gl_shader_cache: Use generic shader cache

Trivially port the generic shader cache to OpenGL.
master
ReinUsesLisp 2020-05-22 20:55:38 +07:00
parent dc27252352
commit b96f65b62b
4 changed files with 80 additions and 93 deletions

@ -30,6 +30,7 @@
#include "video_core/renderer_opengl/gl_shader_cache.h"
#include "video_core/renderer_opengl/maxwell_to_gl.h"
#include "video_core/renderer_opengl/renderer_opengl.h"
#include "video_core/shader_cache.h"
namespace OpenGL {
@ -282,7 +283,7 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
continue;
}
Shader shader{shader_cache.GetStageProgram(program)};
Shader* const shader = shader_cache.GetStageProgram(program);
if (device.UseAssemblyShaders()) {
// Check for ARB limitation. We only have 16 SSBOs per context state. To workaround this
@ -842,7 +843,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
return true;
}
void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, const Shader& shader) {
void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, Shader* shader) {
static constexpr std::array PARAMETER_LUT = {
GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV, GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,
GL_TESS_EVALUATION_PROGRAM_PARAMETER_BUFFER_NV, GL_GEOMETRY_PROGRAM_PARAMETER_BUFFER_NV,
@ -872,7 +873,7 @@ void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, const Shad
}
}
void RasterizerOpenGL::SetupComputeConstBuffers(const Shader& kernel) {
void RasterizerOpenGL::SetupComputeConstBuffers(Shader* kernel) {
MICROPROFILE_SCOPE(OpenGL_UBO);
const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
const auto& entries = kernel->GetEntries();
@ -941,7 +942,7 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
}
}
void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, const Shader& shader) {
void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, Shader* shader) {
auto& gpu{system.GPU()};
auto& memory_manager{gpu.MemoryManager()};
const auto cbufs{gpu.Maxwell3D().state.shader_stages[stage_index]};
@ -956,7 +957,7 @@ void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, const Shad
}
}
void RasterizerOpenGL::SetupComputeGlobalMemory(const Shader& kernel) {
void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
auto& gpu{system.GPU()};
auto& memory_manager{gpu.MemoryManager()};
const auto cbufs{gpu.KeplerCompute().launch_description.const_buffer_config};
@ -979,7 +980,7 @@ void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& e
static_cast<GLsizeiptr>(size));
}
void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, const Shader& shader) {
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;
@ -992,7 +993,7 @@ void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, const Shader&
}
}
void RasterizerOpenGL::SetupComputeTextures(const Shader& kernel) {
void RasterizerOpenGL::SetupComputeTextures(Shader* kernel) {
MICROPROFILE_SCOPE(OpenGL_Texture);
const auto& compute = system.GPU().KeplerCompute();
u32 binding = 0;
@ -1021,7 +1022,7 @@ void RasterizerOpenGL::SetupTexture(u32 binding, const Tegra::Texture::FullTextu
}
}
void RasterizerOpenGL::SetupDrawImages(std::size_t stage_index, const Shader& shader) {
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) {
@ -1031,7 +1032,7 @@ void RasterizerOpenGL::SetupDrawImages(std::size_t stage_index, const Shader& sh
}
}
void RasterizerOpenGL::SetupComputeImages(const Shader& shader) {
void RasterizerOpenGL::SetupComputeImages(Shader* shader) {
const auto& compute = system.GPU().KeplerCompute();
u32 binding = 0;
for (const auto& entry : shader->GetEntries().images) {

@ -100,10 +100,10 @@ private:
void ConfigureClearFramebuffer(bool using_color, bool using_depth_stencil);
/// Configures the current constbuffers to use for the draw command.
void SetupDrawConstBuffers(std::size_t stage_index, const Shader& shader);
void SetupDrawConstBuffers(std::size_t stage_index, Shader* shader);
/// Configures the current constbuffers to use for the kernel invocation.
void SetupComputeConstBuffers(const Shader& kernel);
void SetupComputeConstBuffers(Shader* kernel);
/// Configures a constant buffer.
