Merge 974f7eb721 into cf87f24587

3rdparty/7zip/CMakeLists.txt

@@ -59,11 +59,11 @@ if(WIN32 OR APPLE)
 		7zip/C/XzEnc.c
 		7zip/C/XzIn.c
 		7zip/C/ZstdDec.c)
-	target_include_directories(3rdparty_7zip INTERFACE
+	target_include_directories(3rdparty_7zip SYSTEM INTERFACE
 		$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/7zip/C>
 		$<INSTALL_INTERFACE:/7zip/C>)
 
-	target_include_directories(3rdparty_7zip INTERFACE 7zip)
+	target_include_directories(3rdparty_7zip SYSTEM INTERFACE 7zip)
 
 	set_property(TARGET 3rdparty_7zip PROPERTY FOLDER "3rdparty/")
 

3rdparty/CMakeLists.txt

@@ -35,7 +35,7 @@ if (USE_SYSTEM_FLATBUFFERS)
 		message(FATAL_ERROR "flatc failed to regenerate flatbuffers headers.")
 	endif()
 else()
-	target_include_directories(3rdparty_flatbuffers INTERFACE flatbuffers/include)
+	target_include_directories(3rdparty_flatbuffers SYSTEM INTERFACE flatbuffers/include)
 endif()
 
 # libPNG
@@ -56,7 +56,7 @@ if (USE_SYSTEM_VULKAN_MEMORY_ALLOCATOR)
 	add_library(3rdparty::vulkanmemoryallocator ALIAS GPUOpen::VulkanMemoryAllocator)
 else()
 	add_library(3rdparty_vulkanmemoryallocator INTERFACE)
-	target_include_directories(3rdparty_vulkanmemoryallocator INTERFACE GPUOpen/VulkanMemoryAllocator/include)
+	target_include_directories(3rdparty_vulkanmemoryallocator SYSTEM INTERFACE GPUOpen/VulkanMemoryAllocator/include)
 	add_library(3rdparty::vulkanmemoryallocator ALIAS 3rdparty_vulkanmemoryallocator)
 endif()
 
@@ -111,7 +111,7 @@ if (NOT ANDROID)
 	find_package(OpenGL REQUIRED OPTIONAL_COMPONENTS EGL)
 
 	add_library(3rdparty_opengl INTERFACE)
-	target_include_directories(3rdparty_opengl INTERFACE GL)
+	target_include_directories(3rdparty_opengl SYSTEM INTERFACE GL)
 
 	if (WIN32)
 		if(NOT MSVC)
@@ -204,7 +204,7 @@ if(USE_VULKAN)
 			find_package(Wayland)
 			if (WAYLAND_FOUND)
 				target_include_directories(3rdparty_vulkan
-					INTERFACE ${WAYLAND_INCLUDE_DIR})
+					SYSTEM INTERFACE ${WAYLAND_INCLUDE_DIR})
 			endif()
 		endif()
 
@@ -298,7 +298,7 @@ if(NOT ANDROID)
 		message(STATUS "RPCS3: using shared ffmpeg")
 		find_package(FFMPEG REQUIRED)
 
-		target_include_directories(3rdparty_ffmpeg INTERFACE ${FFMPEG_INCLUDE_DIR})
+		target_include_directories(3rdparty_ffmpeg SYSTEM INTERFACE ${FFMPEG_INCLUDE_DIR})
 		target_link_libraries(3rdparty_ffmpeg INTERFACE ${FFMPEG_LIBRARIES})
 	else()
 		message(STATUS "RPCS3: using builtin ffmpeg")
@@ -328,7 +328,7 @@ if(NOT ANDROID)
 			${FFMPEG_LIB_SWSCALE}
 			${FFMPEG_LIB_SWRESAMPLE}
 		)
-		target_include_directories(3rdparty_ffmpeg INTERFACE "ffmpeg/include")
+		target_include_directories(3rdparty_ffmpeg SYSTEM INTERFACE "ffmpeg/include")
 	endif()
 endif()
 

3rdparty/SoundTouch/CMakeLists.txt

@@ -11,11 +11,11 @@ add_library(soundtouch STATIC EXCLUDE_FROM_ALL
 	soundtouch/source/SoundTouch/TDStretch.cpp
 )
 
-target_include_directories(soundtouch PRIVATE
+target_include_directories(soundtouch SYSTEM PRIVATE
 	soundtouch/source/SoundTouch
 	soundtouch/include)
 
-target_include_directories(soundtouch INTERFACE
+target_include_directories(soundtouch SYSTEM INTERFACE
 	$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/soundtouch/include>
 	$<INSTALL_INTERFACE:/soundtouch/include>)
 

3rdparty/asmjit/CMakeLists.txt

@@ -9,7 +9,7 @@ set(ASMJIT_DIR "${CMAKE_CURRENT_SOURCE_DIR}/asmjit" CACHE PATH "Location of 'asm
 include("${ASMJIT_DIR}/CMakeLists.txt")
 
 add_library(asmjit ${ASMJIT_SRC})
-target_include_directories(asmjit PUBLIC ${ASMJIT_DIR}/src)
+target_include_directories(asmjit SYSTEM PUBLIC ${ASMJIT_DIR}/src)
 target_link_libraries(asmjit PRIVATE ${ASMJIT_DEPS})
 
 # ASMJIT should have a option for disabling installing and this wouldnt

3rdparty/discord-rpc/CMakeLists.txt

@@ -9,7 +9,7 @@ if (USE_DISCORD_RPC AND (WIN32 OR CMAKE_SYSTEM MATCHES "Linux" OR APPLE))
 	set(WARNINGS_AS_ERRORS FALSE CACHE BOOL "When enabled, compiles with `-Werror` (on *nix platforms).")
 
