Merge pull request #2763 from lioncash/map-phys

kernel/vm_manager: Minor cleanup

src/core/hle/kernel/vm_manager.cpp

@@ -296,12 +296,6 @@ ResultVal<VAddr> VMManager::SetHeapSize(u64 size) {
 }
 
 ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
-    const auto end_addr = target + size;
-    const auto last_addr = end_addr - 1;
-    VAddr cur_addr = target;
-
-    ResultCode result = RESULT_SUCCESS;
-
     // Check how much memory we've already mapped.
     const auto mapped_size_result = SizeOfAllocatedVMAsInRange(target, size);
     if (mapped_size_result.Failed()) {
@@ -324,13 +318,16 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
 
     // Keep track of the memory regions we unmap.
     std::vector<std::pair<u64, u64>> mapped_regions;
+    ResultCode result = RESULT_SUCCESS;
 
     // Iterate, trying to map memory.
     {
-        cur_addr = target;
+        const auto end_addr = target + size;
+        const auto last_addr = end_addr - 1;
+        VAddr cur_addr = target;
 
         auto iter = FindVMA(target);
-        ASSERT_MSG(iter != vma_map.end(), "MapPhysicalMemory iter != end");
+        ASSERT(iter != vma_map.end());
 
         while (true) {
             const auto& vma = iter->second;
@@ -342,7 +339,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
             const auto map_size = std::min(end_addr - cur_addr, vma_end - cur_addr);
             if (vma.state == MemoryState::Unmapped) {
                 const auto map_res =
-                    MapMemoryBlock(cur_addr, std::make_shared<PhysicalMemory>(map_size, 0), 0,
+                    MapMemoryBlock(cur_addr, std::make_shared<PhysicalMemory>(map_size), 0,
                                    map_size, MemoryState::Heap, VMAPermission::ReadWrite);
                 result = map_res.Code();
                 if (result.IsError()) {
@@ -360,7 +357,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
             // Advance to the next block.
             cur_addr = vma_end;
             iter = FindVMA(cur_addr);
-            ASSERT_MSG(iter != vma_map.end(), "MapPhysicalMemory iter != end");
+            ASSERT(iter != vma_map.end());
         }
     }
 
@@ -368,7 +365,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
     if (result.IsError()) {
         for (const auto [unmap_address, unmap_size] : mapped_regions) {
             ASSERT_MSG(UnmapRange(unmap_address, unmap_size).IsSuccess(),
-                       "MapPhysicalMemory un-map on error");
+                       "Failed to unmap memory range.");
         }
 
         return result;
@@ -381,12 +378,6 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
 }
 
 ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
-    const auto end_addr = target + size;
-    const auto last_addr = end_addr - 1;
-    VAddr cur_addr = target;
-
-    ResultCode result = RESULT_SUCCESS;
-
     // Check how much memory is currently mapped.
     const auto mapped_size_result = SizeOfUnmappablePhysicalMemoryInRange(target, size);
     if (mapped_size_result.Failed()) {
@@ -401,13 +392,16 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
 
     // Keep track of the memory regions we unmap.
     std::vector<std::pair<u64, u64>> unmapped_regions;
+    ResultCode result = RESULT_SUCCESS;
 
     // Try to unmap regions.
     {
-        cur_addr = target;
+        const auto end_addr = target + size;
+        const auto last_addr = end_addr - 1;
+        VAddr cur_addr = target;
 
         auto iter = FindVMA(target);
-        ASSERT_MSG(iter != vma_map.end(), "UnmapPhysicalMemory iter != end");
+        ASSERT(iter != vma_map.end());
 
         while (true) {
             const auto& vma = iter->second;
@@ -434,7 +428,7 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
             // Advance to the next block.
             cur_addr = vma_end;
             iter = FindVMA(cur_addr);
-            ASSERT_MSG(iter != vma_map.end(), "UnmapPhysicalMemory iter != end");
+            ASSERT(iter != vma_map.end());
         }
     }
 
@@ -443,10 +437,12 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
     if (result.IsError()) {
         for (const auto [map_address, map_size] : unmapped_regions) {
             const auto remap_res =
-                MapMemoryBlock(map_address, std::make_shared<PhysicalMemory>(map_size, 0), 0,
-                               map_size, MemoryState::Heap, VMAPermission::None);
-            ASSERT_MSG(remap_res.Succeeded(), "UnmapPhysicalMemory re-map on error");
+                MapMemoryBlock(map_address, std::make_shared<PhysicalMemory>(map_size), 0, map_size,
+                               MemoryState::Heap, VMAPermission::None);
+            ASSERT_MSG(remap_res.Succeeded(), "Failed to remap a memory block.");
         }
+
+        return result;
     }
 