void SetupConstBuffer(GLenum stage, u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
@ -111,30 +111,30 @@ private:
std::size_t unified_offset);
/// Configures the current global memory entries to use for the draw command.
void SetupDrawGlobalMemory(std::size_t stage_index, const Shader& shader);
void SetupDrawGlobalMemory(std::size_t stage_index, Shader* shader);
/// Configures the current global memory entries to use for the kernel invocation.
void SetupComputeGlobalMemory(const Shader& kernel);
void SetupComputeGlobalMemory(Shader* kernel);
/// Configures a constant buffer.
void SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry, GPUVAddr gpu_addr,
std::size_t size);
/// Configures the current textures to use for the draw command.
void SetupDrawTextures(std::size_t stage_index, const Shader& shader);
void SetupDrawTextures(std::size_t stage_index, Shader* shader);
/// Configures the textures used in a compute shader.
void SetupComputeTextures(const Shader& kernel);
void SetupComputeTextures(Shader* kernel);
/// Configures a texture.
void SetupTexture(u32 binding, const Tegra::Texture::FullTextureInfo& texture,
const SamplerEntry& entry);
/// Configures images in a graphics shader.
void SetupDrawImages(std::size_t stage_index, const Shader& shader);
void SetupDrawImages(std::size_t stage_index, Shader* shader);
/// Configures images in a compute shader.
void SetupComputeImages(const Shader& shader);
void SetupComputeImages(Shader* shader);
/// Configures an image.
void SetupImage(u32 binding, const Tegra::Texture::TICEntry& tic, const ImageEntry& entry);

@ -29,6 +29,7 @@
#include "video_core/shader/memory_util.h"
#include "video_core/shader/registry.h"
#include "video_core/shader/shader_ir.h"
#include "video_core/shader_cache.h"
namespace OpenGL {
@ -194,12 +195,9 @@ std::unordered_set<GLenum> GetSupportedFormats() {
} // Anonymous namespace
CachedShader::CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
std::shared_ptr<VideoCommon::Shader::Registry> registry,
ShaderEntries entries, ProgramSharedPtr program_)
: RasterizerCacheObject{cpu_addr}, registry{std::move(registry)}, entries{std::move(entries)},
size_in_bytes{size_in_bytes}, program{std::move(program_)} {
// Assign either the assembly program or source program. We can't have both.
Shader::Shader(std::shared_ptr<VideoCommon::Shader::Registry> registry, ShaderEntries entries,
ProgramSharedPtr program)
: registry{std::move(registry)}, entries{std::move(entries)}, program{std::move(program)} {
handle = program->assembly_program.handle;
if (handle == 0) {
handle = program->source_program.handle;
@ -207,16 +205,16 @@ CachedShader::CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
ASSERT(handle != 0);
}
CachedShader::~CachedShader() = default;
Shader::~Shader() = default;
GLuint CachedShader::GetHandle() const {
GLuint Shader::GetHandle() const {
DEBUG_ASSERT(registry->IsConsistent());
return handle;
}
Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
Maxwell::ShaderProgram program_type, ProgramCode code,
ProgramCode code_b) {
std::unique_ptr<Shader> Shader::CreateStageFromMemory(const ShaderParameters& params,
Maxwell::ShaderProgram program_type,
ProgramCode code, ProgramCode code_b) {
const auto shader_type = GetShaderType(program_type);
const std::size_t size_in_bytes = code.size() * sizeof(u64);
@ -241,12 +239,12 @@ Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
entry.bindless_samplers = registry->GetBindlessSamplers();
params.disk_cache.SaveEntry(std::move(entry));
return std::shared_ptr<CachedShader>(
new CachedShader(params.cpu_addr, size_in_bytes, std::move(registry),
MakeEntries(params.device, ir, shader_type), std::move(program)));
return std::unique_ptr<Shader>(new Shader(
std::move(registry), MakeEntries(params.device, ir, shader_type), std::move(program)));
}
Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code) {
std::unique_ptr<Shader> Shader::CreateKernelFromMemory(const ShaderParameters& params,
ProgramCode code) {
const std::size_t size_in_bytes = code.size() * sizeof(u64);
auto& engine = params.system.GPU().KeplerCompute();
