Merge pull request #4633 from ReinUsesLisp/gpu-init

video_core: Remove all Core::System references in renderer
This commit is contained in:
Rodrigo Locatti 2020-09-10 02:28:54 +00:00 • committed by GitHub
commit 8fcafa37a8
53 changed files with 573 additions and 633 deletions

View file

@ -188,7 +188,6 @@ struct System::Impl {
if (!gpu_core) {
return ResultStatus::ErrorVideoCore;
}
gpu_core->Renderer().Rasterizer().SetupDirtyFlags();
is_powered_on = true;
exit_lock = false;

View file

@ -51,46 +51,43 @@ public:
bool is_written = false, bool use_fast_cbuf = false) {
std::lock_guard lock{mutex};
auto& memory_manager = system.GPU().MemoryManager();
const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr);
if (!cpu_addr_opt) {
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
return GetEmptyBuffer(size);
}
const VAddr cpu_addr = *cpu_addr_opt;
// Cache management is a big overhead, so only cache entries with a given size.
// TODO: Figure out which size is the best for given games.
constexpr std::size_t max_stream_size = 0x800;
if (use_fast_cbuf || size < max_stream_size) {
if (!is_written && !IsRegionWritten(cpu_addr, cpu_addr + size - 1)) {
const bool is_granular = memory_manager.IsGranularRange(gpu_addr, size);
if (!is_written && !IsRegionWritten(*cpu_addr, *cpu_addr + size - 1)) {
const bool is_granular = gpu_memory.IsGranularRange(gpu_addr, size);
if (use_fast_cbuf) {
u8* dest;
if (is_granular) {
dest = memory_manager.GetPointer(gpu_addr);
dest = gpu_memory.GetPointer(gpu_addr);
} else {
staging_buffer.resize(size);
dest = staging_buffer.data();
memory_manager.ReadBlockUnsafe(gpu_addr, dest, size);
gpu_memory.ReadBlockUnsafe(gpu_addr, dest, size);
}
return ConstBufferUpload(dest, size);
}
if (is_granular) {
u8* const host_ptr = memory_manager.GetPointer(gpu_addr);
u8* const host_ptr = gpu_memory.GetPointer(gpu_addr);
return StreamBufferUpload(size, alignment, [host_ptr, size](u8* dest) {
std::memcpy(dest, host_ptr, size);
});
} else {
return StreamBufferUpload(
size, alignment, [&memory_manager, gpu_addr, size](u8* dest) {
memory_manager.ReadBlockUnsafe(gpu_addr, dest, size);
});
return StreamBufferUpload(size, alignment, [this, gpu_addr, size](u8* dest) {
gpu_memory.ReadBlockUnsafe(gpu_addr, dest, size);
});
}
}
}
Buffer* const block = GetBlock(cpu_addr, size);
MapInterval* const map = MapAddress(block, gpu_addr, cpu_addr, size);
Buffer* const block = GetBlock(*cpu_addr, size);
MapInterval* const map = MapAddress(block, gpu_addr, *cpu_addr, size);
if (!map) {
return GetEmptyBuffer(size);
}
@ -106,7 +103,7 @@ public:
}
}
return BufferInfo{block->Handle(), block->Offset(cpu_addr), block->Address()};
return BufferInfo{block->Handle(), block->Offset(*cpu_addr), block->Address()};
}
/// Uploads from a host memory. Returns the OpenGL buffer where it's located and its offset.
@ -262,9 +259,11 @@ public:
virtual BufferInfo GetEmptyBuffer(std::size_t size) = 0;
protected:
explicit BufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
std::unique_ptr<StreamBuffer> stream_buffer)
: rasterizer{rasterizer}, system{system}, stream_buffer{std::move(stream_buffer)} {}
explicit BufferCache(VideoCore::RasterizerInterface& rasterizer_,
Tegra::MemoryManager& gpu_memory_, Core::Memory::Memory& cpu_memory_,
std::unique_ptr<StreamBuffer> stream_buffer_)
: rasterizer{rasterizer_}, gpu_memory{gpu_memory_}, cpu_memory{cpu_memory_},
stream_buffer{std::move(stream_buffer_)}, stream_buffer_handle{stream_buffer->Handle()} {}
~BufferCache() = default;
@ -326,14 +325,13 @@ private:
MapInterval* MapAddress(Buffer* block, GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size) {
const VectorMapInterval overlaps = GetMapsInRange(cpu_addr, size);
if (overlaps.empty()) {
auto& memory_manager = system.GPU().MemoryManager();
const VAddr cpu_addr_end = cpu_addr + size;
if (memory_manager.IsGranularRange(gpu_addr, size)) {
u8* host_ptr = memory_manager.GetPointer(gpu_addr);
if (gpu_memory.IsGranularRange(gpu_addr, size)) {
u8* const host_ptr = gpu_memory.GetPointer(gpu_addr);
block->Upload(block->Offset(cpu_addr), size, host_ptr);
} else {
staging_buffer.resize(size);
memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
gpu_memory.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
block->Upload(block->Offset(cpu_addr), size, staging_buffer.data());
}
return Register(MapInterval(cpu_addr, cpu_addr_end, gpu_addr));
@ -392,7 +390,7 @@ private:
continue;
}
staging_buffer.resize(size);
system.Memory().ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size);
cpu_memory.ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size);
block->Upload(block->Offset(interval.lower()), size, staging_buffer.data());
}
}
@ -431,7 +429,7 @@ private:
const std::size_t size = map->end - map->start;
staging_buffer.resize(size);
block->Download(block->Offset(map->start), size, staging_buffer.data());
system.Memory().WriteBlockUnsafe(map->start, staging_buffer.data(), size);
cpu_memory.WriteBlockUnsafe(map->start, staging_buffer.data(), size);
map->MarkAsModified(false, 0);
}
@ -567,7 +565,8 @@ private:
}
VideoCore::RasterizerInterface& rasterizer;
Core::System& system;
Tegra::MemoryManager& gpu_memory;
Core::Memory::Memory& cpu_memory;
std::unique_ptr<StreamBuffer> stream_buffer;
BufferType stream_buffer_handle;

View file

@ -74,8 +74,6 @@ public:
}
void WaitPendingFences() {
auto& gpu{system.GPU()};
auto& memory_manager{gpu.MemoryManager()};
while (!fences.empty()) {
TFence& current_fence = fences.front();
if (ShouldWait()) {
@ -83,8 +81,8 @@ public:
}
PopAsyncFlushes();
if (current_fence->IsSemaphore()) {
memory_manager.template Write<u32>(current_fence->GetAddress(),
current_fence->GetPayload());
gpu_memory.template Write<u32>(current_fence->GetAddress(),
current_fence->GetPayload());
} else {
gpu.IncrementSyncPoint(current_fence->GetPayload());
}
@ -93,13 +91,13 @@ public:
}
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} {}
explicit FenceManager(VideoCore::RasterizerInterface& rasterizer_, Tegra::GPU& gpu_,
TTextureCache& texture_cache_, TTBufferCache& buffer_cache_,
TQueryCache& query_cache_)
: rasterizer{rasterizer_}, gpu{gpu_}, gpu_memory{gpu.MemoryManager()},
texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, query_cache{query_cache_} {}
virtual ~FenceManager() {}
virtual ~FenceManager() = default;
/// Creates a Sync Point Fence Interface, does not create a backend fence if 'is_stubbed' is
/// true
@ -113,16 +111,15 @@ protected:
/// Waits until a fence has been signalled by the host GPU.
virtual void WaitFence(TFence& fence) = 0;
Core::System& system;
VideoCore::RasterizerInterface& rasterizer;
Tegra::GPU& gpu;
Tegra::MemoryManager& gpu_memory;
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)) {
@ -130,8 +127,8 @@ private:
}
PopAsyncFlushes();
if (current_fence->IsSemaphore()) {
memory_manager.template Write<u32>(current_fence->GetAddress(),
current_fence->GetPayload());
gpu_memory.template Write<u32>(current_fence->GetAddress(),
current_fence->GetPayload());
} else {
gpu.IncrementSyncPoint(current_fence->GetPayload());
}

