Core: Add Vulkan GPU renderer

src/core/CMakeLists.txt

@@ -36,6 +36,8 @@ add_library(core
     gpu_hw_opengl.h
     gpu_hw_shadergen.cpp
     gpu_hw_shadergen.h
+    gpu_hw_vulkan.cpp
+    gpu_hw_vulkan.h
     gpu_sw.cpp
     gpu_sw.h
     gte.cpp
@@ -91,7 +93,7 @@ set(RECOMPILER_SRCS
 
 target_include_directories(core PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/..")
 target_include_directories(core PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}/..")
-target_link_libraries(core PUBLIC Threads::Threads common imgui tinyxml2 zlib)
+target_link_libraries(core PUBLIC Threads::Threads common imgui tinyxml2 zlib vulkan-loader)
 target_link_libraries(core PRIVATE glad stb)
 
 if(WIN32)

src/core/core.vcxproj

@@ -64,6 +64,7 @@
     <ClCompile Include="gpu_commands.cpp" />
     <ClCompile Include="gpu_hw_d3d11.cpp" />
     <ClCompile Include="gpu_hw_shadergen.cpp" />
+    <ClCompile Include="gpu_hw_vulkan.cpp" />
     <ClCompile Include="gpu_sw.cpp" />
     <ClCompile Include="gte.cpp" />
     <ClCompile Include="dma.cpp" />
@@ -105,6 +106,7 @@
     <ClInclude Include="game_list.h" />
     <ClInclude Include="gpu_hw_d3d11.h" />
     <ClInclude Include="gpu_hw_shadergen.h" />
+    <ClInclude Include="gpu_hw_vulkan.h" />
     <ClInclude Include="gpu_sw.h" />
     <ClInclude Include="gte.h" />
     <ClInclude Include="cpu_types.h" />
@@ -296,7 +298,7 @@
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_CRT_SECURE_NO_WARNINGS;WIN32;_DEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <SDLCheck>true</SDLCheck>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <MinimalRebuild>false</MinimalRebuild>
       <LanguageStandard>stdcpp17</LanguageStandard>
@@ -321,7 +323,7 @@
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_CRT_SECURE_NO_WARNINGS;WIN32;_DEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <SDLCheck>true</SDLCheck>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <MinimalRebuild>false</MinimalRebuild>
       <LanguageStandard>stdcpp17</LanguageStandard>
@@ -346,7 +348,7 @@
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_ITERATOR_DEBUG_LEVEL=1;_CRT_SECURE_NO_WARNINGS;WIN32;_DEBUGFAST;_DEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <SDLCheck>true</SDLCheck>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <BasicRuntimeChecks>Default</BasicRuntimeChecks>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <MinimalRebuild>false</MinimalRebuild>
@@ -374,7 +376,7 @@
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_ITERATOR_DEBUG_LEVEL=1;_CRT_SECURE_NO_WARNINGS;WIN32;_DEBUGFAST;_DEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <SDLCheck>true</SDLCheck>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <BasicRuntimeChecks>Default</BasicRuntimeChecks>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <MinimalRebuild>false</MinimalRebuild>
@@ -401,7 +403,7 @@
       <Optimization>MaxSpeed</Optimization>
       <IntrinsicFunctions>true</IntrinsicFunctions>
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_CRT_SECURE_NO_WARNINGS;WIN32;NDEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <WholeProgramOptimization>false</WholeProgramOptimization>
       <LanguageStandard>stdcpp17</LanguageStandard>
@@ -427,7 +429,7 @@
       <Optimization>MaxSpeed</Optimization>
       <IntrinsicFunctions>true</IntrinsicFunctions>
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_CRT_SECURE_NO_WARNINGS;WIN32;NDEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <WholeProgramOptimization>true</WholeProgramOptimization>
       <LanguageStandard>stdcpp17</LanguageStandard>
@@ -454,7 +456,7 @@
       <Optimization>MaxSpeed</Optimization>
       <IntrinsicFunctions>true</IntrinsicFunctions>
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_CRT_SECURE_NO_WARNINGS;WIN32;NDEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <WholeProgramOptimization>false</WholeProgramOptimization>
       <LanguageStandard>stdcpp17</LanguageStandard>
@@ -480,7 +482,7 @@
       <Optimization>MaxSpeed</Optimization>
       <IntrinsicFunctions>true</IntrinsicFunctions>
       <PreprocessorDefinitions>WITH_RECOMPILER=1;_CRT_SECURE_NO_WARNINGS;WIN32;NDEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
-      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(SolutionDir)dep\msvc\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xbyak\xbyak;$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\vulkan-loader\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
       <WholeProgramOptimization>true</WholeProgramOptimization>
       <LanguageStandard>stdcpp17</LanguageStandard>

