aboutsummaryrefslogtreecommitdiff
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/gpu_thread.cpp2
-rw-r--r--src/video_core/renderer_opengl/gl_device.cpp54
-rw-r--r--src/video_core/renderer_opengl/gl_device.h6
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp66
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.h1
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp61
-rw-r--r--src/video_core/renderer_opengl/gl_shader_gen.cpp19
-rw-r--r--src/video_core/renderer_vulkan/vk_device.cpp136
-rw-r--r--src/video_core/renderer_vulkan/vk_device.h58
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.cpp104
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.h8
11 files changed, 356 insertions, 159 deletions
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index 1e2ff46b0..3f0939ec9 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -75,7 +75,7 @@ void ThreadManager::StartThread(VideoCore::RendererBase& renderer, Tegra::DmaPus
void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
const u64 fence{PushCommand(SubmitListCommand(std::move(entries)))};
- const s64 synchronization_ticks{Core::Timing::usToCycles(9000)};
+ const s64 synchronization_ticks{Core::Timing::usToCycles(std::chrono::microseconds{9000})};
system.CoreTiming().ScheduleEvent(synchronization_ticks, synchronization_event, fence);
}
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index 1d1581f49..65a88b06c 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -2,11 +2,14 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include <array>
#include <cstddef>
#include <glad/glad.h>
#include "common/logging/log.h"
+#include "common/scope_exit.h"
#include "video_core/renderer_opengl/gl_device.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
namespace OpenGL {
@@ -24,6 +27,7 @@ Device::Device() {
max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);
max_varyings = GetInteger<u32>(GL_MAX_VARYING_VECTORS);
has_variable_aoffi = TestVariableAoffi();
+ has_component_indexing_bug = TestComponentIndexingBug();
}
Device::Device(std::nullptr_t) {
@@ -31,6 +35,7 @@ Device::Device(std::nullptr_t) {
max_vertex_attributes = 16;
max_varyings = 15;
has_variable_aoffi = true;
+ has_component_indexing_bug = false;
}
bool Device::TestVariableAoffi() {
@@ -52,4 +57,53 @@ void main() {
return supported;
}
+bool Device::TestComponentIndexingBug() {
+ constexpr char log_message[] = "Renderer_ComponentIndexingBug: {}";
+ const GLchar* COMPONENT_TEST = R"(#version 430 core
+layout (std430, binding = 0) buffer OutputBuffer {
+ uint output_value;
+};
+layout (std140, binding = 0) uniform InputBuffer {
+ uvec4 input_value[4096];
+};
+layout (location = 0) uniform uint idx;
+void main() {
+ output_value = input_value[idx >> 2][idx & 3];
+})";
+ const GLuint shader{glCreateShaderProgramv(GL_VERTEX_SHADER, 1, &COMPONENT_TEST)};
+ SCOPE_EXIT({ glDeleteProgram(shader); });
+ glUseProgram(shader);
+
+ OGLVertexArray vao;
+ vao.Create();
+ glBindVertexArray(vao.handle);
+
+ constexpr std::array<GLuint, 8> values{0, 0, 0, 0, 0x1236327, 0x985482, 0x872753, 0x2378432};
+ OGLBuffer ubo;
+ ubo.Create();
+ glNamedBufferData(ubo.handle, sizeof(values), values.data(), GL_STATIC_DRAW);
+ glBindBufferBase(GL_UNIFORM_BUFFER, 0, ubo.handle);
+
+ OGLBuffer ssbo;
+ ssbo.Create();
+ glNamedBufferStorage(ssbo.handle, sizeof(GLuint), nullptr, GL_CLIENT_STORAGE_BIT);
+
+ for (GLuint index = 4; index < 8; ++index) {
+ glInvalidateBufferData(ssbo.handle);
+ glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, ssbo.handle);
+
+ glProgramUniform1ui(shader, 0, index);
+ glDrawArrays(GL_POINTS, 0, 1);
+
+ GLuint result;
+ glGetNamedBufferSubData(ssbo.handle, 0, sizeof(result), &result);
+ if (result != values.at(index)) {
+ LOG_INFO(Render_OpenGL, log_message, true);
+ return true;
+ }
+ }
+ LOG_INFO(Render_OpenGL, log_message, false);
+ return false;
+}
+
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index de8490682..8c8c93760 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -30,13 +30,19 @@ public:
return has_variable_aoffi;
}
+ bool HasComponentIndexingBug() const {
+ return has_component_indexing_bug;
+ }
+
private:
static bool TestVariableAoffi();
+ static bool TestComponentIndexingBug();
std::size_t uniform_buffer_alignment{};
u32 max_vertex_attributes{};
u32 max_varyings{};
bool has_variable_aoffi{};
+ bool has_component_indexing_bug{};
};
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index d66252224..ac8a9e6b7 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -35,8 +35,8 @@ struct UnspecializedShader {
namespace {
/// Gets the address for the specified shader stage program
-GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
- const auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
+GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) {
+ const auto& gpu{system.GPU().Maxwell3D()};
const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