View file

@ -28,8 +28,8 @@ namespace Tegra {
MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
GPU::GPU(Core::System& system_, bool is_async_)
: system{system_}, dma_pusher{std::make_unique<Tegra::DmaPusher>(system, *this)},
memory_manager{std::make_unique<Tegra::MemoryManager>(system)},
: system{system_}, memory_manager{std::make_unique<Tegra::MemoryManager>(system)},
dma_pusher{std::make_unique<Tegra::DmaPusher>(system, *this)},
maxwell_3d{std::make_unique<Engines::Maxwell3D>(system, *memory_manager)},
fermi_2d{std::make_unique<Engines::Fermi2D>()},
kepler_compute{std::make_unique<Engines::KeplerCompute>(system, *memory_manager)},

View file

@ -347,12 +347,11 @@ private:
protected:
Core::System& system;
std::unique_ptr<Tegra::MemoryManager> memory_manager;
std::unique_ptr<Tegra::DmaPusher> dma_pusher;
std::unique_ptr<VideoCore::RendererBase> renderer;
private:
std::unique_ptr<Tegra::MemoryManager> memory_manager;
/// Mapping of command subchannels to their bound engine ids
std::array<EngineID, 8> bound_engines = {};
/// 3D engine

View file

@ -95,10 +95,12 @@ template <class QueryCache, class CachedQuery, class CounterStream, class HostCo
class QueryPool>
class QueryCacheBase {
public:
explicit QueryCacheBase(Core::System& system, VideoCore::RasterizerInterface& rasterizer)
: system{system}, rasterizer{rasterizer}, streams{{CounterStream{
static_cast<QueryCache&>(*this),
VideoCore::QueryType::SamplesPassed}}} {}
explicit QueryCacheBase(VideoCore::RasterizerInterface& rasterizer_,
Tegra::Engines::Maxwell3D& maxwell3d_,
Tegra::MemoryManager& gpu_memory_)
: rasterizer{rasterizer_}, maxwell3d{maxwell3d_},
gpu_memory{gpu_memory_}, streams{{CounterStream{static_cast<QueryCache&>(*this),
VideoCore::QueryType::SamplesPassed}}} {}
void InvalidateRegion(VAddr addr, std::size_t size) {
std::unique_lock lock{mutex};
@ -118,29 +120,27 @@ public:
*/
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) {
std::unique_lock lock{mutex};
auto& memory_manager = system.GPU().MemoryManager();
const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr);
ASSERT(cpu_addr_opt);
VAddr cpu_addr = *cpu_addr_opt;
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr);
ASSERT(cpu_addr);
CachedQuery* query = TryGet(cpu_addr);
CachedQuery* query = TryGet(*cpu_addr);
if (!query) {
ASSERT_OR_EXECUTE(cpu_addr_opt, return;);
const auto host_ptr = memory_manager.GetPointer(gpu_addr);
ASSERT_OR_EXECUTE(cpu_addr, return;);
u8* const host_ptr = gpu_memory.GetPointer(gpu_addr);
query = Register(type, cpu_addr, host_ptr, timestamp.has_value());
query = Register(type, *cpu_addr, host_ptr, timestamp.has_value());
}
query->BindCounter(Stream(type).Current(), timestamp);
if (Settings::values.use_asynchronous_gpu_emulation.GetValue()) {
AsyncFlushQuery(cpu_addr);
AsyncFlushQuery(*cpu_addr);
}
}
/// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
void UpdateCounters() {
std::unique_lock lock{mutex};
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
Stream(VideoCore::QueryType::SamplesPassed).Update(regs.samplecnt_enable);
}
@ -270,8 +270,9 @@ private:
static constexpr std::uintptr_t PAGE_SIZE = 4096;
static constexpr unsigned PAGE_BITS = 12;
Core::System& system;
VideoCore::RasterizerInterface& rasterizer;
Tegra::Engines::Maxwell3D& maxwell3d;
Tegra::MemoryManager& gpu_memory;
std::recursive_mutex mutex;

View file

@ -106,11 +106,8 @@ public:
virtual void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {}
/// Initialize disk cached resources for the game being emulated
virtual void LoadDiskResources(const std::atomic_bool& stop_loading = false,
const DiskResourceLoadCallback& callback = {}) {}
/// Initializes renderer dirty flags
virtual void SetupDirtyFlags() {}
virtual void LoadDiskResources(u64 title_id, const std::atomic_bool& stop_loading,
const DiskResourceLoadCallback& callback) {}
/// Grant access to the Guest Driver Profile for recording/obtaining info on the guest driver.
GuestDriverProfile& AccessGuestDriverProfile() {

View file

@ -59,9 +59,10 @@ void Buffer::CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst
static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
}
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
OGLBufferCache::OGLBufferCache(VideoCore::RasterizerInterface& rasterizer,
Tegra::MemoryManager& gpu_memory, Core::Memory::Memory& cpu_memory,
const Device& device_, std::size_t stream_size)
: GenericBufferCache{rasterizer, system,
: GenericBufferCache{rasterizer, gpu_memory, cpu_memory,
std::make_unique<OGLStreamBuffer>(device_, stream_size, true)},
device{device_} {
if (!device.HasFastBufferSubData()) {

View file

@ -52,7 +52,8 @@ private:
using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer>;
class OGLBufferCache final : public GenericBufferCache {
public:
explicit OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
explicit OGLBufferCache(VideoCore::RasterizerInterface& rasterizer,
Tegra::MemoryManager& gpu_memory, Core::Memory::Memory& cpu_memory,
const Device& device, std::size_t stream_size);
~OGLBufferCache();

View file

@ -45,11 +45,10 @@ void GLInnerFence::Wait() {
glClientWaitSync(sync_object.handle, 0, GL_TIMEOUT_IGNORED);
}
FenceManagerOpenGL::FenceManagerOpenGL(Core::System& system,
VideoCore::RasterizerInterface& rasterizer,
FenceManagerOpenGL::FenceManagerOpenGL(VideoCore::RasterizerInterface& rasterizer, Tegra::GPU& gpu,
TextureCacheOpenGL& texture_cache,
OGLBufferCache& buffer_cache, QueryCache& query_cache)
: GenericFenceManager(system, rasterizer, texture_cache, buffer_cache, query_cache) {}
: GenericFenceManager{rasterizer, gpu, texture_cache, buffer_cache, query_cache} {}
Fence FenceManagerOpenGL::CreateFence(u32 value, bool is_stubbed) {
return std::make_shared<GLInnerFence>(value, is_stubbed);

