Bump CI clang version to 15 + workaround for unsafe fiber optimizatio…

…ns (#982)

.github/workflows/build.yml

@@ -16,7 +16,7 @@ env:
 
 jobs:
   build-ubuntu:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     steps:
     - name: "Checkout repo"
       uses: actions/checkout@v3
@@ -53,7 +53,7 @@ jobs:
     - name: "Install system dependencies"
       run: |
         sudo apt update -qq
-        sudo apt install -y clang-12 cmake freeglut3-dev libgcrypt20-dev libglm-dev libgtk-3-dev libpulse-dev libsecret-1-dev libsystemd-dev libudev-dev nasm ninja-build
+        sudo apt install -y clang-15 cmake freeglut3-dev libgcrypt20-dev libglm-dev libgtk-3-dev libpulse-dev libsecret-1-dev libsystemd-dev libudev-dev nasm ninja-build
 
     - name: "Bootstrap vcpkg"
       run: |
@@ -75,7 +75,7 @@ jobs:
         
     - name: "cmake"
       run: |
-        cmake -S . -B build ${{ env.BUILD_FLAGS }} -DCMAKE_BUILD_TYPE=${{ env.BUILD_MODE }} -DPORTABLE=OFF -DCMAKE_C_COMPILER=/usr/bin/clang-12 -DCMAKE_CXX_COMPILER=/usr/bin/clang++-12 -G Ninja -DCMAKE_MAKE_PROGRAM=/usr/bin/ninja
+        cmake -S . -B build ${{ env.BUILD_FLAGS }} -DCMAKE_BUILD_TYPE=${{ env.BUILD_MODE }} -DPORTABLE=OFF -DCMAKE_C_COMPILER=/usr/bin/clang-15 -DCMAKE_CXX_COMPILER=/usr/bin/clang++-15 -G Ninja -DCMAKE_MAKE_PROGRAM=/usr/bin/ninja
         
     - name: "Build Cemu"
       run: |
@@ -93,7 +93,7 @@ jobs:
         path: ./bin/Cemu
 
   build-appimage:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     needs: build-ubuntu
     steps:
     - name: Checkout Upstream Repo
@@ -107,7 +107,7 @@ jobs:
     - name: "Install system dependencies"
       run: |
         sudo apt update -qq
-        sudo apt install -y clang-12 cmake freeglut3-dev libgcrypt20-dev libglm-dev libgtk-3-dev libpulse-dev libsecret-1-dev libsystemd-dev nasm ninja-build appstream
+        sudo apt install -y clang-15 cmake freeglut3-dev libgcrypt20-dev libglm-dev libgtk-3-dev libpulse-dev libsecret-1-dev libsystemd-dev nasm ninja-build appstream
 
     - name: "Build AppImage"
       run: |

.github/workflows/deploy_experimental_release.yml

@@ -10,7 +10,7 @@ jobs:
       experimentalversion: ${{ github.run_number }}
   deploy:
     name: Deploy experimental release
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     needs: call-release-build
     steps:
       - uses: actions/checkout@v3
@@ -72,7 +72,7 @@ jobs:
           ls ./bin/
           cp -R ./bin ./${{ env.CEMU_FOLDER_NAME }}
           mv cemu-bin-linux-x64/Cemu ./${{ env.CEMU_FOLDER_NAME }}/Cemu
-          zip -9 -r upload/cemu-${{ env.CEMU_VERSION }}-ubuntu-20.04-x64.zip ${{ env.CEMU_FOLDER_NAME }}
+          zip -9 -r upload/cemu-${{ env.CEMU_VERSION }}-ubuntu-22.04-x64.zip ${{ env.CEMU_FOLDER_NAME }}
           rm -r ./${{ env.CEMU_FOLDER_NAME }}
           
       - name: Create release from macos-bin

.gitignore

@@ -17,7 +17,7 @@
 .idea/
 
 build/
-cmake-build-*-*/
+cmake-build-*/
 out/
 .cache/
 bin/Cemu_*

BUILD.md

@@ -19,17 +19,17 @@ Any other IDE should also work as long as it has CMake and MSVC support. CLion a
 
 ## Linux
 
-To compile Cemu, a recent enough compiler and STL with C++20 support is required! clang-12 or higher is what we recommend.
+To compile Cemu, a recent enough compiler and STL with C++20 support is required! clang-15 or higher is what we recommend.
 