return gpu.regs.code_address.CodeAddress() + shader_config.offset;
}
@@ -350,7 +350,8 @@ ShaderDiskCacheUsage CachedShader::GetUsage(GLenum primitive_mode,
ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
Core::Frontend::EmuWindow& emu_window, const Device& device)
- : RasterizerCache{rasterizer}, emu_window{emu_window}, device{device}, disk_cache{system} {}
+ : RasterizerCache{rasterizer}, system{system}, emu_window{emu_window}, device{device},
+ disk_cache{system} {}
void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
const VideoCore::DiskResourceLoadCallback& callback) {
@@ -546,42 +547,45 @@ std::unordered_map<u64, UnspecializedShader> ShaderCacheOpenGL::GenerateUnspecia
}
Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
- if (!Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.shaders) {
- return last_shaders[static_cast<u32>(program)];
+ if (!system.GPU().Maxwell3D().dirty_flags.shaders) {
+ return last_shaders[static_cast<std::size_t>(program)];
}
- auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
- const GPUVAddr program_addr{GetShaderAddress(program)};
+ auto& memory_manager{system.GPU().MemoryManager()};
+ const GPUVAddr program_addr{GetShaderAddress(system, program)};
// Look up shader in the cache based on address
- const auto& host_ptr{memory_manager.GetPointer(program_addr)};
+ const auto host_ptr{memory_manager.GetPointer(program_addr)};
Shader shader{TryGet(host_ptr)};
+ if (shader) {
+ return last_shaders[static_cast<std::size_t>(program)] = shader;
+ }
- if (!shader) {
- // No shader found - create a new one
- ProgramCode program_code{GetShaderCode(memory_manager, program_addr, host_ptr)};
- ProgramCode program_code_b;
- if (program == Maxwell::ShaderProgram::VertexA) {
- const GPUVAddr program_addr_b{GetShaderAddress(Maxwell::ShaderProgram::VertexB)};
- program_code_b = GetShaderCode(memory_manager, program_addr_b,
- memory_manager.GetPointer(program_addr_b));
- }
- const u64 unique_identifier = GetUniqueIdentifier(program, program_code, program_code_b);
- const VAddr cpu_addr{*memory_manager.GpuToCpuAddress(program_addr)};
- const auto found = precompiled_shaders.find(unique_identifier);
- if (found != precompiled_shaders.end()) {
- shader =
- std::make_shared<CachedShader>(cpu_addr, unique_identifier, program, disk_cache,
- precompiled_programs, found->second, host_ptr);
- } else {
- shader = std::make_shared<CachedShader>(
- device, cpu_addr, unique_identifier, program, disk_cache, precompiled_programs,
- std::move(program_code), std::move(program_code_b), host_ptr);
- }
- Register(shader);
+ // No shader found - create a new one
+ ProgramCode program_code{GetShaderCode(memory_manager, program_addr, host_ptr)};
+ ProgramCode program_code_b;
+ if (program == Maxwell::ShaderProgram::VertexA) {
+ const GPUVAddr program_addr_b{GetShaderAddress(system, Maxwell::ShaderProgram::VertexB)};
+ program_code_b = GetShaderCode(memory_manager, program_addr_b,
+ memory_manager.GetPointer(program_addr_b));
+ }
+
+ const u64 unique_identifier = GetUniqueIdentifier(program, program_code, program_code_b);
+ const VAddr cpu_addr{*memory_manager.GpuToCpuAddress(program_addr)};
+ const auto found = precompiled_shaders.find(unique_identifier);
+ if (found != precompiled_shaders.end()) {
+ // Create a shader from the cache
+ shader = std::make_shared<CachedShader>(cpu_addr, unique_identifier, program, disk_cache,
+ precompiled_programs, found->second, host_ptr);
+ } else {
+ // Create a shader from guest memory
+ shader = std::make_shared<CachedShader>(
+ device, cpu_addr, unique_identifier, program, disk_cache, precompiled_programs,
+ std::move(program_code), std::move(program_code_b), host_ptr);
}
+ Register(shader);
- return last_shaders[static_cast<u32>(program)] = shader;
+ return last_shaders[static_cast<std::size_t>(program)] = shader;
}
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 64e5a5594..09bd0761d 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -137,6 +137,7 @@ private:
CachedProgram GeneratePrecompiledProgram(const ShaderDiskCacheDump& dump,
const std::set<GLenum>& supported_formats);
+ Core::System& system;
Core::Frontend::EmuWindow& emu_window;
const Device& device;
ShaderDiskCacheOpenGL disk_cache;
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index e9f8d40db..29de5c9db 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -45,7 +45,6 @@ struct TextureAoffi {};
using TextureArgument = std::pair<Type, Node>;
using TextureIR = std::variant<TextureAoffi, TextureArgument>;
-enum : u32 { POSITION_VARYING_LOCATION = 0, GENERIC_VARYING_START_LOCATION = 1 };
constexpr u32 MAX_CONSTBUFFER_ELEMENTS =
static_cast<u32>(RasterizerOpenGL::MaxConstbufferSize) / (4 * sizeof(float));
@@ -247,6 +246,12 @@ private:
code.AddLine("layout ({}, max_vertices = {}) out;", topology, max_vertices);
code.AddNewLine();
+ code.AddLine("in gl_PerVertex {{");
+ ++code.scope;
+ code.AddLine("vec4 gl_Position;");
+ --code.scope;
+ code.AddLine("}} gl_in[];");
