summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/CMakeLists.txt2
-rw-r--r--src/video_core/engines/const_buffer_engine_interface.h14
-rw-r--r--src/video_core/engines/shader_bytecode.h20
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp7
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp87
-rw-r--r--src/video_core/renderer_vulkan/maxwell_to_vk.cpp20
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.cpp31
-rw-r--r--src/video_core/renderer_vulkan/wrapper.cpp750
-rw-r--r--src/video_core/renderer_vulkan/wrapper.h987
-rw-r--r--src/video_core/shader/decode/arithmetic_integer.cpp15
-rw-r--r--src/video_core/shader/decode/conversion.cpp15
-rw-r--r--src/video_core/shader/decode/memory.cpp60
-rw-r--r--src/video_core/shader/decode/xmad.cpp9
-rw-r--r--src/video_core/shader/node.h16
-rw-r--r--src/video_core/shader/node_helper.cpp14
-rw-r--r--src/video_core/shader/shader_ir.cpp85
-rw-r--r--src/video_core/shader/shader_ir.h8
17 files changed, 2028 insertions, 112 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 91df062d7..effe76a63 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -210,6 +210,8 @@ if (ENABLE_VULKAN)
renderer_vulkan/vk_texture_cache.h
renderer_vulkan/vk_update_descriptor.cpp
renderer_vulkan/vk_update_descriptor.h
+ renderer_vulkan/wrapper.cpp
+ renderer_vulkan/wrapper.h
)
target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
diff --git a/src/video_core/engines/const_buffer_engine_interface.h b/src/video_core/engines/const_buffer_engine_interface.h
index 724ee0fd6..ebe139504 100644
--- a/src/video_core/engines/const_buffer_engine_interface.h
+++ b/src/video_core/engines/const_buffer_engine_interface.h
@@ -18,10 +18,14 @@ struct SamplerDescriptor {
union {
u32 raw = 0;
BitField<0, 2, Tegra::Shader::TextureType> texture_type;
- BitField<2, 3, Tegra::Texture::ComponentType> component_type;
+ BitField<2, 3, Tegra::Texture::ComponentType> r_type;
BitField<5, 1, u32> is_array;
BitField<6, 1, u32> is_buffer;
BitField<7, 1, u32> is_shadow;
+ BitField<8, 3, Tegra::Texture::ComponentType> g_type;
+ BitField<11, 3, Tegra::Texture::ComponentType> b_type;
+ BitField<14, 3, Tegra::Texture::ComponentType> a_type;
+ BitField<17, 7, Tegra::Texture::TextureFormat> format;
};
bool operator==(const SamplerDescriptor& rhs) const noexcept {
@@ -36,9 +40,11 @@ struct SamplerDescriptor {
using Tegra::Shader::TextureType;
SamplerDescriptor result;
- // This is going to be used to determine the shading language type.
- // Because of that we don't care about all component types on color textures.
- result.component_type.Assign(tic.r_type.Value());
+ result.format.Assign(tic.format.Value());
+ result.r_type.Assign(tic.r_type.Value());
+ result.g_type.Assign(tic.g_type.Value());
+ result.b_type.Assign(tic.b_type.Value());
+ result.a_type.Assign(tic.a_type.Value());
switch (tic.texture_type.Value()) {
case Tegra::Texture::TextureType::Texture1D:
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index eba42deb4..930b605af 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -82,6 +82,10 @@ union Attribute {
Position = 7,
Attribute_0 = 8,
Attribute_31 = 39,
+ FrontColor = 40,
+ FrontSecondaryColor = 41,
+ BackColor = 42,
+ BackSecondaryColor = 43,
ClipDistances0123 = 44,
ClipDistances4567 = 45,
PointCoord = 46,
@@ -89,6 +93,8 @@ union Attribute {
// shader, and a tuple of (TessCoord.x, TessCoord.y, TessCoord.z, ~) when inside a Tess Eval
// shader.
TessCoordInstanceIDVertexID = 47,
+ TexCoord_0 = 48,
+ TexCoord_7 = 55,
// This attribute contains a tuple of (Unk, Unk, Unk, gl_FrontFacing) when inside a fragment
// shader. It is unknown what the other values contain.
FrontFacing = 63,
@@ -225,18 +231,6 @@ enum class AtomicOp : u64 {
Or = 6,
Xor = 7,
Exch = 8,
-};
-
-enum class GlobalAtomicOp : u64 {
- Add = 0,
- Min = 1,
- Max = 2,
- Inc = 3,
- Dec = 4,
- And = 5,
- Or = 6,
- Xor = 7,
- Exch = 8,
SafeAdd = 10,
};
@@ -995,7 +989,7 @@ union Instruction {
} stg;
union {
- BitField<52, 4, GlobalAtomicOp> operation;
+ BitField<52, 4, AtomicOp> operation;
BitField<49, 3, GlobalAtomicType> type;
BitField<28, 20, s64> offset;
} atom;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 826eee7df..31add708f 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -444,6 +444,7 @@ void RasterizerOpenGL::Clear() {
}
SyncRasterizeEnable();
+ SyncStencilTestState();
if (regs.clear_flags.scissor) {
SyncScissorTest();
@@ -1052,12 +1053,8 @@ void RasterizerOpenGL::SyncStencilTestState() {
flags[Dirty::StencilTest] = false;
const auto& regs = gpu.regs;
- if (!regs.stencil_enable) {
- glDisable(GL_STENCIL_TEST);
- return;
- }
+ oglEnable(GL_STENCIL_TEST, regs.stencil_enable);
- glEnable(GL_STENCIL_TEST);
glStencilFuncSeparate(GL_FRONT, MaxwellToGL::ComparisonOp(regs.stencil_front_func_func),
regs.stencil_front_func_ref, regs.stencil_front_func_mask);
glStencilOpSeparate(GL_FRONT, MaxwellToGL::StencilOp(regs.stencil_front_op_fail),
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 2c38f57fd..c7d24cf14 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -366,10 +366,19 @@ constexpr bool IsGenericAttribute(Attribute::Index index) {
return index >= Attribute::Index::Attribute_0 && index <= Attribute::Index::Attribute_31;
}
+constexpr bool IsLegacyTexCoord(Attribute::Index index) {
+ return static_cast<int>(index) >= static_cast<int>(Attribute::Index::TexCoord_0) &&
+ static_cast<int>(index) <= static_cast<int>(Attribute::Index::TexCoord_7);
+}
+
constexpr Attribute::Index ToGenericAttribute(u64 value) {
return static_cast<Attribute::Index>(value + static_cast<u64>(Attribute::Index::Attribute_0));
}
+constexpr int GetLegacyTexCoordIndex(Attribute::Index index) {
+ return static_cast<int>(index) - static_cast<int>(Attribute::Index::TexCoord_0);
+}
+
u32 GetGenericAttributeIndex(Attribute::Index index) {
ASSERT(IsGenericAttribute(index));
return static_cast<u32>(index) - static_cast<u32>(Attribute::Index::Attribute_0);
@@ -498,7 +507,7 @@ private:
if (!identifier.empty()) {
code.AddLine("// {}", identifier);
}
- code.AddLine("#version 440 core");
+ code.AddLine("#version 440 {}", ir.UsesLegacyVaryings() ? "compatibility" : "core");
code.AddLine("#extension GL_ARB_separate_shader_objects : enable");
if (device.HasShaderBallot()) {
code.AddLine("#extension GL_ARB_shader_ballot : require");
@@ -561,6 +570,16 @@ private:
if (stage != ShaderType::Fragment) {
return;
}
+ if (ir.UsesLegacyVaryings()) {
+ code.AddLine("in gl_PerFragment {{");
+ ++code.scope;
+ code.AddLine("vec4 gl_TexCoord[8];");
+ code.AddLine("vec4 gl_Color;");
+ code.AddLine("vec4 gl_SecondaryColor;");
+ --code.scope;
+ code.AddLine("}};");
+ }
+
for (u32 rt = 0; rt < Maxwell::NumRenderTargets; ++rt) {
code.AddLine("layout (location = {}) out vec4 frag_color{};", rt, rt);
}
@@ -617,12 +636,12 @@ private:
code.AddLine("float gl_PointSize;");
}
- if (ir.UsesInstanceId()) {
- code.AddLine("int gl_InstanceID;");
- }
-
- if (ir.UsesVertexId()) {
- code.AddLine("int gl_VertexID;");
+ if (ir.UsesLegacyVaryings()) {
+ code.AddLine("vec4 gl_TexCoord[8];");
+ code.AddLine("vec4 gl_FrontColor;");
+ code.AddLine("vec4 gl_FrontSecondaryColor;");
+ code.AddLine("vec4 gl_BackColor;");
+ code.AddLine("vec4 gl_BackSecondaryColor;");
}
--code.scope;
@@ -1128,6 +1147,10 @@ private:
default:
UNREACHABLE();
}
+ case Attribute::Index::FrontColor:
+ return {"gl_Color"s + GetSwizzle(element), Type::Float};
+ case Attribute::Index::FrontSecondaryColor:
+ return {"gl_SecondaryColor"s + GetSwizzle(element), Type::Float};
case Attribute::Index::PointCoord:
switch (element) {
case 0:
@@ -1168,6 +1191,12 @@ private:
return {GeometryPass(GetGenericInputAttribute(attribute)) + GetSwizzle(element),
Type::Float};
}
+ if (IsLegacyTexCoord(attribute)) {
+ UNIMPLEMENTED_IF(stage == ShaderType::Geometry);
+ return {fmt::format("gl_TexCoord[{}]{}", GetLegacyTexCoordIndex(attribute),
+ GetSwizzle(element)),
+ Type::Float};
+ }
break;
}
UNIMPLEMENTED_MSG("Unhandled input attribute: {}", static_cast<u32>(attribute));
@@ -1206,11 +1235,12 @@ private:
}
std::optional<Expression> GetOutputAttribute(const AbufNode* abuf) {
+ const u32 element = abuf->GetElement();
switch (const auto attribute = abuf->GetIndex()) {
case Attribute::Index::Position:
- return {{"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float}};
+ return {{"gl_Position"s + GetSwizzle(element), Type::Float}};
case Attribute::Index::LayerViewportPointSize:
- switch (abuf->GetElement()) {
+ switch (element) {
case 0:
UNIMPLEMENTED();
return {};
@@ -1228,13 +1258,26 @@ private:
return {{"gl_PointSize", Type::Float}};
}
return {};
+ case Attribute::Index::FrontColor:
+ return {{"gl_FrontColor"s + GetSwizzle(element), Type::Float}};
+ case Attribute::Index::FrontSecondaryColor:
+ return {{"gl_FrontSecondaryColor"s + GetSwizzle(element), Type::Float}};
+ case Attribute::Index::BackColor:
+ return {{"gl_BackColor"s + GetSwizzle(element), Type::Float}};
+ case Attribute::Index::BackSecondaryColor:
+ return {{"gl_BackSecondaryColor"s + GetSwizzle(element), Type::Float}};
case Attribute::Index::ClipDistances0123:
- return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float}};
+ return {{fmt::format("gl_ClipDistance[{}]", element), Type::Float}};
case Attribute::Index::ClipDistances4567:
- return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float}};
+ return {{fmt::format("gl_ClipDistance[{}]", element + 4), Type::Float}};
default:
if (IsGenericAttribute(attribute)) {
- return {{GetGenericOutputAttribute(attribute, abuf->GetElement()), Type::Float}};
+ return {{GetGenericOutputAttribute(attribute, element), Type::Float}};
+ }
+ if (IsLegacyTexCoord(attribute)) {
+ return {{fmt::format("gl_TexCoord[{}]{}", GetLegacyTexCoordIndex(attribute),
+ GetSwizzle(element)),
+ Type::Float}};
}
UNIMPLEMENTED_MSG("Unhandled output attribute: {}", static_cast<u32>(attribute));
return {};
@@ -2071,6 +2114,10 @@ private:
template <const std::string_view& opname, Type type>
Expression Atomic(Operation operation) {
+ if ((opname == Func::Min || opname == Func::Max) && type == Type::Int) {
+ UNIMPLEMENTED_MSG("Unimplemented Min & Max for atomic operations");
+ return {};
+ }
return {fmt::format("atomic{}({}, {})", opname, Visit(operation[0]).GetCode(),
Visit(operation[1]).As(type)),
type};
@@ -2264,6 +2311,8 @@ private:
~Func() = delete;
static constexpr std::string_view Add = "Add";
+ static constexpr std::string_view Min = "Min";
+ static constexpr std::string_view Max = "Max";
static constexpr std::string_view And = "And";
static constexpr std::string_view Or = "Or";
static constexpr std::string_view Xor = "Xor";
@@ -2414,7 +2463,21 @@ private:
&GLSLDecompiler::AtomicImage<Func::Xor>,
&GLSLDecompiler::AtomicImage<Func::Exchange>,
+ &GLSLDecompiler::Atomic<Func::Exchange, Type::Uint>,
&GLSLDecompiler::Atomic<Func::Add, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Min, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Max, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::And, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Or, Type::Uint>,
+ &GLSLDecompiler::Atomic<Func::Xor, Type::Uint>,
+
+ &GLSLDecompiler::Atomic<Func::Exchange, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Add, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Min, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Max, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::And, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Or, Type::Int>,
+ &GLSLDecompiler::Atomic<Func::Xor, Type::Int>,
&GLSLDecompiler::Branch,
&GLSLDecompiler::BranchIndirect,
