mirror of
https://github.com/yuzu-emu/yuzu-android.git
synced 2025-06-14 11:07:56 -05:00
OpenGL: Add GL_PRIMITIVES_GENERATED and GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN queries
This commit is contained in:
@ -18,16 +18,27 @@ namespace OpenGL {
|
||||
|
||||
namespace {
|
||||
|
||||
constexpr std::array<GLenum, VideoCore::NumQueryTypes> QueryTargets = {GL_SAMPLES_PASSED};
|
||||
|
||||
constexpr GLenum GetTarget(VideoCore::QueryType type) {
|
||||
return QueryTargets[static_cast<std::size_t>(type)];
|
||||
switch (type) {
|
||||
case VideoCore::QueryType::SamplesPassed:
|
||||
return GL_SAMPLES_PASSED;
|
||||
case VideoCore::QueryType::PrimitivesGenerated:
|
||||
return GL_PRIMITIVES_GENERATED;
|
||||
case VideoCore::QueryType::TfbPrimitivesWritten:
|
||||
return GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN;
|
||||
default:
|
||||
break;
|
||||
}
|
||||
UNIMPLEMENTED_MSG("Query type {}", type);
|
||||
return 0;
|
||||
}
|
||||
|
||||
} // Anonymous namespace
|
||||
|
||||
QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
|
||||
: QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {}
|
||||
: QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {
|
||||
EnableCounters();
|
||||
}
|
||||
|
||||
QueryCache::~QueryCache() = default;
|
||||
|
||||
@ -103,13 +114,13 @@ u64 CachedQuery::Flush([[maybe_unused]] bool async) {
|
||||
auto& stream = cache->Stream(type);
|
||||
const bool slice_counter = WaitPending() && stream.IsEnabled();
|
||||
if (slice_counter) {
|
||||
stream.Update(false);
|
||||
stream.Disable();
|
||||
}
|
||||
|
||||
auto result = VideoCommon::CachedQueryBase<HostCounter>::Flush();
|
||||
|
||||
if (slice_counter) {
|
||||
stream.Update(true);
|
||||
stream.Enable();
|
||||
}
|
||||
|
||||
return result;
|
||||
|
@ -51,6 +51,22 @@ constexpr size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
|
||||
void oglEnable(GLenum cap, bool state) {
|
||||
(state ? glEnable : glDisable)(cap);
|
||||
}
|
||||
|
||||
std::optional<VideoCore::QueryType> MaxwellToVideoCoreQuery(VideoCommon::QueryType type) {
|
||||
switch (type) {
|
||||
case VideoCommon::QueryType::PrimitivesGenerated:
|
||||
case VideoCommon::QueryType::VtgPrimitivesOut:
|
||||
return VideoCore::QueryType::PrimitivesGenerated;
|
||||
case VideoCommon::QueryType::ZPassPixelCount64:
|
||||
return VideoCore::QueryType::SamplesPassed;
|
||||
case VideoCommon::QueryType::StreamingPrimitivesSucceeded:
|
||||
// case VideoCommon::QueryType::StreamingByteCount:
|
||||
// TODO: StreamingByteCount = StreamingPrimitivesSucceeded * num_verts * vert_stride
|
||||
return VideoCore::QueryType::TfbPrimitivesWritten;
|
||||
default:
|
||||
return std::nullopt;
|
||||
}
|
||||
}
|
||||
} // Anonymous namespace
|
||||
|
||||
RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
|
||||
@ -212,7 +228,6 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
|
||||
|
||||
SCOPE_EXIT({ gpu.TickWork(); });
|
||||
gpu_memory->FlushCaching();
|
||||
query_cache.UpdateCounters();
|
||||
|
||||
GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
|
||||
if (!pipeline) {
|
||||
@ -330,7 +345,6 @@ void RasterizerOpenGL::DrawTexture() {
|
||||
MICROPROFILE_SCOPE(OpenGL_Drawing);
|
||||
|
||||
SCOPE_EXIT({ gpu.TickWork(); });
|
||||
query_cache.UpdateCounters();
|
||||
|
||||
texture_cache.SynchronizeGraphicsDescriptors();
|
||||
texture_cache.UpdateRenderTargets(false);
|
||||
@ -397,21 +411,28 @@ void RasterizerOpenGL::DispatchCompute() {
|
||||
}
|
||||
|
||||
void RasterizerOpenGL::ResetCounter(VideoCommon::QueryType type) {
|
||||
if (type == VideoCommon::QueryType::ZPassPixelCount64) {
|
||||
query_cache.ResetCounter(VideoCore::QueryType::SamplesPassed);
|
||||
const auto query_cache_type = MaxwellToVideoCoreQuery(type);
|
||||
if (!query_cache_type.has_value()) {
|
||||
UNIMPLEMENTED_MSG("Reset query type: {}", type);
|
||||
return;
|
||||
}
|
||||
query_cache.ResetCounter(*query_cache_type);
|
||||
}
|
||||
|
||||
void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
|
||||
VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
|
||||
if (type == VideoCommon::QueryType::ZPassPixelCount64) {
|
||||
if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
|
||||
query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, {gpu.GetTicks()});
|
||||
} else {
|
||||
query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, std::nullopt);
|
||||
}
|
||||
return;
|
||||
const auto query_cache_type = MaxwellToVideoCoreQuery(type);
|
||||
if (!query_cache_type.has_value()) {
|
||||
return QueryFallback(gpu_addr, type, flags, payload, subreport);
|
||||
}
|
||||
const bool has_timeout = True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout);
|
||||
const auto timestamp = has_timeout ? std::optional<u64>{gpu.GetTicks()} : std::nullopt;
|
||||
query_cache.Query(gpu_addr, *query_cache_type, timestamp);
|
||||
}
|
||||
|
||||
void RasterizerOpenGL::QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
|
||||
VideoCommon::QueryPropertiesFlags flags, u32 payload,
|
||||
u32 subreport) {
|
||||
if (type != VideoCommon::QueryType::Payload) {
|
||||
payload = 1u;
|
||||
}
|
||||
@ -1294,15 +1315,13 @@ void RasterizerOpenGL::BeginTransformFeedback(GraphicsPipeline* program, GLenum
|
||||
program->ConfigureTransformFeedback();
|
||||
|
||||
UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderType::TessellationInit) ||
|
||||
regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation) ||
|
||||
regs.IsShaderConfigEnabled(Maxwell::ShaderType::Geometry));
|
||||
UNIMPLEMENTED_IF(primitive_mode != GL_POINTS);
|
||||
regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation));
|
||||
|
||||
// We may have to call BeginTransformFeedbackNV here since they seem to call different
|
||||
// implementations on Nvidia's driver (the pointer is different) but we are using
|
||||
// ARB_transform_feedback3 features with NV_transform_feedback interactions and the ARB
|
||||
// extension doesn't define BeginTransformFeedback (without NV) interactions. It just works.
|
||||
glBeginTransformFeedback(GL_POINTS);
|
||||
glBeginTransformFeedback(primitive_mode);
|
||||
}
|
||||
|
||||
void RasterizerOpenGL::EndTransformFeedback() {
|
||||
|
@ -225,6 +225,9 @@ private:
|
||||
/// End a transform feedback
|
||||
void EndTransformFeedback();
|
||||
|
||||
void QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
|
||||
VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport);
|
||||
|
||||
Tegra::GPU& gpu;
|
||||
|
||||
const Device& device;
|
||||
|
Reference in New Issue
Block a user