 	add_subdirectory(discord-rpc EXCLUDE_FROM_ALL)
-	target_include_directories(3rdparty_discordRPC INTERFACE discord-rpc/include)
+	target_include_directories(3rdparty_discordRPC SYSTEM INTERFACE discord-rpc/include)
 	target_compile_definitions(3rdparty_discordRPC INTERFACE -DWITH_DISCORD_RPC)
 	target_link_libraries(3rdparty_discordRPC INTERFACE discord-rpc)
 endif()

3rdparty/feralinteractive/CMakeLists.txt

@@ -3,7 +3,7 @@
 add_library(3rdparty_feralinteractive INTERFACE)
 
 if (CMAKE_SYSTEM MATCHES "Linux")
-	target_include_directories(3rdparty_feralinteractive INTERFACE feralinteractive/lib)
+	target_include_directories(3rdparty_feralinteractive SYSTEM INTERFACE feralinteractive/lib)
 	target_compile_definitions(3rdparty_feralinteractive INTERFACE -DGAMEMODE_AVAILABLE)
 	target_link_libraries(3rdparty_feralinteractive INTERFACE feralinteractive)
 endif()

3rdparty/glslang/CMakeLists.txt

@@ -7,7 +7,7 @@ if(USE_SYSTEM_GLSLANG)
   target_link_libraries(3rdparty_glslang INTERFACE glslang::SPIRV)
   get_target_property(SPIRV_INCLUDE_DIRS glslang::SPIRV INTERFACE_INCLUDE_DIRECTORIES)
   list(TRANSFORM SPIRV_INCLUDE_DIRS APPEND "/glslang")
-  target_include_directories(3rdparty_glslang INTERFACE ${SPIRV_INCLUDE_DIRS})
+  target_include_directories(3rdparty_glslang SYSTEM INTERFACE ${SPIRV_INCLUDE_DIRS})
 else()
   set(ENABLE_PCH OFF CACHE BOOL "Enables Precompiled header" FORCE)
   set(BUILD_EXTERNAL OFF CACHE BOOL "Build external dependencies in /External" FORCE)

3rdparty/hidapi/CMakeLists.txt

@@ -4,7 +4,7 @@ if(USE_SYSTEM_HIDAPI)
 	pkg_check_modules(hidapi-hidraw REQUIRED IMPORTED_TARGET hidapi-hidraw)
 	add_library(3rdparty_hidapi INTERFACE)
 	target_link_libraries(3rdparty_hidapi INTERFACE PkgConfig::hidapi-hidraw)
-	target_include_directories(3rdparty_hidapi INTERFACE PkgConfig::hidapi-hidraw)
+	target_include_directories(3rdparty_hidapi SYSTEM INTERFACE PkgConfig::hidapi-hidraw)
 else()
 	set(BUILD_SHARED_LIBS FALSE CACHE BOOL "Don't build shared libs")
 	set(HIDAPI_INSTALL_TARGETS FALSE CACHE BOOL "Don't install anything")

3rdparty/libpng/CMakeLists.txt

@@ -6,14 +6,14 @@ if (NOT USE_SYSTEM_LIBPNG)
 	set(PNG_TESTS OFF CACHE BOOL "Build libpng tests")
 	set(SKIP_INSTALL_ALL ON)
 	add_subdirectory(libpng EXCLUDE_FROM_ALL)
-	target_include_directories(png_static INTERFACE "${libpng_BINARY_DIR}" "${libpng_SOURCE_DIR}")
+	target_include_directories(png_static SYSTEM INTERFACE "${libpng_BINARY_DIR}" "${libpng_SOURCE_DIR}")
 
 	set(LIBPNG_TARGET png_static PARENT_SCOPE)
 else()
 	find_package(PNG REQUIRED)
 
 	add_library(3rdparty_system_libpng INTERFACE)
-	target_include_directories(3rdparty_system_libpng INTERFACE ${PNG_INCLUDE_DIR})
+	target_include_directories(3rdparty_system_libpng SYSTEM INTERFACE ${PNG_INCLUDE_DIR})
 	target_link_libraries(3rdparty_system_libpng INTERFACE ${PNG_LIBRARY})
 	target_compile_definitions(3rdparty_system_libpng INTERFACE ${PNG_DEFINITIONS})
 

3rdparty/llvm/CMakeLists.txt

@@ -107,7 +107,7 @@ if(WITH_LLVM)
 
 	add_library(3rdparty_llvm INTERFACE)
 	target_link_libraries(3rdparty_llvm INTERFACE ${LLVM_LIBS})
-	target_include_directories(3rdparty_llvm INTERFACE ${LLVM_INCLUDE_DIRS})
+	target_include_directories(3rdparty_llvm SYSTEM INTERFACE ${LLVM_INCLUDE_DIRS})
 	separate_arguments(LLVM_DEFINITIONS_LIST NATIVE_COMMAND ${LLVM_DEFINITIONS})
 	target_compile_definitions(3rdparty_llvm INTERFACE ${LLVM_DEFINITIONS_LIST} LLVM_AVAILABLE)
 

3rdparty/miniupnp/CMakeLists.txt

@@ -3,9 +3,9 @@ if(USE_SYSTEM_MINIUPNPC)
   pkg_check_modules(MiniUPnPc REQUIRED IMPORTED_TARGET miniupnpc>=2.3.3)
   add_library(3rdparty_miniupnpc INTERFACE)
   target_link_libraries(3rdparty_miniupnpc INTERFACE PkgConfig::MiniUPnPc)
-  target_include_directories(3rdparty_miniupnpc INTERFACE PkgConfig::MiniUPnPc)
+  target_include_directories(3rdparty_miniupnpc SYSTEM INTERFACE PkgConfig::MiniUPnPc)
   list(TRANSFORM MiniUPnPc_INCLUDE_DIRS APPEND "/miniupnpc")
-  target_include_directories(3rdparty_miniupnpc INTERFACE ${MiniUPnPc_INCLUDE_DIRS})
+  target_include_directories(3rdparty_miniupnpc SYSTEM INTERFACE ${MiniUPnPc_INCLUDE_DIRS})
 else()
   option (UPNPC_BUILD_STATIC "Build static library" TRUE)
   option (UPNPC_BUILD_SHARED "Build shared library" FALSE)
@@ -17,5 +17,5 @@ else()
   add_subdirectory(miniupnp/miniupnpc EXCLUDE_FROM_ALL)
   add_library(3rdparty_miniupnpc INTERFACE)
   target_link_libraries(3rdparty_miniupnpc INTERFACE libminiupnpc-static)
-  target_include_directories(3rdparty_miniupnpc INTERFACE libminiupnpc-static)
+  target_include_directories(3rdparty_miniupnpc SYSTEM INTERFACE libminiupnpc-static)
 endif()

