early-access version 3912

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 3911.
+This is the source code for early-access 3912.
 
 ## Legal Notice
 

src/video_core/host_shaders/convert_d32f_to_abgr8.frag

@@ -10,5 +10,5 @@ layout(location = 0) out vec4 color;
 void main() {
     ivec2 coord = ivec2(gl_FragCoord.xy);
     float depth = textureLod(depth_tex, coord, 0).r;
-    color = vec4(depth, depth*depth, sqrt(depth), 1.0); // Convert depth to color
+    color = vec4(depth, depth, depth, 1.0);
 }

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -116,6 +116,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> FORMAT_TAB
     {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV},                // E5B9G9R9_FLOAT
     {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT},            // D32_FLOAT
     {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT},    // D16_UNORM
+    {GL_DEPTH_COMPONENT24, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT_24_8}, // X8_D24_UNORM
     {GL_STENCIL_INDEX8, GL_STENCIL, GL_UNSIGNED_BYTE},                // S8_UINT
     {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8},    // D24_UNORM_S8_UINT
     {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8},    // S8_UINT_D24_UNORM

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -216,6 +216,7 @@ struct FormatTuple {
     // Depth formats
     {VK_FORMAT_D32_SFLOAT, Attachable},          // D32_FLOAT
     {VK_FORMAT_D16_UNORM, Attachable},           // D16_UNORM
+    {VK_FORMAT_X8_D24_UNORM_PACK32, Attachable}, // X8_D24_UNORM
 
     // Stencil formats
     {VK_FORMAT_S8_UINT, Attachable}, // S8_UINT

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -96,6 +96,7 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) {
 VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) {
     switch (framebuffer.pixel_format) {
     case Service::android::PixelFormat::Rgba8888:
+    case Service::android::PixelFormat::Rgbx8888:
         return VK_FORMAT_A8B8G8R8_UNORM_PACK32;
     case Service::android::PixelFormat::Rgb565:
         return VK_FORMAT_R5G6B5_UNORM_PACK16;

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -238,6 +238,7 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
         return any_r ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT;
     case PixelFormat::D16_UNORM:
     case PixelFormat::D32_FLOAT:
+    case PixelFormat::X8_D24_UNORM:
         return VK_IMAGE_ASPECT_DEPTH_BIT;
     case PixelFormat::S8_UINT:
         return VK_IMAGE_ASPECT_STENCIL_BIT;

src/video_core/surface.cpp

@@ -85,6 +85,8 @@ PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
         return PixelFormat::S8_UINT;
     case Tegra::DepthFormat::Z32_FLOAT_X24S8_UINT:
         return PixelFormat::D32_FLOAT_S8_UINT;
+    case Tegra::DepthFormat::X8Z24_UNORM:
+        return PixelFormat::X8_D24_UNORM;
     default:
         UNIMPLEMENTED_MSG("Unimplemented format={}", format);
         return PixelFormat::S8_UINT_D24_UNORM;
@@ -202,6 +204,7 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
 PixelFormat PixelFormatFromGPUPixelFormat(Service::android::PixelFormat format) {
     switch (format) {
     case Service::android::PixelFormat::Rgba8888:
+    case Service::android::PixelFormat::Rgbx8888:
         return PixelFormat::A8B8G8R8_UNORM;
     case Service::android::PixelFormat::Rgb565:
         return PixelFormat::R5G6B5_UNORM;

src/video_core/surface.h

@@ -115,6 +115,7 @@ enum class PixelFormat {
     // Depth formats
     D32_FLOAT = MaxColorFormat,
     D16_UNORM,
+    X8_D24_UNORM,
 
     MaxDepthFormat,
 
@@ -251,6 +252,7 @@ constexpr std::array<u8, MaxPixelFormat> BLOCK_WIDTH_TABLE = {{
     1,  // E5B9G9R9_FLOAT
     1,  // D32_FLOAT
     1,  // D16_UNORM
+    1,  // X8_D24_UNORM
     1,  // S8_UINT
     1,  // D24_UNORM_S8_UINT
     1,  // S8_UINT_D24_UNORM
@@ -360,6 +362,7 @@ constexpr std::array<u8, MaxPixelFormat> BLOCK_HEIGHT_TABLE = {{
     1,  // E5B9G9R9_FLOAT
     1,  // D32_FLOAT
     1,  // D16_UNORM
+    1,  // X8_D24_UNORM
     1,  // S8_UINT
     1,  // D24_UNORM_S8_UINT
     1,  // S8_UINT_D24_UNORM
@@ -469,6 +472,7 @@ constexpr std::array<u8, MaxPixelFormat> BITS_PER_BLOCK_TABLE = {{
     32,  // E5B9G9R9_FLOAT
     32,  // D32_FLOAT
     16,  // D16_UNORM
+    32,  // X8_D24_UNORM
     8,   // S8_UINT
     32,  // D24_UNORM_S8_UINT
     32,  // S8_UINT_D24_UNORM

src/video_core/texture_cache/format_lookup_table.cpp

@@ -142,6 +142,10 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red,
         return PixelFormat::D16_UNORM;
     case Hash(TextureFormat::Z16, UNORM, UINT, UINT, UINT, LINEAR):
         return PixelFormat::D16_UNORM;
+    case Hash(TextureFormat::X8Z24, UNORM):
+        return PixelFormat::X8_D24_UNORM;
+    case Hash(TextureFormat::X8Z24, UNORM, UINT, UINT, UINT, LINEAR):
+        return PixelFormat::X8_D24_UNORM;
     case Hash(TextureFormat::Z24S8, UINT, UNORM, UNORM, UNORM, LINEAR):
         return PixelFormat::S8_UINT_D24_UNORM;
     case Hash(TextureFormat::Z24S8, UINT, UNORM, UINT, UINT, LINEAR):

