::this_thread::yield();
+    }
+}
+
+bool InnerFence::IsEventSignalled() const {
+    switch (const VkResult result = event.GetStatus()) {
+    case VK_EVENT_SET:
+        return true;
+    case VK_EVENT_RESET:
+        return false;
+    default:
+        throw vk::Exception(result);
+    }
+}
+
+VKFenceManager::VKFenceManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                               const VKDevice& device, VKScheduler& scheduler,
+                               VKTextureCache& texture_cache, VKBufferCache& buffer_cache,
+                               VKQueryCache& query_cache)
+    : GenericFenceManager(system, rasterizer, texture_cache, buffer_cache, query_cache),
+      device{device}, scheduler{scheduler} {}
+
+Fence VKFenceManager::CreateFence(u32 value, bool is_stubbed) {
+    return std::make_shared<InnerFence>(device, scheduler, value, is_stubbed);
+}
+
+Fence VKFenceManager::CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) {
+    return std::make_shared<InnerFence>(device, scheduler, addr, value, is_stubbed);
+}
+
+void VKFenceManager::QueueFence(Fence& fence) {
+    fence->Queue();
+}
+
+bool VKFenceManager::IsFenceSignaled(Fence& fence) const {
+    return fence->IsSignaled();
+}
+
+void VKFenceManager::WaitFence(Fence& fence) {
+    fence->Wait();
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_fence_manager.h

@@ -0,0 +1,74 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "video_core/fence_manager.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Core {
+class System;
+}
+
+namespace VideoCore {
+class RasterizerInterface;
+}
+
+namespace Vulkan {
+
+class VKBufferCache;
+class VKDevice;
+class VKQueryCache;
+class VKScheduler;
+class VKTextureCache;
+
+class InnerFence : public VideoCommon::FenceBase {
+public:
+    explicit InnerFence(const VKDevice& device, VKScheduler& scheduler, u32 payload,
+                        bool is_stubbed);
+    explicit InnerFence(const VKDevice& device, VKScheduler& scheduler, GPUVAddr address,
+                        u32 payload, bool is_stubbed);
+    ~InnerFence();
+
+    void Queue();
+
+    bool IsSignaled() const;
+
+    void Wait();
+
+private:
+    bool IsEventSignalled() const;
+
+    const VKDevice& device;
+    VKScheduler& scheduler;
+    vk::Event event;
+    u64 ticks = 0;
+};
+using Fence = std::shared_ptr<InnerFence>;
+
+using GenericFenceManager =
+    VideoCommon::FenceManager<Fence, VKTextureCache, VKBufferCache, VKQueryCache>;
+
+class VKFenceManager final : public GenericFenceManager {
+public:
+    explicit VKFenceManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                            const VKDevice& device, VKScheduler& scheduler,
+                            VKTextureCache& texture_cache, VKBufferCache& buffer_cache,
+                            VKQueryCache& query_cache);
+
+protected:
+    Fence CreateFence(u32 value, bool is_stubbed) override;
+    Fence CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) override;
+    void QueueFence(Fence& fence) override;
+    bool IsFenceSignaled(Fence& fence) const override;
+    void WaitFence(Fence& fence) override;
+
+private:
+    const VKDevice& device;
+    VKScheduler& scheduler;
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -288,7 +288,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
     depth_stencil_ci.maxDepthBounds = 0.0f;
 
     std::array<VkPipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments;
-    const std::size_t num_attachments = renderpass_params.color_attachments.size();
+    const auto num_attachments = static_cast<std::size_t>(renderpass_params.num_color_attachments);
     for (std::size_t index = 0; index < num_attachments; ++index) {
         static constexpr std::array COMPONENT_TABLE = {
             VK_COLOR_COMPONENT_R_BIT, VK_COLOR_COMPONENT_G_BIT, VK_COLOR_COMPONENT_B_BIT,

src/video_core/renderer_vulkan/vk_memory_manager.cpp

@@ -118,8 +118,7 @@ private:
 };
 
 VKMemoryManager::VKMemoryManager(const VKDevice& device)
-    : device{device}, properties{device.GetPhysical().GetMemoryProperties()},
-      is_memory_unified{GetMemoryUnified(properties)} {}
+    : device{device}, properties{device.GetPhysical().GetMemoryProperties()} {}
 
 VKMemoryManager::~VKMemoryManager() = default;
 
@@ -209,16 +208,6 @@ VKMemoryCommit VKMemoryManager::TryAllocCommit(const VkMemoryRequirements& requi
     return {};
 }
 
-bool VKMemoryManager::GetMemoryUnified(const VkPhysicalDeviceMemoryProperties& properties) {
-    for (u32 heap_index = 0; heap_index < properties.memoryHeapCount; ++heap_index) {
-        if (!(properties.memoryHeaps[heap_index].flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT)) {
-            // Memory is considered unified when heaps are device local only.
-            return false;
-        }
-    }
-    return true;
-}
-
 VKMemoryCommitImpl::VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation,
                                        const vk::DeviceMemory& memory, u64 begin, u64 end)
     : device{device}, memory{memory}, interval{begin, end}, allocation{allocation} {}

src/video_core/renderer_vulkan/vk_memory_manager.h

@@ -40,11 +40,6 @@ public:
     /// Commits memory required by the image and binds it.
     VKMemoryCommit Commit(const vk::Image& image, bool host_visible);
 
-    /// Returns true if the memory allocations are done always in host visible and coherent memory.
-    bool IsMemoryUnified() const {
-        return is_memory_unified;
-    }
-
 private:
     /// Allocates a chunk of memory.
     bool AllocMemory(VkMemoryPropertyFlags wanted_properties, u32 type_mask, u64 size);
@@ -53,12 +48,8 @@ private:
     VKMemoryCommit TryAllocCommit(const VkMemoryRequirements& requirements,
                                   VkMemoryPropertyFlags wanted_properties);
 