@ -266,23 +264,23 @@ Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, Prog
entry.bindless_samplers = registry->GetBindlessSamplers();
params.disk_cache.SaveEntry(std::move(entry));
return std::shared_ptr<CachedShader>(
new CachedShader(params.cpu_addr, size_in_bytes, std::move(registry),
MakeEntries(params.device, ir, ShaderType::Compute), std::move(program)));
return std::unique_ptr<Shader>(new Shader(std::move(registry),
MakeEntries(params.device, ir, ShaderType::Compute),
std::move(program)));
}
Shader CachedShader::CreateFromCache(const ShaderParameters& params,
const PrecompiledShader& precompiled_shader,
std::size_t size_in_bytes) {
return std::shared_ptr<CachedShader>(
new CachedShader(params.cpu_addr, size_in_bytes, precompiled_shader.registry,
precompiled_shader.entries, precompiled_shader.program));
std::unique_ptr<Shader> Shader::CreateFromCache(const ShaderParameters& params,
const PrecompiledShader& precompiled_shader) {
return std::unique_ptr<Shader>(new Shader(
precompiled_shader.registry, precompiled_shader.entries, precompiled_shader.program));
}
ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
Core::Frontend::EmuWindow& emu_window, const Device& device)
: RasterizerCache{rasterizer}, system{system}, emu_window{emu_window}, device{device},
disk_cache{system} {}
: VideoCommon::ShaderCache<Shader>{rasterizer}, system{system},
emu_window{emu_window}, device{device}, disk_cache{system} {}
ShaderCacheOpenGL::~ShaderCacheOpenGL() = default;
void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback) {
@ -436,7 +434,7 @@ ProgramSharedPtr ShaderCacheOpenGL::GeneratePrecompiledProgram(
return program;
}
Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
if (!system.GPU().Maxwell3D().dirty.flags[Dirty::Shaders]) {
return last_shaders[static_cast<std::size_t>(program)];
}
@ -446,8 +444,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
// Look up shader in the cache based on address
const auto cpu_addr{memory_manager.GpuToCpuAddress(address)};
Shader shader{cpu_addr ? TryGet(*cpu_addr) : null_shader};
if (shader) {
if (Shader* const shader{cpu_addr ? TryGet(*cpu_addr) : null_shader.get()}) {
return last_shaders[static_cast<std::size_t>(program)] = shader;
}
@ -468,30 +465,29 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
const ShaderParameters params{system, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
std::unique_ptr<Shader> shader;
const auto found = runtime_cache.find(unique_identifier);
if (found == runtime_cache.end()) {
shader = CachedShader::CreateStageFromMemory(params, program, std::move(code),
std::move(code_b));
shader = Shader::CreateStageFromMemory(params, program, std::move(code), std::move(code_b));
} else {
const std::size_t size_in_bytes = code.size() * sizeof(u64);
shader = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
shader = Shader::CreateFromCache(params, found->second);
}
Shader* const result = shader.get();
if (cpu_addr) {
Register(shader);
Register(std::move(shader), *cpu_addr, code.size() * sizeof(u64));
} else {
null_shader = shader;
null_shader = std::move(shader);
}
return last_shaders[static_cast<std::size_t>(program)] = shader;
return last_shaders[static_cast<std::size_t>(program)] = result;
}
Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
Shader* ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
auto& memory_manager{system.GPU().MemoryManager()};
const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)};
auto kernel = cpu_addr ? TryGet(*cpu_addr) : null_kernel;
if (kernel) {
if (Shader* const kernel = cpu_addr ? TryGet(*cpu_addr) : null_kernel.get()) {
return kernel;
}
@ -503,20 +499,21 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
const ShaderParameters params{system, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
std::unique_ptr<Shader> kernel;
const auto found = runtime_cache.find(unique_identifier);
if (found == runtime_cache.end()) {
kernel = CachedShader::CreateKernelFromMemory(params, std::move(code));
kernel = Shader::CreateKernelFromMemory(params, std::move(code));
} else {
const std::size_t size_in_bytes = code.size() * sizeof(u64);