View file

@ -37,9 +37,9 @@ using GenericFenceManager =
class FenceManagerOpenGL final : public GenericFenceManager {
public:
FenceManagerOpenGL(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
TextureCacheOpenGL& texture_cache, OGLBufferCache& buffer_cache,
QueryCache& query_cache);
explicit FenceManagerOpenGL(VideoCore::RasterizerInterface& rasterizer, Tegra::GPU& gpu,
TextureCacheOpenGL& texture_cache, OGLBufferCache& buffer_cache,
QueryCache& query_cache);
protected:
Fence CreateFence(u32 value, bool is_stubbed) override;

View file

@ -30,12 +30,13 @@ constexpr GLenum GetTarget(VideoCore::QueryType type) {
} // Anonymous namespace
QueryCache::QueryCache(Core::System& system, RasterizerOpenGL& gl_rasterizer)
QueryCache::QueryCache(RasterizerOpenGL& rasterizer, Tegra::Engines::Maxwell3D& maxwell3d,
Tegra::MemoryManager& gpu_memory)
: VideoCommon::QueryCacheBase<
QueryCache, CachedQuery, CounterStream, HostCounter,
std::vector<OGLQuery>>{system,
static_cast<VideoCore::RasterizerInterface&>(gl_rasterizer)},
gl_rasterizer{gl_rasterizer} {}
std::vector<OGLQuery>>{static_cast<VideoCore::RasterizerInterface&>(rasterizer),
maxwell3d, gpu_memory},
gl_rasterizer{rasterizer} {}
QueryCache::~QueryCache() = default;

View file

@ -29,7 +29,8 @@ using CounterStream = VideoCommon::CounterStreamBase<QueryCache, HostCounter>;
class QueryCache final : public VideoCommon::QueryCacheBase<QueryCache, CachedQuery, CounterStream,
HostCounter, std::vector<OGLQuery>> {
public:
explicit QueryCache(Core::System& system, RasterizerOpenGL& rasterizer);
explicit QueryCache(RasterizerOpenGL& rasterizer, Tegra::Engines::Maxwell3D& maxwell3d,
Tegra::MemoryManager& gpu_memory);
~QueryCache();
OGLQuery AllocateQuery(VideoCore::QueryType type);