src/core/core.vcxproj.filters

@@ -44,6 +44,7 @@
     <ClCompile Include="namco_guncon.cpp" />
     <ClCompile Include="playstation_mouse.cpp" />
     <ClCompile Include="negcon.cpp" />
+    <ClCompile Include="gpu_hw_vulkan.cpp" />
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="types.h" />
@@ -89,6 +90,7 @@
     <ClInclude Include="namco_guncon.h" />
     <ClInclude Include="playstation_mouse.h" />
     <ClInclude Include="negcon.h" />
+    <ClInclude Include="gpu_hw_vulkan.h" />
   </ItemGroup>
   <ItemGroup>
     <None Include="cpu_core.inl" />

src/core/gpu.h

@@ -164,6 +164,9 @@ public:
   // gpu_hw_opengl.cpp
   static std::unique_ptr<GPU> CreateHardwareOpenGLRenderer();
 
+  // gpu_hw_vulkan.cpp
+  static std::unique_ptr<GPU> CreateHardwareVulkanRenderer();
+
   // gpu_sw.cpp
   static std::unique_ptr<GPU> CreateSoftwareRenderer();
 
@@ -602,7 +605,7 @@ protected:
     void ClearTextureWindowChangedFlag() { texture_window_changed = false; }
   } m_draw_mode = {};
 
-  Common::Rectangle<u32> m_drawing_area;
+  Common::Rectangle<u32> m_drawing_area{0, 0, VRAM_WIDTH, VRAM_HEIGHT};
 
   struct DrawingOffset
   {

src/core/gpu_hw.cpp

@@ -527,6 +527,36 @@ bool GPU_HW::UseVRAMCopyShader(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 w
             .Intersects(Common::Rectangle<u32>::FromExtents(dst_x, dst_y, width, height)));
 }
 
+GPU_HW::VRAMFillUBOData GPU_HW::GetVRAMFillUBOData(u32 x, u32 y, u32 width, u32 height, u32 color) const
+{
+  // drop precision unless true colour is enabled
+  if (!m_true_color)
+    color = RGBA5551ToRGBA8888(RGBA8888ToRGBA5551(color));
+
+  VRAMFillUBOData uniforms;
+  std::tie(uniforms.u_fill_color[0], uniforms.u_fill_color[1], uniforms.u_fill_color[2], uniforms.u_fill_color[3]) =
+    RGBA8ToFloat(color);
+  uniforms.u_interlaced_displayed_field = GetActiveLineLSB();
+  return uniforms;
+}
+
+GPU_HW::VRAMCopyUBOData GPU_HW::GetVRAMCopyUBOData(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width,
+                                                   u32 height) const
+{
+  const VRAMCopyUBOData uniforms = {src_x * m_resolution_scale,
+                                    src_y * m_resolution_scale,
+                                    dst_x * m_resolution_scale,
+                                    dst_y * m_resolution_scale,
+                                    ((dst_x + width) % VRAM_WIDTH) * m_resolution_scale,
+                                    ((dst_y + height) % VRAM_HEIGHT) * m_resolution_scale,
+                                    width * m_resolution_scale,
+                                    height * m_resolution_scale,
+                                    m_GPUSTAT.set_mask_while_drawing ? 1u : 0u,
+                                    GetCurrentNormalizedBatchVertexDepthID()};
+
+  return uniforms;
+}
+
 GPU_HW::BatchPrimitive GPU_HW::GetPrimitiveForCommand(RenderCommand rc)
 {
   if (rc.primitive == Primitive::Line)