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index f93447610..7480cb7c3 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -401,6 +401,26 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
}
break;
case Maxwell::VertexAttribute::Type::SignedScaled:
+ switch (size) {
+ case Maxwell::VertexAttribute::Size::Size_8:
+ return vk::Format::eR8Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_8_8:
+ return vk::Format::eR8G8Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_8_8_8:
+ return vk::Format::eR8G8B8Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+ return vk::Format::eR8G8B8A8Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_16:
+ return vk::Format::eR16Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_16_16:
+ return vk::Format::eR16G16Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_16_16_16:
+ return vk::Format::eR16G16B16Sscaled;
+ case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+ return vk::Format::eR16G16B16A16Sscaled;
+ default:
+ break;
+ }
break;
case Maxwell::VertexAttribute::Type::Float:
switch (size) {
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 51ecb5567..d67f08cf9 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -1941,7 +1941,11 @@ private:
return {};
}
- Expression AtomicAdd(Operation operation) {
+ template <Id (Module::*func)(Id, Id, Id, Id, Id), Type result_type,
+ Type value_type = result_type>
+ Expression Atomic(Operation operation) {
+ const Id type_def = GetTypeDefinition(result_type);
+
Id pointer;
if (const auto smem = std::get_if<SmemNode>(&*operation[0])) {
pointer = GetSharedMemoryPointer(*smem);
@@ -1949,14 +1953,15 @@ private:
pointer = GetGlobalMemoryPointer(*gmem);
} else {
UNREACHABLE();
- return {Constant(t_uint, 0), Type::Uint};
+ return {Constant(type_def, 0), result_type};
}
+ const Id value = As(Visit(operation[1]), value_type);
+
const Id scope = Constant(t_uint, static_cast<u32>(spv::Scope::Device));
- const Id semantics = Constant(t_uint, 0U);
+ const Id semantics = Constant(type_def, 0);
- const Id value = AsUint(Visit(operation[1]));
- return {OpAtomicIAdd(t_uint, pointer, scope, semantics, value), Type::Uint};
+ return {(this->*func)(type_def, pointer, scope, semantics, value), result_type};
}
Expression Branch(Operation operation) {
@@ -2545,7 +2550,21 @@ private:
&SPIRVDecompiler::AtomicImageXor,
&SPIRVDecompiler::AtomicImageExchange,
- &SPIRVDecompiler::AtomicAdd,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicExchange, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicIAdd, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicUMin, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicUMax, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicAnd, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicOr, Type::Uint>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicXor, Type::Uint>,
+
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicExchange, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicIAdd, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicSMin, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicSMax, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicAnd, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicOr, Type::Int>,
+ &SPIRVDecompiler::Atomic<&Module::OpAtomicXor, Type::Int>,
&SPIRVDecompiler::Branch,
&SPIRVDecompiler::BranchIndirect,
diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp
new file mode 100644
index 000000000..9b94dfff1
--- /dev/null
+++ b/src/video_core/renderer_vulkan/wrapper.cpp
@@ -0,0 +1,750 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <exception>
+#include <memory>
+#include <optional>
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Vulkan::vk {
+
+namespace {
+
+template <typename T>
+bool Proc(T& result, const InstanceDispatch& dld, const char* proc_name,
+ VkInstance instance = nullptr) noexcept {
+ result = reinterpret_cast<T>(dld.vkGetInstanceProcAddr(instance, proc_name));
+ return result != nullptr;
+}
+
+template <typename T>
+void Proc(T& result, const DeviceDispatch& dld, const char* proc_name, VkDevice device) noexcept {
+ result = reinterpret_cast<T>(dld.vkGetDeviceProcAddr(device, proc_name));
+}
+
+void Load(VkDevice device, DeviceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name, device)
+ X(vkAcquireNextImageKHR);
+ X(vkAllocateCommandBuffers);
+ X(vkAllocateDescriptorSets);
+ X(vkAllocateMemory);
+ X(vkBeginCommandBuffer);
+ X(vkBindBufferMemory);
+ X(vkBindImageMemory);
+ X(vkCmdBeginQuery);
+ X(vkCmdBeginRenderPass);
+ X(vkCmdBeginTransformFeedbackEXT);
+ X(vkCmdBindDescriptorSets);
+ X(vkCmdBindIndexBuffer);
+ X(vkCmdBindPipeline);
+ X(vkCmdBindTransformFeedbackBuffersEXT);
+ X(vkCmdBindVertexBuffers);
+ X(vkCmdBlitImage);
+ X(vkCmdClearAttachments);
+ X(vkCmdCopyBuffer);
+ X(vkCmdCopyBufferToImage);
+ X(vkCmdCopyImage);
+ X(vkCmdCopyImageToBuffer);
+ X(vkCmdDispatch);
+ X(vkCmdDraw);
+ X(vkCmdDrawIndexed);
+ X(vkCmdEndQuery);
+ X(vkCmdEndRenderPass);
+ X(vkCmdEndTransformFeedbackEXT);
+ X(vkCmdFillBuffer);
+ X(vkCmdPipelineBarrier);
+ X(vkCmdPushConstants);
+ X(vkCmdSetBlendConstants);
+ X(vkCmdSetCheckpointNV);
+ X(vkCmdSetDepthBias);
+ X(vkCmdSetDepthBounds);
+ X(vkCmdSetScissor);
+ X(vkCmdSetStencilCompareMask);
+ X(vkCmdSetStencilReference);
+ X(vkCmdSetStencilWriteMask);
+ X(vkCmdSetViewport);
+ X(vkCreateBuffer);
+ X(vkCreateBufferView);
+ X(vkCreateCommandPool);
+ X(vkCreateComputePipelines);
+ X(vkCreateDescriptorPool);
+ X(vkCreateDescriptorSetLayout);
+ X(vkCreateDescriptorUpdateTemplateKHR);
+ X(vkCreateFence);
+ X(vkCreateFramebuffer);
+ X(vkCreateGraphicsPipelines);
+ X(vkCreateImage);
+ X(vkCreateImageView);
+ X(vkCreatePipelineLayout);
+ X(vkCreateQueryPool);
+ X(vkCreateRenderPass);
+ X(vkCreateSampler);
+ X(vkCreateSemaphore);
+ X(vkCreateShaderModule);
+ X(vkCreateSwapchainKHR);
+ X(vkDestroyBuffer);
+ X(vkDestroyBufferView);
+ X(vkDestroyCommandPool);
+ X(vkDestroyDescriptorPool);
+ X(vkDestroyDescriptorSetLayout);
+ X(vkDestroyDescriptorUpdateTemplateKHR);
+ X(vkDestroyFence);
+ X(vkDestroyFramebuffer);
+ X(vkDestroyImage);
+ X(vkDestroyImageView);
+ X(vkDestroyPipeline);
+ X(vkDestroyPipelineLayout);
+ X(vkDestroyQueryPool);
+ X(vkDestroyRenderPass);
+ X(vkDestroySampler);
+ X(vkDestroySemaphore);
+ X(vkDestroyShaderModule);
+ X(vkDestroySwapchainKHR);
+ X(vkDeviceWaitIdle);
+ X(vkEndCommandBuffer);
+ X(vkFreeCommandBuffers);
+ X(vkFreeDescriptorSets);
+ X(vkFreeMemory);
+ X(vkGetBufferMemoryRequirements);
+ X(vkGetDeviceQueue);
+ X(vkGetFenceStatus);
+ X(vkGetImageMemoryRequirements);
+ X(vkGetQueryPoolResults);
+ X(vkGetQueueCheckpointDataNV);
+ X(vkMapMemory);
+ X(vkQueueSubmit);
+ X(vkResetFences);
+ X(vkResetQueryPoolEXT);
+ X(vkUnmapMemory);
+ X(vkUpdateDescriptorSetWithTemplateKHR);
+ X(vkUpdateDescriptorSets);
+ X(vkWaitForFences);
+#undef X
+}
+
+} // Anonymous namespace
+
+bool Load(InstanceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name)
+ return X(vkCreateInstance) && X(vkEnumerateInstanceExtensionProperties);
+#undef X
+}
+
+bool Load(VkInstance instance, InstanceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name, instance)
+ // These functions may fail to load depending on the enabled extensions.
+ // Don't return a failure on these.
+ X(vkCreateDebugUtilsMessengerEXT);
+ X(vkDestroyDebugUtilsMessengerEXT);
+ X(vkDestroySurfaceKHR);
+ X(vkGetPhysicalDeviceFeatures2KHR);
+ X(vkGetPhysicalDeviceProperties2KHR);
+ X(vkGetPhysicalDeviceSurfaceCapabilitiesKHR);
+ X(vkGetPhysicalDeviceSurfaceFormatsKHR);
+ X(vkGetPhysicalDeviceSurfacePresentModesKHR);
+ X(vkGetPhysicalDeviceSurfaceSupportKHR);
+ X(vkGetSwapchainImagesKHR);
+ X(vkQueuePresentKHR);
+
+ return X(vkCreateDevice) && X(vkDestroyDevice) && X(vkDestroyDevice) &&
+ X(vkEnumerateDeviceExtensionProperties) && X(vkEnumeratePhysicalDevices) &&
+ X(vkGetDeviceProcAddr) && X(vkGetPhysicalDeviceFormatProperties) &&
+ X(vkGetPhysicalDeviceMemoryProperties) && X(vkGetPhysicalDeviceProperties) &&
+ X(vkGetPhysicalDeviceQueueFamilyProperties);
+#undef X
+}
+
+const char* Exception::what() const noexcept {
+ return ToString(result);
+}
+
+const char* ToString(VkResult result) noexcept {
+ switch (result) {
+ case VkResult::VK_SUCCESS:
+ return "VK_SUCCESS";
+ case VkResult::VK_NOT_READY:
+ return "VK_NOT_READY";
+ case VkResult::VK_TIMEOUT:
+ return "VK_TIMEOUT";
+ case VkResult::VK_EVENT_SET:
+ return "VK_EVENT_SET";
+ case VkResult::VK_EVENT_RESET:
+ return "VK_EVENT_RESET";
+ case VkResult::VK_INCOMPLETE:
+ return "VK_INCOMPLETE";
+ case VkResult::VK_ERROR_OUT_OF_HOST_MEMORY:
+ return "VK_ERROR_OUT_OF_HOST_MEMORY";
+ case VkResult::VK_ERROR_OUT_OF_DEVICE_MEMORY:
+ return "VK_ERROR_OUT_OF_DEVICE_MEMORY";
+ case VkResult::VK_ERROR_INITIALIZATION_FAILED:
+ return "VK_ERROR_INITIALIZATION_FAILED";
+ case VkResult::VK_ERROR_DEVICE_LOST:
+ return "VK_ERROR_DEVICE_LOST";
+ case VkResult::VK_ERROR_MEMORY_MAP_FAILED:
+ return "VK_ERROR_MEMORY_MAP_FAILED";
+ case VkResult::VK_ERROR_LAYER_NOT_PRESENT:
+ return "VK_ERROR_LAYER_NOT_PRESENT";
+ case VkResult::VK_ERROR_EXTENSION_NOT_PRESENT:
+ return "VK_ERROR_EXTENSION_NOT_PRESENT";
+ case VkResult::VK_ERROR_FEATURE_NOT_PRESENT:
+ return "VK_ERROR_FEATURE_NOT_PRESENT";
+ case VkResult::VK_ERROR_INCOMPATIBLE_DRIVER:
+ return "VK_ERROR_INCOMPATIBLE_DRIVER";
+ case VkResult::VK_ERROR_TOO_MANY_OBJECTS:
+ return "VK_ERROR_TOO_MANY_OBJECTS";
+ case VkResult::VK_ERROR_FORMAT_NOT_SUPPORTED:
+ return "VK_ERROR_FORMAT_NOT_SUPPORTED";
+ case VkResult::VK_ERROR_FRAGMENTED_POOL:
+ return "VK_ERROR_FRAGMENTED_POOL";
+ case VkResult::VK_ERROR_OUT_OF_POOL_MEMORY:
+ return "VK_ERROR_OUT_OF_POOL_MEMORY";
+ case VkResult::VK_ERROR_INVALID_EXTERNAL_HANDLE:
+ return "VK_ERROR_INVALID_EXTERNAL_HANDLE";
+ case VkResult::VK_ERROR_SURFACE_LOST_KHR:
+ return "VK_ERROR_SURFACE_LOST_KHR";
+ case VkResult::VK_ERROR_NATIVE_WINDOW_IN_USE_KHR:
+ return "VK_ERROR_NATIVE_WINDOW_IN_USE_KHR";
+ case VkResult::VK_SUBOPTIMAL_KHR:
+ return "VK_SUBOPTIMAL_KHR";
+ case VkResult::VK_ERROR_OUT_OF_DATE_KHR:
+ return "VK_ERROR_OUT_OF_DATE_KHR";
+ case VkResult::VK_ERROR_INCOMPATIBLE_DISPLAY_KHR:
+ return "VK_ERROR_INCOMPATIBLE_DISPLAY_KHR";
+ case VkResult::VK_ERROR_VALIDATION_FAILED_EXT:
+ return "VK_ERROR_VALIDATION_FAILED_EXT";
+ case VkResult::VK_ERROR_INVALID_SHADER_NV:
+ return "VK_ERROR_INVALID_SHADER_NV";
+ case VkResult::VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT:
+ return "VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT";
+ case VkResult::VK_ERROR_FRAGMENTATION_EXT:
+ return "VK_ERROR_FRAGMENTATION_EXT";
+ case VkResult::VK_ERROR_NOT_PERMITTED_EXT:
+ return "VK_ERROR_NOT_PERMITTED_EXT";
+ case VkResult::VK_ERROR_INVALID_DEVICE_ADDRESS_EXT:
+ return "VK_ERROR_INVALID_DEVICE_ADDRESS_EXT";
+ case VkResult::VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT:
+ return "VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT";
+ }
+ return "Unknown";
+}
+
+void Destroy(VkInstance instance, const InstanceDispatch& dld) noexcept {
+ dld.vkDestroyInstance(instance, nullptr);
+}
+
+void Destroy(VkDevice device, const InstanceDispatch& dld) noexcept {
+ dld.vkDestroyDevice(device, nullptr);
+}
+