+
DeclareVertexRedeclarations();
}
@@ -349,7 +354,7 @@ private:
}
void DeclareInputAttribute(Attribute::Index index, bool skip_unused) {
- const u32 generic_index{GetGenericAttributeIndex(index)};
+ const u32 location{GetGenericAttributeIndex(index)};
std::string name{GetInputAttribute(index)};
if (stage == ShaderStage::Geometry) {
@@ -358,19 +363,13 @@ private:
std::string suffix;
if (stage == ShaderStage::Fragment) {
- const auto input_mode{header.ps.GetAttributeUse(generic_index)};
+ const auto input_mode{header.ps.GetAttributeUse(location)};
if (skip_unused && input_mode == AttributeUse::Unused) {
return;
}
suffix = GetInputFlags(input_mode);
}
- u32 location = generic_index;
- if (stage != ShaderStage::Vertex) {
- // If inputs are varyings, add an offset
- location += GENERIC_VARYING_START_LOCATION;
- }
-
code.AddLine("layout (location = {}) {} in vec4 {};", location, suffix, name);
}
@@ -395,7 +394,7 @@ private:
}
void DeclareOutputAttribute(Attribute::Index index) {
- const u32 location{GetGenericAttributeIndex(index) + GENERIC_VARYING_START_LOCATION};
+ const u32 location{GetGenericAttributeIndex(index)};
code.AddLine("layout (location = {}) out vec4 {};", location, GetOutputAttribute(index));
}
@@ -577,9 +576,26 @@ private:
if (std::holds_alternative<OperationNode>(*offset)) {
// Indirect access
const std::string final_offset = code.GenerateTemporary();
- code.AddLine("uint {} = (ftou({}) / 4);", final_offset, Visit(offset));
- return fmt::format("{}[{} / 4][{} % 4]", GetConstBuffer(cbuf->GetIndex()),
- final_offset, final_offset);
+ code.AddLine("uint {} = ftou({}) >> 2;", final_offset, Visit(offset));
+
+ if (!device.HasComponentIndexingBug()) {
+ return fmt::format("{}[{} >> 2][{} & 3]", GetConstBuffer(cbuf->GetIndex()),
+ final_offset, final_offset);
+ }
+
+ // AMD's proprietary GLSL compiler emits ill code for variable component access.
+ // To bypass this driver bug generate 4 ifs, one per each component.
+ const std::string pack = code.GenerateTemporary();
+ code.AddLine("vec4 {} = {}[{} >> 2];", pack, GetConstBuffer(cbuf->GetIndex()),
+ final_offset);
+
+ const std::string result = code.GenerateTemporary();
+ code.AddLine("float {};", result);
+ for (u32 swizzle = 0; swizzle < 4; ++swizzle) {
+ code.AddLine("if (({} & 3) == {}) {} = {}{};", final_offset, swizzle, result,
+ pack, GetSwizzle(swizzle));
+ }
+ return result;
}
UNREACHABLE_MSG("Unmanaged offset node type");
@@ -633,10 +649,14 @@ private:
switch (attribute) {
case Attribute::Index::Position:
- if (stage != ShaderStage::Fragment) {
- return GeometryPass("position") + GetSwizzle(element);
- } else {
+ switch (stage) {
+ case ShaderStage::Geometry:
+ return fmt::format("gl_in[ftou({})].gl_Position{}", Visit(buffer),
+ GetSwizzle(element));
+ case ShaderStage::Fragment:
return element == 3 ? "1.0f" : ("gl_FragCoord"s + GetSwizzle(element));
+ default:
+ UNREACHABLE();
}
case Attribute::Index::PointCoord:
switch (element) {
@@ -921,7 +941,7 @@ private:
target = [&]() -> std::string {
switch (const auto attribute = abuf->GetIndex(); abuf->GetIndex()) {
case Attribute::Index::Position:
- return "position"s + GetSwizzle(abuf->GetElement());
+ return "gl_Position"s + GetSwizzle(abuf->GetElement());
case Attribute::Index::PointSize:
return "gl_PointSize";
case Attribute::Index::ClipDistances0123:
@@ -1506,9 +1526,7 @@ private:
// If a geometry shader is attached, it will always flip (it's the last stage before
// fragment). For more info about flipping, refer to gl_shader_gen.cpp.
- code.AddLine("position.xy *= viewport_flip.xy;");
- code.AddLine("gl_Position = position;");
- code.AddLine("position.w = 1.0;");
+ code.AddLine("gl_Position.xy *= viewport_flip.xy;");
code.AddLine("EmitVertex();");
return {};
}
@@ -1746,8 +1764,7 @@ private:
}
u32 GetNumPhysicalVaryings() const {
- return std::min<u32>(device.GetMaxVaryings() - GENERIC_VARYING_START_LOCATION,
- Maxwell::NumVaryings);
+ return std::min<u32>(device.GetMaxVaryings(), Maxwell::NumVaryings);
}
const Device& device;
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp
index d2bb705a9..c845b29aa 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp
@@ -23,8 +23,6 @@ ProgramResult GenerateVertexShader(const Device& device, const ShaderSetup& setu
out += GetCommonDeclarations();
out += R"(
-layout (location = 0) out vec4 position;
-
layout (std140, binding = EMULATION_UBO_BINDING) uniform vs_config {
vec4 viewport_flip;
uvec4 config_pack; // instance_id, flip_stage, y_direction, padding
@@ -48,7 +46,6 @@ layout (std140, binding = EMULATION_UBO_BINDING) uniform vs_config {
out += R"(
void main() {
- position = vec4(0.0, 0.0, 0.0, 0.0);
execute_vertex();
)";
@@ -59,19 +56,12 @@ void main() {
out += R"(
// Set Position Y direction
- position.y *= utof(config_pack[2]);
+ gl_Position.y *= utof(config_pack[2]);
// Check if the flip stage is VertexB
// Config pack's second value is flip_stage
if (config_pack[1] == 1) {