src/core/gpu_hw.h

@@ -14,9 +14,7 @@ public:
   enum class BatchPrimitive : u8
   {
     Lines = 0,
-    LineStrip = 1,
-    Triangles = 2,
-    TriangleStrip = 3
+    Triangles = 1
   };
 
   enum class BatchRenderMode : u8
@@ -119,6 +117,12 @@ protected:
     u32 u_set_mask_while_drawing;
   };
 
+  struct VRAMFillUBOData
+  {
+    float u_fill_color[4];
+    u32 u_interlaced_displayed_field;
+  };
+
   struct VRAMWriteUBOData
   {
     u32 u_base_coords[2];
@@ -230,6 +234,9 @@ protected:
   /// Returns true if the VRAM copy shader should be used (oversized copies, masking).
   bool UseVRAMCopyShader(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height) const;
 
+  VRAMFillUBOData GetVRAMFillUBOData(u32 x, u32 y, u32 width, u32 height, u32 color) const;
+  VRAMCopyUBOData GetVRAMCopyUBOData(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height) const;
+
   /// Handles quads with flipped texture coordinate directions.
   static void HandleFlippedQuadTextureCoordinates(BatchVertex* vertices);
 

src/core/gpu_hw_d3d11.cpp

@@ -534,9 +534,8 @@ void GPU_HW_D3D11::DrawBatchVertices(BatchRenderMode render_mode, u32 base_verte
 {
   const bool textured = (m_batch.texture_mode != TextureMode::Disabled);
 
-  static constexpr std::array<D3D11_PRIMITIVE_TOPOLOGY, 4> d3d_primitives = {
-    {D3D11_PRIMITIVE_TOPOLOGY_LINELIST, D3D11_PRIMITIVE_TOPOLOGY_LINESTRIP, D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST,
-     D3D11_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP}};
+  static constexpr std::array<D3D11_PRIMITIVE_TOPOLOGY, 2> d3d_primitives = {
+    {D3D11_PRIMITIVE_TOPOLOGY_LINELIST, D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST}};
   m_context->IASetPrimitiveTopology(d3d_primitives[static_cast<u8>(m_batch.primitive)]);
 
   m_context->VSSetShader(m_batch_vertex_shaders[BoolToUInt8(textured)].Get(), nullptr, 0);

src/core/gpu_hw_opengl.cpp

@@ -535,7 +535,7 @@ void GPU_HW_OpenGL::DrawBatchVertices(BatchRenderMode render_mode, u32 base_vert
 
   glDepthFunc(m_GPUSTAT.check_mask_before_draw ? GL_GEQUAL : GL_ALWAYS);
 
-  static constexpr std::array<GLenum, 4> gl_primitives = {{GL_LINES, GL_LINE_STRIP, GL_TRIANGLES, GL_TRIANGLE_STRIP}};
+  static constexpr std::array<GLenum, 2> gl_primitives = {{GL_LINES, GL_TRIANGLES}};
   glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, num_vertices);
 }
 

src/core/gpu_hw_shadergen.cpp

@@ -14,10 +14,11 @@ GPU_HW_ShaderGen::GPU_HW_ShaderGen(HostDisplay::RenderAPI render_api, u32 resolu
 {
   if (m_glsl)
   {
-    SetGLSLVersionString();
+    if (m_render_api == HostDisplay::RenderAPI::OpenGL || m_render_api == HostDisplay::RenderAPI::OpenGLES)
+      SetGLSLVersionString();
 