+void Destroy(VkDevice device, VkBuffer handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyBuffer(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkBufferView handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyBufferView(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkCommandPool handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyCommandPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorPool handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyDescriptorPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorSetLayout handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyDescriptorSetLayout(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorUpdateTemplateKHR handle,
+ const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyDescriptorUpdateTemplateKHR(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDeviceMemory handle, const DeviceDispatch& dld) noexcept {
+ dld.vkFreeMemory(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkFence handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyFence(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkFramebuffer handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyFramebuffer(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkImage handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyImage(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkImageView handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyImageView(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkPipeline handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyPipeline(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkPipelineLayout handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyPipelineLayout(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkQueryPool handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyQueryPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkRenderPass handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyRenderPass(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSampler handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroySampler(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSwapchainKHR handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroySwapchainKHR(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSemaphore handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroySemaphore(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkShaderModule handle, const DeviceDispatch& dld) noexcept {
+ dld.vkDestroyShaderModule(device, handle, nullptr);
+}
+
+void Destroy(VkInstance instance, VkDebugUtilsMessengerEXT handle,
+ const InstanceDispatch& dld) noexcept {
+ dld.vkDestroyDebugUtilsMessengerEXT(instance, handle, nullptr);
+}
+
+void Destroy(VkInstance instance, VkSurfaceKHR handle, const InstanceDispatch& dld) noexcept {
+ dld.vkDestroySurfaceKHR(instance, handle, nullptr);
+}
+
+VkResult Free(VkDevice device, VkDescriptorPool handle, Span<VkDescriptorSet> sets,
+ const DeviceDispatch& dld) noexcept {
+ return dld.vkFreeDescriptorSets(device, handle, sets.size(), sets.data());
+}
+
+VkResult Free(VkDevice device, VkCommandPool handle, Span<VkCommandBuffer> buffers,
+ const DeviceDispatch& dld) noexcept {
+ dld.vkFreeCommandBuffers(device, handle, buffers.size(), buffers.data());
+ return VK_SUCCESS;
+}
+
+Instance Instance::Create(Span<const char*> layers, Span<const char*> extensions,
+ InstanceDispatch& dld) noexcept {
+ VkApplicationInfo application_info;
+ application_info.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
+ application_info.pNext = nullptr;
+ application_info.pApplicationName = "yuzu Emulator";
+ application_info.applicationVersion = VK_MAKE_VERSION(0, 1, 0);
+ application_info.pEngineName = "yuzu Emulator";
+ application_info.engineVersion = VK_MAKE_VERSION(0, 1, 0);
+ application_info.apiVersion = VK_API_VERSION_1_1;
+
+ VkInstanceCreateInfo ci;
+ ci.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO;
+ ci.pNext = nullptr;
+ ci.flags = 0;
+ ci.pApplicationInfo = &application_info;
+ ci.enabledLayerCount = layers.size();
+ ci.ppEnabledLayerNames = layers.data();
+ ci.enabledExtensionCount = extensions.size();
+ ci.ppEnabledExtensionNames = extensions.data();
+
+ VkInstance instance;
+ if (dld.vkCreateInstance(&ci, nullptr, &instance) != VK_SUCCESS) {
+ // Failed to create the instance.
+ return {};
+ }
+ if (!Proc(dld.vkDestroyInstance, dld, "vkDestroyInstance", instance)) {
+ // We successfully created an instance but the destroy function couldn't be loaded.
+ // This is a good moment to panic.
+ return {};
+ }
+
+ return Instance(instance, dld);
+}
+
+std::optional<std::vector<VkPhysicalDevice>> Instance::EnumeratePhysicalDevices() {
+ u32 num;
+ if (dld->vkEnumeratePhysicalDevices(handle, &num, nullptr) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ std::vector<VkPhysicalDevice> physical_devices(num);
+ if (dld->vkEnumeratePhysicalDevices(handle, &num, physical_devices.data()) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ return physical_devices;
+}
+
+DebugCallback Instance::TryCreateDebugCallback(
+ PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept {
+ VkDebugUtilsMessengerCreateInfoEXT ci;
+ ci.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT;
+ ci.pNext = nullptr;
+ ci.flags = 0;
+ ci.messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT;
+ ci.messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT |
+ VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT;
+ ci.pfnUserCallback = callback;
+ ci.pUserData = nullptr;
+
+ VkDebugUtilsMessengerEXT messenger;
+ if (dld->vkCreateDebugUtilsMessengerEXT(handle, &ci, nullptr, &messenger) != VK_SUCCESS) {
+ return {};
+ }
+ return DebugCallback(messenger, handle, *dld);
+}
+
+std::vector<VkCheckpointDataNV> Queue::GetCheckpointDataNV(const DeviceDispatch& dld) const {
+ if (!dld.vkGetQueueCheckpointDataNV) {
+ return {};
+ }
+ u32 num;
+ dld.vkGetQueueCheckpointDataNV(queue, &num, nullptr);
+ std::vector<VkCheckpointDataNV> checkpoints(num);
+ dld.vkGetQueueCheckpointDataNV(queue, &num, checkpoints.data());
+ return checkpoints;
+}
+
+void Buffer::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
+ Check(dld->vkBindBufferMemory(owner, handle, memory, offset));
+}
+
+void Image::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
+ Check(dld->vkBindImageMemory(owner, handle, memory, offset));
+}
+
+DescriptorSets DescriptorPool::Allocate(const VkDescriptorSetAllocateInfo& ai) const {
+ const std::size_t num = ai.descriptorSetCount;
+ std::unique_ptr sets = std::make_unique<VkDescriptorSet[]>(num);
+ switch (const VkResult result = dld->vkAllocateDescriptorSets(owner, &ai, sets.get())) {
+ case VK_SUCCESS:
+ return DescriptorSets(std::move(sets), num, owner, handle, *dld);
+ case VK_ERROR_OUT_OF_POOL_MEMORY:
+ return {};
+ default:
+ throw Exception(result);
+ }
+}
+
+CommandBuffers CommandPool::Allocate(std::size_t num_buffers, VkCommandBufferLevel level) const {
+ VkCommandBufferAllocateInfo ai;
+ ai.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
+ ai.pNext = nullptr;
+ ai.commandPool = handle;
+ ai.level = level;
+ ai.commandBufferCount = static_cast<u32>(num_buffers);
+
+ std::unique_ptr buffers = std::make_unique<VkCommandBuffer[]>(num_buffers);
+ switch (const VkResult result = dld->vkAllocateCommandBuffers(owner, &ai, buffers.get())) {
+ case VK_SUCCESS:
+ return CommandBuffers(std::move(buffers), num_buffers, owner, handle, *dld);
+ case VK_ERROR_OUT_OF_POOL_MEMORY:
+ return {};
+ default:
+ throw Exception(result);
+ }
+}
+
+std::vector<VkImage> SwapchainKHR::GetImages() const {
+ u32 num;
+ Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, nullptr));
+ std::vector<VkImage> images(num);
+ Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, images.data()));
+ return images;
+}
+
+Device Device::Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
+ Span<const char*> enabled_extensions,
+ const VkPhysicalDeviceFeatures2& enabled_features,
+ DeviceDispatch& dld) noexcept {
+ VkDeviceCreateInfo ci;
+ ci.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO;
+ ci.pNext = &enabled_features;
+ ci.flags = 0;
+ ci.queueCreateInfoCount = queues_ci.size();
+ ci.pQueueCreateInfos = queues_ci.data();
+ ci.enabledLayerCount = 0;
+ ci.ppEnabledLayerNames = nullptr;
+ ci.enabledExtensionCount = enabled_extensions.size();
+ ci.ppEnabledExtensionNames = enabled_extensions.data();
+ ci.pEnabledFeatures = nullptr;
+
+ VkDevice device;
+ if (dld.vkCreateDevice(physical_device, &ci, nullptr, &device) != VK_SUCCESS) {
+ return {};
+ }
+ Load(device, dld);
+ return Device(device, dld);
+}
+
+Queue Device::GetQueue(u32 family_index) const noexcept {
+ VkQueue queue;
+ dld->vkGetDeviceQueue(handle, family_index, 0, &queue);
+ return Queue(queue, *dld);
+}
+
+Buffer Device::CreateBuffer(const VkBufferCreateInfo& ci) const {
+ VkBuffer object;
+ Check(dld->vkCreateBuffer(handle, &ci, nullptr, &object));
+ return Buffer(object, handle, *dld);
+}
+
+BufferView Device::CreateBufferView(const VkBufferViewCreateInfo& ci) const {
+ VkBufferView object;
+ Check(dld->vkCreateBufferView(handle, &ci, nullptr, &object));
+ return BufferView(object, handle, *dld);
+}
+
+Image Device::CreateImage(const VkImageCreateInfo& ci) const {
+ VkImage object;
+ Check(dld->vkCreateImage(handle, &ci, nullptr, &object));
+ return Image(object, handle, *dld);
+}
+
+ImageView Device::CreateImageView(const VkImageViewCreateInfo& ci) const {
+ VkImageView object;
+ Check(dld->vkCreateImageView(handle, &ci, nullptr, &object));
+ return ImageView(object, handle, *dld);
+}
+
+Semaphore Device::CreateSemaphore() const {
+ VkSemaphoreCreateInfo ci;
+ ci.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
+ ci.pNext = nullptr;
+ ci.flags = 0;
+
+ VkSemaphore object;
+ Check(dld->vkCreateSemaphore(handle, &ci, nullptr, &object));
+ return Semaphore(object, handle, *dld);
+}
+
+Fence Device::CreateFence(const VkFenceCreateInfo& ci) const {
+ VkFence object;
+ Check(dld->vkCreateFence(handle, &ci, nullptr, &object));
+ return Fence(object, handle, *dld);
+}
+
+DescriptorPool Device::CreateDescriptorPool(const VkDescriptorPoolCreateInfo& ci) const {
+ VkDescriptorPool object;
+ Check(dld->vkCreateDescriptorPool(handle, &ci, nullptr, &object));
+ return DescriptorPool(object, handle, *dld);
+}
+
+RenderPass Device::CreateRenderPass(const VkRenderPassCreateInfo& ci) const {
+ VkRenderPass object;
+ Check(dld->vkCreateRenderPass(handle, &ci, nullptr, &object));
+ return RenderPass(object, handle, *dld);
+}
+
+DescriptorSetLayout Device::CreateDescriptorSetLayout(
+ const VkDescriptorSetLayoutCreateInfo& ci) const {
+ VkDescriptorSetLayout object;
+ Check(dld->vkCreateDescriptorSetLayout(handle, &ci, nullptr, &object));
+ return DescriptorSetLayout(object, handle, *dld);
+}
+
+PipelineLayout Device::CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const {
+ VkPipelineLayout object;
+ Check(dld->vkCreatePipelineLayout(handle, &ci, nullptr, &object));
+ return PipelineLayout(object, handle, *dld);
+}
+
+Pipeline Device::CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const {
+ VkPipeline object;
+ Check(dld->vkCreateGraphicsPipelines(handle, nullptr, 1, &ci, nullptr, &object));
+ return Pipeline(object, handle, *dld);
+}
+
+Pipeline Device::CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const {
+ VkPipeline object;
+ Check(dld->vkCreateComputePipelines(handle, nullptr, 1, &ci, nullptr, &object));
+ return Pipeline(object, handle, *dld);
+}
+
+Sampler Device::CreateSampler(const VkSamplerCreateInfo& ci) const {
+ VkSampler object;
+ Check(dld->vkCreateSampler(handle, &ci, nullptr, &object));
+ return Sampler(object, handle, *dld);
+}
+
+Framebuffer Device::CreateFramebuffer(const VkFramebufferCreateInfo& ci) const {
+ VkFramebuffer object;
+ Check(dld->vkCreateFramebuffer(handle, &ci, nullptr, &object));