View file

@ -153,16 +153,19 @@ void UpdateBindlessPointers(GLenum target, GLuint64EXT* pointers, std::size_t nu
} // Anonymous namespace
RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
const Device& device, ScreenInfo& info,
ProgramManager& program_manager, StateTracker& state_tracker)
: RasterizerAccelerated{system.Memory()}, device{device}, texture_cache{system, *this, device,
state_tracker},
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},
async_shaders{emu_window} {
RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gpu_,
Core::Memory::Memory& cpu_memory, const Device& device_,
ScreenInfo& screen_info_, ProgramManager& program_manager_,
StateTracker& state_tracker_)
: RasterizerAccelerated{cpu_memory}, gpu(gpu_), maxwell3d(gpu.Maxwell3D()),
kepler_compute(gpu.KeplerCompute()), gpu_memory(gpu.MemoryManager()), device(device_),
screen_info(screen_info_), program_manager(program_manager_), state_tracker(state_tracker_),
texture_cache(*this, maxwell3d, gpu_memory, device, state_tracker),
shader_cache(*this, emu_window, gpu, maxwell3d, kepler_compute, gpu_memory, device),
query_cache(*this, maxwell3d, gpu_memory),
buffer_cache(*this, gpu_memory, cpu_memory, device, STREAM_BUFFER_SIZE),
fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache),
async_shaders(emu_window) {
CheckExtensions();
unified_uniform_buffer.Create();
@ -196,8 +199,7 @@ void RasterizerOpenGL::CheckExtensions() {
}
void RasterizerOpenGL::SetupVertexFormat() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::VertexFormats]) {
return;
}
@ -217,7 +219,7 @@ void RasterizerOpenGL::SetupVertexFormat() {
}
flags[Dirty::VertexFormat0 + index] = false;
const auto attrib = gpu.regs.vertex_attrib_format[index];
const auto attrib = maxwell3d.regs.vertex_attrib_format[index];
const auto gl_index = static_cast<GLuint>(index);
// Disable constant attributes.
@ -241,8 +243,7 @@ void RasterizerOpenGL::SetupVertexFormat() {
}
void RasterizerOpenGL::SetupVertexBuffer() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::VertexBuffers]) {
return;
}
@ -253,7 +254,7 @@ void RasterizerOpenGL::SetupVertexBuffer() {
const bool use_unified_memory = device.HasVertexBufferUnifiedMemory();
// Upload all guest vertex arrays sequentially to our buffer
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_BINDINGS; ++index) {
if (!flags[Dirty::VertexBuffer0 + index]) {
continue;
@ -290,14 +291,13 @@ void RasterizerOpenGL::SetupVertexBuffer() {
}
void RasterizerOpenGL::SetupVertexInstances() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::VertexInstances]) {
return;
}
flags[Dirty::VertexInstances] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_ATTRIBUTES; ++index) {
if (!flags[Dirty::VertexInstance0 + index]) {
continue;
@ -313,7 +313,7 @@ void RasterizerOpenGL::SetupVertexInstances() {
GLintptr RasterizerOpenGL::SetupIndexBuffer() {
MICROPROFILE_SCOPE(OpenGL_Index);
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
const std::size_t size = CalculateIndexBufferSize();
const auto info = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, info.handle);
@ -322,15 +322,14 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
MICROPROFILE_SCOPE(OpenGL_Shader);
auto& gpu = system.GPU().Maxwell3D();
u32 clip_distances = 0;
for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
const auto& shader_config = gpu.regs.shader_config[index];
const auto& shader_config = maxwell3d.regs.shader_config[index];
const auto program{static_cast<Maxwell::ShaderProgram>(index)};
// Skip stages that are not enabled
if (!gpu.regs.IsShaderConfigEnabled(index)) {
if (!maxwell3d.regs.IsShaderConfigEnabled(index)) {
switch (program) {
case Maxwell::ShaderProgram::Geometry:
program_manager.UseGeometryShader(0);
@ -391,11 +390,11 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
}
SyncClipEnabled(clip_distances);
gpu.dirty.flags[Dirty::Shaders] = false;
maxwell3d.dirty.flags[Dirty::Shaders] = false;
}
std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
std::size_t size = 0;
for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
@ -413,34 +412,27 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
}
std::size_t RasterizerOpenGL::CalculateIndexBufferSize() const {
const auto& regs = system.GPU().Maxwell3D().regs;
return static_cast<std::size_t>(regs.index_array.count) *
static_cast<std::size_t>(regs.index_array.FormatSizeInBytes());
return static_cast<std::size_t>(maxwell3d.regs.index_array.count) *
static_cast<std::size_t>(maxwell3d.regs.index_array.FormatSizeInBytes());
}
void RasterizerOpenGL::LoadDiskResources(const std::atomic_bool& stop_loading,
void RasterizerOpenGL::LoadDiskResources(u64 title_id, const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback) {
shader_cache.LoadDiskCache(stop_loading, callback);
}
void RasterizerOpenGL::SetupDirtyFlags() {
state_tracker.Initialize();
shader_cache.LoadDiskCache(title_id, stop_loading, callback);
}
void RasterizerOpenGL::ConfigureFramebuffers() {
MICROPROFILE_SCOPE(OpenGL_Framebuffer);
auto& gpu = system.GPU().Maxwell3D();
if (!gpu.dirty.flags[VideoCommon::Dirty::RenderTargets]) {
if (!maxwell3d.dirty.flags[VideoCommon::Dirty::RenderTargets]) {
return;
}
gpu.dirty.flags[VideoCommon::Dirty::RenderTargets] = false;
maxwell3d.dirty.flags[VideoCommon::Dirty::RenderTargets] = false;
texture_cache.GuardRenderTargets(true);
View depth_surface = texture_cache.GetDepthBufferSurface(true);
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0);
// Bind the framebuffer surfaces
@ -472,8 +464,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
}
void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color, bool using_depth_stencil) {
auto& gpu = system.GPU().Maxwell3D();
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
texture_cache.GuardRenderTargets(true);
View color_surface;
@ -523,12 +514,11 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color, bool using_de
}
void RasterizerOpenGL::Clear() {
const auto& gpu = system.GPU().Maxwell3D();
if (!gpu.ShouldExecute()) {
if (!maxwell3d.ShouldExecute()) {
return;
}
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
bool use_color{};
bool use_depth{};
bool use_stencil{};
@ -593,7 +583,6 @@ void RasterizerOpenGL::Clear() {
void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
MICROPROFILE_SCOPE(OpenGL_Drawing);
auto& gpu = system.GPU().Maxwell3D();
query_cache.UpdateCounters();
@ -641,7 +630,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
if (invalidated) {
// When the stream buffer has been invalidated, we have to consider vertex buffers as dirty
auto& dirty = gpu.dirty.flags;
auto& dirty = maxwell3d.dirty.flags;
dirty[Dirty::VertexBuffers] = true;
for (int index = Dirty::VertexBuffer0; index <= Dirty::VertexBuffer31; ++index) {
dirty[index] = true;
@ -662,7 +651,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
// Setup emulation uniform buffer.
if (!device.UseAssemblyShaders()) {
MaxwellUniformData ubo;
ubo.SetFromRegs(gpu);
ubo.SetFromRegs(maxwell3d);
const auto info =
buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, info.handle, info.offset,
@ -671,7 +660,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
// Setup shaders and their used resources.
texture_cache.GuardSamplers(true);
const GLenum primitive_mode = MaxwellToGL::PrimitiveTopology(gpu.regs.draw.topology);
const GLenum primitive_mode = MaxwellToGL::PrimitiveTopology(maxwell3d.regs.draw.topology);
SetupShaders(primitive_mode);
texture_cache.GuardSamplers(false);
@ -688,14 +677,14 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
BeginTransformFeedback(primitive_mode);
const GLuint base_instance = static_cast<GLuint>(gpu.regs.vb_base_instance);
const GLuint base_instance = static_cast<GLuint>(maxwell3d.regs.vb_base_instance);
const GLsizei num_instances =
static_cast<GLsizei>(is_instanced ? gpu.mme_draw.instance_count : 1);
static_cast<GLsizei>(is_instanced ? maxwell3d.mme_draw.instance_count : 1);
if (is_indexed) {
const GLint base_vertex = static_cast<GLint>(gpu.regs.vb_element_base);
const GLsizei num_vertices = static_cast<GLsizei>(gpu.regs.index_array.count);
const GLint base_vertex = static_cast<GLint>(maxwell3d.regs.vb_element_base);
const GLsizei num_vertices = static_cast<GLsizei>(maxwell3d.regs.index_array.count);
const GLvoid* offset = reinterpret_cast<const GLvoid*>(index_buffer_offset);
