diff options
52 files changed, 2881 insertions, 2272 deletions
diff --git a/externals/Vulkan-Headers b/externals/Vulkan-Headers | |||
| Subproject d42d0747ee1b7a6726fb8948444b4993f9dcd2e | Subproject 0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5 | ||
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index f7febd6a2..258d58eba 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt | |||
| @@ -156,7 +156,6 @@ add_library(video_core STATIC | |||
| 156 | 156 | ||
| 157 | if (ENABLE_VULKAN) | 157 | if (ENABLE_VULKAN) |
| 158 | target_sources(video_core PRIVATE | 158 | target_sources(video_core PRIVATE |
| 159 | renderer_vulkan/declarations.h | ||
| 160 | renderer_vulkan/fixed_pipeline_state.cpp | 159 | renderer_vulkan/fixed_pipeline_state.cpp |
| 161 | renderer_vulkan/fixed_pipeline_state.h | 160 | renderer_vulkan/fixed_pipeline_state.h |
| 162 | renderer_vulkan/maxwell_to_vk.cpp | 161 | renderer_vulkan/maxwell_to_vk.cpp |
diff --git a/src/video_core/renderer_vulkan/declarations.h b/src/video_core/renderer_vulkan/declarations.h deleted file mode 100644 index 89a035ca4..000000000 --- a/src/video_core/renderer_vulkan/declarations.h +++ /dev/null | |||
| @@ -1,60 +0,0 @@ | |||
| 1 | // Copyright 2019 yuzu Emulator Project | ||
| 2 | // Licensed under GPLv2 or any later version | ||
| 3 | // Refer to the license.txt file included. | ||
| 4 | |||
| 5 | #pragma once | ||
| 6 | |||
| 7 | namespace vk { | ||
| 8 | class DispatchLoaderDynamic; | ||
| 9 | } | ||
| 10 | |||
| 11 | namespace Vulkan { | ||
| 12 | constexpr vk::DispatchLoaderDynamic* dont_use_me_dld = nullptr; | ||
| 13 | } | ||
| 14 | |||
| 15 | #define VULKAN_HPP_DEFAULT_DISPATCHER (*::Vulkan::dont_use_me_dld) | ||
| 16 | #define VULKAN_HPP_ENABLE_DYNAMIC_LOADER_TOOL 0 | ||
| 17 | #define VULKAN_HPP_DISPATCH_LOADER_DYNAMIC 1 | ||
| 18 | #include <vulkan/vulkan.hpp> | ||
| 19 | |||
| 20 | namespace Vulkan { | ||
| 21 | |||
| 22 | // vulkan.hpp unique handlers use DispatchLoaderStatic | ||
| 23 | template <typename T> | ||
| 24 | using UniqueHandle = vk::UniqueHandle<T, vk::DispatchLoaderDynamic>; | ||
| 25 | |||
| 26 | using UniqueAccelerationStructureNV = UniqueHandle<vk::AccelerationStructureNV>; | ||
| 27 | using UniqueBuffer = UniqueHandle<vk::Buffer>; | ||
| 28 | using UniqueBufferView = UniqueHandle<vk::BufferView>; | ||
| 29 | using UniqueCommandBuffer = UniqueHandle<vk::CommandBuffer>; | ||
| 30 | using UniqueCommandPool = UniqueHandle<vk::CommandPool>; | ||
| 31 | using UniqueDescriptorPool = UniqueHandle<vk::DescriptorPool>; | ||
| 32 | using UniqueDescriptorSet = UniqueHandle<vk::DescriptorSet>; | ||
| 33 | using UniqueDescriptorSetLayout = UniqueHandle<vk::DescriptorSetLayout>; | ||
| 34 | using UniqueDescriptorUpdateTemplate = UniqueHandle<vk::DescriptorUpdateTemplate>; | ||
| 35 | using UniqueDevice = UniqueHandle<vk::Device>; | ||
| 36 | using UniqueDeviceMemory = UniqueHandle<vk::DeviceMemory>; | ||
| 37 | using UniqueEvent = UniqueHandle<vk::Event>; | ||
| 38 | using UniqueFence = UniqueHandle<vk::Fence>; | ||
| 39 | using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>; | ||
| 40 | using UniqueImage = UniqueHandle<vk::Image>; | ||
| 41 | using UniqueImageView = UniqueHandle<vk::ImageView>; | ||
| 42 | using UniqueInstance = UniqueHandle<vk::Instance>; | ||
| 43 | using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>; | ||
| 44 | using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>; | ||
| 45 | using UniquePipeline = UniqueHandle<vk::Pipeline>; | ||
| 46 | using UniquePipelineCache = UniqueHandle<vk::PipelineCache>; | ||
| 47 | using UniquePipelineLayout = UniqueHandle<vk::PipelineLayout>; | ||
| 48 | using UniqueQueryPool = UniqueHandle<vk::QueryPool>; | ||
| 49 | using UniqueRenderPass = UniqueHandle<vk::RenderPass>; | ||
| 50 | using UniqueSampler = UniqueHandle<vk::Sampler>; | ||
| 51 | using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>; | ||
| 52 | using UniqueSemaphore = UniqueHandle<vk::Semaphore>; | ||
| 53 | using UniqueShaderModule = UniqueHandle<vk::ShaderModule>; | ||
| 54 | using UniqueSurfaceKHR = UniqueHandle<vk::SurfaceKHR>; | ||
| 55 | using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>; | ||
| 56 | using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>; | ||
| 57 | using UniqueDebugReportCallbackEXT = UniqueHandle<vk::DebugReportCallbackEXT>; | ||
| 58 | using UniqueDebugUtilsMessengerEXT = UniqueHandle<vk::DebugUtilsMessengerEXT>; | ||
| 59 | |||
| 60 | } // namespace Vulkan | ||
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index 7480cb7c3..143478863 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp | |||
| @@ -2,13 +2,15 @@ | |||
| 2 | // Licensed under GPLv2 or any later version | 2 | // Licensed under GPLv2 or any later version |
| 3 | // Refer to the license.txt file included. | 3 | // Refer to the license.txt file included. |
| 4 | 4 | ||
| 5 | #include <iterator> | ||
| 6 | |||
| 5 | #include "common/assert.h" | 7 | #include "common/assert.h" |
| 6 | #include "common/common_types.h" | 8 | #include "common/common_types.h" |
| 7 | #include "common/logging/log.h" | 9 | #include "common/logging/log.h" |
| 8 | #include "video_core/engines/maxwell_3d.h" | 10 | #include "video_core/engines/maxwell_3d.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 11 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 11 | #include "video_core/renderer_vulkan/vk_device.h" | 12 | #include "video_core/renderer_vulkan/vk_device.h" |
| 13 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 12 | #include "video_core/surface.h" | 14 | #include "video_core/surface.h" |
| 13 | 15 | ||
| 14 | namespace Vulkan::MaxwellToVK { | 16 | namespace Vulkan::MaxwellToVK { |
| @@ -17,88 +19,89 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs; | |||
| 17 | 19 | ||
| 18 | namespace Sampler { | 20 | namespace Sampler { |
| 19 | 21 | ||
| 20 | vk::Filter Filter(Tegra::Texture::TextureFilter filter) { | 22 | VkFilter Filter(Tegra::Texture::TextureFilter filter) { |
| 21 | switch (filter) { | 23 | switch (filter) { |
| 22 | case Tegra::Texture::TextureFilter::Linear: | 24 | case Tegra::Texture::TextureFilter::Linear: |
| 23 | return vk::Filter::eLinear; | 25 | return VK_FILTER_LINEAR; |
| 24 | case Tegra::Texture::TextureFilter::Nearest: | 26 | case Tegra::Texture::TextureFilter::Nearest: |
| 25 | return vk::Filter::eNearest; | 27 | return VK_FILTER_NEAREST; |
| 26 | } | 28 | } |
| 27 | UNIMPLEMENTED_MSG("Unimplemented sampler filter={}", static_cast<u32>(filter)); | 29 | UNIMPLEMENTED_MSG("Unimplemented sampler filter={}", static_cast<u32>(filter)); |
| 28 | return {}; | 30 | return {}; |
| 29 | } | 31 | } |
| 30 | 32 | ||
| 31 | vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) { | 33 | VkSamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) { |
| 32 | switch (mipmap_filter) { | 34 | switch (mipmap_filter) { |
| 33 | case Tegra::Texture::TextureMipmapFilter::None: | 35 | case Tegra::Texture::TextureMipmapFilter::None: |
| 34 | // TODO(Rodrigo): None seems to be mapped to OpenGL's mag and min filters without mipmapping | 36 | // TODO(Rodrigo): None seems to be mapped to OpenGL's mag and min filters without mipmapping |
| 35 | // (e.g. GL_NEAREST and GL_LINEAR). Vulkan doesn't have such a thing, find out if we have to | 37 | // (e.g. GL_NEAREST and GL_LINEAR). Vulkan doesn't have such a thing, find out if we have to |
| 36 | // use an image view with a single mipmap level to emulate this. | 38 | // use an image view with a single mipmap level to emulate this. |
| 37 | return vk::SamplerMipmapMode::eLinear; | 39 | return VK_SAMPLER_MIPMAP_MODE_LINEAR; |
| 40 | ; | ||
| 38 | case Tegra::Texture::TextureMipmapFilter::Linear: | 41 | case Tegra::Texture::TextureMipmapFilter::Linear: |
| 39 | return vk::SamplerMipmapMode::eLinear; | 42 | return VK_SAMPLER_MIPMAP_MODE_LINEAR; |
| 40 | case Tegra::Texture::TextureMipmapFilter::Nearest: | 43 | case Tegra::Texture::TextureMipmapFilter::Nearest: |
| 41 | return vk::SamplerMipmapMode::eNearest; | 44 | return VK_SAMPLER_MIPMAP_MODE_NEAREST; |
| 42 | } | 45 | } |
| 43 | UNIMPLEMENTED_MSG("Unimplemented sampler mipmap mode={}", static_cast<u32>(mipmap_filter)); | 46 | UNIMPLEMENTED_MSG("Unimplemented sampler mipmap mode={}", static_cast<u32>(mipmap_filter)); |
| 44 | return {}; | 47 | return {}; |
| 45 | } | 48 | } |
| 46 | 49 | ||
| 47 | vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, | 50 | VkSamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, |
| 48 | Tegra::Texture::TextureFilter filter) { | 51 | Tegra::Texture::TextureFilter filter) { |
| 49 | switch (wrap_mode) { | 52 | switch (wrap_mode) { |
| 50 | case Tegra::Texture::WrapMode::Wrap: | 53 | case Tegra::Texture::WrapMode::Wrap: |
| 51 | return vk::SamplerAddressMode::eRepeat; | 54 | return VK_SAMPLER_ADDRESS_MODE_REPEAT; |
| 52 | case Tegra::Texture::WrapMode::Mirror: | 55 | case Tegra::Texture::WrapMode::Mirror: |
| 53 | return vk::SamplerAddressMode::eMirroredRepeat; | 56 | return VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT; |
| 54 | case Tegra::Texture::WrapMode::ClampToEdge: | 57 | case Tegra::Texture::WrapMode::ClampToEdge: |
| 55 | return vk::SamplerAddressMode::eClampToEdge; | 58 | return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; |
| 56 | case Tegra::Texture::WrapMode::Border: | 59 | case Tegra::Texture::WrapMode::Border: |
| 57 | return vk::SamplerAddressMode::eClampToBorder; | 60 | return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; |
| 58 | case Tegra::Texture::WrapMode::Clamp: | 61 | case Tegra::Texture::WrapMode::Clamp: |
| 59 | if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) { | 62 | if (device.GetDriverID() == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) { |
| 60 | // Nvidia's Vulkan driver defaults to GL_CLAMP on invalid enumerations, we can hack this | 63 | // Nvidia's Vulkan driver defaults to GL_CLAMP on invalid enumerations, we can hack this |
| 61 | // by sending an invalid enumeration. | 64 | // by sending an invalid enumeration. |
| 62 | return static_cast<vk::SamplerAddressMode>(0xcafe); | 65 | return static_cast<VkSamplerAddressMode>(0xcafe); |
| 63 | } | 66 | } |
| 64 | // TODO(Rodrigo): Emulate GL_CLAMP properly on other vendors | 67 | // TODO(Rodrigo): Emulate GL_CLAMP properly on other vendors |
| 65 | switch (filter) { | 68 | switch (filter) { |
| 66 | case Tegra::Texture::TextureFilter::Nearest: | 69 | case Tegra::Texture::TextureFilter::Nearest: |
| 67 | return vk::SamplerAddressMode::eClampToEdge; | 70 | return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; |
| 68 | case Tegra::Texture::TextureFilter::Linear: | 71 | case Tegra::Texture::TextureFilter::Linear: |
| 69 | return vk::SamplerAddressMode::eClampToBorder; | 72 | return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; |
| 70 | } | 73 | } |
| 71 | UNREACHABLE(); | 74 | UNREACHABLE(); |
| 72 | return vk::SamplerAddressMode::eClampToEdge; | 75 | return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; |
| 73 | case Tegra::Texture::WrapMode::MirrorOnceClampToEdge: | 76 | case Tegra::Texture::WrapMode::MirrorOnceClampToEdge: |
| 74 | return vk::SamplerAddressMode::eMirrorClampToEdge; | 77 | return VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE; |
| 75 | case Tegra::Texture::WrapMode::MirrorOnceBorder: | 78 | case Tegra::Texture::WrapMode::MirrorOnceBorder: |
| 76 | UNIMPLEMENTED(); | 79 | UNIMPLEMENTED(); |
| 77 | return vk::SamplerAddressMode::eMirrorClampToEdge; | 80 | return VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE; |
| 78 | default: | 81 | default: |
| 79 | UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode)); | 82 | UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode)); |
| 80 | return {}; | 83 | return {}; |
| 81 | } | 84 | } |
| 82 | } | 85 | } |
| 83 | 86 | ||
| 84 | vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) { | 87 | VkCompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) { |
| 85 | switch (depth_compare_func) { | 88 | switch (depth_compare_func) { |
| 86 | case Tegra::Texture::DepthCompareFunc::Never: | 89 | case Tegra::Texture::DepthCompareFunc::Never: |
| 87 | return vk::CompareOp::eNever; | 90 | return VK_COMPARE_OP_NEVER; |
| 88 | case Tegra::Texture::DepthCompareFunc::Less: | 91 | case Tegra::Texture::DepthCompareFunc::Less: |
| 89 | return vk::CompareOp::eLess; | 92 | return VK_COMPARE_OP_LESS; |
| 90 | case Tegra::Texture::DepthCompareFunc::LessEqual: | 93 | case Tegra::Texture::DepthCompareFunc::LessEqual: |
| 91 | return vk::CompareOp::eLessOrEqual; | 94 | return VK_COMPARE_OP_LESS_OR_EQUAL; |
| 92 | case Tegra::Texture::DepthCompareFunc::Equal: | 95 | case Tegra::Texture::DepthCompareFunc::Equal: |
| 93 | return vk::CompareOp::eEqual; | 96 | return VK_COMPARE_OP_EQUAL; |
| 94 | case Tegra::Texture::DepthCompareFunc::NotEqual: | 97 | case Tegra::Texture::DepthCompareFunc::NotEqual: |
| 95 | return vk::CompareOp::eNotEqual; | 98 | return VK_COMPARE_OP_NOT_EQUAL; |
| 96 | case Tegra::Texture::DepthCompareFunc::Greater: | 99 | case Tegra::Texture::DepthCompareFunc::Greater: |
| 97 | return vk::CompareOp::eGreater; | 100 | return VK_COMPARE_OP_GREATER; |
| 98 | case Tegra::Texture::DepthCompareFunc::GreaterEqual: | 101 | case Tegra::Texture::DepthCompareFunc::GreaterEqual: |
| 99 | return vk::CompareOp::eGreaterOrEqual; | 102 | return VK_COMPARE_OP_GREATER_OR_EQUAL; |
| 100 | case Tegra::Texture::DepthCompareFunc::Always: | 103 | case Tegra::Texture::DepthCompareFunc::Always: |
| 101 | return vk::CompareOp::eAlways; | 104 | return VK_COMPARE_OP_ALWAYS; |
| 102 | } | 105 | } |
| 103 | UNIMPLEMENTED_MSG("Unimplemented sampler depth compare function={}", | 106 | UNIMPLEMENTED_MSG("Unimplemented sampler depth compare function={}", |
| 104 | static_cast<u32>(depth_compare_func)); | 107 | static_cast<u32>(depth_compare_func)); |
| @@ -112,92 +115,92 @@ namespace { | |||
| 112 | enum : u32 { Attachable = 1, Storage = 2 }; | 115 | enum : u32 { Attachable = 1, Storage = 2 }; |
| 113 | 116 | ||
| 114 | struct FormatTuple { | 117 | struct FormatTuple { |
| 115 | vk::Format format; ///< Vulkan format | 118 | VkFormat format; ///< Vulkan format |
| 116 | int usage; ///< Describes image format usage | 119 | int usage = 0; ///< Describes image format usage |
| 117 | } constexpr tex_format_tuples[] = { | 120 | } constexpr tex_format_tuples[] = { |
| 118 | {vk::Format::eA8B8G8R8UnormPack32, Attachable | Storage}, // ABGR8U | 121 | {VK_FORMAT_A8B8G8R8_UNORM_PACK32, Attachable | Storage}, // ABGR8U |
| 119 | {vk::Format::eA8B8G8R8SnormPack32, Attachable | Storage}, // ABGR8S | 122 | {VK_FORMAT_A8B8G8R8_SNORM_PACK32, Attachable | Storage}, // ABGR8S |
| 120 | {vk::Format::eA8B8G8R8UintPack32, Attachable | Storage}, // ABGR8UI | 123 | {VK_FORMAT_A8B8G8R8_UINT_PACK32, Attachable | Storage}, // ABGR8UI |
| 121 | {vk::Format::eB5G6R5UnormPack16, {}}, // B5G6R5U | 124 | {VK_FORMAT_B5G6R5_UNORM_PACK16}, // B5G6R5U |
| 122 | {vk::Format::eA2B10G10R10UnormPack32, Attachable | Storage}, // A2B10G10R10U | 125 | {VK_FORMAT_A2B10G10R10_UNORM_PACK32, Attachable | Storage}, // A2B10G10R10U |
| 123 | {vk::Format::eA1R5G5B5UnormPack16, Attachable}, // A1B5G5R5U (flipped with swizzle) | 126 | {VK_FORMAT_A1R5G5B5_UNORM_PACK16, Attachable}, // A1B5G5R5U (flipped with swizzle) |
| 124 | {vk::Format::eR8Unorm, Attachable | Storage}, // R8U | 127 | {VK_FORMAT_R8_UNORM, Attachable | Storage}, // R8U |
| 125 | {vk::Format::eR8Uint, Attachable | Storage}, // R8UI | 128 | {VK_FORMAT_R8_UINT, Attachable | Storage}, // R8UI |
| 126 | {vk::Format::eR16G16B16A16Sfloat, Attachable | Storage}, // RGBA16F | 129 | {VK_FORMAT_R16G16B16A16_SFLOAT, Attachable | Storage}, // RGBA16F |
| 127 | {vk::Format::eR16G16B16A16Unorm, Attachable | Storage}, // RGBA16U | 130 | {VK_FORMAT_R16G16B16A16_UNORM, Attachable | Storage}, // RGBA16U |
| 128 | {vk::Format::eR16G16B16A16Snorm, Attachable | Storage}, // RGBA16S | 131 | {VK_FORMAT_R16G16B16A16_SNORM, Attachable | Storage}, // RGBA16S |
| 129 | {vk::Format::eR16G16B16A16Uint, Attachable | Storage}, // RGBA16UI | 132 | {VK_FORMAT_R16G16B16A16_UINT, Attachable | Storage}, // RGBA16UI |
| 130 | {vk::Format::eB10G11R11UfloatPack32, Attachable | Storage}, // R11FG11FB10F | 133 | {VK_FORMAT_B10G11R11_UFLOAT_PACK32, Attachable | Storage}, // R11FG11FB10F |
| 131 | {vk::Format::eR32G32B32A32Uint, Attachable | Storage}, // RGBA32UI | 134 | {VK_FORMAT_R32G32B32A32_UINT, Attachable | Storage}, // RGBA32UI |
| 132 | {vk::Format::eBc1RgbaUnormBlock, {}}, // DXT1 | 135 | {VK_FORMAT_BC1_RGBA_UNORM_BLOCK}, // DXT1 |
| 133 | {vk::Format::eBc2UnormBlock, {}}, // DXT23 | 136 | {VK_FORMAT_BC2_UNORM_BLOCK}, // DXT23 |
| 134 | {vk::Format::eBc3UnormBlock, {}}, // DXT45 | 137 | {VK_FORMAT_BC3_UNORM_BLOCK}, // DXT45 |
| 135 | {vk::Format::eBc4UnormBlock, {}}, // DXN1 | 138 | {VK_FORMAT_BC4_UNORM_BLOCK}, // DXN1 |
| 136 | {vk::Format::eBc5UnormBlock, {}}, // DXN2UNORM | 139 | {VK_FORMAT_BC5_UNORM_BLOCK}, // DXN2UNORM |
| 137 | {vk::Format::eBc5SnormBlock, {}}, // DXN2SNORM | 140 | {VK_FORMAT_BC5_SNORM_BLOCK}, // DXN2SNORM |
| 138 | {vk::Format::eBc7UnormBlock, {}}, // BC7U | 141 | {VK_FORMAT_BC7_UNORM_BLOCK}, // BC7U |
| 139 | {vk::Format::eBc6HUfloatBlock, {}}, // BC6H_UF16 | 142 | {VK_FORMAT_BC6H_UFLOAT_BLOCK}, // BC6H_UF16 |
| 140 | {vk::Format::eBc6HSfloatBlock, {}}, // BC6H_SF16 | 143 | {VK_FORMAT_BC6H_SFLOAT_BLOCK}, // BC6H_SF16 |
| 141 | {vk::Format::eAstc4x4UnormBlock, {}}, // ASTC_2D_4X4 | 144 | {VK_FORMAT_ASTC_4x4_UNORM_BLOCK}, // ASTC_2D_4X4 |
| 142 | {vk::Format::eB8G8R8A8Unorm, {}}, // BGRA8 | 145 | {VK_FORMAT_B8G8R8A8_UNORM}, // BGRA8 |
| 143 | {vk::Format::eR32G32B32A32Sfloat, Attachable | Storage}, // RGBA32F | 146 | {VK_FORMAT_R32G32B32A32_SFLOAT, Attachable | Storage}, // RGBA32F |
| 144 | {vk::Format::eR32G32Sfloat, Attachable | Storage}, // RG32F | 147 | {VK_FORMAT_R32G32_SFLOAT, Attachable | Storage}, // RG32F |
| 145 | {vk::Format::eR32Sfloat, Attachable | Storage}, // R32F | 148 | {VK_FORMAT_R32_SFLOAT, Attachable | Storage}, // R32F |
| 146 | {vk::Format::eR16Sfloat, Attachable | Storage}, // R16F | 149 | {VK_FORMAT_R16_SFLOAT, Attachable | Storage}, // R16F |
| 147 | {vk::Format::eR16Unorm, Attachable | Storage}, // R16U | 150 | {VK_FORMAT_R16_UNORM, Attachable | Storage}, // R16U |
| 148 | {vk::Format::eUndefined, {}}, // R16S | 151 | {VK_FORMAT_UNDEFINED}, // R16S |
| 149 | {vk::Format::eUndefined, {}}, // R16UI | 152 | {VK_FORMAT_UNDEFINED}, // R16UI |
| 150 | {vk::Format::eUndefined, {}}, // R16I | 153 | {VK_FORMAT_UNDEFINED}, // R16I |
| 151 | {vk::Format::eR16G16Unorm, Attachable | Storage}, // RG16 | 154 | {VK_FORMAT_R16G16_UNORM, Attachable | Storage}, // RG16 |
| 152 | {vk::Format::eR16G16Sfloat, Attachable | Storage}, // RG16F | 155 | {VK_FORMAT_R16G16_SFLOAT, Attachable | Storage}, // RG16F |
| 153 | {vk::Format::eUndefined, {}}, // RG16UI | 156 | {VK_FORMAT_UNDEFINED}, // RG16UI |
| 154 | {vk::Format::eUndefined, {}}, // RG16I | 157 | {VK_FORMAT_UNDEFINED}, // RG16I |
| 155 | {vk::Format::eR16G16Snorm, Attachable | Storage}, // RG16S | 158 | {VK_FORMAT_R16G16_SNORM, Attachable | Storage}, // RG16S |
| 156 | {vk::Format::eUndefined, {}}, // RGB32F | 159 | {VK_FORMAT_UNDEFINED}, // RGB32F |
| 157 | {vk::Format::eR8G8B8A8Srgb, Attachable}, // RGBA8_SRGB | 160 | {VK_FORMAT_R8G8B8A8_SRGB, Attachable}, // RGBA8_SRGB |
| 158 | {vk::Format::eR8G8Unorm, Attachable | Storage}, // RG8U | 161 | {VK_FORMAT_R8G8_UNORM, Attachable | Storage}, // RG8U |
| 159 | {vk::Format::eR8G8Snorm, Attachable | Storage}, // RG8S | 162 | {VK_FORMAT_R8G8_SNORM, Attachable | Storage}, // RG8S |
| 160 | {vk::Format::eR32G32Uint, Attachable | Storage}, // RG32UI | 163 | {VK_FORMAT_R32G32_UINT, Attachable | Storage}, // RG32UI |
| 161 | {vk::Format::eUndefined, {}}, // RGBX16F | 164 | {VK_FORMAT_UNDEFINED}, // RGBX16F |
| 162 | {vk::Format::eR32Uint, Attachable | Storage}, // R32UI | 165 | {VK_FORMAT_R32_UINT, Attachable | Storage}, // R32UI |
| 163 | {vk::Format::eR32Sint, Attachable | Storage}, // R32I | 166 | {VK_FORMAT_R32_SINT, Attachable | Storage}, // R32I |
| 164 | {vk::Format::eAstc8x8UnormBlock, {}}, // ASTC_2D_8X8 | 167 | {VK_FORMAT_ASTC_8x8_UNORM_BLOCK}, // ASTC_2D_8X8 |
| 165 | {vk::Format::eUndefined, {}}, // ASTC_2D_8X5 | 168 | {VK_FORMAT_UNDEFINED}, // ASTC_2D_8X5 |
| 166 | {vk::Format::eUndefined, {}}, // ASTC_2D_5X4 | 169 | {VK_FORMAT_UNDEFINED}, // ASTC_2D_5X4 |
| 167 | {vk::Format::eUndefined, {}}, // BGRA8_SRGB | 170 | {VK_FORMAT_UNDEFINED}, // BGRA8_SRGB |
| 168 | {vk::Format::eBc1RgbaSrgbBlock, {}}, // DXT1_SRGB | 171 | {VK_FORMAT_BC1_RGBA_SRGB_BLOCK}, // DXT1_SRGB |
| 169 | {vk::Format::eBc2SrgbBlock, {}}, // DXT23_SRGB | 172 | {VK_FORMAT_BC2_SRGB_BLOCK}, // DXT23_SRGB |
| 170 | {vk::Format::eBc3SrgbBlock, {}}, // DXT45_SRGB | 173 | {VK_FORMAT_BC3_SRGB_BLOCK}, // DXT45_SRGB |
| 171 | {vk::Format::eBc7SrgbBlock, {}}, // BC7U_SRGB | 174 | {VK_FORMAT_BC7_SRGB_BLOCK}, // BC7U_SRGB |
| 172 | {vk::Format::eR4G4B4A4UnormPack16, Attachable}, // R4G4B4A4U | 175 | {VK_FORMAT_R4G4B4A4_UNORM_PACK16, Attachable}, // R4G4B4A4U |
| 173 | {vk::Format::eAstc4x4SrgbBlock, {}}, // ASTC_2D_4X4_SRGB | 176 | {VK_FORMAT_ASTC_4x4_SRGB_BLOCK}, // ASTC_2D_4X4_SRGB |
| 174 | {vk::Format::eAstc8x8SrgbBlock, {}}, // ASTC_2D_8X8_SRGB | 177 | {VK_FORMAT_ASTC_8x8_SRGB_BLOCK}, // ASTC_2D_8X8_SRGB |
| 175 | {vk::Format::eAstc8x5SrgbBlock, {}}, // ASTC_2D_8X5_SRGB | 178 | {VK_FORMAT_ASTC_8x5_SRGB_BLOCK}, // ASTC_2D_8X5_SRGB |
| 176 | {vk::Format::eAstc5x4SrgbBlock, {}}, // ASTC_2D_5X4_SRGB | 179 | {VK_FORMAT_ASTC_5x4_SRGB_BLOCK}, // ASTC_2D_5X4_SRGB |
| 177 | {vk::Format::eAstc5x5UnormBlock, {}}, // ASTC_2D_5X5 | 180 | {VK_FORMAT_ASTC_5x5_UNORM_BLOCK}, // ASTC_2D_5X5 |
| 178 | {vk::Format::eAstc5x5SrgbBlock, {}}, // ASTC_2D_5X5_SRGB | 181 | {VK_FORMAT_ASTC_5x5_SRGB_BLOCK}, // ASTC_2D_5X5_SRGB |
| 179 | {vk::Format::eAstc10x8UnormBlock, {}}, // ASTC_2D_10X8 | 182 | {VK_FORMAT_ASTC_10x8_UNORM_BLOCK}, // ASTC_2D_10X8 |
| 180 | {vk::Format::eAstc10x8SrgbBlock, {}}, // ASTC_2D_10X8_SRGB | 183 | {VK_FORMAT_ASTC_10x8_SRGB_BLOCK}, // ASTC_2D_10X8_SRGB |
| 181 | {vk::Format::eAstc6x6UnormBlock, {}}, // ASTC_2D_6X6 | 184 | {VK_FORMAT_ASTC_6x6_UNORM_BLOCK}, // ASTC_2D_6X6 |
| 182 | {vk::Format::eAstc6x6SrgbBlock, {}}, // ASTC_2D_6X6_SRGB | 185 | {VK_FORMAT_ASTC_6x6_SRGB_BLOCK}, // ASTC_2D_6X6_SRGB |
| 183 | {vk::Format::eAstc10x10UnormBlock, {}}, // ASTC_2D_10X10 | 186 | {VK_FORMAT_ASTC_10x10_UNORM_BLOCK}, // ASTC_2D_10X10 |
| 184 | {vk::Format::eAstc10x10SrgbBlock, {}}, // ASTC_2D_10X10_SRGB | 187 | {VK_FORMAT_ASTC_10x10_SRGB_BLOCK}, // ASTC_2D_10X10_SRGB |
| 185 | {vk::Format::eAstc12x12UnormBlock, {}}, // ASTC_2D_12X12 | 188 | {VK_FORMAT_ASTC_12x12_UNORM_BLOCK}, // ASTC_2D_12X12 |
| 186 | {vk::Format::eAstc12x12SrgbBlock, {}}, // ASTC_2D_12X12_SRGB | 189 | {VK_FORMAT_ASTC_12x12_SRGB_BLOCK}, // ASTC_2D_12X12_SRGB |
| 187 | {vk::Format::eAstc8x6UnormBlock, {}}, // ASTC_2D_8X6 | 190 | {VK_FORMAT_ASTC_8x6_UNORM_BLOCK}, // ASTC_2D_8X6 |
| 188 | {vk::Format::eAstc8x6SrgbBlock, {}}, // ASTC_2D_8X6_SRGB | 191 | {VK_FORMAT_ASTC_8x6_SRGB_BLOCK}, // ASTC_2D_8X6_SRGB |
| 189 | {vk::Format::eAstc6x5UnormBlock, {}}, // ASTC_2D_6X5 | 192 | {VK_FORMAT_ASTC_6x5_UNORM_BLOCK}, // ASTC_2D_6X5 |
| 190 | {vk::Format::eAstc6x5SrgbBlock, {}}, // ASTC_2D_6X5_SRGB | 193 | {VK_FORMAT_ASTC_6x5_SRGB_BLOCK}, // ASTC_2D_6X5_SRGB |
| 191 | {vk::Format::eE5B9G9R9UfloatPack32, {}}, // E5B9G9R9F | 194 | {VK_FORMAT_E5B9G9R9_UFLOAT_PACK32}, // E5B9G9R9F |
| 192 | 195 | ||
| 193 | // Depth formats | 196 | // Depth formats |
| 194 | {vk::Format::eD32Sfloat, Attachable}, // Z32F | 197 | {VK_FORMAT_D32_SFLOAT, Attachable}, // Z32F |
| 195 | {vk::Format::eD16Unorm, Attachable}, // Z16 | 198 | {VK_FORMAT_D16_UNORM, Attachable}, // Z16 |
| 196 | 199 | ||
| 197 | // DepthStencil formats | 200 | // DepthStencil formats |
| 198 | {vk::Format::eD24UnormS8Uint, Attachable}, // Z24S8 | 201 | {VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // Z24S8 |
| 199 | {vk::Format::eD24UnormS8Uint, Attachable}, // S8Z24 (emulated) | 202 | {VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // S8Z24 (emulated) |
| 200 | {vk::Format::eD32SfloatS8Uint, Attachable}, // Z32FS8 | 203 | {VK_FORMAT_D32_SFLOAT_S8_UINT, Attachable}, // Z32FS8 |
| 201 | }; | 204 | }; |
| 202 | static_assert(std::size(tex_format_tuples) == VideoCore::Surface::MaxPixelFormat); | 205 | static_assert(std::size(tex_format_tuples) == VideoCore::Surface::MaxPixelFormat); |
| 203 | 206 | ||
| @@ -212,106 +215,106 @@ FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFo | |||
| 212 | ASSERT(static_cast<std::size_t>(pixel_format) < std::size(tex_format_tuples)); | 215 | ASSERT(static_cast<std::size_t>(pixel_format) < std::size(tex_format_tuples)); |
| 213 | 216 | ||
| 214 | auto tuple = tex_format_tuples[static_cast<std::size_t>(pixel_format)]; | 217 | auto tuple = tex_format_tuples[static_cast<std::size_t>(pixel_format)]; |
| 215 | if (tuple.format == vk::Format::eUndefined) { | 218 | if (tuple.format == VK_FORMAT_UNDEFINED) { |
| 216 | UNIMPLEMENTED_MSG("Unimplemented texture format with pixel format={}", | 219 | UNIMPLEMENTED_MSG("Unimplemented texture format with pixel format={}", |
| 217 | static_cast<u32>(pixel_format)); | 220 | static_cast<u32>(pixel_format)); |
| 218 | return {vk::Format::eA8B8G8R8UnormPack32, true, true}; | 221 | return {VK_FORMAT_A8B8G8R8_UNORM_PACK32, true, true}; |
| 219 | } | 222 | } |
| 220 | 223 | ||
| 221 | // Use ABGR8 on hardware that doesn't support ASTC natively | 224 | // Use ABGR8 on hardware that doesn't support ASTC natively |
| 222 | if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) { | 225 | if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) { |
| 223 | tuple.format = VideoCore::Surface::IsPixelFormatSRGB(pixel_format) | 226 | tuple.format = VideoCore::Surface::IsPixelFormatSRGB(pixel_format) |
| 224 | ? vk::Format::eA8B8G8R8SrgbPack32 | 227 | ? VK_FORMAT_A8B8G8R8_SRGB_PACK32 |
| 225 | : vk::Format::eA8B8G8R8UnormPack32; | 228 | : VK_FORMAT_A8B8G8R8_UNORM_PACK32; |
| 226 | } | 229 | } |
| 227 | const bool attachable = tuple.usage & Attachable; | 230 | const bool attachable = tuple.usage & Attachable; |
| 228 | const bool storage = tuple.usage & Storage; | 231 | const bool storage = tuple.usage & Storage; |
| 229 | 232 | ||
| 230 | vk::FormatFeatureFlags usage; | 233 | VkFormatFeatureFlags usage; |
| 231 | if (format_type == FormatType::Buffer) { | 234 | if (format_type == FormatType::Buffer) { |
| 232 | usage = vk::FormatFeatureFlagBits::eStorageTexelBuffer | | 235 | usage = |
| 233 | vk::FormatFeatureFlagBits::eUniformTexelBuffer; | 236 | VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_BIT | VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT; |
| 234 | } else { | 237 | } else { |
| 235 | usage = vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eTransferDst | | 238 | usage = VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_TRANSFER_DST_BIT | |
| 236 | vk::FormatFeatureFlagBits::eTransferSrc; | 239 | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT; |
| 237 | if (attachable) { | 240 | if (attachable) { |
| 238 | usage |= IsZetaFormat(pixel_format) ? vk::FormatFeatureFlagBits::eDepthStencilAttachment | 241 | usage |= IsZetaFormat(pixel_format) ? VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT |
| 239 | : vk::FormatFeatureFlagBits::eColorAttachment; | 242 | : VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT; |
| 240 | } | 243 | } |
| 241 | if (storage) { | 244 | if (storage) { |
| 242 | usage |= vk::FormatFeatureFlagBits::eStorageImage; | 245 | usage |= VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT; |
| 243 | } | 246 | } |
| 244 | } | 247 | } |
| 245 | return {device.GetSupportedFormat(tuple.format, usage, format_type), attachable, storage}; | 248 | return {device.GetSupportedFormat(tuple.format, usage, format_type), attachable, storage}; |
| 246 | } | 249 | } |
| 247 | 250 | ||
| 248 | vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) { | 251 | VkShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) { |
| 249 | switch (stage) { | 252 | switch (stage) { |
| 250 | case Tegra::Engines::ShaderType::Vertex: | 253 | case Tegra::Engines::ShaderType::Vertex: |
| 251 | return vk::ShaderStageFlagBits::eVertex; | 254 | return VK_SHADER_STAGE_VERTEX_BIT; |
| 252 | case Tegra::Engines::ShaderType::TesselationControl: | 255 | case Tegra::Engines::ShaderType::TesselationControl: |
| 253 | return vk::ShaderStageFlagBits::eTessellationControl; | 256 | return VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT; |
| 254 | case Tegra::Engines::ShaderType::TesselationEval: | 257 | case Tegra::Engines::ShaderType::TesselationEval: |
| 255 | return vk::ShaderStageFlagBits::eTessellationEvaluation; | 258 | return VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT; |
| 256 | case Tegra::Engines::ShaderType::Geometry: | 259 | case Tegra::Engines::ShaderType::Geometry: |
| 257 | return vk::ShaderStageFlagBits::eGeometry; | 260 | return VK_SHADER_STAGE_GEOMETRY_BIT; |
| 258 | case Tegra::Engines::ShaderType::Fragment: | 261 | case Tegra::Engines::ShaderType::Fragment: |
| 259 | return vk::ShaderStageFlagBits::eFragment; | 262 | return VK_SHADER_STAGE_FRAGMENT_BIT; |
| 260 | case Tegra::Engines::ShaderType::Compute: | 263 | case Tegra::Engines::ShaderType::Compute: |
| 261 | return vk::ShaderStageFlagBits::eCompute; | 264 | return VK_SHADER_STAGE_COMPUTE_BIT; |
| 262 | } | 265 | } |
| 263 | UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage)); | 266 | UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage)); |
| 264 | return {}; | 267 | return {}; |
| 265 | } | 268 | } |
| 266 | 269 | ||
| 267 | vk::PrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device, | 270 | VkPrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device, |
| 268 | Maxwell::PrimitiveTopology topology) { | 271 | Maxwell::PrimitiveTopology topology) { |
| 269 | switch (topology) { | 272 | switch (topology) { |
| 270 | case Maxwell::PrimitiveTopology::Points: | 273 | case Maxwell::PrimitiveTopology::Points: |
| 271 | return vk::PrimitiveTopology::ePointList; | 274 | return VK_PRIMITIVE_TOPOLOGY_POINT_LIST; |
| 272 | case Maxwell::PrimitiveTopology::Lines: | 275 | case Maxwell::PrimitiveTopology::Lines: |
| 273 | return vk::PrimitiveTopology::eLineList; | 276 | return VK_PRIMITIVE_TOPOLOGY_LINE_LIST; |
| 274 | case Maxwell::PrimitiveTopology::LineStrip: | 277 | case Maxwell::PrimitiveTopology::LineStrip: |
| 275 | return vk::PrimitiveTopology::eLineStrip; | 278 | return VK_PRIMITIVE_TOPOLOGY_LINE_STRIP; |
| 276 | case Maxwell::PrimitiveTopology::Triangles: | 279 | case Maxwell::PrimitiveTopology::Triangles: |
| 277 | return vk::PrimitiveTopology::eTriangleList; | 280 | return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; |
| 278 | case Maxwell::PrimitiveTopology::TriangleStrip: | 281 | case Maxwell::PrimitiveTopology::TriangleStrip: |
| 279 | return vk::PrimitiveTopology::eTriangleStrip; | 282 | return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP; |
| 280 | case Maxwell::PrimitiveTopology::TriangleFan: | 283 | case Maxwell::PrimitiveTopology::TriangleFan: |
| 281 | return vk::PrimitiveTopology::eTriangleFan; | 284 | return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN; |
| 282 | case Maxwell::PrimitiveTopology::Quads: | 285 | case Maxwell::PrimitiveTopology::Quads: |
| 283 | // TODO(Rodrigo): Use VK_PRIMITIVE_TOPOLOGY_QUAD_LIST_EXT whenever it releases | 286 | // TODO(Rodrigo): Use VK_PRIMITIVE_TOPOLOGY_QUAD_LIST_EXT whenever it releases |
| 284 | return vk::PrimitiveTopology::eTriangleList; | 287 | return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; |
| 285 | case Maxwell::PrimitiveTopology::Patches: | 288 | case Maxwell::PrimitiveTopology::Patches: |
| 286 | return vk::PrimitiveTopology::ePatchList; | 289 | return VK_PRIMITIVE_TOPOLOGY_PATCH_LIST; |
| 287 | default: | 290 | default: |
| 288 | UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology)); | 291 | UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology)); |
| 289 | return {}; | 292 | return {}; |
| 290 | } | 293 | } |
| 291 | } | 294 | } |
| 292 | 295 | ||
| 293 | vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) { | 296 | VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) { |
| 294 | switch (type) { | 297 | switch (type) { |
| 295 | case Maxwell::VertexAttribute::Type::SignedNorm: | 298 | case Maxwell::VertexAttribute::Type::SignedNorm: |
| 296 | switch (size) { | 299 | switch (size) { |
| 297 | case Maxwell::VertexAttribute::Size::Size_8: | 300 | case Maxwell::VertexAttribute::Size::Size_8: |
| 298 | return vk::Format::eR8Snorm; | 301 | return VK_FORMAT_R8_SNORM; |
| 299 | case Maxwell::VertexAttribute::Size::Size_8_8: | 302 | case Maxwell::VertexAttribute::Size::Size_8_8: |
| 300 | return vk::Format::eR8G8Snorm; | 303 | return VK_FORMAT_R8G8_SNORM; |
| 301 | case Maxwell::VertexAttribute::Size::Size_8_8_8: | 304 | case Maxwell::VertexAttribute::Size::Size_8_8_8: |
| 302 | return vk::Format::eR8G8B8Snorm; | 305 | return VK_FORMAT_R8G8B8_SNORM; |
| 303 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: | 306 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: |
| 304 | return vk::Format::eR8G8B8A8Snorm; | 307 | return VK_FORMAT_R8G8B8A8_SNORM; |
| 305 | case Maxwell::VertexAttribute::Size::Size_16: | 308 | case Maxwell::VertexAttribute::Size::Size_16: |
| 306 | return vk::Format::eR16Snorm; | 309 | return VK_FORMAT_R16_SNORM; |
| 307 | case Maxwell::VertexAttribute::Size::Size_16_16: | 310 | case Maxwell::VertexAttribute::Size::Size_16_16: |
| 308 | return vk::Format::eR16G16Snorm; | 311 | return VK_FORMAT_R16G16_SNORM; |
| 309 | case Maxwell::VertexAttribute::Size::Size_16_16_16: | 312 | case Maxwell::VertexAttribute::Size::Size_16_16_16: |
| 310 | return vk::Format::eR16G16B16Snorm; | 313 | return VK_FORMAT_R16G16B16_SNORM; |
| 311 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: | 314 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: |
| 312 | return vk::Format::eR16G16B16A16Snorm; | 315 | return VK_FORMAT_R16G16B16A16_SNORM; |
| 313 | case Maxwell::VertexAttribute::Size::Size_10_10_10_2: | 316 | case Maxwell::VertexAttribute::Size::Size_10_10_10_2: |
| 314 | return vk::Format::eA2B10G10R10SnormPack32; | 317 | return VK_FORMAT_A2B10G10R10_SNORM_PACK32; |
| 315 | default: | 318 | default: |
| 316 | break; | 319 | break; |
| 317 | } | 320 | } |
| @@ -319,23 +322,23 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr | |||
| 319 | case Maxwell::VertexAttribute::Type::UnsignedNorm: | 322 | case Maxwell::VertexAttribute::Type::UnsignedNorm: |
| 320 | switch (size) { | 323 | switch (size) { |