// Viewport can be flipped, which is unsupported by glViewport
- position.xy *= viewport_flip.xy;
- }
- gl_Position = position;
-
- // TODO(bunnei): This is likely a hack, position.w should be interpolated as 1.0
- // For now, this is here to bring order in lieu of proper emulation
- if (config_pack[1] == 1) {
- position.w = 1.0;
+ gl_Position.xy *= viewport_flip.xy;
}
})";
@@ -85,9 +75,6 @@ ProgramResult GenerateGeometryShader(const Device& device, const ShaderSetup& se
out += GetCommonDeclarations();
out += R"(
-layout (location = 0) in vec4 gs_position[];
-layout (location = 0) out vec4 position;
-
layout (std140, binding = EMULATION_UBO_BINDING) uniform gs_config {
vec4 viewport_flip;
uvec4 config_pack; // instance_id, flip_stage, y_direction, padding
@@ -124,8 +111,6 @@ layout (location = 5) out vec4 FragColor5;
layout (location = 6) out vec4 FragColor6;
layout (location = 7) out vec4 FragColor7;
-layout (location = 0) in noperspective vec4 position;
-
layout (std140, binding = EMULATION_UBO_BINDING) uniform fs_config {
vec4 viewport_flip;
uvec4 config_pack; // instance_id, flip_stage, y_direction, padding
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index 00242ecbe..3b966ddc3 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -18,6 +18,7 @@ constexpr std::array<vk::Format, 3> Depth24UnormS8Uint = {
vk::Format::eD32SfloatS8Uint, vk::Format::eD16UnormS8Uint, {}};
constexpr std::array<vk::Format, 3> Depth16UnormS8Uint = {
vk::Format::eD24UnormS8Uint, vk::Format::eD32SfloatS8Uint, {}};
+constexpr std::array<vk::Format, 2> Astc = {vk::Format::eA8B8G8R8UnormPack32, {}};
} // namespace Alternatives
@@ -51,15 +52,19 @@ VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice phy
: physical{physical}, format_properties{GetFormatProperties(dldi, physical)} {
SetupFamilies(dldi, surface);
SetupProperties(dldi);
+ SetupFeatures(dldi);
}
VKDevice::~VKDevice() = default;
bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
- const auto queue_cis = GetDeviceQueueCreateInfos();
- vk::PhysicalDeviceFeatures device_features{};
+ vk::PhysicalDeviceFeatures device_features;
+ device_features.vertexPipelineStoresAndAtomics = true;
+ device_features.independentBlend = true;
+ device_features.textureCompressionASTC_LDR = is_optimal_astc_supported;
- const std::vector<const char*> extensions = {VK_KHR_SWAPCHAIN_EXTENSION_NAME};
+ const auto queue_cis = GetDeviceQueueCreateInfos();
+ const std::vector<const char*> extensions = LoadExtensions(dldi);
const vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(),
0, nullptr, static_cast<u32>(extensions.size()),
extensions.data(), &device_features);
@@ -90,7 +95,7 @@ vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
LOG_CRITICAL(Render_Vulkan,
"Format={} with usage={} and type={} has no defined alternatives and host "
"hardware does not support it",
- static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
+ vk::to_string(wanted_format), vk::to_string(wanted_usage),
static_cast<u32>(format_type));
UNREACHABLE();
return wanted_format;
@@ -118,6 +123,30 @@ vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
return wanted_format;
}
+bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
+ const vk::DispatchLoaderDynamic& dldi) const {
+ if (!features.textureCompressionASTC_LDR) {
+ return false;
+ }
+ const auto format_feature_usage{
+ vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eBlitSrc |
+ vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc |
+ vk::FormatFeatureFlagBits::eTransferDst};
+ constexpr std::array<vk::Format, 9> astc_formats = {
+ vk::Format::eAstc4x4UnormBlock, vk::Format::eAstc4x4SrgbBlock,
+ vk::Format::eAstc8x8SrgbBlock, vk::Format::eAstc8x6SrgbBlock,
+ vk::Format::eAstc5x4SrgbBlock, vk::Format::eAstc5x5UnormBlock,
+ vk::Format::eAstc5x5SrgbBlock, vk::Format::eAstc10x8UnormBlock,
+ vk::Format::eAstc10x8SrgbBlock};
+ for (const auto format : astc_formats) {
+ const auto format_properties{physical.getFormatProperties(format, dldi)};
+ if (!(format_properties.optimalTilingFeatures & format_feature_usage)) {
+ return false;
+ }
+ }
+ return true;
+}
+
bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
FormatType format_type) const {
const auto it = format_properties.find(wanted_format);
@@ -132,11 +161,9 @@ bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlag
bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
vk::SurfaceKHR surface) {
- const std::string swapchain_extension = VK_KHR_SWAPCHAIN_EXTENSION_NAME;
-
bool has_swapchain{};
for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
- has_swapchain |= prop.extensionName == swapchain_extension;
+ has_swapchain |= prop.extensionName == std::string(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
}
if (!has_swapchain) {
// The device doesn't support creating swapchains.
@@ -160,8 +187,14 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
// TODO(Rodrigo): Check if the device matches all requeriments.