+ return Framebuffer(object, handle, *dld);
+}
+
+CommandPool Device::CreateCommandPool(const VkCommandPoolCreateInfo& ci) const {
+ VkCommandPool object;
+ Check(dld->vkCreateCommandPool(handle, &ci, nullptr, &object));
+ return CommandPool(object, handle, *dld);
+}
+
+DescriptorUpdateTemplateKHR Device::CreateDescriptorUpdateTemplateKHR(
+ const VkDescriptorUpdateTemplateCreateInfoKHR& ci) const {
+ VkDescriptorUpdateTemplateKHR object;
+ Check(dld->vkCreateDescriptorUpdateTemplateKHR(handle, &ci, nullptr, &object));
+ return DescriptorUpdateTemplateKHR(object, handle, *dld);
+}
+
+QueryPool Device::CreateQueryPool(const VkQueryPoolCreateInfo& ci) const {
+ VkQueryPool object;
+ Check(dld->vkCreateQueryPool(handle, &ci, nullptr, &object));
+ return QueryPool(object, handle, *dld);
+}
+
+ShaderModule Device::CreateShaderModule(const VkShaderModuleCreateInfo& ci) const {
+ VkShaderModule object;
+ Check(dld->vkCreateShaderModule(handle, &ci, nullptr, &object));
+ return ShaderModule(object, handle, *dld);
+}
+
+SwapchainKHR Device::CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const {
+ VkSwapchainKHR object;
+ Check(dld->vkCreateSwapchainKHR(handle, &ci, nullptr, &object));
+ return SwapchainKHR(object, handle, *dld);
+}
+
+DeviceMemory Device::TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept {
+ VkDeviceMemory memory;
+ if (dld->vkAllocateMemory(handle, &ai, nullptr, &memory) != VK_SUCCESS) {
+ return {};
+ }
+ return DeviceMemory(memory, handle, *dld);
+}
+
+DeviceMemory Device::AllocateMemory(const VkMemoryAllocateInfo& ai) const {
+ VkDeviceMemory memory;
+ Check(dld->vkAllocateMemory(handle, &ai, nullptr, &memory));
+ return DeviceMemory(memory, handle, *dld);
+}
+
+VkMemoryRequirements Device::GetBufferMemoryRequirements(VkBuffer buffer) const noexcept {
+ VkMemoryRequirements requirements;
+ dld->vkGetBufferMemoryRequirements(handle, buffer, &requirements);
+ return requirements;
+}
+
+VkMemoryRequirements Device::GetImageMemoryRequirements(VkImage image) const noexcept {
+ VkMemoryRequirements requirements;
+ dld->vkGetImageMemoryRequirements(handle, image, &requirements);
+ return requirements;
+}
+
+void Device::UpdateDescriptorSets(Span<VkWriteDescriptorSet> writes,
+ Span<VkCopyDescriptorSet> copies) const noexcept {
+ dld->vkUpdateDescriptorSets(handle, writes.size(), writes.data(), copies.size(), copies.data());
+}
+
+VkPhysicalDeviceProperties PhysicalDevice::GetProperties() const noexcept {
+ VkPhysicalDeviceProperties properties;
+ dld->vkGetPhysicalDeviceProperties(physical_device, &properties);
+ return properties;
+}
+
+void PhysicalDevice::GetProperties2KHR(VkPhysicalDeviceProperties2KHR& properties) const noexcept {
+ dld->vkGetPhysicalDeviceProperties2KHR(physical_device, &properties);
+}
+
+VkPhysicalDeviceFeatures PhysicalDevice::GetFeatures() const noexcept {
+ VkPhysicalDeviceFeatures2KHR features2;
+ features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR;
+ features2.pNext = nullptr;
+ dld->vkGetPhysicalDeviceFeatures2KHR(physical_device, &features2);
+ return features2.features;
+}
+
+void PhysicalDevice::GetFeatures2KHR(VkPhysicalDeviceFeatures2KHR& features) const noexcept {
+ dld->vkGetPhysicalDeviceFeatures2KHR(physical_device, &features);
+}
+
+VkFormatProperties PhysicalDevice::GetFormatProperties(VkFormat format) const noexcept {
+ VkFormatProperties properties;
+ dld->vkGetPhysicalDeviceFormatProperties(physical_device, format, &properties);
+ return properties;
+}
+
+std::vector<VkExtensionProperties> PhysicalDevice::EnumerateDeviceExtensionProperties() const {
+ u32 num;
+ dld->vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &num, nullptr);
+ std::vector<VkExtensionProperties> properties(num);
+ dld->vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &num, properties.data());
+ return properties;
+}
+
+std::vector<VkQueueFamilyProperties> PhysicalDevice::GetQueueFamilyProperties() const {
+ u32 num;
+ dld->vkGetPhysicalDeviceQueueFamilyProperties(physical_device, &num, nullptr);
+ std::vector<VkQueueFamilyProperties> properties(num);
+ dld->vkGetPhysicalDeviceQueueFamilyProperties(physical_device, &num, properties.data());
+ return properties;
+}
+
+bool PhysicalDevice::GetSurfaceSupportKHR(u32 queue_family_index, VkSurfaceKHR surface) const {
+ VkBool32 supported;
+ Check(dld->vkGetPhysicalDeviceSurfaceSupportKHR(physical_device, queue_family_index, surface,
+ &supported));
+ return supported == VK_TRUE;
+}
+
+VkSurfaceCapabilitiesKHR PhysicalDevice::GetSurfaceCapabilitiesKHR(VkSurfaceKHR surface) const
+ noexcept {
+ VkSurfaceCapabilitiesKHR capabilities;
+ Check(dld->vkGetPhysicalDeviceSurfaceCapabilitiesKHR(physical_device, surface, &capabilities));
+ return capabilities;
+}
+
+std::vector<VkSurfaceFormatKHR> PhysicalDevice::GetSurfaceFormatsKHR(VkSurfaceKHR surface) const {
+ u32 num;
+ Check(dld->vkGetPhysicalDeviceSurfaceFormatsKHR(physical_device, surface, &num, nullptr));
+ std::vector<VkSurfaceFormatKHR> formats(num);
+ Check(
+ dld->vkGetPhysicalDeviceSurfaceFormatsKHR(physical_device, surface, &num, formats.data()));
+ return formats;
+}
+
+std::vector<VkPresentModeKHR> PhysicalDevice::GetSurfacePresentModesKHR(
+ VkSurfaceKHR surface) const {
+ u32 num;
+ Check(dld->vkGetPhysicalDeviceSurfacePresentModesKHR(physical_device, surface, &num, nullptr));
+ std::vector<VkPresentModeKHR> modes(num);
+ Check(dld->vkGetPhysicalDeviceSurfacePresentModesKHR(physical_device, surface, &num,
+ modes.data()));
+ return modes;
+}
+
+VkPhysicalDeviceMemoryProperties PhysicalDevice::GetMemoryProperties() const noexcept {
+ VkPhysicalDeviceMemoryProperties properties;
+ dld->vkGetPhysicalDeviceMemoryProperties(physical_device, &properties);
+ return properties;
+}
+
+std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
+ const InstanceDispatch& dld) {
+ u32 num;
+ if (dld.vkEnumerateInstanceExtensionProperties(nullptr, &num, nullptr) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ std::vector<VkExtensionProperties> properties(num);
+ if (dld.vkEnumerateInstanceExtensionProperties(nullptr, &num, properties.data()) !=
+ VK_SUCCESS) {
+ return std::nullopt;
+ }
+ return properties;
+}
+
+} // namespace Vulkan::vk
diff --git a/src/video_core/renderer_vulkan/wrapper.h b/src/video_core/renderer_vulkan/wrapper.h
new file mode 100644
index 000000000..fb3657819
--- /dev/null
+++ b/src/video_core/renderer_vulkan/wrapper.h
@@ -0,0 +1,987 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <exception>
+#include <iterator>
+#include <limits>
+#include <memory>
+#include <optional>
+#include <type_traits>
+#include <utility>
+#include <vector>
+
+#define VK_NO_PROTOTYPES
+#include <vulkan/vulkan.h>
+
+#include "common/common_types.h"
+
+namespace Vulkan::vk {
+
+/**
+ * Span for Vulkan arrays.
+ * Based on std::span but optimized for array access instead of iterators.
+ * Size returns uint32_t instead of size_t to ease interaction with Vulkan functions.
+ */
+template <typename T>
+class Span {
+public:
+ using value_type = T;
+ using size_type = u32;
+ using difference_type = std::ptrdiff_t;
+ using reference = const T&;
+ using const_reference = const T&;
+ using pointer = const T*;
+ using const_pointer = const T*;
+ using iterator = const T*;
+ using const_iterator = const T*;
+
+ /// Construct an empty span.
+ constexpr Span() noexcept = default;
+
+ /// Construct a span from a single element.
+ constexpr Span(const T& value) noexcept : ptr{&value}, num{1} {}
+
+ /// Construct a span from a range.
+ template <typename Range>
+ // requires std::data(const Range&)
+ // requires std::size(const Range&)
+ constexpr Span(const Range& range) : ptr{std::data(range)}, num{std::size(range)} {}
+
+ /// Construct a span from a pointer and a size.
+ /// This is inteded for subranges.
+ constexpr Span(const T* ptr, std::size_t num) noexcept : ptr{ptr}, num{num} {}
+
+ /// Returns the data pointer by the span.
+ constexpr const T* data() const noexcept {
+ return ptr;
+ }
+
+ /// Returns the number of elements in the span.
+ /// @note Returns a 32 bits integer because most Vulkan functions expect this type.
+ constexpr u32 size() const noexcept {
+ return static_cast<u32>(num);
+ }
+
+ /// Returns true when the span is empty.
+ constexpr bool empty() const noexcept {
+ return num == 0;
+ }
+
+ /// Returns a reference to the element in the passed index.
+ /// @pre: index < size()
+ constexpr const T& operator[](std::size_t index) const noexcept {
+ return ptr[index];
+ }
+
+ /// Returns an iterator to the beginning of the span.
+ constexpr const T* begin() const noexcept {
+ return ptr;
+ }
+
+ /// Returns an iterator to the end of the span.
+ constexpr const T* end() const noexcept {
+ return ptr + num;
+ }
+
+ /// Returns an iterator to the beginning of the span.
+ constexpr const T* cbegin() const noexcept {
+ return ptr;
+ }
+
+ /// Returns an iterator to the end of the span.
+ constexpr const T* cend() const noexcept {
+ return ptr + num;
+ }
+
+private:
+ const T* ptr = nullptr;
+ std::size_t num = 0;
+};
+
+/// Vulkan exception generated from a VkResult.
+class Exception final : public std::exception {
+public:
+ /// Construct the exception with a result.
+ /// @pre result != VK_SUCCESS
+ explicit Exception(VkResult result_) : result{result_} {}
+ virtual ~Exception() = default;
+
+ const char* what() const noexcept override;
+
+private:
+ VkResult result;
+};
+
+/// Converts a VkResult enum into a rodata string
+const char* ToString(VkResult) noexcept;
+
+/// Throws a Vulkan exception if result is not success.
+inline void Check(VkResult result) {
+ if (result != VK_SUCCESS) {
+ throw Exception(result);
+ }
+}
+
+/// Throws a Vulkan exception if result is an error.
+/// @return result
+inline VkResult Filter(VkResult result) {
+ if (result < 0) {
+ throw Exception(result);
+ }
+ return result;
+}
+
+/// Table holding Vulkan instance function pointers.
+struct InstanceDispatch {
+ PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr;
+
+ PFN_vkCreateInstance vkCreateInstance;
+ PFN_vkDestroyInstance vkDestroyInstance;
+ PFN_vkEnumerateInstanceExtensionProperties vkEnumerateInstanceExtensionProperties;
+
+ PFN_vkCreateDebugUtilsMessengerEXT vkCreateDebugUtilsMessengerEXT;
+ PFN_vkCreateDevice vkCreateDevice;
+ PFN_vkDestroyDebugUtilsMessengerEXT vkDestroyDebugUtilsMessengerEXT;
+ PFN_vkDestroyDevice vkDestroyDevice;
+ PFN_vkDestroySurfaceKHR vkDestroySurfaceKHR;
+ PFN_vkEnumerateDeviceExtensionProperties vkEnumerateDeviceExtensionProperties;
+ PFN_vkEnumeratePhysicalDevices vkEnumeratePhysicalDevices;
+ PFN_vkGetDeviceProcAddr vkGetDeviceProcAddr;
+ PFN_vkGetPhysicalDeviceFeatures2KHR vkGetPhysicalDeviceFeatures2KHR;
+ PFN_vkGetPhysicalDeviceFormatProperties vkGetPhysicalDeviceFormatProperties;
+ PFN_vkGetPhysicalDeviceMemoryProperties vkGetPhysicalDeviceMemoryProperties;
+ PFN_vkGetPhysicalDeviceProperties vkGetPhysicalDeviceProperties;
+ PFN_vkGetPhysicalDeviceProperties2KHR vkGetPhysicalDeviceProperties2KHR;
+ PFN_vkGetPhysicalDeviceQueueFamilyProperties vkGetPhysicalDeviceQueueFamilyProperties;
+ PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR;
+ PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR;
+ PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR;
+ PFN_vkGetPhysicalDeviceSurfaceSupportKHR vkGetPhysicalDeviceSurfaceSupportKHR;
+ PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR;
+ PFN_vkQueuePresentKHR vkQueuePresentKHR;
+};
+
+/// Table holding Vulkan device function pointers.