-    /// Returns true if the device uses an unified memory model.
-    static bool GetMemoryUnified(const VkPhysicalDeviceMemoryProperties& properties);
-
-    const VKDevice& device;                            ///< Device handler.
-    const VkPhysicalDeviceMemoryProperties properties; ///< Physical device properties.
-    const bool is_memory_unified;                      ///< True if memory model is unified.
+    const VKDevice& device;                                       ///< Device handler.
+    const VkPhysicalDeviceMemoryProperties properties;            ///< Physical device properties.
     std::vector<std::unique_ptr<VKMemoryAllocation>> allocations; ///< Current allocations.
 };
 

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -27,12 +27,18 @@
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/renderer_vulkan/wrapper.h"
 #include "video_core/shader/compiler_settings.h"
+#include "video_core/shader/memory_util.h"
 
 namespace Vulkan {
 
 MICROPROFILE_DECLARE(Vulkan_PipelineCache);
 
 using Tegra::Engines::ShaderType;
+using VideoCommon::Shader::GetShaderAddress;
+using VideoCommon::Shader::GetShaderCode;
+using VideoCommon::Shader::KERNEL_MAIN_OFFSET;
+using VideoCommon::Shader::ProgramCode;
+using VideoCommon::Shader::STAGE_MAIN_OFFSET;
 
 namespace {
 
@@ -45,60 +51,6 @@ constexpr VkDescriptorType STORAGE_IMAGE = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
 constexpr VideoCommon::Shader::CompilerSettings compiler_settings{
     VideoCommon::Shader::CompileDepth::FullDecompile};
 
-/// Gets the address for the specified shader stage program
-GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) {
-    const auto& gpu{system.GPU().Maxwell3D()};
-    const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
-    return gpu.regs.code_address.CodeAddress() + shader_config.offset;
-}
-
-/// Gets if the current instruction offset is a scheduler instruction
-constexpr bool IsSchedInstruction(std::size_t offset, std::size_t main_offset) {
-    // Sched instructions appear once every 4 instructions.
-    constexpr std::size_t SchedPeriod = 4;
-    const std::size_t absolute_offset = offset - main_offset;
-    return (absolute_offset % SchedPeriod) == 0;
-}
-
-/// Calculates the size of a program stream
-std::size_t CalculateProgramSize(const ProgramCode& program, bool is_compute) {
-    const std::size_t start_offset = is_compute ? 0 : 10;
-    // This is the encoded version of BRA that jumps to itself. All Nvidia
-    // shaders end with one.
-    constexpr u64 self_jumping_branch = 0xE2400FFFFF07000FULL;
-    constexpr u64 mask = 0xFFFFFFFFFF7FFFFFULL;
-    std::size_t offset = start_offset;
-    while (offset < program.size()) {
-        const u64 instruction = program[offset];
-        if (!IsSchedInstruction(offset, start_offset)) {
-            if ((instruction & mask) == self_jumping_branch) {
-                // End on Maxwell's "nop" instruction
-                break;
-            }
-            if (instruction == 0) {
-                break;
-            }
-        }
-        ++offset;
-    }
-    // The last instruction is included in the program size
-    return std::min(offset + 1, program.size());
-}
-
-/// Gets the shader program code from memory for the specified address
-ProgramCode GetShaderCode(Tegra::MemoryManager& memory_manager, const GPUVAddr gpu_addr,
-                          const u8* host_ptr, bool is_compute) {
-    ProgramCode program_code(VideoCommon::Shader::MAX_PROGRAM_LENGTH);
-    ASSERT_OR_EXECUTE(host_ptr != nullptr, {
-        std::fill(program_code.begin(), program_code.end(), 0);
-        return program_code;
-    });
-    memory_manager.ReadBlockUnsafe(gpu_addr, program_code.data(),
-                                   program_code.size() * sizeof(u64));
-    program_code.resize(CalculateProgramSize(program_code, is_compute));
-    return program_code;
-}
-
 constexpr std::size_t GetStageFromProgram(std::size_t program) {
     return program == 0 ? 0 : program - 1;
 }
@@ -161,6 +113,24 @@ u32 FillDescriptorLayout(const ShaderEntries& entries,
 
 } // Anonymous namespace
 
+std::size_t GraphicsPipelineCacheKey::Hash() const noexcept {
+    const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), sizeof *this);
+    return static_cast<std::size_t>(hash);
+}
+
+bool GraphicsPipelineCacheKey::operator==(const GraphicsPipelineCacheKey& rhs) const noexcept {
+    return std::memcmp(&rhs, this, sizeof *this) == 0;
+}
+
+std::size_t ComputePipelineCacheKey::Hash() const noexcept {
+    const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), sizeof *this);
+    return static_cast<std::size_t>(hash);
+}
+
+bool ComputePipelineCacheKey::operator==(const ComputePipelineCacheKey& rhs) const noexcept {
+    return std::memcmp(&rhs, this, sizeof *this) == 0;
+}
+
 CachedShader::CachedShader(Core::System& system, Tegra::Engines::ShaderType stage,
                            GPUVAddr gpu_addr, VAddr cpu_addr, ProgramCode program_code,
                            u32 main_offset)
@@ -207,18 +177,22 @@ std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
         const GPUVAddr program_addr{GetShaderAddress(system, program)};
         const std::optional cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
         ASSERT(cpu_addr);
-        auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr;
+        auto shader = cpu_addr ? TryGet(*cpu_addr) : null_shader;
         if (!shader) {
             const auto host_ptr{memory_manager.GetPointer(program_addr)};
 
             // No shader found - create a new one
-            constexpr u32 stage_offset = 10;
+            constexpr u32 stage_offset = STAGE_MAIN_OFFSET;
             const auto stage = static_cast<Tegra::Engines::ShaderType>(index == 0 ? 0 : index - 1);
-            auto code = GetShaderCode(memory_manager, program_addr, host_ptr, false);
+            ProgramCode code = GetShaderCode(memory_manager, program_addr, host_ptr, false);
 
             shader = std::make_shared<CachedShader>(system, stage, program_addr, *cpu_addr,
                                                     std::move(code), stage_offset);
-            Register(shader);
+            if (cpu_addr) {
+                Register(shader);
+            } else {
+                null_shader = shader;
+            }
         }
         shaders[index] = std::move(shader);
     }
@@ -261,17 +235,20 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach
     const auto cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
     ASSERT(cpu_addr);
 