| 321 | case Maxwell::VertexAttribute::Size::Size_8: | 324 | case Maxwell::VertexAttribute::Size::Size_8: |
| 322 | return vk::Format::eR8Unorm; | 325 | return VK_FORMAT_R8_UNORM; |
| 323 | case Maxwell::VertexAttribute::Size::Size_8_8: | 326 | case Maxwell::VertexAttribute::Size::Size_8_8: |
| 324 | return vk::Format::eR8G8Unorm; | 327 | return VK_FORMAT_R8G8_UNORM; |
| 325 | case Maxwell::VertexAttribute::Size::Size_8_8_8: | 328 | case Maxwell::VertexAttribute::Size::Size_8_8_8: |
| 326 | return vk::Format::eR8G8B8Unorm; | 329 | return VK_FORMAT_R8G8B8_UNORM; |
| 327 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: | 330 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: |
| 328 | return vk::Format::eR8G8B8A8Unorm; | 331 | return VK_FORMAT_R8G8B8A8_UNORM; |
| 329 | case Maxwell::VertexAttribute::Size::Size_16: | 332 | case Maxwell::VertexAttribute::Size::Size_16: |
| 330 | return vk::Format::eR16Unorm; | 333 | return VK_FORMAT_R16_UNORM; |
| 331 | case Maxwell::VertexAttribute::Size::Size_16_16: | 334 | case Maxwell::VertexAttribute::Size::Size_16_16: |
| 332 | return vk::Format::eR16G16Unorm; | 335 | return VK_FORMAT_R16G16_UNORM; |
| 333 | case Maxwell::VertexAttribute::Size::Size_16_16_16: | 336 | case Maxwell::VertexAttribute::Size::Size_16_16_16: |
| 334 | return vk::Format::eR16G16B16Unorm; | 337 | return VK_FORMAT_R16G16B16_UNORM; |
| 335 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: | 338 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: |
| 336 | return vk::Format::eR16G16B16A16Unorm; | 339 | return VK_FORMAT_R16G16B16A16_UNORM; |
| 337 | case Maxwell::VertexAttribute::Size::Size_10_10_10_2: | 340 | case Maxwell::VertexAttribute::Size::Size_10_10_10_2: |
| 338 | return vk::Format::eA2B10G10R10UnormPack32; | 341 | return VK_FORMAT_A2B10G10R10_UNORM_PACK32; |
| 339 | default: | 342 | default: |
| 340 | break; | 343 | break; |
| 341 | } | 344 | } |
| @@ -343,59 +346,59 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr | |||
| 343 | case Maxwell::VertexAttribute::Type::SignedInt: | 346 | case Maxwell::VertexAttribute::Type::SignedInt: |
| 344 | switch (size) { | 347 | switch (size) { |
| 345 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: | 348 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: |
| 346 | return vk::Format::eR16G16B16A16Sint; | 349 | return VK_FORMAT_R16G16B16A16_SINT; |
| 347 | case Maxwell::VertexAttribute::Size::Size_8: | 350 | case Maxwell::VertexAttribute::Size::Size_8: |
| 348 | return vk::Format::eR8Sint; | 351 | return VK_FORMAT_R8_SINT; |
| 349 | case Maxwell::VertexAttribute::Size::Size_8_8: | 352 | case Maxwell::VertexAttribute::Size::Size_8_8: |
| 350 | return vk::Format::eR8G8Sint; | 353 | return VK_FORMAT_R8G8_SINT; |
| 351 | case Maxwell::VertexAttribute::Size::Size_8_8_8: | 354 | case Maxwell::VertexAttribute::Size::Size_8_8_8: |
| 352 | return vk::Format::eR8G8B8Sint; | 355 | return VK_FORMAT_R8G8B8_SINT; |
| 353 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: | 356 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: |
| 354 | return vk::Format::eR8G8B8A8Sint; | 357 | return VK_FORMAT_R8G8B8A8_SINT; |
| 355 | case Maxwell::VertexAttribute::Size::Size_32: | 358 | case Maxwell::VertexAttribute::Size::Size_32: |
| 356 | return vk::Format::eR32Sint; | 359 | return VK_FORMAT_R32_SINT; |
| 357 | default: | 360 | default: |
| 358 | break; | 361 | break; |
| 359 | } | 362 | } |
| 360 | case Maxwell::VertexAttribute::Type::UnsignedInt: | 363 | case Maxwell::VertexAttribute::Type::UnsignedInt: |
| 361 | switch (size) { | 364 | switch (size) { |
| 362 | case Maxwell::VertexAttribute::Size::Size_8: | 365 | case Maxwell::VertexAttribute::Size::Size_8: |
| 363 | return vk::Format::eR8Uint; | 366 | return VK_FORMAT_R8_UINT; |
| 364 | case Maxwell::VertexAttribute::Size::Size_8_8: | 367 | case Maxwell::VertexAttribute::Size::Size_8_8: |
| 365 | return vk::Format::eR8G8Uint; | 368 | return VK_FORMAT_R8G8_UINT; |
| 366 | case Maxwell::VertexAttribute::Size::Size_8_8_8: | 369 | case Maxwell::VertexAttribute::Size::Size_8_8_8: |
| 367 | return vk::Format::eR8G8B8Uint; | 370 | return VK_FORMAT_R8G8B8_UINT; |
| 368 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: | 371 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: |
| 369 | return vk::Format::eR8G8B8A8Uint; | 372 | return VK_FORMAT_R8G8B8A8_UINT; |
| 370 | case Maxwell::VertexAttribute::Size::Size_32: | 373 | case Maxwell::VertexAttribute::Size::Size_32: |
| 371 | return vk::Format::eR32Uint; | 374 | return VK_FORMAT_R32_UINT; |
| 372 | case Maxwell::VertexAttribute::Size::Size_32_32: | 375 | case Maxwell::VertexAttribute::Size::Size_32_32: |
| 373 | return vk::Format::eR32G32Uint; | 376 | return VK_FORMAT_R32G32_UINT; |
| 374 | case Maxwell::VertexAttribute::Size::Size_32_32_32: | 377 | case Maxwell::VertexAttribute::Size::Size_32_32_32: |
| 375 | return vk::Format::eR32G32B32Uint; | 378 | return VK_FORMAT_R32G32B32_UINT; |
| 376 | case Maxwell::VertexAttribute::Size::Size_32_32_32_32: | 379 | case Maxwell::VertexAttribute::Size::Size_32_32_32_32: |
| 377 | return vk::Format::eR32G32B32A32Uint; | 380 | return VK_FORMAT_R32G32B32A32_UINT; |
| 378 | default: | 381 | default: |
| 379 | break; | 382 | break; |
| 380 | } | 383 | } |
| 381 | case Maxwell::VertexAttribute::Type::UnsignedScaled: | 384 | case Maxwell::VertexAttribute::Type::UnsignedScaled: |
| 382 | switch (size) { | 385 | switch (size) { |
| 383 | case Maxwell::VertexAttribute::Size::Size_8: | 386 | case Maxwell::VertexAttribute::Size::Size_8: |
| 384 | return vk::Format::eR8Uscaled; | 387 | return VK_FORMAT_R8_USCALED; |
| 385 | case Maxwell::VertexAttribute::Size::Size_8_8: | 388 | case Maxwell::VertexAttribute::Size::Size_8_8: |
| 386 | return vk::Format::eR8G8Uscaled; | 389 | return VK_FORMAT_R8G8_USCALED; |
| 387 | case Maxwell::VertexAttribute::Size::Size_8_8_8: | 390 | case Maxwell::VertexAttribute::Size::Size_8_8_8: |
| 388 | return vk::Format::eR8G8B8Uscaled; | 391 | return VK_FORMAT_R8G8B8_USCALED; |
| 389 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: | 392 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: |
| 390 | return vk::Format::eR8G8B8A8Uscaled; | 393 | return VK_FORMAT_R8G8B8A8_USCALED; |
| 391 | case Maxwell::VertexAttribute::Size::Size_16: | 394 | case Maxwell::VertexAttribute::Size::Size_16: |
| 392 | return vk::Format::eR16Uscaled; | 395 | return VK_FORMAT_R16_USCALED; |
| 393 | case Maxwell::VertexAttribute::Size::Size_16_16: | 396 | case Maxwell::VertexAttribute::Size::Size_16_16: |
| 394 | return vk::Format::eR16G16Uscaled; | 397 | return VK_FORMAT_R16G16_USCALED; |
| 395 | case Maxwell::VertexAttribute::Size::Size_16_16_16: | 398 | case Maxwell::VertexAttribute::Size::Size_16_16_16: |
| 396 | return vk::Format::eR16G16B16Uscaled; | 399 | return VK_FORMAT_R16G16B16_USCALED; |
| 397 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: | 400 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: |
| 398 | return vk::Format::eR16G16B16A16Uscaled; | 401 | return VK_FORMAT_R16G16B16A16_USCALED; |
| 399 | default: | 402 | default: |
| 400 | break; | 403 | break; |
| 401 | } | 404 | } |
| @@ -403,21 +406,21 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr | |||
| 403 | case Maxwell::VertexAttribute::Type::SignedScaled: | 406 | case Maxwell::VertexAttribute::Type::SignedScaled: |
| 404 | switch (size) { | 407 | switch (size) { |
| 405 | case Maxwell::VertexAttribute::Size::Size_8: | 408 | case Maxwell::VertexAttribute::Size::Size_8: |
| 406 | return vk::Format::eR8Sscaled; | 409 | return VK_FORMAT_R8_SSCALED; |
| 407 | case Maxwell::VertexAttribute::Size::Size_8_8: | 410 | case Maxwell::VertexAttribute::Size::Size_8_8: |
| 408 | return vk::Format::eR8G8Sscaled; | 411 | return VK_FORMAT_R8G8_SSCALED; |
| 409 | case Maxwell::VertexAttribute::Size::Size_8_8_8: | 412 | case Maxwell::VertexAttribute::Size::Size_8_8_8: |
| 410 | return vk::Format::eR8G8B8Sscaled; | 413 | return VK_FORMAT_R8G8B8_SSCALED; |
| 411 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: | 414 | case Maxwell::VertexAttribute::Size::Size_8_8_8_8: |
| 412 | return vk::Format::eR8G8B8A8Sscaled; | 415 | return VK_FORMAT_R8G8B8A8_SSCALED; |
| 413 | case Maxwell::VertexAttribute::Size::Size_16: | 416 | case Maxwell::VertexAttribute::Size::Size_16: |
| 414 | return vk::Format::eR16Sscaled; | 417 | return VK_FORMAT_R16_SSCALED; |
| 415 | case Maxwell::VertexAttribute::Size::Size_16_16: | 418 | case Maxwell::VertexAttribute::Size::Size_16_16: |
| 416 | return vk::Format::eR16G16Sscaled; | 419 | return VK_FORMAT_R16G16_SSCALED; |
| 417 | case Maxwell::VertexAttribute::Size::Size_16_16_16: | 420 | case Maxwell::VertexAttribute::Size::Size_16_16_16: |
| 418 | return vk::Format::eR16G16B16Sscaled; | 421 | return VK_FORMAT_R16G16B16_SSCALED; |
| 419 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: | 422 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: |
| 420 | return vk::Format::eR16G16B16A16Sscaled; | 423 | return VK_FORMAT_R16G16B16A16_SSCALED; |
| 421 | default: | 424 | default: |
| 422 | break; | 425 | break; |
| 423 | } | 426 | } |
| @@ -425,21 +428,21 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr | |||
| 425 | case Maxwell::VertexAttribute::Type::Float: | 428 | case Maxwell::VertexAttribute::Type::Float: |
| 426 | switch (size) { | 429 | switch (size) { |
| 427 | case Maxwell::VertexAttribute::Size::Size_32: | 430 | case Maxwell::VertexAttribute::Size::Size_32: |
| 428 | return vk::Format::eR32Sfloat; | 431 | return VK_FORMAT_R32_SFLOAT; |
| 429 | case Maxwell::VertexAttribute::Size::Size_32_32: | 432 | case Maxwell::VertexAttribute::Size::Size_32_32: |
| 430 | return vk::Format::eR32G32Sfloat; | 433 | return VK_FORMAT_R32G32_SFLOAT; |
| 431 | case Maxwell::VertexAttribute::Size::Size_32_32_32: | 434 | case Maxwell::VertexAttribute::Size::Size_32_32_32: |
| 432 | return vk::Format::eR32G32B32Sfloat; | 435 | return VK_FORMAT_R32G32B32_SFLOAT; |
| 433 | case Maxwell::VertexAttribute::Size::Size_32_32_32_32: | 436 | case Maxwell::VertexAttribute::Size::Size_32_32_32_32: |
| 434 | return vk::Format::eR32G32B32A32Sfloat; | 437 | return VK_FORMAT_R32G32B32A32_SFLOAT; |
| 435 | case Maxwell::VertexAttribute::Size::Size_16: | 438 | case Maxwell::VertexAttribute::Size::Size_16: |
| 436 | return vk::Format::eR16Sfloat; | 439 | return VK_FORMAT_R16_SFLOAT; |
| 437 | case Maxwell::VertexAttribute::Size::Size_16_16: | 440 | case Maxwell::VertexAttribute::Size::Size_16_16: |
| 438 | return vk::Format::eR16G16Sfloat; | 441 | return VK_FORMAT_R16G16_SFLOAT; |
| 439 | case Maxwell::VertexAttribute::Size::Size_16_16_16: | 442 | case Maxwell::VertexAttribute::Size::Size_16_16_16: |
| 440 | return vk::Format::eR16G16B16Sfloat; | 443 | return VK_FORMAT_R16G16B16_SFLOAT; |
| 441 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: | 444 | case Maxwell::VertexAttribute::Size::Size_16_16_16_16: |
| 442 | return vk::Format::eR16G16B16A16Sfloat; | 445 | return VK_FORMAT_R16G16B16A16_SFLOAT; |
| 443 | default: | 446 | default: |
| 444 | break; | 447 | break; |
| 445 | } | 448 | } |
| @@ -450,210 +453,210 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr | |||
| 450 | return {}; | 453 | return {}; |
| 451 | } | 454 | } |
| 452 | 455 | ||
| 453 | vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison) { | 456 | VkCompareOp ComparisonOp(Maxwell::ComparisonOp comparison) { |
| 454 | switch (comparison) { | 457 | switch (comparison) { |
| 455 | case Maxwell::ComparisonOp::Never: | 458 | case Maxwell::ComparisonOp::Never: |
| 456 | case Maxwell::ComparisonOp::NeverOld: | 459 | case Maxwell::ComparisonOp::NeverOld: |
| 457 | return vk::CompareOp::eNever; | 460 | return VK_COMPARE_OP_NEVER; |
| 458 | case Maxwell::ComparisonOp::Less: | 461 | case Maxwell::ComparisonOp::Less: |
| 459 | case Maxwell::ComparisonOp::LessOld: | 462 | case Maxwell::ComparisonOp::LessOld: |
| 460 | return vk::CompareOp::eLess; | 463 | return VK_COMPARE_OP_LESS; |
| 461 | case Maxwell::ComparisonOp::Equal: | 464 | case Maxwell::ComparisonOp::Equal: |
| 462 | case Maxwell::ComparisonOp::EqualOld: | 465 | case Maxwell::ComparisonOp::EqualOld: |
| 463 | return vk::CompareOp::eEqual; | 466 | return VK_COMPARE_OP_EQUAL; |
| 464 | case Maxwell::ComparisonOp::LessEqual: | 467 | case Maxwell::ComparisonOp::LessEqual: |
| 465 | case Maxwell::ComparisonOp::LessEqualOld: | 468 | case Maxwell::ComparisonOp::LessEqualOld: |
| 466 | return vk::CompareOp::eLessOrEqual; | 469 | return VK_COMPARE_OP_LESS_OR_EQUAL; |
| 467 | case Maxwell::ComparisonOp::Greater: | 470 | case Maxwell::ComparisonOp::Greater: |
| 468 | case Maxwell::ComparisonOp::GreaterOld: | 471 | case Maxwell::ComparisonOp::GreaterOld: |
| 469 | return vk::CompareOp::eGreater; | 472 | return VK_COMPARE_OP_GREATER; |
| 470 | case Maxwell::ComparisonOp::NotEqual: | 473 | case Maxwell::ComparisonOp::NotEqual: |
| 471 | case Maxwell::ComparisonOp::NotEqualOld: | 474 | case Maxwell::ComparisonOp::NotEqualOld: |
| 472 | return vk::CompareOp::eNotEqual; | 475 | return VK_COMPARE_OP_NOT_EQUAL; |
| 473 | case Maxwell::ComparisonOp::GreaterEqual: | 476 | case Maxwell::ComparisonOp::GreaterEqual: |
| 474 | case Maxwell::ComparisonOp::GreaterEqualOld: | 477 | case Maxwell::ComparisonOp::GreaterEqualOld: |
| 475 | return vk::CompareOp::eGreaterOrEqual; | 478 | return VK_COMPARE_OP_GREATER_OR_EQUAL; |
| 476 | case Maxwell::ComparisonOp::Always: | 479 | case Maxwell::ComparisonOp::Always: |
| 477 | case Maxwell::ComparisonOp::AlwaysOld: | 480 | case Maxwell::ComparisonOp::AlwaysOld: |
| 478 | return vk::CompareOp::eAlways; | 481 | return VK_COMPARE_OP_ALWAYS; |
| 479 | } | 482 | } |
| 480 | UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison)); | 483 | UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison)); |
| 481 | return {}; | 484 | return {}; |
| 482 | } | 485 | } |
| 483 | 486 | ||
| 484 | vk::IndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format) { | 487 | VkIndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format) { |
| 485 | switch (index_format) { | 488 | switch (index_format) { |
| 486 | case Maxwell::IndexFormat::UnsignedByte: | 489 | case Maxwell::IndexFormat::UnsignedByte: |
| 487 | if (!device.IsExtIndexTypeUint8Supported()) { | 490 | if (!device.IsExtIndexTypeUint8Supported()) { |
| 488 | UNIMPLEMENTED_MSG("Native uint8 indices are not supported on this device"); | 491 | UNIMPLEMENTED_MSG("Native uint8 indices are not supported on this device"); |
| 489 | return vk::IndexType::eUint16; | 492 | return VK_INDEX_TYPE_UINT16; |
| 490 | } | 493 | } |
| 491 | return vk::IndexType::eUint8EXT; | 494 | return VK_INDEX_TYPE_UINT8_EXT; |
| 492 | case Maxwell::IndexFormat::UnsignedShort: | 495 | case Maxwell::IndexFormat::UnsignedShort: |
| 493 | return vk::IndexType::eUint16; | 496 | return VK_INDEX_TYPE_UINT16; |
| 494 | case Maxwell::IndexFormat::UnsignedInt: | 497 | case Maxwell::IndexFormat::UnsignedInt: |
| 495 | return vk::IndexType::eUint32; | 498 | return VK_INDEX_TYPE_UINT32; |
| 496 | } | 499 | } |
| 497 | UNIMPLEMENTED_MSG("Unimplemented index_format={}", static_cast<u32>(index_format)); | 500 | UNIMPLEMENTED_MSG("Unimplemented index_format={}", static_cast<u32>(index_format)); |
| 498 | return {}; | 501 | return {}; |
| 499 | } | 502 | } |
| 500 | 503 | ||
| 501 | vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op) { | 504 | VkStencilOp StencilOp(Maxwell::StencilOp stencil_op) { |
| 502 | switch (stencil_op) { | 505 | switch (stencil_op) { |
| 503 | case Maxwell::StencilOp::Keep: | 506 | case Maxwell::StencilOp::Keep: |
| 504 | case Maxwell::StencilOp::KeepOGL: | 507 | case Maxwell::StencilOp::KeepOGL: |
| 505 | return vk::StencilOp::eKeep; | 508 | return VK_STENCIL_OP_KEEP; |
| 506 | case Maxwell::StencilOp::Zero: | 509 | case Maxwell::StencilOp::Zero: |
| 507 | case Maxwell::StencilOp::ZeroOGL: | 510 | case Maxwell::StencilOp::ZeroOGL: |
| 508 | return vk::StencilOp::eZero; | 511 | return VK_STENCIL_OP_ZERO; |
| 509 | case Maxwell::StencilOp::Replace: | 512 | case Maxwell::StencilOp::Replace: |
| 510 | case Maxwell::StencilOp::ReplaceOGL: | 513 | case Maxwell::StencilOp::ReplaceOGL: |
| 511 | return vk::StencilOp::eReplace; | 514 | return VK_STENCIL_OP_REPLACE; |
| 512 | case Maxwell::StencilOp::Incr: | 515 | case Maxwell::StencilOp::Incr: |
| 513 | case Maxwell::StencilOp::IncrOGL: | 516 | case Maxwell::StencilOp::IncrOGL: |
| 514 | return vk::StencilOp::eIncrementAndClamp; | 517 | return VK_STENCIL_OP_INCREMENT_AND_CLAMP; |
| 515 | case Maxwell::StencilOp::Decr: | 518 | case Maxwell::StencilOp::Decr: |
| 516 | case Maxwell::StencilOp::DecrOGL: | 519 | case Maxwell::StencilOp::DecrOGL: |
| 517 | return vk::StencilOp::eDecrementAndClamp; | 520 | return VK_STENCIL_OP_DECREMENT_AND_CLAMP; |
| 518 | case Maxwell::StencilOp::Invert: | 521 | case Maxwell::StencilOp::Invert: |
| 519 | case Maxwell::StencilOp::InvertOGL: | 522 | case Maxwell::StencilOp::InvertOGL: |
| 520 | return vk::StencilOp::eInvert; | 523 | return VK_STENCIL_OP_INVERT; |
| 521 | case Maxwell::StencilOp::IncrWrap: | 524 | case Maxwell::StencilOp::IncrWrap: |
| 522 | case Maxwell::StencilOp::IncrWrapOGL: | 525 | case Maxwell::StencilOp::IncrWrapOGL: |
| 523 | return vk::StencilOp::eIncrementAndWrap; | 526 | return VK_STENCIL_OP_INCREMENT_AND_WRAP; |
| 524 | case Maxwell::StencilOp::DecrWrap: | 527 | case Maxwell::StencilOp::DecrWrap: |
| 525 | case Maxwell::StencilOp::DecrWrapOGL: | 528 | case Maxwell::StencilOp::DecrWrapOGL: |
| 526 | return vk::StencilOp::eDecrementAndWrap; | 529 | return VK_STENCIL_OP_DECREMENT_AND_WRAP; |
| 527 | } | 530 | } |
| 528 | UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil_op)); | 531 | UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil_op)); |
| 529 | return {}; | 532 | return {}; |
| 530 | } | 533 | } |
| 531 | 534 | ||
| 532 | vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation) { | 535 | VkBlendOp BlendEquation(Maxwell::Blend::Equation equation) { |
| 533 | switch (equation) { | 536 | switch (equation) { |
| 534 | case Maxwell::Blend::Equation::Add: | 537 | case Maxwell::Blend::Equation::Add: |
| 535 | case Maxwell::Blend::Equation::AddGL: | 538 | case Maxwell::Blend::Equation::AddGL: |
| 536 | return vk::BlendOp::eAdd; | 539 | return VK_BLEND_OP_ADD; |
| 537 | case Maxwell::Blend::Equation::Subtract: | 540 | case Maxwell::Blend::Equation::Subtract: |
| 538 | case Maxwell::Blend::Equation::SubtractGL: | 541 | case Maxwell::Blend::Equation::SubtractGL: |
| 539 | return vk::BlendOp::eSubtract; | 542 | return VK_BLEND_OP_SUBTRACT; |
| 540 | case Maxwell::Blend::Equation::ReverseSubtract: | 543 | case Maxwell::Blend::Equation::ReverseSubtract: |
| 541 | case Maxwell::Blend::Equation::ReverseSubtractGL: | 544 | case Maxwell::Blend::Equation::ReverseSubtractGL: |
| 542 | return vk::BlendOp::eReverseSubtract; | 545 | return VK_BLEND_OP_REVERSE_SUBTRACT; |
| 543 | case Maxwell::Blend::Equation::Min: | 546 | case Maxwell::Blend::Equation::Min: |
| 544 | case Maxwell::Blend::Equation::MinGL: | 547 | case Maxwell::Blend::Equation::MinGL: |
| 545 | return vk::BlendOp::eMin; | 548 | return VK_BLEND_OP_MIN; |
| 546 | case Maxwell::Blend::Equation::Max: | 549 | case Maxwell::Blend::Equation::Max: |
| 547 | case Maxwell::Blend::Equation::MaxGL: | 550 | case Maxwell::Blend::Equation::MaxGL: |
| 548 | return vk::BlendOp::eMax; | 551 | return VK_BLEND_OP_MAX; |
| 549 | } | 552 | } |
| 550 | UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation)); | 553 | UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation)); |
| 551 | return {}; | 554 | return {}; |
| 552 | } | 555 | } |
| 553 | 556 | ||
| 554 | vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) { | 557 | VkBlendFactor BlendFactor(Maxwell::Blend::Factor factor) { |
| 555 | switch (factor) { | 558 | switch (factor) { |
| 556 | case Maxwell::Blend::Factor::Zero: | 559 | case Maxwell::Blend::Factor::Zero: |
| 557 | case Maxwell::Blend::Factor::ZeroGL: | 560 | case Maxwell::Blend::Factor::ZeroGL: |
| 558 | return vk::BlendFactor::eZero; | 561 | return VK_BLEND_FACTOR_ZERO; |
| 559 | case Maxwell::Blend::Factor::One: | 562 | case Maxwell::Blend::Factor::One: |
| 560 | case Maxwell::Blend::Factor::OneGL: | 563 | case Maxwell::Blend::Factor::OneGL: |
| 561 | return vk::BlendFactor::eOne; | 564 | return VK_BLEND_FACTOR_ONE; |
| 562 | case Maxwell::Blend::Factor::SourceColor: | 565 | case Maxwell::Blend::Factor::SourceColor: |
| 563 | case Maxwell::Blend::Factor::SourceColorGL: | 566 | case Maxwell::Blend::Factor::SourceColorGL: |
| 564 | return vk::BlendFactor::eSrcColor; | 567 | return VK_BLEND_FACTOR_SRC_COLOR; |
| 565 | case Maxwell::Blend::Factor::OneMinusSourceColor: | 568 | case Maxwell::Blend::Factor::OneMinusSourceColor: |
| 566 | case Maxwell::Blend::Factor::OneMinusSourceColorGL: | 569 | case Maxwell::Blend::Factor::OneMinusSourceColorGL: |
| 567 | return vk::BlendFactor::eOneMinusSrcColor; | 570 | return VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR; |
| 568 | case Maxwell::Blend::Factor::SourceAlpha: | 571 | case Maxwell::Blend::Factor::SourceAlpha: |
| 569 | case Maxwell::Blend::Factor::SourceAlphaGL: | 572 | case Maxwell::Blend::Factor::SourceAlphaGL: |
| 570 | return vk::BlendFactor::eSrcAlpha; | 573 | return VK_BLEND_FACTOR_SRC_ALPHA; |
| 571 | case Maxwell::Blend::Factor::OneMinusSourceAlpha: | 574 | case Maxwell::Blend::Factor::OneMinusSourceAlpha: |
| 572 | case Maxwell::Blend::Factor::OneMinusSourceAlphaGL: | 575 | case Maxwell::Blend::Factor::OneMinusSourceAlphaGL: |
| 573 | return vk::BlendFactor::eOneMinusSrcAlpha; | 576 | return VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; |
| 574 | case Maxwell::Blend::Factor::DestAlpha: | 577 | case Maxwell::Blend::Factor::DestAlpha: |
| 575 | case Maxwell::Blend::Factor::DestAlphaGL: | 578 | case Maxwell::Blend::Factor::DestAlphaGL: |
| 576 | return vk::BlendFactor::eDstAlpha; | 579 | return VK_BLEND_FACTOR_DST_ALPHA; |
| 577 | case Maxwell::Blend::Factor::OneMinusDestAlpha: | 580 | case Maxwell::Blend::Factor::OneMinusDestAlpha: |
| 578 | case Maxwell::Blend::Factor::OneMinusDestAlphaGL: | 581 | case Maxwell::Blend::Factor::OneMinusDestAlphaGL: |
| 579 | return vk::BlendFactor::eOneMinusDstAlpha; | 582 | return VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA; |
| 580 | case Maxwell::Blend::Factor::DestColor: | 583 | case Maxwell::Blend::Factor::DestColor: |
| 581 | case Maxwell::Blend::Factor::DestColorGL: | 584 | case Maxwell::Blend::Factor::DestColorGL: |
| 582 | return vk::BlendFactor::eDstColor; | 585 | return VK_BLEND_FACTOR_DST_COLOR; |
| 583 | case Maxwell::Blend::Factor::OneMinusDestColor: | 586 | case Maxwell::Blend::Factor::OneMinusDestColor: |
| 584 | case Maxwell::Blend::Factor::OneMinusDestColorGL: | 587 | case Maxwell::Blend::Factor::OneMinusDestColorGL: |
| 585 | return vk::BlendFactor::eOneMinusDstColor; | 588 | return VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR; |
| 586 | case Maxwell::Blend::Factor::SourceAlphaSaturate: | 589 | case Maxwell::Blend::Factor::SourceAlphaSaturate: |
| 587 | case Maxwell::Blend::Factor::SourceAlphaSaturateGL: | 590 | case Maxwell::Blend::Factor::SourceAlphaSaturateGL: |
| 588 | return vk::BlendFactor::eSrcAlphaSaturate; | 591 | return VK_BLEND_FACTOR_SRC_ALPHA_SATURATE; |
| 589 | case Maxwell::Blend::Factor::Source1Color: | 592 | case Maxwell::Blend::Factor::Source1Color: |
| 590 | case Maxwell::Blend::Factor::Source1ColorGL: | 593 | case Maxwell::Blend::Factor::Source1ColorGL: |
| 591 | return vk::BlendFactor::eSrc1Color; | 594 | return VK_BLEND_FACTOR_SRC1_COLOR; |
| 592 | case Maxwell::Blend::Factor::OneMinusSource1Color: | 595 | case Maxwell::Blend::Factor::OneMinusSource1Color: |
| 593 | case Maxwell::Blend::Factor::OneMinusSource1ColorGL: | 596 | case Maxwell::Blend::Factor::OneMinusSource1ColorGL: |
| 594 | return vk::BlendFactor::eOneMinusSrc1Color; | 597 | return VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR; |
| 595 | case Maxwell::Blend::Factor::Source1Alpha: | 598 | case Maxwell::Blend::Factor::Source1Alpha: |
| 596 | case Maxwell::Blend::Factor::Source1AlphaGL: | 599 | case Maxwell::Blend::Factor::Source1AlphaGL: |
| 597 | return vk::BlendFactor::eSrc1Alpha; | 600 | return VK_BLEND_FACTOR_SRC1_ALPHA; |
| 598 | case Maxwell::Blend::Factor::OneMinusSource1Alpha: | 601 | case Maxwell::Blend::Factor::OneMinusSource1Alpha: |
| 599 | case Maxwell::Blend::Factor::OneMinusSource1AlphaGL: | 602 | case Maxwell::Blend::Factor::OneMinusSource1AlphaGL: |
| 600 | return vk::BlendFactor::eOneMinusSrc1Alpha; | 603 | return VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA; |
| 601 | case Maxwell::Blend::Factor::ConstantColor: | 604 | case Maxwell::Blend::Factor::ConstantColor: |
| 602 | case Maxwell::Blend::Factor::ConstantColorGL: | 605 | case Maxwell::Blend::Factor::ConstantColorGL: |
| 603 | return vk::BlendFactor::eConstantColor; | 606 | return VK_BLEND_FACTOR_CONSTANT_COLOR; |
| 604 | case Maxwell::Blend::Factor::OneMinusConstantColor: | 607 | case Maxwell::Blend::Factor::OneMinusConstantColor: |
| 605 | case Maxwell::Blend::Factor::OneMinusConstantColorGL: | 608 | case Maxwell::Blend::Factor::OneMinusConstantColorGL: |
| 606 | return vk::BlendFactor::eOneMinusConstantColor; | 609 | return VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR; |
| 607 | case Maxwell::Blend::Factor::ConstantAlpha: | 610 | case Maxwell::Blend::Factor::ConstantAlpha: |
| 608 | case Maxwell::Blend::Factor::ConstantAlphaGL: | 611 | case Maxwell::Blend::Factor::ConstantAlphaGL: |
| 609 | return vk::BlendFactor::eConstantAlpha; | 612 | return VK_BLEND_FACTOR_CONSTANT_ALPHA; |
| 610 | case Maxwell::Blend::Factor::OneMinusConstantAlpha: | 613 | case Maxwell::Blend::Factor::OneMinusConstantAlpha: |
| 611 | case Maxwell::Blend::Factor::OneMinusConstantAlphaGL: | 614 | case Maxwell::Blend::Factor::OneMinusConstantAlphaGL: |
| 612 | return vk::BlendFactor::eOneMinusConstantAlpha; | 615 | return VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA; |
| 613 | } | 616 | } |
| 614 | UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor)); | 617 | UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor)); |
| 615 | return {}; | 618 | return {}; |
| 616 | } | 619 | } |
| 617 | 620 | ||
| 618 | vk::FrontFace FrontFace(Maxwell::FrontFace front_face) { | 621 | VkFrontFace FrontFace(Maxwell::FrontFace front_face) { |
| 619 | switch (front_face) { | 622 | switch (front_face) { |
| 620 | case Maxwell::FrontFace::ClockWise: | 623 | case Maxwell::FrontFace::ClockWise: |
| 621 | return vk::FrontFace::eClockwise; | 624 | return VK_FRONT_FACE_CLOCKWISE; |
| 622 | case Maxwell::FrontFace::CounterClockWise: | 625 | case Maxwell::FrontFace::CounterClockWise: |
| 623 | return vk::FrontFace::eCounterClockwise; | 626 | return VK_FRONT_FACE_COUNTER_CLOCKWISE; |
| 624 | } | 627 | } |
| 625 | UNIMPLEMENTED_MSG("Unimplemented front face={}", static_cast<u32>(front_face)); | 628 | UNIMPLEMENTED_MSG("Unimplemented front face={}", static_cast<u32>(front_face)); |
| 626 | return {}; | 629 | return {}; |
| 627 | } | 630 | } |
| 628 | 631 | ||
| 629 | vk::CullModeFlags CullFace(Maxwell::CullFace cull_face) { | 632 | VkCullModeFlags CullFace(Maxwell::CullFace cull_face) { |
| 630 | switch (cull_face) { | 633 | switch (cull_face) { |
| 631 | case Maxwell::CullFace::Front: | 634 | case Maxwell::CullFace::Front: |
| 632 | return vk::CullModeFlagBits::eFront; | 635 | return VK_CULL_MODE_FRONT_BIT; |
| 633 | case Maxwell::CullFace::Back: | 636 | case Maxwell::CullFace::Back: |
| 634 | return vk::CullModeFlagBits::eBack; | 637 | return VK_CULL_MODE_BACK_BIT; |
| 635 | case Maxwell::CullFace::FrontAndBack: | 638 | case Maxwell::CullFace::FrontAndBack: |
| 636 | return vk::CullModeFlagBits::eFrontAndBack; | 639 | return VK_CULL_MODE_FRONT_AND_BACK; |
| 637 | } | 640 | } |
| 638 | UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face)); | 641 | UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face)); |
| 639 | return {}; | 642 | return {}; |
| 640 | } | 643 | } |
| 641 | 644 | ||
| 642 | vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) { | 645 | VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) { |
| 643 | switch (swizzle) { | 646 | switch (swizzle) { |
| 644 | case Tegra::Texture::SwizzleSource::Zero: | 647 | case Tegra::Texture::SwizzleSource::Zero: |
| 645 | return vk::ComponentSwizzle::eZero; | 648 | return VK_COMPONENT_SWIZZLE_ZERO; |
| 646 | case Tegra::Texture::SwizzleSource::R: | 649 | case Tegra::Texture::SwizzleSource::R: |
| 647 | return vk::ComponentSwizzle::eR; | 650 | return VK_COMPONENT_SWIZZLE_R; |
| 648 | case Tegra::Texture::SwizzleSource::G: | 651 | case Tegra::Texture::SwizzleSource::G: |
| 649 | return vk::ComponentSwizzle::eG; | 652 | return VK_COMPONENT_SWIZZLE_G; |
| 650 | case Tegra::Texture::SwizzleSource::B: | 653 | case Tegra::Texture::SwizzleSource::B: |
| 651 | return vk::ComponentSwizzle::eB; | 654 | return VK_COMPONENT_SWIZZLE_B; |
| 652 | case Tegra::Texture::SwizzleSource::A: | 655 | case Tegra::Texture::SwizzleSource::A: |
| 653 | return vk::ComponentSwizzle::eA; | 656 | return VK_COMPONENT_SWIZZLE_A; |
| 654 | case Tegra::Texture::SwizzleSource::OneInt: | 657 | case Tegra::Texture::SwizzleSource::OneInt: |
| 655 | case Tegra::Texture::SwizzleSource::OneFloat: | 658 | case Tegra::Texture::SwizzleSource::OneFloat: |
| 656 | return vk::ComponentSwizzle::eOne; | 659 | return VK_COMPONENT_SWIZZLE_ONE; |
| 657 | } | 660 | } |
| 658 | UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(swizzle)); | 661 | UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(swizzle)); |
| 659 | return {}; | 662 | return {}; |
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.h b/src/video_core/renderer_vulkan/maxwell_to_vk.h index 24f6ab544..81bce4c6c 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.h +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.h | |||
| @@ -6,8 +6,8 @@ | |||
| 6 | 6 | ||
| 7 | #include "common/common_types.h" | 7 | #include "common/common_types.h" |
| 8 | #include "video_core/engines/maxwell_3d.h" | 8 | #include "video_core/engines/maxwell_3d.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/vk_device.h" | 9 | #include "video_core/renderer_vulkan/vk_device.h" |
| 10 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 11 | #include "video_core/surface.h" | 11 | #include "video_core/surface.h" |
| 12 | #include "video_core/textures/texture.h" | 12 | #include "video_core/textures/texture.h" |
| 13 | 13 | ||
| @@ -18,46 +18,45 @@ using PixelFormat = VideoCore::Surface::PixelFormat; | |||
| 18 | 18 | ||
| 19 | namespace Sampler { | 19 | namespace Sampler { |
| 20 | 20 | ||
| 21 | vk::Filter Filter(Tegra::Texture::TextureFilter filter); | 21 | VkFilter Filter(Tegra::Texture::TextureFilter filter); |
| 22 | 22 | ||
| 23 | vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter); | 23 | VkSamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter); |
| 24 | 24 | ||
| 25 | vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, | 25 | VkSamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, |
| 26 | Tegra::Texture::TextureFilter filter); | 26 | Tegra::Texture::TextureFilter filter); |
| 27 | 27 | ||
| 28 | vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func); | 28 | VkCompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func); |
| 29 | 29 | ||
| 30 | } // namespace Sampler | 30 | } // namespace Sampler |
| 31 | 31 | ||
| 32 | struct FormatInfo { | 32 | struct FormatInfo { |
| 33 | vk::Format format; | 33 | VkFormat format; |
| 34 | bool attachable; | 34 | bool attachable; |
| 35 | bool storage; | 35 | bool storage; |
| 36 | }; | 36 | }; |
| 37 | 37 | ||
| 38 | FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFormat pixel_format); | 38 | FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFormat pixel_format); |
| 39 | 39 | ||
| 40 | vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage); | 40 | VkShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage); |
| 41 | 41 | ||
| 42 | vk::PrimitiveTopology PrimitiveTopology(const VKDevice& device, | 42 | VkPrimitiveTopology PrimitiveTopology(const VKDevice& device, Maxwell::PrimitiveTopology topology); |
| 43 | Maxwell::PrimitiveTopology topology); | ||
| 44 | 43 | ||
| 45 | vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size); | 44 | VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size); |
| 46 | 45 | ||
| 47 | vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison); | 46 | VkCompareOp ComparisonOp(Maxwell::ComparisonOp comparison); |
| 48 | 47 | ||
| 49 | vk::IndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format); | 48 | VkIndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format); |
| 50 | 49 | ||
| 51 | vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op); | 50 | VkStencilOp StencilOp(Maxwell::StencilOp stencil_op); |
| 52 | 51 | ||
| 53 | vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation); | 52 | VkBlendOp BlendEquation(Maxwell::Blend::Equation equation); |
| 54 | 53 | ||
| 55 | vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor); | 54 | VkBlendFactor BlendFactor(Maxwell::Blend::Factor factor); |
| 56 | 55 | ||
| 57 | vk::FrontFace FrontFace(Maxwell::FrontFace front_face); | 56 | VkFrontFace FrontFace(Maxwell::FrontFace front_face); |
| 58 | 57 | ||
| 59 | vk::CullModeFlags CullFace(Maxwell::CullFace cull_face); | 58 | VkCullModeFlags CullFace(Maxwell::CullFace cull_face); |
| 60 | 59 | ||
| 61 | vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle); | 60 | VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle); |
| 62 | 61 | ||
| 63 | } // namespace Vulkan::MaxwellToVK | 62 | } // namespace Vulkan::MaxwellToVK |
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 9cdb4b627..dd590c38b 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp | |||
| @@ -24,7 +24,6 @@ | |||
| 24 | #include "core/settings.h" | 24 | #include "core/settings.h" |
| 25 | #include "core/telemetry_session.h" | 25 | #include "core/telemetry_session.h" |
| 26 | #include "video_core/gpu.h" | 26 | #include "video_core/gpu.h" |
| 27 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 28 | #include "video_core/renderer_vulkan/renderer_vulkan.h" | 27 | #include "video_core/renderer_vulkan/renderer_vulkan.h" |
| 29 | #include "video_core/renderer_vulkan/vk_blit_screen.h" | 28 | #include "video_core/renderer_vulkan/vk_blit_screen.h" |
| 30 | #include "video_core/renderer_vulkan/vk_device.h" | 29 | #include "video_core/renderer_vulkan/vk_device.h" |
| @@ -34,8 +33,9 @@ | |||
| 34 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 33 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 35 | #include "video_core/renderer_vulkan/vk_state_tracker.h" | 34 | #include "video_core/renderer_vulkan/vk_state_tracker.h" |