const GLenum format = MaxwellToGL::IndexFormat(gpu.regs.index_array.format);
const GLenum format = MaxwellToGL::IndexFormat(maxwell3d.regs.index_array.format);
if (num_instances == 1 && base_instance == 0 && base_vertex == 0) {
glDrawElements(primitive_mode, num_vertices, format, offset);
} else if (num_instances == 1 && base_instance == 0) {
@ -714,8 +703,8 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
base_instance);
}
} else {
const GLint base_vertex = static_cast<GLint>(gpu.regs.vertex_buffer.first);
const GLsizei num_vertices = static_cast<GLsizei>(gpu.regs.vertex_buffer.count);
const GLint base_vertex = static_cast<GLint>(maxwell3d.regs.vertex_buffer.first);
const GLsizei num_vertices = static_cast<GLsizei>(maxwell3d.regs.vertex_buffer.count);
if (num_instances == 1 && base_instance == 0) {
glDrawArrays(primitive_mode, base_vertex, num_vertices);
} else if (base_instance == 0) {
@ -730,7 +719,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
++num_queued_commands;
system.GPU().TickWork();
gpu.TickWork();
}
void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
@ -753,7 +742,8 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
buffer_cache.Unmap();
const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
const auto& launch_desc = kepler_compute.launch_description;
program_manager.BindCompute(kernel->GetHandle());
glDispatchCompute(launch_desc.grid_dim_x, launch_desc.grid_dim_y, launch_desc.grid_dim_z);
++num_queued_commands;
}
@ -815,17 +805,14 @@ void RasterizerOpenGL::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);
gpu_memory.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;
@ -834,7 +821,6 @@ void RasterizerOpenGL::SignalSyncPoint(u32 value) {
}
void RasterizerOpenGL::ReleaseFences() {
auto& gpu{system.GPU()};
if (!gpu.IsAsync()) {
return;
}
@ -920,7 +906,7 @@ void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, Shader* sh
GL_FRAGMENT_PROGRAM_PARAMETER_BUFFER_NV};
MICROPROFILE_SCOPE(OpenGL_UBO);
const auto& stages = system.GPU().Maxwell3D().state.shader_stages;
const auto& stages = maxwell3d.state.shader_stages;
const auto& shader_stage = stages[stage_index];
const auto& entries = shader->GetEntries();
const bool use_unified = entries.use_unified_uniforms;
@ -945,7 +931,7 @@ void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, Shader* sh
void RasterizerOpenGL::SetupComputeConstBuffers(Shader* kernel) {
MICROPROFILE_SCOPE(OpenGL_UBO);
const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
const auto& launch_desc = kepler_compute.launch_description;
const auto& entries = kernel->GetEntries();
const bool use_unified = entries.use_unified_uniforms;
@ -1018,9 +1004,7 @@ void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, Shader* sh
GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV,
};
auto& gpu{system.GPU()};
auto& memory_manager{gpu.MemoryManager()};
const auto& cbufs{gpu.Maxwell3D().state.shader_stages[stage_index]};
const auto& cbufs{maxwell3d.state.shader_stages[stage_index]};
const auto& entries{shader->GetEntries().global_memory_entries};
std::array<GLuint64EXT, 32> pointers;
@ -1030,8 +1014,8 @@ void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, Shader* sh
u32 binding = assembly_shaders ? 0 : device.GetBaseBindings(stage_index).shader_storage_buffer;
for (const auto& entry : entries) {
const GPUVAddr addr{cbufs.const_buffers[entry.cbuf_index].address + entry.cbuf_offset};
const GPUVAddr gpu_addr{memory_manager.Read<u64>(addr)};
const u32 size{memory_manager.Read<u32>(addr + 8)};
const GPUVAddr gpu_addr{gpu_memory.Read<u64>(addr)};
const u32 size{gpu_memory.Read<u32>(addr + 8)};
SetupGlobalMemory(binding, entry, gpu_addr, size, &pointers[binding]);
++binding;
}
@ -1041,9 +1025,7 @@ void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, Shader* sh
}
void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
auto& gpu{system.GPU()};
auto& memory_manager{gpu.MemoryManager()};
const auto& cbufs{gpu.KeplerCompute().launch_description.const_buffer_config};
const auto& cbufs{kepler_compute.launch_description.const_buffer_config};
const auto& entries{kernel->GetEntries().global_memory_entries};
std::array<GLuint64EXT, 32> pointers;
@ -1052,8 +1034,8 @@ void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
u32 binding = 0;
for (const auto& entry : entries) {
const GPUVAddr addr{cbufs[entry.cbuf_index].Address() + entry.cbuf_offset};
const GPUVAddr gpu_addr{memory_manager.Read<u64>(addr)};
const u32 size{memory_manager.Read<u32>(addr + 8)};
const GPUVAddr gpu_addr{gpu_memory.Read<u64>(addr)};
const u32 size{gpu_memory.Read<u32>(addr + 8)};
SetupGlobalMemory(binding, entry, gpu_addr, size, &pointers[binding]);
++binding;
}
@ -1077,7 +1059,6 @@ void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& e
void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, Shader* shader) {
MICROPROFILE_SCOPE(OpenGL_Texture);
const auto& maxwell3d = system.GPU().Maxwell3D();
u32 binding = device.GetBaseBindings(stage_index).sampler;
for (const auto& entry : shader->GetEntries().samplers) {
const auto shader_type = static_cast<ShaderType>(stage_index);
@ -1090,11 +1071,10 @@ void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, Shader* shader
void RasterizerOpenGL::SetupComputeTextures(Shader* kernel) {
MICROPROFILE_SCOPE(OpenGL_Texture);
const auto& compute = system.GPU().KeplerCompute();
u32 binding = 0;
for (const auto& entry : kernel->GetEntries().samplers) {
for (std::size_t i = 0; i < entry.size; ++i) {
const auto texture = GetTextureInfo(compute, entry, ShaderType::Compute, i);
const auto texture = GetTextureInfo(kepler_compute, entry, ShaderType::Compute, i);
SetupTexture(binding++, texture, entry);
}
}
@ -1118,20 +1098,18 @@ void RasterizerOpenGL::SetupTexture(u32 binding, const Tegra::Texture::FullTextu
}
void RasterizerOpenGL::SetupDrawImages(std::size_t stage_index, Shader* shader) {
const auto& maxwell3d = system.GPU().Maxwell3D();
u32 binding = device.GetBaseBindings(stage_index).image;
for (const auto& entry : shader->GetEntries().images) {
const auto shader_type = static_cast<Tegra::Engines::ShaderType>(stage_index);
const auto shader_type = static_cast<ShaderType>(stage_index);
const auto tic = GetTextureInfo(maxwell3d, entry, shader_type).tic;
SetupImage(binding++, tic, entry);
}
}
void RasterizerOpenGL::SetupComputeImages(Shader* shader) {
const auto& compute = system.GPU().KeplerCompute();
u32 binding = 0;
for (const auto& entry : shader->GetEntries().images) {
const auto tic = GetTextureInfo(compute, entry, Tegra::Engines::ShaderType::Compute).tic;
const auto tic = GetTextureInfo(kepler_compute, entry, ShaderType::Compute).tic;
SetupImage(binding++, tic, entry);
}
}
@ -1151,9 +1129,8 @@ void RasterizerOpenGL::SetupImage(u32 binding, const Tegra::Texture::TICEntry& t
}
void RasterizerOpenGL::SyncViewport() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
const auto& regs = gpu.regs;
auto& flags = maxwell3d.dirty.flags;
const auto& regs = maxwell3d.regs;
const bool dirty_viewport = flags[Dirty::Viewports];
const bool dirty_clip_control = flags[Dirty::ClipControl];
@ -1225,25 +1202,23 @@ void RasterizerOpenGL::SyncViewport() {
}
void RasterizerOpenGL::SyncDepthClamp() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::DepthClampEnabled]) {
return;
}
flags[Dirty::DepthClampEnabled] = false;
oglEnable(GL_DEPTH_CLAMP, gpu.regs.view_volume_clip_control.depth_clamp_disabled == 0);
oglEnable(GL_DEPTH_CLAMP, maxwell3d.regs.view_volume_clip_control.depth_clamp_disabled == 0);
}
void RasterizerOpenGL::SyncClipEnabled(u32 clip_mask) {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::ClipDistances] && !flags[Dirty::Shaders]) {
return;
}
flags[Dirty::ClipDistances] = false;
clip_mask &= gpu.regs.clip_distance_enabled;
clip_mask &= maxwell3d.regs.clip_distance_enabled;
if (clip_mask == last_clip_distance_mask) {
return;
}
@ -1259,9 +1234,8 @@ void RasterizerOpenGL::SyncClipCoef() {
}
void RasterizerOpenGL::SyncCullMode() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
const auto& regs = gpu.regs;
auto& flags = maxwell3d.dirty.flags;
const auto& regs = maxwell3d.regs;
if (flags[Dirty::CullTest]) {
flags[Dirty::CullTest] = false;
@ -1276,26 +1250,24 @@ void RasterizerOpenGL::SyncCullMode() {
}
void RasterizerOpenGL::SyncPrimitiveRestart() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::PrimitiveRestart]) {