3rdparty/rtmidi/CMakeLists.txt

@@ -3,7 +3,7 @@ if(USE_SYSTEM_RTMIDI)
   pkg_check_modules(RtMidi REQUIRED IMPORTED_TARGET rtmidi>=6.0.0)
   add_library(rtmidi INTERFACE)
   target_link_libraries(rtmidi INTERFACE PkgConfig::RtMidi)
-  target_include_directories(rtmidi INTERFACE PkgConfig::RtMidi)
+  target_include_directories(rtmidi SYSTEM INTERFACE PkgConfig::RtMidi)
 else()
   option(RTMIDI_API_JACK "Compile with JACK support." OFF)
   option(RTMIDI_BUILD_TESTING "Build test programs" OFF)

3rdparty/stblib/CMakeLists.txt

@@ -1,2 +1,2 @@
 add_library(3rdparty_stblib INTERFACE)
-target_include_directories(3rdparty_stblib INTERFACE stb)
+target_include_directories(3rdparty_stblib SYSTEM INTERFACE stb)

3rdparty/zlib/CMakeLists.txt

@@ -13,6 +13,6 @@ else()
 
 	add_library(3rdparty_zlib INTERFACE)
 	target_link_libraries(3rdparty_zlib INTERFACE zlibstatic)
-	target_include_directories(3rdparty_zlib INTERFACE zlib ${CMAKE_CURRENT_BINARY_DIR}/zlib)
+	target_include_directories(3rdparty_zlib SYSTEM INTERFACE zlib ${CMAKE_CURRENT_BINARY_DIR}/zlib)
 	target_compile_definitions(3rdparty_zlib INTERFACE -DZLIB_CONST=1)
 endif()

3rdparty/zstd/CMakeLists.txt

@@ -3,7 +3,7 @@ if(USE_SYSTEM_ZSTD)
   pkg_check_modules(zstd REQUIRED IMPORTED_TARGET libzstd)
   add_library(3rdparty_zstd INTERFACE)
   target_link_libraries(3rdparty_zstd INTERFACE PkgConfig::zstd)
-  target_include_directories(3rdparty_zstd INTERFACE PkgConfig::RtMidi)
+  target_include_directories(3rdparty_zstd SYSTEM INTERFACE PkgConfig::RtMidi)
 else()
   option(ZSTD_BUILD_PROGRAMS "BUILD PROGRAMS" OFF)
   option(ZSTD_BUILD_SHARED "BUILD SHARED LIBRARIES" OFF)

rpcs3/Emu/CPU/CPUTranslator.cpp

@@ -71,7 +71,11 @@ cpu_translator::cpu_translator(llvm::Module* _module, bool is_be)
 			result = m_ir->CreateInsertElement(v, m_ir->CreateExtractElement(data0, m_ir->CreateExtractElement(mask, i)), i);
 			v->addIncoming(result, loop);
 			m_ir->CreateCondBr(m_ir->CreateICmpULT(i, m_ir->getInt32(16)), loop, next);
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+			m_ir->SetInsertPoint(next->getFirstNonPHIIt());
+#else
 			m_ir->SetInsertPoint(next->getFirstNonPHI());
+#endif
 			result = m_ir->CreateSelect(m_ir->CreateICmpSLT(index, zeros), zeros, result);
 
 			return result;

rpcs3/Emu/CPU/CPUTranslator.h

@@ -1149,7 +1149,11 @@ struct llvm_fshl
 	static llvm::Function* get_fshl(llvm::IRBuilder<>* ir)
 	{
 		const auto _module = ir->GetInsertBlock()->getParent()->getParent();
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+		return llvm::Intrinsic::getOrInsertDeclaration(_module, llvm::Intrinsic::fshl, {llvm_value_t<T>::get_type(ir->getContext())});
+#else
 		return llvm::Intrinsic::getDeclaration(_module, llvm::Intrinsic::fshl, {llvm_value_t<T>::get_type(ir->getContext())});
+#endif
 	}
 
 	static llvm::Value* fold(llvm::IRBuilder<>* ir, llvm::Value* v1, llvm::Value* v2, llvm::Value* v3)
@@ -1221,7 +1225,11 @@ struct llvm_fshr
 	static llvm::Function* get_fshr(llvm::IRBuilder<>* ir)
 	{
 		const auto _module = ir->GetInsertBlock()->getParent()->getParent();
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+		return llvm::Intrinsic::getOrInsertDeclaration(_module, llvm::Intrinsic::fshr, {llvm_value_t<T>::get_type(ir->getContext())});
+#else
 		return llvm::Intrinsic::getDeclaration(_module, llvm::Intrinsic::fshr, {llvm_value_t<T>::get_type(ir->getContext())});
+#endif
 	}
 
 	static llvm::Value* fold(llvm::IRBuilder<>* ir, llvm::Value* v1, llvm::Value* v2, llvm::Value* v3)
@@ -2220,7 +2228,11 @@ struct llvm_add_sat
 	static llvm::Function* get_add_sat(llvm::IRBuilder<>* ir)
 	{
 		const auto _module = ir->GetInsertBlock()->getParent()->getParent();
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+		return llvm::Intrinsic::getOrInsertDeclaration(_module, intr, {llvm_value_t<T>::get_type(ir->getContext())});
+#else
 		return llvm::Intrinsic::getDeclaration(_module, intr, {llvm_value_t<T>::get_type(ir->getContext())});
+#endif
 	}
 