-    auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr;
+    auto shader = cpu_addr ? TryGet(*cpu_addr) : null_kernel;
     if (!shader) {
         // No shader found - create a new one
         const auto host_ptr = memory_manager.GetPointer(program_addr);
 
-        auto code = GetShaderCode(memory_manager, program_addr, host_ptr, true);
-        constexpr u32 kernel_main_offset = 0;
+        ProgramCode code = GetShaderCode(memory_manager, program_addr, host_ptr, true);
         shader = std::make_shared<CachedShader>(system, Tegra::Engines::ShaderType::Compute,
                                                 program_addr, *cpu_addr, std::move(code),
-                                                kernel_main_offset);
-        Register(shader);
+                                                KERNEL_MAIN_OFFSET);
+        if (cpu_addr) {
+            Register(shader);
+        } else {
+            null_kernel = shader;
+        }
     }
 
     Specialization specialization;

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -7,7 +7,6 @@
 #include <array>
 #include <cstddef>
 #include <memory>
-#include <tuple>
 #include <type_traits>
 #include <unordered_map>
 #include <utility>
@@ -25,6 +24,7 @@
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_shader_decompiler.h"
 #include "video_core/renderer_vulkan/wrapper.h"
+#include "video_core/shader/memory_util.h"
 #include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 #include "video_core/surface.h"
@@ -47,46 +47,40 @@ class CachedShader;
 using Shader = std::shared_ptr<CachedShader>;
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
-using ProgramCode = std::vector<u64>;
-
 struct GraphicsPipelineCacheKey {
     FixedPipelineState fixed_state;
-    std::array<GPUVAddr, Maxwell::MaxShaderProgram> shaders;
     RenderPassParams renderpass_params;
+    std::array<GPUVAddr, Maxwell::MaxShaderProgram> shaders;
+    u64 padding; // This is necessary for unique object representations
 
-    std::size_t Hash() const noexcept {
-        std::size_t hash = fixed_state.Hash();
-        for (const auto& shader : shaders) {
-            boost::hash_combine(hash, shader);
-        }
-        boost::hash_combine(hash, renderpass_params.Hash());
-        return hash;
-    }
+    std::size_t Hash() const noexcept;
 
-    bool operator==(const GraphicsPipelineCacheKey& rhs) const noexcept {
-        return std::tie(fixed_state, shaders, renderpass_params) ==
-               std::tie(rhs.fixed_state, rhs.shaders, rhs.renderpass_params);
+    bool operator==(const GraphicsPipelineCacheKey& rhs) const noexcept;
+
+    bool operator!=(const GraphicsPipelineCacheKey& rhs) const noexcept {
+        return !operator==(rhs);
     }
 };
+static_assert(std::has_unique_object_representations_v<GraphicsPipelineCacheKey>);
+static_assert(std::is_trivially_copyable_v<GraphicsPipelineCacheKey>);
+static_assert(std::is_trivially_constructible_v<GraphicsPipelineCacheKey>);
 
 struct ComputePipelineCacheKey {
-    GPUVAddr shader{};
-    u32 shared_memory_size{};
-    std::array<u32, 3> workgroup_size{};
+    GPUVAddr shader;
+    u32 shared_memory_size;
+    std::array<u32, 3> workgroup_size;
 
-    std::size_t Hash() const noexcept {
-        return static_cast<std::size_t>(shader) ^
-               ((static_cast<std::size_t>(shared_memory_size) >> 7) << 40) ^
-               static_cast<std::size_t>(workgroup_size[0]) ^
-               (static_cast<std::size_t>(workgroup_size[1]) << 16) ^
-               (static_cast<std::size_t>(workgroup_size[2]) << 24);
-    }
+    std::size_t Hash() const noexcept;
 
-    bool operator==(const ComputePipelineCacheKey& rhs) const noexcept {
-        return std::tie(shader, shared_memory_size, workgroup_size) ==
-               std::tie(rhs.shader, rhs.shared_memory_size, rhs.workgroup_size);
+    bool operator==(const ComputePipelineCacheKey& rhs) const noexcept;
+
+    bool operator!=(const ComputePipelineCacheKey& rhs) const noexcept {
+        return !operator==(rhs);
     }
 };
+static_assert(std::has_unique_object_representations_v<ComputePipelineCacheKey>);
+static_assert(std::is_trivially_copyable_v<ComputePipelineCacheKey>);
+static_assert(std::is_trivially_constructible_v<ComputePipelineCacheKey>);
 
 } // namespace Vulkan
 
@@ -113,7 +107,8 @@ namespace Vulkan {
 class CachedShader final : public RasterizerCacheObject {
 public:
     explicit CachedShader(Core::System& system, Tegra::Engines::ShaderType stage, GPUVAddr gpu_addr,
-                          VAddr cpu_addr, ProgramCode program_code, u32 main_offset);
+                          VAddr cpu_addr, VideoCommon::Shader::ProgramCode program_code,
+                          u32 main_offset);
     ~CachedShader();
 
     GPUVAddr GetGpuAddr() const {
@@ -145,7 +140,7 @@ private:
                                                                  Tegra::Engines::ShaderType stage);
 
     GPUVAddr gpu_addr{};
-    ProgramCode program_code;
+    VideoCommon::Shader::ProgramCode program_code;
     VideoCommon::Shader::Registry registry;
     VideoCommon::Shader::ShaderIR shader_ir;
     ShaderEntries entries;
@@ -182,6 +177,9 @@ private:
     VKUpdateDescriptorQueue& update_descriptor_queue;
     VKRenderPassCache& renderpass_cache;
 
+    Shader null_shader{};
+    Shader null_kernel{};
+
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 
     GraphicsPipelineCacheKey last_graphics_key;