return;
}
flags[Dirty::PrimitiveRestart] = false;
if (gpu.regs.primitive_restart.enabled) {
if (maxwell3d.regs.primitive_restart.enabled) {
glEnable(GL_PRIMITIVE_RESTART);
glPrimitiveRestartIndex(gpu.regs.primitive_restart.index);
glPrimitiveRestartIndex(maxwell3d.regs.primitive_restart.index);
} else {
glDisable(GL_PRIMITIVE_RESTART);
}
}
void RasterizerOpenGL::SyncDepthTestState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
const auto& regs = maxwell3d.regs;
const auto& regs = gpu.regs;
if (flags[Dirty::DepthMask]) {
flags[Dirty::DepthMask] = false;
glDepthMask(regs.depth_write_enabled ? GL_TRUE : GL_FALSE);
@ -1313,14 +1285,13 @@ void RasterizerOpenGL::SyncDepthTestState() {
}
void RasterizerOpenGL::SyncStencilTestState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::StencilTest]) {
return;
}
flags[Dirty::StencilTest] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
oglEnable(GL_STENCIL_TEST, regs.stencil_enable);
glStencilFuncSeparate(GL_FRONT, MaxwellToGL::ComparisonOp(regs.stencil_front_func_func),
@ -1345,25 +1316,24 @@ void RasterizerOpenGL::SyncStencilTestState() {
}
void RasterizerOpenGL::SyncRasterizeEnable() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::RasterizeEnable]) {
return;
}
flags[Dirty::RasterizeEnable] = false;
oglEnable(GL_RASTERIZER_DISCARD, gpu.regs.rasterize_enable == 0);
oglEnable(GL_RASTERIZER_DISCARD, maxwell3d.regs.rasterize_enable == 0);
}
void RasterizerOpenGL::SyncPolygonModes() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::PolygonModes]) {
return;
}
flags[Dirty::PolygonModes] = false;
if (gpu.regs.fill_rectangle) {
const auto& regs = maxwell3d.regs;
if (regs.fill_rectangle) {
if (!GLAD_GL_NV_fill_rectangle) {
LOG_ERROR(Render_OpenGL, "GL_NV_fill_rectangle used and not supported");
glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
@ -1376,27 +1346,26 @@ void RasterizerOpenGL::SyncPolygonModes() {
return;
}
if (gpu.regs.polygon_mode_front == gpu.regs.polygon_mode_back) {
if (regs.polygon_mode_front == regs.polygon_mode_back) {
flags[Dirty::PolygonModeFront] = false;
flags[Dirty::PolygonModeBack] = false;
glPolygonMode(GL_FRONT_AND_BACK, MaxwellToGL::PolygonMode(gpu.regs.polygon_mode_front));
glPolygonMode(GL_FRONT_AND_BACK, MaxwellToGL::PolygonMode(regs.polygon_mode_front));
return;
}
if (flags[Dirty::PolygonModeFront]) {
flags[Dirty::PolygonModeFront] = false;
glPolygonMode(GL_FRONT, MaxwellToGL::PolygonMode(gpu.regs.polygon_mode_front));
glPolygonMode(GL_FRONT, MaxwellToGL::PolygonMode(regs.polygon_mode_front));
}
if (flags[Dirty::PolygonModeBack]) {
flags[Dirty::PolygonModeBack] = false;
glPolygonMode(GL_BACK, MaxwellToGL::PolygonMode(gpu.regs.polygon_mode_back));
glPolygonMode(GL_BACK, MaxwellToGL::PolygonMode(regs.polygon_mode_back));
}
}
void RasterizerOpenGL::SyncColorMask() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::ColorMasks]) {
return;
}
@ -1405,7 +1374,7 @@ void RasterizerOpenGL::SyncColorMask() {
const bool force = flags[Dirty::ColorMaskCommon];
flags[Dirty::ColorMaskCommon] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
if (regs.color_mask_common) {
if (!force && !flags[Dirty::ColorMask0]) {
return;
@ -1430,33 +1399,30 @@ void RasterizerOpenGL::SyncColorMask() {
}
void RasterizerOpenGL::SyncMultiSampleState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::MultisampleControl]) {
return;
}
flags[Dirty::MultisampleControl] = false;
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
oglEnable(GL_SAMPLE_ALPHA_TO_COVERAGE, regs.multisample_control.alpha_to_coverage);
oglEnable(GL_SAMPLE_ALPHA_TO_ONE, regs.multisample_control.alpha_to_one);
}
void RasterizerOpenGL::SyncFragmentColorClampState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::FragmentClampColor]) {
return;
}
flags[Dirty::FragmentClampColor] = false;
glClampColor(GL_CLAMP_FRAGMENT_COLOR, gpu.regs.frag_color_clamp ? GL_TRUE : GL_FALSE);
glClampColor(GL_CLAMP_FRAGMENT_COLOR, maxwell3d.regs.frag_color_clamp ? GL_TRUE : GL_FALSE);
}
void RasterizerOpenGL::SyncBlendState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
const auto& regs = gpu.regs;
auto& flags = maxwell3d.dirty.flags;
const auto& regs = maxwell3d.regs;
if (flags[Dirty::BlendColor]) {
flags[Dirty::BlendColor] = false;
@ -1513,14 +1479,13 @@ void RasterizerOpenGL::SyncBlendState() {
}
void RasterizerOpenGL::SyncLogicOpState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::LogicOp]) {
return;
}
flags[Dirty::LogicOp] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
if (regs.logic_op.enable) {
glEnable(GL_COLOR_LOGIC_OP);
glLogicOp(MaxwellToGL::LogicOp(regs.logic_op.operation));
@ -1530,14 +1495,13 @@ void RasterizerOpenGL::SyncLogicOpState() {
}
void RasterizerOpenGL::SyncScissorTest() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::Scissors]) {
return;
}
flags[Dirty::Scissors] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
for (std::size_t index = 0; index < Maxwell::NumViewports; ++index) {
if (!flags[Dirty::Scissor0 + index]) {
continue;
@ -1556,16 +1520,15 @@ void RasterizerOpenGL::SyncScissorTest() {
}
void RasterizerOpenGL::SyncPointState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::PointSize]) {
return;
}
flags[Dirty::PointSize] = false;
oglEnable(GL_POINT_SPRITE, gpu.regs.point_sprite_enable);
oglEnable(GL_POINT_SPRITE, maxwell3d.regs.point_sprite_enable);
if (gpu.regs.vp_point_size.enable) {
if (maxwell3d.regs.vp_point_size.enable) {
// By definition of GL_POINT_SIZE, it only matters if GL_PROGRAM_POINT_SIZE is disabled.
glEnable(GL_PROGRAM_POINT_SIZE);
return;
@ -1573,32 +1536,30 @@ void RasterizerOpenGL::SyncPointState() {
// Limit the point size to 1 since nouveau sometimes sets a point size of 0 (and that's invalid
// in OpenGL).
glPointSize(std::max(1.0f, gpu.regs.point_size));
glPointSize(std::max(1.0f, maxwell3d.regs.point_size));
glDisable(GL_PROGRAM_POINT_SIZE);
}
void RasterizerOpenGL::SyncLineState() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::LineWidth]) {
return;
}
flags[Dirty::LineWidth] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
oglEnable(GL_LINE_SMOOTH, regs.line_smooth_enable);
glLineWidth(regs.line_smooth_enable ? regs.line_width_smooth : regs.line_width_aliased);
}
void RasterizerOpenGL::SyncPolygonOffset() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::PolygonOffset]) {
return;
}
flags[Dirty::PolygonOffset] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
oglEnable(GL_POLYGON_OFFSET_FILL, regs.polygon_offset_fill_enable);
oglEnable(GL_POLYGON_OFFSET_LINE, regs.polygon_offset_line_enable);
oglEnable(GL_POLYGON_OFFSET_POINT, regs.polygon_offset_point_enable);
@ -1612,14 +1573,13 @@ void RasterizerOpenGL::SyncPolygonOffset() {
}
void RasterizerOpenGL::SyncAlphaTest() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::AlphaTest]) {
return;
}
flags[Dirty::AlphaTest] = false;
const auto& regs = gpu.regs;
const auto& regs = maxwell3d.regs;
if (regs.alpha_test_enabled && regs.rt_control.count > 1) {
LOG_WARNING(Render_OpenGL, "Alpha testing with more than one render target is not tested");
}
@ -1633,20 +1593,19 @@ void RasterizerOpenGL::SyncAlphaTest() {
}
void RasterizerOpenGL::SyncFramebufferSRGB() {
auto& gpu = system.GPU().Maxwell3D();
auto& flags = gpu.dirty.flags;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::FramebufferSRGB]) {
return;
}
flags[Dirty::FramebufferSRGB] = false;
oglEnable(GL_FRAMEBUFFER_SRGB, gpu.regs.framebuffer_srgb);
oglEnable(GL_FRAMEBUFFER_SRGB, maxwell3d.regs.framebuffer_srgb);
}
void RasterizerOpenGL::SyncTransformFeedback() {
// TODO(Rodrigo): Inject SKIP_COMPONENTS*_NV when required. An unimplemented message will signal
// when this is required.
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
static constexpr std::size_t STRIDE = 3;
std::array<GLint, 128 * STRIDE * Maxwell::NumTransformFeedbackBuffers> attribs;
@ -1698,7 +1657,7 @@ void RasterizerOpenGL::SyncTransformFeedback() {
}
void RasterizerOpenGL::BeginTransformFeedback(GLenum primitive_mode) {
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
if (regs.tfb_enabled == 0) {
return;
}
@ -1741,7 +1700,7 @@ void RasterizerOpenGL::BeginTransformFeedback(GLenum primitive_mode) {
}
void RasterizerOpenGL::EndTransformFeedback() {
const auto& regs = system.GPU().Maxwell3D().regs;
const auto& regs = maxwell3d.regs;
if (regs.tfb_enabled == 0) {
return;
}