 ### Installing dependencies
 
 #### For Ubuntu and derivatives:
-`sudo apt install -y cmake curl freeglut3-dev git libgcrypt20-dev libglm-dev libgtk-3-dev libpulse-dev libsecret-1-dev libsystemd-dev nasm ninja-build` 
+`sudo apt install -y cmake curl clang-15 freeglut3-dev git libgcrypt20-dev libglm-dev libgtk-3-dev libpulse-dev libsecret-1-dev libsystemd-dev nasm ninja-build`
 
-*Additionally, for Ubuntu 22.04 only:*
- - `sudo apt install -y clang-12`
- - At step 3 while building, use  
-   `cmake -S . -B build -DCMAKE_BUILD_TYPE=release -DCMAKE_C_COMPILER=/usr/bin/clang-12 -DCMAKE_CXX_COMPILER=/usr/bin/clang++-12 -G Ninja -DCMAKE_MAKE_PROGRAM=/usr/bin/ninja`
+You may also need to install `libusb-1.0-0-dev` as a workaround for an issue with the vcpkg hidapi package.
+
+At step 3 while building, use:
+   `cmake -S . -B build -DCMAKE_BUILD_TYPE=release -DCMAKE_C_COMPILER=/usr/bin/clang-15 -DCMAKE_CXX_COMPILER=/usr/bin/clang++-15 -G Ninja -DCMAKE_MAKE_PROGRAM=/usr/bin/ninja`
 
 #### For Arch and derivatives:
 `sudo pacman -S --needed base-devel clang cmake freeglut git glm gtk3 libgcrypt libpulse libsecret linux-headers llvm nasm ninja systemd unzip zip`

src/Cafe/HW/Espresso/Debugger/Debugger.cpp

@@ -210,7 +210,8 @@ void debugger_handleSingleStepException(uint64 dr6)
 	}
 	if (catchBP)
 	{
-		debugger_createCodeBreakpoint(ppcInterpreterCurrentInstance->instructionPointer + 4, DEBUGGER_BP_T_ONE_SHOT);
+		PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
+		debugger_createCodeBreakpoint(hCPU->instructionPointer + 4, DEBUGGER_BP_T_ONE_SHOT);
 	}
 }
 

src/Cafe/HW/Espresso/Debugger/GDBStub.cpp

@@ -959,8 +959,9 @@ void GDBServer::HandleAccessException(uint64 dr6)
 
 	if (!response.empty())
 	{
+		PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
 		cemuLog_logDebug(LogType::Force, "Received matching breakpoint exception: {}", response);
-		auto nextInstructions = findNextInstruction(ppcInterpreterCurrentInstance->instructionPointer, ppcInterpreterCurrentInstance->spr.LR, ppcInterpreterCurrentInstance->spr.CTR);
+		auto nextInstructions = findNextInstruction(hCPU->instructionPointer, hCPU->spr.LR, hCPU->spr.CTR);
 		for (MPTR nextInstr : nextInstructions)
 		{
 			auto bpIt = m_patchedInstructions.find(nextInstr);

src/Cafe/HW/Espresso/Interpreter/PPCInterpreterMain.cpp

@@ -6,7 +6,6 @@
 thread_local PPCInterpreter_t* ppcInterpreterCurrentInstance;
 