| 36 | #include "video_core/renderer_vulkan/vk_swapchain.h" | 35 | #include "video_core/renderer_vulkan/vk_swapchain.h" |
| 36 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 37 | 37 | ||
| 38 | // Include these late to avoid changing Vulkan-Hpp's dynamic dispatcher size | 38 | // Include these late to avoid polluting previous headers |
| 39 | #ifdef _WIN32 | 39 | #ifdef _WIN32 |
| 40 | #include <windows.h> | 40 | #include <windows.h> |
| 41 | // ensure include order | 41 | // ensure include order |
| @@ -54,20 +54,19 @@ namespace { | |||
| 54 | 54 | ||
| 55 | using Core::Frontend::WindowSystemType; | 55 | using Core::Frontend::WindowSystemType; |
| 56 | 56 | ||
| 57 | VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_, | 57 | VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity, |
| 58 | VkDebugUtilsMessageTypeFlagsEXT type, | 58 | VkDebugUtilsMessageTypeFlagsEXT type, |
| 59 | const VkDebugUtilsMessengerCallbackDataEXT* data, | 59 | const VkDebugUtilsMessengerCallbackDataEXT* data, |
| 60 | [[maybe_unused]] void* user_data) { | 60 | [[maybe_unused]] void* user_data) { |
| 61 | const auto severity{static_cast<vk::DebugUtilsMessageSeverityFlagBitsEXT>(severity_)}; | ||
| 62 | const char* message{data->pMessage}; | 61 | const char* message{data->pMessage}; |
| 63 | 62 | ||
| 64 | if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eError) { | 63 | if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT) { |
| 65 | LOG_CRITICAL(Render_Vulkan, "{}", message); | 64 | LOG_CRITICAL(Render_Vulkan, "{}", message); |
| 66 | } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning) { | 65 | } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT) { |
| 67 | LOG_WARNING(Render_Vulkan, "{}", message); | 66 | LOG_WARNING(Render_Vulkan, "{}", message); |
| 68 | } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo) { | 67 | } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT) { |
| 69 | LOG_INFO(Render_Vulkan, "{}", message); | 68 | LOG_INFO(Render_Vulkan, "{}", message); |
| 70 | } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose) { | 69 | } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT) { |
| 71 | LOG_DEBUG(Render_Vulkan, "{}", message); | 70 | LOG_DEBUG(Render_Vulkan, "{}", message); |
| 72 | } | 71 | } |
| 73 | return VK_FALSE; | 72 | return VK_FALSE; |
| @@ -94,22 +93,24 @@ Common::DynamicLibrary OpenVulkanLibrary() { | |||
| 94 | return library; | 93 | return library; |
| 95 | } | 94 | } |
| 96 | 95 | ||
| 97 | UniqueInstance CreateInstance(Common::DynamicLibrary& library, vk::DispatchLoaderDynamic& dld, | 96 | vk::Instance CreateInstance(Common::DynamicLibrary& library, vk::InstanceDispatch& dld, |
| 98 | WindowSystemType window_type = WindowSystemType::Headless, | 97 | WindowSystemType window_type = WindowSystemType::Headless, |
| 99 | bool enable_layers = false) { | 98 | bool enable_layers = false) { |
| 100 | if (!library.IsOpen()) { | 99 | if (!library.IsOpen()) { |
| 101 | LOG_ERROR(Render_Vulkan, "Vulkan library not available"); | 100 | LOG_ERROR(Render_Vulkan, "Vulkan library not available"); |
| 102 | return UniqueInstance{}; | 101 | return {}; |
| 103 | } | 102 | } |
| 104 | PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr; | 103 | if (!library.GetSymbol("vkGetInstanceProcAddr", &dld.vkGetInstanceProcAddr)) { |
| 105 | if (!library.GetSymbol("vkGetInstanceProcAddr", &vkGetInstanceProcAddr)) { | ||
| 106 | LOG_ERROR(Render_Vulkan, "vkGetInstanceProcAddr not present in Vulkan"); | 104 | LOG_ERROR(Render_Vulkan, "vkGetInstanceProcAddr not present in Vulkan"); |
| 107 | return UniqueInstance{}; | 105 | return {}; |
| 106 | } | ||
| 107 | if (!vk::Load(dld)) { | ||
| 108 | LOG_ERROR(Render_Vulkan, "Failed to load Vulkan function pointers"); | ||
| 109 | return {}; | ||
| 108 | } | 110 | } |
| 109 | dld.init(vkGetInstanceProcAddr); | ||
| 110 | 111 | ||
| 111 | std::vector<const char*> extensions; | 112 | std::vector<const char*> extensions; |
| 112 | extensions.reserve(4); | 113 | extensions.reserve(6); |
| 113 | switch (window_type) { | 114 | switch (window_type) { |
| 114 | case Core::Frontend::WindowSystemType::Headless: | 115 | case Core::Frontend::WindowSystemType::Headless: |
| 115 | break; | 116 | break; |
| @@ -136,45 +137,39 @@ UniqueInstance CreateInstance(Common::DynamicLibrary& library, vk::DispatchLoade | |||
| 136 | if (enable_layers) { | 137 | if (enable_layers) { |
| 137 | extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME); | 138 | extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME); |
| 138 | } | 139 | } |
| 140 | extensions.push_back(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME); | ||
| 139 | 141 | ||
| 140 | u32 num_properties; | 142 | const std::optional properties = vk::EnumerateInstanceExtensionProperties(dld); |
| 141 | if (vk::enumerateInstanceExtensionProperties(nullptr, &num_properties, nullptr, dld) != | 143 | if (!properties) { |
| 142 | vk::Result::eSuccess) { | ||
| 143 | LOG_ERROR(Render_Vulkan, "Failed to query number of extension properties"); | ||
| 144 | return UniqueInstance{}; | ||
| 145 | } | ||
| 146 | std::vector<vk::ExtensionProperties> properties(num_properties); | ||
| 147 | if (vk::enumerateInstanceExtensionProperties(nullptr, &num_properties, properties.data(), | ||
| 148 | dld) != vk::Result::eSuccess) { | ||
| 149 | LOG_ERROR(Render_Vulkan, "Failed to query extension properties"); | 144 | LOG_ERROR(Render_Vulkan, "Failed to query extension properties"); |
| 150 | return UniqueInstance{}; | 145 | return {}; |
| 151 | } | 146 | } |
| 152 | 147 | ||
| 153 | for (const char* extension : extensions) { | 148 | for (const char* extension : extensions) { |
| 154 | const auto it = | 149 | const auto it = |
| 155 | std::find_if(properties.begin(), properties.end(), [extension](const auto& prop) { | 150 | std::find_if(properties->begin(), properties->end(), [extension](const auto& prop) { |
| 156 | return !std::strcmp(extension, prop.extensionName); | 151 | return !std::strcmp(extension, prop.extensionName); |
| 157 | }); | 152 | }); |
| 158 | if (it == properties.end()) { | 153 | if (it == properties->end()) { |
| 159 | LOG_ERROR(Render_Vulkan, "Required instance extension {} is not available", extension); | 154 | LOG_ERROR(Render_Vulkan, "Required instance extension {} is not available", extension); |
| 160 | return UniqueInstance{}; | 155 | return {}; |
| 161 | } | 156 | } |
| 162 | } | 157 | } |
| 163 | 158 | ||
| 164 | const vk::ApplicationInfo application_info("yuzu Emulator", VK_MAKE_VERSION(0, 1, 0), | 159 | static constexpr std::array layers_data{"VK_LAYER_LUNARG_standard_validation"}; |
| 165 | "yuzu Emulator", VK_MAKE_VERSION(0, 1, 0), | 160 | vk::Span<const char*> layers = layers_data; |
| 166 | VK_API_VERSION_1_1); | 161 | if (!enable_layers) { |
| 167 | const std::array layers = {"VK_LAYER_LUNARG_standard_validation"}; | 162 | layers = {}; |
| 168 | const vk::InstanceCreateInfo instance_ci( | 163 | } |
| 169 | {}, &application_info, enable_layers ? static_cast<u32>(layers.size()) : 0, layers.data(), | 164 | vk::Instance instance = vk::Instance::Create(layers, extensions, dld); |
| 170 | static_cast<u32>(extensions.size()), extensions.data()); | 165 | if (!instance) { |
| 171 | vk::Instance unsafe_instance; | ||
| 172 | if (vk::createInstance(&instance_ci, nullptr, &unsafe_instance, dld) != vk::Result::eSuccess) { | ||
| 173 | LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance"); | 166 | LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance"); |
| 174 | return UniqueInstance{}; | 167 | return {}; |
| 168 | } | ||
| 169 | if (!vk::Load(*instance, dld)) { | ||
| 170 | LOG_ERROR(Render_Vulkan, "Failed to load Vulkan instance function pointers"); | ||
| 175 | } | 171 | } |
| 176 | dld.init(unsafe_instance); | 172 | return instance; |
| 177 | return UniqueInstance(unsafe_instance, {nullptr, dld}); | ||
| 178 | } | 173 | } |
| 179 | 174 | ||
| 180 | std::string GetReadableVersion(u32 version) { | 175 | std::string GetReadableVersion(u32 version) { |
| @@ -187,14 +182,14 @@ std::string GetDriverVersion(const VKDevice& device) { | |||
| 187 | // https://github.com/SaschaWillems/vulkan.gpuinfo.org/blob/5dddea46ea1120b0df14eef8f15ff8e318e35462/functions.php#L308-L314 | 182 | // https://github.com/SaschaWillems/vulkan.gpuinfo.org/blob/5dddea46ea1120b0df14eef8f15ff8e318e35462/functions.php#L308-L314 |
| 188 | const u32 version = device.GetDriverVersion(); | 183 | const u32 version = device.GetDriverVersion(); |
| 189 | 184 | ||
| 190 | if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) { | 185 | if (device.GetDriverID() == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) { |
| 191 | const u32 major = (version >> 22) & 0x3ff; | 186 | const u32 major = (version >> 22) & 0x3ff; |
| 192 | const u32 minor = (version >> 14) & 0x0ff; | 187 | const u32 minor = (version >> 14) & 0x0ff; |
| 193 | const u32 secondary = (version >> 6) & 0x0ff; | 188 | const u32 secondary = (version >> 6) & 0x0ff; |
| 194 | const u32 tertiary = version & 0x003f; | 189 | const u32 tertiary = version & 0x003f; |
| 195 | return fmt::format("{}.{}.{}.{}", major, minor, secondary, tertiary); | 190 | return fmt::format("{}.{}.{}.{}", major, minor, secondary, tertiary); |
| 196 | } | 191 | } |
| 197 | if (device.GetDriverID() == vk::DriverIdKHR::eIntelProprietaryWindows) { | 192 | if (device.GetDriverID() == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS_KHR) { |
| 198 | const u32 major = version >> 14; | 193 | const u32 major = version >> 14; |
| 199 | const u32 minor = version & 0x3fff; | 194 | const u32 minor = version & 0x3fff; |
| 200 | return fmt::format("{}.{}", major, minor); | 195 | return fmt::format("{}.{}", major, minor); |
| @@ -307,10 +302,8 @@ void RendererVulkan::ShutDown() { | |||
| 307 | if (!device) { | 302 | if (!device) { |
| 308 | return; | 303 | return; |
| 309 | } | 304 | } |
| 310 | const auto dev = device->GetLogical(); | 305 | if (const auto& dev = device->GetLogical()) { |
| 311 | const auto& dld = device->GetDispatchLoader(); | 306 | dev.WaitIdle(); |
| 312 | if (dev && dld.vkDeviceWaitIdle) { | ||
| 313 | dev.waitIdle(dld); | ||
| 314 | } | 307 | } |
| 315 | 308 | ||
| 316 | rasterizer.reset(); | 309 | rasterizer.reset(); |
| @@ -326,23 +319,11 @@ bool RendererVulkan::CreateDebugCallback() { | |||
| 326 | if (!Settings::values.renderer_debug) { | 319 | if (!Settings::values.renderer_debug) { |
| 327 | return true; | 320 | return true; |
| 328 | } | 321 | } |
| 329 | const vk::DebugUtilsMessengerCreateInfoEXT callback_ci( | 322 | debug_callback = instance.TryCreateDebugCallback(DebugCallback); |
| 330 | {}, | 323 | if (!debug_callback) { |
| 331 | vk::DebugUtilsMessageSeverityFlagBitsEXT::eError | | ||
| 332 | vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning | | ||
| 333 | vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo | | ||
| 334 | vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose, | ||
| 335 | vk::DebugUtilsMessageTypeFlagBitsEXT::eGeneral | | ||
| 336 | vk::DebugUtilsMessageTypeFlagBitsEXT::eValidation | | ||
| 337 | vk::DebugUtilsMessageTypeFlagBitsEXT::ePerformance, | ||
| 338 | &DebugCallback, nullptr); | ||
| 339 | vk::DebugUtilsMessengerEXT unsafe_callback; | ||
| 340 | if (instance->createDebugUtilsMessengerEXT(&callback_ci, nullptr, &unsafe_callback, dld) != | ||
| 341 | vk::Result::eSuccess) { | ||
| 342 | LOG_ERROR(Render_Vulkan, "Failed to create debug callback"); | 324 | LOG_ERROR(Render_Vulkan, "Failed to create debug callback"); |
| 343 | return false; | 325 | return false; |
| 344 | } | 326 | } |
| 345 | debug_callback = UniqueDebugUtilsMessengerEXT(unsafe_callback, {*instance, nullptr, dld}); | ||
| 346 | return true; | 327 | return true; |
| 347 | } | 328 | } |
| 348 | 329 | ||
| @@ -357,8 +338,8 @@ bool RendererVulkan::CreateSurface() { | |||
| 357 | nullptr, 0, nullptr, hWnd}; | 338 | nullptr, 0, nullptr, hWnd}; |
| 358 | const auto vkCreateWin32SurfaceKHR = reinterpret_cast<PFN_vkCreateWin32SurfaceKHR>( | 339 | const auto vkCreateWin32SurfaceKHR = reinterpret_cast<PFN_vkCreateWin32SurfaceKHR>( |
| 359 | dld.vkGetInstanceProcAddr(*instance, "vkCreateWin32SurfaceKHR")); | 340 | dld.vkGetInstanceProcAddr(*instance, "vkCreateWin32SurfaceKHR")); |
| 360 | if (!vkCreateWin32SurfaceKHR || vkCreateWin32SurfaceKHR(instance.get(), &win32_ci, nullptr, | 341 | if (!vkCreateWin32SurfaceKHR || |
| 361 | &unsafe_surface) != VK_SUCCESS) { | 342 | vkCreateWin32SurfaceKHR(*instance, &win32_ci, nullptr, &unsafe_surface) != VK_SUCCESS) { |
| 362 | LOG_ERROR(Render_Vulkan, "Failed to initialize Win32 surface"); | 343 | LOG_ERROR(Render_Vulkan, "Failed to initialize Win32 surface"); |
| 363 | return false; | 344 | return false; |
| 364 | } | 345 | } |
| @@ -372,8 +353,8 @@ bool RendererVulkan::CreateSurface() { | |||
| 372 | reinterpret_cast<Window>(window_info.render_surface)}; | 353 | reinterpret_cast<Window>(window_info.render_surface)}; |
| 373 | const auto vkCreateXlibSurfaceKHR = reinterpret_cast<PFN_vkCreateXlibSurfaceKHR>( | 354 | const auto vkCreateXlibSurfaceKHR = reinterpret_cast<PFN_vkCreateXlibSurfaceKHR>( |
| 374 | dld.vkGetInstanceProcAddr(*instance, "vkCreateXlibSurfaceKHR")); | 355 | dld.vkGetInstanceProcAddr(*instance, "vkCreateXlibSurfaceKHR")); |
| 375 | if (!vkCreateXlibSurfaceKHR || vkCreateXlibSurfaceKHR(instance.get(), &xlib_ci, nullptr, | 356 | if (!vkCreateXlibSurfaceKHR || |
| 376 | &unsafe_surface) != VK_SUCCESS) { | 357 | vkCreateXlibSurfaceKHR(*instance, &xlib_ci, nullptr, &unsafe_surface) != VK_SUCCESS) { |
| 377 | LOG_ERROR(Render_Vulkan, "Failed to initialize Xlib surface"); | 358 | LOG_ERROR(Render_Vulkan, "Failed to initialize Xlib surface"); |
| 378 | return false; | 359 | return false; |
| 379 | } | 360 | } |
| @@ -386,7 +367,7 @@ bool RendererVulkan::CreateSurface() { | |||
| 386 | const auto vkCreateWaylandSurfaceKHR = reinterpret_cast<PFN_vkCreateWaylandSurfaceKHR>( | 367 | const auto vkCreateWaylandSurfaceKHR = reinterpret_cast<PFN_vkCreateWaylandSurfaceKHR>( |
| 387 | dld.vkGetInstanceProcAddr(*instance, "vkCreateWaylandSurfaceKHR")); | 368 | dld.vkGetInstanceProcAddr(*instance, "vkCreateWaylandSurfaceKHR")); |
| 388 | if (!vkCreateWaylandSurfaceKHR || | 369 | if (!vkCreateWaylandSurfaceKHR || |
| 389 | vkCreateWaylandSurfaceKHR(instance.get(), &wayland_ci, nullptr, &unsafe_surface) != | 370 | vkCreateWaylandSurfaceKHR(*instance, &wayland_ci, nullptr, &unsafe_surface) != |
| 390 | VK_SUCCESS) { | 371 | VK_SUCCESS) { |
| 391 | LOG_ERROR(Render_Vulkan, "Failed to initialize Wayland surface"); | 372 | LOG_ERROR(Render_Vulkan, "Failed to initialize Wayland surface"); |
| 392 | return false; | 373 | return false; |
| @@ -398,26 +379,30 @@ bool RendererVulkan::CreateSurface() { | |||
| 398 | return false; | 379 | return false; |
| 399 | } | 380 | } |
| 400 | 381 | ||
| 401 | surface = UniqueSurfaceKHR(unsafe_surface, {*instance, nullptr, dld}); | 382 | surface = vk::SurfaceKHR(unsafe_surface, *instance, dld); |
| 402 | return true; | 383 | return true; |
| 403 | } | 384 | } |
| 404 | 385 | ||
| 405 | bool RendererVulkan::PickDevices() { | 386 | bool RendererVulkan::PickDevices() { |
| 406 | const auto devices = instance->enumeratePhysicalDevices(dld); | 387 | const auto devices = instance.EnumeratePhysicalDevices(); |
| 388 | if (!devices) { | ||
| 389 | LOG_ERROR(Render_Vulkan, "Failed to enumerate physical devices"); | ||
| 390 | return false; | ||
| 391 | } | ||
| 407 | 392 | ||
| 408 | const s32 device_index = Settings::values.vulkan_device; | 393 | const s32 device_index = Settings::values.vulkan_device; |
| 409 | if (device_index < 0 || device_index >= static_cast<s32>(devices.size())) { | 394 | if (device_index < 0 || device_index >= static_cast<s32>(devices->size())) { |
| 410 | LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index); | 395 | LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index); |
| 411 | return false; | 396 | return false; |
| 412 | } | 397 | } |
| 413 | const vk::PhysicalDevice physical_device = devices[static_cast<std::size_t>(device_index)]; | 398 | const vk::PhysicalDevice physical_device((*devices)[static_cast<std::size_t>(device_index)], |
| 414 | 399 | dld); | |
| 415 | if (!VKDevice::IsSuitable(physical_device, *surface, dld)) { | 400 | if (!VKDevice::IsSuitable(physical_device, *surface)) { |
| 416 | return false; | 401 | return false; |
| 417 | } | 402 | } |
| 418 | 403 | ||
| 419 | device = std::make_unique<VKDevice>(dld, physical_device, *surface); | 404 | device = std::make_unique<VKDevice>(*instance, physical_device, *surface, dld); |
| 420 | return device->Create(*instance); | 405 | return device->Create(); |
| 421 | } | 406 | } |
| 422 | 407 | ||
| 423 | void RendererVulkan::Report() const { | 408 | void RendererVulkan::Report() const { |
| @@ -444,30 +429,22 @@ void RendererVulkan::Report() const { | |||
| 444 | } | 429 | } |
| 445 | 430 | ||
| 446 | std::vector<std::string> RendererVulkan::EnumerateDevices() { | 431 | std::vector<std::string> RendererVulkan::EnumerateDevices() { |
| 447 | // Avoid putting DispatchLoaderDynamic, it's too large | 432 | vk::InstanceDispatch dld; |
| 448 | auto dld_memory = std::make_unique<vk::DispatchLoaderDynamic>(); | ||
| 449 | auto& dld = *dld_memory; | ||
| 450 | |||
| 451 | Common::DynamicLibrary library = OpenVulkanLibrary(); | 433 | Common::DynamicLibrary library = OpenVulkanLibrary(); |
| 452 | UniqueInstance instance = CreateInstance(library, dld); | 434 | vk::Instance instance = CreateInstance(library, dld); |
| 453 | if (!instance) { | 435 | if (!instance) { |
| 454 | return {}; | 436 | return {}; |
| 455 | } | 437 | } |
| 456 | 438 | ||
| 457 | u32 num_devices; | 439 | const std::optional physical_devices = instance.EnumeratePhysicalDevices(); |
| 458 | if (instance->enumeratePhysicalDevices(&num_devices, nullptr, dld) != vk::Result::eSuccess) { | 440 | if (!physical_devices) { |
| 459 | return {}; | ||
| 460 | } | ||
| 461 | std::vector<vk::PhysicalDevice> devices(num_devices); | ||
| 462 | if (instance->enumeratePhysicalDevices(&num_devices, devices.data(), dld) != | ||
| 463 | vk::Result::eSuccess) { | ||
| 464 | return {}; | 441 | return {}; |
| 465 | } | 442 | } |
| 466 | 443 | ||
| 467 | std::vector<std::string> names; | 444 | std::vector<std::string> names; |
| 468 | names.reserve(num_devices); | 445 | names.reserve(physical_devices->size()); |
| 469 | for (auto& device : devices) { | 446 | for (const auto& device : *physical_devices) { |
| 470 | names.push_back(device.getProperties(dld).deviceName); | 447 | names.push_back(vk::PhysicalDevice(device, dld).GetProperties().deviceName); |
| 471 | } | 448 | } |
| 472 | return names; | 449 | return names; |
| 473 | } | 450 | } |
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h index 42e253de5..18270909b 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.h +++ b/src/video_core/renderer_vulkan/renderer_vulkan.h | |||
| @@ -12,7 +12,7 @@ | |||
| 12 | #include "common/dynamic_library.h" | 12 | #include "common/dynamic_library.h" |
| 13 | 13 | ||
| 14 | #include "video_core/renderer_base.h" | 14 | #include "video_core/renderer_base.h" |
| 15 | #include "video_core/renderer_vulkan/declarations.h" | 15 | #include "video_core/renderer_vulkan/wrapper.h" |
| 16 | 16 | ||
| 17 | namespace Core { | 17 | namespace Core { |
| 18 | class System; | 18 | class System; |
| @@ -61,14 +61,14 @@ private: | |||
| 61 | Core::System& system; | 61 | Core::System& system; |
| 62 | 62 | ||
| 63 | Common::DynamicLibrary library; | 63 | Common::DynamicLibrary library; |
| 64 | vk::DispatchLoaderDynamic dld; | 64 | vk::InstanceDispatch dld; |
| 65 | 65 | ||
| 66 | UniqueInstance instance; | 66 | vk::Instance instance; |
| 67 | UniqueSurfaceKHR surface; | 67 | vk::SurfaceKHR surface; |
| 68 | 68 | ||
| 69 | VKScreenInfo screen_info; | 69 | VKScreenInfo screen_info; |
| 70 | 70 | ||
| 71 | UniqueDebugUtilsMessengerEXT debug_callback; | 71 | vk::DebugCallback debug_callback; |
| 72 | std::unique_ptr<VKDevice> device; | 72 | std::unique_ptr<VKDevice> device; |
| 73 | std::unique_ptr<VKSwapchain> swapchain; | 73 | std::unique_ptr<VKSwapchain> swapchain; |
| 74 | std::unique_ptr<VKMemoryManager> memory_manager; | 74 | std::unique_ptr<VKMemoryManager> memory_manager; |
diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index 855cfc883..21644a7e7 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp | |||
| @@ -20,7 +20,6 @@ | |||
| 20 | #include "video_core/gpu.h" | 20 | #include "video_core/gpu.h" |
| 21 | #include "video_core/morton.h" | 21 | #include "video_core/morton.h" |
| 22 | #include "video_core/rasterizer_interface.h" | 22 | #include "video_core/rasterizer_interface.h" |
| 23 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 24 | #include "video_core/renderer_vulkan/renderer_vulkan.h" | 23 | #include "video_core/renderer_vulkan/renderer_vulkan.h" |
| 25 | #include "video_core/renderer_vulkan/vk_blit_screen.h" | 24 | #include "video_core/renderer_vulkan/vk_blit_screen.h" |
| 26 | #include "video_core/renderer_vulkan/vk_device.h" | 25 | #include "video_core/renderer_vulkan/vk_device.h" |
| @@ -30,6 +29,7 @@ | |||
| 30 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 29 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 31 | #include "video_core/renderer_vulkan/vk_shader_util.h" | 30 | #include "video_core/renderer_vulkan/vk_shader_util.h" |
| 32 | #include "video_core/renderer_vulkan/vk_swapchain.h" | 31 | #include "video_core/renderer_vulkan/vk_swapchain.h" |
| 32 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 33 | #include "video_core/surface.h" | 33 | #include "video_core/surface.h" |
| 34 | 34 | ||
| 35 | namespace Vulkan { | 35 | namespace Vulkan { |
| @@ -140,16 +140,25 @@ struct ScreenRectVertex { | |||
| 140 | std::array<f32, 2> position; | 140 | std::array<f32, 2> position; |
| 141 | std::array<f32, 2> tex_coord; | 141 | std::array<f32, 2> tex_coord; |
| 142 | 142 | ||
| 143 | static vk::VertexInputBindingDescription GetDescription() { | 143 | static VkVertexInputBindingDescription GetDescription() { |
| 144 | return vk::VertexInputBindingDescription(0, sizeof(ScreenRectVertex), | 144 | VkVertexInputBindingDescription description; |
| 145 | vk::VertexInputRate::eVertex); | 145 | description.binding = 0; |
| 146 | description.stride = sizeof(ScreenRectVertex); | ||
| 147 | description.inputRate = VK_VERTEX_INPUT_RATE_VERTEX; | ||
| 148 | return description; | ||
| 146 | } | 149 | } |
| 147 | 150 | ||
| 148 | static std::array<vk::VertexInputAttributeDescription, 2> GetAttributes() { | 151 | static std::array<VkVertexInputAttributeDescription, 2> GetAttributes() { |
| 149 | return {vk::VertexInputAttributeDescription(0, 0, vk::Format::eR32G32Sfloat, | 152 | std::array<VkVertexInputAttributeDescription, 2> attributes; |
| 150 | offsetof(ScreenRectVertex, position)), | 153 | attributes[0].location = 0; |
| 151 | vk::VertexInputAttributeDescription(1, 0, vk::Format::eR32G32Sfloat, | 154 | attributes[0].binding = 0; |
| 152 | offsetof(ScreenRectVertex, tex_coord))}; | 155 | attributes[0].format = VK_FORMAT_R32G32_SFLOAT; |
| 156 | attributes[0].offset = offsetof(ScreenRectVertex, position); | ||
| 157 | attributes[1].location = 1; | ||
| 158 | attributes[1].binding = 0; | ||
| 159 | attributes[1].format = VK_FORMAT_R32G32_SFLOAT; | ||
| 160 | attributes[1].offset = offsetof(ScreenRectVertex, tex_coord); | ||
| 161 | return attributes; | ||
| 153 | } | 162 | } |
| 154 | }; | 163 | }; |
| 155 | 164 | ||
| @@ -172,16 +181,16 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) { | |||
| 172 | static_cast<std::size_t>(framebuffer.height) * GetBytesPerPixel(framebuffer); | 181 | static_cast<std::size_t>(framebuffer.height) * GetBytesPerPixel(framebuffer); |
| 173 | } | 182 | } |
| 174 | 183 | ||
| 175 | vk::Format GetFormat(const Tegra::FramebufferConfig& framebuffer) { | 184 | VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) { |
| 176 | switch (framebuffer.pixel_format) { | 185 | switch (framebuffer.pixel_format) { |
| 177 | case Tegra::FramebufferConfig::PixelFormat::ABGR8: | 186 | case Tegra::FramebufferConfig::PixelFormat::ABGR8: |
| 178 | return vk::Format::eA8B8G8R8UnormPack32; | 187 | return VK_FORMAT_A8B8G8R8_UNORM_PACK32; |
| 179 | case Tegra::FramebufferConfig::PixelFormat::RGB565: | 188 | case Tegra::FramebufferConfig::PixelFormat::RGB565: |
| 180 | return vk::Format::eR5G6B5UnormPack16; | 189 | return VK_FORMAT_R5G6B5_UNORM_PACK16; |
| 181 | default: | 190 | default: |
| 182 | UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", | 191 | UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", |
| 183 | static_cast<u32>(framebuffer.pixel_format)); | 192 | static_cast<u32>(framebuffer.pixel_format)); |
| 184 | return vk::Format::eA8B8G8R8UnormPack32; | 193 | return VK_FORMAT_A8B8G8R8_UNORM_PACK32; |
| 185 | } | 194 | } |
| 186 | } | 195 | } |
| 187 | 196 | ||
| @@ -219,8 +228,8 @@ void VKBlitScreen::Recreate() { | |||
| 219 | CreateDynamicResources(); | 228 | CreateDynamicResources(); |
| 220 | } | 229 | } |
| 221 | 230 | ||
| 222 | std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, | 231 | std::tuple<VKFence&, VkSemaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, |
| 223 | bool use_accelerated) { | 232 | bool use_accelerated) { |
| 224 | RefreshResources(framebuffer); | 233 | RefreshResources(framebuffer); |
| 225 | 234 | ||
| 226 | // Finish any pending renderpass | 235 | // Finish any pending renderpass |
| @@ -255,46 +264,76 @@ std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferC | |||
| 255 | framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1, | 264 | framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1, |
| 256 | map.GetAddress() + image_offset, host_ptr); | 265 | map.GetAddress() + image_offset, host_ptr); |
| 257 | 266 | ||
| 258 | blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eTransfer, | 267 | blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_TRANSFER_BIT, |
| 259 | vk::AccessFlagBits::eTransferWrite, | 268 | VK_ACCESS_TRANSFER_WRITE_BIT, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL); |
| 260 | vk::ImageLayout::eTransferDstOptimal); | 269 | |
| 261 | 270 | VkBufferImageCopy copy; | |
| 262 | const vk::BufferImageCopy copy(image_offset, 0, 0, | 271 | copy.bufferOffset = image_offset; |
| 263 | {vk::ImageAspectFlagBits::eColor, 0, 0, 1}, {0, 0, 0}, | 272 | copy.bufferRowLength = 0; |
| 264 | {framebuffer.width, framebuffer.height, 1}); | 273 | copy.bufferImageHeight = 0; |
| 265 | scheduler.Record([buffer_handle = *buffer, image = blit_image->GetHandle(), | 274 | copy.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; |
| 266 | copy](auto cmdbuf, auto& dld) { | 275 | copy.imageSubresource.mipLevel = 0; |
| 267 | cmdbuf.copyBufferToImage(buffer_handle, image, vk::ImageLayout::eTransferDstOptimal, | 276 | copy.imageSubresource.baseArrayLayer = 0; |
| 268 | {copy}, dld); | 277 | copy.imageSubresource.layerCount = 1; |
| 269 | }); | 278 | copy.imageOffset.x = 0; |
| 279 | copy.imageOffset.y = 0; | ||
| 280 | copy.imageOffset.z = 0; | ||
| 281 | copy.imageExtent.width = framebuffer.width; | ||
| 282 | copy.imageExtent.height = framebuffer.height; | ||
| 283 | copy.imageExtent.depth = 1; | ||
| 284 | scheduler.Record( | ||
| 285 | [buffer = *buffer, image = *blit_image->GetHandle(), copy](vk::CommandBuffer cmdbuf) { | ||
| 286 | cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy); | ||
| 287 | }); | ||
| 270 | } | 288 | } |
| 271 | map.Release(); | 289 | map.Release(); |
| 272 | 290 | ||
| 273 | blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eFragmentShader, | 291 | blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, |
| 274 | vk::AccessFlagBits::eShaderRead, | 292 | VK_ACCESS_SHADER_READ_BIT, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL); |
| 275 | vk::ImageLayout::eShaderReadOnlyOptimal); | ||
| 276 | 293 | ||
| 277 | scheduler.Record([renderpass = *renderpass, framebuffer = *framebuffers[image_index], | 294 | scheduler.Record([renderpass = *renderpass, framebuffer = *framebuffers[image_index], |
| 278 | descriptor_set = descriptor_sets[image_index], buffer = *buffer, | 295 | descriptor_set = descriptor_sets[image_index], buffer = *buffer, |
| 279 | size = swapchain.GetSize(), pipeline = *pipeline, | 296 | size = swapchain.GetSize(), pipeline = *pipeline, |
| 280 | layout = *pipeline_layout](auto cmdbuf, auto& dld) { | 297 | layout = *pipeline_layout](vk::CommandBuffer cmdbuf) { |
| 281 | const vk::ClearValue clear_color{std::array{0.0f, 0.0f, 0.0f, 1.0f}}; | 298 | VkClearValue clear_color; |
| 282 | const vk::RenderPassBeginInfo renderpass_bi(renderpass, framebuffer, {{0, 0}, size}, 1, | 299 | clear_color.color.float32[0] = 0.0f; |
| 283 | &clear_color); | 300 | clear_color.color.float32[1] = 0.0f; |
| 284 | 301 | clear_color.color.float32[2] = 0.0f; | |
| 285 | cmdbuf.beginRenderPass(renderpass_bi, vk::SubpassContents::eInline, dld); | 302 | clear_color.color.float32[3] = 0.0f; |
| 286 | cmdbuf.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline, dld); | 303 | |
| 287 | cmdbuf.setViewport( | 304 | VkRenderPassBeginInfo renderpass_bi; |
| 288 | 0, | 305 | renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; |
| 289 | {{0.0f, 0.0f, static_cast<f32>(size.width), static_cast<f32>(size.height), 0.0f, 1.0f}}, | 306 | renderpass_bi.pNext = nullptr; |
| 290 | dld); | 307 | renderpass_bi.renderPass = renderpass; |
| 291 | cmdbuf.setScissor(0, {{{0, 0}, size}}, dld); | 308 | renderpass_bi.framebuffer = framebuffer; |
| 292 | 309 | renderpass_bi.renderArea.offset.x = 0; | |
| 293 | cmdbuf.bindVertexBuffers(0, {buffer}, {offsetof(BufferData, vertices)}, dld); | 310 | renderpass_bi.renderArea.offset.y = 0; |
| 294 | cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, layout, 0, {descriptor_set}, {}, | 311 | renderpass_bi.renderArea.extent = size; |
| 295 | dld); | 312 | renderpass_bi.clearValueCount = 1; |
| 296 | cmdbuf.draw(4, 1, 0, 0, dld); | 313 | renderpass_bi.pClearValues = &clear_color; |
| 297 | cmdbuf.endRenderPass(dld); | 314 | |
| 315 | VkViewport viewport; | ||
| 316 | viewport.x = 0.0f; | ||
| 317 | viewport.y = 0.0f; | ||
| 318 | viewport.width = static_cast<float>(size.width); | ||
| 319 | viewport.height = static_cast<float>(size.height); | ||
| 320 | viewport.minDepth = 0.0f; | ||
| 321 | viewport.maxDepth = 1.0f; | ||
| 322 | |||
| 323 | VkRect2D scissor; | ||
| 324 | scissor.offset.x = 0; | ||
| 325 | scissor.offset.y = 0; | ||
| 326 | scissor.extent = size; | ||
| 327 | |||
| 328 | cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE); | ||
| 329 | cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); | ||
| 330 | cmdbuf.SetViewport(0, viewport); | ||
| 331 | cmdbuf.SetScissor(0, scissor); | ||
| 332 | |||
| 333 | cmdbuf.BindVertexBuffer(0, buffer, offsetof(BufferData, vertices)); | ||
| 334 | cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set, {}); | ||
| 335 | cmdbuf.Draw(4, 1, 0, 0); | ||
| 336 | cmdbuf.EndRenderPass(); | ||
| 298 | }); | 337 | }); |
| 299 | 338 | ||
| 300 | return {scheduler.GetFence(), *semaphores[image_index]}; | 339 | return {scheduler.GetFence(), *semaphores[image_index]}; |
| @@ -334,165 +373,295 @@ void VKBlitScreen::CreateShaders() { | |||
| 334 | } | 373 | } |
| 335 | 374 | ||
| 336 | void VKBlitScreen::CreateSemaphores() { | 375 | void VKBlitScreen::CreateSemaphores() { |
| 337 | const auto dev = device.GetLogical(); | ||
| 338 | const auto& dld = device.GetDispatchLoader(); | ||
| 339 | |||
| 340 | semaphores.resize(image_count); | 376 | semaphores.resize(image_count); |
| 341 | for (std::size_t i = 0; i < image_count; ++i) { | 377 | std::generate(semaphores.begin(), semaphores.end(), |
| 342 | semaphores[i] = dev.createSemaphoreUnique({}, nullptr, dld); | 378 | [this] { return device.GetLogical().CreateSemaphore(); }); |
| 343 | } | ||
| 344 | } | 379 | } |
| 345 | 380 | ||
| 346 | void VKBlitScreen::CreateDescriptorPool() { | 381 | void VKBlitScreen::CreateDescriptorPool() { |
| 347 | const std::array<vk::DescriptorPoolSize, 2> pool_sizes{ | 382 | std::array<VkDescriptorPoolSize, 2> pool_sizes; |
| 348 | vk::DescriptorPoolSize{vk::DescriptorType::eUniformBuffer, static_cast<u32>(image_count)}, | 383 | pool_sizes[0].type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; |
| 349 | vk::DescriptorPoolSize{vk::DescriptorType::eCombinedImageSampler, | 384 | pool_sizes[0].descriptorCount = static_cast<u32>(image_count); |
| 350 | static_cast<u32>(image_count)}}; | 385 | pool_sizes[1].type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; |
| 351 | const vk::DescriptorPoolCreateInfo pool_ci( | 386 | pool_sizes[1].descriptorCount = static_cast<u32>(image_count); |
| 352 | {}, static_cast<u32>(image_count), static_cast<u32>(pool_sizes.size()), pool_sizes.data()); | 387 | |
| 353 | const auto dev = device.GetLogical(); | 388 | VkDescriptorPoolCreateInfo ci; |
| 354 | descriptor_pool = dev.createDescriptorPoolUnique(pool_ci, nullptr, device.GetDispatchLoader()); | 389 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO; |
| 390 | ci.pNext = nullptr; | ||
| 391 | ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; | ||
| 392 | ci.maxSets = static_cast<u32>(image_count); | ||
| 393 | ci.poolSizeCount = static_cast<u32>(pool_sizes.size()); | ||
| 394 | ci.pPoolSizes = pool_sizes.data(); | ||
| 395 | descriptor_pool = device.GetLogical().CreateDescriptorPool(ci); | ||
| 355 | } | 396 | } |
| 356 | 397 | ||
| 357 | void VKBlitScreen::CreateRenderPass() { | 398 | void VKBlitScreen::CreateRenderPass() { |
| 358 | const vk::AttachmentDescription color_attachment( | 399 | VkAttachmentDescription color_attachment; |
| 359 | {}, swapchain.GetImageFormat(), vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eClear, | 400 | color_attachment.flags = 0; |
| 360 | vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eDontCare, | 401 | color_attachment.format = swapchain.GetImageFormat(); |
| 361 | vk::AttachmentStoreOp::eDontCare, vk::ImageLayout::eUndefined, | 402 | color_attachment.samples = VK_SAMPLE_COUNT_1_BIT; |
| 362 | vk::ImageLayout::ePresentSrcKHR); | 403 | color_attachment.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR; |
| 363 | 404 | color_attachment.storeOp = VK_ATTACHMENT_STORE_OP_STORE; | |
| 364 | const vk::AttachmentReference color_attachment_ref(0, vk::ImageLayout::eColorAttachmentOptimal); | 405 | color_attachment.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; |
| 365 | 406 | color_attachment.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; | |