+struct DeviceDispatch : public InstanceDispatch {
+ PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR;
+ PFN_vkAllocateCommandBuffers vkAllocateCommandBuffers;
+ PFN_vkAllocateDescriptorSets vkAllocateDescriptorSets;
+ PFN_vkAllocateMemory vkAllocateMemory;
+ PFN_vkBeginCommandBuffer vkBeginCommandBuffer;
+ PFN_vkBindBufferMemory vkBindBufferMemory;
+ PFN_vkBindImageMemory vkBindImageMemory;
+ PFN_vkCmdBeginQuery vkCmdBeginQuery;
+ PFN_vkCmdBeginRenderPass vkCmdBeginRenderPass;
+ PFN_vkCmdBeginTransformFeedbackEXT vkCmdBeginTransformFeedbackEXT;
+ PFN_vkCmdBindDescriptorSets vkCmdBindDescriptorSets;
+ PFN_vkCmdBindIndexBuffer vkCmdBindIndexBuffer;
+ PFN_vkCmdBindPipeline vkCmdBindPipeline;
+ PFN_vkCmdBindTransformFeedbackBuffersEXT vkCmdBindTransformFeedbackBuffersEXT;
+ PFN_vkCmdBindVertexBuffers vkCmdBindVertexBuffers;
+ PFN_vkCmdBlitImage vkCmdBlitImage;
+ PFN_vkCmdClearAttachments vkCmdClearAttachments;
+ PFN_vkCmdCopyBuffer vkCmdCopyBuffer;
+ PFN_vkCmdCopyBufferToImage vkCmdCopyBufferToImage;
+ PFN_vkCmdCopyImage vkCmdCopyImage;
+ PFN_vkCmdCopyImageToBuffer vkCmdCopyImageToBuffer;
+ PFN_vkCmdDispatch vkCmdDispatch;
+ PFN_vkCmdDraw vkCmdDraw;
+ PFN_vkCmdDrawIndexed vkCmdDrawIndexed;
+ PFN_vkCmdEndQuery vkCmdEndQuery;
+ PFN_vkCmdEndRenderPass vkCmdEndRenderPass;
+ PFN_vkCmdEndTransformFeedbackEXT vkCmdEndTransformFeedbackEXT;
+ PFN_vkCmdFillBuffer vkCmdFillBuffer;
+ PFN_vkCmdPipelineBarrier vkCmdPipelineBarrier;
+ PFN_vkCmdPushConstants vkCmdPushConstants;
+ PFN_vkCmdSetBlendConstants vkCmdSetBlendConstants;
+ PFN_vkCmdSetCheckpointNV vkCmdSetCheckpointNV;
+ PFN_vkCmdSetDepthBias vkCmdSetDepthBias;
+ PFN_vkCmdSetDepthBounds vkCmdSetDepthBounds;
+ PFN_vkCmdSetScissor vkCmdSetScissor;
+ PFN_vkCmdSetStencilCompareMask vkCmdSetStencilCompareMask;
+ PFN_vkCmdSetStencilReference vkCmdSetStencilReference;
+ PFN_vkCmdSetStencilWriteMask vkCmdSetStencilWriteMask;
+ PFN_vkCmdSetViewport vkCmdSetViewport;
+ PFN_vkCreateBuffer vkCreateBuffer;
+ PFN_vkCreateBufferView vkCreateBufferView;
+ PFN_vkCreateCommandPool vkCreateCommandPool;
+ PFN_vkCreateComputePipelines vkCreateComputePipelines;
+ PFN_vkCreateDescriptorPool vkCreateDescriptorPool;
+ PFN_vkCreateDescriptorSetLayout vkCreateDescriptorSetLayout;
+ PFN_vkCreateDescriptorUpdateTemplateKHR vkCreateDescriptorUpdateTemplateKHR;
+ PFN_vkCreateFence vkCreateFence;
+ PFN_vkCreateFramebuffer vkCreateFramebuffer;
+ PFN_vkCreateGraphicsPipelines vkCreateGraphicsPipelines;
+ PFN_vkCreateImage vkCreateImage;
+ PFN_vkCreateImageView vkCreateImageView;
+ PFN_vkCreatePipelineLayout vkCreatePipelineLayout;
+ PFN_vkCreateQueryPool vkCreateQueryPool;
+ PFN_vkCreateRenderPass vkCreateRenderPass;
+ PFN_vkCreateSampler vkCreateSampler;
+ PFN_vkCreateSemaphore vkCreateSemaphore;
+ PFN_vkCreateShaderModule vkCreateShaderModule;
+ PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR;
+ PFN_vkDestroyBuffer vkDestroyBuffer;
+ PFN_vkDestroyBufferView vkDestroyBufferView;
+ PFN_vkDestroyCommandPool vkDestroyCommandPool;
+ PFN_vkDestroyDescriptorPool vkDestroyDescriptorPool;
+ PFN_vkDestroyDescriptorSetLayout vkDestroyDescriptorSetLayout;
+ PFN_vkDestroyDescriptorUpdateTemplateKHR vkDestroyDescriptorUpdateTemplateKHR;
+ PFN_vkDestroyFence vkDestroyFence;
+ PFN_vkDestroyFramebuffer vkDestroyFramebuffer;
+ PFN_vkDestroyImage vkDestroyImage;
+ PFN_vkDestroyImageView vkDestroyImageView;
+ PFN_vkDestroyPipeline vkDestroyPipeline;
+ PFN_vkDestroyPipelineLayout vkDestroyPipelineLayout;
+ PFN_vkDestroyQueryPool vkDestroyQueryPool;
+ PFN_vkDestroyRenderPass vkDestroyRenderPass;
+ PFN_vkDestroySampler vkDestroySampler;
+ PFN_vkDestroySemaphore vkDestroySemaphore;
+ PFN_vkDestroyShaderModule vkDestroyShaderModule;
+ PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR;
+ PFN_vkDeviceWaitIdle vkDeviceWaitIdle;
+ PFN_vkEndCommandBuffer vkEndCommandBuffer;
+ PFN_vkFreeCommandBuffers vkFreeCommandBuffers;
+ PFN_vkFreeDescriptorSets vkFreeDescriptorSets;
+ PFN_vkFreeMemory vkFreeMemory;
+ PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements;
+ PFN_vkGetDeviceQueue vkGetDeviceQueue;
+ PFN_vkGetFenceStatus vkGetFenceStatus;
+ PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements;
+ PFN_vkGetQueryPoolResults vkGetQueryPoolResults;
+ PFN_vkGetQueueCheckpointDataNV vkGetQueueCheckpointDataNV;
+ PFN_vkMapMemory vkMapMemory;
+ PFN_vkQueueSubmit vkQueueSubmit;
+ PFN_vkResetFences vkResetFences;
+ PFN_vkResetQueryPoolEXT vkResetQueryPoolEXT;
+ PFN_vkUnmapMemory vkUnmapMemory;
+ PFN_vkUpdateDescriptorSetWithTemplateKHR vkUpdateDescriptorSetWithTemplateKHR;
+ PFN_vkUpdateDescriptorSets vkUpdateDescriptorSets;
+ PFN_vkWaitForFences vkWaitForFences;
+};
+
+/// Loads instance agnostic function pointers.
+/// @return True on success, false on error.
+bool Load(InstanceDispatch&) noexcept;
+
+/// Loads instance function pointers.
+/// @return True on success, false on error.
+bool Load(VkInstance, InstanceDispatch&) noexcept;
+
+void Destroy(VkInstance, const InstanceDispatch&) noexcept;
+void Destroy(VkDevice, const InstanceDispatch&) noexcept;
+
+void Destroy(VkDevice, VkBuffer, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkBufferView, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkCommandPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorSetLayout, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorUpdateTemplateKHR, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDeviceMemory, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkFence, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkFramebuffer, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkImage, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkImageView, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipeline, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipelineLayout, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkQueryPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkRenderPass, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSampler, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSwapchainKHR, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSemaphore, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkShaderModule, const DeviceDispatch&) noexcept;
+void Destroy(VkInstance, VkDebugUtilsMessengerEXT, const InstanceDispatch&) noexcept;
+void Destroy(VkInstance, VkSurfaceKHR, const InstanceDispatch&) noexcept;
+
+VkResult Free(VkDevice, VkDescriptorPool, Span<VkDescriptorSet>, const DeviceDispatch&) noexcept;
+VkResult Free(VkDevice, VkCommandPool, Span<VkCommandBuffer>, const DeviceDispatch&) noexcept;
+
+template <typename Type, typename OwnerType, typename Dispatch>
+class Handle;
+
+/// Handle with an owning type.
+/// Analogue to std::unique_ptr.
+template <typename Type, typename OwnerType, typename Dispatch>
+class Handle {
+public:
+ /// Construct a handle and hold it's ownership.
+ explicit Handle(Type handle_, OwnerType owner_, const Dispatch& dld_) noexcept
+ : handle{handle_}, owner{owner_}, dld{&dld_} {}
+
+ /// Construct an empty handle.
+ Handle() = default;
+
+ /// Copying Vulkan objects is not supported and will never be.
+ Handle(const Handle&) = delete;
+ Handle& operator=(const Handle&) = delete;
+
+ /// Construct a handle transfering the ownership from another handle.
+ Handle(Handle&& rhs) noexcept
+ : handle{std::exchange(rhs.handle, nullptr)}, owner{rhs.owner}, dld{rhs.dld} {}
+
+ /// Assign the current handle transfering the ownership from another handle.
+ /// Destroys any previously held object.
+ Handle& operator=(Handle&& rhs) noexcept {
+ Release();
+ handle = std::exchange(rhs.handle, nullptr);
+ owner = rhs.owner;
+ dld = rhs.dld;
+ return *this;
+ }
+
+ /// Destroys the current handle if it existed.
+ ~Handle() noexcept {
+ Release();
+ }
+
+ /// Destroys any held object.
+ void reset() noexcept {
+ Release();
+ handle = nullptr;
+ }
+
+ /// Returns the address of the held object.
+ /// Intended for Vulkan structures that expect a pointer to an array.
+ const Type* address() const noexcept {
+ return std::addressof(handle);
+ }
+
+ /// Returns the held Vulkan handle.
+ Type operator*() const noexcept {
+ return handle;
+ }
+
+ /// Returns true when there's a held object.
+ explicit operator bool() const noexcept {
+ return handle != nullptr;
+ }
+
+protected:
+ Type handle = nullptr;
+ OwnerType owner = nullptr;
+ const Dispatch* dld = nullptr;
+
+private:
+ /// Destroys the held object if it exists.
+ void Release() noexcept {
+ if (handle) {
+ Destroy(owner, handle, *dld);
+ }
+ }
+};
+
+/// Dummy type used to specify a handle has no owner.
+struct NoOwner {};
+
+/// Handle without an owning type.
+/// Analogue to std::unique_ptr
+template <typename Type, typename Dispatch>
+class Handle<Type, NoOwner, Dispatch> {
+public:
+ /// Construct a handle and hold it's ownership.
+ explicit Handle(Type handle_, const Dispatch& dld_) noexcept : handle{handle_}, dld{&dld_} {}
+
+ /// Construct an empty handle.
+ Handle() noexcept = default;
+
+ /// Copying Vulkan objects is not supported and will never be.
+ Handle(const Handle&) = delete;
+ Handle& operator=(const Handle&) = delete;
+
+ /// Construct a handle transfering ownership from another handle.
+ Handle(Handle&& rhs) noexcept : handle{std::exchange(rhs.handle, nullptr)}, dld{rhs.dld} {}
+
+ /// Assign the current handle transfering the ownership from another handle.
+ /// Destroys any previously held object.
+ Handle& operator=(Handle&& rhs) noexcept {
+ Release();
+ handle = std::exchange(rhs.handle, nullptr);
+ dld = rhs.dld;
+ return *this;
+ }
+
+ /// Destroys the current handle if it existed.
+ ~Handle() noexcept {
+ Release();
+ }
+
+ /// Destroys any held object.
+ void reset() noexcept {
+ Release();
+ handle = nullptr;
+ }
+
+ /// Returns the address of the held object.
+ /// Intended for Vulkan structures that expect a pointer to an array.
+ const Type* address() const noexcept {
+ return std::addressof(handle);
+ }
+
+ /// Returns the held Vulkan handle.
+ Type operator*() const noexcept {
+ return handle;
+ }
+
+ /// Returns true when there's a held object.
+ operator bool() const noexcept {
+ return handle != nullptr;
+ }
+
+protected:
+ Type handle = nullptr;
+ const Dispatch* dld = nullptr;
+
+private:
+ /// Destroys the held object if it exists.
+ void Release() noexcept {
+ if (handle) {
+ Destroy(handle, *dld);
+ }
+ }
+};
+
+/// Array of a pool allocation.
+/// Analogue to std::vector
+template <typename AllocationType, typename PoolType>
+class PoolAllocations {
+public:
+ /// Construct an empty allocation.