 	llvm::Value* eval(llvm::IRBuilder<>* ir) const
@@ -2303,7 +2315,11 @@ struct llvm_sub_sat
 	static llvm::Function* get_sub_sat(llvm::IRBuilder<>* ir)
 	{
 		const auto _module = ir->GetInsertBlock()->getParent()->getParent();
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+		return llvm::Intrinsic::getOrInsertDeclaration(_module, intr, {llvm_value_t<T>::get_type(ir->getContext())});
+#else
 		return llvm::Intrinsic::getDeclaration(_module, intr, {llvm_value_t<T>::get_type(ir->getContext())});
+#endif
 	}
 
 	llvm::Value* eval(llvm::IRBuilder<>* ir) const
@@ -3592,7 +3608,11 @@ public:
 	llvm::Function* get_intrinsic(llvm::Intrinsic::ID id)
 	{
 		const auto _module = m_ir->GetInsertBlock()->getParent()->getParent();
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+		return llvm::Intrinsic::getOrInsertDeclaration(_module, id, {get_type<Types>()...});
+#else
 		return llvm::Intrinsic::getDeclaration(_module, id, {get_type<Types>()...});
+#endif
 	}
 
 	template <typename T1, typename T2>

rpcs3/Emu/Cell/Modules/cellGem.cpp

@@ -626,9 +626,9 @@ public:
 			cellGem.notice("Could not load mouse gem config. Using defaults.");
 		}
 
-		cellGem.notice("Real gem config=\n", g_cfg_gem_real.to_string());
-		cellGem.notice("Fake gem config=\n", g_cfg_gem_fake.to_string());
-		cellGem.notice("Mouse gem config=\n", g_cfg_gem_mouse.to_string());
+		cellGem.notice("Real gem config=%s", g_cfg_gem_real.to_string());
+		cellGem.notice("Fake gem config=%s", g_cfg_gem_fake.to_string());
+		cellGem.notice("Mouse gem config=%s", g_cfg_gem_mouse.to_string());
 	}
 };
 
@@ -719,43 +719,120 @@ namespace gem
 		constexpr u32 in_pitch = 640;
 		constexpr u32 out_pitch = 640 * 4;
 
-		for (u32 y = 0; y < 480 - 1; y += 2)
+		// Hamilton–Adams demosaicing
+		for (s32 y = 0; y < 480; y++)
 		{
-			const u8* src0 = src + y * in_pitch;
-			const u8* src1 = src0 + in_pitch;
+			const bool is_even_y = (y % 2) == 0;
+			const u8* srcc = src + y * in_pitch;
+			const u8* srcu = src + std::max(0, y - 1) * in_pitch;
+			const u8* srcd = src + std::min(480 - 1, y + 1) * in_pitch;
 
 			u8* dst0 = dst + y * out_pitch;
-			u8* dst1 = dst0 + out_pitch;
 
-			for (u32 x = 0; x < 640 - 1; x += 2, src0 += 2, src1 += 2, dst0 += 8, dst1 += 8)
+			// Split loops (roughly twice the performance by removing one condition)
+			if (is_even_y)
 			{
-				u8 b  = src0[0];
-				u8 g0 = src0[1];
-				u8 g1 = src1[0];
-				u8 r  = src1[1];
-
-				if constexpr (use_gain)
+				for (s32 x = 0; x < 640; x++, dst0 += 4)
 				{
-					b  = static_cast<u8>(std::clamp(b  * gain_b, 0.0f, 255.0f));
-					g0 = static_cast<u8>(std::clamp(g0 * gain_g, 0.0f, 255.0f));
-					g1 = static_cast<u8>(std::clamp(g1 * gain_g, 0.0f, 255.0f));
-					r  = static_cast<u8>(std::clamp(r  * gain_r, 0.0f, 255.0f));
+					const bool is_even_x = (x % 2) == 0;
+					const int xl = std::max(0, x - 1);
+					const int xr = std::min(640 - 1, x + 1);
+
+					u8 r, b, g;
+
+					if (is_even_x)
+					{
+						// Blue pixel
+						const u8 up = srcu[x];
+						const u8 down = srcd[x];
+						const u8 left = srcc[xl];
+						const u8 right = srcc[xr];
+						const int dh = std::abs(int(left) - int(right));
+						const int dv = std::abs(int(up)   - int(down));
+
+						r = (srcu[xl] + srcu[xr] + srcd[xl] + srcd[xr]) / 4;
+						if (dh < dv)
+							g = (left + right) / 2;
+						else if (dv < dh)
+							g = (up + down) / 2;
+						else
+							g = (up + down + left + right) / 4;
+						b = srcc[x];
+					}
+					else
+					{
+						// Green (on blue row)
+						r = (srcu[x] + srcd[x]) / 2;
+						g = srcc[x];
+						b = (srcc[xl] + srcc[xr]) / 2;
+					}
+
+					if constexpr (use_gain)
+					{
+						dst0[0] = static_cast<u8>(std::clamp(r * gain_r, 0.0f, 255.0f));
+						dst0[1] = static_cast<u8>(std::clamp(b * gain_b, 0.0f, 255.0f));
+						dst0[2] = static_cast<u8>(std::clamp(g * gain_g, 0.0f, 255.0f));
+					}
+					else
+					{
+						dst0[0] = r;
+						dst0[1] = g;
+						dst0[2] = b;
+					}
+					dst0[3] = alpha;
 				}
+			}
+			else
+			{
+				for (s32 x = 0; x < 640; x++, dst0 += 4)
+				{
+					const bool is_even_x = (x % 2) == 0;
+					const int xl = std::max(0, x - 1);
+					const int xr = std::min(640 - 1, x + 1);
 