     // Update mapped amount
@@ -757,20 +753,26 @@ void VMManager::MergeAdjacentVMA(VirtualMemoryArea& left, const VirtualMemoryAre
     // Always merge allocated memory blocks, even when they don't share the same backing block.
     if (left.type == VMAType::AllocatedMemoryBlock &&
         (left.backing_block != right.backing_block || left.offset + left.size != right.offset)) {
+        const auto right_begin = right.backing_block->begin() + right.offset;
+        const auto right_end = right_begin + right.size;
+
         // Check if we can save work.
         if (left.offset == 0 && left.size == left.backing_block->size()) {
             // Fast case: left is an entire backing block.
-            left.backing_block->insert(left.backing_block->end(),
-                                       right.backing_block->begin() + right.offset,
-                                       right.backing_block->begin() + right.offset + right.size);
+            left.backing_block->insert(left.backing_block->end(), right_begin, right_end);
         } else {
             // Slow case: make a new memory block for left and right.
+            const auto left_begin = left.backing_block->begin() + left.offset;
+            const auto left_end = left_begin + left.size;
+            const auto left_size = static_cast<std::size_t>(std::distance(left_begin, left_end));
+            const auto right_size = static_cast<std::size_t>(std::distance(right_begin, right_end));
+
             auto new_memory = std::make_shared<PhysicalMemory>();
-            new_memory->insert(new_memory->end(), left.backing_block->begin() + left.offset,
-                               left.backing_block->begin() + left.offset + left.size);
-            new_memory->insert(new_memory->end(), right.backing_block->begin() + right.offset,
-                               right.backing_block->begin() + right.offset + right.size);
-            left.backing_block = new_memory;
+            new_memory->reserve(left_size + right_size);
+            new_memory->insert(new_memory->end(), left_begin, left_end);
+            new_memory->insert(new_memory->end(), right_begin, right_end);
+
+            left.backing_block = std::move(new_memory);
             left.offset = 0;
         }
 
@@ -965,7 +967,7 @@ ResultVal<std::size_t> VMManager::SizeOfAllocatedVMAsInRange(VAddr address,
 
     VAddr cur_addr = address;
     auto iter = FindVMA(cur_addr);
-    ASSERT_MSG(iter != vma_map.end(), "SizeOfAllocatedVMAsInRange iter != end");
+    ASSERT(iter != vma_map.end());
 
     while (true) {
         const auto& vma = iter->second;
@@ -986,7 +988,7 @@ ResultVal<std::size_t> VMManager::SizeOfAllocatedVMAsInRange(VAddr address,
         // Advance to the next block.
         cur_addr = vma_end;
         iter = std::next(iter);
-        ASSERT_MSG(iter != vma_map.end(), "SizeOfAllocatedVMAsInRange iter != end");
+        ASSERT(iter != vma_map.end());
     }
 
     return MakeResult(mapped_size);
@@ -1000,7 +1002,7 @@ ResultVal<std::size_t> VMManager::SizeOfUnmappablePhysicalMemoryInRange(VAddr ad
 
     VAddr cur_addr = address;
     auto iter = FindVMA(cur_addr);
-    ASSERT_MSG(iter != vma_map.end(), "SizeOfUnmappablePhysicalMemoryInRange iter != end");
+    ASSERT(iter != vma_map.end());
 
     while (true) {
         const auto& vma = iter->second;
@@ -1029,7 +1031,7 @@ ResultVal<std::size_t> VMManager::SizeOfUnmappablePhysicalMemoryInRange(VAddr ad
         // Advance to the next block.
         cur_addr = vma_end;
         iter = std::next(iter);
-        ASSERT_MSG(iter != vma_map.end(), "SizeOfUnmappablePhysicalMemoryInRange iter != end");
+        ASSERT(iter != vma_map.end());
     }
 
     return MakeResult(mapped_size);

src/core/hle/kernel/vm_manager.h

@@ -454,8 +454,8 @@ public:
 
     /// Maps memory at a given address.
     ///
-    /// @param addr The virtual address to map memory at.
-    /// @param size The amount of memory to map.
+    /// @param target The virtual address to map memory at.
+    /// @param size   The amount of memory to map.
     ///
     /// @note The destination address must lie within the Map region.
     ///
@@ -468,8 +468,8 @@ public:
 
     /// Unmaps memory at a given address.
     ///
-    /// @param addr The virtual address to unmap memory at.
-    /// @param size The amount of memory to unmap.
+    /// @param target The virtual address to unmap memory at.
+    /// @param size   The amount of memory to unmap.
     ///
     /// @note The destination address must lie within the Map region.
     ///