| 366 | const vk::SubpassDescription subpass_description({}, vk::PipelineBindPoint::eGraphics, 0, | 407 | color_attachment.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; |
| 367 | nullptr, 1, &color_attachment_ref, nullptr, | 408 | color_attachment.finalLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR; |
| 368 | nullptr, 0, nullptr); | 409 | |
| 369 | 410 | VkAttachmentReference color_attachment_ref; | |
| 370 | const vk::SubpassDependency dependency( | 411 | color_attachment_ref.attachment = 0; |
| 371 | VK_SUBPASS_EXTERNAL, 0, vk::PipelineStageFlagBits::eColorAttachmentOutput, | 412 | color_attachment_ref.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; |
| 372 | vk::PipelineStageFlagBits::eColorAttachmentOutput, {}, | 413 | |
| 373 | vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite, {}); | 414 | VkSubpassDescription subpass_description; |
| 374 | 415 | subpass_description.flags = 0; | |
| 375 | const vk::RenderPassCreateInfo renderpass_ci({}, 1, &color_attachment, 1, &subpass_description, | 416 | subpass_description.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; |
| 376 | 1, &dependency); | 417 | subpass_description.inputAttachmentCount = 0; |
| 377 | 418 | subpass_description.pInputAttachments = nullptr; | |
| 378 | const auto dev = device.GetLogical(); | 419 | subpass_description.colorAttachmentCount = 1; |
| 379 | renderpass = dev.createRenderPassUnique(renderpass_ci, nullptr, device.GetDispatchLoader()); | 420 | subpass_description.pColorAttachments = &color_attachment_ref; |
| 421 | subpass_description.pResolveAttachments = nullptr; | ||
| 422 | subpass_description.pDepthStencilAttachment = nullptr; | ||
| 423 | subpass_description.preserveAttachmentCount = 0; | ||
| 424 | subpass_description.pPreserveAttachments = nullptr; | ||
| 425 | |||
| 426 | VkSubpassDependency dependency; | ||
| 427 | dependency.srcSubpass = VK_SUBPASS_EXTERNAL; | ||
| 428 | dependency.dstSubpass = 0; | ||
| 429 | dependency.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; | ||
| 430 | dependency.dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; | ||
| 431 | dependency.srcAccessMask = 0; | ||
| 432 | dependency.dstAccessMask = | ||
| 433 | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; | ||
| 434 | dependency.dependencyFlags = 0; | ||
| 435 | |||
| 436 | VkRenderPassCreateInfo renderpass_ci; | ||
| 437 | renderpass_ci.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; | ||
| 438 | renderpass_ci.pNext = nullptr; | ||
| 439 | renderpass_ci.flags = 0; | ||
| 440 | renderpass_ci.attachmentCount = 1; | ||
| 441 | renderpass_ci.pAttachments = &color_attachment; | ||
| 442 | renderpass_ci.subpassCount = 1; | ||
| 443 | renderpass_ci.pSubpasses = &subpass_description; | ||
| 444 | renderpass_ci.dependencyCount = 1; | ||
| 445 | renderpass_ci.pDependencies = &dependency; | ||
| 446 | |||
| 447 | renderpass = device.GetLogical().CreateRenderPass(renderpass_ci); | ||
| 380 | } | 448 | } |
| 381 | 449 | ||
| 382 | void VKBlitScreen::CreateDescriptorSetLayout() { | 450 | void VKBlitScreen::CreateDescriptorSetLayout() { |
| 383 | const std::array<vk::DescriptorSetLayoutBinding, 2> layout_bindings{ | 451 | std::array<VkDescriptorSetLayoutBinding, 2> layout_bindings; |
| 384 | vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eUniformBuffer, 1, | 452 | layout_bindings[0].binding = 0; |
| 385 | vk::ShaderStageFlagBits::eVertex, nullptr), | 453 | layout_bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; |
| 386 | vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eCombinedImageSampler, 1, | 454 | layout_bindings[0].descriptorCount = 1; |
| 387 | vk::ShaderStageFlagBits::eFragment, nullptr)}; | 455 | layout_bindings[0].stageFlags = VK_SHADER_STAGE_VERTEX_BIT; |
| 388 | const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci( | 456 | layout_bindings[0].pImmutableSamplers = nullptr; |
| 389 | {}, static_cast<u32>(layout_bindings.size()), layout_bindings.data()); | 457 | layout_bindings[1].binding = 1; |
| 390 | 458 | layout_bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; | |
| 391 | const auto dev = device.GetLogical(); | 459 | layout_bindings[1].descriptorCount = 1; |
| 392 | const auto& dld = device.GetDispatchLoader(); | 460 | layout_bindings[1].stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT; |
| 393 | descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld); | 461 | layout_bindings[1].pImmutableSamplers = nullptr; |
| 462 | |||
| 463 | VkDescriptorSetLayoutCreateInfo ci; | ||
| 464 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; | ||
| 465 | ci.pNext = nullptr; | ||
| 466 | ci.flags = 0; | ||
| 467 | ci.bindingCount = static_cast<u32>(layout_bindings.size()); | ||
| 468 | ci.pBindings = layout_bindings.data(); | ||
| 469 | |||
| 470 | descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(ci); | ||
| 394 | } | 471 | } |
| 395 | 472 | ||
| 396 | void VKBlitScreen::CreateDescriptorSets() { | 473 | void VKBlitScreen::CreateDescriptorSets() { |
| 397 | const auto dev = device.GetLogical(); | 474 | const std::vector layouts(image_count, *descriptor_set_layout); |
| 398 | const auto& dld = device.GetDispatchLoader(); | 475 | |
| 399 | 476 | VkDescriptorSetAllocateInfo ai; | |
| 400 | descriptor_sets.resize(image_count); | 477 | ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO; |
| 401 | for (std::size_t i = 0; i < image_count; ++i) { | 478 | ai.pNext = nullptr; |
| 402 | const vk::DescriptorSetLayout layout = *descriptor_set_layout; | 479 | ai.descriptorPool = *descriptor_pool; |
| 403 | const vk::DescriptorSetAllocateInfo descriptor_set_ai(*descriptor_pool, 1, &layout); | 480 | ai.descriptorSetCount = static_cast<u32>(image_count); |
| 404 | const vk::Result result = | 481 | ai.pSetLayouts = layouts.data(); |
| 405 | dev.allocateDescriptorSets(&descriptor_set_ai, &descriptor_sets[i], dld); | 482 | descriptor_sets = descriptor_pool.Allocate(ai); |
| 406 | ASSERT(result == vk::Result::eSuccess); | ||
| 407 | } | ||
| 408 | } | 483 | } |
| 409 | 484 | ||
| 410 | void VKBlitScreen::CreatePipelineLayout() { | 485 | void VKBlitScreen::CreatePipelineLayout() { |
| 411 | const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &descriptor_set_layout.get(), 0, | 486 | VkPipelineLayoutCreateInfo ci; |
| 412 | nullptr); | 487 | ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; |
| 413 | const auto dev = device.GetLogical(); | 488 | ci.pNext = nullptr; |
| 414 | const auto& dld = device.GetDispatchLoader(); | 489 | ci.flags = 0; |
| 415 | pipeline_layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld); | 490 | ci.setLayoutCount = 1; |
| 491 | ci.pSetLayouts = descriptor_set_layout.address(); | ||
| 492 | ci.pushConstantRangeCount = 0; | ||
| 493 | ci.pPushConstantRanges = nullptr; | ||
| 494 | pipeline_layout = device.GetLogical().CreatePipelineLayout(ci); | ||
| 416 | } | 495 | } |
| 417 | 496 | ||
| 418 | void VKBlitScreen::CreateGraphicsPipeline() { | 497 | void VKBlitScreen::CreateGraphicsPipeline() { |
| 419 | const std::array shader_stages = { | 498 | std::array<VkPipelineShaderStageCreateInfo, 2> shader_stages; |
| 420 | vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eVertex, *vertex_shader, | 499 | shader_stages[0].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; |
| 421 | "main", nullptr), | 500 | shader_stages[0].pNext = nullptr; |
| 422 | vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eFragment, *fragment_shader, | 501 | shader_stages[0].flags = 0; |
| 423 | "main", nullptr)}; | 502 | shader_stages[0].stage = VK_SHADER_STAGE_VERTEX_BIT; |
| 503 | shader_stages[0].module = *vertex_shader; | ||
| 504 | shader_stages[0].pName = "main"; | ||
| 505 | shader_stages[0].pSpecializationInfo = nullptr; | ||
| 506 | shader_stages[1].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; | ||
| 507 | shader_stages[1].pNext = nullptr; | ||
| 508 | shader_stages[1].flags = 0; | ||
| 509 | shader_stages[1].stage = VK_SHADER_STAGE_FRAGMENT_BIT; | ||
| 510 | shader_stages[1].module = *fragment_shader; | ||
| 511 | shader_stages[1].pName = "main"; | ||
| 512 | shader_stages[1].pSpecializationInfo = nullptr; | ||
| 424 | 513 | ||
| 425 | const auto vertex_binding_description = ScreenRectVertex::GetDescription(); | 514 | const auto vertex_binding_description = ScreenRectVertex::GetDescription(); |
| 426 | const auto vertex_attrs_description = ScreenRectVertex::GetAttributes(); | 515 | const auto vertex_attrs_description = ScreenRectVertex::GetAttributes(); |
| 427 | const vk::PipelineVertexInputStateCreateInfo vertex_input( | ||
| 428 | {}, 1, &vertex_binding_description, static_cast<u32>(vertex_attrs_description.size()), | ||
| 429 | vertex_attrs_description.data()); | ||
| 430 | |||
| 431 | const vk::PipelineInputAssemblyStateCreateInfo input_assembly( | ||
| 432 | {}, vk::PrimitiveTopology::eTriangleStrip, false); | ||
| 433 | |||
| 434 | // Set a dummy viewport, it's going to be replaced by dynamic states. | ||
| 435 | const vk::Viewport viewport(0.0f, 0.0f, 1.0f, 1.0f, 0.0f, 1.0f); | ||
| 436 | const vk::Rect2D scissor({0, 0}, {1, 1}); | ||
| 437 | 516 | ||
| 438 | const vk::PipelineViewportStateCreateInfo viewport_state({}, 1, &viewport, 1, &scissor); | 517 | VkPipelineVertexInputStateCreateInfo vertex_input_ci; |
| 439 | 518 | vertex_input_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; | |
| 440 | const vk::PipelineRasterizationStateCreateInfo rasterizer( | 519 | vertex_input_ci.pNext = nullptr; |
| 441 | {}, false, false, vk::PolygonMode::eFill, vk::CullModeFlagBits::eNone, | 520 | vertex_input_ci.flags = 0; |
| 442 | vk::FrontFace::eClockwise, false, 0.0f, 0.0f, 0.0f, 1.0f); | 521 | vertex_input_ci.vertexBindingDescriptionCount = 1; |
| 443 | 522 | vertex_input_ci.pVertexBindingDescriptions = &vertex_binding_description; | |
| 444 | const vk::PipelineMultisampleStateCreateInfo multisampling({}, vk::SampleCountFlagBits::e1, | 523 | vertex_input_ci.vertexAttributeDescriptionCount = u32{vertex_attrs_description.size()}; |
| 445 | false, 0.0f, nullptr, false, false); | 524 | vertex_input_ci.pVertexAttributeDescriptions = vertex_attrs_description.data(); |
| 446 | 525 | ||
| 447 | const vk::PipelineColorBlendAttachmentState color_blend_attachment( | 526 | VkPipelineInputAssemblyStateCreateInfo input_assembly_ci; |
| 448 | false, vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd, | 527 | input_assembly_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO; |
| 449 | vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd, | 528 | input_assembly_ci.pNext = nullptr; |
| 450 | vk::ColorComponentFlagBits::eR | vk::ColorComponentFlagBits::eG | | 529 | input_assembly_ci.flags = 0; |
| 451 | vk::ColorComponentFlagBits::eB | vk::ColorComponentFlagBits::eA); | 530 | input_assembly_ci.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP; |
| 452 | 531 | input_assembly_ci.primitiveRestartEnable = VK_FALSE; | |
| 453 | const vk::PipelineColorBlendStateCreateInfo color_blending( | 532 | |
| 454 | {}, false, vk::LogicOp::eCopy, 1, &color_blend_attachment, {0.0f, 0.0f, 0.0f, 0.0f}); | 533 | VkPipelineViewportStateCreateInfo viewport_state_ci; |
| 455 | 534 | viewport_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; | |
| 456 | const std::array<vk::DynamicState, 2> dynamic_states = {vk::DynamicState::eViewport, | 535 | viewport_state_ci.pNext = nullptr; |
| 457 | vk::DynamicState::eScissor}; | 536 | viewport_state_ci.flags = 0; |
| 458 | 537 | viewport_state_ci.viewportCount = 1; | |
| 459 | const vk::PipelineDynamicStateCreateInfo dynamic_state( | 538 | viewport_state_ci.scissorCount = 1; |
| 460 | {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data()); | 539 | |
| 461 | 540 | VkPipelineRasterizationStateCreateInfo rasterization_ci; | |
| 462 | const vk::GraphicsPipelineCreateInfo pipeline_ci( | 541 | rasterization_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; |
| 463 | {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input, | 542 | rasterization_ci.pNext = nullptr; |
| 464 | &input_assembly, nullptr, &viewport_state, &rasterizer, &multisampling, nullptr, | 543 | rasterization_ci.flags = 0; |
| 465 | &color_blending, &dynamic_state, *pipeline_layout, *renderpass, 0, nullptr, 0); | 544 | rasterization_ci.depthClampEnable = VK_FALSE; |
| 466 | 545 | rasterization_ci.rasterizerDiscardEnable = VK_FALSE; | |
| 467 | const auto dev = device.GetLogical(); | 546 | rasterization_ci.polygonMode = VK_POLYGON_MODE_FILL; |
| 468 | const auto& dld = device.GetDispatchLoader(); | 547 | rasterization_ci.cullMode = VK_CULL_MODE_NONE; |
| 469 | pipeline = dev.createGraphicsPipelineUnique({}, pipeline_ci, nullptr, dld); | 548 | rasterization_ci.frontFace = VK_FRONT_FACE_CLOCKWISE; |
| 549 | rasterization_ci.depthBiasEnable = VK_FALSE; | ||
| 550 | rasterization_ci.depthBiasConstantFactor = 0.0f; | ||
| 551 | rasterization_ci.depthBiasClamp = 0.0f; | ||
| 552 | rasterization_ci.depthBiasSlopeFactor = 0.0f; | ||
| 553 | rasterization_ci.lineWidth = 1.0f; | ||
| 554 | |||
| 555 | VkPipelineMultisampleStateCreateInfo multisampling_ci; | ||
| 556 | multisampling_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; | ||
| 557 | multisampling_ci.pNext = nullptr; | ||
| 558 | multisampling_ci.flags = 0; | ||
| 559 | multisampling_ci.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT; | ||
| 560 | multisampling_ci.sampleShadingEnable = VK_FALSE; | ||
| 561 | multisampling_ci.minSampleShading = 0.0f; | ||
| 562 | multisampling_ci.pSampleMask = nullptr; | ||
| 563 | multisampling_ci.alphaToCoverageEnable = VK_FALSE; | ||
| 564 | multisampling_ci.alphaToOneEnable = VK_FALSE; | ||
| 565 | |||
| 566 | VkPipelineColorBlendAttachmentState color_blend_attachment; | ||
| 567 | color_blend_attachment.blendEnable = VK_FALSE; | ||
| 568 | color_blend_attachment.srcColorBlendFactor = VK_BLEND_FACTOR_ZERO; | ||
| 569 | color_blend_attachment.dstColorBlendFactor = VK_BLEND_FACTOR_ZERO; | ||
| 570 | color_blend_attachment.colorBlendOp = VK_BLEND_OP_ADD; | ||
| 571 | color_blend_attachment.srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO; | ||
| 572 | color_blend_attachment.dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO; | ||
| 573 | color_blend_attachment.alphaBlendOp = VK_BLEND_OP_ADD; | ||
| 574 | color_blend_attachment.colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT | | ||
| 575 | VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT; | ||
| 576 | |||
| 577 | VkPipelineColorBlendStateCreateInfo color_blend_ci; | ||
| 578 | color_blend_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; | ||
| 579 | color_blend_ci.flags = 0; | ||
| 580 | color_blend_ci.pNext = nullptr; | ||
| 581 | color_blend_ci.logicOpEnable = VK_FALSE; | ||
| 582 | color_blend_ci.logicOp = VK_LOGIC_OP_COPY; | ||
| 583 | color_blend_ci.attachmentCount = 1; | ||
| 584 | color_blend_ci.pAttachments = &color_blend_attachment; | ||
| 585 | color_blend_ci.blendConstants[0] = 0.0f; | ||
| 586 | color_blend_ci.blendConstants[1] = 0.0f; | ||
| 587 | color_blend_ci.blendConstants[2] = 0.0f; | ||
| 588 | color_blend_ci.blendConstants[3] = 0.0f; | ||
| 589 | |||
| 590 | static constexpr std::array dynamic_states = {VK_DYNAMIC_STATE_VIEWPORT, | ||
| 591 | VK_DYNAMIC_STATE_SCISSOR}; | ||
| 592 | VkPipelineDynamicStateCreateInfo dynamic_state_ci; | ||
| 593 | dynamic_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO; | ||
| 594 | dynamic_state_ci.pNext = nullptr; | ||
| 595 | dynamic_state_ci.flags = 0; | ||
| 596 | dynamic_state_ci.dynamicStateCount = static_cast<u32>(dynamic_states.size()); | ||
| 597 | dynamic_state_ci.pDynamicStates = dynamic_states.data(); | ||
| 598 | |||
| 599 | VkGraphicsPipelineCreateInfo pipeline_ci; | ||
| 600 | pipeline_ci.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; | ||
| 601 | pipeline_ci.pNext = nullptr; | ||
| 602 | pipeline_ci.flags = 0; | ||
| 603 | pipeline_ci.stageCount = static_cast<u32>(shader_stages.size()); | ||
| 604 | pipeline_ci.pStages = shader_stages.data(); | ||
| 605 | pipeline_ci.pVertexInputState = &vertex_input_ci; | ||
| 606 | pipeline_ci.pInputAssemblyState = &input_assembly_ci; | ||
| 607 | pipeline_ci.pTessellationState = nullptr; | ||
| 608 | pipeline_ci.pViewportState = &viewport_state_ci; | ||
| 609 | pipeline_ci.pRasterizationState = &rasterization_ci; | ||
| 610 | pipeline_ci.pMultisampleState = &multisampling_ci; | ||
| 611 | pipeline_ci.pDepthStencilState = nullptr; | ||
| 612 | pipeline_ci.pColorBlendState = &color_blend_ci; | ||
| 613 | pipeline_ci.pDynamicState = &dynamic_state_ci; | ||
| 614 | pipeline_ci.layout = *pipeline_layout; | ||
| 615 | pipeline_ci.renderPass = *renderpass; | ||
| 616 | pipeline_ci.subpass = 0; | ||
| 617 | pipeline_ci.basePipelineHandle = 0; | ||
| 618 | pipeline_ci.basePipelineIndex = 0; | ||
| 619 | |||
| 620 | pipeline = device.GetLogical().CreateGraphicsPipeline(pipeline_ci); | ||
| 470 | } | 621 | } |
| 471 | 622 | ||
| 472 | void VKBlitScreen::CreateSampler() { | 623 | void VKBlitScreen::CreateSampler() { |
| 473 | const auto dev = device.GetLogical(); | 624 | VkSamplerCreateInfo ci; |
| 474 | const auto& dld = device.GetDispatchLoader(); | 625 | ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO; |
| 475 | const vk::SamplerCreateInfo sampler_ci( | 626 | ci.pNext = nullptr; |
| 476 | {}, vk::Filter::eLinear, vk::Filter::eLinear, vk::SamplerMipmapMode::eLinear, | 627 | ci.flags = 0; |
| 477 | vk::SamplerAddressMode::eClampToBorder, vk::SamplerAddressMode::eClampToBorder, | 628 | ci.magFilter = VK_FILTER_LINEAR; |
| 478 | vk::SamplerAddressMode::eClampToBorder, 0.0f, false, 0.0f, false, vk::CompareOp::eNever, | 629 | ci.minFilter = VK_FILTER_NEAREST; |
| 479 | 0.0f, 0.0f, vk::BorderColor::eFloatOpaqueBlack, false); | 630 | ci.mipmapMode = VK_SAMPLER_MIPMAP_MODE_LINEAR; |
| 480 | sampler = dev.createSamplerUnique(sampler_ci, nullptr, dld); | 631 | ci.addressModeU = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; |
| 632 | ci.addressModeV = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; | ||
| 633 | ci.addressModeW = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; | ||
| 634 | ci.mipLodBias = 0.0f; | ||
| 635 | ci.anisotropyEnable = VK_FALSE; | ||
| 636 | ci.maxAnisotropy = 0.0f; | ||
| 637 | ci.compareEnable = VK_FALSE; | ||
| 638 | ci.compareOp = VK_COMPARE_OP_NEVER; | ||
| 639 | ci.minLod = 0.0f; | ||
| 640 | ci.maxLod = 0.0f; | ||
| 641 | ci.borderColor = VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK; | ||
| 642 | ci.unnormalizedCoordinates = VK_FALSE; | ||
| 643 | |||
| 644 | sampler = device.GetLogical().CreateSampler(ci); | ||
| 481 | } | 645 | } |
| 482 | 646 | ||
| 483 | void VKBlitScreen::CreateFramebuffers() { | 647 | void VKBlitScreen::CreateFramebuffers() { |
| 484 | const vk::Extent2D size{swapchain.GetSize()}; | 648 | const VkExtent2D size{swapchain.GetSize()}; |
| 485 | framebuffers.clear(); | ||
| 486 | framebuffers.resize(image_count); | 649 | framebuffers.resize(image_count); |
| 487 | 650 | ||
| 488 | const auto dev = device.GetLogical(); | 651 | VkFramebufferCreateInfo ci; |
| 489 | const auto& dld = device.GetDispatchLoader(); | 652 | ci.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; |
| 653 | ci.pNext = nullptr; | ||
| 654 | ci.flags = 0; | ||
| 655 | ci.renderPass = *renderpass; | ||
| 656 | ci.attachmentCount = 1; | ||
| 657 | ci.width = size.width; | ||
| 658 | ci.height = size.height; | ||
| 659 | ci.layers = 1; | ||
| 490 | 660 | ||
| 491 | for (std::size_t i = 0; i < image_count; ++i) { | 661 | for (std::size_t i = 0; i < image_count; ++i) { |
| 492 | const vk::ImageView image_view{swapchain.GetImageViewIndex(i)}; | 662 | const VkImageView image_view{swapchain.GetImageViewIndex(i)}; |
| 493 | const vk::FramebufferCreateInfo framebuffer_ci({}, *renderpass, 1, &image_view, size.width, | 663 | ci.pAttachments = &image_view; |
| 494 | size.height, 1); | 664 | framebuffers[i] = device.GetLogical().CreateFramebuffer(ci); |
| 495 | framebuffers[i] = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld); | ||
| 496 | } | 665 | } |
| 497 | } | 666 | } |
| 498 | 667 | ||
| @@ -507,54 +676,86 @@ void VKBlitScreen::ReleaseRawImages() { | |||
| 507 | } | 676 | } |
| 508 | 677 | ||
| 509 | void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) { | 678 | void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) { |
| 510 | const auto dev = device.GetLogical(); | 679 | VkBufferCreateInfo ci; |
| 511 | const auto& dld = device.GetDispatchLoader(); | 680 | ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; |
| 512 | 681 | ci.pNext = nullptr; | |
| 513 | const vk::BufferCreateInfo buffer_ci({}, CalculateBufferSize(framebuffer), | 682 | ci.flags = 0; |
| 514 | vk::BufferUsageFlagBits::eTransferSrc | | 683 | ci.size = CalculateBufferSize(framebuffer); |
| 515 | vk::BufferUsageFlagBits::eVertexBuffer | | 684 | ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | |
| 516 | vk::BufferUsageFlagBits::eUniformBuffer, | 685 | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT; |
| 517 | vk::SharingMode::eExclusive, 0, nullptr); | 686 | ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; |
| 518 | buffer = dev.createBufferUnique(buffer_ci, nullptr, dld); | 687 | ci.queueFamilyIndexCount = 0; |
| 519 | buffer_commit = memory_manager.Commit(*buffer, true); | 688 | ci.pQueueFamilyIndices = nullptr; |
| 689 | |||
| 690 | buffer = device.GetLogical().CreateBuffer(ci); | ||
| 691 | buffer_commit = memory_manager.Commit(buffer, true); | ||
| 520 | } | 692 | } |
| 521 | 693 | ||
| 522 | void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) { | 694 | void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) { |
| 523 | raw_images.resize(image_count); | 695 | raw_images.resize(image_count); |
| 524 | raw_buffer_commits.resize(image_count); | 696 | raw_buffer_commits.resize(image_count); |
| 525 | 697 | ||
| 526 | const auto format = GetFormat(framebuffer); | 698 | VkImageCreateInfo ci; |
| 699 | ci.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO; | ||
| 700 | ci.pNext = nullptr; | ||
| 701 | ci.flags = 0; | ||
| 702 | ci.imageType = VK_IMAGE_TYPE_2D; | ||
| 703 | ci.format = GetFormat(framebuffer); | ||
| 704 | ci.extent.width = framebuffer.width; | ||
| 705 | ci.extent.height = framebuffer.height; | ||
| 706 | ci.extent.depth = 1; | ||
| 707 | ci.mipLevels = 1; | ||
| 708 | ci.arrayLayers = 1; | ||
| 709 | ci.samples = VK_SAMPLE_COUNT_1_BIT; | ||
| 710 | ci.tiling = VK_IMAGE_TILING_LINEAR; | ||
| 711 | ci.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT; | ||
| 712 | ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; | ||
| 713 | ci.queueFamilyIndexCount = 0; | ||
| 714 | ci.pQueueFamilyIndices = nullptr; | ||
| 715 | ci.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; | ||
| 716 | |||
| 527 | for (std::size_t i = 0; i < image_count; ++i) { | 717 | for (std::size_t i = 0; i < image_count; ++i) { |
| 528 | const vk::ImageCreateInfo image_ci( | 718 | raw_images[i] = std::make_unique<VKImage>(device, scheduler, ci, VK_IMAGE_ASPECT_COLOR_BIT); |
| 529 | {}, vk::ImageType::e2D, format, {framebuffer.width, framebuffer.height, 1}, 1, 1, | ||
| 530 | vk::SampleCountFlagBits::e1, vk::ImageTiling::eOptimal, | ||
| 531 | vk::ImageUsageFlagBits::eTransferDst | vk::ImageUsageFlagBits::eSampled, | ||
| 532 | vk::SharingMode::eExclusive, 0, nullptr, vk::ImageLayout::eUndefined); | ||
| 533 | |||
| 534 | raw_images[i] = | ||
| 535 | std::make_unique<VKImage>(device, scheduler, image_ci, vk::ImageAspectFlagBits::eColor); | ||
| 536 | raw_buffer_commits[i] = memory_manager.Commit(raw_images[i]->GetHandle(), false); | 719 | raw_buffer_commits[i] = memory_manager.Commit(raw_images[i]->GetHandle(), false); |
| 537 | } | 720 | } |
| 538 | } | 721 | } |
| 539 | 722 | ||
| 540 | void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const { | 723 | void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, VkImageView image_view) const { |
| 541 | const vk::DescriptorSet descriptor_set = descriptor_sets[image_index]; | 724 | VkDescriptorBufferInfo buffer_info; |
| 542 | 725 | buffer_info.buffer = *buffer; | |
| 543 | const vk::DescriptorBufferInfo buffer_info(*buffer, offsetof(BufferData, uniform), | 726 | buffer_info.offset = offsetof(BufferData, uniform); |
| 544 | sizeof(BufferData::uniform)); | 727 | buffer_info.range = sizeof(BufferData::uniform); |
| 545 | const vk::WriteDescriptorSet ubo_write(descriptor_set, 0, 0, 1, | 728 | |
| 546 | vk::DescriptorType::eUniformBuffer, nullptr, | 729 | VkWriteDescriptorSet ubo_write; |
| 547 | &buffer_info, nullptr); | 730 | ubo_write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET; |
| 548 | 731 | ubo_write.pNext = nullptr; | |
| 549 | const vk::DescriptorImageInfo image_info(*sampler, image_view, | 732 | ubo_write.dstSet = descriptor_sets[image_index]; |
| 550 | vk::ImageLayout::eShaderReadOnlyOptimal); | 733 | ubo_write.dstBinding = 0; |
| 551 | const vk::WriteDescriptorSet sampler_write(descriptor_set, 1, 0, 1, | 734 | ubo_write.dstArrayElement = 0; |
| 552 | vk::DescriptorType::eCombinedImageSampler, | 735 | ubo_write.descriptorCount = 1; |
| 553 | &image_info, nullptr, nullptr); | 736 | ubo_write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; |
| 554 | 737 | ubo_write.pImageInfo = nullptr; | |
| 555 | const auto dev = device.GetLogical(); | 738 | ubo_write.pBufferInfo = &buffer_info; |
| 556 | const auto& dld = device.GetDispatchLoader(); | 739 | ubo_write.pTexelBufferView = nullptr; |
| 557 | dev.updateDescriptorSets({ubo_write, sampler_write}, {}, dld); | 740 | |
| 741 | VkDescriptorImageInfo image_info; | ||
| 742 | image_info.sampler = *sampler; | ||
| 743 | image_info.imageView = image_view; | ||
| 744 | image_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; | ||
| 745 | |||
| 746 | VkWriteDescriptorSet sampler_write; | ||
| 747 | sampler_write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET; | ||
| 748 | sampler_write.pNext = nullptr; | ||
| 749 | sampler_write.dstSet = descriptor_sets[image_index]; | ||
| 750 | sampler_write.dstBinding = 1; | ||
| 751 | sampler_write.dstArrayElement = 0; | ||
| 752 | sampler_write.descriptorCount = 1; | ||
| 753 | sampler_write.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; | ||
| 754 | sampler_write.pImageInfo = &image_info; | ||
| 755 | sampler_write.pBufferInfo = nullptr; | ||
| 756 | sampler_write.pTexelBufferView = nullptr; | ||
| 757 | |||
| 758 | device.GetLogical().UpdateDescriptorSets(std::array{ubo_write, sampler_write}, {}); | ||
| 558 | } | 759 | } |
| 559 | 760 | ||
| 560 | void VKBlitScreen::SetUniformData(BufferData& data, | 761 | void VKBlitScreen::SetUniformData(BufferData& data, |
diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.h b/src/video_core/renderer_vulkan/vk_blit_screen.h index ea680b3f5..5eb544aea 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.h +++ b/src/video_core/renderer_vulkan/vk_blit_screen.h | |||
| @@ -8,9 +8,9 @@ | |||
| 8 | #include <memory> | 8 | #include <memory> |
| 9 | #include <tuple> | 9 | #include <tuple> |
| 10 | 10 | ||
| 11 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 12 | #include "video_core/renderer_vulkan/vk_memory_manager.h" | 11 | #include "video_core/renderer_vulkan/vk_memory_manager.h" |
| 13 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 12 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 13 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 14 | 14 | ||
| 15 | namespace Core { | 15 | namespace Core { |
| 16 | class System; | 16 | class System; |
| @@ -49,8 +49,8 @@ public: | |||
| 49 | 49 | ||
| 50 | void Recreate(); | 50 | void Recreate(); |
| 51 | 51 | ||
| 52 | std::tuple<VKFence&, vk::Semaphore> Draw(const Tegra::FramebufferConfig& framebuffer, | 52 | std::tuple<VKFence&, VkSemaphore> Draw(const Tegra::FramebufferConfig& framebuffer, |
| 53 | bool use_accelerated); | 53 | bool use_accelerated); |
| 54 | 54 | ||
| 55 | private: | 55 | private: |
| 56 | struct BufferData; | 56 | struct BufferData; |
| @@ -74,7 +74,7 @@ private: | |||
| 74 | void CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer); | 74 | void CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer); |
| 75 | void CreateRawImages(const Tegra::FramebufferConfig& framebuffer); | 75 | void CreateRawImages(const Tegra::FramebufferConfig& framebuffer); |
| 76 | 76 | ||
| 77 | void UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const; | 77 | void UpdateDescriptorSet(std::size_t image_index, VkImageView image_view) const; |
| 78 | void SetUniformData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const; | 78 | void SetUniformData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const; |
| 79 | void SetVertexData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const; | 79 | void SetVertexData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const; |
| 80 | 80 | ||
| @@ -93,23 +93,23 @@ private: | |||
| 93 | const std::size_t image_count; | 93 | const std::size_t image_count; |
| 94 | const VKScreenInfo& screen_info; | 94 | const VKScreenInfo& screen_info; |
| 95 | 95 | ||
| 96 | UniqueShaderModule vertex_shader; | 96 | vk::ShaderModule vertex_shader; |
| 97 | UniqueShaderModule fragment_shader; | 97 | vk::ShaderModule fragment_shader; |
| 98 | UniqueDescriptorPool descriptor_pool; | 98 | vk::DescriptorPool descriptor_pool; |
| 99 | UniqueDescriptorSetLayout descriptor_set_layout; | 99 | vk::DescriptorSetLayout descriptor_set_layout; |
| 100 | UniquePipelineLayout pipeline_layout; | 100 | vk::PipelineLayout pipeline_layout; |
| 101 | UniquePipeline pipeline; | 101 | vk::Pipeline pipeline; |
| 102 | UniqueRenderPass renderpass; | 102 | vk::RenderPass renderpass; |
| 103 | std::vector<UniqueFramebuffer> framebuffers; | 103 | std::vector<vk::Framebuffer> framebuffers; |
| 104 | std::vector<vk::DescriptorSet> descriptor_sets; | 104 | vk::DescriptorSets descriptor_sets; |
| 105 | UniqueSampler sampler; | 105 | vk::Sampler sampler; |
| 106 | 106 | ||
| 107 | UniqueBuffer buffer; | 107 | vk::Buffer buffer; |
| 108 | VKMemoryCommit buffer_commit; | 108 | VKMemoryCommit buffer_commit; |
| 109 | 109 | ||
| 110 | std::vector<std::unique_ptr<VKFenceWatch>> watches; | 110 | std::vector<std::unique_ptr<VKFenceWatch>> watches; |
| 111 | 111 | ||
| 112 | std::vector<UniqueSemaphore> semaphores; | 112 | std::vector<vk::Semaphore> semaphores; |
| 113 | std::vector<std::unique_ptr<VKImage>> raw_images; | 113 | std::vector<std::unique_ptr<VKImage>> raw_images; |
| 114 | std::vector<VKMemoryCommit> raw_buffer_commits; | 114 | std::vector<VKMemoryCommit> raw_buffer_commits; |
| 115 | u32 raw_width = 0; | 115 | u32 raw_width = 0; |
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp index 326d74f29..0d167afbd 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp | |||
| @@ -11,32 +11,31 @@ | |||
| 11 | #include "common/assert.h" | 11 | #include "common/assert.h" |
| 12 | #include "common/bit_util.h" | 12 | #include "common/bit_util.h" |
| 13 | #include "core/core.h" | 13 | #include "core/core.h" |
| 14 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 15 | #include "video_core/renderer_vulkan/vk_buffer_cache.h" | 14 | #include "video_core/renderer_vulkan/vk_buffer_cache.h" |
| 16 | #include "video_core/renderer_vulkan/vk_device.h" | 15 | #include "video_core/renderer_vulkan/vk_device.h" |
| 17 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 16 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 18 | #include "video_core/renderer_vulkan/vk_stream_buffer.h" | 17 | #include "video_core/renderer_vulkan/vk_stream_buffer.h" |
| 18 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 19 | 19 | ||
| 20 | namespace Vulkan { | 20 | namespace Vulkan { |
| 21 | 21 | ||
| 22 | namespace { | 22 | namespace { |
| 23 | 23 | ||
| 24 | const auto BufferUsage = | 24 | constexpr VkBufferUsageFlags BUFFER_USAGE = |
| 25 | vk::BufferUsageFlagBits::eVertexBuffer | vk::BufferUsageFlagBits::eIndexBuffer | | 25 | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT | |
| 26 | vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer; | 26 | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT; |
| 27 | 27 | ||
| 28 | const auto UploadPipelineStage = | 28 | constexpr VkPipelineStageFlags UPLOAD_PIPELINE_STAGE = |
| 29 | vk::PipelineStageFlagBits::eTransfer | vk::PipelineStageFlagBits::eVertexInput | | 29 | VK_PIPELINE_STAGE_TRANSFER_BIT | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | |
| 30 | vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader | | 30 | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | |
| 31 | vk::PipelineStageFlagBits::eComputeShader; | 31 | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; |
| 32 | 32 | ||
| 33 | const auto UploadAccessBarriers = | 33 | constexpr VkAccessFlags UPLOAD_ACCESS_BARRIERS = |
| 34 | vk::AccessFlagBits::eTransferRead | vk::AccessFlagBits::eShaderRead | | 34 | VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_UNIFORM_READ_BIT | |
| 35 | vk::AccessFlagBits::eUniformRead | vk::AccessFlagBits::eVertexAttributeRead | | 35 | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDEX_READ_BIT; |
| 36 | vk::AccessFlagBits::eIndexRead; | ||
| 37 | 36 | ||
| 38 | auto CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) { | 37 | std::unique_ptr<VKStreamBuffer> CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) { |
| 39 | return std::make_unique<VKStreamBuffer>(device, scheduler, BufferUsage); | 38 | return std::make_unique<VKStreamBuffer>(device, scheduler, BUFFER_USAGE); |
| 40 | } | 39 | } |
| 41 | 40 | ||
| 42 | } // Anonymous namespace | 41 | } // Anonymous namespace |
| @@ -44,15 +43,18 @@ auto CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) { | |||
| 44 | CachedBufferBlock::CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager, | 43 | CachedBufferBlock::CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager, |
| 45 | VAddr cpu_addr, std::size_t size) | 44 | VAddr cpu_addr, std::size_t size) |
| 46 | : VideoCommon::BufferBlock{cpu_addr, size} { | 45 | : VideoCommon::BufferBlock{cpu_addr, size} { |
| 47 | const vk::BufferCreateInfo buffer_ci({}, static_cast<vk::DeviceSize>(size), | 46 | VkBufferCreateInfo ci; |
| 48 | BufferUsage | vk::BufferUsageFlagBits::eTransferSrc | | 47 | ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; |
| 49 | vk::BufferUsageFlagBits::eTransferDst, | 48 | ci.pNext = nullptr; |
| 50 | vk::SharingMode::eExclusive, 0, nullptr); | 49 | ci.flags = 0; |
| 51 | 50 | ci.size = static_cast<VkDeviceSize>(size); | |
| 52 | const auto& dld{device.GetDispatchLoader()}; | 51 | ci.usage = BUFFER_USAGE | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; |
| 53 | const auto dev{device.GetLogical()}; | 52 | ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; |