-				const u8 top[4] = { r, g0, b, alpha };
-				const u8 bottom[4] = { r, g1, b, alpha };
+					u8 r, b, g;
 
-				// Top-Left
-				std::memcpy(dst0, top, 4);
+					if (is_even_x)
+					{
+						// Green (on red row)
+						r = (srcc[xl] + srcc[xr]) / 2;
+						g = srcc[x];
+						b = (srcu[x] + srcd[x]) / 2;
+					}
+					else
+					{
+						// Red pixel
+						const u8 up = srcu[x];
+						const u8 down = srcd[x];
+						const u8 left = srcc[xl];
+						const u8 right = srcc[xr];
+						const int dh = std::abs(int(left) - int(right));
+						const int dv = std::abs(int(up)   - int(down));
 
-				// Top-Right Pixel
-				std::memcpy(dst0 + 4, top, 4);
+						r = srcc[x];
+						if (dh < dv)
+							g = (left + right) / 2;
+						else if (dv < dh)
+							g = (up + down) / 2;
+						else
+							g = (up + down + left + right) / 4;
+						b = (srcu[xl] + srcu[xr] + srcd[xl] + srcd[xr]) / 4;
+					}
 
-				// Bottom-Left Pixel
-				std::memcpy(dst1, bottom, 4);
-
-				// Bottom-Right Pixel
-				std::memcpy(dst1 + 4, bottom, 4);
+					if constexpr (use_gain)
+					{
+						dst0[0] = static_cast<u8>(std::clamp(r * gain_r, 0.0f, 255.0f));
+						dst0[1] = static_cast<u8>(std::clamp(b * gain_b, 0.0f, 255.0f));
+						dst0[2] = static_cast<u8>(std::clamp(g * gain_g, 0.0f, 255.0f));
+					}
+					else
+					{
+						dst0[0] = r;
+						dst0[1] = g;
+						dst0[2] = b;
+					}
+					dst0[3] = alpha;
+				}
 			}
 		}
 	}

rpcs3/Emu/Cell/SPULLVMRecompiler.cpp

@@ -2617,7 +2617,11 @@ public:
 						{
 							if (b2 != bqbi)
 							{
+#if LLVM_VERSION_MAJOR >= 21 || (LLVM_VERSION_MAJOR == 20 && LLVM_VERSION_MINOR >= 1)
+								auto ins = &*b2->block->getFirstNonPHIIt();
+#else
 								auto ins = b2->block->getFirstNonPHI();
+#endif
 
 								if (b2->bb->preds.size() == 1)
 								{

rpcs3/Emu/RSX/VK/vkutils/buffer_object.cpp

@@ -50,6 +50,7 @@ namespace vk
 		: m_device(dev)
 	{
 		const bool nullable = !!(flags & VK_BUFFER_CREATE_ALLOW_NULL_RPCS3);
+		const bool no_vmem_recovery = !!(flags & VK_BUFFER_CREATE_IGNORE_VMEM_PRESSURE_RPCS3);
 
 		info.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
 		info.flags = flags & ~VK_BUFFER_CREATE_SPECIAL_FLAGS_RPCS3;
@@ -69,18 +70,27 @@ namespace vk
 			fmt::throw_exception("No compatible memory type was found!");
 		}
 
-		memory = std::make_unique<memory_block>(m_device, memory_reqs.size, memory_reqs.alignment, allocation_type_info, allocation_pool, nullable);
+		memory_allocation_request request
+		{
+			.size = memory_reqs.size,
+			.alignment = memory_reqs.alignment,
+			.memory_type = &allocation_type_info,
+			.pool = allocation_pool,
+			.throw_on_fail = !nullable,
+			.recover_vmem_on_fail = !no_vmem_recovery
+		};
+		memory = std::make_unique<memory_block>(m_device, request);
+
 		if (auto device_memory = memory->get_vk_device_memory();
 			device_memory != VK_NULL_HANDLE)
 		{
 			vkBindBufferMemory(dev, value, device_memory, memory->get_vk_device_memory_offset());
+			return;
 		}
-		else
-		{
-			ensure(nullable);
-			vkDestroyBuffer(m_device, value, nullptr);
-			value = VK_NULL_HANDLE;
-		}
+
+		ensure(nullable);
+		vkDestroyBuffer(m_device, value, nullptr);
+		value = VK_NULL_HANDLE;
 	}
 
 	buffer::buffer(const vk::render_device& dev, VkBufferUsageFlags usage, void* host_pointer, u64 size)

rpcs3/Emu/RSX/VK/vkutils/buffer_object.h

@@ -9,9 +9,10 @@ namespace vk
 {
 	enum : u32
 	{
-		VK_BUFFER_CREATE_ALLOW_NULL_RPCS3 = 0x80000000,
+		VK_BUFFER_CREATE_ALLOW_NULL_RPCS3           = 0x10000000,   // If we cannot allocate memory for the buffer, just return an empty but valid object with a null handle.
+		VK_BUFFER_CREATE_IGNORE_VMEM_PRESSURE_RPCS3 = 0x20000000,   // If we cannot allocate memory for the buffer, do not run recovery routine to recover VRAM. Crash or return empty handle immediately instead.
 