kernel = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
kernel = Shader::CreateFromCache(params, found->second);
}
Shader* const result = kernel.get();
if (cpu_addr) {
Register(kernel);
Register(std::move(kernel), *cpu_addr, code.size() * sizeof(u64));
} else {
null_kernel = kernel;
null_kernel = std::move(kernel);
}
return kernel;
return result;
}
} // namespace OpenGL

@ -18,12 +18,12 @@
#include "common/common_types.h"
#include "video_core/engines/shader_type.h"
#include "video_core/rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_shader_decompiler.h"
#include "video_core/renderer_opengl/gl_shader_disk_cache.h"
#include "video_core/shader/registry.h"
#include "video_core/shader/shader_ir.h"
#include "video_core/shader_cache.h"
namespace Core {
class System;
@ -35,12 +35,10 @@ class EmuWindow;
namespace OpenGL {
class CachedShader;
class Device;
class RasterizerOpenGL;
struct UnspecializedShader;
using Shader = std::shared_ptr<CachedShader>;
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
struct ProgramHandle {
@ -64,62 +62,53 @@ struct ShaderParameters {
u64 unique_identifier;
};
class CachedShader final : public RasterizerCacheObject {
class Shader final {
public:
~CachedShader();
~Shader();
/// Gets the GL program handle for the shader
GLuint GetHandle() const;
/// Returns the size in bytes of the shader
std::size_t GetSizeInBytes() const override {
return size_in_bytes;
}
/// Gets the shader entries for the shader
const ShaderEntries& GetEntries() const {
return entries;
}
static Shader CreateStageFromMemory(const ShaderParameters& params,
Maxwell::ShaderProgram program_type,
ProgramCode program_code, ProgramCode program_code_b);
static Shader CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code);
static std::unique_ptr<Shader> CreateStageFromMemory(const ShaderParameters& params,
Maxwell::ShaderProgram program_type,
ProgramCode program_code,
ProgramCode program_code_b);
static std::unique_ptr<Shader> CreateKernelFromMemory(const ShaderParameters& params,
ProgramCode code);
static Shader CreateFromCache(const ShaderParameters& params,
const PrecompiledShader& precompiled_shader,
std::size_t size_in_bytes);
static std::unique_ptr<Shader> CreateFromCache(const ShaderParameters& params,
const PrecompiledShader& precompiled_shader);
private:
explicit CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
std::shared_ptr<VideoCommon::Shader::Registry> registry,
ShaderEntries entries, ProgramSharedPtr program);
explicit Shader(std::shared_ptr<VideoCommon::Shader::Registry> registry, ShaderEntries entries,
ProgramSharedPtr program);
std::shared_ptr<VideoCommon::Shader::Registry> registry;
ShaderEntries entries;
std::size_t size_in_bytes = 0;
ProgramSharedPtr program;
GLuint handle = 0;
};
class ShaderCacheOpenGL final : public RasterizerCache<Shader> {
class ShaderCacheOpenGL final : public VideoCommon::ShaderCache<Shader> {
public:
explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
Core::Frontend::EmuWindow& emu_window, const Device& device);
~ShaderCacheOpenGL() override;
/// Loads disk cache for the current game
void LoadDiskCache(const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback);
/// Gets the current specified shader stage program
Shader GetStageProgram(Maxwell::ShaderProgram program);
Shader* GetStageProgram(Maxwell::ShaderProgram program);
/// Gets a compute kernel in the passed address
Shader GetComputeKernel(GPUVAddr code_addr);
protected:
// We do not have to flush this cache as things in it are never modified by us.
void FlushObjectInner(const Shader& object) override {}
Shader* GetComputeKernel(GPUVAddr code_addr);
private:
ProgramSharedPtr GeneratePrecompiledProgram(
@ -132,10 +121,10 @@ private:
ShaderDiskCacheOpenGL disk_cache;
std::unordered_map<u64, PrecompiledShader> runtime_cache;
Shader null_shader{};
Shader null_kernel{};
std::unique_ptr<Shader> null_shader;
std::unique_ptr<Shader> null_kernel;
std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
std::array<Shader*, Maxwell::MaxShaderProgram> last_shaders{};
};
} // namespace OpenGL