View file

@ -36,8 +36,8 @@
#include "video_core/shader/async_shaders.h"
#include "video_core/textures/texture.h"
namespace Core {
class System;
namespace Core::Memory {
class Memory;
}
namespace Core::Frontend {
@ -55,9 +55,10 @@ struct DrawParameters;
class RasterizerOpenGL : public VideoCore::RasterizerAccelerated {
public:
explicit RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
const Device& device, ScreenInfo& info,
ProgramManager& program_manager, StateTracker& state_tracker);
explicit RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gpu,
Core::Memory::Memory& cpu_memory, const Device& device,
ScreenInfo& screen_info, ProgramManager& program_manager,
StateTracker& state_tracker);
~RasterizerOpenGL() override;
void Draw(bool is_indexed, bool is_instanced) override;
@ -83,9 +84,8 @@ public:
const Tegra::Engines::Fermi2D::Config& copy_config) override;
bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
u32 pixel_stride) override;
void LoadDiskResources(const std::atomic_bool& stop_loading,
void LoadDiskResources(u64 title_id, const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback) override;
void SetupDirtyFlags() override;
/// Returns true when there are commands queued to the OpenGL server.
bool AnyCommandQueued() const {
@ -237,7 +237,15 @@ private:
void SetupShaders(GLenum primitive_mode);
Tegra::GPU& gpu;
Tegra::Engines::Maxwell3D& maxwell3d;
Tegra::Engines::KeplerCompute& kepler_compute;
Tegra::MemoryManager& gpu_memory;
const Device& device;
ScreenInfo& screen_info;
ProgramManager& program_manager;
StateTracker& state_tracker;
TextureCacheOpenGL texture_cache;
ShaderCacheOpenGL shader_cache;
@ -247,10 +255,6 @@ private:
OGLBufferCache buffer_cache;
FenceManagerOpenGL fence_manager;
Core::System& system;
ScreenInfo& screen_info;
ProgramManager& program_manager;
StateTracker& state_tracker;
VideoCommon::Shader::AsyncShaders async_shaders;
static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;

View file

@ -239,12 +239,11 @@ std::unique_ptr<Shader> Shader::CreateStageFromMemory(
ProgramCode code_b, VideoCommon::Shader::AsyncShaders& async_shaders, VAddr cpu_addr) {
const auto shader_type = GetShaderType(program_type);
auto& gpu = params.system.GPU();
auto& gpu = params.gpu;
gpu.ShaderNotify().MarkSharderBuilding();
auto registry = std::make_shared<Registry>(shader_type, gpu.Maxwell3D());
if (!async_shaders.IsShaderAsync(params.system.GPU()) ||
!params.device.UseAsynchronousShaders()) {
if (!async_shaders.IsShaderAsync(gpu) || !params.device.UseAsynchronousShaders()) {
const ShaderIR ir(code, STAGE_MAIN_OFFSET, COMPILER_SETTINGS, *registry);
// TODO(Rodrigo): Handle VertexA shaders
// std::optional<ShaderIR> ir_b;
@ -287,11 +286,10 @@ std::unique_ptr<Shader> Shader::CreateStageFromMemory(
std::unique_ptr<Shader> Shader::CreateKernelFromMemory(const ShaderParameters& params,
ProgramCode code) {
auto& gpu = params.system.GPU();
auto& gpu = params.gpu;
gpu.ShaderNotify().MarkSharderBuilding();
auto& engine = gpu.KeplerCompute();
auto registry = std::make_shared<Registry>(ShaderType::Compute, engine);
auto registry = std::make_shared<Registry>(ShaderType::Compute, params.engine);
const ShaderIR ir(code, KERNEL_MAIN_OFFSET, COMPILER_SETTINGS, *registry);
const u64 uid = params.unique_identifier;
auto program = BuildShader(params.device, ShaderType::Compute, uid, ir, *registry);
@ -320,15 +318,20 @@ std::unique_ptr<Shader> Shader::CreateFromCache(const ShaderParameters& params,
precompiled_shader.registry, precompiled_shader.entries, precompiled_shader.program));
}
ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
Core::Frontend::EmuWindow& emu_window, const Device& device)
: VideoCommon::ShaderCache<Shader>{rasterizer}, system{system},
emu_window{emu_window}, device{device}, disk_cache{system} {}
ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer,
Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
Tegra::Engines::Maxwell3D& maxwell3d_,
Tegra::Engines::KeplerCompute& kepler_compute_,
Tegra::MemoryManager& gpu_memory_, const Device& device_)
: VideoCommon::ShaderCache<Shader>{rasterizer}, emu_window{emu_window_}, gpu{gpu_},
gpu_memory{gpu_memory_}, maxwell3d{maxwell3d_},
kepler_compute{kepler_compute_}, device{device_} {}
ShaderCacheOpenGL::~ShaderCacheOpenGL() = default;
void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
void ShaderCacheOpenGL::LoadDiskCache(u64 title_id, const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback) {
disk_cache.BindTitleID(title_id);
const std::optional transferable = disk_cache.LoadTransferable();
if (!transferable) {
return;
@ -481,21 +484,19 @@ ProgramSharedPtr ShaderCacheOpenGL::GeneratePrecompiledProgram(
Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program,
VideoCommon::Shader::AsyncShaders& async_shaders) {
if (!system.GPU().Maxwell3D().dirty.flags[Dirty::Shaders]) {
if (!maxwell3d.dirty.flags[Dirty::Shaders]) {
auto* last_shader = last_shaders[static_cast<std::size_t>(program)];
if (last_shader->IsBuilt()) {
return last_shader;
}
}
auto& memory_manager{system.GPU().MemoryManager()};
const GPUVAddr address{GetShaderAddress(system, program)};
const GPUVAddr address{GetShaderAddress(maxwell3d, program)};
if (device.UseAsynchronousShaders() && async_shaders.HasCompletedWork()) {
auto completed_work = async_shaders.GetCompletedWork();
for (auto& work : completed_work) {
Shader* shader = TryGet(work.cpu_address);
auto& gpu = system.GPU();
gpu.ShaderNotify().MarkShaderComplete();
if (shader == nullptr) {
continue;
@ -507,14 +508,13 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program,
shader->AsyncGLASMBuilt(std::move(work.program.glasm));
}
auto& registry = shader->GetRegistry();
ShaderDiskCacheEntry entry;
entry.type = work.shader_type;
entry.code = std::move(work.code);
entry.code_b = std::move(work.code_b);
entry.unique_identifier = work.uid;
auto& registry = shader->GetRegistry();
entry.bound_buffer = registry.GetBoundBuffer();
entry.graphics_info = registry.GetGraphicsInfo();
entry.keys = registry.GetKeys();
@ -525,28 +525,28 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program,
}
// Look up shader in the cache based on address
const auto cpu_addr{memory_manager.GpuToCpuAddress(address)};
const std::optional<VAddr> cpu_addr{gpu_memory.GpuToCpuAddress(address)};
if (Shader* const shader{cpu_addr ? TryGet(*cpu_addr) : null_shader.get()}) {
return last_shaders[static_cast<std::size_t>(program)] = shader;
}
const auto host_ptr{memory_manager.GetPointer(address)};
const u8* const host_ptr{gpu_memory.GetPointer(address)};
// No shader found - create a new one
ProgramCode code{GetShaderCode(memory_manager, address, host_ptr, false)};
ProgramCode code{GetShaderCode(gpu_memory, address, host_ptr, false)};
ProgramCode code_b;
if (program == Maxwell::ShaderProgram::VertexA) {
const GPUVAddr address_b{GetShaderAddress(system, Maxwell::ShaderProgram::VertexB)};
const u8* host_ptr_b = memory_manager.GetPointer(address_b);
code_b = GetShaderCode(memory_manager, address_b, host_ptr_b, false);
const GPUVAddr address_b{GetShaderAddress(maxwell3d, Maxwell::ShaderProgram::VertexB)};
const u8* host_ptr_b = gpu_memory.GetPointer(address_b);
code_b = GetShaderCode(gpu_memory, address_b, host_ptr_b, false);
}
const std::size_t code_size = code.size() * sizeof(u64);
const u64 unique_identifier = GetUniqueIdentifier(
GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b);
const ShaderParameters params{system, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
const ShaderParameters params{gpu, maxwell3d, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
std::unique_ptr<Shader> shader;
const auto found = runtime_cache.find(unique_identifier);
@ -568,21 +568,20 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program,
}
Shader* ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
auto& memory_manager{system.GPU().MemoryManager()};
const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)};
const std::optional<VAddr> cpu_addr{gpu_memory.GpuToCpuAddress(code_addr)};
if (Shader* const kernel = cpu_addr ? TryGet(*cpu_addr) : null_kernel.get()) {
return kernel;
}
const auto host_ptr{memory_manager.GetPointer(code_addr)};
// No kernel found, create a new one
ProgramCode code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
const u8* host_ptr{gpu_memory.GetPointer(code_addr)};
ProgramCode code{GetShaderCode(gpu_memory, code_addr, host_ptr, true)};
const std::size_t code_size{code.size() * sizeof(u64)};
const u64 unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
const ShaderParameters params{system, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
const ShaderParameters params{gpu, kepler_compute, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
std::unique_ptr<Shader> kernel;
const auto found = runtime_cache.find(unique_identifier);