+ PoolAllocations() = default;
+
+ /// Construct an allocation. Errors are reported through IsOutOfPoolMemory().
+ explicit PoolAllocations(std::unique_ptr<AllocationType[]> allocations, std::size_t num,
+ VkDevice device, PoolType pool, const DeviceDispatch& dld) noexcept
+ : allocations{std::move(allocations)}, num{num}, device{device}, pool{pool}, dld{&dld} {}
+
+ /// Copying Vulkan allocations is not supported and will never be.
+ PoolAllocations(const PoolAllocations&) = delete;
+ PoolAllocations& operator=(const PoolAllocations&) = delete;
+
+ /// Construct an allocation transfering ownership from another allocation.
+ PoolAllocations(PoolAllocations&& rhs) noexcept
+ : allocations{std::move(rhs.allocations)}, num{rhs.num}, device{rhs.device}, pool{rhs.pool},
+ dld{rhs.dld} {}
+
+ /// Assign an allocation transfering ownership from another allocation.
+ /// Releases any previously held allocation.
+ PoolAllocations& operator=(PoolAllocations&& rhs) noexcept {
+ Release();
+ allocations = std::move(rhs.allocations);
+ num = rhs.num;
+ device = rhs.device;
+ pool = rhs.pool;
+ dld = rhs.dld;
+ return *this;
+ }
+
+ /// Destroys any held allocation.
+ ~PoolAllocations() {
+ Release();
+ }
+
+ /// Returns the number of allocations.
+ std::size_t size() const noexcept {
+ return num;
+ }
+
+ /// Returns a pointer to the array of allocations.
+ AllocationType const* data() const noexcept {
+ return allocations.get();
+ }
+
+ /// Returns the allocation in the specified index.
+ /// @pre index < size()
+ AllocationType operator[](std::size_t index) const noexcept {
+ return allocations[index];
+ }
+
+ /// True when a pool fails to construct.
+ bool IsOutOfPoolMemory() const noexcept {
+ return !device;
+ }
+
+private:
+ /// Destroys the held allocations if they exist.
+ void Release() noexcept {
+ if (!allocations) {
+ return;
+ }
+ const Span<AllocationType> span(allocations.get(), num);
+ const VkResult result = Free(device, pool, span, *dld);
+ // There's no way to report errors from a destructor.
+ if (result != VK_SUCCESS) {
+ std::terminate();
+ }
+ }
+
+ std::unique_ptr<AllocationType[]> allocations;
+ std::size_t num = 0;
+ VkDevice device = nullptr;
+ PoolType pool = nullptr;
+ const DeviceDispatch* dld = nullptr;
+};
+
+using BufferView = Handle<VkBufferView, VkDevice, DeviceDispatch>;
+using DebugCallback = Handle<VkDebugUtilsMessengerEXT, VkInstance, InstanceDispatch>;
+using DescriptorSetLayout = Handle<VkDescriptorSetLayout, VkDevice, DeviceDispatch>;
+using DescriptorUpdateTemplateKHR = Handle<VkDescriptorUpdateTemplateKHR, VkDevice, DeviceDispatch>;
+using Framebuffer = Handle<VkFramebuffer, VkDevice, DeviceDispatch>;
+using ImageView = Handle<VkImageView, VkDevice, DeviceDispatch>;
+using Pipeline = Handle<VkPipeline, VkDevice, DeviceDispatch>;
+using PipelineLayout = Handle<VkPipelineLayout, VkDevice, DeviceDispatch>;
+using QueryPool = Handle<VkQueryPool, VkDevice, DeviceDispatch>;
+using RenderPass = Handle<VkRenderPass, VkDevice, DeviceDispatch>;
+using Sampler = Handle<VkSampler, VkDevice, DeviceDispatch>;
+using Semaphore = Handle<VkSemaphore, VkDevice, DeviceDispatch>;
+using ShaderModule = Handle<VkShaderModule, VkDevice, DeviceDispatch>;
+using SurfaceKHR = Handle<VkSurfaceKHR, VkInstance, InstanceDispatch>;
+
+using DescriptorSets = PoolAllocations<VkDescriptorSet, VkDescriptorPool>;
+using CommandBuffers = PoolAllocations<VkCommandBuffer, VkCommandPool>;
+
+/// Vulkan instance owning handle.
+class Instance : public Handle<VkInstance, NoOwner, InstanceDispatch> {
+ using Handle<VkInstance, NoOwner, InstanceDispatch>::Handle;
+
+public:
+ /// Creates a Vulkan instance. Use "operator bool" for error handling.
+ static Instance Create(Span<const char*> layers, Span<const char*> extensions,
+ InstanceDispatch& dld) noexcept;
+
+ /// Enumerates physical devices.
+ /// @return Physical devices and an empty handle on failure.
+ std::optional<std::vector<VkPhysicalDevice>> EnumeratePhysicalDevices();
+
+ /// Tries to create a debug callback messenger. Returns an empty handle on failure.
+ DebugCallback TryCreateDebugCallback(PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept;
+};
+
+class Queue {
+public:
+ /// Construct an empty queue handle.
+ constexpr Queue() noexcept = default;
+
+ /// Construct a queue handle.
+ constexpr Queue(VkQueue queue, const DeviceDispatch& dld) noexcept : queue{queue}, dld{&dld} {}
+
+ /// Returns the checkpoint data.
+ /// @note Returns an empty vector when the function pointer is not present.
+ std::vector<VkCheckpointDataNV> GetCheckpointDataNV(const DeviceDispatch& dld) const;
+
+ void Submit(Span<VkSubmitInfo> submit_infos, VkFence fence) const {
+ Check(dld->vkQueueSubmit(queue, submit_infos.size(), submit_infos.data(), fence));
+ }
+
+ VkResult Present(const VkPresentInfoKHR& present_info) const noexcept {
+ return dld->vkQueuePresentKHR(queue, &present_info);
+ }
+
+private:
+ VkQueue queue = nullptr;
+ const DeviceDispatch* dld = nullptr;
+};
+
+class Buffer : public Handle<VkBuffer, VkDevice, DeviceDispatch> {
+ using Handle<VkBuffer, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Attaches a memory allocation.
+ void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+};
+
+class Image : public Handle<VkImage, VkDevice, DeviceDispatch> {
+ using Handle<VkImage, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Attaches a memory allocation.
+ void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+};
+
+class DeviceMemory : public Handle<VkDeviceMemory, VkDevice, DeviceDispatch> {
+ using Handle<VkDeviceMemory, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ u8* Map(VkDeviceSize offset, VkDeviceSize size) const {
+ void* data;
+ Check(dld->vkMapMemory(owner, handle, offset, size, 0, &data));
+ return static_cast<u8*>(data);
+ }
+
+ void Unmap() const noexcept {
+ dld->vkUnmapMemory(owner, handle);
+ }
+};
+
+class Fence : public Handle<VkFence, VkDevice, DeviceDispatch> {
+ using Handle<VkFence, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ VkResult Wait(u64 timeout = std::numeric_limits<u64>::max()) const noexcept {
+ return dld->vkWaitForFences(owner, 1, &handle, true, timeout);
+ }
+
+ VkResult GetStatus() const noexcept {
+ return dld->vkGetFenceStatus(owner, handle);
+ }
+
+ void Reset() const {
+ Check(dld->vkResetFences(owner, 1, &handle));
+ }
+};
+
+class DescriptorPool : public Handle<VkDescriptorPool, VkDevice, DeviceDispatch> {
+ using Handle<VkDescriptorPool, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ DescriptorSets Allocate(const VkDescriptorSetAllocateInfo& ai) const;
+};
+
+class CommandPool : public Handle<VkCommandPool, VkDevice, DeviceDispatch> {
+ using Handle<VkCommandPool, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ CommandBuffers Allocate(std::size_t num_buffers,
+ VkCommandBufferLevel level = VK_COMMAND_BUFFER_LEVEL_PRIMARY) const;
+};
+
+class SwapchainKHR : public Handle<VkSwapchainKHR, VkDevice, DeviceDispatch> {
+ using Handle<VkSwapchainKHR, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ std::vector<VkImage> GetImages() const;
+};
+
+class Device : public Handle<VkDevice, NoOwner, DeviceDispatch> {
+ using Handle<VkDevice, NoOwner, DeviceDispatch>::Handle;
+
+public:
+ static Device Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
+ Span<const char*> enabled_extensions,
+ const VkPhysicalDeviceFeatures2& enabled_features,
+ DeviceDispatch& dld) noexcept;
+
+ Queue GetQueue(u32 family_index) const noexcept;
+
+ Buffer CreateBuffer(const VkBufferCreateInfo& ci) const;
+
+ BufferView CreateBufferView(const VkBufferViewCreateInfo& ci) const;
+
+ Image CreateImage(const VkImageCreateInfo& ci) const;
+
+ ImageView CreateImageView(const VkImageViewCreateInfo& ci) const;
+
+ Semaphore CreateSemaphore() const;
+
+ Fence CreateFence(const VkFenceCreateInfo& ci) const;
+
+ DescriptorPool CreateDescriptorPool(const VkDescriptorPoolCreateInfo& ci) const;
+
+ RenderPass CreateRenderPass(const VkRenderPassCreateInfo& ci) const;
+
+ DescriptorSetLayout CreateDescriptorSetLayout(const VkDescriptorSetLayoutCreateInfo& ci) const;
+
+ PipelineLayout CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const;
+
+ Pipeline CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const;
+
+ Pipeline CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const;
+
+ Sampler CreateSampler(const VkSamplerCreateInfo& ci) const;
+
+ Framebuffer CreateFramebuffer(const VkFramebufferCreateInfo& ci) const;
+
+ CommandPool CreateCommandPool(const VkCommandPoolCreateInfo& ci) const;
+
+ DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplateKHR(
+ const VkDescriptorUpdateTemplateCreateInfoKHR& ci) const;
+
+ QueryPool CreateQueryPool(const VkQueryPoolCreateInfo& ci) const;
+
+ ShaderModule CreateShaderModule(const VkShaderModuleCreateInfo& ci) const;
+
+ SwapchainKHR CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const;
+
+ DeviceMemory TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept;
+
+ DeviceMemory AllocateMemory(const VkMemoryAllocateInfo& ai) const;
+
+ VkMemoryRequirements GetBufferMemoryRequirements(VkBuffer buffer) const noexcept;
+
+ VkMemoryRequirements GetImageMemoryRequirements(VkImage image) const noexcept;
+
+ void UpdateDescriptorSets(Span<VkWriteDescriptorSet> writes,
+ Span<VkCopyDescriptorSet> copies) const noexcept;
+
+ void UpdateDescriptorSet(VkDescriptorSet set, VkDescriptorUpdateTemplateKHR update_template,
+ const void* data) const noexcept {
+ dld->vkUpdateDescriptorSetWithTemplateKHR(handle, set, update_template, data);
+ }
+
+ VkResult AcquireNextImageKHR(VkSwapchainKHR swapchain, u64 timeout, VkSemaphore semaphore,
+ VkFence fence, u32* image_index) const noexcept {
+ return dld->vkAcquireNextImageKHR(handle, swapchain, timeout, semaphore, fence,
+ image_index);
+ }
+
+ VkResult WaitIdle() const noexcept {
+ return dld->vkDeviceWaitIdle(handle);
+ }
+
+ void ResetQueryPoolEXT(VkQueryPool query_pool, u32 first, u32 count) const noexcept {
+ dld->vkResetQueryPoolEXT(handle, query_pool, first, count);
+ }
+
+ void GetQueryResults(VkQueryPool query_pool, u32 first, u32 count, std::size_t data_size,
+ void* data, VkDeviceSize stride, VkQueryResultFlags flags) const {
+ Check(dld->vkGetQueryPoolResults(handle, query_pool, first, count, data_size, data, stride,
+ flags));
+ }
+
+ template <typename T>
+ T GetQueryResult(VkQueryPool query_pool, u32 first, VkQueryResultFlags flags) const {
+ static_assert(std::is_trivially_copyable_v<T>);
+ T value;
+ GetQueryResults(query_pool, first, 1, sizeof(T), &value, sizeof(T), flags);
+ return value;
+ }
+};
+
+class PhysicalDevice {
+public:
+ constexpr PhysicalDevice() noexcept = default;
+
+ constexpr PhysicalDevice(VkPhysicalDevice physical_device, const InstanceDispatch& dld) noexcept
+ : physical_device{physical_device}, dld{&dld} {}
+
+ constexpr operator VkPhysicalDevice() const noexcept {
+ return physical_device;
