diff options
Diffstat (limited to 'src/video_core/vulkan_common')
| -rw-r--r-- | src/video_core/vulkan_common/vulkan_device.cpp | 56 | ||||
| -rw-r--r-- | src/video_core/vulkan_common/vulkan_memory_allocator.cpp | 21 | ||||
| -rw-r--r-- | src/video_core/vulkan_common/vulkan_memory_allocator.h | 14 |
3 files changed, 75 insertions, 16 deletions
diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index 3960b135a..876cec2e8 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/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) { - case NvidiaArchitecture::AmpereOrNewer: + if (arch >= NvidiaArchitecture::AmpereOrNewer) { LOG_WARNING(Render_Vulkan, "Ampere and newer have broken float16 math"); features.shader_float16_int8.shaderFloat16 = false; - break; - case NvidiaArchitecture::Turing: - break; - case NvidiaArchitecture::VoltaOrOlder: + } else if (arch <= NvidiaArchitecture::Volta) { 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"); diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp index 3ef381a38..82767fdf0 100644 --- a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp +++ b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp @@ -9,6 +9,7 @@ #include "common/alignment.h" #include "common/assert.h" #include "common/common_types.h" +#include "common/literals.h" #include "common/logging/log.h" #include "common/polyfill_ranges.h" #include "video_core/vulkan_common/vma.h" @@ -69,8 +70,7 @@ struct Range { case MemoryUsage::Download: return VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT; case MemoryUsage::DeviceLocal: - return VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT | - VMA_ALLOCATION_CREATE_HOST_ACCESS_ALLOW_TRANSFER_INSTEAD_BIT; + return {}; } return {}; } @@ -212,7 +212,20 @@ MemoryAllocator::MemoryAllocator(const Device& device_) : device{device_}, allocator{device.GetAllocator()}, properties{device_.GetPhysical().GetMemoryProperties().memoryProperties}, buffer_image_granularity{ - device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {} + device_.GetPhysical().GetProperties().limits.bufferImageGranularity} { + // GPUs not supporting rebar may only have a region with less than 256MB host visible/device + // local memory. In that case, opening 2 RenderDoc captures side-by-side is not possible due to + // the heap running out of memory. With RenderDoc attached and only a small host/device region, + // only allow the stream buffer in this memory heap. + if (device.HasDebuggingToolAttached()) { + using namespace Common::Literals; + ForEachDeviceLocalHostVisibleHeap(device, [this](size_t index, VkMemoryHeap& heap) { + if (heap.size <= 256_MiB) { + valid_memory_types &= ~(1u << index); + } + }); + } +} MemoryAllocator::~MemoryAllocator() = default; @@ -244,7 +257,7 @@ vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo& ci, MemoryUsa .usage = MemoryUsageVma(usage), .requiredFlags = 0, .preferredFlags = MemoryUsagePreferedVmaFlags(usage), - .memoryTypeBits = 0, + .memoryTypeBits = usage == MemoryUsage::Stream ? 0u : valid_memory_types, .pool = VK_NULL_HANDLE, .pUserData = nullptr, .priority = 0.f, diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.h b/src/video_core/vulkan_common/vulkan_memory_allocator.h index f449bc8d0..38a182bcb 100644 --- a/src/video_core/vulkan_common/vulkan_memory_allocator.h +++ b/src/video_core/vulkan_common/vulkan_memory_allocator.h @@ -7,6 +7,7 @@ #include <span> #include <vector> #include "common/common_types.h" +#include "video_core/vulkan_common/vulkan_device.h" #include "video_core/vulkan_common/vulkan_wrapper.h" VK_DEFINE_HANDLE(VmaAllocator) @@ -26,6 +27,18 @@ enum class MemoryUsage { Stream, ///< Requests device local host visible buffer, falling back host memory. }; +template <typename F> +void ForEachDeviceLocalHostVisibleHeap(const Device& device, F&& f) { + auto memory_props = device.GetPhysical().GetMemoryProperties().memoryProperties; + for (size_t i = 0; i < memory_props.memoryTypeCount; i++) { + auto& memory_type = memory_props.memoryTypes[i]; + if ((memory_type.propertyFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT) && + (memory_type.propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT)) { + f(memory_type.heapIndex, memory_props.memoryHeaps[memory_type.heapIndex]); + } + } +} + /// Ownership handle of a memory commitment. /// Points to a subregion of a memory allocation. class MemoryCommit { @@ -124,6 +137,7 @@ private: std::vector<std::unique_ptr<MemoryAllocation>> allocations; ///< Current allocations. VkDeviceSize buffer_image_granularity; // The granularity for adjacent offsets between buffers // and optimal images + u32 valid_memory_types{~0u}; }; } // namespace Vulkan |