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -17,6 +17,7 @@
 #include "common/microprofile.h"
 #include "core/core.h"
 #include "core/memory.h"
+#include "core/settings.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/fixed_pipeline_state.h"
@@ -299,7 +300,9 @@ RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWind
       pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue,
                      renderpass_cache),
       buffer_cache(*this, system, device, memory_manager, scheduler, staging_pool),
-      sampler_cache(device), query_cache(system, *this, device, scheduler) {
+      sampler_cache(device),
+      fence_manager(system, *this, device, scheduler, texture_cache, buffer_cache, query_cache),
+      query_cache(system, *this, device, scheduler) {
     scheduler.SetQueryCache(query_cache);
 }
 
@@ -313,7 +316,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     query_cache.UpdateCounters();
 
     const auto& gpu = system.GPU().Maxwell3D();
-    GraphicsPipelineCacheKey key{GetFixedPipelineState(gpu.regs)};
+    GraphicsPipelineCacheKey key;
+    key.fixed_state.Fill(gpu.regs);
 
     buffer_cache.Map(CalculateGraphicsStreamBufferSize(is_indexed));
 
@@ -331,10 +335,11 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
 
     buffer_cache.Unmap();
 
-    const auto texceptions = UpdateAttachments();
+    const Texceptions texceptions = UpdateAttachments();
     SetupImageTransitions(texceptions, color_attachments, zeta_attachment);
 
     key.renderpass_params = GetRenderPassParams(texceptions);
+    key.padding = 0;
 
     auto& pipeline = pipeline_cache.GetGraphicsPipeline(key);
     scheduler.BindGraphicsPipeline(pipeline.GetHandle());
@@ -360,6 +365,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     });
 
     EndTransformFeedback();
+
+    system.GPU().TickWork();
 }
 
 void RasterizerVulkan::Clear() {
@@ -448,10 +455,12 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
     query_cache.UpdateCounters();
 
     const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
-    const ComputePipelineCacheKey key{
-        code_addr,
-        launch_desc.shared_alloc,
-        {launch_desc.block_dim_x, launch_desc.block_dim_y, launch_desc.block_dim_z}};
+    ComputePipelineCacheKey key;
+    key.shader = code_addr;
+    key.shared_memory_size = launch_desc.shared_alloc;
+    key.workgroup_size = {launch_desc.block_dim_x, launch_desc.block_dim_y,
+                          launch_desc.block_dim_z};
+
     auto& pipeline = pipeline_cache.GetComputePipeline(key);
 
     // Compute dispatches can't be executed inside a renderpass
@@ -504,6 +513,13 @@ void RasterizerVulkan::FlushRegion(VAddr addr, u64 size) {
     query_cache.FlushRegion(addr, size);
 }
 
+bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size) {
+    if (!Settings::IsGPULevelHigh()) {
+        return buffer_cache.MustFlushRegion(addr, size);
+    }
+    return texture_cache.MustFlushRegion(addr, size) || buffer_cache.MustFlushRegion(addr, size);
+}
+
 void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
     if (addr == 0 || size == 0) {
         return;
@@ -514,6 +530,47 @@ void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
     query_cache.InvalidateRegion(addr, size);
 }
 
+void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
+    if (addr == 0 || size == 0) {
+        return;
+    }
+    texture_cache.OnCPUWrite(addr, size);
+    pipeline_cache.InvalidateRegion(addr, size);
+    buffer_cache.OnCPUWrite(addr, size);
+    query_cache.InvalidateRegion(addr, size);
+}
+
+void RasterizerVulkan::SyncGuestHost() {
+    texture_cache.SyncGuestHost();
+    buffer_cache.SyncGuestHost();
+}
+
+void RasterizerVulkan::SignalSemaphore(GPUVAddr addr, u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        gpu.MemoryManager().Write<u32>(addr, value);
+        return;
+    }
+    fence_manager.SignalSemaphore(addr, value);
+}
+
+void RasterizerVulkan::SignalSyncPoint(u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        gpu.IncrementSyncPoint(value);
+        return;
+    }
+    fence_manager.SignalSyncPoint(value);
+}
+
+void RasterizerVulkan::ReleaseFences() {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        return;
+    }
+    fence_manager.WaitPendingFences();
+}
+
 void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     FlushRegion(addr, size);
     InvalidateRegion(addr, size);
@@ -635,7 +692,7 @@ std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers(
     FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(),
                             std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()};
 
-    const auto try_push = [&](const View& view) {
+    const auto try_push = [&key](const View& view) {
         if (!view) {
             return false;
         }
@@ -646,7 +703,9 @@ std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers(
         return true;
     };
 
-    for (std::size_t index = 0; index < std::size(color_attachments); ++index) {
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    const std::size_t num_attachments = static_cast<std::size_t>(regs.rt_control.count);
+    for (std::size_t index = 0; index < num_attachments; ++index) {
         if (try_push(color_attachments[index])) {
             texture_cache.MarkColorBufferInUse(index);
         }
@@ -824,8 +883,12 @@ void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex
         const GPUVAddr start{vertex_array.StartAddress()};
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
 
-        ASSERT(end > start);
-        const std::size_t size{end - start + 1};
+        ASSERT(end >= start);
+        const std::size_t size{end - start};
+        if (size == 0) {
+            buffer_bindings.AddVertexBinding(DefaultBuffer(), 0);
+            continue;
+        }
         const auto [buffer, offset] = buffer_cache.UploadMemory(start, size);
         buffer_bindings.AddVertexBinding(buffer, offset);
     }
@@ -980,8 +1043,7 @@ void RasterizerVulkan::SetupConstBuffer(const ConstBufferEntry& entry,
                                         const Tegra::Engines::ConstBufferInfo& buffer) {
     if (!buffer.enabled) {
         // Set values to zero to unbind buffers
-        update_descriptor_queue.AddBuffer(buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
-                                          sizeof(float));
+        update_descriptor_queue.AddBuffer(DefaultBuffer(), 0, DEFAULT_BUFFER_SIZE);
         return;
     }
 