- const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
- if (props.limits.maxUniformBufferRange < 65536) {
+ const auto properties{physical.getProperties(dldi)};
+ const auto limits{properties.limits};
+ if (limits.maxUniformBufferRange < 65536) {
+ return false;
+ }
+
+ const vk::PhysicalDeviceFeatures features{physical.getFeatures(dldi)};
+ if (!features.vertexPipelineStoresAndAtomics || !features.independentBlend) {
return false;
}
@@ -169,6 +202,30 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
return true;
}
+std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynamic& dldi) {
+ std::vector<const char*> extensions;
+ extensions.reserve(2);
+ extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
+
+ const auto Test = [&](const vk::ExtensionProperties& extension,
+ std::optional<std::reference_wrapper<bool>> status, const char* name,
+ u32 revision) {
+ if (extension.extensionName != std::string(name)) {
+ return;
+ }
+ extensions.push_back(name);
+ if (status) {
+ status->get() = true;
+ }
+ };
+
+ for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+ Test(extension, ext_scalar_block_layout, VK_EXT_SCALAR_BLOCK_LAYOUT_EXTENSION_NAME, 1);
+ }
+
+ return extensions;
+}
+
void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) {
std::optional<u32> graphics_family_, present_family_;
@@ -196,10 +253,16 @@ void VKDevice::SetupProperties(const vk::DispatchLoaderDynamic& dldi) {
const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
device_type = props.deviceType;
uniform_buffer_alignment = static_cast<u64>(props.limits.minUniformBufferOffsetAlignment);
+ max_storage_buffer_range = static_cast<u64>(props.limits.maxStorageBufferRange);
+}
+
+void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) {
+ const auto supported_features{physical.getFeatures(dldi)};
+ is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi);
}
std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const {
- static const float QUEUE_PRIORITY = 1.f;
+ static const float QUEUE_PRIORITY = 1.0f;
std::set<u32> unique_queue_families = {graphics_family, present_family};
std::vector<vk::DeviceQueueCreateInfo> queue_cis;
@@ -212,26 +275,43 @@ std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() con
std::map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
+ static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
+ vk::Format::eB5G6R5UnormPack16,
+ vk::Format::eA2B10G10R10UnormPack32,
+ vk::Format::eR32G32B32A32Sfloat,
+ vk::Format::eR16G16Unorm,
+ vk::Format::eR16G16Snorm,
+ vk::Format::eR8G8B8A8Srgb,
+ vk::Format::eR8Unorm,
+ vk::Format::eB10G11R11UfloatPack32,
+ vk::Format::eR32Sfloat,
+ vk::Format::eR16Sfloat,
+ vk::Format::eR16G16B16A16Sfloat,
+ vk::Format::eD32Sfloat,
+ vk::Format::eD16Unorm,
+ vk::Format::eD16UnormS8Uint,
+ vk::Format::eD24UnormS8Uint,
+ vk::Format::eD32SfloatS8Uint,
+ vk::Format::eBc1RgbaUnormBlock,
+ vk::Format::eBc2UnormBlock,
+ vk::Format::eBc3UnormBlock,
+ vk::Format::eBc4UnormBlock,
+ vk::Format::eBc5UnormBlock,
+ vk::Format::eBc5SnormBlock,
+ vk::Format::eBc7UnormBlock,
+ vk::Format::eAstc4x4UnormBlock,
+ vk::Format::eAstc4x4SrgbBlock,
+ vk::Format::eAstc8x8SrgbBlock,
+ vk::Format::eAstc8x6SrgbBlock,
+ vk::Format::eAstc5x4SrgbBlock,
+ vk::Format::eAstc5x5UnormBlock,
+ vk::Format::eAstc5x5SrgbBlock,
+ vk::Format::eAstc10x8UnormBlock,
+ vk::Format::eAstc10x8SrgbBlock};
std::map<vk::Format, vk::FormatProperties> format_properties;
-
- const auto AddFormatQuery = [&format_properties, &dldi, physical](vk::Format format) {
+ for (const auto format : formats) {
format_properties.emplace(format, physical.getFormatProperties(format, dldi));
- };
- AddFormatQuery(vk::Format::eA8B8G8R8UnormPack32);
- AddFormatQuery(vk::Format::eB5G6R5UnormPack16);
- AddFormatQuery(vk::Format::eA2B10G10R10UnormPack32);
- AddFormatQuery(vk::Format::eR8G8B8A8Srgb);
- AddFormatQuery(vk::Format::eR8Unorm);
- AddFormatQuery(vk::Format::eD32Sfloat);
- AddFormatQuery(vk::Format::eD16Unorm);
- AddFormatQuery(vk::Format::eD16UnormS8Uint);
- AddFormatQuery(vk::Format::eD24UnormS8Uint);
- AddFormatQuery(vk::Format::eD32SfloatS8Uint);
- AddFormatQuery(vk::Format::eBc1RgbaUnormBlock);
- AddFormatQuery(vk::Format::eBc2UnormBlock);
- AddFormatQuery(vk::Format::eBc3UnormBlock);
- AddFormatQuery(vk::Format::eBc4UnormBlock);
-
+ }
return format_properties;
}
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index e87c7a508..537825d8b 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -11,7 +11,7 @@
namespace Vulkan {
-/// Format usage descriptor
+/// Format usage descriptor.
enum class FormatType { Linear, Optimal, Buffer };
/// Handles data specific to a physical device.
@@ -34,12 +34,12 @@ public:
vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
FormatType format_type) const;
- /// Returns the dispatch loader with direct function pointers of the device
+ /// Returns the dispatch loader with direct function pointers of the device.
const vk::DispatchLoaderDynamic& GetDispatchLoader() const {
return dld;
}
- /// Returns the logical device
+ /// Returns the logical device.
vk::Device GetLogical() const {
return logical.get();
}
@@ -69,30 +69,55 @@ public:
return present_family;
}
- /// Returns if the device is integrated with the host CPU
+ /// Returns if the device is integrated with the host CPU.
bool IsIntegrated() const {
return device_type == vk::PhysicalDeviceType::eIntegratedGpu;
}
- /// Returns uniform buffer alignment requeriment
+ /// Returns uniform buffer alignment requeriment.
u64 GetUniformBufferAlignment() const {
return uniform_buffer_alignment;
}
+ /// Returns the maximum range for storage buffers.