 // main thread instruction counter and timing
-volatile uint64 ppcMainThreadCycleCounter = 0;
 uint64 ppcMainThreadDECCycleValue = 0; // value that was set to dec register
 uint64 ppcMainThreadDECCycleStart = 0; // at which cycle the dec register was set, if == 0 -> dec is 0
 uint64 ppcCyclesSince2000 = 0;
@@ -29,11 +28,16 @@ PPCInterpreter_t* PPCInterpreter_createInstance(unsigned int Entrypoint)
 	return pData;
 }
 
-PPCInterpreter_t* PPCInterpreter_getCurrentInstance()
+TLS_WORKAROUND_NOINLINE PPCInterpreter_t* PPCInterpreter_getCurrentInstance()
 {
 	return ppcInterpreterCurrentInstance;
 }
 
+TLS_WORKAROUND_NOINLINE void PPCInterpreter_setCurrentInstance(PPCInterpreter_t* hCPU)
+{
+	ppcInterpreterCurrentInstance = hCPU;
+}
+
 uint64 PPCInterpreter_getMainCoreCycleCounter()
 {
 	return PPCTimer_getFromRDTSC();
@@ -78,24 +82,25 @@ uint32 PPCInterpreter_getCoreIndex(PPCInterpreter_t* hCPU)
 
 uint32 PPCInterpreter_getCurrentCoreIndex()
 {
-	return ppcInterpreterCurrentInstance->spr.UPIR;
+	return PPCInterpreter_getCurrentInstance()->spr.UPIR;
 };
 
 uint8* PPCInterpreterGetStackPointer()
 {
-	return memory_getPointerFromVirtualOffset(ppcInterpreterCurrentInstance->gpr[1]);
+	return memory_getPointerFromVirtualOffset(PPCInterpreter_getCurrentInstance()->gpr[1]);
 }
 
 uint8* PPCInterpreterGetAndModifyStackPointer(sint32 offset)
 {
-	uint8* result = memory_getPointerFromVirtualOffset(ppcInterpreterCurrentInstance->gpr[1] - offset);
-	ppcInterpreterCurrentInstance->gpr[1] -= offset;
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
+	uint8* result = memory_getPointerFromVirtualOffset(hCPU->gpr[1] - offset);
+	hCPU->gpr[1] -= offset;
 	return result;
 }
 
 void PPCInterpreterModifyStackPointer(sint32 offset)
 {
-	ppcInterpreterCurrentInstance->gpr[1] -= offset;
+	PPCInterpreter_getCurrentInstance()->gpr[1] -= offset;
 }
 
 uint32 RPLLoader_MakePPCCallable(void(*ppcCallableExport)(PPCInterpreter_t* hCPU));

src/Cafe/HW/Espresso/PPCCallback.h

@@ -18,19 +18,20 @@ uint32 PPCCoreCallback(MPTR function, PPCCoreCallbackData_t& data, T currentArg,
 {
 	cemu_assert_debug(data.gprCount <= 8);
 	cemu_assert_debug(data.floatCount <= 8);
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
 	if constexpr (std::is_pointer_v<T>)
 	{
-		ppcInterpreterCurrentInstance->gpr[3 + data.gprCount] = MEMPTR(currentArg).GetMPTR();
+		hCPU->gpr[3 + data.gprCount] = MEMPTR(currentArg).GetMPTR();
 		data.gprCount++;
 	}
 	else if constexpr (std::is_base_of_v<MEMPTRBase, std::remove_reference_t<T>>)
 	{
-		ppcInterpreterCurrentInstance->gpr[3 + data.gprCount] = currentArg.GetMPTR();
+		hCPU->gpr[3 + data.gprCount] = currentArg.GetMPTR();
 		data.gprCount++;
 	}
 	else if constexpr (std::is_reference_v<T>)
 	{
-		ppcInterpreterCurrentInstance->gpr[3 + data.gprCount] = MEMPTR(&currentArg).GetMPTR();
+		hCPU->gpr[3 + data.gprCount] = MEMPTR(&currentArg).GetMPTR();
 		data.gprCount++;
 	}
 	else if constexpr(std::is_enum_v<T>)
@@ -40,19 +41,19 @@ uint32 PPCCoreCallback(MPTR function, PPCCoreCallbackData_t& data, T currentArg,
 	}
 	else if constexpr (std::is_floating_point_v<T>)
 	{
-		ppcInterpreterCurrentInstance->fpr[1 + data.floatCount].fpr = (double)currentArg;
+		hCPU->fpr[1 + data.floatCount].fpr = (double)currentArg;
 		data.floatCount++;
 	}
 	else if constexpr (std::is_integral_v<T> && sizeof(T) == sizeof(uint64))
 	{
-		ppcInterpreterCurrentInstance->gpr[3 + data.gprCount] = (uint32)(currentArg >> 32); // high
-		ppcInterpreterCurrentInstance->gpr[3 + data.gprCount + 1] = (uint32)currentArg; // low
+		hCPU->gpr[3 + data.gprCount] = (uint32)(currentArg >> 32); // high
+		hCPU->gpr[3 + data.gprCount + 1] = (uint32)currentArg; // low
 