@@ -1004,7 +1066,9 @@ void RasterizerVulkan::SetupGlobalBuffer(const GlobalBufferEntry& entry, GPUVAdd
     if (size == 0) {
         // Sometimes global memory pointers don't have a proper size. Upload a dummy entry
         // because Vulkan doesn't like empty buffers.
-        constexpr std::size_t dummy_size = 4;
+        // Note: Do *not* use DefaultBuffer() here, storage buffers can be written breaking the
+        // default buffer.
+        static constexpr std::size_t dummy_size = 4;
         const auto buffer = buffer_cache.GetEmptyBuffer(dummy_size);
         update_descriptor_queue.AddBuffer(buffer, 0, dummy_size);
         return;
@@ -1169,7 +1233,7 @@ std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
         DEBUG_ASSERT(end >= start);
 
-        size += (end - start + 1) * regs.vertex_array[index].enable;
+        size += (end - start) * regs.vertex_array[index].enable;
     }
     return size;
 }
@@ -1192,28 +1256,54 @@ std::size_t RasterizerVulkan::CalculateConstBufferSize(
 }
 
 RenderPassParams RasterizerVulkan::GetRenderPassParams(Texceptions texceptions) const {
-    using namespace VideoCore::Surface;
-
     const auto& regs = system.GPU().Maxwell3D().regs;
-    RenderPassParams renderpass_params;
+    const std::size_t num_attachments = static_cast<std::size_t>(regs.rt_control.count);
 
-    for (std::size_t rt = 0; rt < static_cast<std::size_t>(regs.rt_control.count); ++rt) {
+    RenderPassParams params;
+    params.color_formats = {};
+    std::size_t color_texceptions = 0;
+
+    std::size_t index = 0;
+    for (std::size_t rt = 0; rt < num_attachments; ++rt) {
         const auto& rendertarget = regs.rt[rt];
         if (rendertarget.Address() == 0 || rendertarget.format == Tegra::RenderTargetFormat::NONE) {
             continue;
         }
-        renderpass_params.color_attachments.push_back(RenderPassParams::ColorAttachment{
-            static_cast<u32>(rt), PixelFormatFromRenderTargetFormat(rendertarget.format),
-            texceptions[rt]});
+        params.color_formats[index] = static_cast<u8>(rendertarget.format);
+        color_texceptions |= (texceptions[rt] ? 1ULL : 0ULL) << index;
+        ++index;
+    }
+    params.num_color_attachments = static_cast<u8>(index);
+    params.texceptions = static_cast<u8>(color_texceptions);
+
+    params.zeta_format = regs.zeta_enable ? static_cast<u8>(regs.zeta.format) : 0;
+    params.zeta_texception = texceptions[ZETA_TEXCEPTION_INDEX];
+    return params;
+}
+
+VkBuffer RasterizerVulkan::DefaultBuffer() {
+    if (default_buffer) {
+        return *default_buffer;
     }
 
-    renderpass_params.has_zeta = regs.zeta_enable;
-    if (renderpass_params.has_zeta) {
-        renderpass_params.zeta_pixel_format = PixelFormatFromDepthFormat(regs.zeta.format);
-        renderpass_params.zeta_texception = texceptions[ZETA_TEXCEPTION_INDEX];
-    }
+    VkBufferCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.size = DEFAULT_BUFFER_SIZE;
+    ci.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT |
+               VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
+    ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
+    ci.queueFamilyIndexCount = 0;
+    ci.pQueueFamilyIndices = nullptr;
+    default_buffer = device.GetLogical().CreateBuffer(ci);
+    default_buffer_commit = memory_manager.Commit(default_buffer, false);
 
-    return renderpass_params;
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([buffer = *default_buffer](vk::CommandBuffer cmdbuf) {
+        cmdbuf.FillBuffer(buffer, 0, DEFAULT_BUFFER_SIZE, 0);
+    });
+    return *default_buffer;
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -21,6 +21,7 @@
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_fence_manager.h"
 #include "video_core/renderer_vulkan/vk_memory_manager.h"
 #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
 #include "video_core/renderer_vulkan/vk_query_cache.h"
@@ -118,7 +119,13 @@ public:
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
     void FlushRegion(VAddr addr, u64 size) override;
+    bool MustFlushRegion(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size) override;
+    void OnCPUWrite(VAddr addr, u64 size) override;
+    void SyncGuestHost() override;
+    void SignalSemaphore(GPUVAddr addr, u32 value) override;
+    void SignalSyncPoint(u32 value) override;
+    void ReleaseFences() override;
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void FlushCommands() override;
     void TickFrame() override;
@@ -148,6 +155,7 @@ private:
     using Texceptions = std::bitset<Maxwell::NumRenderTargets + 1>;
 
     static constexpr std::size_t ZETA_TEXCEPTION_INDEX = 8;
+    static constexpr VkDeviceSize DEFAULT_BUFFER_SIZE = 4 * sizeof(float);
 
     void FlushWork();
 
@@ -240,6 +248,8 @@ private:
 
     RenderPassParams GetRenderPassParams(Texceptions texceptions) const;
 
+    VkBuffer DefaultBuffer();
+
     Core::System& system;
     Core::Frontend::EmuWindow& render_window;
     VKScreenInfo& screen_info;
@@ -261,8 +271,12 @@ private:
     VKPipelineCache pipeline_cache;
     VKBufferCache buffer_cache;
     VKSamplerCache sampler_cache;
+    VKFenceManager fence_manager;
     VKQueryCache query_cache;
 
+    vk::Buffer default_buffer;
+    VKMemoryCommit default_buffer_commit;
+
     std::array<View, Maxwell::NumRenderTargets> color_attachments;
     View zeta_attachment;
 

src/video_core/renderer_vulkan/vk_renderpass_cache.cpp

@@ -2,9 +2,11 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <cstring>
 #include <memory>
 #include <vector>
 