+ u64 GetMaxStorageBufferRange() const {
+ return max_storage_buffer_range;
+ }
+
+ /// Returns true if ASTC is natively supported.
+ bool IsOptimalAstcSupported() const {
+ return is_optimal_astc_supported;
+ }
+
+ /// Returns true if the device supports VK_EXT_scalar_block_layout.
+ bool IsExtScalarBlockLayoutSupported() const {
+ return ext_scalar_block_layout;
+ }
+
/// Checks if the physical device is suitable.
static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
vk::SurfaceKHR surface);
private:
+ /// Loads extensions into a vector and stores available ones in this object.
+ std::vector<const char*> LoadExtensions(const vk::DispatchLoaderDynamic& dldi);
+
/// Sets up queue families.
void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface);
/// Sets up device properties.
void SetupProperties(const vk::DispatchLoaderDynamic& dldi);
+ /// Sets up device features.
+ void SetupFeatures(const vk::DispatchLoaderDynamic& dldi);
+
/// Returns a list of queue initialization descriptors.
std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
+ /// Returns true if ASTC textures are natively supported.
+ bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
+ const vk::DispatchLoaderDynamic& dldi) const;
+
/// Returns true if a format is supported.
bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
FormatType format_type) const;
@@ -101,16 +126,19 @@ private:
static std::map<vk::Format, vk::FormatProperties> GetFormatProperties(
const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
- const vk::PhysicalDevice physical; ///< Physical device
- vk::DispatchLoaderDynamic dld; ///< Device function pointers
- UniqueDevice logical; ///< Logical device
- vk::Queue graphics_queue; ///< Main graphics queue
- vk::Queue present_queue; ///< Main present queue
- u32 graphics_family{}; ///< Main graphics queue family index
- u32 present_family{}; ///< Main present queue family index
- vk::PhysicalDeviceType device_type; ///< Physical device type
- u64 uniform_buffer_alignment{}; ///< Uniform buffer alignment requeriment
- std::map<vk::Format, vk::FormatProperties> format_properties; ///< Format properties dictionary
+ const vk::PhysicalDevice physical; ///< Physical device.
+ vk::DispatchLoaderDynamic dld; ///< Device function pointers.
+ UniqueDevice logical; ///< Logical device.
+ vk::Queue graphics_queue; ///< Main graphics queue.
+ vk::Queue present_queue; ///< Main present queue.
+ u32 graphics_family{}; ///< Main graphics queue family index.
+ u32 present_family{}; ///< Main present queue family index.
+ vk::PhysicalDeviceType device_type; ///< Physical device type.
+ u64 uniform_buffer_alignment{}; ///< Uniform buffer alignment requeriment.
+ u64 max_storage_buffer_range{}; ///< Max storage buffer size.
+ bool is_optimal_astc_supported{}; ///< Support for native ASTC.
+ bool ext_scalar_block_layout{}; ///< Support for VK_EXT_scalar_block_layout.
+ std::map<vk::Format, vk::FormatProperties> format_properties; ///< Format properties dictionary.
};
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index a5b25aeff..a85fcae5a 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -17,6 +17,7 @@
#include "video_core/engines/maxwell_3d.h"
#include "video_core/engines/shader_bytecode.h"
#include "video_core/engines/shader_header.h"
+#include "video_core/renderer_vulkan/vk_device.h"
#include "video_core/renderer_vulkan/vk_shader_decompiler.h"
#include "video_core/shader/shader_ir.h"
@@ -33,7 +34,8 @@ using ShaderStage = Tegra::Engines::Maxwell3D::Regs::ShaderStage;
using Operation = const OperationNode&;
// TODO(Rodrigo): Use rasterizer's value
-constexpr u32 MAX_CONSTBUFFER_ELEMENTS = 0x1000;
+constexpr u32 MAX_CONSTBUFFER_FLOATS = 0x4000;
+constexpr u32 MAX_CONSTBUFFER_ELEMENTS = MAX_CONSTBUFFER_FLOATS / 4;
constexpr u32 STAGE_BINDING_STRIDE = 0x100;
enum class Type { Bool, Bool2, Float, Int, Uint, HalfFloat };
@@ -87,8 +89,8 @@ bool IsPrecise(Operation operand) {
class SPIRVDecompiler : public Sirit::Module {
public:
- explicit SPIRVDecompiler(const ShaderIR& ir, ShaderStage stage)
- : Module(0x00010300), ir{ir}, stage{stage}, header{ir.GetHeader()} {
+ explicit SPIRVDecompiler(const VKDevice& device, const ShaderIR& ir, ShaderStage stage)
+ : Module(0x00010300), device{device}, ir{ir}, stage{stage}, header{ir.GetHeader()} {
AddCapability(spv::Capability::Shader);
AddExtension("SPV_KHR_storage_buffer_storage_class");
AddExtension("SPV_KHR_variable_pointers");
@@ -195,7 +197,9 @@ public:
entries.samplers.emplace_back(sampler);
}
for (const auto& attribute : ir.GetInputAttributes()) {
- entries.attributes.insert(GetGenericAttributeLocation(attribute));
+ if (IsGenericAttribute(attribute)) {
+ entries.attributes.insert(GetGenericAttributeLocation(attribute));