 		data.gprCount += 2;
 	}
 	else
 	{
-		ppcInterpreterCurrentInstance->gpr[3 + data.gprCount] = (uint32)currentArg;
+		hCPU->gpr[3 + data.gprCount] = (uint32)currentArg;
 		data.gprCount++;
 	}
 

src/Cafe/HW/Espresso/PPCScheduler.cpp

@@ -11,21 +11,24 @@ uint32 ppcThreadQuantum = 45000; // execute 45000 instructions before thread res
 
 void PPCInterpreter_relinquishTimeslice()
 {
-	if( ppcInterpreterCurrentInstance->remainingCycles >= 0 )
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
+	if( hCPU->remainingCycles >= 0 )
 	{
-		ppcInterpreterCurrentInstance->skippedCycles = ppcInterpreterCurrentInstance->remainingCycles + 1;
-		ppcInterpreterCurrentInstance->remainingCycles = -1;
+		hCPU->skippedCycles = hCPU->remainingCycles + 1;
+		hCPU->remainingCycles = -1;
 	}
 }
 
 void PPCCore_boostQuantum(sint32 numCycles)
 {
-	ppcInterpreterCurrentInstance->remainingCycles += numCycles;
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
+	hCPU->remainingCycles += numCycles;
 }
 
 void PPCCore_deboostQuantum(sint32 numCycles)
 {
-	ppcInterpreterCurrentInstance->remainingCycles -= numCycles;
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
+	hCPU->remainingCycles -= numCycles;
 }
 
 namespace coreinit
@@ -36,7 +39,7 @@ namespace coreinit
 void PPCCore_switchToScheduler()
 {
 	cemu_assert_debug(__OSHasSchedulerLock() == false); // scheduler lock must not be hold past thread time slice
-	cemu_assert_debug(ppcInterpreterCurrentInstance->coreInterruptMask != 0 || CafeSystem::GetForegroundTitleId() == 0x000500001019e600);
+	cemu_assert_debug(PPCInterpreter_getCurrentInstance()->coreInterruptMask != 0 || CafeSystem::GetForegroundTitleId() == 0x000500001019e600);
 	__OSLockScheduler();
 	coreinit::__OSThreadSwitchToNext();
 	__OSUnlockScheduler();
@@ -45,7 +48,7 @@ void PPCCore_switchToScheduler()
 void PPCCore_switchToSchedulerWithLock()
 {
 	cemu_assert_debug(__OSHasSchedulerLock() == true); // scheduler lock must be hold
-	cemu_assert_debug(ppcInterpreterCurrentInstance->coreInterruptMask != 0 || CafeSystem::GetForegroundTitleId() == 0x000500001019e600);
+	cemu_assert_debug(PPCInterpreter_getCurrentInstance()->coreInterruptMask != 0 || CafeSystem::GetForegroundTitleId() == 0x000500001019e600);
 	coreinit::__OSThreadSwitchToNext();
 }
 