-    m_use_glsl_interface_blocks = (GLAD_GL_ES_VERSION_3_2 || GLAD_GL_VERSION_3_2);
-    m_use_glsl_binding_layout = UseGLSLBindingLayout();
+    m_use_glsl_interface_blocks = (IsVulkan() || GLAD_GL_ES_VERSION_3_2 || GLAD_GL_VERSION_3_2);
+    m_use_glsl_binding_layout = (IsVulkan() || UseGLSLBindingLayout());
   }
 }
 
@@ -82,6 +83,8 @@ void GPU_HW_ShaderGen::WriteHeader(std::stringstream& ss)
 {
   if (m_render_api == HostDisplay::RenderAPI::OpenGL || m_render_api == HostDisplay::RenderAPI::OpenGLES)
     ss << m_glsl_version_string << "\n\n";
+  else if (m_render_api == HostDisplay::RenderAPI::Vulkan)
+    ss << "#version 450 core\n\n";
 
   // Extension enabling for OpenGL.
   if (m_render_api == HostDisplay::RenderAPI::OpenGLES)
@@ -107,6 +110,7 @@ void GPU_HW_ShaderGen::WriteHeader(std::stringstream& ss)
   DefineMacro(ss, "API_OPENGL", m_render_api == HostDisplay::RenderAPI::OpenGL);
   DefineMacro(ss, "API_OPENGL_ES", m_render_api == HostDisplay::RenderAPI::OpenGLES);
   DefineMacro(ss, "API_D3D11", m_render_api == HostDisplay::RenderAPI::D3D11);
+  DefineMacro(ss, "API_VULKAN", m_render_api == HostDisplay::RenderAPI::Vulkan);
 
   if (m_render_api == HostDisplay::RenderAPI::OpenGLES)
   {
@@ -219,9 +223,17 @@ float4 RGBA5551ToRGBA8(uint v)
 )";
 }
 
-void GPU_HW_ShaderGen::DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members)
+void GPU_HW_ShaderGen::DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members,
+                                            bool push_constant_on_vulkan)
 {
-  if (m_glsl)
+  if (IsVulkan())
+  {
+    if (push_constant_on_vulkan)
+      ss << "layout(push_constant) uniform PushConstants\n";
+    else
+      ss << "layout(std140, set = 0, binding = 0) uniform UBOBlock\n";
+  }
+  else if (m_glsl)
   {
     if (m_use_glsl_binding_layout)
       ss << "layout(std140, binding = 1) uniform UBOBlock\n";
@@ -243,7 +255,9 @@ void GPU_HW_ShaderGen::DeclareTexture(std::stringstream& ss, const char* name, u
 {
   if (m_glsl)
   {
-    if (m_use_glsl_binding_layout)
+    if (IsVulkan())
+      ss << "layout(set = 0, binding = " << (index + 1u) << ") ";
+    else if (m_use_glsl_binding_layout)
       ss << "layout(binding = " << index << ") ";
 
     ss << "uniform sampler2D " << name << ";\n";
@@ -260,7 +274,9 @@ void GPU_HW_ShaderGen::DeclareTextureBuffer(std::stringstream& ss, const char* n
 {
   if (m_glsl)
   {
-    if (m_use_glsl_binding_layout)
+    if (IsVulkan())
+      ss << "layout(set = 0, binding = " << index << ") ";
+    else if (m_use_glsl_binding_layout)
       ss << "layout(binding = " << index << ") ";
 
     ss << "uniform " << (is_int ? (is_unsigned ? "u" : "i") : "") << "samplerBuffer " << name << ";\n";
@@ -296,6 +312,9 @@ void GPU_HW_ShaderGen::DeclareVertexEntryPoint(
 
     if (m_use_glsl_interface_blocks)
     {
+      if (IsVulkan())
+        ss << "layout(location = 0) ";
+
       ss << "out VertexData {\n";
       for (u32 i = 0; i < num_color_outputs; i++)
         ss << "  float4 v_col" << i << ";\n";
@@ -321,7 +340,12 @@ void GPU_HW_ShaderGen::DeclareVertexEntryPoint(
 