+ }
}
entries.clip_distances = ir.GetClipDistances();
entries.shader_length = ir.GetLength();
@@ -210,7 +214,6 @@ private:
std::array<OperationDecompilerFn, static_cast<std::size_t>(OperationCode::Amount)>;
static constexpr auto INTERNAL_FLAGS_COUNT = static_cast<std::size_t>(InternalFlag::Amount);
- static constexpr u32 CBUF_STRIDE = 16;
void AllocateBindings() {
const u32 binding_base = static_cast<u32>(stage) * STAGE_BINDING_STRIDE;
@@ -315,6 +318,7 @@ private:
constexpr std::array<const char*, INTERNAL_FLAGS_COUNT> names = {"zero", "sign", "carry",
"overflow"};
for (std::size_t flag = 0; flag < INTERNAL_FLAGS_COUNT; ++flag) {
+ const auto flag_code = static_cast<InternalFlag>(flag);
const Id id = OpVariable(t_prv_bool, spv::StorageClass::Private, v_false);
internal_flags[flag] = AddGlobalVariable(Name(id, names[flag]));
}
@@ -374,7 +378,9 @@ private:
u32 binding = const_buffers_base_binding;
for (const auto& entry : ir.GetConstantBuffers()) {
const auto [index, size] = entry;
- const Id id = OpVariable(t_cbuf_ubo, spv::StorageClass::Uniform);
+ const Id type =
+ device.IsExtScalarBlockLayoutSupported() ? t_cbuf_scalar_ubo : t_cbuf_std140_ubo;
+ const Id id = OpVariable(type, spv::StorageClass::Uniform);
AddGlobalVariable(Name(id, fmt::format("cbuf_{}", index)));
Decorate(id, spv::Decoration::Binding, binding++);
@@ -569,33 +575,35 @@ private:
const Node offset = cbuf->GetOffset();
const Id buffer_id = constant_buffers.at(cbuf->GetIndex());
- Id buffer_index{};
- Id buffer_element{};
-
- if (const auto immediate = std::get_if<ImmediateNode>(offset)) {
- // Direct access
- const u32 offset_imm = immediate->GetValue();
- ASSERT(offset_imm % 4 == 0);
- buffer_index = Constant(t_uint, offset_imm / 16);
- buffer_element = Constant(t_uint, (offset_imm / 4) % 4);
-
- } else if (std::holds_alternative<OperationNode>(*offset)) {
- // Indirect access
- // TODO(Rodrigo): Use a uniform buffer stride of 4 and drop this slow math (which
- // emits sub-optimal code on GLSL from my testing).
- const Id offset_id = BitcastTo<Type::Uint>(Visit(offset));
- const Id unsafe_offset = Emit(OpUDiv(t_uint, offset_id, Constant(t_uint, 4)));
- const Id final_offset = Emit(
- OpUMod(t_uint, unsafe_offset, Constant(t_uint, MAX_CONSTBUFFER_ELEMENTS - 1)));
- buffer_index = Emit(OpUDiv(t_uint, final_offset, Constant(t_uint, 4)));
- buffer_element = Emit(OpUMod(t_uint, final_offset, Constant(t_uint, 4)));
-
+ Id pointer{};
+ if (device.IsExtScalarBlockLayoutSupported()) {
+ const Id buffer_offset = Emit(OpShiftRightLogical(
+ t_uint, BitcastTo<Type::Uint>(Visit(offset)), Constant(t_uint, 2u)));
+ pointer = Emit(
+ OpAccessChain(t_cbuf_float, buffer_id, Constant(t_uint, 0u), buffer_offset));
} else {
- UNREACHABLE_MSG("Unmanaged offset node type");
+ Id buffer_index{};
+ Id buffer_element{};
+ if (const auto immediate = std::get_if<ImmediateNode>(offset)) {
+ // Direct access
+ const u32 offset_imm = immediate->GetValue();
+ ASSERT(offset_imm % 4 == 0);
+ buffer_index = Constant(t_uint, offset_imm / 16);
+ buffer_element = Constant(t_uint, (offset_imm / 4) % 4);
+ } else if (std::holds_alternative<OperationNode>(*offset)) {
+ // Indirect access
+ const Id offset_id = BitcastTo<Type::Uint>(Visit(offset));
+ const Id unsafe_offset = Emit(OpUDiv(t_uint, offset_id, Constant(t_uint, 4)));
+ const Id final_offset = Emit(OpUMod(
+ t_uint, unsafe_offset, Constant(t_uint, MAX_CONSTBUFFER_ELEMENTS - 1)));
+ buffer_index = Emit(OpUDiv(t_uint, final_offset, Constant(t_uint, 4)));
+ buffer_element = Emit(OpUMod(t_uint, final_offset, Constant(t_uint, 4)));
+ } else {
+ UNREACHABLE_MSG("Unmanaged offset node type");
+ }
+ pointer = Emit(OpAccessChain(t_cbuf_float, buffer_id, Constant(t_uint, 0),
+ buffer_index, buffer_element));
}
-
- const Id pointer = Emit(OpAccessChain(t_cbuf_float, buffer_id, Constant(t_uint, 0),
- buffer_index, buffer_element));
return Emit(OpLoad(t_float, pointer));
} else if (const auto gmem = std::get_if<GmemNode>(node)) {
@@ -612,7 +620,9 @@ private:
// It's invalid to call conditional on nested nodes, use an operation instead
const Id true_label = OpLabel();
const Id skip_label = OpLabel();
- Emit(OpBranchConditional(Visit(conditional->GetCondition()), true_label, skip_label));
+ const Id condition = Visit(conditional->GetCondition());
+ Emit(OpSelectionMerge(skip_label, spv::SelectionControlMask::MaskNone));
+ Emit(OpBranchConditional(condition, true_label, skip_label));
Emit(true_label);
VisitBasicBlock(conditional->GetCode());
@@ -968,11 +978,11 @@ private:
case ShaderStage::Vertex: {
// TODO(Rodrigo): We should use VK_EXT_depth_range_unrestricted instead, but it doesn't
// seem to be working on Nvidia's drivers and Intel (mesa and blob) doesn't support it.
