diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp index 74f06427dd..dce6c55cd0 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp @@ -10,6 +10,7 @@ #include #include +#include "common/settings.h" #include "video_core/buffer_cache/buffer_cache_base.h" #include "video_core/renderer_vulkan/vk_buffer_cache.h" @@ -334,6 +335,9 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m staging_pool{staging_pool_}, guest_descriptor_queue{guest_descriptor_queue_}, quad_index_pass(device, scheduler, descriptor_pool, staging_pool, compute_pass_descriptor_queue) { + use_vertex_input_dynamic_state = device.IsExtVertexInputDynamicStateSupported() && + Settings::values.vertex_input_dynamic_state.GetValue(); + const VkDriverIdKHR driver_id = device.GetDriverID(); limit_dynamic_storage_buffers = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY || driver_id == VK_DRIVER_ID_ARM_PROPRIETARY; @@ -556,7 +560,10 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset if (index >= device.GetMaxVertexInputBindings()) { return; } - if (device.IsExtExtendedDynamicStateSupported()) { + const bool use_bind_vertex_buffers2_stride = + device.IsExtExtendedDynamicStateSupported() && !use_vertex_input_dynamic_state; + + if (use_bind_vertex_buffers2_stride) { scheduler.Record([index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf) { const VkDeviceSize vk_offset = buffer != VK_NULL_HANDLE ? offset : 0; const VkDeviceSize vk_size = buffer != VK_NULL_HANDLE ? size : VK_WHOLE_SIZE; @@ -596,7 +603,10 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings& bi if (binding_count == 0) { return; } - if (device.IsExtExtendedDynamicStateSupported()) { + const bool use_bind_vertex_buffers2_stride = + device.IsExtExtendedDynamicStateSupported() && !use_vertex_input_dynamic_state; + + if (use_bind_vertex_buffers2_stride) { scheduler.Record([bindings_ = std::move(bindings), buffer_handles_ = std::move(buffer_handles), binding_count](vk::CommandBuffer cmdbuf) { cmdbuf.BindVertexBuffers2EXT(bindings_.min_index, binding_count, buffer_handles_.data(), bindings_.offsets.data(), bindings_.sizes.data(), bindings_.strides.data()); }); diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h index b73fcd162b..c8dd68e3dc 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.h +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h @@ -183,6 +183,7 @@ private: std::unique_ptr uint8_pass; QuadIndexedPass quad_index_pass; + bool use_vertex_input_dynamic_state = false; bool limit_dynamic_storage_buffers = false; u32 max_dynamic_storage_buffers = (std::numeric_limits::max)(); }; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index ba7b5d8c1b..ad431c9ac5 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -1732,39 +1732,54 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs) } dirty[Dirty::VertexInput] = false; + const size_t max_vertex_attributes = + (std::min)(Maxwell::NumVertexAttributes, + static_cast(device.GetMaxVertexInputAttributes())); + const size_t max_vertex_bindings = + (std::min)(Maxwell::NumVertexArrays, + static_cast(device.GetMaxVertexInputBindings())); + + // Dynamic state path must respect device limits, same as static pipeline creation. + for (size_t index = max_vertex_attributes; index < Maxwell::NumVertexAttributes; ++index) { + dirty[Dirty::VertexAttribute0 + index] = false; + } + for (size_t index = max_vertex_bindings; index < Maxwell::NumVertexArrays; ++index) { + dirty[Dirty::VertexBinding0 + index] = false; + } + boost::container::static_vector bindings; boost::container::static_vector attributes; - // There seems to be a bug on Nvidia's driver where updating only higher attributes ends up - // generating dirty state. Track the highest dirty attribute and update all attributes until - // that one. - size_t highest_dirty_attr{}; - for (size_t index = 0; index < Maxwell::NumVertexAttributes; ++index) { - if (dirty[Dirty::VertexAttribute0 + index]) { - highest_dirty_attr = index; - } - } - for (size_t index = 0; index < highest_dirty_attr; ++index) { - const Maxwell::VertexAttribute attribute{regs.vertex_attrib_format[index]}; - const u32 binding{attribute.buffer}; + std::array used_bindings{}; + for (size_t index = 0; index < max_vertex_attributes; ++index) { dirty[Dirty::VertexAttribute0 + index] = false; - dirty[Dirty::VertexBinding0 + static_cast(binding)] = true; - if (!attribute.constant) { - attributes.push_back({ - .sType = VK_STRUCTURE_TYPE_VERTEX_INPUT_ATTRIBUTE_DESCRIPTION_2_EXT, - .pNext = nullptr, - .location = static_cast(index), - .binding = binding, - .format = MaxwellToVK::VertexFormat(device, attribute.type, attribute.size), - .offset = attribute.offset, - }); - } - } - for (size_t index = 0; index < Maxwell::NumVertexAttributes; ++index) { - if (!dirty[Dirty::VertexBinding0 + index]) { + + const Maxwell::VertexAttribute attribute{regs.vertex_attrib_format[index]}; + if (attribute.constant) { continue; } + + const size_t binding{attribute.buffer}; + if (binding >= max_vertex_bindings) { + continue; + } + + used_bindings[binding] = true; + attributes.push_back({ + .sType = VK_STRUCTURE_TYPE_VERTEX_INPUT_ATTRIBUTE_DESCRIPTION_2_EXT, + .pNext = nullptr, + .location = static_cast(index), + .binding = static_cast(binding), + .format = MaxwellToVK::VertexFormat(device, attribute.type, attribute.size), + .offset = attribute.offset, + }); + } + + for (size_t index = 0; index < max_vertex_bindings; ++index) { dirty[Dirty::VertexBinding0 + index] = false; + if (!used_bindings[index]) { + continue; + } const u32 binding{static_cast(index)}; const auto& input_binding{regs.vertex_streams[binding]}; @@ -1778,6 +1793,10 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs) .divisor = is_instanced ? input_binding.frequency : 1, }); } + + ASSERT(attributes.size() <= max_vertex_attributes); + ASSERT(bindings.size() <= max_vertex_bindings); + scheduler.Record([bindings, attributes](vk::CommandBuffer cmdbuf) { cmdbuf.SetVertexInputEXT(bindings, attributes); }); diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 48aa5ec476..2dbaab4127 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -2300,23 +2300,18 @@ vk::ImageView ImageView::MakeView(VkFormat vk_format, VkImageAspectFlags aspect_ Sampler::Sampler(TextureCacheRuntime& runtime, const Tegra::Texture::TSCEntry& tsc) { const auto& device = runtime.device; - // Check if custom border colors are supported - const bool has_custom_border_colors = runtime.device.IsCustomBorderColorsSupported(); - const bool has_format_undefined = runtime.device.IsCustomBorderColorWithoutFormatSupported(); + const bool has_custom_border_extension = runtime.device.IsExtCustomBorderColorSupported(); + const bool has_format_undefined = + has_custom_border_extension && runtime.device.IsCustomBorderColorWithoutFormatSupported(); + const bool has_custom_border_colors = + has_format_undefined && runtime.device.IsCustomBorderColorsSupported(); const auto color = tsc.BorderColor(); - // Determine border format based on available features: - // - If customBorderColorWithoutFormat is available: use VK_FORMAT_UNDEFINED (most flexible) - // - If only customBorderColors is available: use concrete format (R8G8B8A8_UNORM) - // - If neither is available: use standard border colors (handled by ConvertBorderColor) - const VkFormat border_format = has_format_undefined ? VK_FORMAT_UNDEFINED - : VK_FORMAT_R8G8B8A8_UNORM; - const VkSamplerCustomBorderColorCreateInfoEXT border_ci{ .sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT, .pNext = nullptr, .customBorderColor = std::bit_cast(color), - .format = border_format, + .format = VK_FORMAT_UNDEFINED, }; const void* pnext = nullptr; if (has_custom_border_colors) { diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index 5075a79bcd..05ac56f011 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -919,6 +919,17 @@ bool Device::GetSuitability(bool requires_swapchain) { FOR_EACH_VK_FEATURE_EXT(FEATURE_EXTENSION); FOR_EACH_VK_EXTENSION(EXTENSION); + if (supported_extensions.contains(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME)) { + loaded_extensions.erase(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME); + loaded_extensions.insert(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME); + extensions.robustness_2 = true; + } else if (supported_extensions.contains(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME)) { + loaded_extensions.insert(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME); + extensions.robustness_2 = true; + } else { + extensions.robustness_2 = false; + } + #undef FEATURE_EXTENSION #undef EXTENSION @@ -1101,38 +1112,13 @@ bool Device::GetSuitability(bool requires_swapchain) { // VK_DYNAMIC_STATE // Driver detection variables for workarounds in GetSuitability - const VkDriverId driver_id = properties.driver.driverID; // VK_EXT_extended_dynamic_state2 below this will appear drivers that need workarounds. // VK_EXT_extended_dynamic_state3 below this will appear drivers that need workarounds. - // Samsung: Broken extendedDynamicState3ColorBlendEquation - // Disable blend equation dynamic state, force static pipeline state - if (extensions.extended_dynamic_state3 && - (driver_id == VK_DRIVER_ID_SAMSUNG_PROPRIETARY)) { - LOG_WARNING(Render_Vulkan, - "Samsung: Disabling broken extendedDynamicState3ColorBlendEquation"); - features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable = false; - features.extended_dynamic_state3.extendedDynamicState3ColorBlendEquation = false; - } - - // Intel Windows < 27.20.100.0: Broken VertexInputDynamicState - // Same for NVIDIA Proprietary < 580.119.02, unknown when VIDS was first NOT broken - // Disable VertexInputDynamicState on old Intel Windows drivers - if (extensions.vertex_input_dynamic_state) { - const u32 version = (properties.properties.driverVersion << 3) >> 3; - if ((driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS && version < VK_MAKE_API_VERSION(27, 20, 100, 0)) - || (driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY && version < VK_MAKE_API_VERSION(580, 119, 02, 0))) { - LOG_WARNING(Render_Vulkan, "Disabling broken VK_EXT_vertex_input_dynamic_state"); - RemoveExtensionFeature(extensions.vertex_input_dynamic_state, features.vertex_input_dynamic_state, VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); - } - } - if (u32(Settings::values.dyna_state.GetValue()) == 0) { LOG_INFO(Render_Vulkan, "Extended Dynamic State disabled by user setting, clearing all EDS features"); - features.custom_border_color.customBorderColors = false; - features.custom_border_color.customBorderColorWithoutFormat = false; features.extended_dynamic_state.extendedDynamicState = false; features.extended_dynamic_state2.extendedDynamicState2 = false; features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable = false; @@ -1148,24 +1134,13 @@ bool Device::GetSuitability(bool requires_swapchain) { void Device::RemoveUnsuitableExtensions() { // VK_EXT_custom_border_color - // Enable extension if driver supports it, then check individual features - // - customBorderColors: Required to use VK_BORDER_COLOR_FLOAT_CUSTOM_EXT - // - customBorderColorWithoutFormat: Optional, allows VK_FORMAT_UNDEFINED - // If only customBorderColors is available, we must provide a specific format if (extensions.custom_border_color) { - // Verify that at least customBorderColors is available - if (!features.custom_border_color.customBorderColors) { - LOG_WARNING(Render_Vulkan, - "VK_EXT_custom_border_color reported but customBorderColors feature not available, disabling"); - extensions.custom_border_color = false; - } + extensions.custom_border_color = + features.custom_border_color.customBorderColors && + features.custom_border_color.customBorderColorWithoutFormat; } RemoveExtensionFeatureIfUnsuitable(extensions.custom_border_color, features.custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); - // VK_KHR_unified_image_layouts - extensions.unified_image_layouts = features.unified_image_layouts.unifiedImageLayouts; - RemoveExtensionFeatureIfUnsuitable(extensions.unified_image_layouts, features.unified_image_layouts, - VK_KHR_UNIFIED_IMAGE_LAYOUTS_EXTENSION_NAME); // VK_EXT_depth_bias_control extensions.depth_bias_control = @@ -1251,16 +1226,22 @@ void Device::RemoveUnsuitableExtensions() { VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME); // VK_EXT_robustness2 - extensions.robustness_2 = features.robustness2.robustBufferAccess2 || - features.robustness2.robustImageAccess2 || - features.robustness2.nullDescriptor; + features.robustness2.robustBufferAccess2 = VK_FALSE; + features.robustness2.robustImageAccess2 = VK_FALSE; + extensions.robustness_2 = features.robustness2.nullDescriptor; + + const char* robustness2_extension_name = + loaded_extensions.contains(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME) + ? VK_KHR_ROBUSTNESS_2_EXTENSION_NAME + : VK_EXT_ROBUSTNESS_2_EXTENSION_NAME; RemoveExtensionFeatureIfUnsuitable(extensions.robustness_2, features.robustness2, - VK_EXT_ROBUSTNESS_2_EXTENSION_NAME); + robustness2_extension_name); - // VK_EXT_image_robustness - extensions.image_robustness = features.image_robustness.robustImageAccess; - RemoveExtensionFeatureIfUnsuitable(extensions.image_robustness, features.image_robustness, + // Image robustness + extensions.robust_image_access = features.robust_image_access.robustImageAccess; + RemoveExtensionFeatureIfUnsuitable(extensions.robust_image_access, + features.robust_image_access, VK_EXT_IMAGE_ROBUSTNESS_EXTENSION_NAME); // VK_KHR_shader_atomic_int64 diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h index ad9d53ce16..8914ef0eb3 100644 --- a/src/video_core/vulkan_common/vulkan_device.h +++ b/src/video_core/vulkan_common/vulkan_device.h @@ -38,7 +38,7 @@ VK_DEFINE_HANDLE(VmaAllocator) FEATURE(KHR, TimelineSemaphore, TIMELINE_SEMAPHORE, timeline_semaphore) #define FOR_EACH_VK_FEATURE_1_3(FEATURE) \ - FEATURE(EXT, ImageRobustness, IMAGE_ROBUSTNESS, image_robustness) \ + FEATURE(EXT, ImageRobustness, IMAGE_ROBUSTNESS, robust_image_access) \ FEATURE(EXT, ShaderDemoteToHelperInvocation, SHADER_DEMOTE_TO_HELPER_INVOCATION, \ shader_demote_to_helper_invocation) \ FEATURE(EXT, SubgroupSizeControl, SUBGROUP_SIZE_CONTROL, subgroup_size_control) \ @@ -68,8 +68,7 @@ VK_DEFINE_HANDLE(VmaAllocator) FEATURE(KHR, PipelineExecutableProperties, PIPELINE_EXECUTABLE_PROPERTIES, \ pipeline_executable_properties) \ FEATURE(KHR, WorkgroupMemoryExplicitLayout, WORKGROUP_MEMORY_EXPLICIT_LAYOUT, \ - workgroup_memory_explicit_layout) \ - FEATURE(KHR, UnifiedImageLayouts, UNIFIED_IMAGE_LAYOUTS, unified_image_layouts) + workgroup_memory_explicit_layout) // Define miscellaneous extensions which may be used by the implementation here. @@ -124,7 +123,6 @@ VK_DEFINE_HANDLE(VmaAllocator) EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_EXTERNAL_MEMORY_HOST_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_4444_FORMATS_EXTENSION_NAME) \ - EXTENSION_NAME(VK_EXT_IMAGE_ROBUSTNESS_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME) \ @@ -174,13 +172,11 @@ VK_DEFINE_HANDLE(VmaAllocator) FEATURE_NAME(depth_bias_control, depthBiasExact) \ FEATURE_NAME(extended_dynamic_state, extendedDynamicState) \ FEATURE_NAME(format_a4b4g4r4, formatA4B4G4R4) \ - FEATURE_NAME(image_robustness, robustImageAccess) \ + FEATURE_NAME(robust_image_access, robustImageAccess) \ FEATURE_NAME(index_type_uint8, indexTypeUint8) \ FEATURE_NAME(primitive_topology_list_restart, primitiveTopologyListRestart) \ FEATURE_NAME(provoking_vertex, provokingVertexLast) \ FEATURE_NAME(robustness2, nullDescriptor) \ - FEATURE_NAME(robustness2, robustBufferAccess2) \ - FEATURE_NAME(robustness2, robustImageAccess2) \ FEATURE_NAME(shader_float16_int8, shaderFloat16) \ FEATURE_NAME(shader_float16_int8, shaderInt8) \ FEATURE_NAME(timeline_semaphore, timelineSemaphore) \ @@ -552,36 +548,6 @@ public: return extensions.custom_border_color; } - /// Returns true if the device supports VK_EXT_image_robustness. - bool IsExtImageRobustnessSupported() const { - return extensions.image_robustness; - } - - /// Returns true if robustImageAccess is supported. - bool IsRobustImageAccessSupported() const { - return features.image_robustness.robustImageAccess; - } - - /// Returns true if the device supports VK_EXT_robustness2. - bool IsExtRobustness2Supported() const { - return extensions.robustness_2; - } - - /// Returns true if robustBufferAccess2 is supported. - bool IsRobustBufferAccess2Supported() const { - return features.robustness2.robustBufferAccess2; - } - - /// Returns true if robustImageAccess2 is supported. - bool IsRobustImageAccess2Supported() const { - return features.robustness2.robustImageAccess2; - } - - /// Returns true if nullDescriptor is supported. - bool IsNullDescriptorSupported() const { - return features.robustness2.nullDescriptor; - } - /// Returns true if customBorderColors feature is available. bool IsCustomBorderColorsSupported() const { return features.custom_border_color.customBorderColors;