     ss << "#define v_pos gl_Position\n\n";
     if (declare_vertex_id)
-      ss << "#define v_id uint(gl_VertexID)\n";
+    {
+      if (IsVulkan())
+        ss << "#define v_id uint(gl_VertexIndex)\n";
+      else
+        ss << "#define v_id uint(gl_VertexID)\n";
+    }
 
     ss << "\n";
     ss << "void main()\n";
@@ -366,6 +390,9 @@ void GPU_HW_ShaderGen::DeclareFragmentEntryPoint(
   {
     if (m_use_glsl_interface_blocks)
     {
+      if (IsVulkan())
+        ss << "layout(location = 0) ";
+
       ss << "in VertexData {\n";
       for (u32 i = 0; i < num_color_inputs; i++)
         ss << "  float4 v_col" << i << ";\n";
@@ -467,7 +494,8 @@ void GPU_HW_ShaderGen::WriteBatchUniformBuffer(std::stringstream& ss)
   DeclareUniformBuffer(ss,
                        {"uint2 u_texture_window_mask", "uint2 u_texture_window_offset", "float u_src_alpha_factor",
                         "float u_dst_alpha_factor", "uint u_interlaced_displayed_field", "uint u_base_vertex_depth_id",
-                        "bool u_check_mask_before_draw", "bool u_set_mask_while_drawing"});
+                        "bool u_check_mask_before_draw", "bool u_set_mask_while_drawing"},
+                       false);
 }
 
 std::string GPU_HW_ShaderGen::GenerateBatchVertexShader(bool textured)
@@ -507,6 +535,12 @@ std::string GPU_HW_ShaderGen::GenerateBatchVertexShader(bool textured)
 #if API_OPENGL || API_OPENGL_ES
   pos_y += EPSILON;
 #endif
+
+  // NDC space Y flip in Vulkan.
+#if API_VULKAN
+  pos_y = -pos_y;
+#endif
+
   v_pos = float4(pos_x, pos_y, 0.0, 1.0);
 
 #if API_D3D11
@@ -861,13 +895,18 @@ CONSTANT float2 WIDTH = (1.0 / float2(VRAM_SIZE)) * float2(RESOLUTION_SCALE, RES
   // GS is a pain, too different between HLSL and GLSL...
   if (m_glsl)
   {
-    ss << R"(
-in VertexData {
+    if (IsVulkan())
+      ss << "layout(location = 0) ";
+
+    ss << R"(in VertexData {
   float4 v_col0;
   nointerpolation float v_depth;
-} in_data[];
+} in_data[];)";
 
-out VertexData {
+    if (IsVulkan())
+      ss << "layout(location = 0) ";
+
+    ss << R"(out VertexData {
   float4 v_col0;
   nointerpolation float v_depth;
 } out_data;
@@ -968,8 +1007,8 @@ std::string GPU_HW_ShaderGen::GenerateScreenQuadVertexShader()
 {
   v_tex0 = float2(float((v_id << 1) & 2u), float(v_id & 2u));
   v_pos = float4(v_tex0 * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);
-  #if API_OPENGL || API_OPENGL_ES
-    v_pos.y = -gl_Position.y;
+  #if API_OPENGL || API_OPENGL_ES || API_VULKAN
+    v_pos.y = -v_pos.y;
   #endif
 }
 )";
@@ -981,7 +1020,7 @@ std::string GPU_HW_ShaderGen::GenerateFillFragmentShader()
 {
   std::stringstream ss;
   WriteHeader(ss);
-  DeclareUniformBuffer(ss, {"float4 u_fill_color"});
+  DeclareUniformBuffer(ss, {"float4 u_fill_color"}, true);
   DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1, true);
 