- const Id position = AccessElement(t_float4, per_vertex, position_index);
- Id depth = Emit(OpLoad(t_float, AccessElement(t_out_float, position, 2)));
+ const Id z_pointer = AccessElement(t_out_float, per_vertex, position_index, 2u);
+ Id depth = Emit(OpLoad(t_float, z_pointer));
depth = Emit(OpFAdd(t_float, depth, Constant(t_float, 1.0f)));
depth = Emit(OpFMul(t_float, depth, Constant(t_float, 0.5f)));
- Emit(OpStore(AccessElement(t_out_float, position, 2), depth));
+ Emit(OpStore(z_pointer, depth));
break;
}
case ShaderStage::Fragment: {
@@ -1311,6 +1321,7 @@ private:
&SPIRVDecompiler::WorkGroupId<2>,
};
+ const VKDevice& device;
const ShaderIR& ir;
const ShaderStage stage;
const Tegra::Shader::Header header;
@@ -1349,12 +1360,18 @@ private:
const Id t_out_float4 = Name(TypePointer(spv::StorageClass::Output, t_float4), "out_float4");
const Id t_cbuf_float = TypePointer(spv::StorageClass::Uniform, t_float);
- const Id t_cbuf_array =
- Decorate(Name(TypeArray(t_float4, Constant(t_uint, MAX_CONSTBUFFER_ELEMENTS)), "CbufArray"),
- spv::Decoration::ArrayStride, CBUF_STRIDE);
- const Id t_cbuf_struct = MemberDecorate(
- Decorate(TypeStruct(t_cbuf_array), spv::Decoration::Block), 0, spv::Decoration::Offset, 0);
- const Id t_cbuf_ubo = TypePointer(spv::StorageClass::Uniform, t_cbuf_struct);
+ const Id t_cbuf_std140 = Decorate(
+ Name(TypeArray(t_float4, Constant(t_uint, MAX_CONSTBUFFER_ELEMENTS)), "CbufStd140Array"),
+ spv::Decoration::ArrayStride, 16u);
+ const Id t_cbuf_scalar = Decorate(
+ Name(TypeArray(t_float, Constant(t_uint, MAX_CONSTBUFFER_FLOATS)), "CbufScalarArray"),
+ spv::Decoration::ArrayStride, 4u);
+ const Id t_cbuf_std140_struct = MemberDecorate(
+ Decorate(TypeStruct(t_cbuf_std140), spv::Decoration::Block), 0, spv::Decoration::Offset, 0);
+ const Id t_cbuf_scalar_struct = MemberDecorate(
+ Decorate(TypeStruct(t_cbuf_scalar), spv::Decoration::Block), 0, spv::Decoration::Offset, 0);
+ const Id t_cbuf_std140_ubo = TypePointer(spv::StorageClass::Uniform, t_cbuf_std140_struct);
+ const Id t_cbuf_scalar_ubo = TypePointer(spv::StorageClass::Uniform, t_cbuf_scalar_struct);
const Id t_gmem_float = TypePointer(spv::StorageClass::StorageBuffer, t_float);
const Id t_gmem_array =
@@ -1403,8 +1420,9 @@ private:
std::map<u32, Id> labels;
};
-DecompilerResult Decompile(const VideoCommon::Shader::ShaderIR& ir, Maxwell::ShaderStage stage) {
- auto decompiler = std::make_unique<SPIRVDecompiler>(ir, stage);
+DecompilerResult Decompile(const VKDevice& device, const VideoCommon::Shader::ShaderIR& ir,
+ Maxwell::ShaderStage stage) {
+ auto decompiler = std::make_unique<SPIRVDecompiler>(device, ir, stage);
decompiler->Decompile();
return {std::move(decompiler), decompiler->GetShaderEntries()};
}
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.h b/src/video_core/renderer_vulkan/vk_shader_decompiler.h
index 329d8fa38..f90541cc1 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.h
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.h
@@ -20,10 +20,13 @@ namespace VideoCommon::Shader {
class ShaderIR;
}
+namespace Vulkan {
+class VKDevice;
+}
+
namespace Vulkan::VKShader {
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
using SamplerEntry = VideoCommon::Shader::Sampler;
constexpr u32 DESCRIPTOR_SET = 0;
@@ -75,6 +78,7 @@ struct ShaderEntries {
using DecompilerResult = std::pair<std::unique_ptr<Sirit::Module>, ShaderEntries>;
-DecompilerResult Decompile(const VideoCommon::Shader::ShaderIR& ir, Maxwell::ShaderStage stage);
+DecompilerResult Decompile(const VKDevice& device, const VideoCommon::Shader::ShaderIR& ir,
+ Maxwell::ShaderStage stage);
} // namespace Vulkan::VKShader
generated by cgit v1.2.3 (git 2.25.1) at 2025-12-28 02:49:09 +0000