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -537,8 +537,7 @@ std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers(
         texture_cache.MarkDepthBufferInUse();
 
         fbkey.zeta = depth_surface;
-        fbkey.stencil_enable = regs.stencil_enable &&
-                               depth_surface->GetSurfaceParams().type == SurfaceType::DepthStencil;
+        fbkey.stencil_enable = depth_surface->GetSurfaceParams().type == SurfaceType::DepthStencil;
     }
 
     texture_cache.GuardRenderTargets(false);
@@ -577,16 +576,15 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(OpenGLState& current_state, boo
     if (depth_surface) {
         const auto& params = depth_surface->GetSurfaceParams();
         switch (params.type) {
-        case VideoCore::Surface::SurfaceType::Depth: {
+        case VideoCore::Surface::SurfaceType::Depth:
             depth_surface->Attach(GL_DEPTH_ATTACHMENT, GL_DRAW_FRAMEBUFFER);
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
             break;
-        }
-        case VideoCore::Surface::SurfaceType::DepthStencil: {
-            depth_surface->Attach(GL_DEPTH_ATTACHMENT, GL_DRAW_FRAMEBUFFER);
+        case VideoCore::Surface::SurfaceType::DepthStencil:
+            depth_surface->Attach(GL_DEPTH_STENCIL_ATTACHMENT, GL_DRAW_FRAMEBUFFER);
             break;
-        }
-        default: { UNIMPLEMENTED(); }
+        default:
+            UNIMPLEMENTED();
         }
     } else {
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
@@ -639,6 +637,7 @@ void RasterizerOpenGL::Clear() {
         ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear stencil but buffer is not enabled!");
         use_stencil = true;
         clear_state.stencil.test_enabled = true;
+
         if (regs.clear_flags.stencil) {
             // Stencil affects the clear so fill it with the used masks
             clear_state.stencil.front.test_func = GL_ALWAYS;
@@ -1119,9 +1118,12 @@ void RasterizerOpenGL::SyncStencilTestState() {
     if (!maxwell3d.dirty.stencil_test) {
         return;
     }
-    const auto& regs = maxwell3d.regs;
+    maxwell3d.dirty.stencil_test = false;
 
+    const auto& regs = maxwell3d.regs;
     state.stencil.test_enabled = regs.stencil_enable != 0;
+    state.MarkDirtyStencilState();
+
     if (!regs.stencil_enable) {
         return;
     }
@@ -1150,8 +1152,6 @@ void RasterizerOpenGL::SyncStencilTestState() {
         state.stencil.back.action_depth_fail = GL_KEEP;
         state.stencil.back.action_depth_pass = GL_KEEP;
     }
-    state.MarkDirtyStencilState();
-    maxwell3d.dirty.stencil_test = false;
 }
 
 void RasterizerOpenGL::SyncColorMask() {

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -348,23 +348,16 @@ Shader CachedShader::CreateKernelFromCache(const ShaderParameters& params,
 }
 
 std::tuple<GLuint, BaseBindings> CachedShader::GetProgramHandle(const ProgramVariant& variant) {
-    GLuint handle{};
-    if (program_type == ProgramType::Geometry) {
-        handle = GetGeometryShader(variant);
-    } else {
-        const auto [entry, is_cache_miss] = programs.try_emplace(variant);
-        auto& program = entry->second;
-        if (is_cache_miss) {
-            program = TryLoadProgram(variant);
-            if (!program) {
-                program = SpecializeShader(code, entries, program_type, variant);
-                disk_cache.SaveUsage(GetUsage(variant));
-            }
-
-            LabelGLObject(GL_PROGRAM, program->handle, cpu_addr);
+    const auto [entry, is_cache_miss] = programs.try_emplace(variant);
+    auto& program = entry->second;
+    if (is_cache_miss) {
+        program = TryLoadProgram(variant);
+        if (!program) {
+            program = SpecializeShader(code, entries, program_type, variant);
+            disk_cache.SaveUsage(GetUsage(variant));
         }
 
-        handle = program->handle;
+        LabelGLObject(GL_PROGRAM, program->handle, cpu_addr);
     }
 
     auto base_bindings = variant.base_bindings;
@@ -375,52 +368,9 @@ std::tuple<GLuint, BaseBindings> CachedShader::GetProgramHandle(const ProgramVar
     base_bindings.gmem += static_cast<u32>(entries.global_memory_entries.size());
     base_bindings.sampler += static_cast<u32>(entries.samplers.size());
 
-    return {handle, base_bindings};
+    return {program->handle, base_bindings};
 }
 