+ }
+
+ VkPhysicalDeviceProperties GetProperties() const noexcept;
+
+ void GetProperties2KHR(VkPhysicalDeviceProperties2KHR&) const noexcept;
+
+ VkPhysicalDeviceFeatures GetFeatures() const noexcept;
+
+ void GetFeatures2KHR(VkPhysicalDeviceFeatures2KHR&) const noexcept;
+
+ VkFormatProperties GetFormatProperties(VkFormat) const noexcept;
+
+ std::vector<VkExtensionProperties> EnumerateDeviceExtensionProperties() const;
+
+ std::vector<VkQueueFamilyProperties> GetQueueFamilyProperties() const;
+
+ bool GetSurfaceSupportKHR(u32 queue_family_index, VkSurfaceKHR) const;
+
+ VkSurfaceCapabilitiesKHR GetSurfaceCapabilitiesKHR(VkSurfaceKHR) const noexcept;
+
+ std::vector<VkSurfaceFormatKHR> GetSurfaceFormatsKHR(VkSurfaceKHR) const;
+
+ std::vector<VkPresentModeKHR> GetSurfacePresentModesKHR(VkSurfaceKHR) const;
+
+ VkPhysicalDeviceMemoryProperties GetMemoryProperties() const noexcept;
+
+private:
+ VkPhysicalDevice physical_device = nullptr;
+ const InstanceDispatch* dld = nullptr;
+};
+
+class CommandBuffer {
+public:
+ CommandBuffer() noexcept = default;
+
+ explicit CommandBuffer(VkCommandBuffer handle, const DeviceDispatch& dld) noexcept
+ : handle{handle}, dld{&dld} {}
+
+ const VkCommandBuffer* address() const noexcept {
+ return &handle;
+ }
+
+ void Begin(const VkCommandBufferBeginInfo& begin_info) const {
+ Check(dld->vkBeginCommandBuffer(handle, &begin_info));
+ }
+
+ void End() const {
+ Check(dld->vkEndCommandBuffer(handle));
+ }
+
+ void BeginRenderPass(const VkRenderPassBeginInfo& renderpass_bi,
+ VkSubpassContents contents) const noexcept {
+ dld->vkCmdBeginRenderPass(handle, &renderpass_bi, contents);
+ }
+
+ void EndRenderPass() const noexcept {
+ dld->vkCmdEndRenderPass(handle);
+ }
+
+ void BeginQuery(VkQueryPool query_pool, u32 query, VkQueryControlFlags flags) const noexcept {
+ dld->vkCmdBeginQuery(handle, query_pool, query, flags);
+ }
+
+ void EndQuery(VkQueryPool query_pool, u32 query) const noexcept {
+ dld->vkCmdEndQuery(handle, query_pool, query);
+ }
+
+ void BindDescriptorSets(VkPipelineBindPoint bind_point, VkPipelineLayout layout, u32 first,
+ Span<VkDescriptorSet> sets, Span<u32> dynamic_offsets) const noexcept {
+ dld->vkCmdBindDescriptorSets(handle, bind_point, layout, first, sets.size(), sets.data(),
+ dynamic_offsets.size(), dynamic_offsets.data());
+ }
+
+ void BindPipeline(VkPipelineBindPoint bind_point, VkPipeline pipeline) const noexcept {
+ dld->vkCmdBindPipeline(handle, bind_point, pipeline);
+ }
+
+ void BindIndexBuffer(VkBuffer buffer, VkDeviceSize offset, VkIndexType index_type) const
+ noexcept {
+ dld->vkCmdBindIndexBuffer(handle, buffer, offset, index_type);
+ }
+
+ void BindVertexBuffers(u32 first, u32 count, const VkBuffer* buffers,
+ const VkDeviceSize* offsets) const noexcept {
+ dld->vkCmdBindVertexBuffers(handle, first, count, buffers, offsets);
+ }
+
+ void BindVertexBuffer(u32 binding, VkBuffer buffer, VkDeviceSize offset) const noexcept {
+ BindVertexBuffers(binding, 1, &buffer, &offset);
+ }
+
+ void Draw(u32 vertex_count, u32 instance_count, u32 first_vertex, u32 first_instance) const
+ noexcept {
+ dld->vkCmdDraw(handle, vertex_count, instance_count, first_vertex, first_instance);
+ }
+
+ void DrawIndexed(u32 index_count, u32 instance_count, u32 first_index, u32 vertex_offset,
+ u32 first_instance) const noexcept {
+ dld->vkCmdDrawIndexed(handle, index_count, instance_count, first_index, vertex_offset,
+ first_instance);
+ }
+
+ void ClearAttachments(Span<VkClearAttachment> attachments, Span<VkClearRect> rects) const
+ noexcept {
+ dld->vkCmdClearAttachments(handle, attachments.size(), attachments.data(), rects.size(),
+ rects.data());
+ }
+
+ void BlitImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+ VkImageLayout dst_layout, Span<VkImageBlit> regions, VkFilter filter) const
+ noexcept {
+ dld->vkCmdBlitImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+ regions.data(), filter);
+ }
+
+ void Dispatch(u32 x, u32 y, u32 z) const noexcept {
+ dld->vkCmdDispatch(handle, x, y, z);
+ }
+
+ void PipelineBarrier(VkPipelineStageFlags src_stage_mask, VkPipelineStageFlags dst_stage_mask,
+ VkDependencyFlags dependency_flags, Span<VkMemoryBarrier> memory_barriers,
+ Span<VkBufferMemoryBarrier> buffer_barriers,
+ Span<VkImageMemoryBarrier> image_barriers) const noexcept {
+ dld->vkCmdPipelineBarrier(handle, src_stage_mask, dst_stage_mask, dependency_flags,
+ memory_barriers.size(), memory_barriers.data(),
+ buffer_barriers.size(), buffer_barriers.data(),
+ image_barriers.size(), image_barriers.data());
+ }
+
+ void CopyBufferToImage(VkBuffer src_buffer, VkImage dst_image, VkImageLayout dst_image_layout,
+ Span<VkBufferImageCopy> regions) const noexcept {
+ dld->vkCmdCopyBufferToImage(handle, src_buffer, dst_image, dst_image_layout, regions.size(),
+ regions.data());
+ }
+
+ void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer, Span<VkBufferCopy> regions) const
+ noexcept {
+ dld->vkCmdCopyBuffer(handle, src_buffer, dst_buffer, regions.size(), regions.data());
+ }
+
+ void CopyImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+ VkImageLayout dst_layout, Span<VkImageCopy> regions) const noexcept {
+ dld->vkCmdCopyImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+ regions.data());
+ }
+
+ void CopyImageToBuffer(VkImage src_image, VkImageLayout src_layout, VkBuffer dst_buffer,
+ Span<VkBufferImageCopy> regions) const noexcept {
+ dld->vkCmdCopyImageToBuffer(handle, src_image, src_layout, dst_buffer, regions.size(),
+ regions.data());
+ }
+
+ void FillBuffer(VkBuffer dst_buffer, VkDeviceSize dst_offset, VkDeviceSize size, u32 data) const
+ noexcept {
+ dld->vkCmdFillBuffer(handle, dst_buffer, dst_offset, size, data);
+ }
+
+ void PushConstants(VkPipelineLayout layout, VkShaderStageFlags flags, u32 offset, u32 size,
+ const void* values) const noexcept {
+ dld->vkCmdPushConstants(handle, layout, flags, offset, size, values);
+ }
+
+ void SetCheckpointNV(const void* checkpoint_marker) const noexcept {
+ dld->vkCmdSetCheckpointNV(handle, checkpoint_marker);
+ }
+
+ void SetViewport(u32 first, Span<VkViewport> viewports) const noexcept {
+ dld->vkCmdSetViewport(handle, first, viewports.size(), viewports.data());
+ }
+
+ void SetScissor(u32 first, Span<VkRect2D> scissors) const noexcept {
+ dld->vkCmdSetScissor(handle, first, scissors.size(), scissors.data());
+ }
+
+ void SetBlendConstants(const float blend_constants[4]) const noexcept {
+ dld->vkCmdSetBlendConstants(handle, blend_constants);
+ }
+
+ void SetStencilCompareMask(VkStencilFaceFlags face_mask, u32 compare_mask) const noexcept {
+ dld->vkCmdSetStencilCompareMask(handle, face_mask, compare_mask);
+ }
+
+ void SetStencilReference(VkStencilFaceFlags face_mask, u32 reference) const noexcept {
+ dld->vkCmdSetStencilReference(handle, face_mask, reference);
+ }
+
+ void SetStencilWriteMask(VkStencilFaceFlags face_mask, u32 write_mask) const noexcept {
+ dld->vkCmdSetStencilWriteMask(handle, face_mask, write_mask);
+ }
+
+ void SetDepthBias(float constant_factor, float clamp, float slope_factor) const noexcept {
+ dld->vkCmdSetDepthBias(handle, constant_factor, clamp, slope_factor);
+ }
+
+ void SetDepthBounds(float min_depth_bounds, float max_depth_bounds) const noexcept {
+ dld->vkCmdSetDepthBounds(handle, min_depth_bounds, max_depth_bounds);
+ }
+
+ void BindTransformFeedbackBuffersEXT(u32 first, u32 count, const VkBuffer* buffers,
+ const VkDeviceSize* offsets,
+ const VkDeviceSize* sizes) const noexcept {
+ dld->vkCmdBindTransformFeedbackBuffersEXT(handle, first, count, buffers, offsets, sizes);
+ }
+
+ void BeginTransformFeedbackEXT(u32 first_counter_buffer, u32 counter_buffers_count,
+ const VkBuffer* counter_buffers,
+ const VkDeviceSize* counter_buffer_offsets) const noexcept {
+ dld->vkCmdBeginTransformFeedbackEXT(handle, first_counter_buffer, counter_buffers_count,
+ counter_buffers, counter_buffer_offsets);
+ }
+
+ void EndTransformFeedbackEXT(u32 first_counter_buffer, u32 counter_buffers_count,
+ const VkBuffer* counter_buffers,
+ const VkDeviceSize* counter_buffer_offsets) const noexcept {
+ dld->vkCmdEndTransformFeedbackEXT(handle, first_counter_buffer, counter_buffers_count,
+ counter_buffers, counter_buffer_offsets);
+ }
+
+private:
+ VkCommandBuffer handle;
+ const DeviceDispatch* dld;
+};
+
+std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
+ const InstanceDispatch& dld);
+
+} // namespace Vulkan::vk
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 2fe787d6f..0f4c3103a 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -235,34 +235,30 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
case OpCode::Id::LEA_IMM:
case OpCode::Id::LEA_RZ:
case OpCode::Id::LEA_HI: {
- const auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
+ auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
switch (opcode->get().GetId()) {
case OpCode::Id::LEA_R2: {
return {GetRegister(instr.gpr20), GetRegister(instr.gpr39),
Immediate(static_cast<u32>(instr.lea.r2.entry_a))};
}
-
case OpCode::Id::LEA_R1: {
const bool neg = instr.lea.r1.neg != 0;
return {GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
GetRegister(instr.gpr20),
Immediate(static_cast<u32>(instr.lea.r1.entry_a))};
}
-
case OpCode::Id::LEA_IMM: {
const bool neg = instr.lea.imm.neg != 0;
return {Immediate(static_cast<u32>(instr.lea.imm.entry_a)),
GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
Immediate(static_cast<u32>(instr.lea.imm.entry_b))};
}
-
case OpCode::Id::LEA_RZ: {
const bool neg = instr.lea.rz.neg != 0;
return {GetConstBuffer(instr.lea.rz.cb_index, instr.lea.rz.cb_offset),
GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
Immediate(static_cast<u32>(instr.lea.rz.entry_a))};
}
-
case OpCode::Id::LEA_HI:
default:
UNIMPLEMENTED_MSG("Unhandled LEA subinstruction: {}", opcode->get().GetName());
@@ -275,12 +271,9 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
UNIMPLEMENTED_IF_MSG(instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex),
"Unhandled LEA Predicate");
- const Node shifted_c =
- Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, Immediate(1), op_c);
- const Node mul_bc = Operation(OperationCode::IMul, NO_PRECISE, op_b, shifted_c);
- const Node value = Operation(OperationCode::IAdd, NO_PRECISE, op_a, mul_bc);
-
- SetRegister(bb, instr.gpr0, value);
+ Node value = Operation(OperationCode::ILogicalShiftLeft, std::move(op_a), std::move(op_c));
+ value = Operation(OperationCode::IAdd, std::move(op_b), std::move(value));
+ SetRegister(bb, instr.gpr0, std::move(value));
break;
}
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index 6ead42070..c72690b2b 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -138,18 +138,23 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
- value = [&]() {
+ value = [&] {
+ if (instr.conversion.src_size != instr.conversion.dst_size) {
+ // Rounding operations only matter when the source and destination conversion size
+ // is the same.