-		VK_BUFFER_CREATE_SPECIAL_FLAGS_RPCS3 = (VK_BUFFER_CREATE_ALLOW_NULL_RPCS3)
+		VK_BUFFER_CREATE_SPECIAL_FLAGS_RPCS3 = (VK_BUFFER_CREATE_ALLOW_NULL_RPCS3 | VK_BUFFER_CREATE_IGNORE_VMEM_PRESSURE_RPCS3)
 	};
 
 	struct buffer_view : public unique_resource

rpcs3/Emu/RSX/VK/vkutils/data_heap.cpp

@@ -53,7 +53,7 @@ namespace vk
 		VkFlags create_flags = 0;
 		if (m_prefer_writethrough)
 		{
-			create_flags |= VK_BUFFER_CREATE_ALLOW_NULL_RPCS3;
+			create_flags |= (VK_BUFFER_CREATE_ALLOW_NULL_RPCS3 | VK_BUFFER_CREATE_IGNORE_VMEM_PRESSURE_RPCS3);
 		}
 
 		heap = std::make_unique<buffer>(*g_render_device, size, memory_index, memory_flags, usage, create_flags, VMM_ALLOCATION_POOL_SYSTEM);
@@ -146,7 +146,7 @@ namespace vk
 		VkFlags create_flags = 0;
 		if (m_prefer_writethrough)
 		{
-			create_flags |= VK_BUFFER_CREATE_ALLOW_NULL_RPCS3;
+			create_flags |= (VK_BUFFER_CREATE_ALLOW_NULL_RPCS3 | VK_BUFFER_CREATE_IGNORE_VMEM_PRESSURE_RPCS3);
 		}
 
 		heap = std::make_unique<buffer>(*g_render_device, aligned_new_size, memory_index, memory_flags, usage, create_flags, VMM_ALLOCATION_POOL_SYSTEM);

rpcs3/Emu/RSX/VK/vkutils/image.cpp

@@ -128,7 +128,16 @@ namespace vk
 			fmt::throw_exception("No compatible memory type was found!");
 		}
 
-		memory = std::make_shared<vk::memory_block>(m_device, memory_req.size, memory_req.alignment, allocation_type_info, allocation_pool, nullable);
+		memory_allocation_request alloc_request
+		{
+			.size = memory_req.size,
+			.alignment = memory_req.alignment,
+			.memory_type = &allocation_type_info,
+			.pool = allocation_pool,
+			.throw_on_fail = !nullable
+		};
+		memory = std::make_shared<vk::memory_block>(m_device, alloc_request);
+
 		if (auto device_mem = memory->get_vk_device_memory();
 			device_mem != VK_NULL_HANDLE) [[likely]]
 		{

rpcs3/Emu/RSX/VK/vkutils/memory.cpp

@@ -224,7 +224,7 @@ namespace vk
 		vmaDestroyAllocator(m_allocator);
 	}
 
-	mem_allocator_vk::mem_handle_t mem_allocator_vma::alloc(u64 block_sz, u64 alignment, const memory_type_info& memory_type, vmm_allocation_pool pool, bool throw_on_fail)
+	mem_allocator_vk::mem_handle_t mem_allocator_vma::alloc(const memory_allocation_request& request)
 	{
 		VmaAllocation vma_alloc;
 		VkMemoryRequirements mem_req = {};
@@ -233,11 +233,11 @@ namespace vk
 
 		auto do_vma_alloc = [&]() -> std::tuple<VkResult, u32>
 		{
-			for (const auto& memory_type_index : memory_type)
+			for (const auto& memory_type_index : *request.memory_type)
 			{
 				mem_req.memoryTypeBits = 1u << memory_type_index;
-				mem_req.size = ::align2(block_sz, alignment);
-				mem_req.alignment = alignment;
+				mem_req.size = ::align2(request.size, request.alignment);
+				mem_req.alignment = request.alignment;
 				create_info.memoryTypeBits = 1u << memory_type_index;
 				create_info.flags = m_allocation_flags;
 
@@ -256,26 +256,29 @@ namespace vk
 			const auto [status, type] = do_vma_alloc();
 			if (status == VK_SUCCESS)
 			{
-				vmm_notify_memory_allocated(vma_alloc, type, block_sz, pool);
+				vmm_notify_memory_allocated(vma_alloc, type, request.size, request.pool);
 				return vma_alloc;
 			}
 		}
 
-		const auto severity = (throw_on_fail) ? rsx::problem_severity::fatal : rsx::problem_severity::severe;
-		if (error_code == VK_ERROR_OUT_OF_DEVICE_MEMORY &&
-			vmm_handle_memory_pressure(severity))
+		if (request.recover_vmem_on_fail)
 		{
-			// Out of memory. Try again.
-			const auto [status, type] = do_vma_alloc();
-			if (status == VK_SUCCESS)
+			const auto severity = (request.throw_on_fail) ? rsx::problem_severity::fatal : rsx::problem_severity::severe;
+			if (error_code == VK_ERROR_OUT_OF_DEVICE_MEMORY &&
+				vmm_handle_memory_pressure(severity))
 			{
-				rsx_log.warning("Renderer ran out of video memory but successfully recovered.");
-				vmm_notify_memory_allocated(vma_alloc, type, block_sz, pool);
-				return vma_alloc;
+				// Out of memory. Try again.
+				const auto [status, type] = do_vma_alloc();
+				if (status == VK_SUCCESS)
+				{
+					rsx_log.warning("Renderer ran out of video memory but successfully recovered.");
+					vmm_notify_memory_allocated(vma_alloc, type, request.size, request.pool);
+					return vma_alloc;
+				}
 			}
 		}
 
-		if (!throw_on_fail)
+		if (!request.throw_on_fail)
 		{
 			return VK_NULL_HANDLE;
 		}
@@ -361,18 +364,18 @@ namespace vk
 		m_allocation_flags = VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT;
 	}
 
-	mem_allocator_vk::mem_handle_t mem_allocator_vk::alloc(u64 block_sz, u64 /*alignment*/, const memory_type_info& memory_type, vmm_allocation_pool pool, bool throw_on_fail)
+	mem_allocator_vk::mem_handle_t mem_allocator_vk::alloc(const memory_allocation_request& request)
 	{
 		VkResult error_code = VK_ERROR_UNKNOWN;
 		VkDeviceMemory memory;
 