| 54 | buffer.handle = dev.createBufferUnique(buffer_ci, nullptr, dld); | 53 | ci.queueFamilyIndexCount = 0; |
| 55 | buffer.commit = memory_manager.Commit(*buffer.handle, false); | 54 | ci.pQueueFamilyIndices = nullptr; |
| 55 | |||
| 56 | buffer.handle = device.GetLogical().CreateBuffer(ci); | ||
| 57 | buffer.commit = memory_manager.Commit(buffer.handle, false); | ||
| 56 | } | 58 | } |
| 57 | 59 | ||
| 58 | CachedBufferBlock::~CachedBufferBlock() = default; | 60 | CachedBufferBlock::~CachedBufferBlock() = default; |
| @@ -60,9 +62,9 @@ CachedBufferBlock::~CachedBufferBlock() = default; | |||
| 60 | VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system, | 62 | VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system, |
| 61 | const VKDevice& device, VKMemoryManager& memory_manager, | 63 | const VKDevice& device, VKMemoryManager& memory_manager, |
| 62 | VKScheduler& scheduler, VKStagingBufferPool& staging_pool) | 64 | VKScheduler& scheduler, VKStagingBufferPool& staging_pool) |
| 63 | : VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer>{rasterizer, system, | 65 | : VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer>{rasterizer, system, |
| 64 | CreateStreamBuffer(device, | 66 | CreateStreamBuffer(device, |
| 65 | scheduler)}, | 67 | scheduler)}, |
| 66 | device{device}, memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{ | 68 | device{device}, memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{ |
| 67 | staging_pool} {} | 69 | staging_pool} {} |
| 68 | 70 | ||
| @@ -72,18 +74,18 @@ Buffer VKBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) { | |||
| 72 | return std::make_shared<CachedBufferBlock>(device, memory_manager, cpu_addr, size); | 74 | return std::make_shared<CachedBufferBlock>(device, memory_manager, cpu_addr, size); |
| 73 | } | 75 | } |
| 74 | 76 | ||
| 75 | const vk::Buffer* VKBufferCache::ToHandle(const Buffer& buffer) { | 77 | const VkBuffer* VKBufferCache::ToHandle(const Buffer& buffer) { |
| 76 | return buffer->GetHandle(); | 78 | return buffer->GetHandle(); |
| 77 | } | 79 | } |
| 78 | 80 | ||
| 79 | const vk::Buffer* VKBufferCache::GetEmptyBuffer(std::size_t size) { | 81 | const VkBuffer* VKBufferCache::GetEmptyBuffer(std::size_t size) { |
| 80 | size = std::max(size, std::size_t(4)); | 82 | size = std::max(size, std::size_t(4)); |
| 81 | const auto& empty = staging_pool.GetUnusedBuffer(size, false); | 83 | const auto& empty = staging_pool.GetUnusedBuffer(size, false); |
| 82 | scheduler.RequestOutsideRenderPassOperationContext(); | 84 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 83 | scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf, auto& dld) { | 85 | scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf) { |
| 84 | cmdbuf.fillBuffer(buffer, 0, size, 0, dld); | 86 | cmdbuf.FillBuffer(buffer, 0, size, 0); |
| 85 | }); | 87 | }); |
| 86 | return &*empty.handle; | 88 | return empty.handle.address(); |
| 87 | } | 89 | } |
| 88 | 90 | ||
| 89 | void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, | 91 | void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, |
| @@ -93,14 +95,21 @@ void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, st | |||
| 93 | 95 | ||
| 94 | scheduler.RequestOutsideRenderPassOperationContext(); | 96 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 95 | scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset, | 97 | scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset, |
| 96 | size](auto cmdbuf, auto& dld) { | 98 | size](vk::CommandBuffer cmdbuf) { |
| 97 | cmdbuf.copyBuffer(staging, buffer, {{0, offset, size}}, dld); | 99 | cmdbuf.CopyBuffer(staging, buffer, VkBufferCopy{0, offset, size}); |
| 98 | cmdbuf.pipelineBarrier( | 100 | |
| 99 | vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {}, | 101 | VkBufferMemoryBarrier barrier; |
| 100 | {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers, | 102 | barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; |
| 101 | VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, | 103 | barrier.pNext = nullptr; |
| 102 | offset, size)}, | 104 | barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT; |
| 103 | {}, dld); | 105 | barrier.dstAccessMask = UPLOAD_ACCESS_BARRIERS; |
| 106 | barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 107 | barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 108 | barrier.buffer = buffer; | ||
| 109 | barrier.offset = offset; | ||
| 110 | barrier.size = size; | ||
| 111 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {}, | ||
| 112 | barrier, {}); | ||
| 104 | }); | 113 | }); |
| 105 | } | 114 | } |
| 106 | 115 | ||
| @@ -109,16 +118,23 @@ void VKBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, | |||
| 109 | const auto& staging = staging_pool.GetUnusedBuffer(size, true); | 118 | const auto& staging = staging_pool.GetUnusedBuffer(size, true); |
| 110 | scheduler.RequestOutsideRenderPassOperationContext(); | 119 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 111 | scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset, | 120 | scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset, |
| 112 | size](auto cmdbuf, auto& dld) { | 121 | size](vk::CommandBuffer cmdbuf) { |
| 113 | cmdbuf.pipelineBarrier( | 122 | VkBufferMemoryBarrier barrier; |
| 114 | vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader | | 123 | barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; |
| 115 | vk::PipelineStageFlagBits::eComputeShader, | 124 | barrier.pNext = nullptr; |
| 116 | vk::PipelineStageFlagBits::eTransfer, {}, {}, | 125 | barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; |
| 117 | {vk::BufferMemoryBarrier(vk::AccessFlagBits::eShaderWrite, | 126 | barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT; |
| 118 | vk::AccessFlagBits::eTransferRead, VK_QUEUE_FAMILY_IGNORED, | 127 | barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 119 | VK_QUEUE_FAMILY_IGNORED, buffer, offset, size)}, | 128 | barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 120 | {}, dld); | 129 | barrier.buffer = buffer; |
| 121 | cmdbuf.copyBuffer(buffer, staging, {{offset, 0, size}}, dld); | 130 | barrier.offset = offset; |
| 131 | barrier.size = size; | ||
| 132 | |||
| 133 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | | ||
| 134 | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | | ||
| 135 | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, | ||
| 136 | VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {}, barrier, {}); | ||
| 137 | cmdbuf.CopyBuffer(buffer, staging, VkBufferCopy{offset, 0, size}); | ||
| 122 | }); | 138 | }); |
| 123 | scheduler.Finish(); | 139 | scheduler.Finish(); |
| 124 | 140 | ||
| @@ -129,17 +145,30 @@ void VKBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t | |||
| 129 | std::size_t dst_offset, std::size_t size) { | 145 | std::size_t dst_offset, std::size_t size) { |
| 130 | scheduler.RequestOutsideRenderPassOperationContext(); | 146 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 131 | scheduler.Record([src_buffer = *src->GetHandle(), dst_buffer = *dst->GetHandle(), src_offset, | 147 | scheduler.Record([src_buffer = *src->GetHandle(), dst_buffer = *dst->GetHandle(), src_offset, |
| 132 | dst_offset, size](auto cmdbuf, auto& dld) { | 148 | dst_offset, size](vk::CommandBuffer cmdbuf) { |
| 133 | cmdbuf.copyBuffer(src_buffer, dst_buffer, {{src_offset, dst_offset, size}}, dld); | 149 | cmdbuf.CopyBuffer(src_buffer, dst_buffer, VkBufferCopy{src_offset, dst_offset, size}); |
| 134 | cmdbuf.pipelineBarrier( | 150 | |
| 135 | vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {}, | 151 | std::array<VkBufferMemoryBarrier, 2> barriers; |
| 136 | {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferRead, | 152 | barriers[0].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; |
| 137 | vk::AccessFlagBits::eShaderWrite, VK_QUEUE_FAMILY_IGNORED, | 153 | barriers[0].pNext = nullptr; |
| 138 | VK_QUEUE_FAMILY_IGNORED, src_buffer, src_offset, size), | 154 | barriers[0].srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT; |
| 139 | vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers, | 155 | barriers[0].dstAccessMask = VK_ACCESS_SHADER_WRITE_BIT; |
| 140 | VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, dst_buffer, | 156 | barriers[0].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 141 | dst_offset, size)}, | 157 | barriers[0].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 142 | {}, dld); | 158 | barriers[0].buffer = src_buffer; |
| 159 | barriers[0].offset = src_offset; | ||
| 160 | barriers[0].size = size; | ||
| 161 | barriers[1].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; | ||
| 162 | barriers[1].pNext = nullptr; | ||
| 163 | barriers[1].srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT; | ||
| 164 | barriers[1].dstAccessMask = UPLOAD_ACCESS_BARRIERS; | ||
| 165 | barriers[1].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 166 | barriers[1].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 167 | barriers[1].buffer = dst_buffer; | ||
| 168 | barriers[1].offset = dst_offset; | ||
| 169 | barriers[1].size = size; | ||
| 170 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {}, | ||
| 171 | barriers, {}); | ||
| 143 | }); | 172 | }); |
| 144 | } | 173 | } |
| 145 | 174 | ||
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h index 508214618..d3c23da98 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.h +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h | |||
| @@ -11,11 +11,11 @@ | |||
| 11 | #include "common/common_types.h" | 11 | #include "common/common_types.h" |
| 12 | #include "video_core/buffer_cache/buffer_cache.h" | 12 | #include "video_core/buffer_cache/buffer_cache.h" |
| 13 | #include "video_core/rasterizer_cache.h" | 13 | #include "video_core/rasterizer_cache.h" |
| 14 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 15 | #include "video_core/renderer_vulkan/vk_memory_manager.h" | 14 | #include "video_core/renderer_vulkan/vk_memory_manager.h" |
| 16 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 15 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 17 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" | 16 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" |
| 18 | #include "video_core/renderer_vulkan/vk_stream_buffer.h" | 17 | #include "video_core/renderer_vulkan/vk_stream_buffer.h" |
| 18 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 19 | 19 | ||
| 20 | namespace Core { | 20 | namespace Core { |
| 21 | class System; | 21 | class System; |
| @@ -33,8 +33,8 @@ public: | |||
| 33 | VAddr cpu_addr, std::size_t size); | 33 | VAddr cpu_addr, std::size_t size); |
| 34 | ~CachedBufferBlock(); | 34 | ~CachedBufferBlock(); |
| 35 | 35 | ||
| 36 | const vk::Buffer* GetHandle() const { | 36 | const VkBuffer* GetHandle() const { |
| 37 | return &*buffer.handle; | 37 | return buffer.handle.address(); |
| 38 | } | 38 | } |
| 39 | 39 | ||
| 40 | private: | 40 | private: |
| @@ -43,21 +43,21 @@ private: | |||
| 43 | 43 | ||
| 44 | using Buffer = std::shared_ptr<CachedBufferBlock>; | 44 | using Buffer = std::shared_ptr<CachedBufferBlock>; |
| 45 | 45 | ||
| 46 | class VKBufferCache final : public VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer> { | 46 | class VKBufferCache final : public VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer> { |
| 47 | public: | 47 | public: |
| 48 | explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system, | 48 | explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system, |
| 49 | const VKDevice& device, VKMemoryManager& memory_manager, | 49 | const VKDevice& device, VKMemoryManager& memory_manager, |
| 50 | VKScheduler& scheduler, VKStagingBufferPool& staging_pool); | 50 | VKScheduler& scheduler, VKStagingBufferPool& staging_pool); |
| 51 | ~VKBufferCache(); | 51 | ~VKBufferCache(); |
| 52 | 52 | ||
| 53 | const vk::Buffer* GetEmptyBuffer(std::size_t size) override; | 53 | const VkBuffer* GetEmptyBuffer(std::size_t size) override; |
| 54 | 54 | ||
| 55 | protected: | 55 | protected: |
| 56 | void WriteBarrier() override {} | 56 | void WriteBarrier() override {} |
| 57 | 57 | ||
| 58 | Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override; | 58 | Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override; |
| 59 | 59 | ||
| 60 | const vk::Buffer* ToHandle(const Buffer& buffer) override; | 60 | const VkBuffer* ToHandle(const Buffer& buffer) override; |
| 61 | 61 | ||
| 62 | void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, | 62 | void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, |
| 63 | const u8* data) override; | 63 | const u8* data) override; |
diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp index 7bdda3d79..9d92305f4 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp | |||
| @@ -10,13 +10,13 @@ | |||
| 10 | #include "common/alignment.h" | 10 | #include "common/alignment.h" |
| 11 | #include "common/assert.h" | 11 | #include "common/assert.h" |
| 12 | #include "common/common_types.h" | 12 | #include "common/common_types.h" |
| 13 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 14 | #include "video_core/renderer_vulkan/vk_compute_pass.h" | 13 | #include "video_core/renderer_vulkan/vk_compute_pass.h" |
| 15 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 14 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| 16 | #include "video_core/renderer_vulkan/vk_device.h" | 15 | #include "video_core/renderer_vulkan/vk_device.h" |
| 17 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 16 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 18 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" | 17 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" |
| 19 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 18 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 19 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 20 | 20 | ||
| 21 | namespace Vulkan { | 21 | namespace Vulkan { |
| 22 | 22 | ||
| @@ -114,6 +114,35 @@ constexpr u8 quad_array[] = { | |||
| 114 | 0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00, | 114 | 0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00, |
| 115 | 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00}; | 115 | 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00}; |
| 116 | 116 | ||
| 117 | VkDescriptorSetLayoutBinding BuildQuadArrayPassDescriptorSetLayoutBinding() { | ||
| 118 | VkDescriptorSetLayoutBinding binding; | ||
| 119 | binding.binding = 0; | ||
| 120 | binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; | ||
| 121 | binding.descriptorCount = 1; | ||
| 122 | binding.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 123 | binding.pImmutableSamplers = nullptr; | ||
| 124 | return binding; | ||
| 125 | } | ||
| 126 | |||
| 127 | VkDescriptorUpdateTemplateEntryKHR BuildQuadArrayPassDescriptorUpdateTemplateEntry() { | ||
| 128 | VkDescriptorUpdateTemplateEntryKHR entry; | ||
| 129 | entry.dstBinding = 0; | ||
| 130 | entry.dstArrayElement = 0; | ||
| 131 | entry.descriptorCount = 1; | ||
| 132 | entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; | ||
| 133 | entry.offset = 0; | ||
| 134 | entry.stride = sizeof(DescriptorUpdateEntry); | ||
| 135 | return entry; | ||
| 136 | } | ||
| 137 | |||
| 138 | VkPushConstantRange BuildQuadArrayPassPushConstantRange() { | ||
| 139 | VkPushConstantRange range; | ||
| 140 | range.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 141 | range.offset = 0; | ||
| 142 | range.size = sizeof(u32); | ||
| 143 | return range; | ||
| 144 | } | ||
| 145 | |||
| 117 | // Uint8 SPIR-V module. Generated from the "shaders/" directory. | 146 | // Uint8 SPIR-V module. Generated from the "shaders/" directory. |
| 118 | constexpr u8 uint8_pass[] = { | 147 | constexpr u8 uint8_pass[] = { |
| 119 | 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x2f, 0x00, 0x00, 0x00, | 148 | 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x2f, 0x00, 0x00, 0x00, |
| @@ -191,53 +220,111 @@ constexpr u8 uint8_pass[] = { | |||
| 191 | 0xf9, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, | 220 | 0xf9, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, |
| 192 | 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00}; | 221 | 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00}; |
| 193 | 222 | ||
| 223 | std::array<VkDescriptorSetLayoutBinding, 2> BuildUint8PassDescriptorSetBindings() { | ||
| 224 | std::array<VkDescriptorSetLayoutBinding, 2> bindings; | ||
| 225 | bindings[0].binding = 0; | ||
| 226 | bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; | ||
| 227 | bindings[0].descriptorCount = 1; | ||
| 228 | bindings[0].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 229 | bindings[0].pImmutableSamplers = nullptr; | ||
| 230 | bindings[1].binding = 1; | ||
| 231 | bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; | ||
| 232 | bindings[1].descriptorCount = 1; | ||
| 233 | bindings[1].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 234 | bindings[1].pImmutableSamplers = nullptr; | ||
| 235 | return bindings; | ||
| 236 | } | ||
| 237 | |||
| 238 | VkDescriptorUpdateTemplateEntryKHR BuildUint8PassDescriptorUpdateTemplateEntry() { | ||
| 239 | VkDescriptorUpdateTemplateEntryKHR entry; | ||
| 240 | entry.dstBinding = 0; | ||
| 241 | entry.dstArrayElement = 0; | ||
| 242 | entry.descriptorCount = 2; | ||
| 243 | entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; | ||
| 244 | entry.offset = 0; | ||
| 245 | entry.stride = sizeof(DescriptorUpdateEntry); | ||
| 246 | return entry; | ||
| 247 | } | ||
| 248 | |||
| 194 | } // Anonymous namespace | 249 | } // Anonymous namespace |
| 195 | 250 | ||
| 196 | VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool, | 251 | VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool, |
| 197 | const std::vector<vk::DescriptorSetLayoutBinding>& bindings, | 252 | vk::Span<VkDescriptorSetLayoutBinding> bindings, |
| 198 | const std::vector<vk::DescriptorUpdateTemplateEntry>& templates, | 253 | vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates, |
| 199 | const std::vector<vk::PushConstantRange> push_constants, | 254 | vk::Span<VkPushConstantRange> push_constants, std::size_t code_size, |
| 200 | std::size_t code_size, const u8* code) { | 255 | const u8* code) { |
| 201 | const auto dev = device.GetLogical(); | 256 | VkDescriptorSetLayoutCreateInfo descriptor_layout_ci; |
| 202 | const auto& dld = device.GetDispatchLoader(); | 257 | descriptor_layout_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; |
| 203 | 258 | descriptor_layout_ci.pNext = nullptr; | |
| 204 | const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci( | 259 | descriptor_layout_ci.flags = 0; |
| 205 | {}, static_cast<u32>(bindings.size()), bindings.data()); | 260 | descriptor_layout_ci.bindingCount = bindings.size(); |
| 206 | descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld); | 261 | descriptor_layout_ci.pBindings = bindings.data(); |
| 207 | 262 | descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(descriptor_layout_ci); | |
| 208 | const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout, | 263 | |
| 209 | static_cast<u32>(push_constants.size()), | 264 | VkPipelineLayoutCreateInfo pipeline_layout_ci; |
| 210 | push_constants.data()); | 265 | pipeline_layout_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; |
| 211 | layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld); | 266 | pipeline_layout_ci.pNext = nullptr; |
| 267 | pipeline_layout_ci.flags = 0; | ||
| 268 | pipeline_layout_ci.setLayoutCount = 1; | ||
| 269 | pipeline_layout_ci.pSetLayouts = descriptor_set_layout.address(); | ||
| 270 | pipeline_layout_ci.pushConstantRangeCount = push_constants.size(); | ||
| 271 | pipeline_layout_ci.pPushConstantRanges = push_constants.data(); | ||
| 272 | layout = device.GetLogical().CreatePipelineLayout(pipeline_layout_ci); | ||
| 212 | 273 | ||
| 213 | if (!templates.empty()) { | 274 | if (!templates.empty()) { |
| 214 | const vk::DescriptorUpdateTemplateCreateInfo template_ci( | 275 | VkDescriptorUpdateTemplateCreateInfoKHR template_ci; |
| 215 | {}, static_cast<u32>(templates.size()), templates.data(), | 276 | template_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR; |
| 216 | vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout, | 277 | template_ci.pNext = nullptr; |
| 217 | vk::PipelineBindPoint::eGraphics, *layout, 0); | 278 | template_ci.flags = 0; |
| 218 | descriptor_template = dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld); | 279 | template_ci.descriptorUpdateEntryCount = templates.size(); |
| 280 | template_ci.pDescriptorUpdateEntries = templates.data(); | ||
| 281 | template_ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR; | ||
| 282 | template_ci.descriptorSetLayout = *descriptor_set_layout; | ||
| 283 | template_ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; | ||
| 284 | template_ci.pipelineLayout = *layout; | ||
| 285 | template_ci.set = 0; | ||
| 286 | descriptor_template = device.GetLogical().CreateDescriptorUpdateTemplateKHR(template_ci); | ||
| 219 | 287 | ||
| 220 | descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout); | 288 | descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout); |
| 221 | } | 289 | } |
| 222 | 290 | ||
| 223 | auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1); | 291 | auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1); |
| 224 | std::memcpy(code_copy.get(), code, code_size); | 292 | std::memcpy(code_copy.get(), code, code_size); |
| 225 | const vk::ShaderModuleCreateInfo module_ci({}, code_size, code_copy.get()); | ||
| 226 | module = dev.createShaderModuleUnique(module_ci, nullptr, dld); | ||
| 227 | 293 | ||
| 228 | const vk::PipelineShaderStageCreateInfo stage_ci({}, vk::ShaderStageFlagBits::eCompute, *module, | 294 | VkShaderModuleCreateInfo module_ci; |
| 229 | "main", nullptr); | 295 | module_ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; |
| 296 | module_ci.pNext = nullptr; | ||
| 297 | module_ci.flags = 0; | ||
| 298 | module_ci.codeSize = code_size; | ||
| 299 | module_ci.pCode = code_copy.get(); | ||
| 300 | module = device.GetLogical().CreateShaderModule(module_ci); | ||
| 301 | |||
| 302 | VkComputePipelineCreateInfo pipeline_ci; | ||
| 303 | pipeline_ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO; | ||
| 304 | pipeline_ci.pNext = nullptr; | ||
| 305 | pipeline_ci.flags = 0; | ||
| 306 | pipeline_ci.layout = *layout; | ||
| 307 | pipeline_ci.basePipelineHandle = nullptr; | ||
| 308 | pipeline_ci.basePipelineIndex = 0; | ||
| 230 | 309 | ||
| 231 | const vk::ComputePipelineCreateInfo pipeline_ci({}, stage_ci, *layout, nullptr, 0); | 310 | VkPipelineShaderStageCreateInfo& stage_ci = pipeline_ci.stage; |
| 232 | pipeline = dev.createComputePipelineUnique(nullptr, pipeline_ci, nullptr, dld); | 311 | stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; |
| 312 | stage_ci.pNext = nullptr; | ||
| 313 | stage_ci.flags = 0; | ||
| 314 | stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 315 | stage_ci.module = *module; | ||
| 316 | stage_ci.pName = "main"; | ||
| 317 | stage_ci.pSpecializationInfo = nullptr; | ||
| 318 | |||
| 319 | pipeline = device.GetLogical().CreateComputePipeline(pipeline_ci); | ||
| 233 | } | 320 | } |
| 234 | 321 | ||
| 235 | VKComputePass::~VKComputePass() = default; | 322 | VKComputePass::~VKComputePass() = default; |
| 236 | 323 | ||
| 237 | vk::DescriptorSet VKComputePass::CommitDescriptorSet( | 324 | VkDescriptorSet VKComputePass::CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue, |
| 238 | VKUpdateDescriptorQueue& update_descriptor_queue, VKFence& fence) { | 325 | VKFence& fence) { |
| 239 | if (!descriptor_template) { | 326 | if (!descriptor_template) { |
| 240 | return {}; | 327 | return nullptr; |
| 241 | } | 328 | } |
| 242 | const auto set = descriptor_allocator->Commit(fence); | 329 | const auto set = descriptor_allocator->Commit(fence); |
| 243 | update_descriptor_queue.Send(*descriptor_template, set); | 330 | update_descriptor_queue.Send(*descriptor_template, set); |
| @@ -248,25 +335,21 @@ QuadArrayPass::QuadArrayPass(const VKDevice& device, VKScheduler& scheduler, | |||
| 248 | VKDescriptorPool& descriptor_pool, | 335 | VKDescriptorPool& descriptor_pool, |
| 249 | VKStagingBufferPool& staging_buffer_pool, | 336 | VKStagingBufferPool& staging_buffer_pool, |
| 250 | VKUpdateDescriptorQueue& update_descriptor_queue) | 337 | VKUpdateDescriptorQueue& update_descriptor_queue) |
| 251 | : VKComputePass(device, descriptor_pool, | 338 | : VKComputePass(device, descriptor_pool, BuildQuadArrayPassDescriptorSetLayoutBinding(), |
| 252 | {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1, | 339 | BuildQuadArrayPassDescriptorUpdateTemplateEntry(), |
| 253 | vk::ShaderStageFlagBits::eCompute, nullptr)}, | 340 | BuildQuadArrayPassPushConstantRange(), std::size(quad_array), quad_array), |
| 254 | {vk::DescriptorUpdateTemplateEntry(0, 0, 1, vk::DescriptorType::eStorageBuffer, | ||
| 255 | 0, sizeof(DescriptorUpdateEntry))}, | ||
| 256 | {vk::PushConstantRange(vk::ShaderStageFlagBits::eCompute, 0, sizeof(u32))}, | ||
| 257 | std::size(quad_array), quad_array), | ||
| 258 | scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool}, | 341 | scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool}, |
| 259 | update_descriptor_queue{update_descriptor_queue} {} | 342 | update_descriptor_queue{update_descriptor_queue} {} |
| 260 | 343 | ||
| 261 | QuadArrayPass::~QuadArrayPass() = default; | 344 | QuadArrayPass::~QuadArrayPass() = default; |
| 262 | 345 | ||
| 263 | std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) { | 346 | std::pair<const VkBuffer*, VkDeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) { |
| 264 | const u32 num_triangle_vertices = num_vertices * 6 / 4; | 347 | const u32 num_triangle_vertices = num_vertices * 6 / 4; |
| 265 | const std::size_t staging_size = num_triangle_vertices * sizeof(u32); | 348 | const std::size_t staging_size = num_triangle_vertices * sizeof(u32); |
| 266 | auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false); | 349 | auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false); |
| 267 | 350 | ||
| 268 | update_descriptor_queue.Acquire(); | 351 | update_descriptor_queue.Acquire(); |
| 269 | update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size); | 352 | update_descriptor_queue.AddBuffer(buffer.handle.address(), 0, staging_size); |
| 270 | const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence()); | 353 | const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence()); |
| 271 | 354 | ||
| 272 | scheduler.RequestOutsideRenderPassOperationContext(); | 355 | scheduler.RequestOutsideRenderPassOperationContext(); |
| @@ -274,66 +357,72 @@ std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_ver | |||
| 274 | ASSERT(num_vertices % 4 == 0); | 357 | ASSERT(num_vertices % 4 == 0); |
| 275 | const u32 num_quads = num_vertices / 4; | 358 | const u32 num_quads = num_vertices / 4; |
| 276 | scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, num_quads, | 359 | scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, num_quads, |
| 277 | first, set](auto cmdbuf, auto& dld) { | 360 | first, set](vk::CommandBuffer cmdbuf) { |
| 278 | constexpr u32 dispatch_size = 1024; | 361 | constexpr u32 dispatch_size = 1024; |
| 279 | cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld); | 362 | cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline); |
| 280 | cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld); | 363 | cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {}); |
| 281 | cmdbuf.pushConstants(layout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(first), &first, | 364 | cmdbuf.PushConstants(layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(first), &first); |
| 282 | dld); | 365 | cmdbuf.Dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1); |
| 283 | cmdbuf.dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1, dld); | 366 | |
| 284 | 367 | VkBufferMemoryBarrier barrier; | |
| 285 | const vk::BufferMemoryBarrier barrier( | 368 | barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; |
| 286 | vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead, | 369 | barrier.pNext = nullptr; |
| 287 | VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0, | 370 | barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; |
| 288 | static_cast<vk::DeviceSize>(num_quads) * 6 * sizeof(u32)); | 371 | barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT; |
| 289 | cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, | 372 | barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 290 | vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld); | 373 | barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 374 | barrier.buffer = buffer; | ||
| 375 | barrier.offset = 0; | ||
| 376 | barrier.size = static_cast<VkDeviceSize>(num_quads) * 6 * sizeof(u32); | ||
| 377 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, | ||
| 378 | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, {barrier}, {}); | ||
| 291 | }); | 379 | }); |
| 292 | return {*buffer.handle, 0}; | 380 | return {buffer.handle.address(), 0}; |
| 293 | } | 381 | } |
| 294 | 382 | ||
| 295 | Uint8Pass::Uint8Pass(const VKDevice& device, VKScheduler& scheduler, | 383 | Uint8Pass::Uint8Pass(const VKDevice& device, VKScheduler& scheduler, |
| 296 | VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool, | 384 | VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool, |
| 297 | VKUpdateDescriptorQueue& update_descriptor_queue) | 385 | VKUpdateDescriptorQueue& update_descriptor_queue) |
| 298 | : VKComputePass(device, descriptor_pool, | 386 | : VKComputePass(device, descriptor_pool, BuildUint8PassDescriptorSetBindings(), |
| 299 | {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1, | 387 | BuildUint8PassDescriptorUpdateTemplateEntry(), {}, std::size(uint8_pass), |
| 300 | vk::ShaderStageFlagBits::eCompute, nullptr), | 388 | uint8_pass), |
| 301 | vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eStorageBuffer, 1, | ||
| 302 | vk::ShaderStageFlagBits::eCompute, nullptr)}, | ||
| 303 | {vk::DescriptorUpdateTemplateEntry(0, 0, 2, vk::DescriptorType::eStorageBuffer, | ||
| 304 | 0, sizeof(DescriptorUpdateEntry))}, | ||
| 305 | {}, std::size(uint8_pass), uint8_pass), | ||
| 306 | scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool}, | 389 | scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool}, |
| 307 | update_descriptor_queue{update_descriptor_queue} {} | 390 | update_descriptor_queue{update_descriptor_queue} {} |
| 308 | 391 | ||
| 309 | Uint8Pass::~Uint8Pass() = default; | 392 | Uint8Pass::~Uint8Pass() = default; |
| 310 | 393 | ||
| 311 | std::pair<const vk::Buffer*, u64> Uint8Pass::Assemble(u32 num_vertices, vk::Buffer src_buffer, | 394 | std::pair<const VkBuffer*, u64> Uint8Pass::Assemble(u32 num_vertices, VkBuffer src_buffer, |
| 312 | u64 src_offset) { | 395 | u64 src_offset) { |
| 313 | const auto staging_size = static_cast<u32>(num_vertices * sizeof(u16)); | 396 | const auto staging_size = static_cast<u32>(num_vertices * sizeof(u16)); |
| 314 | auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false); | 397 | auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false); |
| 315 | 398 | ||
| 316 | update_descriptor_queue.Acquire(); | 399 | update_descriptor_queue.Acquire(); |
| 317 | update_descriptor_queue.AddBuffer(&src_buffer, src_offset, num_vertices); | 400 | update_descriptor_queue.AddBuffer(&src_buffer, src_offset, num_vertices); |
| 318 | update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size); | 401 | update_descriptor_queue.AddBuffer(buffer.handle.address(), 0, staging_size); |
| 319 | const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence()); | 402 | const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence()); |
| 320 | 403 | ||
| 321 | scheduler.RequestOutsideRenderPassOperationContext(); | 404 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 322 | scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, set, | 405 | scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, set, |
| 323 | num_vertices](auto cmdbuf, auto& dld) { | 406 | num_vertices](vk::CommandBuffer cmdbuf) { |
| 324 | constexpr u32 dispatch_size = 1024; | 407 | constexpr u32 dispatch_size = 1024; |
| 325 | cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld); | 408 | cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline); |
| 326 | cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld); | 409 | cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {}); |
| 327 | cmdbuf.dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1, dld); | 410 | cmdbuf.Dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1); |
| 328 | 411 | ||
| 329 | const vk::BufferMemoryBarrier barrier( | 412 | VkBufferMemoryBarrier barrier; |
| 330 | vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead, | 413 | barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; |
| 331 | VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0, | 414 | barrier.pNext = nullptr; |
| 332 | static_cast<vk::DeviceSize>(num_vertices) * sizeof(u16)); | 415 | barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; |
| 333 | cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, | 416 | barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT; |
| 334 | vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld); | 417 | barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; |
| 418 | barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 419 | barrier.buffer = buffer; | ||
| 420 | barrier.offset = 0; | ||
| 421 | barrier.size = static_cast<VkDeviceSize>(num_vertices * sizeof(u16)); | ||
| 422 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, | ||
| 423 | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, barrier, {}); | ||
| 335 | }); | 424 | }); |
| 336 | return {&*buffer.handle, 0}; | 425 | return {buffer.handle.address(), 0}; |
| 337 | } | 426 | } |
| 338 | 427 | ||
| 339 | } // namespace Vulkan | 428 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.h b/src/video_core/renderer_vulkan/vk_compute_pass.h index 7057eb837..c62516bff 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.h +++ b/src/video_core/renderer_vulkan/vk_compute_pass.h | |||
| @@ -8,8 +8,8 @@ | |||
| 8 | #include <utility> | 8 | #include <utility> |
| 9 | #include <vector> | 9 | #include <vector> |
| 10 | #include "common/common_types.h" | 10 | #include "common/common_types.h" |
| 11 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 12 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 11 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -22,24 +22,24 @@ class VKUpdateDescriptorQueue; | |||