+ return value;
+ }
switch (instr.conversion.f2f.GetRoundingMode()) {
case Tegra::Shader::F2fRoundingOp::None:
return value;
case Tegra::Shader::F2fRoundingOp::Round:
- return Operation(OperationCode::FRoundEven, PRECISE, value);
+ return Operation(OperationCode::FRoundEven, value);
case Tegra::Shader::F2fRoundingOp::Floor:
- return Operation(OperationCode::FFloor, PRECISE, value);
+ return Operation(OperationCode::FFloor, value);
case Tegra::Shader::F2fRoundingOp::Ceil:
- return Operation(OperationCode::FCeil, PRECISE, value);
+ return Operation(OperationCode::FCeil, value);
case Tegra::Shader::F2fRoundingOp::Trunc:
- return Operation(OperationCode::FTrunc, PRECISE, value);
+ return Operation(OperationCode::FTrunc, value);
default:
UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
static_cast<u32>(instr.conversion.f2f.rounding.Value()));
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index b5fbc4d58..28a49addd 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -19,7 +19,6 @@ namespace VideoCommon::Shader {
using Tegra::Shader::AtomicOp;
using Tegra::Shader::AtomicType;
using Tegra::Shader::Attribute;
-using Tegra::Shader::GlobalAtomicOp;
using Tegra::Shader::GlobalAtomicType;
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
@@ -28,6 +27,28 @@ using Tegra::Shader::StoreType;
namespace {
+Node GetAtomOperation(AtomicOp op, bool is_signed, Node memory, Node data) {
+ const OperationCode operation_code = [op] {
+ switch (op) {
+ case AtomicOp::Add:
+ return OperationCode::AtomicIAdd;
+ case AtomicOp::Min:
+ return OperationCode::AtomicIMin;
+ case AtomicOp::Max:
+ return OperationCode::AtomicIMax;
+ case AtomicOp::And:
+ return OperationCode::AtomicIAnd;
+ case AtomicOp::Or:
+ return OperationCode::AtomicIOr;
+ case AtomicOp::Xor:
+ return OperationCode::AtomicIXor;
+ case AtomicOp::Exch:
+ return OperationCode::AtomicIExchange;
+ }
+ }();
+ return SignedOperation(operation_code, is_signed, std::move(memory), std::move(data));
+}
+
bool IsUnaligned(Tegra::Shader::UniformType uniform_type) {
return uniform_type == Tegra::Shader::UniformType::UnsignedByte ||
uniform_type == Tegra::Shader::UniformType::UnsignedShort;
@@ -363,10 +384,13 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
break;
}
case OpCode::Id::ATOM: {
- UNIMPLEMENTED_IF_MSG(instr.atom.operation != GlobalAtomicOp::Add, "operation={}",
- static_cast<int>(instr.atom.operation.Value()));
- UNIMPLEMENTED_IF_MSG(instr.atom.type != GlobalAtomicType::S32, "type={}",
- static_cast<int>(instr.atom.type.Value()));
+ UNIMPLEMENTED_IF_MSG(instr.atom.operation == AtomicOp::Inc ||
+ instr.atom.operation == AtomicOp::Dec ||
+ instr.atom.operation == AtomicOp::SafeAdd,
+ "operation={}", static_cast<int>(instr.atom.operation.Value()));
+ UNIMPLEMENTED_IF_MSG(instr.atom.type == GlobalAtomicType::S64 ||
+ instr.atom.type == GlobalAtomicType::U64,
+ "type={}", static_cast<int>(instr.atom.type.Value()));
const auto [real_address, base_address, descriptor] =
TrackGlobalMemory(bb, instr, true, true);
@@ -375,25 +399,29 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
break;
}
+ const bool is_signed =
+ instr.atoms.type == AtomicType::S32 || instr.atoms.type == AtomicType::S64;
Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
- Node value = Operation(OperationCode::AtomicAdd, std::move(gmem), GetRegister(instr.gpr20));
+ Node value = GetAtomOperation(static_cast<AtomicOp>(instr.atom.operation), is_signed, gmem,
+ GetRegister(instr.gpr20));
SetRegister(bb, instr.gpr0, std::move(value));
break;
}
case OpCode::Id::ATOMS: {
- UNIMPLEMENTED_IF_MSG(instr.atoms.operation != AtomicOp::Add, "operation={}",
- static_cast<int>(instr.atoms.operation.Value()));
- UNIMPLEMENTED_IF_MSG(instr.atoms.type != AtomicType::U32, "type={}",
- static_cast<int>(instr.atoms.type.Value()));
-
+ UNIMPLEMENTED_IF_MSG(instr.atoms.operation == AtomicOp::Inc ||
+ instr.atoms.operation == AtomicOp::Dec,
+ "operation={}", static_cast<int>(instr.atoms.operation.Value()));
+ UNIMPLEMENTED_IF_MSG(instr.atoms.type == AtomicType::S64 ||
+ instr.atoms.type == AtomicType::U64,
+ "type={}", static_cast<int>(instr.atoms.type.Value()));
+ const bool is_signed =
+ instr.atoms.type == AtomicType::S32 || instr.atoms.type == AtomicType::S64;
const s32 offset = instr.atoms.GetImmediateOffset();
Node address = GetRegister(instr.gpr8);
address = Operation(OperationCode::IAdd, std::move(address), Immediate(offset));
-
- Node memory = GetSharedMemory(std::move(address));
- Node data = GetRegister(instr.gpr20);
-
- Node value = Operation(OperationCode::AtomicAdd, std::move(memory), std::move(data));
+ Node value =
+ GetAtomOperation(static_cast<AtomicOp>(instr.atoms.operation), is_signed,
+ GetSharedMemory(std::move(address)), GetRegister(instr.gpr20));
SetRegister(bb, instr.gpr0, std::move(value));
break;
}
diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp
index fbd7e9a17..6191ffba1 100644
--- a/src/video_core/shader/decode/xmad.cpp
+++ b/src/video_core/shader/decode/xmad.cpp
@@ -31,7 +31,7 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
const bool is_signed_b = instr.xmad.sign_b == 1;
const bool is_signed_c = is_signed_a;
- auto [is_merge, is_psl, is_high_b, mode, op_b,
+ auto [is_merge, is_psl, is_high_b, mode, op_b_binding,
op_c] = [&]() -> std::tuple<bool, bool, bool, Tegra::Shader::XmadMode, Node, Node> {
switch (opcode->get().GetId()) {
case OpCode::Id::XMAD_CR:
@@ -67,9 +67,10 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
op_a = SignedOperation(OperationCode::IBitfieldExtract, is_signed_a, std::move(op_a),
instr.xmad.high_a ? Immediate(16) : Immediate(0), Immediate(16));
- const Node original_b = op_b;
- op_b = SignedOperation(OperationCode::IBitfieldExtract, is_signed_b, std::move(op_b),
- is_high_b ? Immediate(16) : Immediate(0), Immediate(16));
+ const Node original_b = op_b_binding;
+ const Node op_b =
+ SignedOperation(OperationCode::IBitfieldExtract, is_signed_b, std::move(op_b_binding),
+ is_high_b ? Immediate(16) : Immediate(0), Immediate(16));
// we already check sign_a and sign_b is difference or not before so just use one in here.
Node product = SignedOperation(OperationCode::IMul, is_signed_a, op_a, op_b);
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index a1828546e..5fcc9da60 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -162,7 +162,21 @@ enum class OperationCode {
AtomicImageXor, /// (MetaImage, int[N] coords) -> void
AtomicImageExchange, /// (MetaImage, int[N] coords) -> void
- AtomicAdd, /// (memory, {u}int) -> {u}int
+ AtomicUExchange, /// (memory, uint) -> uint
+ AtomicUAdd, /// (memory, uint) -> uint
+ AtomicUMin, /// (memory, uint) -> uint
+ AtomicUMax, /// (memory, uint) -> uint
+ AtomicUAnd, /// (memory, uint) -> uint
+ AtomicUOr, /// (memory, uint) -> uint
+ AtomicUXor, /// (memory, uint) -> uint
+
+ AtomicIExchange, /// (memory, int) -> int
+ AtomicIAdd, /// (memory, int) -> int
+ AtomicIMin, /// (memory, int) -> int
+ AtomicIMax, /// (memory, int) -> int
+ AtomicIAnd, /// (memory, int) -> int
+ AtomicIOr, /// (memory, int) -> int
+ AtomicIXor, /// (memory, int) -> int
Branch, /// (uint branch_target) -> void
BranchIndirect, /// (uint branch_target) -> void
diff --git a/src/video_core/shader/node_helper.cpp b/src/video_core/shader/node_helper.cpp
index 76c56abb5..7bf4ff387 100644
--- a/src/video_core/shader/node_helper.cpp
+++ b/src/video_core/shader/node_helper.cpp
@@ -86,6 +86,20 @@ OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed)
return OperationCode::LogicalUNotEqual;
case OperationCode::LogicalIGreaterEqual:
return OperationCode::LogicalUGreaterEqual;
+ case OperationCode::AtomicIExchange:
+ return OperationCode::AtomicUExchange;
+ case OperationCode::AtomicIAdd:
+ return OperationCode::AtomicUAdd;
+ case OperationCode::AtomicIMin:
+ return OperationCode::AtomicUMin;
+ case OperationCode::AtomicIMax:
+ return OperationCode::AtomicUMax;
+ case OperationCode::AtomicIAnd:
+ return OperationCode::AtomicUAnd;
+ case OperationCode::AtomicIOr:
+ return OperationCode::AtomicUOr;
+ case OperationCode::AtomicIXor:
+ return OperationCode::AtomicUXor;
case OperationCode::INegate:
UNREACHABLE_MSG("Can't negate an unsigned integer");
return {};
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index 425927777..baf7188d2 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -96,6 +96,7 @@ Node ShaderIR::GetPredicate(bool immediate) {
}
Node ShaderIR::GetInputAttribute(Attribute::Index index, u64 element, Node buffer) {
+ MarkAttributeUsage(index, element);
used_input_attributes.emplace(index);
return MakeNode<AbufNode>(index, static_cast<u32>(element), std::move(buffer));
}
@@ -106,42 +107,8 @@ Node ShaderIR::GetPhysicalInputAttribute(Tegra::Shader::Register physical_addres
}
Node ShaderIR::GetOutputAttribute(Attribute::Index index, u64 element, Node buffer) {
- if (index == Attribute::Index::LayerViewportPointSize) {
- switch (element) {
- case 0:
- UNIMPLEMENTED();
- break;
- case 1:
- uses_layer = true;
- break;
- case 2:
- uses_viewport_index = true;
- break;
- case 3:
- uses_point_size = true;
- break;
- }
- }
- if (index == Attribute::Index::TessCoordInstanceIDVertexID) {
- switch (element) {
- case 2:
- uses_instance_id = true;
- break;
- case 3:
- uses_vertex_id = true;
- break;
- default:
- break;
- }
- }
- if (index == Attribute::Index::ClipDistances0123 ||
- index == Attribute::Index::ClipDistances4567) {
- const auto clip_index =
- static_cast<u32>((index == Attribute::Index::ClipDistances4567 ? 1 : 0) + element);
- used_clip_distances.at(clip_index) = true;
- }
+ MarkAttributeUsage(index, element);
used_output_attributes.insert(index);
-
return MakeNode<AbufNode>(index, static_cast<u32>(element), std::move(buffer));
}
@@ -452,6 +419,54 @@ Node ShaderIR::BitfieldInsert(Node base, Node insert, u32 offset, u32 bits) {
Immediate(bits));
}
+void ShaderIR::MarkAttributeUsage(Attribute::Index index, u64 element) {
+ switch (index) {
+ case Attribute::Index::LayerViewportPointSize:
+ switch (element) {
+ case 0:
+ UNIMPLEMENTED();
+ break;
+ case 1:
+ uses_layer = true;
+ break;
+ case 2:
+ uses_viewport_index = true;
+ break;
+ case 3:
+ uses_point_size = true;
+ break;
+ }
+ break;
+ case Attribute::Index::TessCoordInstanceIDVertexID:
+ switch (element) {
+ case 2:
+ uses_instance_id = true;
+ break;
+ case 3:
+ uses_vertex_id = true;
+ break;
+ }
+ break;
+ case Attribute::Index::ClipDistances0123:
+ case Attribute::Index::ClipDistances4567: {
+ const u64 clip_index = (index == Attribute::Index::ClipDistances4567 ? 4 : 0) + element;
+ used_clip_distances.at(clip_index) = true;
+ break;
+ }
+ case Attribute::Index::FrontColor:
+ case Attribute::Index::FrontSecondaryColor:
+ case Attribute::Index::BackColor:
+ case Attribute::Index::BackSecondaryColor:
+ uses_legacy_varyings = true;
+ break;
+ default:
+ if (index >= Attribute::Index::TexCoord_0 && index <= Attribute::Index::TexCoord_7) {
+ uses_legacy_varyings = true;
+ }
+ break;
+ }
+}
+
std::size_t ShaderIR::DeclareAmend(Node new_amend) {
const std::size_t id = amend_code.size();
amend_code.push_back(new_amend);
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index dde036b40..80fc9b82c 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -137,6 +137,10 @@ public:
return uses_vertex_id;
}
+ bool UsesLegacyVaryings() const {
+ return uses_legacy_varyings;
+ }
+
bool UsesWarps() const {
return uses_warps;
}
@@ -343,6 +347,9 @@ private:
/// Inserts a sequence of bits from a node
Node BitfieldInsert(Node base, Node insert, u32 offset, u32 bits);
+ /// Marks the usage of a input or output attribute.
+ void MarkAttributeUsage(Tegra::Shader::Attribute::Index index, u64 element);
+
void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
const Node4& components);
@@ -443,6 +450,7 @@ private:
bool uses_physical_attributes{}; // Shader uses AL2P or physical attribute read/writes
bool uses_instance_id{};
bool uses_vertex_id{};
+ bool uses_legacy_varyings{};
bool uses_warps{};
bool uses_indexed_samplers{};
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-09 14:09:22 +0200