@@ -58,7 +61,7 @@ void _PPCCore_callbackExit(PPCInterpreter_t* hCPU)
 PPCInterpreter_t* PPCCore_executeCallbackInternal(uint32 functionMPTR)
 {
 	cemu_assert_debug(functionMPTR != 0);
-	PPCInterpreter_t* hCPU = ppcInterpreterCurrentInstance;
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
 	// remember LR and instruction pointer
 	uint32 lr = hCPU->spr.LR;
 	uint32 ip = hCPU->instructionPointer;

src/Cafe/HW/Espresso/PPCSchedulerLLE.cpp

@@ -220,7 +220,7 @@ void PPCCoreLLE_startSingleCoreScheduler(uint32 entrypoint)
 		for (uint32 coreIndex = 0; coreIndex < 3; coreIndex++)
 		{
 			PPCInterpreter_t* hCPU = cpuContext->cores+coreIndex;
-			ppcInterpreterCurrentInstance = hCPU;
+			PPCInterpreter_setCurrentInstance(hCPU);
 			if (coreIndex == 1)
 			{
 				// check SCR core 1 enable bit

src/Cafe/HW/Espresso/PPCState.h

@@ -149,6 +149,7 @@ static uint64 PPCInterpreter_getCallParamU64(PPCInterpreter_t* hCPU, uint32 inde
 
 PPCInterpreter_t* PPCInterpreter_createInstance(unsigned int Entrypoint);
 PPCInterpreter_t* PPCInterpreter_getCurrentInstance();
+void PPCInterpreter_setCurrentInstance(PPCInterpreter_t* hCPU);
 
 uint64 PPCInterpreter_getMainCoreCycleCounter();
 
@@ -192,7 +193,6 @@ uint32 PPCInterpreter_getCurrentCoreIndex();
 void PPCInterpreter_setDEC(PPCInterpreter_t* hCPU, uint32 newValue);
 
 // timing for main processor
-extern volatile uint64 ppcMainThreadCycleCounter;
 extern uint64 ppcCyclesSince2000; // on init this is set to the cycles that passed since 1.1.2000
 extern uint64 ppcCyclesSince2000TimerClock; // on init this is set to the cycles that passed since 1.1.2000 / 20
 extern uint64 ppcCyclesSince2000_UTC;
@@ -213,7 +213,6 @@ void PPCTimer_start();
 // core info and control
 extern uint32 ppcThreadQuantum;
 
-extern thread_local PPCInterpreter_t *ppcInterpreterCurrentInstance;
 uint8* PPCInterpreterGetAndModifyStackPointer(sint32 offset);
 uint8* PPCInterpreterGetStackPointer();
 void PPCInterpreterModifyStackPointer(sint32 offset);

src/Cafe/HW/Espresso/Recompiler/PPCRecompilerX64.cpp

@@ -100,7 +100,7 @@ void* ATTR_MS_ABI PPCRecompiler_virtualHLE(PPCInterpreter_t* hCPU, uint32 hleFun
 		hCPU->remainingCycles -= 500; // let subtract about 500 cycles for each HLE call
 		hCPU->gpr[3] = 0;
 		PPCInterpreter_nextInstruction(hCPU);
-		return ppcInterpreterCurrentInstance;
+		return hCPU;
 	}
 	else
 	{
@@ -109,7 +109,7 @@ void* ATTR_MS_ABI PPCRecompiler_virtualHLE(PPCInterpreter_t* hCPU, uint32 hleFun
 		hleCall(hCPU);
 	}
 	hCPU->rspTemp = prevRSPTemp;
-	return ppcInterpreterCurrentInstance;
+	return PPCInterpreter_getCurrentInstance();
 }
 