 		VkMemoryAllocateInfo info = {};
 		info.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
-		info.allocationSize = block_sz;
+		info.allocationSize = request.size;
 
 		auto do_vk_alloc = [&]() -> std::tuple<VkResult, u32>
 		{
-			for (const auto& memory_type_index : memory_type)
+			for (const auto& memory_type_index : *request.memory_type)
 			{
 				info.memoryTypeIndex = memory_type_index;
 				error_code = vkAllocateMemory(m_device, &info, nullptr, &memory);
@@ -389,26 +392,29 @@ namespace vk
 			const auto [status, type] = do_vk_alloc();
 			if (status == VK_SUCCESS)
 			{
-				vmm_notify_memory_allocated(memory, type, block_sz, pool);
+				vmm_notify_memory_allocated(memory, type, request.size, request.pool);
 				return memory;
 			}
 		}
 
-		const auto severity = (throw_on_fail) ? rsx::problem_severity::fatal : rsx::problem_severity::severe;
-		if (error_code == VK_ERROR_OUT_OF_DEVICE_MEMORY &&
-			vmm_handle_memory_pressure(severity))
+		if (request.recover_vmem_on_fail)
 		{
-			// Out of memory. Try again.
-			const auto [status, type] = do_vk_alloc();
-			if (status == VK_SUCCESS)
+			const auto severity = (request.throw_on_fail) ? rsx::problem_severity::fatal : rsx::problem_severity::severe;
+			if (error_code == VK_ERROR_OUT_OF_DEVICE_MEMORY &&
+				vmm_handle_memory_pressure(severity))
 			{
-				rsx_log.warning("Renderer ran out of video memory but successfully recovered.");
-				vmm_notify_memory_allocated(memory, type, block_sz, pool);
-				return memory;
+				// Out of memory. Try again.
+				const auto [status, type] = do_vk_alloc();
+				if (status == VK_SUCCESS)
+				{
+					rsx_log.warning("Renderer ran out of video memory but successfully recovered.");
+					vmm_notify_memory_allocated(memory, type, request.size, request.pool);
+					return memory;
+				}
 			}
 		}
 
-		if (!throw_on_fail)
+		if (!request.throw_on_fail)
 		{
 			return VK_NULL_HANDLE;
 		}
@@ -455,11 +461,11 @@ namespace vk
 		return g_render_device->get_allocator();
 	}
 
-	memory_block::memory_block(VkDevice dev, u64 block_sz, u64 alignment, const memory_type_info& memory_type, vmm_allocation_pool pool, bool nullable)
-		: m_device(dev), m_size(block_sz)
+	memory_block::memory_block(VkDevice dev, const memory_allocation_request& alloc_request)
+		: m_device(dev), m_size(alloc_request.size)
 	{
 		m_mem_allocator = get_current_mem_allocator();
-		m_mem_handle    = m_mem_allocator->alloc(block_sz, alignment, memory_type, pool, !nullable);
+		m_mem_handle    = m_mem_allocator->alloc(alloc_request);
 	}
 
 	memory_block::~memory_block()

rpcs3/Emu/RSX/VK/vkutils/memory.h

@@ -66,6 +66,16 @@ namespace vk
 		u64 size;
 	};
 
+	struct memory_allocation_request
+	{
+		u64 size = 0;
+		u64 alignment = 1;
+		const memory_type_info* memory_type = nullptr;
+		vmm_allocation_pool pool = VMM_ALLOCATION_POOL_UNDEFINED;
+		bool throw_on_fail = true;
+		bool recover_vmem_on_fail = true;
+	};
+
 	class mem_allocator_base
 	{
 	public:
@@ -76,7 +86,7 @@ namespace vk
 
 		virtual void destroy() = 0;
 
-		virtual mem_handle_t alloc(u64 block_sz, u64 alignment, const memory_type_info& memory_type, vmm_allocation_pool pool, bool throw_on_fail) = 0;
+		virtual mem_handle_t alloc(const memory_allocation_request& request) = 0;
 		virtual void free(mem_handle_t mem_handle) = 0;
 		virtual void* map(mem_handle_t mem_handle, u64 offset, u64 size) = 0;
 		virtual void unmap(mem_handle_t mem_handle) = 0;
@@ -104,7 +114,7 @@ namespace vk
 
 		void destroy() override;
 
-		mem_handle_t alloc(u64 block_sz, u64 alignment, const memory_type_info& memory_type, vmm_allocation_pool pool, bool throw_on_fail) override;
+		mem_handle_t alloc(const memory_allocation_request& request) override;
 
 		void free(mem_handle_t mem_handle) override;
 		void* map(mem_handle_t mem_handle, u64 offset, u64 /*size*/) override;
@@ -134,7 +144,7 @@ namespace vk
 
 		void destroy() override {}
 
-		mem_handle_t alloc(u64 block_sz, u64 /*alignment*/, const memory_type_info& memory_type, vmm_allocation_pool pool, bool throw_on_fail) override;
+		mem_handle_t alloc(const memory_allocation_request& request) override;
 
 		void free(mem_handle_t mem_handle) override;
 		void* map(mem_handle_t mem_handle, u64 offset, u64 size) override;
@@ -147,7 +157,7 @@ namespace vk
 
 	struct memory_block
 	{
-		memory_block(VkDevice dev, u64 block_sz, u64 alignment, const memory_type_info& memory_type, vmm_allocation_pool pool, bool nullable = false);
+		memory_block(VkDevice dev, const memory_allocation_request& alloc_request);
 		virtual ~memory_block();
 
 		virtual VkDeviceMemory get_vk_device_memory();

rpcs3/rpcs3qt/qt_camera_video_sink.cpp

@@ -114,45 +114,60 @@ bool qt_camera_video_sink::present(const QVideoFrame& frame)
 		case CELL_CAMERA_RAW8: // The game seems to expect BGGR
 		{
 			// Let's use a very simple algorithm to convert the image to raw BGGR
-			const auto convert_to_bggr = [&image_buffer, &image, width, height](u32 y_begin, u32 y_end)
+			const auto convert_to_bggr = [this, &image_buffer, &image, width, height](u32 y_begin, u32 y_end)
 			{
 				u8* dst = &image_buffer.data[image_buffer.width * y_begin];
 
 				for (u32 y = y_begin; y < height && y < y_end; y++)
 				{
 					const u8* src = image.constScanLine(y);
+					const u8* srcu = image.constScanLine(std::max<s32>(0, y - 1));
+					const u8* srcd = image.constScanLine(std::min(height - 1, y + 1));
 					const bool is_top_pixel = (y % 2) == 0;
 