   ss << R"(
@@ -999,7 +1038,7 @@ std::string GPU_HW_ShaderGen::GenerateInterlacedFillFragmentShader()
   std::stringstream ss;
   WriteHeader(ss);
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(ss, {"float4 u_fill_color", "uint u_interlaced_displayed_field"});
+  DeclareUniformBuffer(ss, {"float4 u_fill_color", "uint u_interlaced_displayed_field"}, true);
   DeclareFragmentEntryPoint(ss, 0, 1, {}, true, 1, true);
 
   ss << R"(
@@ -1019,7 +1058,7 @@ std::string GPU_HW_ShaderGen::GenerateCopyFragmentShader()
 {
   std::stringstream ss;
   WriteHeader(ss);
-  DeclareUniformBuffer(ss, {"float4 u_src_rect"});
+  DeclareUniformBuffer(ss, {"float4 u_src_rect"}, true);
   DeclareTexture(ss, "samp0", 0);
   DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1);
 
@@ -1043,7 +1082,7 @@ std::string GPU_HW_ShaderGen::GenerateDisplayFragmentShader(bool depth_24bit,
   DefineMacro(ss, "INTERLEAVED", interlace_mode == GPU_HW::InterlacedRenderMode::InterleavedFields);
 
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(ss, {"uint2 u_vram_offset", "uint u_crop_left", "uint u_field_offset"});
+  DeclareUniformBuffer(ss, {"uint2 u_vram_offset", "uint u_crop_left", "uint u_field_offset"}, true);
   DeclareTexture(ss, "samp0", 0);
 
   DeclareFragmentEntryPoint(ss, 0, 1, {}, true, 1);
@@ -1093,7 +1132,7 @@ std::string GPU_HW_ShaderGen::GenerateVRAMReadFragmentShader()
   std::stringstream ss;
   WriteHeader(ss);
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(ss, {"uint2 u_base_coords", "uint2 u_size"});
+  DeclareUniformBuffer(ss, {"uint2 u_base_coords", "uint2 u_size"}, true);
 
   DeclareTexture(ss, "samp0", 0);
 
@@ -1146,8 +1185,10 @@ std::string GPU_HW_ShaderGen::GenerateVRAMWriteFragmentShader()
   std::stringstream ss;
   WriteHeader(ss);
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(ss, {"uint2 u_base_coords", "uint2 u_size", "uint u_buffer_base_offset", "uint u_mask_or_bits",
-                            "float u_depth_value"});
+  DeclareUniformBuffer(
+    ss,
+    {"uint2 u_base_coords", "uint2 u_size", "uint u_buffer_base_offset", "uint u_mask_or_bits", "float u_depth_value"},
+    true);
 
   DeclareTextureBuffer(ss, "samp0", 0, true, true);
   DeclareFragmentEntryPoint(ss, 0, 1, {}, true, 1, true);
@@ -1176,8 +1217,10 @@ std::string GPU_HW_ShaderGen::GenerateVRAMCopyFragmentShader()
   std::stringstream ss;
   WriteHeader(ss);
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(ss, {"uint2 u_src_coords", "uint2 u_dst_coords", "uint2 u_end_coords", "uint2 u_size",
-                            "bool u_set_mask_bit", "float u_depth_value"});
+  DeclareUniformBuffer(ss,
+                       {"uint2 u_src_coords", "uint2 u_dst_coords", "uint2 u_end_coords", "uint2 u_size",
+                        "bool u_set_mask_bit", "float u_depth_value"},
+                       true);
 
   DeclareTexture(ss, "samp0", 0);
   DeclareFragmentEntryPoint(ss, 0, 1, {}, true, 1, true);

src/core/gpu_hw_shadergen.h

@@ -28,9 +28,12 @@ public:
   std::string GenerateVRAMUpdateDepthFragmentShader();
 
 private:
+  ALWAYS_INLINE bool IsVulkan() const { return (m_render_api == HostDisplay::RenderAPI::Vulkan); }
+
   void SetGLSLVersionString();
   void WriteHeader(std::stringstream& ss);
-  void DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members);
+  void DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members,
+                            bool push_constant_on_vulkan);
   void DeclareTexture(std::stringstream& ss, const char* name, u32 index);
   void DeclareTextureBuffer(std::stringstream& ss, const char* name, u32 index, bool is_int, bool is_unsigned);
   void DeclareVertexEntryPoint(std::stringstream& ss, const std::initializer_list<const char*>& attributes,