 void ATTR_MS_ABI PPCRecompiler_getTBL(PPCInterpreter_t* hCPU, uint32 gprIndex)

src/Cafe/IOSU/legacy/iosu_ioctl.cpp

@@ -23,7 +23,7 @@ sint32 iosuIoctl_pushAndWait(uint32 ioctlHandle, ioQueueEntry_t* ioQueueEntry)
 	}
 	__OSLockScheduler();
 	ioctlMutex.lock();
-	ioQueueEntry->ppcThread = coreinitThread_getCurrentThreadDepr(ppcInterpreterCurrentInstance);
+	ioQueueEntry->ppcThread = coreinitThread_getCurrentThreadDepr(PPCInterpreter_getCurrentInstance());
 
 	_ioctlRingbuffer[ioctlHandle].Push(ioQueueEntry);
 	ioctlMutex.unlock();

src/Cafe/OS/common/OSUtil.h

@@ -65,7 +65,7 @@ class cafeExportParamWrapper
 		}
 		else if constexpr (std::is_floating_point_v<T>)
 		{
-			v = (T)ppcInterpreterCurrentInstance->fpr[1 + fprIndex].fpr;
+			v = (T)hCPU->fpr[1 + fprIndex].fpr;
 			fprIndex++;
 		}
 		else

src/Cafe/OS/libs/coreinit/coreinit.cpp

@@ -204,7 +204,7 @@ namespace coreinit
 {
 	sint32 OSGetCoreId()
 	{
-		return PPCInterpreter_getCoreIndex(ppcInterpreterCurrentInstance);
+		return PPCInterpreter_getCoreIndex(PPCInterpreter_getCurrentInstance());
 	}
 
 	uint32 OSGetCoreCount()
@@ -239,7 +239,7 @@ namespace coreinit
 
 	uint32 OSGetStackPointer()
 	{
-		return ppcInterpreterCurrentInstance->gpr[1];
+		return PPCInterpreter_getCurrentInstance()->gpr[1];
 	}
 
 	void coreinitExport_ENVGetEnvironmentVariable(PPCInterpreter_t* hCPU)

src/Cafe/OS/libs/coreinit/coreinit_FS.cpp

@@ -691,7 +691,7 @@ namespace coreinit
 			while (OSSendMessage(ioMsgQueue, &fsCmdBlockBody->asyncResult.msgUnion.osMsg, 0) == 0)
 			{
 				cemuLog_log(LogType::Force, "FS driver: Failed to add message to result queue. Retrying...");
-				if (ppcInterpreterCurrentInstance)
+				if (PPCInterpreter_getCurrentInstance())
 					PPCCore_switchToScheduler();
 				else
 					std::this_thread::sleep_for(std::chrono::milliseconds(10));

src/Cafe/OS/libs/coreinit/coreinit_Init.cpp

@@ -142,10 +142,11 @@ void CafeInit()
 		}
 	}
 	// setup UGQR
-	ppcInterpreterCurrentInstance->spr.UGQR[0 + 2] = 0x00040004;
-	ppcInterpreterCurrentInstance->spr.UGQR[0 + 3] = 0x00050005;
-	ppcInterpreterCurrentInstance->spr.UGQR[0 + 4] = 0x00060006;
-	ppcInterpreterCurrentInstance->spr.UGQR[0 + 5] = 0x00070007;
+	PPCInterpreter_t* hCPU = PPCInterpreter_getCurrentInstance();
+	hCPU->spr.UGQR[0 + 2] = 0x00040004;
+	hCPU->spr.UGQR[0 + 3] = 0x00050005;
+	hCPU->spr.UGQR[0 + 4] = 0x00060006;
+	hCPU->spr.UGQR[0 + 5] = 0x00070007;
 	coreinit::InitForegroundBucket();
 	coreinit::InitSysHeap();
 }