src/video_core/texture_cache/formatter.h

@@ -211,6 +211,8 @@ struct fmt::formatter<VideoCore::Surface::PixelFormat> : fmt::formatter<fmt::str
                 return "D32_FLOAT";
             case PixelFormat::D16_UNORM:
                 return "D16_UNORM";
+            case PixelFormat::X8_D24_UNORM:
+                return "X8_D24_UNORM";
             case PixelFormat::S8_UINT:
                 return "S8_UINT";
             case PixelFormat::D24_UNORM_S8_UINT:

src/video_core/texture_cache/image_view_base.cpp

@@ -85,6 +85,7 @@ bool ImageViewBase::SupportsAnisotropy() const noexcept {
     // Depth formats
     case PixelFormat::D32_FLOAT:
     case PixelFormat::D16_UNORM:
+    case PixelFormat::X8_D24_UNORM:
     // Stencil formats
     case PixelFormat::S8_UINT:
     // DepthStencil formats

src/video_core/vulkan_common/vulkan_device.cpp

@@ -84,9 +84,12 @@ constexpr std::array VK_FORMAT_A4B4G4R4_UNORM_PACK16{
 } // namespace Alternatives
 
 enum class NvidiaArchitecture {
-    AmpereOrNewer,
+    KeplerOrOlder,
+    Maxwell,
+    Pascal,
+    Volta,
     Turing,
-    VoltaOrOlder,
+    AmpereOrNewer,
 };
 
 template <typename T>
@@ -200,6 +203,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(vk::Physica
         VK_FORMAT_BC7_UNORM_BLOCK,
         VK_FORMAT_D16_UNORM,
         VK_FORMAT_D16_UNORM_S8_UINT,
+        VK_FORMAT_X8_D24_UNORM_PACK32,
         VK_FORMAT_D24_UNORM_S8_UINT,
         VK_FORMAT_D32_SFLOAT,
         VK_FORMAT_D32_SFLOAT_S8_UINT,
@@ -321,13 +325,38 @@ NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical,
         physical.GetProperties2(physical_properties);
         if (shading_rate_props.primitiveFragmentShadingRateWithMultipleViewports) {
             // Only Ampere and newer support this feature
+            // TODO: Find a way to differentiate Ampere and Ada
             return NvidiaArchitecture::AmpereOrNewer;
         }
-    }
-    if (exts.contains(VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME)) {
         return NvidiaArchitecture::Turing;
     }
-    return NvidiaArchitecture::VoltaOrOlder;
+
+    if (exts.contains(VK_EXT_BLEND_OPERATION_ADVANCED_EXTENSION_NAME)) {
+        VkPhysicalDeviceBlendOperationAdvancedPropertiesEXT advanced_blending_props{};
+        advanced_blending_props.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_BLEND_OPERATION_ADVANCED_PROPERTIES_EXT;
+        VkPhysicalDeviceProperties2 physical_properties{};
+        physical_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2;
+        physical_properties.pNext = &advanced_blending_props;
+        physical.GetProperties2(physical_properties);
+        if (advanced_blending_props.advancedBlendMaxColorAttachments == 1) {
+            return NvidiaArchitecture::Maxwell;
+        }
+
+        if (exts.contains(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME)) {
+            VkPhysicalDeviceConservativeRasterizationPropertiesEXT conservative_raster_props{};
+            conservative_raster_props.sType =
+                VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONSERVATIVE_RASTERIZATION_PROPERTIES_EXT;
+            physical_properties.pNext = &conservative_raster_props;
+            physical.GetProperties2(physical_properties);
+            if (conservative_raster_props.degenerateLinesRasterized) {
+                return NvidiaArchitecture::Volta;
+            }
+            return NvidiaArchitecture::Pascal;
+        }
+    }
+
+    return NvidiaArchitecture::KeplerOrOlder;
 }
 
 std::vector<const char*> ExtensionListForVulkan(
@@ -504,19 +533,14 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
     if (is_nvidia) {
         const u32 nv_major_version = (properties.properties.driverVersion >> 22) & 0x3ff;
         const auto arch = GetNvidiaArchitecture(physical, supported_extensions);
-        switch (arch) {
+        if (arch >= NvidiaArchitecture::AmpereOrNewer) {
-        case NvidiaArchitecture::AmpereOrNewer:
             LOG_WARNING(Render_Vulkan, "Ampere and newer have broken float16 math");
             features.shader_float16_int8.shaderFloat16 = false;
-            break;
+        } else if (arch <= NvidiaArchitecture::Volta) {
-        case NvidiaArchitecture::Turing:
-            break;
-        case NvidiaArchitecture::VoltaOrOlder:
             if (nv_major_version < 527) {
                 LOG_WARNING(Render_Vulkan, "Volta and older have broken VK_KHR_push_descriptor");
                 RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
             }
-            break;
         }
         if (nv_major_version >= 510) {
             LOG_WARNING(Render_Vulkan, "NVIDIA Drivers >= 510 do not support MSAA image blits");
@@ -661,7 +685,15 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
                         "ANV drivers 22.3.0 to 23.1.0 have broken VK_KHR_push_descriptor");
             RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
         }
+    } else if (extensions.push_descriptor && is_nvidia) {
+        const auto arch = GetNvidiaArchitecture(physical, supported_extensions);
+        if (arch <= NvidiaArchitecture::Pascal) {
+            LOG_WARNING(Render_Vulkan,
+                        "Pascal and older architectures have broken VK_KHR_push_descriptor");
+            RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
         }
+    }
+
     if (is_mvk) {
         LOG_WARNING(Render_Vulkan,
                     "MVK driver breaks when using more than 16 vertex attributes/bindings");