+					// We apply gaussian blur to get better demosaicing results later when debayering again
+					const auto blurred = [&](s32 x, s32 c)
+					{
+						const s32 i = x * 4 + c;
+						const s32 il = std::max(0, x - 1) * 4 + c;
+						const s32 ir = std::min<s32>(width - 1, x + 1) * 4 + c;
+						const s32 sum =
+							              srcu[i] +
+							src[il] + 4 * src[i]  + src[ir] +
+							              srcd[i];
+						return static_cast<u8>(std::clamp((sum + 4) / 8, 0, 255));
+					};
+
 					// Split loops (roughly twice the performance by removing one condition)
 					if (is_top_pixel)
 					{
-						for (u32 x = 0; x < width; x++, dst++, src += 4)
+						for (u32 x = 0; x < width; x++, dst++)
 						{
 							const bool is_left_pixel = (x % 2) == 0;
 
 							if (is_left_pixel)
 							{
-								*dst = src[2]; // Blue
+								*dst = blurred(x, 2); // Blue
 							}
 							else
 							{
-								*dst = src[1]; // Green
+								*dst = blurred(x, 1); // Green
 							}
 						}
 					}
 					else
 					{
-						for (u32 x = 0; x < width; x++, dst++, src += 4)
+						for (u32 x = 0; x < width; x++, dst++)
 						{
 							const bool is_left_pixel = (x % 2) == 0;
 
 							if (is_left_pixel)
 							{
-								*dst = src[1]; // Green
+								*dst = blurred(x, 1); // Green
 							}
 							else
 							{
-								*dst = src[0]; // Red
+								*dst = blurred(x, 0); // Red
 							}
 						}
 					}
@@ -182,13 +197,13 @@ bool qt_camera_video_sink::present(const QVideoFrame& frame)
 			// Simple RGB to Y0_U_Y1_V conversion from stackoverflow.
 			const auto convert_to_yuv422 = [&image_buffer, &image, width, height, format = m_format](u32 y_begin, u32 y_end)
 			{
-				constexpr int yuv_bytes_per_pixel = 2;
-				const int yuv_pitch = image_buffer.width * yuv_bytes_per_pixel;
+				constexpr s32 yuv_bytes_per_pixel = 2;
+				const s32 yuv_pitch = image_buffer.width * yuv_bytes_per_pixel;
 
-				const int y0_offset = (format == CELL_CAMERA_Y0_U_Y1_V) ? 0 : 3;
-				const int u_offset  = (format == CELL_CAMERA_Y0_U_Y1_V) ? 1 : 2;
-				const int y1_offset = (format == CELL_CAMERA_Y0_U_Y1_V) ? 2 : 1;
-				const int v_offset  = (format == CELL_CAMERA_Y0_U_Y1_V) ? 3 : 0;
+				const s32 y0_offset = (format == CELL_CAMERA_Y0_U_Y1_V) ? 0 : 3;
+				const s32 u_offset  = (format == CELL_CAMERA_Y0_U_Y1_V) ? 1 : 2;
+				const s32 y1_offset = (format == CELL_CAMERA_Y0_U_Y1_V) ? 2 : 1;
+				const s32 v_offset  = (format == CELL_CAMERA_Y0_U_Y1_V) ? 3 : 0;
 
 				for (u32 y = y_begin; y < height && y < y_end; y++)
 				{
@@ -197,19 +212,19 @@ bool qt_camera_video_sink::present(const QVideoFrame& frame)
 
 					for (u32 x = 0; x < width - 1; x += 2, src += 8)
 					{
-						const float r1 = src[0];
-						const float g1 = src[1];
-						const float b1 = src[2];
-						const float r2 = src[4];
-						const float g2 = src[5];
-						const float b2 = src[6];
+						const f32 r1 = src[0];
+						const f32 g1 = src[1];
+						const f32 b1 = src[2];
+						const f32 r2 = src[4];
+						const f32 g2 = src[5];
+						const f32 b2 = src[6];
 
-						const int y0 =  (0.257f * r1) + (0.504f * g1) + (0.098f * b1) +  16.0f;
-						const int u  = -(0.148f * r1) - (0.291f * g1) + (0.439f * b1) + 128.0f;
-						const int v  =  (0.439f * r1) - (0.368f * g1) - (0.071f * b1) + 128.0f;
-						const int y1 =  (0.257f * r2) + (0.504f * g2) + (0.098f * b2) +  16.0f;
+						const s32 y0 =  (0.257f * r1) + (0.504f * g1) + (0.098f * b1) +  16.0f;
+						const s32 u  = -(0.148f * r1) - (0.291f * g1) + (0.439f * b1) + 128.0f;
+						const s32 v  =  (0.439f * r1) - (0.368f * g1) - (0.071f * b1) + 128.0f;
+						const s32 y1 =  (0.257f * r2) + (0.504f * g2) + (0.098f * b2) +  16.0f;
 
-						const int yuv_index = x * yuv_bytes_per_pixel;
+						const s32 yuv_index = x * yuv_bytes_per_pixel;
 						yuv_row_ptr[yuv_index + y0_offset] = static_cast<u8>(std::clamp(y0, 0, 255));
 						yuv_row_ptr[yuv_index + u_offset]  = static_cast<u8>(std::clamp( u, 0, 255));
 						yuv_row_ptr[yuv_index + y1_offset] = static_cast<u8>(std::clamp(y1, 0, 255));