+#include "common/cityhash.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/maxwell_to_vk.h"
 #include "video_core/renderer_vulkan/vk_device.h"
@@ -13,6 +15,15 @@
 
 namespace Vulkan {
 
+std::size_t RenderPassParams::Hash() const noexcept {
+    const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), sizeof *this);
+    return static_cast<std::size_t>(hash);
+}
+
+bool RenderPassParams::operator==(const RenderPassParams& rhs) const noexcept {
+    return std::memcmp(&rhs, this, sizeof *this) == 0;
+}
+
 VKRenderPassCache::VKRenderPassCache(const VKDevice& device) : device{device} {}
 
 VKRenderPassCache::~VKRenderPassCache() = default;
@@ -27,20 +38,22 @@ VkRenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) {
 }
 
 vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const {
+    using namespace VideoCore::Surface;
     std::vector<VkAttachmentDescription> descriptors;
     std::vector<VkAttachmentReference> color_references;
 
-    for (std::size_t rt = 0; rt < params.color_attachments.size(); ++rt) {
-        const auto attachment = params.color_attachments[rt];
-        const auto format =
-            MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, attachment.pixel_format);
+    const std::size_t num_attachments = static_cast<std::size_t>(params.num_color_attachments);
+    for (std::size_t rt = 0; rt < num_attachments; ++rt) {
+        const auto guest_format = static_cast<Tegra::RenderTargetFormat>(params.color_formats[rt]);
+        const PixelFormat pixel_format = PixelFormatFromRenderTargetFormat(guest_format);
+        const auto format = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, pixel_format);
         ASSERT_MSG(format.attachable, "Trying to attach a non-attachable format with format={}",
-                   static_cast<u32>(attachment.pixel_format));
+                   static_cast<int>(pixel_format));
 
-        // TODO(Rodrigo): Add eMayAlias when it's needed.
-        const auto color_layout = attachment.is_texception
-                                      ? VK_IMAGE_LAYOUT_GENERAL
-                                      : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
+        // TODO(Rodrigo): Add MAY_ALIAS_BIT when it's needed.
+        const VkImageLayout color_layout = ((params.texceptions >> rt) & 1) != 0
+                                               ? VK_IMAGE_LAYOUT_GENERAL
+                                               : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
         VkAttachmentDescription& descriptor = descriptors.emplace_back();
         descriptor.flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT;
         descriptor.format = format.format;
@@ -58,15 +71,17 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
     }
 
     VkAttachmentReference zeta_attachment_ref;
-    if (params.has_zeta) {
-        const auto format =
-            MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.zeta_pixel_format);
+    const bool has_zeta = params.zeta_format != 0;
+    if (has_zeta) {
+        const auto guest_format = static_cast<Tegra::DepthFormat>(params.zeta_format);
+        const PixelFormat pixel_format = PixelFormatFromDepthFormat(guest_format);
+        const auto format = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, pixel_format);
         ASSERT_MSG(format.attachable, "Trying to attach a non-attachable format with format={}",
-                   static_cast<u32>(params.zeta_pixel_format));
+                   static_cast<int>(pixel_format));
 
-        const auto zeta_layout = params.zeta_texception
-                                     ? VK_IMAGE_LAYOUT_GENERAL
-                                     : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
+        const VkImageLayout zeta_layout = params.zeta_texception != 0
+                                              ? VK_IMAGE_LAYOUT_GENERAL
+                                              : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
         VkAttachmentDescription& descriptor = descriptors.emplace_back();
         descriptor.flags = 0;
         descriptor.format = format.format;
@@ -78,7 +93,7 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
         descriptor.initialLayout = zeta_layout;
         descriptor.finalLayout = zeta_layout;
 
-        zeta_attachment_ref.attachment = static_cast<u32>(params.color_attachments.size());
+        zeta_attachment_ref.attachment = static_cast<u32>(num_attachments);
         zeta_attachment_ref.layout = zeta_layout;
     }
 
@@ -90,7 +105,7 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
     subpass_description.colorAttachmentCount = static_cast<u32>(color_references.size());
     subpass_description.pColorAttachments = color_references.data();
     subpass_description.pResolveAttachments = nullptr;
-    subpass_description.pDepthStencilAttachment = params.has_zeta ? &zeta_attachment_ref : nullptr;
+    subpass_description.pDepthStencilAttachment = has_zeta ? &zeta_attachment_ref : nullptr;
     subpass_description.preserveAttachmentCount = 0;
     subpass_description.pPreserveAttachments = nullptr;
 
@@ -101,7 +116,7 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
         stage |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
     }
 
-    if (params.has_zeta) {
+    if (has_zeta) {
         access |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
                   VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
         stage |= VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;

src/video_core/renderer_vulkan/vk_renderpass_cache.h

@@ -4,8 +4,7 @@
 
 #pragma once
 
-#include <memory>
-#include <tuple>
+#include <type_traits>
 #include <unordered_map>
 
 #include <boost/container/static_vector.hpp>
@@ -19,51 +18,25 @@ namespace Vulkan {
 
 class VKDevice;
 
-// TODO(Rodrigo): Optimize this structure for faster hashing
-
 struct RenderPassParams {
-    struct ColorAttachment {
-        u32 index = 0;
-        VideoCore::Surface::PixelFormat pixel_format = VideoCore::Surface::PixelFormat::Invalid;
-        bool is_texception = false;
+    std::array<u8, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> color_formats;
+    u8 num_color_attachments;
+    u8 texceptions;
 
-        std::size_t Hash() const noexcept {
-            return static_cast<std::size_t>(pixel_format) |
-                   static_cast<std::size_t>(is_texception) << 6 |
-                   static_cast<std::size_t>(index) << 7;
-        }
+    u8 zeta_format;
+    u8 zeta_texception;
 
-        bool operator==(const ColorAttachment& rhs) const noexcept {
-            return std::tie(index, pixel_format, is_texception) ==
-                   std::tie(rhs.index, rhs.pixel_format, rhs.is_texception);
-        }
-    };
+    std::size_t Hash() const noexcept;
 