| 22 | class VKComputePass { | 22 | class VKComputePass { |
| 23 | public: | 23 | public: |
| 24 | explicit VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool, | 24 | explicit VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool, |
| 25 | const std::vector<vk::DescriptorSetLayoutBinding>& bindings, | 25 | vk::Span<VkDescriptorSetLayoutBinding> bindings, |
| 26 | const std::vector<vk::DescriptorUpdateTemplateEntry>& templates, | 26 | vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates, |
| 27 | const std::vector<vk::PushConstantRange> push_constants, | 27 | vk::Span<VkPushConstantRange> push_constants, std::size_t code_size, |
| 28 | std::size_t code_size, const u8* code); | 28 | const u8* code); |
| 29 | ~VKComputePass(); | 29 | ~VKComputePass(); |
| 30 | 30 | ||
| 31 | protected: | 31 | protected: |
| 32 | vk::DescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue, | 32 | VkDescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue, |
| 33 | VKFence& fence); | 33 | VKFence& fence); |
| 34 | 34 | ||
| 35 | UniqueDescriptorUpdateTemplate descriptor_template; | 35 | vk::DescriptorUpdateTemplateKHR descriptor_template; |
| 36 | UniquePipelineLayout layout; | 36 | vk::PipelineLayout layout; |
| 37 | UniquePipeline pipeline; | 37 | vk::Pipeline pipeline; |
| 38 | 38 | ||
| 39 | private: | 39 | private: |
| 40 | UniqueDescriptorSetLayout descriptor_set_layout; | 40 | vk::DescriptorSetLayout descriptor_set_layout; |
| 41 | std::optional<DescriptorAllocator> descriptor_allocator; | 41 | std::optional<DescriptorAllocator> descriptor_allocator; |
| 42 | UniqueShaderModule module; | 42 | vk::ShaderModule module; |
| 43 | }; | 43 | }; |
| 44 | 44 | ||
| 45 | class QuadArrayPass final : public VKComputePass { | 45 | class QuadArrayPass final : public VKComputePass { |
| @@ -50,7 +50,7 @@ public: | |||
| 50 | VKUpdateDescriptorQueue& update_descriptor_queue); | 50 | VKUpdateDescriptorQueue& update_descriptor_queue); |
| 51 | ~QuadArrayPass(); | 51 | ~QuadArrayPass(); |
| 52 | 52 | ||
| 53 | std::pair<const vk::Buffer&, vk::DeviceSize> Assemble(u32 num_vertices, u32 first); | 53 | std::pair<const VkBuffer*, VkDeviceSize> Assemble(u32 num_vertices, u32 first); |
| 54 | 54 | ||
| 55 | private: | 55 | private: |
| 56 | VKScheduler& scheduler; | 56 | VKScheduler& scheduler; |
| @@ -65,8 +65,7 @@ public: | |||
| 65 | VKUpdateDescriptorQueue& update_descriptor_queue); | 65 | VKUpdateDescriptorQueue& update_descriptor_queue); |
| 66 | ~Uint8Pass(); | 66 | ~Uint8Pass(); |
| 67 | 67 | ||
| 68 | std::pair<const vk::Buffer*, u64> Assemble(u32 num_vertices, vk::Buffer src_buffer, | 68 | std::pair<const VkBuffer*, u64> Assemble(u32 num_vertices, VkBuffer src_buffer, u64 src_offset); |
| 69 | u64 src_offset); | ||
| 70 | 69 | ||
| 71 | private: | 70 | private: |
| 72 | VKScheduler& scheduler; | 71 | VKScheduler& scheduler; |
diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp index 60f57d83e..23beafa4f 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp | |||
| @@ -5,7 +5,6 @@ | |||
| 5 | #include <memory> | 5 | #include <memory> |
| 6 | #include <vector> | 6 | #include <vector> |
| 7 | 7 | ||
| 8 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 9 | #include "video_core/renderer_vulkan/vk_compute_pipeline.h" | 8 | #include "video_core/renderer_vulkan/vk_compute_pipeline.h" |
| 10 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 9 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| 11 | #include "video_core/renderer_vulkan/vk_device.h" | 10 | #include "video_core/renderer_vulkan/vk_device.h" |
| @@ -14,6 +13,7 @@ | |||
| 14 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 13 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 15 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" | 14 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" |
| 16 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 15 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 16 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 17 | 17 | ||
| 18 | namespace Vulkan { | 18 | namespace Vulkan { |
| 19 | 19 | ||
| @@ -30,7 +30,7 @@ VKComputePipeline::VKComputePipeline(const VKDevice& device, VKScheduler& schedu | |||
| 30 | 30 | ||
| 31 | VKComputePipeline::~VKComputePipeline() = default; | 31 | VKComputePipeline::~VKComputePipeline() = default; |
| 32 | 32 | ||
| 33 | vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() { | 33 | VkDescriptorSet VKComputePipeline::CommitDescriptorSet() { |
| 34 | if (!descriptor_template) { | 34 | if (!descriptor_template) { |
| 35 | return {}; | 35 | return {}; |
| 36 | } | 36 | } |
| @@ -39,74 +39,109 @@ vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() { | |||
| 39 | return set; | 39 | return set; |
| 40 | } | 40 | } |
| 41 | 41 | ||
| 42 | UniqueDescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const { | 42 | vk::DescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const { |
| 43 | std::vector<vk::DescriptorSetLayoutBinding> bindings; | 43 | std::vector<VkDescriptorSetLayoutBinding> bindings; |
| 44 | u32 binding = 0; | 44 | u32 binding = 0; |
| 45 | const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) { | 45 | const auto add_bindings = [&](VkDescriptorType descriptor_type, std::size_t num_entries) { |
| 46 | // TODO(Rodrigo): Maybe make individual bindings here? | 46 | // TODO(Rodrigo): Maybe make individual bindings here? |
| 47 | for (u32 bindpoint = 0; bindpoint < static_cast<u32>(num_entries); ++bindpoint) { | 47 | for (u32 bindpoint = 0; bindpoint < static_cast<u32>(num_entries); ++bindpoint) { |
| 48 | bindings.emplace_back(binding++, descriptor_type, 1, vk::ShaderStageFlagBits::eCompute, | 48 | VkDescriptorSetLayoutBinding& entry = bindings.emplace_back(); |
| 49 | nullptr); | 49 | entry.binding = binding++; |
| 50 | entry.descriptorType = descriptor_type; | ||
| 51 | entry.descriptorCount = 1; | ||
| 52 | entry.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 53 | entry.pImmutableSamplers = nullptr; | ||
| 50 | } | 54 | } |
| 51 | }; | 55 | }; |
| 52 | AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size()); | 56 | add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, entries.const_buffers.size()); |
| 53 | AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size()); | 57 | add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, entries.global_buffers.size()); |
| 54 | AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size()); | 58 | add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, entries.texel_buffers.size()); |
| 55 | AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size()); | 59 | add_bindings(VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, entries.samplers.size()); |
| 56 | AddBindings(vk::DescriptorType::eStorageImage, entries.images.size()); | 60 | add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, entries.images.size()); |
| 57 | 61 | ||
| 58 | const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci( | 62 | VkDescriptorSetLayoutCreateInfo ci; |
| 59 | {}, static_cast<u32>(bindings.size()), bindings.data()); | 63 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; |
| 60 | 64 | ci.pNext = nullptr; | |
| 61 | const auto dev = device.GetLogical(); | 65 | ci.flags = 0; |
| 62 | const auto& dld = device.GetDispatchLoader(); | 66 | ci.bindingCount = static_cast<u32>(bindings.size()); |
| 63 | return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld); | 67 | ci.pBindings = bindings.data(); |
| 68 | return device.GetLogical().CreateDescriptorSetLayout(ci); | ||
| 64 | } | 69 | } |
| 65 | 70 | ||
| 66 | UniquePipelineLayout VKComputePipeline::CreatePipelineLayout() const { | 71 | vk::PipelineLayout VKComputePipeline::CreatePipelineLayout() const { |
| 67 | const vk::PipelineLayoutCreateInfo layout_ci({}, 1, &*descriptor_set_layout, 0, nullptr); | 72 | VkPipelineLayoutCreateInfo ci; |
| 68 | const auto dev = device.GetLogical(); | 73 | ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; |
| 69 | return dev.createPipelineLayoutUnique(layout_ci, nullptr, device.GetDispatchLoader()); | 74 | ci.pNext = nullptr; |
| 75 | ci.flags = 0; | ||
| 76 | ci.setLayoutCount = 1; | ||
| 77 | ci.pSetLayouts = descriptor_set_layout.address(); | ||
| 78 | ci.pushConstantRangeCount = 0; | ||
| 79 | ci.pPushConstantRanges = nullptr; | ||
| 80 | return device.GetLogical().CreatePipelineLayout(ci); | ||
| 70 | } | 81 | } |
| 71 | 82 | ||
| 72 | UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate() const { | 83 | vk::DescriptorUpdateTemplateKHR VKComputePipeline::CreateDescriptorUpdateTemplate() const { |
| 73 | std::vector<vk::DescriptorUpdateTemplateEntry> template_entries; | 84 | std::vector<VkDescriptorUpdateTemplateEntryKHR> template_entries; |
| 74 | u32 binding = 0; | 85 | u32 binding = 0; |
| 75 | u32 offset = 0; | 86 | u32 offset = 0; |
| 76 | FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries); | 87 | FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries); |
| 77 | if (template_entries.empty()) { | 88 | if (template_entries.empty()) { |
| 78 | // If the shader doesn't use descriptor sets, skip template creation. | 89 | // If the shader doesn't use descriptor sets, skip template creation. |
| 79 | return UniqueDescriptorUpdateTemplate{}; | 90 | return {}; |
| 80 | } | 91 | } |
| 81 | 92 | ||
| 82 | const vk::DescriptorUpdateTemplateCreateInfo template_ci( | 93 | VkDescriptorUpdateTemplateCreateInfoKHR ci; |
| 83 | {}, static_cast<u32>(template_entries.size()), template_entries.data(), | 94 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR; |
| 84 | vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout, | 95 | ci.pNext = nullptr; |
| 85 | vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET); | 96 | ci.flags = 0; |
| 86 | 97 | ci.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size()); | |
| 87 | const auto dev = device.GetLogical(); | 98 | ci.pDescriptorUpdateEntries = template_entries.data(); |
| 88 | const auto& dld = device.GetDispatchLoader(); | 99 | ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR; |
| 89 | return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld); | 100 | ci.descriptorSetLayout = *descriptor_set_layout; |
| 101 | ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; | ||
| 102 | ci.pipelineLayout = *layout; | ||
| 103 | ci.set = DESCRIPTOR_SET; | ||
| 104 | return device.GetLogical().CreateDescriptorUpdateTemplateKHR(ci); | ||
| 90 | } | 105 | } |
| 91 | 106 | ||
| 92 | UniqueShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const { | 107 | vk::ShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const { |
| 93 | const vk::ShaderModuleCreateInfo module_ci({}, code.size() * sizeof(u32), code.data()); | 108 | VkShaderModuleCreateInfo ci; |
| 94 | const auto dev = device.GetLogical(); | 109 | ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; |
| 95 | return dev.createShaderModuleUnique(module_ci, nullptr, device.GetDispatchLoader()); | 110 | ci.pNext = nullptr; |
| 111 | ci.flags = 0; | ||
| 112 | ci.codeSize = code.size() * sizeof(u32); | ||
| 113 | ci.pCode = code.data(); | ||
| 114 | return device.GetLogical().CreateShaderModule(ci); | ||
| 96 | } | 115 | } |
| 97 | 116 | ||
| 98 | UniquePipeline VKComputePipeline::CreatePipeline() const { | 117 | vk::Pipeline VKComputePipeline::CreatePipeline() const { |
| 99 | vk::PipelineShaderStageCreateInfo shader_stage_ci({}, vk::ShaderStageFlagBits::eCompute, | 118 | VkComputePipelineCreateInfo ci; |
| 100 | *shader_module, "main", nullptr); | 119 | VkPipelineShaderStageCreateInfo& stage_ci = ci.stage; |
| 101 | vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; | 120 | stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; |
| 121 | stage_ci.pNext = nullptr; | ||
| 122 | stage_ci.flags = 0; | ||
| 123 | stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT; | ||
| 124 | stage_ci.module = *shader_module; | ||
| 125 | stage_ci.pName = "main"; | ||
| 126 | stage_ci.pSpecializationInfo = nullptr; | ||
| 127 | |||
| 128 | VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; | ||
| 129 | subgroup_size_ci.sType = | ||
| 130 | VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT; | ||
| 131 | subgroup_size_ci.pNext = nullptr; | ||
| 102 | subgroup_size_ci.requiredSubgroupSize = GuestWarpSize; | 132 | subgroup_size_ci.requiredSubgroupSize = GuestWarpSize; |
| 103 | if (entries.uses_warps && device.IsGuestWarpSizeSupported(vk::ShaderStageFlagBits::eCompute)) { | 133 | |
| 104 | shader_stage_ci.pNext = &subgroup_size_ci; | 134 | if (entries.uses_warps && device.IsGuestWarpSizeSupported(VK_SHADER_STAGE_COMPUTE_BIT)) { |
| 135 | stage_ci.pNext = &subgroup_size_ci; | ||
| 105 | } | 136 | } |
| 106 | 137 | ||
| 107 | const vk::ComputePipelineCreateInfo create_info({}, shader_stage_ci, *layout, {}, 0); | 138 | ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO; |
| 108 | const auto dev = device.GetLogical(); | 139 | ci.pNext = nullptr; |
| 109 | return dev.createComputePipelineUnique({}, create_info, nullptr, device.GetDispatchLoader()); | 140 | ci.flags = 0; |
| 141 | ci.layout = *layout; | ||
| 142 | ci.basePipelineHandle = nullptr; | ||
| 143 | ci.basePipelineIndex = 0; | ||
| 144 | return device.GetLogical().CreateComputePipeline(ci); | ||
| 110 | } | 145 | } |
| 111 | 146 | ||
| 112 | } // namespace Vulkan | 147 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.h b/src/video_core/renderer_vulkan/vk_compute_pipeline.h index 22235c6c9..33b9af29e 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.h | |||
| @@ -7,9 +7,9 @@ | |||
| 7 | #include <memory> | 7 | #include <memory> |
| 8 | 8 | ||
| 9 | #include "common/common_types.h" | 9 | #include "common/common_types.h" |
| 10 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 11 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 10 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| 12 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" | 11 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -25,42 +25,42 @@ public: | |||
| 25 | const SPIRVShader& shader); | 25 | const SPIRVShader& shader); |
| 26 | ~VKComputePipeline(); | 26 | ~VKComputePipeline(); |
| 27 | 27 | ||
| 28 | vk::DescriptorSet CommitDescriptorSet(); | 28 | VkDescriptorSet CommitDescriptorSet(); |
| 29 | 29 | ||
| 30 | vk::Pipeline GetHandle() const { | 30 | VkPipeline GetHandle() const { |
| 31 | return *pipeline; | 31 | return *pipeline; |
| 32 | } | 32 | } |
| 33 | 33 | ||
| 34 | vk::PipelineLayout GetLayout() const { | 34 | VkPipelineLayout GetLayout() const { |
| 35 | return *layout; | 35 | return *layout; |
| 36 | } | 36 | } |
| 37 | 37 | ||
| 38 | const ShaderEntries& GetEntries() { | 38 | const ShaderEntries& GetEntries() const { |
| 39 | return entries; | 39 | return entries; |
| 40 | } | 40 | } |
| 41 | 41 | ||
| 42 | private: | 42 | private: |
| 43 | UniqueDescriptorSetLayout CreateDescriptorSetLayout() const; | 43 | vk::DescriptorSetLayout CreateDescriptorSetLayout() const; |
| 44 | 44 | ||
| 45 | UniquePipelineLayout CreatePipelineLayout() const; | 45 | vk::PipelineLayout CreatePipelineLayout() const; |
| 46 | 46 | ||
| 47 | UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate() const; | 47 | vk::DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplate() const; |
| 48 | 48 | ||
| 49 | UniqueShaderModule CreateShaderModule(const std::vector<u32>& code) const; | 49 | vk::ShaderModule CreateShaderModule(const std::vector<u32>& code) const; |
| 50 | 50 | ||
| 51 | UniquePipeline CreatePipeline() const; | 51 | vk::Pipeline CreatePipeline() const; |
| 52 | 52 | ||
| 53 | const VKDevice& device; | 53 | const VKDevice& device; |
| 54 | VKScheduler& scheduler; | 54 | VKScheduler& scheduler; |
| 55 | ShaderEntries entries; | 55 | ShaderEntries entries; |
| 56 | 56 | ||
| 57 | UniqueDescriptorSetLayout descriptor_set_layout; | 57 | vk::DescriptorSetLayout descriptor_set_layout; |
| 58 | DescriptorAllocator descriptor_allocator; | 58 | DescriptorAllocator descriptor_allocator; |
| 59 | VKUpdateDescriptorQueue& update_descriptor_queue; | 59 | VKUpdateDescriptorQueue& update_descriptor_queue; |
| 60 | UniquePipelineLayout layout; | 60 | vk::PipelineLayout layout; |
| 61 | UniqueDescriptorUpdateTemplate descriptor_template; | 61 | vk::DescriptorUpdateTemplateKHR descriptor_template; |
| 62 | UniqueShaderModule shader_module; | 62 | vk::ShaderModule shader_module; |
| 63 | UniquePipeline pipeline; | 63 | vk::Pipeline pipeline; |
| 64 | }; | 64 | }; |
| 65 | 65 | ||
| 66 | } // namespace Vulkan | 66 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp index cc7c281a0..e9d528aa6 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp | |||
| @@ -6,10 +6,10 @@ | |||
| 6 | #include <vector> | 6 | #include <vector> |
| 7 | 7 | ||
| 8 | #include "common/common_types.h" | 8 | #include "common/common_types.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 9 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| 11 | #include "video_core/renderer_vulkan/vk_device.h" | 10 | #include "video_core/renderer_vulkan/vk_device.h" |
| 12 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 11 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -17,19 +17,18 @@ namespace Vulkan { | |||
| 17 | constexpr std::size_t SETS_GROW_RATE = 0x20; | 17 | constexpr std::size_t SETS_GROW_RATE = 0x20; |
| 18 | 18 | ||
| 19 | DescriptorAllocator::DescriptorAllocator(VKDescriptorPool& descriptor_pool, | 19 | DescriptorAllocator::DescriptorAllocator(VKDescriptorPool& descriptor_pool, |
| 20 | vk::DescriptorSetLayout layout) | 20 | VkDescriptorSetLayout layout) |
| 21 | : VKFencedPool{SETS_GROW_RATE}, descriptor_pool{descriptor_pool}, layout{layout} {} | 21 | : VKFencedPool{SETS_GROW_RATE}, descriptor_pool{descriptor_pool}, layout{layout} {} |
| 22 | 22 | ||
| 23 | DescriptorAllocator::~DescriptorAllocator() = default; | 23 | DescriptorAllocator::~DescriptorAllocator() = default; |
| 24 | 24 | ||
| 25 | vk::DescriptorSet DescriptorAllocator::Commit(VKFence& fence) { | 25 | VkDescriptorSet DescriptorAllocator::Commit(VKFence& fence) { |
| 26 | return *descriptors[CommitResource(fence)]; | 26 | const std::size_t index = CommitResource(fence); |
| 27 | return descriptors_allocations[index / SETS_GROW_RATE][index % SETS_GROW_RATE]; | ||
| 27 | } | 28 | } |
| 28 | 29 | ||
| 29 | void DescriptorAllocator::Allocate(std::size_t begin, std::size_t end) { | 30 | void DescriptorAllocator::Allocate(std::size_t begin, std::size_t end) { |
| 30 | auto new_sets = descriptor_pool.AllocateDescriptors(layout, end - begin); | 31 | descriptors_allocations.push_back(descriptor_pool.AllocateDescriptors(layout, end - begin)); |
| 31 | descriptors.insert(descriptors.end(), std::make_move_iterator(new_sets.begin()), | ||
| 32 | std::make_move_iterator(new_sets.end())); | ||
| 33 | } | 32 | } |
| 34 | 33 | ||
| 35 | VKDescriptorPool::VKDescriptorPool(const VKDevice& device) | 34 | VKDescriptorPool::VKDescriptorPool(const VKDevice& device) |
| @@ -37,53 +36,50 @@ VKDescriptorPool::VKDescriptorPool(const VKDevice& device) | |||
| 37 | 36 | ||
| 38 | VKDescriptorPool::~VKDescriptorPool() = default; | 37 | VKDescriptorPool::~VKDescriptorPool() = default; |
| 39 | 38 | ||
| 40 | vk::DescriptorPool VKDescriptorPool::AllocateNewPool() { | 39 | vk::DescriptorPool* VKDescriptorPool::AllocateNewPool() { |
| 41 | static constexpr u32 num_sets = 0x20000; | 40 | static constexpr u32 num_sets = 0x20000; |
| 42 | static constexpr vk::DescriptorPoolSize pool_sizes[] = { | 41 | static constexpr VkDescriptorPoolSize pool_sizes[] = { |
| 43 | {vk::DescriptorType::eUniformBuffer, num_sets * 90}, | 42 | {VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, num_sets * 90}, |
| 44 | {vk::DescriptorType::eStorageBuffer, num_sets * 60}, | 43 | {VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, num_sets * 60}, |
| 45 | {vk::DescriptorType::eUniformTexelBuffer, num_sets * 64}, | 44 | {VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, num_sets * 64}, |
| 46 | {vk::DescriptorType::eCombinedImageSampler, num_sets * 64}, | 45 | {VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, num_sets * 64}, |
| 47 | {vk::DescriptorType::eStorageImage, num_sets * 40}}; | 46 | {VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, num_sets * 40}}; |
| 48 | 47 | ||
| 49 | const vk::DescriptorPoolCreateInfo create_info( | 48 | VkDescriptorPoolCreateInfo ci; |
| 50 | vk::DescriptorPoolCreateFlagBits::eFreeDescriptorSet, num_sets, | 49 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO; |
| 51 | static_cast<u32>(std::size(pool_sizes)), std::data(pool_sizes)); | 50 | ci.pNext = nullptr; |
| 52 | const auto dev = device.GetLogical(); | 51 | ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; |
| 53 | return *pools.emplace_back( | 52 | ci.maxSets = num_sets; |
| 54 | dev.createDescriptorPoolUnique(create_info, nullptr, device.GetDispatchLoader())); | 53 | ci.poolSizeCount = static_cast<u32>(std::size(pool_sizes)); |
| 54 | ci.pPoolSizes = std::data(pool_sizes); | ||
| 55 | return &pools.emplace_back(device.GetLogical().CreateDescriptorPool(ci)); | ||
| 55 | } | 56 | } |
| 56 | 57 | ||
| 57 | std::vector<UniqueDescriptorSet> VKDescriptorPool::AllocateDescriptors( | 58 | vk::DescriptorSets VKDescriptorPool::AllocateDescriptors(VkDescriptorSetLayout layout, |
| 58 | vk::DescriptorSetLayout layout, std::size_t count) { | 59 | std::size_t count) { |
| 59 | std::vector layout_copies(count, layout); | 60 | const std::vector layout_copies(count, layout); |
| 60 | vk::DescriptorSetAllocateInfo allocate_info(active_pool, static_cast<u32>(count), | 61 | VkDescriptorSetAllocateInfo ai; |
| 61 | layout_copies.data()); | 62 | ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO; |
| 62 | 63 | ai.pNext = nullptr; | |
| 63 | std::vector<vk::DescriptorSet> sets(count); | 64 | ai.descriptorPool = **active_pool; |
| 64 | const auto dev = device.GetLogical(); | 65 | ai.descriptorSetCount = static_cast<u32>(count); |
| 65 | const auto& dld = device.GetDispatchLoader(); | 66 | ai.pSetLayouts = layout_copies.data(); |
| 66 | switch (const auto result = dev.allocateDescriptorSets(&allocate_info, sets.data(), dld)) { | 67 | |
| 67 | case vk::Result::eSuccess: | 68 | vk::DescriptorSets sets = active_pool->Allocate(ai); |
| 68 | break; | 69 | if (!sets.IsOutOfPoolMemory()) { |
| 69 | case vk::Result::eErrorOutOfPoolMemory: | 70 | return sets; |
| 70 | active_pool = AllocateNewPool(); | ||
| 71 | allocate_info.descriptorPool = active_pool; | ||
| 72 | if (dev.allocateDescriptorSets(&allocate_info, sets.data(), dld) == vk::Result::eSuccess) { | ||
| 73 | break; | ||
| 74 | } | ||
| 75 | [[fallthrough]]; | ||
| 76 | default: | ||
| 77 | vk::throwResultException(result, "vk::Device::allocateDescriptorSetsUnique"); | ||
| 78 | } | 71 | } |
| 79 | 72 | ||
| 80 | vk::PoolFree deleter(dev, active_pool, dld); | 73 | // Our current pool is out of memory. Allocate a new one and retry |
| 81 | std::vector<UniqueDescriptorSet> unique_sets; | 74 | active_pool = AllocateNewPool(); |
| 82 | unique_sets.reserve(count); | 75 | ai.descriptorPool = **active_pool; |
| 83 | for (const auto set : sets) { | 76 | sets = active_pool->Allocate(ai); |
| 84 | unique_sets.push_back(UniqueDescriptorSet{set, deleter}); | 77 | if (!sets.IsOutOfPoolMemory()) { |
| 78 | return sets; | ||
| 85 | } | 79 | } |
| 86 | return unique_sets; | 80 | |
| 81 | // After allocating a new pool, we are out of memory again. We can't handle this from here. | ||
| 82 | throw vk::Exception(VK_ERROR_OUT_OF_POOL_MEMORY); | ||
| 87 | } | 83 | } |
| 88 | 84 | ||
| 89 | } // namespace Vulkan | 85 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.h b/src/video_core/renderer_vulkan/vk_descriptor_pool.h index a441dbc0f..ab40c70f0 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.h +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.h | |||
| @@ -8,8 +8,8 @@ | |||
| 8 | #include <vector> | 8 | #include <vector> |
| 9 | 9 | ||
| 10 | #include "common/common_types.h" | 10 | #include "common/common_types.h" |
| 11 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 12 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 11 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -17,21 +17,21 @@ class VKDescriptorPool; | |||
| 17 | 17 | ||
| 18 | class DescriptorAllocator final : public VKFencedPool { | 18 | class DescriptorAllocator final : public VKFencedPool { |
| 19 | public: | 19 | public: |
| 20 | explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, vk::DescriptorSetLayout layout); | 20 | explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, VkDescriptorSetLayout layout); |
| 21 | ~DescriptorAllocator() override; | 21 | ~DescriptorAllocator() override; |
| 22 | 22 | ||
| 23 | DescriptorAllocator(const DescriptorAllocator&) = delete; | 23 | DescriptorAllocator(const DescriptorAllocator&) = delete; |
| 24 | 24 | ||
| 25 | vk::DescriptorSet Commit(VKFence& fence); | 25 | VkDescriptorSet Commit(VKFence& fence); |
| 26 | 26 | ||
| 27 | protected: | 27 | protected: |
| 28 | void Allocate(std::size_t begin, std::size_t end) override; | 28 | void Allocate(std::size_t begin, std::size_t end) override; |
| 29 | 29 | ||
| 30 | private: | 30 | private: |
| 31 | VKDescriptorPool& descriptor_pool; | 31 | VKDescriptorPool& descriptor_pool; |
| 32 | const vk::DescriptorSetLayout layout; | 32 | const VkDescriptorSetLayout layout; |
| 33 | 33 | ||
| 34 | std::vector<UniqueDescriptorSet> descriptors; | 34 | std::vector<vk::DescriptorSets> descriptors_allocations; |
| 35 | }; | 35 | }; |
| 36 | 36 | ||
| 37 | class VKDescriptorPool final { | 37 | class VKDescriptorPool final { |
| @@ -42,15 +42,14 @@ public: | |||
| 42 | ~VKDescriptorPool(); | 42 | ~VKDescriptorPool(); |
| 43 | 43 | ||
| 44 | private: | 44 | private: |
| 45 | vk::DescriptorPool AllocateNewPool(); | 45 | vk::DescriptorPool* AllocateNewPool(); |
| 46 | 46 | ||
| 47 | std::vector<UniqueDescriptorSet> AllocateDescriptors(vk::DescriptorSetLayout layout, | 47 | vk::DescriptorSets AllocateDescriptors(VkDescriptorSetLayout layout, std::size_t count); |
| 48 | std::size_t count); | ||
| 49 | 48 | ||
| 50 | const VKDevice& device; | 49 | const VKDevice& device; |
| 51 | 50 | ||
| 52 | std::vector<UniqueDescriptorPool> pools; | 51 | std::vector<vk::DescriptorPool> pools; |
| 53 | vk::DescriptorPool active_pool; | 52 | vk::DescriptorPool* active_pool; |
| 54 | }; | 53 | }; |
| 55 | 54 | ||
| 56 | } // namespace Vulkan \ No newline at end of file | 55 | } // namespace Vulkan \ No newline at end of file |
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index 6f4ae9132..52d29e49d 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp | |||
| @@ -6,15 +6,15 @@ | |||
| 6 | #include <chrono> | 6 | #include <chrono> |
| 7 | #include <cstdlib> | 7 | #include <cstdlib> |
| 8 | #include <optional> | 8 | #include <optional> |
| 9 | #include <set> | ||
| 10 | #include <string_view> | 9 | #include <string_view> |
| 11 | #include <thread> | 10 | #include <thread> |
| 11 | #include <unordered_set> | ||
| 12 | #include <vector> | 12 | #include <vector> |
| 13 | 13 | ||
| 14 | #include "common/assert.h" | 14 | #include "common/assert.h" |
| 15 | #include "core/settings.h" | 15 | #include "core/settings.h" |
| 16 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 17 | #include "video_core/renderer_vulkan/vk_device.h" | 16 | #include "video_core/renderer_vulkan/vk_device.h" |
| 17 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 18 | 18 | ||
| 19 | namespace Vulkan { | 19 | namespace Vulkan { |
| 20 | 20 | ||
| @@ -22,49 +22,43 @@ namespace { | |||
| 22 | 22 | ||
| 23 | namespace Alternatives { | 23 | namespace Alternatives { |
| 24 | 24 | ||
| 25 | constexpr std::array Depth24UnormS8Uint = {vk::Format::eD32SfloatS8Uint, | 25 | constexpr std::array Depth24UnormS8_UINT = {VK_FORMAT_D32_SFLOAT_S8_UINT, |
| 26 | vk::Format::eD16UnormS8Uint, vk::Format{}}; | 26 | VK_FORMAT_D16_UNORM_S8_UINT, VkFormat{}}; |
| 27 | constexpr std::array Depth16UnormS8Uint = {vk::Format::eD24UnormS8Uint, | 27 | constexpr std::array Depth16UnormS8_UINT = {VK_FORMAT_D24_UNORM_S8_UINT, |
| 28 | vk::Format::eD32SfloatS8Uint, vk::Format{}}; | 28 | VK_FORMAT_D32_SFLOAT_S8_UINT, VkFormat{}}; |
| 29 | 29 | ||
| 30 | } // namespace Alternatives | 30 | } // namespace Alternatives |
| 31 | 31 | ||
| 32 | constexpr std::array REQUIRED_EXTENSIONS = { | ||
| 33 | VK_KHR_SWAPCHAIN_EXTENSION_NAME, | ||
| 34 | VK_KHR_16BIT_STORAGE_EXTENSION_NAME, | ||
| 35 | VK_KHR_8BIT_STORAGE_EXTENSION_NAME, | ||
| 36 | VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME, | ||
| 37 | VK_KHR_DESCRIPTOR_UPDATE_TEMPLATE_EXTENSION_NAME, | ||
| 38 | VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME, | ||
| 39 | VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME, | ||
| 40 | VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME, | ||
| 41 | VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME, | ||
| 42 | }; | ||
| 43 | |||
| 32 | template <typename T> | 44 | template <typename T> |
| 33 | void SetNext(void**& next, T& data) { | 45 | void SetNext(void**& next, T& data) { |
| 34 | *next = &data; | 46 | *next = &data; |
| 35 | next = &data.pNext; | 47 | next = &data.pNext; |
| 36 | } | 48 | } |
| 37 | 49 | ||
| 38 | template <typename T> | 50 | constexpr const VkFormat* GetFormatAlternatives(VkFormat format) { |
| 39 | T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dld) { | ||
| 40 | vk::PhysicalDeviceFeatures2 features; | ||
| 41 | T extension_features; | ||
| 42 | features.pNext = &extension_features; | ||
| 43 | physical.getFeatures2(&features, dld); | ||
| 44 | return extension_features; | ||
| 45 | } | ||
| 46 | |||
| 47 | template <typename T> | ||
| 48 | T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dld) { | ||
| 49 | vk::PhysicalDeviceProperties2 properties; | ||
| 50 | T extension_properties; | ||
| 51 | properties.pNext = &extension_properties; | ||
| 52 | physical.getProperties2(&properties, dld); | ||
| 53 | return extension_properties; | ||
| 54 | } | ||
| 55 | |||
| 56 | constexpr const vk::Format* GetFormatAlternatives(vk::Format format) { | ||
| 57 | switch (format) { | 51 | switch (format) { |
| 58 | case vk::Format::eD24UnormS8Uint: | 52 | case VK_FORMAT_D24_UNORM_S8_UINT: |
| 59 | return Alternatives::Depth24UnormS8Uint.data(); | 53 | return Alternatives::Depth24UnormS8_UINT.data(); |
| 60 | case vk::Format::eD16UnormS8Uint: | 54 | case VK_FORMAT_D16_UNORM_S8_UINT: |
| 61 | return Alternatives::Depth16UnormS8Uint.data(); | 55 | return Alternatives::Depth16UnormS8_UINT.data(); |
| 62 | default: | 56 | default: |
| 63 | return nullptr; | 57 | return nullptr; |
| 64 | } | 58 | } |
| 65 | } | 59 | } |
| 66 | 60 | ||
| 67 | vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, FormatType format_type) { | 61 | VkFormatFeatureFlags GetFormatFeatures(VkFormatProperties properties, FormatType format_type) { |
| 68 | switch (format_type) { | 62 | switch (format_type) { |
| 69 | case FormatType::Linear: | 63 | case FormatType::Linear: |
| 70 | return properties.linearTilingFeatures; | 64 | return properties.linearTilingFeatures; |
| @@ -77,79 +71,220 @@ vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, Format | |||
| 77 | } | 71 | } |
| 78 | } | 72 | } |
| 79 | 73 | ||
| 74 | std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties( | ||
| 75 | vk::PhysicalDevice physical, const vk::InstanceDispatch& dld) { | ||
| 76 | static constexpr std::array formats{VK_FORMAT_A8B8G8R8_UNORM_PACK32, | ||
| 77 | VK_FORMAT_A8B8G8R8_UINT_PACK32, | ||
| 78 | VK_FORMAT_A8B8G8R8_SNORM_PACK32, | ||
| 79 | VK_FORMAT_A8B8G8R8_SRGB_PACK32, | ||
| 80 | VK_FORMAT_B5G6R5_UNORM_PACK16, | ||
| 81 | VK_FORMAT_A2B10G10R10_UNORM_PACK32, | ||
| 82 | VK_FORMAT_A1R5G5B5_UNORM_PACK16, | ||
| 83 | VK_FORMAT_R32G32B32A32_SFLOAT, | ||
| 84 | VK_FORMAT_R32G32B32A32_UINT, | ||
| 85 | VK_FORMAT_R32G32_SFLOAT, | ||
| 86 | VK_FORMAT_R32G32_UINT, | ||
| 87 | VK_FORMAT_R16G16B16A16_UINT, | ||
| 88 | VK_FORMAT_R16G16B16A16_SNORM, | ||
| 89 | VK_FORMAT_R16G16B16A16_UNORM, | ||
| 90 | VK_FORMAT_R16G16_UNORM, | ||
| 91 | VK_FORMAT_R16G16_SNORM, | ||
| 92 | VK_FORMAT_R16G16_SFLOAT, | ||
| 93 | VK_FORMAT_R16_UNORM, | ||
| 94 | VK_FORMAT_R8G8B8A8_SRGB, | ||
| 95 | VK_FORMAT_R8G8_UNORM, | ||
| 96 | VK_FORMAT_R8G8_SNORM, | ||
| 97 | VK_FORMAT_R8_UNORM, | ||
| 98 | VK_FORMAT_R8_UINT, | ||
| 99 | VK_FORMAT_B10G11R11_UFLOAT_PACK32, | ||
| 100 | VK_FORMAT_R32_SFLOAT, | ||
| 101 | VK_FORMAT_R32_UINT, | ||
| 102 | VK_FORMAT_R32_SINT, | ||
| 103 | VK_FORMAT_R16_SFLOAT, | ||
| 104 | VK_FORMAT_R16G16B16A16_SFLOAT, | ||
| 105 | VK_FORMAT_B8G8R8A8_UNORM, | ||
| 106 | VK_FORMAT_R4G4B4A4_UNORM_PACK16, | ||
| 107 | VK_FORMAT_D32_SFLOAT, | ||
| 108 | VK_FORMAT_D16_UNORM, | ||
| 109 | VK_FORMAT_D16_UNORM_S8_UINT, | ||
| 110 | VK_FORMAT_D24_UNORM_S8_UINT, | ||
| 111 | VK_FORMAT_D32_SFLOAT_S8_UINT, | ||
| 112 | VK_FORMAT_BC1_RGBA_UNORM_BLOCK, | ||
| 113 | VK_FORMAT_BC2_UNORM_BLOCK, | ||
| 114 | VK_FORMAT_BC3_UNORM_BLOCK, | ||
| 115 | VK_FORMAT_BC4_UNORM_BLOCK, | ||
| 116 | VK_FORMAT_BC5_UNORM_BLOCK, | ||
| 117 | VK_FORMAT_BC5_SNORM_BLOCK, | ||
| 118 | VK_FORMAT_BC7_UNORM_BLOCK, | ||
| 119 | VK_FORMAT_BC6H_UFLOAT_BLOCK, | ||
| 120 | VK_FORMAT_BC6H_SFLOAT_BLOCK, | ||
| 121 | VK_FORMAT_BC1_RGBA_SRGB_BLOCK, | ||
| 122 | VK_FORMAT_BC2_SRGB_BLOCK, | ||
| 123 | VK_FORMAT_BC3_SRGB_BLOCK, | ||
| 124 | VK_FORMAT_BC7_SRGB_BLOCK, | ||
| 125 | VK_FORMAT_ASTC_4x4_SRGB_BLOCK, | ||
| 126 | VK_FORMAT_ASTC_8x8_SRGB_BLOCK, | ||
| 127 | VK_FORMAT_ASTC_8x5_SRGB_BLOCK, | ||
| 128 | VK_FORMAT_ASTC_5x4_SRGB_BLOCK, | ||
| 129 | VK_FORMAT_ASTC_5x5_UNORM_BLOCK, | ||
| 130 | VK_FORMAT_ASTC_5x5_SRGB_BLOCK, | ||
| 131 | VK_FORMAT_ASTC_10x8_UNORM_BLOCK, | ||
| 132 | VK_FORMAT_ASTC_10x8_SRGB_BLOCK, | ||
| 133 | VK_FORMAT_ASTC_6x6_UNORM_BLOCK, | ||
| 134 | VK_FORMAT_ASTC_6x6_SRGB_BLOCK, | ||
| 135 | VK_FORMAT_ASTC_10x10_UNORM_BLOCK, | ||
| 136 | VK_FORMAT_ASTC_10x10_SRGB_BLOCK, | ||
| 137 | VK_FORMAT_ASTC_12x12_UNORM_BLOCK, | ||
| 138 | VK_FORMAT_ASTC_12x12_SRGB_BLOCK, | ||
| 139 | VK_FORMAT_ASTC_8x6_UNORM_BLOCK, | ||
| 140 | VK_FORMAT_ASTC_8x6_SRGB_BLOCK, | ||
| 141 | VK_FORMAT_ASTC_6x5_UNORM_BLOCK, | ||
| 142 | VK_FORMAT_ASTC_6x5_SRGB_BLOCK, | ||
| 143 | VK_FORMAT_E5B9G9R9_UFLOAT_PACK32}; | ||
| 144 | std::unordered_map<VkFormat, VkFormatProperties> format_properties; | ||
| 145 | for (const auto format : formats) { | ||
| 146 | format_properties.emplace(format, physical.GetFormatProperties(format)); | ||
| 147 | } | ||
| 148 | return format_properties; | ||
| 149 | } | ||
| 150 | |||
| 80 | } // Anonymous namespace | 151 | } // Anonymous namespace |
| 81 | 152 | ||