src/core/gpu_hw_vulkan.h

@@ -0,0 +1,124 @@
+#pragma once
+#include "common/dimensional_array.h"
+#include "common/vulkan/staging_texture.h"
+#include "common/vulkan/stream_buffer.h"
+#include "common/vulkan/texture.h"
+#include "gpu_hw.h"
+#include <array>
+#include <memory>
+#include <tuple>
+
+class GPU_HW_Vulkan : public GPU_HW
+{
+public:
+  GPU_HW_Vulkan();
+  ~GPU_HW_Vulkan() override;
+
+  bool Initialize(HostDisplay* host_display, System* system, DMA* dma, InterruptController* interrupt_controller,
+                  Timers* timers) override;
+  void Reset() override;
+
+  void ResetGraphicsAPIState() override;
+  void RestoreGraphicsAPIState() override;
+  void UpdateSettings() override;
+
+protected:
+  void UpdateDisplay() override;
+  void ReadVRAM(u32 x, u32 y, u32 width, u32 height) override;
+  void FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color) override;
+  void UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data) override;
+  void CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height) override;
+  void UpdateVRAMReadTexture() override;
+  void UpdateDepthBufferFromMaskBit() override;
+  void SetScissorFromDrawingArea() override;
+  void MapBatchVertexPointer(u32 required_vertices) override;
+  void UnmapBatchVertexPointer(u32 used_vertices) override;
+  void UploadUniformBuffer(const void* data, u32 data_size) override;
+  void DrawBatchVertices(BatchRenderMode render_mode, u32 base_vertex, u32 num_vertices) override;
+
+private:
+  enum : u32
+  {
+    MAX_PUSH_CONSTANTS_SIZE = 64,
+  };
+  void SetCapabilities();
+  void DestroyResources();
+
+  ALWAYS_INLINE bool InRenderPass() const { return (m_current_render_pass != VK_NULL_HANDLE); }
+  void BeginRenderPass(VkRenderPass render_pass, VkFramebuffer framebuffer, u32 x, u32 y, u32 width, u32 height);
+  void BeginVRAMRenderPass();
+  void EndRenderPass();
+
+  bool CreatePipelineLayouts();
+  bool CreateSamplers();
+
+  bool CreateFramebuffer();
+  void ClearFramebuffer();
+  void DestroyFramebuffer();
+
+  bool CreateVertexBuffer();
+  bool CreateUniformBuffer();
+  bool CreateTextureBuffer();
+
+  bool CompilePipelines();
+  void DestroyPipelines();
+
+  VkRenderPass m_current_render_pass = VK_NULL_HANDLE;
+
+  VkRenderPass m_vram_render_pass = VK_NULL_HANDLE;
+  VkRenderPass m_vram_update_depth_render_pass = VK_NULL_HANDLE;
+  VkRenderPass m_display_render_pass = VK_NULL_HANDLE;
+  VkRenderPass m_vram_readback_render_pass = VK_NULL_HANDLE;
+
+  VkDescriptorSetLayout m_batch_descriptor_set_layout = VK_NULL_HANDLE;
+  VkDescriptorSetLayout m_single_sampler_descriptor_set_layout = VK_NULL_HANDLE;
+  VkDescriptorSetLayout m_vram_write_descriptor_set_layout = VK_NULL_HANDLE;
+
+  VkPipelineLayout m_batch_pipeline_layout = VK_NULL_HANDLE;
+  VkPipelineLayout m_no_samplers_pipeline_layout = VK_NULL_HANDLE;
+  VkPipelineLayout m_single_sampler_pipeline_layout = VK_NULL_HANDLE;
+  VkPipelineLayout m_vram_write_pipeline_layout = VK_NULL_HANDLE;
+
+  Vulkan::Texture m_vram_texture;
+  Vulkan::Texture m_vram_depth_texture;
+  Vulkan::Texture m_vram_read_texture;
+  Vulkan::Texture m_vram_readback_texture;
+  Vulkan::StagingTexture m_vram_readback_staging_texture;
+  Vulkan::Texture m_display_texture;
+
+  VkFramebuffer m_vram_framebuffer = VK_NULL_HANDLE;
+  VkFramebuffer m_vram_update_depth_framebuffer = VK_NULL_HANDLE;
+  VkFramebuffer m_vram_readback_framebuffer = VK_NULL_HANDLE;
+  VkFramebuffer m_display_framebuffer = VK_NULL_HANDLE;
+
+  VkSampler m_point_sampler = VK_NULL_HANDLE;
+  VkSampler m_linear_sampler = VK_NULL_HANDLE;
+
+  VkDescriptorSet m_batch_descriptor_set = VK_NULL_HANDLE;
+  VkDescriptorSet m_vram_copy_descriptor_set = VK_NULL_HANDLE;
+  VkDescriptorSet m_vram_read_descriptor_set = VK_NULL_HANDLE;
+  VkDescriptorSet m_vram_write_descriptor_set = VK_NULL_HANDLE;
+
+  Vulkan::StreamBuffer m_vertex_stream_buffer;
+  Vulkan::StreamBuffer m_uniform_stream_buffer;
+  Vulkan::StreamBuffer m_texture_stream_buffer;
+
+  u32 m_current_uniform_buffer_offset = 0;
+  VkBufferView m_texture_stream_buffer_view = VK_NULL_HANDLE;
+
+  // [primitive][depth_test][render_mode][texture_mode][transparency_mode][dithering][interlacing]
+  DimensionalArray<VkPipeline, 2, 2, 5, 9, 4, 2, 2> m_batch_pipelines{};
+
+  // [interlaced]
+  std::array<VkPipeline, 2> m_vram_fill_pipelines{};
+
+  // [depth_test]
+  std::array<VkPipeline, 2> m_vram_write_pipelines{};
+  std::array<VkPipeline, 2> m_vram_copy_pipelines{};
+
+  VkPipeline m_vram_readback_pipeline = VK_NULL_HANDLE;
+  VkPipeline m_vram_update_depth_pipeline = VK_NULL_HANDLE;
+
+  // [depth_24][interlace_mode]
+  DimensionalArray<VkPipeline, 3, 2> m_display_pipelines{};
+};