-    boost::container::static_vector<ColorAttachment,
-                                    Tegra::Engines::Maxwell3D::Regs::NumRenderTargets>
-        color_attachments{};
-    // TODO(Rodrigo): Unify has_zeta into zeta_pixel_format and zeta_component_type.
-    VideoCore::Surface::PixelFormat zeta_pixel_format = VideoCore::Surface::PixelFormat::Invalid;
-    bool has_zeta = false;
-    bool zeta_texception = false;
+    bool operator==(const RenderPassParams& rhs) const noexcept;
 
-    std::size_t Hash() const noexcept {
-        std::size_t hash = 0;
-        for (const auto& rt : color_attachments) {
-            boost::hash_combine(hash, rt.Hash());
-        }
-        boost::hash_combine(hash, zeta_pixel_format);
-        boost::hash_combine(hash, has_zeta);
-        boost::hash_combine(hash, zeta_texception);
-        return hash;
-    }
-
-    bool operator==(const RenderPassParams& rhs) const {
-        return std::tie(color_attachments, zeta_pixel_format, has_zeta, zeta_texception) ==
-               std::tie(rhs.color_attachments, rhs.zeta_pixel_format, rhs.has_zeta,
-                        rhs.zeta_texception);
+    bool operator!=(const RenderPassParams& rhs) const noexcept {
+        return !operator==(rhs);
     }
 };
+static_assert(std::has_unique_object_representations_v<RenderPassParams>);
+static_assert(std::is_trivially_copyable_v<RenderPassParams>);
+static_assert(std::is_trivially_constructible_v<RenderPassParams>);
 
 } // namespace Vulkan
 

src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp

@@ -39,8 +39,7 @@ VKStagingBufferPool::StagingBuffer& VKStagingBufferPool::StagingBuffer::operator
 
 VKStagingBufferPool::VKStagingBufferPool(const VKDevice& device, VKMemoryManager& memory_manager,
                                          VKScheduler& scheduler)
-    : device{device}, memory_manager{memory_manager}, scheduler{scheduler},
-      is_device_integrated{device.IsIntegrated()} {}
+    : device{device}, memory_manager{memory_manager}, scheduler{scheduler} {}
 
 VKStagingBufferPool::~VKStagingBufferPool() = default;
 
@@ -56,9 +55,7 @@ void VKStagingBufferPool::TickFrame() {
     current_delete_level = (current_delete_level + 1) % NumLevels;
 
     ReleaseCache(true);
-    if (!is_device_integrated) {
-        ReleaseCache(false);
-    }
+    ReleaseCache(false);
 }
 
 VKBuffer* VKStagingBufferPool::TryGetReservedBuffer(std::size_t size, bool host_visible) {
@@ -81,7 +78,7 @@ VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_v
     ci.size = 1ULL << log2;
     ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
                VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
-               VK_BUFFER_USAGE_INDEX_BUFFER_BIT;
+               VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
     ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
     ci.queueFamilyIndexCount = 0;
     ci.pQueueFamilyIndices = nullptr;
@@ -95,7 +92,7 @@ VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_v
 }
 
 VKStagingBufferPool::StagingBuffersCache& VKStagingBufferPool::GetCache(bool host_visible) {
-    return is_device_integrated || host_visible ? host_staging_buffers : device_staging_buffers;
+    return host_visible ? host_staging_buffers : device_staging_buffers;
 }
 
 void VKStagingBufferPool::ReleaseCache(bool host_visible) {

src/video_core/renderer_vulkan/vk_staging_buffer_pool.h

@@ -71,7 +71,6 @@ private:
     const VKDevice& device;
     VKMemoryManager& memory_manager;
     VKScheduler& scheduler;
-    const bool is_device_integrated;
 
     StagingBuffersCache host_staging_buffers;
     StagingBuffersCache device_staging_buffers;

src/video_core/renderer_vulkan/wrapper.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <exception>
 #include <memory>
 #include <optional>
@@ -16,6 +17,23 @@ namespace Vulkan::vk {
 
 namespace {
 
+void SortPhysicalDevices(std::vector<VkPhysicalDevice>& devices, const InstanceDispatch& dld) {
+    std::stable_sort(devices.begin(), devices.end(), [&](auto lhs, auto rhs) {
+        // This will call Vulkan more than needed, but these calls are cheap.
+        const auto lhs_properties = vk::PhysicalDevice(lhs, dld).GetProperties();
+        const auto rhs_properties = vk::PhysicalDevice(rhs, dld).GetProperties();
+
+        // Prefer discrete GPUs, Nvidia over AMD, AMD over Intel, Intel over the rest.
+        const bool preferred =
+            (lhs_properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU &&
+             rhs_properties.deviceType != VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU) ||
+            (lhs_properties.vendorID == 0x10DE && rhs_properties.vendorID != 0x10DE) ||
+            (lhs_properties.vendorID == 0x1002 && rhs_properties.vendorID != 0x1002) ||
+            (lhs_properties.vendorID == 0x8086 && rhs_properties.vendorID != 0x8086);
+        return !preferred;
+    });
+}
+
 template <typename T>
 bool Proc(T& result, const InstanceDispatch& dld, const char* proc_name,
           VkInstance instance = nullptr) noexcept {
@@ -63,6 +81,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCmdSetBlendConstants);
     X(vkCmdSetDepthBias);
     X(vkCmdSetDepthBounds);
+    X(vkCmdSetEvent);
     X(vkCmdSetScissor);
     X(vkCmdSetStencilCompareMask);
     X(vkCmdSetStencilReference);
@@ -75,6 +94,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCreateDescriptorPool);
     X(vkCreateDescriptorSetLayout);
     X(vkCreateDescriptorUpdateTemplateKHR);
+    X(vkCreateEvent);
     X(vkCreateFence);
     X(vkCreateFramebuffer);
     X(vkCreateGraphicsPipelines);
@@ -93,6 +113,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkDestroyDescriptorPool);
     X(vkDestroyDescriptorSetLayout);
     X(vkDestroyDescriptorUpdateTemplateKHR);
+    X(vkDestroyEvent);
     X(vkDestroyFence);
     X(vkDestroyFramebuffer);
     X(vkDestroyImage);
@@ -112,6 +133,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkFreeMemory);
     X(vkGetBufferMemoryRequirements);
     X(vkGetDeviceQueue);
+    X(vkGetEventStatus);
     X(vkGetFenceStatus);
     X(vkGetImageMemoryRequirements);
     X(vkGetQueryPoolResults);
@@ -269,6 +291,10 @@ void Destroy(VkDevice device, VkDeviceMemory handle, const DeviceDispatch& dld)
     dld.vkFreeMemory(device, handle, nullptr);
 }
 