| 82 | VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical, | 153 | VKDevice::VKDevice(VkInstance instance, vk::PhysicalDevice physical, VkSurfaceKHR surface, |
| 83 | vk::SurfaceKHR surface) | 154 | const vk::InstanceDispatch& dld) |
| 84 | : dld{dld}, physical{physical}, properties{physical.getProperties(dld)}, | 155 | : dld{dld}, physical{physical}, properties{physical.GetProperties()}, |
| 85 | format_properties{GetFormatProperties(dld, physical)} { | 156 | format_properties{GetFormatProperties(physical, dld)} { |
| 86 | SetupFamilies(surface); | 157 | SetupFamilies(surface); |
| 87 | SetupFeatures(); | 158 | SetupFeatures(); |
| 88 | } | 159 | } |
| 89 | 160 | ||
| 90 | VKDevice::~VKDevice() = default; | 161 | VKDevice::~VKDevice() = default; |
| 91 | 162 | ||
| 92 | bool VKDevice::Create(vk::Instance instance) { | 163 | bool VKDevice::Create() { |
| 93 | const auto queue_cis = GetDeviceQueueCreateInfos(); | 164 | const auto queue_cis = GetDeviceQueueCreateInfos(); |
| 94 | const std::vector extensions = LoadExtensions(); | 165 | const std::vector extensions = LoadExtensions(); |
| 95 | 166 | ||
| 96 | vk::PhysicalDeviceFeatures2 features2; | 167 | VkPhysicalDeviceFeatures2 features2; |
| 168 | features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; | ||
| 169 | features2.pNext = nullptr; | ||
| 97 | void** next = &features2.pNext; | 170 | void** next = &features2.pNext; |
| 171 | |||
| 98 | auto& features = features2.features; | 172 | auto& features = features2.features; |
| 99 | features.vertexPipelineStoresAndAtomics = true; | 173 | features.robustBufferAccess = false; |
| 174 | features.fullDrawIndexUint32 = false; | ||
| 175 | features.imageCubeArray = false; | ||
| 100 | features.independentBlend = true; | 176 | features.independentBlend = true; |
| 177 | features.geometryShader = true; | ||
| 178 | features.tessellationShader = true; | ||
| 179 | features.sampleRateShading = false; | ||
| 180 | features.dualSrcBlend = false; | ||
| 181 | features.logicOp = false; | ||
| 182 | features.multiDrawIndirect = false; | ||
| 183 | features.drawIndirectFirstInstance = false; | ||
| 101 | features.depthClamp = true; | 184 | features.depthClamp = true; |
| 102 | features.samplerAnisotropy = true; | 185 | features.depthBiasClamp = true; |
| 186 | features.fillModeNonSolid = false; | ||
| 187 | features.depthBounds = false; | ||
| 188 | features.wideLines = false; | ||
| 103 | features.largePoints = true; | 189 | features.largePoints = true; |
| 190 | features.alphaToOne = false; | ||
| 104 | features.multiViewport = true; | 191 | features.multiViewport = true; |
| 105 | features.depthBiasClamp = true; | 192 | features.samplerAnisotropy = true; |
| 106 | features.geometryShader = true; | 193 | features.textureCompressionETC2 = false; |
| 107 | features.tessellationShader = true; | 194 | features.textureCompressionASTC_LDR = is_optimal_astc_supported; |
| 195 | features.textureCompressionBC = false; | ||
| 108 | features.occlusionQueryPrecise = true; | 196 | features.occlusionQueryPrecise = true; |
| 197 | features.pipelineStatisticsQuery = false; | ||
| 198 | features.vertexPipelineStoresAndAtomics = true; | ||
| 109 | features.fragmentStoresAndAtomics = true; | 199 | features.fragmentStoresAndAtomics = true; |
| 200 | features.shaderTessellationAndGeometryPointSize = false; | ||
| 110 | features.shaderImageGatherExtended = true; | 201 | features.shaderImageGatherExtended = true; |
| 202 | features.shaderStorageImageExtendedFormats = false; | ||
| 203 | features.shaderStorageImageMultisample = false; | ||
| 111 | features.shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported; | 204 | features.shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported; |
| 112 | features.shaderStorageImageWriteWithoutFormat = true; | 205 | features.shaderStorageImageWriteWithoutFormat = true; |
| 113 | features.textureCompressionASTC_LDR = is_optimal_astc_supported; | 206 | features.shaderUniformBufferArrayDynamicIndexing = false; |
| 114 | 207 | features.shaderSampledImageArrayDynamicIndexing = false; | |
| 115 | vk::PhysicalDevice16BitStorageFeaturesKHR bit16_storage; | 208 | features.shaderStorageBufferArrayDynamicIndexing = false; |
| 209 | features.shaderStorageImageArrayDynamicIndexing = false; | ||
| 210 | features.shaderClipDistance = false; | ||
| 211 | features.shaderCullDistance = false; | ||
| 212 | features.shaderFloat64 = false; | ||
| 213 | features.shaderInt64 = false; | ||
| 214 | features.shaderInt16 = false; | ||
| 215 | features.shaderResourceResidency = false; | ||
| 216 | features.shaderResourceMinLod = false; | ||
| 217 | features.sparseBinding = false; | ||
| 218 | features.sparseResidencyBuffer = false; | ||
| 219 | features.sparseResidencyImage2D = false; | ||
| 220 | features.sparseResidencyImage3D = false; | ||
| 221 | features.sparseResidency2Samples = false; | ||
| 222 | features.sparseResidency4Samples = false; | ||
| 223 | features.sparseResidency8Samples = false; | ||
| 224 | features.sparseResidency16Samples = false; | ||
| 225 | features.sparseResidencyAliased = false; | ||
| 226 | features.variableMultisampleRate = false; | ||
| 227 | features.inheritedQueries = false; | ||
| 228 | |||
| 229 | VkPhysicalDevice16BitStorageFeaturesKHR bit16_storage; | ||
| 230 | bit16_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR; | ||
| 231 | bit16_storage.pNext = nullptr; | ||
| 232 | bit16_storage.storageBuffer16BitAccess = false; | ||
| 116 | bit16_storage.uniformAndStorageBuffer16BitAccess = true; | 233 | bit16_storage.uniformAndStorageBuffer16BitAccess = true; |
| 234 | bit16_storage.storagePushConstant16 = false; | ||
| 235 | bit16_storage.storageInputOutput16 = false; | ||
| 117 | SetNext(next, bit16_storage); | 236 | SetNext(next, bit16_storage); |
| 118 | 237 | ||
| 119 | vk::PhysicalDevice8BitStorageFeaturesKHR bit8_storage; | 238 | VkPhysicalDevice8BitStorageFeaturesKHR bit8_storage; |
| 239 | bit8_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES_KHR; | ||
| 240 | bit8_storage.pNext = nullptr; | ||
| 241 | bit8_storage.storageBuffer8BitAccess = false; | ||
| 120 | bit8_storage.uniformAndStorageBuffer8BitAccess = true; | 242 | bit8_storage.uniformAndStorageBuffer8BitAccess = true; |
| 243 | bit8_storage.storagePushConstant8 = false; | ||
| 121 | SetNext(next, bit8_storage); | 244 | SetNext(next, bit8_storage); |
| 122 | 245 | ||
| 123 | vk::PhysicalDeviceHostQueryResetFeaturesEXT host_query_reset; | 246 | VkPhysicalDeviceHostQueryResetFeaturesEXT host_query_reset; |
| 247 | host_query_reset.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES_EXT; | ||
| 124 | host_query_reset.hostQueryReset = true; | 248 | host_query_reset.hostQueryReset = true; |
| 125 | SetNext(next, host_query_reset); | 249 | SetNext(next, host_query_reset); |
| 126 | 250 | ||
| 127 | vk::PhysicalDeviceFloat16Int8FeaturesKHR float16_int8; | 251 | VkPhysicalDeviceFloat16Int8FeaturesKHR float16_int8; |
| 128 | if (is_float16_supported) { | 252 | if (is_float16_supported) { |
| 253 | float16_int8.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR; | ||
| 254 | float16_int8.pNext = nullptr; | ||
| 129 | float16_int8.shaderFloat16 = true; | 255 | float16_int8.shaderFloat16 = true; |
| 256 | float16_int8.shaderInt8 = false; | ||
| 130 | SetNext(next, float16_int8); | 257 | SetNext(next, float16_int8); |
| 131 | } else { | 258 | } else { |
| 132 | LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively"); | 259 | LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively"); |
| 133 | } | 260 | } |
| 134 | 261 | ||
| 135 | vk::PhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout; | 262 | VkPhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout; |
| 136 | if (khr_uniform_buffer_standard_layout) { | 263 | if (khr_uniform_buffer_standard_layout) { |
| 264 | std430_layout.sType = | ||
| 265 | VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_UNIFORM_BUFFER_STANDARD_LAYOUT_FEATURES_KHR; | ||
| 266 | std430_layout.pNext = nullptr; | ||
| 137 | std430_layout.uniformBufferStandardLayout = true; | 267 | std430_layout.uniformBufferStandardLayout = true; |
| 138 | SetNext(next, std430_layout); | 268 | SetNext(next, std430_layout); |
| 139 | } else { | 269 | } else { |
| 140 | LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs"); | 270 | LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs"); |
| 141 | } | 271 | } |
| 142 | 272 | ||
| 143 | vk::PhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8; | 273 | VkPhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8; |
| 144 | if (ext_index_type_uint8) { | 274 | if (ext_index_type_uint8) { |
| 275 | index_type_uint8.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_INDEX_TYPE_UINT8_FEATURES_EXT; | ||
| 276 | index_type_uint8.pNext = nullptr; | ||
| 145 | index_type_uint8.indexTypeUint8 = true; | 277 | index_type_uint8.indexTypeUint8 = true; |
| 146 | SetNext(next, index_type_uint8); | 278 | SetNext(next, index_type_uint8); |
| 147 | } else { | 279 | } else { |
| 148 | LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes"); | 280 | LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes"); |
| 149 | } | 281 | } |
| 150 | 282 | ||
| 151 | vk::PhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback; | 283 | VkPhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback; |
| 152 | if (ext_transform_feedback) { | 284 | if (ext_transform_feedback) { |
| 285 | transform_feedback.sType = | ||
| 286 | VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT; | ||
| 287 | transform_feedback.pNext = nullptr; | ||
| 153 | transform_feedback.transformFeedback = true; | 288 | transform_feedback.transformFeedback = true; |
| 154 | transform_feedback.geometryStreams = true; | 289 | transform_feedback.geometryStreams = true; |
| 155 | SetNext(next, transform_feedback); | 290 | SetNext(next, transform_feedback); |
| @@ -161,60 +296,48 @@ bool VKDevice::Create(vk::Instance instance) { | |||
| 161 | LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted"); | 296 | LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted"); |
| 162 | } | 297 | } |
| 163 | 298 | ||
| 164 | vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(), 0, | 299 | logical = vk::Device::Create(physical, queue_cis, extensions, features2, dld); |
| 165 | nullptr, static_cast<u32>(extensions.size()), extensions.data(), | 300 | if (!logical) { |
| 166 | nullptr); | 301 | LOG_ERROR(Render_Vulkan, "Failed to create logical device"); |
| 167 | device_ci.pNext = &features2; | ||
| 168 | |||
| 169 | vk::Device unsafe_logical; | ||
| 170 | if (physical.createDevice(&device_ci, nullptr, &unsafe_logical, dld) != vk::Result::eSuccess) { | ||
| 171 | LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!"); | ||
| 172 | return false; | 302 | return false; |
| 173 | } | 303 | } |
| 174 | dld.init(instance, dld.vkGetInstanceProcAddr, unsafe_logical); | ||
| 175 | logical = UniqueDevice(unsafe_logical, {nullptr, dld}); | ||
| 176 | 304 | ||
| 177 | CollectTelemetryParameters(); | 305 | CollectTelemetryParameters(); |
| 178 | 306 | ||
| 179 | graphics_queue = logical->getQueue(graphics_family, 0, dld); | 307 | graphics_queue = logical.GetQueue(graphics_family); |
| 180 | present_queue = logical->getQueue(present_family, 0, dld); | 308 | present_queue = logical.GetQueue(present_family); |
| 181 | return true; | 309 | return true; |
| 182 | } | 310 | } |
| 183 | 311 | ||
| 184 | vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format, | 312 | VkFormat VKDevice::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, |
| 185 | vk::FormatFeatureFlags wanted_usage, | 313 | FormatType format_type) const { |
| 186 | FormatType format_type) const { | ||
| 187 | if (IsFormatSupported(wanted_format, wanted_usage, format_type)) { | 314 | if (IsFormatSupported(wanted_format, wanted_usage, format_type)) { |
| 188 | return wanted_format; | 315 | return wanted_format; |
| 189 | } | 316 | } |
| 190 | // The wanted format is not supported by hardware, search for alternatives | 317 | // The wanted format is not supported by hardware, search for alternatives |
| 191 | const vk::Format* alternatives = GetFormatAlternatives(wanted_format); | 318 | const VkFormat* alternatives = GetFormatAlternatives(wanted_format); |
| 192 | if (alternatives == nullptr) { | 319 | if (alternatives == nullptr) { |
| 193 | UNREACHABLE_MSG("Format={} with usage={} and type={} has no defined alternatives and host " | 320 | UNREACHABLE_MSG("Format={} with usage={} and type={} has no defined alternatives and host " |
| 194 | "hardware does not support it", | 321 | "hardware does not support it", |
| 195 | vk::to_string(wanted_format), vk::to_string(wanted_usage), | 322 | wanted_format, wanted_usage, format_type); |
| 196 | static_cast<u32>(format_type)); | ||
| 197 | return wanted_format; | 323 | return wanted_format; |
| 198 | } | 324 | } |
| 199 | 325 | ||
| 200 | std::size_t i = 0; | 326 | std::size_t i = 0; |
| 201 | for (vk::Format alternative = alternatives[0]; alternative != vk::Format{}; | 327 | for (VkFormat alternative = *alternatives; alternative; alternative = alternatives[++i]) { |
| 202 | alternative = alternatives[++i]) { | ||
| 203 | if (!IsFormatSupported(alternative, wanted_usage, format_type)) { | 328 | if (!IsFormatSupported(alternative, wanted_usage, format_type)) { |
| 204 | continue; | 329 | continue; |
| 205 | } | 330 | } |
| 206 | LOG_WARNING(Render_Vulkan, | 331 | LOG_WARNING(Render_Vulkan, |
| 207 | "Emulating format={} with alternative format={} with usage={} and type={}", | 332 | "Emulating format={} with alternative format={} with usage={} and type={}", |
| 208 | static_cast<u32>(wanted_format), static_cast<u32>(alternative), | 333 | wanted_format, alternative, wanted_usage, format_type); |
| 209 | static_cast<u32>(wanted_usage), static_cast<u32>(format_type)); | ||
| 210 | return alternative; | 334 | return alternative; |
| 211 | } | 335 | } |
| 212 | 336 | ||
| 213 | // No alternatives found, panic | 337 | // No alternatives found, panic |
| 214 | UNREACHABLE_MSG("Format={} with usage={} and type={} is not supported by the host hardware and " | 338 | UNREACHABLE_MSG("Format={} with usage={} and type={} is not supported by the host hardware and " |
| 215 | "doesn't support any of the alternatives", | 339 | "doesn't support any of the alternatives", |
| 216 | static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage), | 340 | wanted_format, wanted_usage, format_type); |
| 217 | static_cast<u32>(format_type)); | ||
| 218 | return wanted_format; | 341 | return wanted_format; |
| 219 | } | 342 | } |
| 220 | 343 | ||
| @@ -228,38 +351,39 @@ void VKDevice::ReportLoss() const { | |||
| 228 | return; | 351 | return; |
| 229 | } | 352 | } |
| 230 | 353 | ||
| 231 | [[maybe_unused]] const std::vector data = graphics_queue.getCheckpointDataNV(dld); | 354 | [[maybe_unused]] const std::vector data = graphics_queue.GetCheckpointDataNV(dld); |
| 232 | // Catch here in debug builds (or with optimizations disabled) the last graphics pipeline to be | 355 | // Catch here in debug builds (or with optimizations disabled) the last graphics pipeline to be |
| 233 | // executed. It can be done on a debugger by evaluating the expression: | 356 | // executed. It can be done on a debugger by evaluating the expression: |
| 234 | // *(VKGraphicsPipeline*)data[0] | 357 | // *(VKGraphicsPipeline*)data[0] |
| 235 | } | 358 | } |
| 236 | 359 | ||
| 237 | bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features) const { | 360 | bool VKDevice::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const { |
| 238 | // Disable for now to avoid converting ASTC twice. | 361 | // Disable for now to avoid converting ASTC twice. |
| 239 | static constexpr std::array astc_formats = { | 362 | static constexpr std::array astc_formats = { |
| 240 | vk::Format::eAstc4x4UnormBlock, vk::Format::eAstc4x4SrgbBlock, | 363 | VK_FORMAT_ASTC_4x4_UNORM_BLOCK, VK_FORMAT_ASTC_4x4_SRGB_BLOCK, |
| 241 | vk::Format::eAstc5x4UnormBlock, vk::Format::eAstc5x4SrgbBlock, | 364 | VK_FORMAT_ASTC_5x4_UNORM_BLOCK, VK_FORMAT_ASTC_5x4_SRGB_BLOCK, |
| 242 | vk::Format::eAstc5x5UnormBlock, vk::Format::eAstc5x5SrgbBlock, | 365 | VK_FORMAT_ASTC_5x5_UNORM_BLOCK, VK_FORMAT_ASTC_5x5_SRGB_BLOCK, |
| 243 | vk::Format::eAstc6x5UnormBlock, vk::Format::eAstc6x5SrgbBlock, | 366 | VK_FORMAT_ASTC_6x5_UNORM_BLOCK, VK_FORMAT_ASTC_6x5_SRGB_BLOCK, |
| 244 | vk::Format::eAstc6x6UnormBlock, vk::Format::eAstc6x6SrgbBlock, | 367 | VK_FORMAT_ASTC_6x6_UNORM_BLOCK, VK_FORMAT_ASTC_6x6_SRGB_BLOCK, |
| 245 | vk::Format::eAstc8x5UnormBlock, vk::Format::eAstc8x5SrgbBlock, | 368 | VK_FORMAT_ASTC_8x5_UNORM_BLOCK, VK_FORMAT_ASTC_8x5_SRGB_BLOCK, |
| 246 | vk::Format::eAstc8x6UnormBlock, vk::Format::eAstc8x6SrgbBlock, | 369 | VK_FORMAT_ASTC_8x6_UNORM_BLOCK, VK_FORMAT_ASTC_8x6_SRGB_BLOCK, |
| 247 | vk::Format::eAstc8x8UnormBlock, vk::Format::eAstc8x8SrgbBlock, | 370 | VK_FORMAT_ASTC_8x8_UNORM_BLOCK, VK_FORMAT_ASTC_8x8_SRGB_BLOCK, |
| 248 | vk::Format::eAstc10x5UnormBlock, vk::Format::eAstc10x5SrgbBlock, | 371 | VK_FORMAT_ASTC_10x5_UNORM_BLOCK, VK_FORMAT_ASTC_10x5_SRGB_BLOCK, |
| 249 | vk::Format::eAstc10x6UnormBlock, vk::Format::eAstc10x6SrgbBlock, | 372 | VK_FORMAT_ASTC_10x6_UNORM_BLOCK, VK_FORMAT_ASTC_10x6_SRGB_BLOCK, |
| 250 | vk::Format::eAstc10x8UnormBlock, vk::Format::eAstc10x8SrgbBlock, | 373 | VK_FORMAT_ASTC_10x8_UNORM_BLOCK, VK_FORMAT_ASTC_10x8_SRGB_BLOCK, |
| 251 | vk::Format::eAstc10x10UnormBlock, vk::Format::eAstc10x10SrgbBlock, | 374 | VK_FORMAT_ASTC_10x10_UNORM_BLOCK, VK_FORMAT_ASTC_10x10_SRGB_BLOCK, |
| 252 | vk::Format::eAstc12x10UnormBlock, vk::Format::eAstc12x10SrgbBlock, | 375 | VK_FORMAT_ASTC_12x10_UNORM_BLOCK, VK_FORMAT_ASTC_12x10_SRGB_BLOCK, |
| 253 | vk::Format::eAstc12x12UnormBlock, vk::Format::eAstc12x12SrgbBlock}; | 376 | VK_FORMAT_ASTC_12x12_UNORM_BLOCK, VK_FORMAT_ASTC_12x12_SRGB_BLOCK, |
| 377 | }; | ||
| 254 | if (!features.textureCompressionASTC_LDR) { | 378 | if (!features.textureCompressionASTC_LDR) { |
| 255 | return false; | 379 | return false; |
| 256 | } | 380 | } |
| 257 | const auto format_feature_usage{ | 381 | const auto format_feature_usage{ |
| 258 | vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eBlitSrc | | 382 | VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT | |
| 259 | vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc | | 383 | VK_FORMAT_FEATURE_BLIT_DST_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT | |
| 260 | vk::FormatFeatureFlagBits::eTransferDst}; | 384 | VK_FORMAT_FEATURE_TRANSFER_DST_BIT}; |
| 261 | for (const auto format : astc_formats) { | 385 | for (const auto format : astc_formats) { |
| 262 | const auto format_properties{physical.getFormatProperties(format, dld)}; | 386 | const auto format_properties{physical.GetFormatProperties(format)}; |
| 263 | if (!(format_properties.optimalTilingFeatures & format_feature_usage)) { | 387 | if (!(format_properties.optimalTilingFeatures & format_feature_usage)) { |
| 264 | return false; | 388 | return false; |
| 265 | } | 389 | } |
| @@ -267,61 +391,49 @@ bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features | |||
| 267 | return true; | 391 | return true; |
| 268 | } | 392 | } |
| 269 | 393 | ||
| 270 | bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage, | 394 | bool VKDevice::IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, |
| 271 | FormatType format_type) const { | 395 | FormatType format_type) const { |
| 272 | const auto it = format_properties.find(wanted_format); | 396 | const auto it = format_properties.find(wanted_format); |
| 273 | if (it == format_properties.end()) { | 397 | if (it == format_properties.end()) { |
| 274 | UNIMPLEMENTED_MSG("Unimplemented format query={}", vk::to_string(wanted_format)); | 398 | UNIMPLEMENTED_MSG("Unimplemented format query={}", wanted_format); |
| 275 | return true; | 399 | return true; |
| 276 | } | 400 | } |
| 277 | const auto supported_usage = GetFormatFeatures(it->second, format_type); | 401 | const auto supported_usage = GetFormatFeatures(it->second, format_type); |
| 278 | return (supported_usage & wanted_usage) == wanted_usage; | 402 | return (supported_usage & wanted_usage) == wanted_usage; |
| 279 | } | 403 | } |
| 280 | 404 | ||
| 281 | bool VKDevice::IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface, | 405 | bool VKDevice::IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface) { |
| 282 | const vk::DispatchLoaderDynamic& dld) { | ||
| 283 | static constexpr std::array required_extensions = { | ||
| 284 | VK_KHR_SWAPCHAIN_EXTENSION_NAME, | ||
| 285 | VK_KHR_16BIT_STORAGE_EXTENSION_NAME, | ||
| 286 | VK_KHR_8BIT_STORAGE_EXTENSION_NAME, | ||
| 287 | VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME, | ||
| 288 | VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME, | ||
| 289 | VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME, | ||
| 290 | VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME, | ||
| 291 | VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME, | ||
| 292 | }; | ||
| 293 | bool is_suitable = true; | 406 | bool is_suitable = true; |
| 294 | std::bitset<required_extensions.size()> available_extensions{}; | 407 | std::bitset<REQUIRED_EXTENSIONS.size()> available_extensions; |
| 295 | 408 | ||
| 296 | for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dld)) { | 409 | for (const auto& prop : physical.EnumerateDeviceExtensionProperties()) { |
| 297 | for (std::size_t i = 0; i < required_extensions.size(); ++i) { | 410 | for (std::size_t i = 0; i < REQUIRED_EXTENSIONS.size(); ++i) { |
| 298 | if (available_extensions[i]) { | 411 | if (available_extensions[i]) { |
| 299 | continue; | 412 | continue; |
| 300 | } | 413 | } |
| 301 | available_extensions[i] = | 414 | const std::string_view name{prop.extensionName}; |
| 302 | required_extensions[i] == std::string_view{prop.extensionName}; | 415 | available_extensions[i] = name == REQUIRED_EXTENSIONS[i]; |
| 303 | } | 416 | } |
| 304 | } | 417 | } |
| 305 | if (!available_extensions.all()) { | 418 | if (!available_extensions.all()) { |
| 306 | for (std::size_t i = 0; i < required_extensions.size(); ++i) { | 419 | for (std::size_t i = 0; i < REQUIRED_EXTENSIONS.size(); ++i) { |
| 307 | if (available_extensions[i]) { | 420 | if (available_extensions[i]) { |
| 308 | continue; | 421 | continue; |
| 309 | } | 422 | } |
| 310 | LOG_ERROR(Render_Vulkan, "Missing required extension: {}", required_extensions[i]); | 423 | LOG_ERROR(Render_Vulkan, "Missing required extension: {}", REQUIRED_EXTENSIONS[i]); |
| 311 | is_suitable = false; | 424 | is_suitable = false; |
| 312 | } | 425 | } |
| 313 | } | 426 | } |
| 314 | 427 | ||
| 315 | bool has_graphics{}, has_present{}; | 428 | bool has_graphics{}, has_present{}; |
| 316 | const auto queue_family_properties = physical.getQueueFamilyProperties(dld); | 429 | const std::vector queue_family_properties = physical.GetQueueFamilyProperties(); |
| 317 | for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) { | 430 | for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) { |
| 318 | const auto& family = queue_family_properties[i]; | 431 | const auto& family = queue_family_properties[i]; |
| 319 | if (family.queueCount == 0) { | 432 | if (family.queueCount == 0) { |
| 320 | continue; | 433 | continue; |
| 321 | } | 434 | } |
| 322 | has_graphics |= | 435 | has_graphics |= family.queueFlags & VK_QUEUE_GRAPHICS_BIT; |
| 323 | (family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0); | 436 | has_present |= physical.GetSurfaceSupportKHR(i, surface); |
| 324 | has_present |= physical.getSurfaceSupportKHR(i, surface, dld) != 0; | ||
| 325 | } | 437 | } |
| 326 | if (!has_graphics || !has_present) { | 438 | if (!has_graphics || !has_present) { |
| 327 | LOG_ERROR(Render_Vulkan, "Device lacks a graphics and present queue"); | 439 | LOG_ERROR(Render_Vulkan, "Device lacks a graphics and present queue"); |
| @@ -329,7 +441,7 @@ bool VKDevice::IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface, | |||
| 329 | } | 441 | } |
| 330 | 442 | ||
| 331 | // TODO(Rodrigo): Check if the device matches all requeriments. | 443 | // TODO(Rodrigo): Check if the device matches all requeriments. |
| 332 | const auto properties{physical.getProperties(dld)}; | 444 | const auto properties{physical.GetProperties()}; |
| 333 | const auto& limits{properties.limits}; | 445 | const auto& limits{properties.limits}; |
| 334 | 446 | ||
| 335 | constexpr u32 required_ubo_size = 65536; | 447 | constexpr u32 required_ubo_size = 65536; |
| @@ -346,7 +458,7 @@ bool VKDevice::IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface, | |||
| 346 | is_suitable = false; | 458 | is_suitable = false; |
| 347 | } | 459 | } |
| 348 | 460 | ||
| 349 | const auto features{physical.getFeatures(dld)}; | 461 | const auto features{physical.GetFeatures()}; |
| 350 | const std::array feature_report = { | 462 | const std::array feature_report = { |
| 351 | std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"), | 463 | std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"), |
| 352 | std::make_pair(features.independentBlend, "independentBlend"), | 464 | std::make_pair(features.independentBlend, "independentBlend"), |
| @@ -380,7 +492,7 @@ bool VKDevice::IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface, | |||
| 380 | 492 | ||
| 381 | std::vector<const char*> VKDevice::LoadExtensions() { | 493 | std::vector<const char*> VKDevice::LoadExtensions() { |
| 382 | std::vector<const char*> extensions; | 494 | std::vector<const char*> extensions; |
| 383 | const auto Test = [&](const vk::ExtensionProperties& extension, | 495 | const auto Test = [&](const VkExtensionProperties& extension, |
| 384 | std::optional<std::reference_wrapper<bool>> status, const char* name, | 496 | std::optional<std::reference_wrapper<bool>> status, const char* name, |
| 385 | bool push) { | 497 | bool push) { |
| 386 | if (extension.extensionName != std::string_view(name)) { | 498 | if (extension.extensionName != std::string_view(name)) { |
| @@ -394,22 +506,13 @@ std::vector<const char*> VKDevice::LoadExtensions() { | |||
| 394 | } | 506 | } |
| 395 | }; | 507 | }; |
| 396 | 508 | ||
| 397 | extensions.reserve(15); | 509 | extensions.reserve(7 + REQUIRED_EXTENSIONS.size()); |
| 398 | extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME); | 510 | extensions.insert(extensions.begin(), REQUIRED_EXTENSIONS.begin(), REQUIRED_EXTENSIONS.end()); |
| 399 | extensions.push_back(VK_KHR_16BIT_STORAGE_EXTENSION_NAME); | 511 | |
| 400 | extensions.push_back(VK_KHR_8BIT_STORAGE_EXTENSION_NAME); | ||
| 401 | extensions.push_back(VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME); | ||
| 402 | extensions.push_back(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME); | ||
| 403 | extensions.push_back(VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME); | ||
| 404 | extensions.push_back(VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME); | ||
| 405 | extensions.push_back(VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME); | ||
| 406 | |||
| 407 | [[maybe_unused]] const bool nsight = | ||
| 408 | std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED"); | ||
| 409 | bool has_khr_shader_float16_int8{}; | 512 | bool has_khr_shader_float16_int8{}; |
| 410 | bool has_ext_subgroup_size_control{}; | 513 | bool has_ext_subgroup_size_control{}; |
| 411 | bool has_ext_transform_feedback{}; | 514 | bool has_ext_transform_feedback{}; |
| 412 | for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dld)) { | 515 | for (const auto& extension : physical.EnumerateDeviceExtensionProperties()) { |
| 413 | Test(extension, khr_uniform_buffer_standard_layout, | 516 | Test(extension, khr_uniform_buffer_standard_layout, |
| 414 | VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true); | 517 | VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true); |
| 415 | Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, | 518 | Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, |
| @@ -429,38 +532,67 @@ std::vector<const char*> VKDevice::LoadExtensions() { | |||
| 429 | } | 532 | } |
| 430 | } | 533 | } |
| 431 | 534 | ||
| 535 | VkPhysicalDeviceFeatures2KHR features; | ||
| 536 | features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR; | ||
| 537 | |||
| 538 | VkPhysicalDeviceProperties2KHR properties; | ||
| 539 | properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR; | ||
| 540 | |||
| 432 | if (has_khr_shader_float16_int8) { | 541 | if (has_khr_shader_float16_int8) { |
| 433 | is_float16_supported = | 542 | VkPhysicalDeviceFloat16Int8FeaturesKHR float16_int8_features; |
| 434 | GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dld).shaderFloat16; | 543 | float16_int8_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR; |
| 544 | float16_int8_features.pNext = nullptr; | ||
| 545 | features.pNext = &float16_int8_features; | ||
| 546 | |||
| 547 | physical.GetFeatures2KHR(features); | ||
| 548 | is_float16_supported = float16_int8_features.shaderFloat16; | ||
| 435 | extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME); | 549 | extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME); |
| 436 | } | 550 | } |
| 437 | 551 | ||
| 438 | if (has_ext_subgroup_size_control) { | 552 | if (has_ext_subgroup_size_control) { |
| 439 | const auto features = | 553 | VkPhysicalDeviceSubgroupSizeControlFeaturesEXT subgroup_features; |
| 440 | GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dld); | 554 | subgroup_features.sType = |
| 441 | const auto properties = | 555 | VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_FEATURES_EXT; |
| 442 | GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dld); | 556 | subgroup_features.pNext = nullptr; |
| 443 | 557 | features.pNext = &subgroup_features; | |
| 444 | is_warp_potentially_bigger = properties.maxSubgroupSize > GuestWarpSize; | 558 | physical.GetFeatures2KHR(features); |
| 445 | 559 | ||
| 446 | if (features.subgroupSizeControl && properties.minSubgroupSize <= GuestWarpSize && | 560 | VkPhysicalDeviceSubgroupSizeControlPropertiesEXT subgroup_properties; |
| 447 | properties.maxSubgroupSize >= GuestWarpSize) { | 561 | subgroup_properties.sType = |
| 562 | VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_PROPERTIES_EXT; | ||
| 563 | subgroup_properties.pNext = nullptr; | ||
| 564 | properties.pNext = &subgroup_properties; | ||
| 565 | physical.GetProperties2KHR(properties); | ||
| 566 | |||
| 567 | is_warp_potentially_bigger = subgroup_properties.maxSubgroupSize > GuestWarpSize; | ||
| 568 | |||
| 569 | if (subgroup_features.subgroupSizeControl && | ||
| 570 | subgroup_properties.minSubgroupSize <= GuestWarpSize && | ||
| 571 | subgroup_properties.maxSubgroupSize >= GuestWarpSize) { | ||
| 448 | extensions.push_back(VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME); | 572 | extensions.push_back(VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME); |
| 449 | guest_warp_stages = properties.requiredSubgroupSizeStages; | 573 | guest_warp_stages = subgroup_properties.requiredSubgroupSizeStages; |
| 450 | } | 574 | } |
| 451 | } else { | 575 | } else { |
| 452 | is_warp_potentially_bigger = true; | 576 | is_warp_potentially_bigger = true; |
| 453 | } | 577 | } |
| 454 | 578 | ||
| 455 | if (has_ext_transform_feedback) { | 579 | if (has_ext_transform_feedback) { |
| 456 | const auto features = | 580 | VkPhysicalDeviceTransformFeedbackFeaturesEXT tfb_features; |
| 457 | GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dld); | 581 | tfb_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT; |
| 458 | const auto properties = | 582 | tfb_features.pNext = nullptr; |
| 459 | GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dld); | 583 | features.pNext = &tfb_features; |
| 460 | 584 | physical.GetFeatures2KHR(features); | |
| 461 | if (features.transformFeedback && features.geometryStreams && | 585 | |
| 462 | properties.maxTransformFeedbackStreams >= 4 && properties.maxTransformFeedbackBuffers && | 586 | VkPhysicalDeviceTransformFeedbackPropertiesEXT tfb_properties; |
| 463 | properties.transformFeedbackQueries && properties.transformFeedbackDraw) { | 587 | tfb_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT; |
| 588 | tfb_properties.pNext = nullptr; | ||
| 589 | properties.pNext = &tfb_properties; | ||
| 590 | physical.GetProperties2KHR(properties); | ||
| 591 | |||
| 592 | if (tfb_features.transformFeedback && tfb_features.geometryStreams && | ||
| 593 | tfb_properties.maxTransformFeedbackStreams >= 4 && | ||
| 594 | tfb_properties.maxTransformFeedbackBuffers && tfb_properties.transformFeedbackQueries && | ||
| 595 | tfb_properties.transformFeedbackDraw) { | ||
| 464 | extensions.push_back(VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME); | 596 | extensions.push_back(VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME); |
| 465 | ext_transform_feedback = true; | 597 | ext_transform_feedback = true; |
| 466 | } | 598 | } |
| @@ -469,10 +601,10 @@ std::vector<const char*> VKDevice::LoadExtensions() { | |||
| 469 | return extensions; | 601 | return extensions; |
| 470 | } | 602 | } |
| 471 | 603 | ||
| 472 | void VKDevice::SetupFamilies(vk::SurfaceKHR surface) { | 604 | void VKDevice::SetupFamilies(VkSurfaceKHR surface) { |
| 473 | std::optional<u32> graphics_family_, present_family_; | 605 | std::optional<u32> graphics_family_, present_family_; |
| 474 | 606 | ||
| 475 | const auto queue_family_properties = physical.getQueueFamilyProperties(dld); | 607 | const std::vector queue_family_properties = physical.GetQueueFamilyProperties(); |
| 476 | for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) { | 608 | for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) { |
| 477 | if (graphics_family_ && present_family_) | 609 | if (graphics_family_ && present_family_) |
| 478 | break; | 610 | break; |
| @@ -481,10 +613,10 @@ void VKDevice::SetupFamilies(vk::SurfaceKHR surface) { | |||
| 481 | if (queue_family.queueCount == 0) | 613 | if (queue_family.queueCount == 0) |
| 482 | continue; | 614 | continue; |
| 483 | 615 | ||
| 484 | if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics) { | 616 | if (queue_family.queueFlags & VK_QUEUE_GRAPHICS_BIT) { |
| 485 | graphics_family_ = i; | 617 | graphics_family_ = i; |
| 486 | } | 618 | } |
| 487 | if (physical.getSurfaceSupportKHR(i, surface, dld)) { | 619 | if (physical.GetSurfaceSupportKHR(i, surface)) { |
| 488 | present_family_ = i; | 620 | present_family_ = i; |
| 489 | } | 621 | } |
| 490 | } | 622 | } |
| @@ -495,120 +627,48 @@ void VKDevice::SetupFamilies(vk::SurfaceKHR surface) { | |||
| 495 | } | 627 | } |
| 496 | 628 | ||
| 497 | void VKDevice::SetupFeatures() { | 629 | void VKDevice::SetupFeatures() { |