src/core/host_interface.cpp

@@ -314,6 +314,11 @@ void HostInterface::OnRunningGameChanged() {}
 
 void HostInterface::OnControllerTypeChanged(u32 slot) {}
 
+std::string HostInterface::GetShaderCacheDirectory()
+{
+  return GetUserDirectoryRelativePath("cache");
+}
+
 void HostInterface::SetDefaultSettings(SettingsInterface& si)
 {
   si.SetStringValue("Console", "Region", Settings::GetConsoleRegionName(ConsoleRegion::Auto));

src/core/host_interface.h

@@ -114,6 +114,9 @@ protected:
   virtual void OnRunningGameChanged();
   virtual void OnControllerTypeChanged(u32 slot);
 
+  /// Returns the path to the shader cache directory.
+  virtual std::string GetShaderCacheDirectory();
+
   /// Restores all settings to defaults.
   virtual void SetDefaultSettings(SettingsInterface& si);
 

src/core/system.cpp

@@ -288,6 +288,10 @@ bool System::CreateGPU(GPURenderer renderer)
       m_gpu = GPU::CreateHardwareOpenGLRenderer();
       break;
 
+    case GPURenderer::HardwareVulkan:
+      m_gpu = GPU::CreateHardwareVulkanRenderer();
+      break;
+
 #ifdef WIN32
     case GPURenderer::HardwareD3D11:
       m_gpu = GPU::CreateHardwareD3D11Renderer();