View file

@ -25,8 +25,8 @@
#include "video_core/shader/shader_ir.h"
#include "video_core/shader_cache.h"
namespace Core {
class System;
namespace Tegra {
class MemoryManager;
}
namespace Core::Frontend {
@ -57,11 +57,12 @@ struct PrecompiledShader {
};
struct ShaderParameters {
Core::System& system;
Tegra::GPU& gpu;
Tegra::Engines::ConstBufferEngineInterface& engine;
ShaderDiskCacheOpenGL& disk_cache;
const Device& device;
VAddr cpu_addr;
u8* host_ptr;
const u8* host_ptr;
u64 unique_identifier;
};
@ -118,12 +119,14 @@ private:
class ShaderCacheOpenGL final : public VideoCommon::ShaderCache<Shader> {
public:
explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
Core::Frontend::EmuWindow& emu_window, const Device& device);
explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::Frontend::EmuWindow& emu_window,
Tegra::GPU& gpu, Tegra::Engines::Maxwell3D& maxwell3d,
Tegra::Engines::KeplerCompute& kepler_compute,
Tegra::MemoryManager& gpu_memory, const Device& device);
~ShaderCacheOpenGL() override;
/// Loads disk cache for the current game
void LoadDiskCache(const std::atomic_bool& stop_loading,
void LoadDiskCache(u64 title_id, const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback);
/// Gets the current specified shader stage program
@ -138,9 +141,13 @@ private:
const ShaderDiskCacheEntry& entry, const ShaderDiskCachePrecompiled& precompiled_entry,
const std::unordered_set<GLenum>& supported_formats);
Core::System& system;
Core::Frontend::EmuWindow& emu_window;
Tegra::GPU& gpu;
Tegra::MemoryManager& gpu_memory;
Tegra::Engines::Maxwell3D& maxwell3d;
Tegra::Engines::KeplerCompute& kepler_compute;
const Device& device;
ShaderDiskCacheOpenGL disk_cache;
std::unordered_map<u64, PrecompiledShader> runtime_cache;

View file

@ -206,13 +206,17 @@ bool ShaderDiskCacheEntry::Save(Common::FS::IOFile& file) const {
flat_bindless_samplers.size();
}
ShaderDiskCacheOpenGL::ShaderDiskCacheOpenGL(Core::System& system) : system{system} {}
ShaderDiskCacheOpenGL::ShaderDiskCacheOpenGL() = default;
ShaderDiskCacheOpenGL::~ShaderDiskCacheOpenGL() = default;
void ShaderDiskCacheOpenGL::BindTitleID(u64 title_id_) {
title_id = title_id_;
}
std::optional<std::vector<ShaderDiskCacheEntry>> ShaderDiskCacheOpenGL::LoadTransferable() {
// Skip games without title id
const bool has_title_id = system.CurrentProcess()->GetTitleID() != 0;
const bool has_title_id = title_id != 0;
if (!Settings::values.use_disk_shader_cache.GetValue() || !has_title_id) {
return std::nullopt;
}
@ -474,7 +478,7 @@ std::string ShaderDiskCacheOpenGL::GetBaseDir() const {
}
std::string ShaderDiskCacheOpenGL::GetTitleID() const {
return fmt::format("{:016X}", system.CurrentProcess()->GetTitleID());
return fmt::format("{:016X}", title_id);
}
} // namespace OpenGL

View file

@ -21,10 +21,6 @@
#include "video_core/engines/shader_type.h"
#include "video_core/shader/registry.h"
namespace Core {
class System;
}
namespace Common::FS {
class IOFile;
}
@ -70,9 +66,12 @@ struct ShaderDiskCachePrecompiled {
class ShaderDiskCacheOpenGL {
public:
explicit ShaderDiskCacheOpenGL(Core::System& system);
explicit ShaderDiskCacheOpenGL();
~ShaderDiskCacheOpenGL();
/// Binds a title ID for all future operations.
void BindTitleID(u64 title_id);
/// Loads transferable cache. If file has a old version or on failure, it deletes the file.
std::optional<std::vector<ShaderDiskCacheEntry>> LoadTransferable();
@ -157,8 +156,6 @@ private:
return LoadArrayFromPrecompiled(&object, 1);
}
Core::System& system;
// Stores whole precompiled cache which will be read from or saved to the precompiled chache
// file
FileSys::VectorVfsFile precompiled_cache_virtual_file;
@ -168,8 +165,11 @@ private:
// Stored transferable shaders
std::unordered_set<u64> stored_transferable;
/// Title ID to operate on
u64 title_id = 0;
// The cache has been loaded at boot
bool is_usable{};
bool is_usable = false;
};
} // namespace OpenGL

View file

@ -214,10 +214,8 @@ void SetupDirtyMisc(Tables& tables) {
} // Anonymous namespace
StateTracker::StateTracker(Core::System& system) : system{system} {}
void StateTracker::Initialize() {
auto& dirty = system.GPU().Maxwell3D().dirty;
StateTracker::StateTracker(Tegra::GPU& gpu) : flags{gpu.Maxwell3D().dirty.flags} {
auto& dirty = gpu.Maxwell3D().dirty;
auto& tables = dirty.tables;
SetupDirtyRenderTargets(tables);
SetupDirtyColorMasks(tables);

Some files were not shown because too many files have changed in this diff Show more