+void Destroy(VkDevice device, VkEvent handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyEvent(device, handle, nullptr);
+}
+
 void Destroy(VkDevice device, VkFence handle, const DeviceDispatch& dld) noexcept {
     dld.vkDestroyFence(device, handle, nullptr);
 }
@@ -381,7 +407,8 @@ std::optional<std::vector<VkPhysicalDevice>> Instance::EnumeratePhysicalDevices(
     if (dld->vkEnumeratePhysicalDevices(handle, &num, physical_devices.data()) != VK_SUCCESS) {
         return std::nullopt;
     }
-    return physical_devices;
+    SortPhysicalDevices(physical_devices, *dld);
+    return std::make_optional(std::move(physical_devices));
 }
 
 DebugCallback Instance::TryCreateDebugCallback(
@@ -599,6 +626,16 @@ ShaderModule Device::CreateShaderModule(const VkShaderModuleCreateInfo& ci) cons
     return ShaderModule(object, handle, *dld);
 }
 
+Event Device::CreateEvent() const {
+    VkEventCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_EVENT_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    VkEvent object;
+    Check(dld->vkCreateEvent(handle, &ci, nullptr, &object));
+    return Event(object, handle, *dld);
+}
+
 SwapchainKHR Device::CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const {
     VkSwapchainKHR object;
     Check(dld->vkCreateSwapchainKHR(handle, &ci, nullptr, &object));

src/video_core/renderer_vulkan/wrapper.h

@@ -199,6 +199,7 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkCmdSetBlendConstants vkCmdSetBlendConstants;
     PFN_vkCmdSetDepthBias vkCmdSetDepthBias;
     PFN_vkCmdSetDepthBounds vkCmdSetDepthBounds;
+    PFN_vkCmdSetEvent vkCmdSetEvent;
     PFN_vkCmdSetScissor vkCmdSetScissor;
     PFN_vkCmdSetStencilCompareMask vkCmdSetStencilCompareMask;
     PFN_vkCmdSetStencilReference vkCmdSetStencilReference;
@@ -211,6 +212,7 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkCreateDescriptorPool vkCreateDescriptorPool;
     PFN_vkCreateDescriptorSetLayout vkCreateDescriptorSetLayout;
     PFN_vkCreateDescriptorUpdateTemplateKHR vkCreateDescriptorUpdateTemplateKHR;
+    PFN_vkCreateEvent vkCreateEvent;
     PFN_vkCreateFence vkCreateFence;
     PFN_vkCreateFramebuffer vkCreateFramebuffer;
     PFN_vkCreateGraphicsPipelines vkCreateGraphicsPipelines;
@@ -229,6 +231,7 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkDestroyDescriptorPool vkDestroyDescriptorPool;
     PFN_vkDestroyDescriptorSetLayout vkDestroyDescriptorSetLayout;
     PFN_vkDestroyDescriptorUpdateTemplateKHR vkDestroyDescriptorUpdateTemplateKHR;
+    PFN_vkDestroyEvent vkDestroyEvent;
     PFN_vkDestroyFence vkDestroyFence;
     PFN_vkDestroyFramebuffer vkDestroyFramebuffer;
     PFN_vkDestroyImage vkDestroyImage;
@@ -248,6 +251,7 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkFreeMemory vkFreeMemory;
     PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements;
     PFN_vkGetDeviceQueue vkGetDeviceQueue;
+    PFN_vkGetEventStatus vkGetEventStatus;
     PFN_vkGetFenceStatus vkGetFenceStatus;
     PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements;
     PFN_vkGetQueryPoolResults vkGetQueryPoolResults;
@@ -279,6 +283,7 @@ void Destroy(VkDevice, VkDescriptorPool, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkDescriptorSetLayout, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkDescriptorUpdateTemplateKHR, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkDeviceMemory, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkEvent, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkFence, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkFramebuffer, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkImage, const DeviceDispatch&) noexcept;
@@ -648,6 +653,15 @@ public:
     std::vector<VkImage> GetImages() const;
 };
 
+class Event : public Handle<VkEvent, VkDevice, DeviceDispatch> {
+    using Handle<VkEvent, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    VkResult GetStatus() const noexcept {
+        return dld->vkGetEventStatus(owner, handle);
+    }
+};
+
 class Device : public Handle<VkDevice, NoOwner, DeviceDispatch> {
     using Handle<VkDevice, NoOwner, DeviceDispatch>::Handle;
 
@@ -695,6 +709,8 @@ public:
 
     ShaderModule CreateShaderModule(const VkShaderModuleCreateInfo& ci) const;
 
+    Event CreateEvent() const;
+
     SwapchainKHR CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const;
 
     DeviceMemory TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept;
@@ -938,6 +954,10 @@ public:
         dld->vkCmdSetDepthBounds(handle, min_depth_bounds, max_depth_bounds);
     }
 
+    void SetEvent(VkEvent event, VkPipelineStageFlags stage_flags) const noexcept {
+        dld->vkCmdSetEvent(handle, event, stage_flags);
+    }
+
     void BindTransformFeedbackBuffersEXT(u32 first, u32 count, const VkBuffer* buffers,
                                          const VkDeviceSize* offsets,
                                          const VkDeviceSize* sizes) const noexcept {