-GLuint CachedShader::GetGeometryShader(const ProgramVariant& variant) {
-    const auto [entry, is_cache_miss] = geometry_programs.try_emplace(variant);
-    auto& programs = entry->second;
-
-    switch (variant.primitive_mode) {
-    case GL_POINTS:
-        return LazyGeometryProgram(programs.points, variant);
-    case GL_LINES:
-    case GL_LINE_STRIP:
-        return LazyGeometryProgram(programs.lines, variant);
-    case GL_LINES_ADJACENCY:
-    case GL_LINE_STRIP_ADJACENCY:
-        return LazyGeometryProgram(programs.lines_adjacency, variant);
-    case GL_TRIANGLES:
-    case GL_TRIANGLE_STRIP:
-    case GL_TRIANGLE_FAN:
-        return LazyGeometryProgram(programs.triangles, variant);
-    case GL_TRIANGLES_ADJACENCY:
-    case GL_TRIANGLE_STRIP_ADJACENCY:
-        return LazyGeometryProgram(programs.triangles_adjacency, variant);
-    default:
-        UNREACHABLE_MSG("Unknown primitive mode.");
-        return LazyGeometryProgram(programs.points, variant);
-    }
-}
-
-GLuint CachedShader::LazyGeometryProgram(CachedProgram& target_program,
-                                         const ProgramVariant& variant) {
-    if (target_program) {
-        return target_program->handle;
-    }
-    const auto [glsl_name, debug_name, vertices] = GetPrimitiveDescription(variant.primitive_mode);
-    target_program = TryLoadProgram(variant);
-    if (!target_program) {
-        target_program = SpecializeShader(code, entries, program_type, variant);
-        disk_cache.SaveUsage(GetUsage(variant));
-    }
-
-    LabelGLObject(GL_PROGRAM, target_program->handle, cpu_addr, debug_name);
-
-    return target_program->handle;
-};
-
 CachedProgram CachedShader::TryLoadProgram(const ProgramVariant& variant) const {
     const auto found = precompiled_programs.find(GetUsage(variant));
     if (found == precompiled_programs.end()) {

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -86,22 +86,6 @@ private:
     explicit CachedShader(const ShaderParameters& params, ProgramType program_type,
                           GLShader::ProgramResult result);
 
-    // Geometry programs. These are needed because GLSL needs an input topology but it's not
-    // declared by the hardware. Workaround this issue by generating a different shader per input
-    // topology class.
-    struct GeometryPrograms {
-        CachedProgram points;
-        CachedProgram lines;
-        CachedProgram lines_adjacency;
-        CachedProgram triangles;
-        CachedProgram triangles_adjacency;
-    };
-
-    GLuint GetGeometryShader(const ProgramVariant& variant);
-
-    /// Generates a geometry shader or returns one that already exists.
-    GLuint LazyGeometryProgram(CachedProgram& target_program, const ProgramVariant& variant);
-
     CachedProgram TryLoadProgram(const ProgramVariant& variant) const;
 
     ShaderDiskCacheUsage GetUsage(const ProgramVariant& variant) const;
@@ -117,11 +101,6 @@ private:
     std::size_t shader_length{};
 
     std::unordered_map<ProgramVariant, CachedProgram> programs;
-    std::unordered_map<ProgramVariant, GeometryPrograms> geometry_programs;
-
-    std::unordered_map<u32, GLuint> cbuf_resource_cache;
-    std::unordered_map<u32, GLuint> gmem_resource_cache;
-    std::unordered_map<u32, GLint> uniform_cache;
 };
 
 class ShaderCacheOpenGL final : public RasterizerCache<Shader> {

src/yuzu/main.cpp

@@ -6,6 +6,9 @@
 #include <clocale>
 #include <memory>
 #include <thread>
+#ifdef __APPLE__
+#include <unistd.h> // for chdir
+#endif
 
 // VFS includes must be before glad as they will conflict with Windows file api, which uses defines.
 #include "applets/error.h"
@@ -2226,6 +2229,14 @@ int main(int argc, char* argv[]) {
     QCoreApplication::setOrganizationName(QStringLiteral("yuzu team"));
     QCoreApplication::setApplicationName(QStringLiteral("yuzu"));
 
+#ifdef __APPLE__
+    // If you start a bundle (binary) on OSX without the Terminal, the working directory is "/".
+    // But since we require the working directory to be the executable path for the location of the
+    // user folder in the Qt Frontend, we need to cd into that working directory
+    const std::string bin_path = FileUtil::GetBundleDirectory() + DIR_SEP + "..";
+    chdir(bin_path.c_str());
+#endif
+
     // Enables the core to make the qt created contexts current on std::threads
     QCoreApplication::setAttribute(Qt::AA_DontCheckOpenGLContextThreadAffinity);
     QApplication app(argc, argv);