| 498 | const auto supported_features{physical.getFeatures(dld)}; | 630 | const auto supported_features{physical.GetFeatures()}; |
| 499 | is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat; | 631 | is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat; |
| 500 | is_optimal_astc_supported = IsOptimalAstcSupported(supported_features); | 632 | is_optimal_astc_supported = IsOptimalAstcSupported(supported_features); |
| 501 | } | 633 | } |
| 502 | 634 | ||
| 503 | void VKDevice::CollectTelemetryParameters() { | 635 | void VKDevice::CollectTelemetryParameters() { |
| 504 | const auto driver = GetProperties<vk::PhysicalDeviceDriverPropertiesKHR>(physical, dld); | 636 | VkPhysicalDeviceDriverPropertiesKHR driver; |
| 637 | driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR; | ||
| 638 | driver.pNext = nullptr; | ||
| 639 | |||
| 640 | VkPhysicalDeviceProperties2KHR properties; | ||
| 641 | properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR; | ||
| 642 | properties.pNext = &driver; | ||
| 643 | physical.GetProperties2KHR(properties); | ||
| 644 | |||
| 505 | driver_id = driver.driverID; | 645 | driver_id = driver.driverID; |
| 506 | vendor_name = driver.driverName; | 646 | vendor_name = driver.driverName; |
| 507 | 647 | ||
| 508 | const auto extensions = physical.enumerateDeviceExtensionProperties(nullptr, dld); | 648 | const std::vector extensions = physical.EnumerateDeviceExtensionProperties(); |
| 509 | reported_extensions.reserve(std::size(extensions)); | 649 | reported_extensions.reserve(std::size(extensions)); |
| 510 | for (const auto& extension : extensions) { | 650 | for (const auto& extension : extensions) { |
| 511 | reported_extensions.push_back(extension.extensionName); | 651 | reported_extensions.push_back(extension.extensionName); |
| 512 | } | 652 | } |
| 513 | } | 653 | } |
| 514 | 654 | ||
| 515 | std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const { | 655 | std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const { |
| 516 | static const float QUEUE_PRIORITY = 1.0f; | 656 | static constexpr float QUEUE_PRIORITY = 1.0f; |
| 517 | 657 | ||
| 518 | std::set<u32> unique_queue_families = {graphics_family, present_family}; | 658 | std::unordered_set<u32> unique_queue_families = {graphics_family, present_family}; |
| 519 | std::vector<vk::DeviceQueueCreateInfo> queue_cis; | 659 | std::vector<VkDeviceQueueCreateInfo> queue_cis; |
| 520 | 660 | ||
| 521 | for (u32 queue_family : unique_queue_families) | 661 | for (const u32 queue_family : unique_queue_families) { |
| 522 | queue_cis.push_back({{}, queue_family, 1, &QUEUE_PRIORITY}); | 662 | VkDeviceQueueCreateInfo& ci = queue_cis.emplace_back(); |
| 663 | ci.sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO; | ||
| 664 | ci.pNext = nullptr; | ||
| 665 | ci.flags = 0; | ||
| 666 | ci.queueFamilyIndex = queue_family; | ||
| 667 | ci.queueCount = 1; | ||
| 668 | ci.pQueuePriorities = &QUEUE_PRIORITY; | ||
| 669 | } | ||
| 523 | 670 | ||
| 524 | return queue_cis; | 671 | return queue_cis; |
| 525 | } | 672 | } |
| 526 | 673 | ||
| 527 | std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties( | ||
| 528 | const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical) { | ||
| 529 | static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32, | ||
| 530 | vk::Format::eA8B8G8R8UintPack32, | ||
| 531 | vk::Format::eA8B8G8R8SnormPack32, | ||
| 532 | vk::Format::eA8B8G8R8SrgbPack32, | ||
| 533 | vk::Format::eB5G6R5UnormPack16, | ||
| 534 | vk::Format::eA2B10G10R10UnormPack32, | ||
| 535 | vk::Format::eA1R5G5B5UnormPack16, | ||
| 536 | vk::Format::eR32G32B32A32Sfloat, | ||
| 537 | vk::Format::eR32G32B32A32Uint, | ||
| 538 | vk::Format::eR32G32Sfloat, | ||
| 539 | vk::Format::eR32G32Uint, | ||
| 540 | vk::Format::eR16G16B16A16Uint, | ||
| 541 | vk::Format::eR16G16B16A16Snorm, | ||
| 542 | vk::Format::eR16G16B16A16Unorm, | ||
| 543 | vk::Format::eR16G16Unorm, | ||
| 544 | vk::Format::eR16G16Snorm, | ||
| 545 | vk::Format::eR16G16Sfloat, | ||
| 546 | vk::Format::eR16Unorm, | ||
| 547 | vk::Format::eR8G8B8A8Srgb, | ||
| 548 | vk::Format::eR8G8Unorm, | ||
| 549 | vk::Format::eR8G8Snorm, | ||
| 550 | vk::Format::eR8Unorm, | ||
| 551 | vk::Format::eR8Uint, | ||
| 552 | vk::Format::eB10G11R11UfloatPack32, | ||
| 553 | vk::Format::eR32Sfloat, | ||
| 554 | vk::Format::eR32Uint, | ||
| 555 | vk::Format::eR32Sint, | ||
| 556 | vk::Format::eR16Sfloat, | ||
| 557 | vk::Format::eR16G16B16A16Sfloat, | ||
| 558 | vk::Format::eB8G8R8A8Unorm, | ||
| 559 | vk::Format::eR4G4B4A4UnormPack16, | ||
| 560 | vk::Format::eD32Sfloat, | ||
| 561 | vk::Format::eD16Unorm, | ||
| 562 | vk::Format::eD16UnormS8Uint, | ||
| 563 | vk::Format::eD24UnormS8Uint, | ||
| 564 | vk::Format::eD32SfloatS8Uint, | ||
| 565 | vk::Format::eBc1RgbaUnormBlock, | ||
| 566 | vk::Format::eBc2UnormBlock, | ||
| 567 | vk::Format::eBc3UnormBlock, | ||
| 568 | vk::Format::eBc4UnormBlock, | ||
| 569 | vk::Format::eBc5UnormBlock, | ||
| 570 | vk::Format::eBc5SnormBlock, | ||
| 571 | vk::Format::eBc7UnormBlock, | ||
| 572 | vk::Format::eBc6HUfloatBlock, | ||
| 573 | vk::Format::eBc6HSfloatBlock, | ||
| 574 | vk::Format::eBc1RgbaSrgbBlock, | ||
| 575 | vk::Format::eBc2SrgbBlock, | ||
| 576 | vk::Format::eBc3SrgbBlock, | ||
| 577 | vk::Format::eBc7SrgbBlock, | ||
| 578 | vk::Format::eAstc4x4UnormBlock, | ||
| 579 | vk::Format::eAstc4x4SrgbBlock, | ||
| 580 | vk::Format::eAstc5x4UnormBlock, | ||
| 581 | vk::Format::eAstc5x4SrgbBlock, | ||
| 582 | vk::Format::eAstc5x5UnormBlock, | ||
| 583 | vk::Format::eAstc5x5SrgbBlock, | ||
| 584 | vk::Format::eAstc6x5UnormBlock, | ||
| 585 | vk::Format::eAstc6x5SrgbBlock, | ||
| 586 | vk::Format::eAstc6x6UnormBlock, | ||
| 587 | vk::Format::eAstc6x6SrgbBlock, | ||
| 588 | vk::Format::eAstc8x5UnormBlock, | ||
| 589 | vk::Format::eAstc8x5SrgbBlock, | ||
| 590 | vk::Format::eAstc8x6UnormBlock, | ||
| 591 | vk::Format::eAstc8x6SrgbBlock, | ||
| 592 | vk::Format::eAstc8x8UnormBlock, | ||
| 593 | vk::Format::eAstc8x8SrgbBlock, | ||
| 594 | vk::Format::eAstc10x5UnormBlock, | ||
| 595 | vk::Format::eAstc10x5SrgbBlock, | ||
| 596 | vk::Format::eAstc10x6UnormBlock, | ||
| 597 | vk::Format::eAstc10x6SrgbBlock, | ||
| 598 | vk::Format::eAstc10x8UnormBlock, | ||
| 599 | vk::Format::eAstc10x8SrgbBlock, | ||
| 600 | vk::Format::eAstc10x10UnormBlock, | ||
| 601 | vk::Format::eAstc10x10SrgbBlock, | ||
| 602 | vk::Format::eAstc12x10UnormBlock, | ||
| 603 | vk::Format::eAstc12x10SrgbBlock, | ||
| 604 | vk::Format::eAstc12x12UnormBlock, | ||
| 605 | vk::Format::eAstc12x12SrgbBlock, | ||
| 606 | vk::Format::eE5B9G9R9UfloatPack32}; | ||
| 607 | std::unordered_map<vk::Format, vk::FormatProperties> format_properties; | ||
| 608 | for (const auto format : formats) { | ||
| 609 | format_properties.emplace(format, physical.getFormatProperties(format, dld)); | ||
| 610 | } | ||
| 611 | return format_properties; | ||
| 612 | } | ||
| 613 | |||
| 614 | } // namespace Vulkan | 674 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h index d9d809852..60d64572a 100644 --- a/src/video_core/renderer_vulkan/vk_device.h +++ b/src/video_core/renderer_vulkan/vk_device.h | |||
| @@ -8,8 +8,9 @@ | |||
| 8 | #include <string_view> | 8 | #include <string_view> |
| 9 | #include <unordered_map> | 9 | #include <unordered_map> |
| 10 | #include <vector> | 10 | #include <vector> |
| 11 | |||
| 11 | #include "common/common_types.h" | 12 | #include "common/common_types.h" |
| 12 | #include "video_core/renderer_vulkan/declarations.h" | 13 | #include "video_core/renderer_vulkan/wrapper.h" |
| 13 | 14 | ||
| 14 | namespace Vulkan { | 15 | namespace Vulkan { |
| 15 | 16 | ||
| @@ -22,12 +23,12 @@ const u32 GuestWarpSize = 32; | |||
| 22 | /// Handles data specific to a physical device. | 23 | /// Handles data specific to a physical device. |
| 23 | class VKDevice final { | 24 | class VKDevice final { |
| 24 | public: | 25 | public: |
| 25 | explicit VKDevice(const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical, | 26 | explicit VKDevice(VkInstance instance, vk::PhysicalDevice physical, VkSurfaceKHR surface, |
| 26 | vk::SurfaceKHR surface); | 27 | const vk::InstanceDispatch& dld); |
| 27 | ~VKDevice(); | 28 | ~VKDevice(); |
| 28 | 29 | ||
| 29 | /// Initializes the device. Returns true on success. | 30 | /// Initializes the device. Returns true on success. |
| 30 | bool Create(vk::Instance instance); | 31 | bool Create(); |
| 31 | 32 | ||
| 32 | /** | 33 | /** |
| 33 | * Returns a format supported by the device for the passed requeriments. | 34 | * Returns a format supported by the device for the passed requeriments. |
| @@ -36,20 +37,20 @@ public: | |||
| 36 | * @param format_type Format type usage. | 37 | * @param format_type Format type usage. |
| 37 | * @returns A format supported by the device. | 38 | * @returns A format supported by the device. |
| 38 | */ | 39 | */ |
| 39 | vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage, | 40 | VkFormat GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, |
| 40 | FormatType format_type) const; | 41 | FormatType format_type) const; |
| 41 | 42 | ||
| 42 | /// Reports a device loss. | 43 | /// Reports a device loss. |
| 43 | void ReportLoss() const; | 44 | void ReportLoss() const; |
| 44 | 45 | ||
| 45 | /// Returns the dispatch loader with direct function pointers of the device. | 46 | /// Returns the dispatch loader with direct function pointers of the device. |
| 46 | const vk::DispatchLoaderDynamic& GetDispatchLoader() const { | 47 | const vk::DeviceDispatch& GetDispatchLoader() const { |
| 47 | return dld; | 48 | return dld; |
| 48 | } | 49 | } |
| 49 | 50 | ||
| 50 | /// Returns the logical device. | 51 | /// Returns the logical device. |
| 51 | vk::Device GetLogical() const { | 52 | const vk::Device& GetLogical() const { |
| 52 | return logical.get(); | 53 | return logical; |
| 53 | } | 54 | } |
| 54 | 55 | ||
| 55 | /// Returns the physical device. | 56 | /// Returns the physical device. |
| @@ -79,7 +80,7 @@ public: | |||
| 79 | 80 | ||
| 80 | /// Returns true if the device is integrated with the host CPU. | 81 | /// Returns true if the device is integrated with the host CPU. |
| 81 | bool IsIntegrated() const { | 82 | bool IsIntegrated() const { |
| 82 | return properties.deviceType == vk::PhysicalDeviceType::eIntegratedGpu; | 83 | return properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; |
| 83 | } | 84 | } |
| 84 | 85 | ||
| 85 | /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers. | 86 | /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers. |
| @@ -98,27 +99,27 @@ public: | |||
| 98 | } | 99 | } |
| 99 | 100 | ||
| 100 | /// Returns the driver ID. | 101 | /// Returns the driver ID. |
| 101 | vk::DriverIdKHR GetDriverID() const { | 102 | VkDriverIdKHR GetDriverID() const { |
| 102 | return driver_id; | 103 | return driver_id; |
| 103 | } | 104 | } |
| 104 | 105 | ||
| 105 | /// Returns uniform buffer alignment requeriment. | 106 | /// Returns uniform buffer alignment requeriment. |
| 106 | vk::DeviceSize GetUniformBufferAlignment() const { | 107 | VkDeviceSize GetUniformBufferAlignment() const { |
| 107 | return properties.limits.minUniformBufferOffsetAlignment; | 108 | return properties.limits.minUniformBufferOffsetAlignment; |
| 108 | } | 109 | } |
| 109 | 110 | ||
| 110 | /// Returns storage alignment requeriment. | 111 | /// Returns storage alignment requeriment. |
| 111 | vk::DeviceSize GetStorageBufferAlignment() const { | 112 | VkDeviceSize GetStorageBufferAlignment() const { |
| 112 | return properties.limits.minStorageBufferOffsetAlignment; | 113 | return properties.limits.minStorageBufferOffsetAlignment; |
| 113 | } | 114 | } |
| 114 | 115 | ||
| 115 | /// Returns the maximum range for storage buffers. | 116 | /// Returns the maximum range for storage buffers. |
| 116 | vk::DeviceSize GetMaxStorageBufferRange() const { | 117 | VkDeviceSize GetMaxStorageBufferRange() const { |
| 117 | return properties.limits.maxStorageBufferRange; | 118 | return properties.limits.maxStorageBufferRange; |
| 118 | } | 119 | } |
| 119 | 120 | ||
| 120 | /// Returns the maximum size for push constants. | 121 | /// Returns the maximum size for push constants. |
| 121 | vk::DeviceSize GetMaxPushConstantsSize() const { | 122 | VkDeviceSize GetMaxPushConstantsSize() const { |
| 122 | return properties.limits.maxPushConstantsSize; | 123 | return properties.limits.maxPushConstantsSize; |
| 123 | } | 124 | } |
| 124 | 125 | ||
| @@ -138,8 +139,8 @@ public: | |||
| 138 | } | 139 | } |
| 139 | 140 | ||
| 140 | /// Returns true if the device can be forced to use the guest warp size. | 141 | /// Returns true if the device can be forced to use the guest warp size. |
| 141 | bool IsGuestWarpSizeSupported(vk::ShaderStageFlagBits stage) const { | 142 | bool IsGuestWarpSizeSupported(VkShaderStageFlagBits stage) const { |
| 142 | return (guest_warp_stages & stage) != vk::ShaderStageFlags{}; | 143 | return guest_warp_stages & stage; |
| 143 | } | 144 | } |
| 144 | 145 | ||
| 145 | /// Returns true if formatless image load is supported. | 146 | /// Returns true if formatless image load is supported. |
| @@ -188,15 +189,14 @@ public: | |||
| 188 | } | 189 | } |
| 189 | 190 | ||
| 190 | /// Checks if the physical device is suitable. | 191 | /// Checks if the physical device is suitable. |
| 191 | static bool IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface, | 192 | static bool IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface); |
| 192 | const vk::DispatchLoaderDynamic& dld); | ||
| 193 | 193 | ||
| 194 | private: | 194 | private: |
| 195 | /// Loads extensions into a vector and stores available ones in this object. | 195 | /// Loads extensions into a vector and stores available ones in this object. |
| 196 | std::vector<const char*> LoadExtensions(); | 196 | std::vector<const char*> LoadExtensions(); |
| 197 | 197 | ||
| 198 | /// Sets up queue families. | 198 | /// Sets up queue families. |
| 199 | void SetupFamilies(vk::SurfaceKHR surface); | 199 | void SetupFamilies(VkSurfaceKHR surface); |
| 200 | 200 | ||
| 201 | /// Sets up device features. | 201 | /// Sets up device features. |
| 202 | void SetupFeatures(); | 202 | void SetupFeatures(); |
| @@ -205,32 +205,28 @@ private: | |||
| 205 | void CollectTelemetryParameters(); | 205 | void CollectTelemetryParameters(); |
| 206 | 206 | ||
| 207 | /// Returns a list of queue initialization descriptors. | 207 | /// Returns a list of queue initialization descriptors. |
| 208 | std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const; | 208 | std::vector<VkDeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const; |
| 209 | 209 | ||
| 210 | /// Returns true if ASTC textures are natively supported. | 210 | /// Returns true if ASTC textures are natively supported. |
| 211 | bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features) const; | 211 | bool IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const; |
| 212 | 212 | ||
| 213 | /// Returns true if a format is supported. | 213 | /// Returns true if a format is supported. |
| 214 | bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage, | 214 | bool IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, |
| 215 | FormatType format_type) const; | 215 | FormatType format_type) const; |
| 216 | 216 | ||
| 217 | /// Returns the device properties for Vulkan formats. | 217 | vk::DeviceDispatch dld; ///< Device function pointers. |
| 218 | static std::unordered_map<vk::Format, vk::FormatProperties> GetFormatProperties( | 218 | vk::PhysicalDevice physical; ///< Physical device. |
| 219 | const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical); | 219 | VkPhysicalDeviceProperties properties; ///< Device properties. |
| 220 | 220 | vk::Device logical; ///< Logical device. | |
| 221 | vk::DispatchLoaderDynamic dld; ///< Device function pointers. | 221 | vk::Queue graphics_queue; ///< Main graphics queue. |
| 222 | vk::PhysicalDevice physical; ///< Physical device. | 222 | vk::Queue present_queue; ///< Main present queue. |
| 223 | vk::PhysicalDeviceProperties properties; ///< Device properties. | 223 | u32 graphics_family{}; ///< Main graphics queue family index. |
| 224 | UniqueDevice logical; ///< Logical device. | 224 | u32 present_family{}; ///< Main present queue family index. |
| 225 | vk::Queue graphics_queue; ///< Main graphics queue. | 225 | VkDriverIdKHR driver_id{}; ///< Driver ID. |
| 226 | vk::Queue present_queue; ///< Main present queue. | 226 | VkShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced.ed |
| 227 | u32 graphics_family{}; ///< Main graphics queue family index. | 227 | bool is_optimal_astc_supported{}; ///< Support for native ASTC. |
| 228 | u32 present_family{}; ///< Main present queue family index. | 228 | bool is_float16_supported{}; ///< Support for float16 arithmetics. |
| 229 | vk::DriverIdKHR driver_id{}; ///< Driver ID. | 229 | bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. |
| 230 | vk::ShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced.ed | ||
| 231 | bool is_optimal_astc_supported{}; ///< Support for native ASTC. | ||
| 232 | bool is_float16_supported{}; ///< Support for float16 arithmetics. | ||
| 233 | bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. | ||
| 234 | bool is_formatless_image_load_supported{}; ///< Support for shader image read without format. | 230 | bool is_formatless_image_load_supported{}; ///< Support for shader image read without format. |
| 235 | bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs. | 231 | bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs. |
| 236 | bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8. | 232 | bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8. |
| @@ -244,7 +240,7 @@ private: | |||
| 244 | std::vector<std::string> reported_extensions; ///< Reported Vulkan extensions. | 240 | std::vector<std::string> reported_extensions; ///< Reported Vulkan extensions. |
| 245 | 241 | ||
| 246 | /// Format properties dictionary. | 242 | /// Format properties dictionary. |
| 247 | std::unordered_map<vk::Format, vk::FormatProperties> format_properties; | 243 | std::unordered_map<VkFormat, VkFormatProperties> format_properties; |
| 248 | }; | 244 | }; |
| 249 | 245 | ||
| 250 | } // namespace Vulkan | 246 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index 6a02403c1..b540b838d 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp | |||
| @@ -2,11 +2,13 @@ | |||
| 2 | // Licensed under GPLv2 or any later version | 2 | // Licensed under GPLv2 or any later version |
| 3 | // Refer to the license.txt file included. | 3 | // Refer to the license.txt file included. |
| 4 | 4 | ||
| 5 | #include <array> | ||
| 6 | #include <cstring> | ||
| 5 | #include <vector> | 7 | #include <vector> |
| 8 | |||
| 6 | #include "common/assert.h" | 9 | #include "common/assert.h" |
| 7 | #include "common/common_types.h" | 10 | #include "common/common_types.h" |
| 8 | #include "common/microprofile.h" | 11 | #include "common/microprofile.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" | 12 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" |
| 11 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 13 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 12 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 14 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| @@ -16,6 +18,7 @@ | |||
| 16 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" | 18 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" |
| 17 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 19 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 18 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 20 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 21 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 19 | 22 | ||
| 20 | namespace Vulkan { | 23 | namespace Vulkan { |
| 21 | 24 | ||
| @@ -23,21 +26,26 @@ MICROPROFILE_DECLARE(Vulkan_PipelineCache); | |||
| 23 | 26 | ||
| 24 | namespace { | 27 | namespace { |
| 25 | 28 | ||
| 26 | vk::StencilOpState GetStencilFaceState(const FixedPipelineState::StencilFace& face) { | 29 | VkStencilOpState GetStencilFaceState(const FixedPipelineState::StencilFace& face) { |
| 27 | return vk::StencilOpState(MaxwellToVK::StencilOp(face.action_stencil_fail), | 30 | VkStencilOpState state; |
| 28 | MaxwellToVK::StencilOp(face.action_depth_pass), | 31 | state.failOp = MaxwellToVK::StencilOp(face.action_stencil_fail); |
| 29 | MaxwellToVK::StencilOp(face.action_depth_fail), | 32 | state.passOp = MaxwellToVK::StencilOp(face.action_depth_pass); |
| 30 | MaxwellToVK::ComparisonOp(face.test_func), 0, 0, 0); | 33 | state.depthFailOp = MaxwellToVK::StencilOp(face.action_depth_fail); |
| 34 | state.compareOp = MaxwellToVK::ComparisonOp(face.test_func); | ||
| 35 | state.compareMask = 0; | ||
| 36 | state.writeMask = 0; | ||
| 37 | state.reference = 0; | ||
| 38 | return state; | ||
| 31 | } | 39 | } |
| 32 | 40 | ||
| 33 | bool SupportsPrimitiveRestart(vk::PrimitiveTopology topology) { | 41 | bool SupportsPrimitiveRestart(VkPrimitiveTopology topology) { |
| 34 | static constexpr std::array unsupported_topologies = { | 42 | static constexpr std::array unsupported_topologies = { |
| 35 | vk::PrimitiveTopology::ePointList, | 43 | VK_PRIMITIVE_TOPOLOGY_POINT_LIST, |
| 36 | vk::PrimitiveTopology::eLineList, | 44 | VK_PRIMITIVE_TOPOLOGY_LINE_LIST, |
| 37 | vk::PrimitiveTopology::eTriangleList, | 45 | VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST, |
| 38 | vk::PrimitiveTopology::eLineListWithAdjacency, | 46 | VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY, |
| 39 | vk::PrimitiveTopology::eTriangleListWithAdjacency, | 47 | VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY, |
| 40 | vk::PrimitiveTopology::ePatchList}; | 48 | VK_PRIMITIVE_TOPOLOGY_PATCH_LIST}; |
| 41 | return std::find(std::begin(unsupported_topologies), std::end(unsupported_topologies), | 49 | return std::find(std::begin(unsupported_topologies), std::end(unsupported_topologies), |
| 42 | topology) == std::end(unsupported_topologies); | 50 | topology) == std::end(unsupported_topologies); |
| 43 | } | 51 | } |
| @@ -49,7 +57,7 @@ VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device, VKScheduler& sche | |||
| 49 | VKUpdateDescriptorQueue& update_descriptor_queue, | 57 | VKUpdateDescriptorQueue& update_descriptor_queue, |
| 50 | VKRenderPassCache& renderpass_cache, | 58 | VKRenderPassCache& renderpass_cache, |
| 51 | const GraphicsPipelineCacheKey& key, | 59 | const GraphicsPipelineCacheKey& key, |
| 52 | const std::vector<vk::DescriptorSetLayoutBinding>& bindings, | 60 | vk::Span<VkDescriptorSetLayoutBinding> bindings, |
| 53 | const SPIRVProgram& program) | 61 | const SPIRVProgram& program) |
| 54 | : device{device}, scheduler{scheduler}, fixed_state{key.fixed_state}, hash{key.Hash()}, | 62 | : device{device}, scheduler{scheduler}, fixed_state{key.fixed_state}, hash{key.Hash()}, |
| 55 | descriptor_set_layout{CreateDescriptorSetLayout(bindings)}, | 63 | descriptor_set_layout{CreateDescriptorSetLayout(bindings)}, |
| @@ -63,7 +71,7 @@ VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device, VKScheduler& sche | |||
| 63 | 71 | ||
| 64 | VKGraphicsPipeline::~VKGraphicsPipeline() = default; | 72 | VKGraphicsPipeline::~VKGraphicsPipeline() = default; |
| 65 | 73 | ||
| 66 | vk::DescriptorSet VKGraphicsPipeline::CommitDescriptorSet() { | 74 | VkDescriptorSet VKGraphicsPipeline::CommitDescriptorSet() { |
| 67 | if (!descriptor_template) { | 75 | if (!descriptor_template) { |
| 68 | return {}; | 76 | return {}; |
| 69 | } | 77 | } |
| @@ -72,27 +80,32 @@ vk::DescriptorSet VKGraphicsPipeline::CommitDescriptorSet() { | |||
| 72 | return set; | 80 | return set; |
| 73 | } | 81 | } |
| 74 | 82 | ||
| 75 | UniqueDescriptorSetLayout VKGraphicsPipeline::CreateDescriptorSetLayout( | 83 | vk::DescriptorSetLayout VKGraphicsPipeline::CreateDescriptorSetLayout( |
| 76 | const std::vector<vk::DescriptorSetLayoutBinding>& bindings) const { | 84 | vk::Span<VkDescriptorSetLayoutBinding> bindings) const { |
| 77 | const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci( | 85 | VkDescriptorSetLayoutCreateInfo ci; |
| 78 | {}, static_cast<u32>(bindings.size()), bindings.data()); | 86 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; |
| 79 | 87 | ci.pNext = nullptr; | |
| 80 | const auto dev = device.GetLogical(); | 88 | ci.flags = 0; |
| 81 | const auto& dld = device.GetDispatchLoader(); | 89 | ci.bindingCount = bindings.size(); |
| 82 | return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld); | 90 | ci.pBindings = bindings.data(); |
| 91 | return device.GetLogical().CreateDescriptorSetLayout(ci); | ||
| 83 | } | 92 | } |
| 84 | 93 | ||
| 85 | UniquePipelineLayout VKGraphicsPipeline::CreatePipelineLayout() const { | 94 | vk::PipelineLayout VKGraphicsPipeline::CreatePipelineLayout() const { |
| 86 | const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout, 0, | 95 | VkPipelineLayoutCreateInfo ci; |
| 87 | nullptr); | 96 | ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; |
| 88 | const auto dev = device.GetLogical(); | 97 | ci.pNext = nullptr; |
| 89 | const auto& dld = device.GetDispatchLoader(); | 98 | ci.flags = 0; |
| 90 | return dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld); | 99 | ci.setLayoutCount = 1; |
| 100 | ci.pSetLayouts = descriptor_set_layout.address(); | ||
| 101 | ci.pushConstantRangeCount = 0; | ||
| 102 | ci.pPushConstantRanges = nullptr; | ||
| 103 | return device.GetLogical().CreatePipelineLayout(ci); | ||
| 91 | } | 104 | } |
| 92 | 105 | ||
| 93 | UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplate( | 106 | vk::DescriptorUpdateTemplateKHR VKGraphicsPipeline::CreateDescriptorUpdateTemplate( |
| 94 | const SPIRVProgram& program) const { | 107 | const SPIRVProgram& program) const { |
| 95 | std::vector<vk::DescriptorUpdateTemplateEntry> template_entries; | 108 | std::vector<VkDescriptorUpdateTemplateEntry> template_entries; |
| 96 | u32 binding = 0; | 109 | u32 binding = 0; |
| 97 | u32 offset = 0; | 110 | u32 offset = 0; |
| 98 | for (const auto& stage : program) { | 111 | for (const auto& stage : program) { |
| @@ -102,38 +115,47 @@ UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplat | |||
| 102 | } | 115 | } |
| 103 | if (template_entries.empty()) { | 116 | if (template_entries.empty()) { |
| 104 | // If the shader doesn't use descriptor sets, skip template creation. | 117 | // If the shader doesn't use descriptor sets, skip template creation. |
| 105 | return UniqueDescriptorUpdateTemplate{}; | 118 | return {}; |
| 106 | } | 119 | } |
| 107 | 120 | ||
| 108 | const vk::DescriptorUpdateTemplateCreateInfo template_ci( | 121 | VkDescriptorUpdateTemplateCreateInfoKHR ci; |
| 109 | {}, static_cast<u32>(template_entries.size()), template_entries.data(), | 122 | ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR; |
| 110 | vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout, | 123 | ci.pNext = nullptr; |
| 111 | vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET); | 124 | ci.flags = 0; |
| 112 | 125 | ci.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size()); | |
| 113 | const auto dev = device.GetLogical(); | 126 | ci.pDescriptorUpdateEntries = template_entries.data(); |
| 114 | const auto& dld = device.GetDispatchLoader(); | 127 | ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR; |
| 115 | return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld); | 128 | ci.descriptorSetLayout = *descriptor_set_layout; |
| 129 | ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; | ||
| 130 | ci.pipelineLayout = *layout; | ||
| 131 | ci.set = DESCRIPTOR_SET; | ||
| 132 | return device.GetLogical().CreateDescriptorUpdateTemplateKHR(ci); | ||
| 116 | } | 133 | } |
| 117 | 134 | ||
| 118 | std::vector<UniqueShaderModule> VKGraphicsPipeline::CreateShaderModules( | 135 | std::vector<vk::ShaderModule> VKGraphicsPipeline::CreateShaderModules( |
| 119 | const SPIRVProgram& program) const { | 136 | const SPIRVProgram& program) const { |
| 120 | std::vector<UniqueShaderModule> modules; | 137 | VkShaderModuleCreateInfo ci; |
| 121 | const auto dev = device.GetLogical(); | 138 | ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; |
| 122 | const auto& dld = device.GetDispatchLoader(); | 139 | ci.pNext = nullptr; |
| 140 | ci.flags = 0; | ||
| 141 | |||
| 142 | std::vector<vk::ShaderModule> modules; | ||
| 143 | modules.reserve(Maxwell::MaxShaderStage); | ||
| 123 | for (std::size_t i = 0; i < Maxwell::MaxShaderStage; ++i) { | 144 | for (std::size_t i = 0; i < Maxwell::MaxShaderStage; ++i) { |
| 124 | const auto& stage = program[i]; | 145 | const auto& stage = program[i]; |
| 125 | if (!stage) { | 146 | if (!stage) { |
| 126 | continue; | 147 | continue; |
| 127 | } | 148 | } |
| 128 | const vk::ShaderModuleCreateInfo module_ci({}, stage->code.size() * sizeof(u32), | 149 | |
| 129 | stage->code.data()); | 150 | ci.codeSize = stage->code.size() * sizeof(u32); |
| 130 | modules.emplace_back(dev.createShaderModuleUnique(module_ci, nullptr, dld)); | 151 | ci.pCode = stage->code.data(); |
| 152 | modules.push_back(device.GetLogical().CreateShaderModule(ci)); | ||
| 131 | } | 153 | } |
| 132 | return modules; | 154 | return modules; |
| 133 | } | 155 | } |
| 134 | 156 | ||
| 135 | UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params, | 157 | vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params, |
| 136 | const SPIRVProgram& program) const { | 158 | const SPIRVProgram& program) const { |
| 137 | const auto& vi = fixed_state.vertex_input; | 159 | const auto& vi = fixed_state.vertex_input; |
| 138 | const auto& ia = fixed_state.input_assembly; | 160 | const auto& ia = fixed_state.input_assembly; |
| 139 | const auto& ds = fixed_state.depth_stencil; | 161 | const auto& ds = fixed_state.depth_stencil; |
| @@ -141,19 +163,26 @@ UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& render | |||
| 141 | const auto& ts = fixed_state.tessellation; | 163 | const auto& ts = fixed_state.tessellation; |
| 142 | const auto& rs = fixed_state.rasterizer; | 164 | const auto& rs = fixed_state.rasterizer; |
| 143 | 165 | ||
| 144 | std::vector<vk::VertexInputBindingDescription> vertex_bindings; | 166 | std::vector<VkVertexInputBindingDescription> vertex_bindings; |
| 145 | std::vector<vk::VertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors; | 167 | std::vector<VkVertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors; |
| 146 | for (std::size_t i = 0; i < vi.num_bindings; ++i) { | 168 | for (std::size_t i = 0; i < vi.num_bindings; ++i) { |
| 147 | const auto& binding = vi.bindings[i]; | 169 | const auto& binding = vi.bindings[i]; |
| 148 | const bool instanced = binding.divisor != 0; | 170 | const bool instanced = binding.divisor != 0; |
| 149 | const auto rate = instanced ? vk::VertexInputRate::eInstance : vk::VertexInputRate::eVertex; | 171 | const auto rate = instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX; |
| 150 | vertex_bindings.emplace_back(binding.index, binding.stride, rate); | 172 | |
| 173 | auto& vertex_binding = vertex_bindings.emplace_back(); | ||
| 174 | vertex_binding.binding = binding.index; | ||
| 175 | vertex_binding.stride = binding.stride; | ||
| 176 | vertex_binding.inputRate = rate; | ||
| 177 | |||
| 151 | if (instanced) { | 178 | if (instanced) { |
| 152 | vertex_binding_divisors.emplace_back(binding.index, binding.divisor); | 179 | auto& binding_divisor = vertex_binding_divisors.emplace_back(); |
| 180 | binding_divisor.binding = binding.index; | ||
| 181 | binding_divisor.divisor = binding.divisor; | ||
| 153 | } | 182 | } |
| 154 | } | 183 | } |
| 155 | 184 | ||
| 156 | std::vector<vk::VertexInputAttributeDescription> vertex_attributes; | 185 | std::vector<VkVertexInputAttributeDescription> vertex_attributes; |
| 157 | const auto& input_attributes = program[0]->entries.attributes; | 186 | const auto& input_attributes = program[0]->entries.attributes; |
| 158 | for (std::size_t i = 0; i < vi.num_attributes; ++i) { | 187 | for (std::size_t i = 0; i < vi.num_attributes; ++i) { |
| 159 | const auto& attribute = vi.attributes[i]; | 188 | const auto& attribute = vi.attributes[i]; |
| @@ -161,109 +190,194 @@ UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& render | |||
| 161 | // Skip attributes not used by the vertex shaders. | 190 | // Skip attributes not used by the vertex shaders. |
| 162 | continue; | 191 | continue; |
| 163 | } | 192 | } |
| 164 | vertex_attributes.emplace_back(attribute.index, attribute.buffer, | 193 | auto& vertex_attribute = vertex_attributes.emplace_back(); |
| 165 | MaxwellToVK::VertexFormat(attribute.type, attribute.size), | 194 | vertex_attribute.location = attribute.index; |
| 166 | attribute.offset); | 195 | vertex_attribute.binding = attribute.buffer; |
| 196 | vertex_attribute.format = MaxwellToVK::VertexFormat(attribute.type, attribute.size); | ||
| 197 | vertex_attribute.offset = attribute.offset; | ||
| 167 | } | 198 | } |
| 168 | 199 | ||
| 169 | vk::PipelineVertexInputStateCreateInfo vertex_input_ci( | 200 | VkPipelineVertexInputStateCreateInfo vertex_input_ci; |
| 170 | {}, static_cast<u32>(vertex_bindings.size()), vertex_bindings.data(), | 201 | vertex_input_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; |
| 171 | static_cast<u32>(vertex_attributes.size()), vertex_attributes.data()); | 202 | vertex_input_ci.pNext = nullptr; |
| 172 | 203 | vertex_input_ci.flags = 0; | |
| 173 | const vk::PipelineVertexInputDivisorStateCreateInfoEXT vertex_input_divisor_ci( | 204 | vertex_input_ci.vertexBindingDescriptionCount = static_cast<u32>(vertex_bindings.size()); |
| 174 | static_cast<u32>(vertex_binding_divisors.size()), vertex_binding_divisors.data()); | 205 | vertex_input_ci.pVertexBindingDescriptions = vertex_bindings.data(); |
| 206 | vertex_input_ci.vertexAttributeDescriptionCount = static_cast<u32>(vertex_attributes.size()); | ||
| 207 | vertex_input_ci.pVertexAttributeDescriptions = vertex_attributes.data(); | ||
| 208 | |||
| 209 | VkPipelineVertexInputDivisorStateCreateInfoEXT input_divisor_ci; | ||
| 210 | input_divisor_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT; | ||
| 211 | input_divisor_ci.pNext = nullptr; | ||
| 212 | input_divisor_ci.vertexBindingDivisorCount = static_cast<u32>(vertex_binding_divisors.size()); | ||
| 213 | input_divisor_ci.pVertexBindingDivisors = vertex_binding_divisors.data(); | ||
| 175 | if (!vertex_binding_divisors.empty()) { | 214 | if (!vertex_binding_divisors.empty()) { |
| 176 | vertex_input_ci.pNext = &vertex_input_divisor_ci; | 215 | vertex_input_ci.pNext = &input_divisor_ci; |
| 177 | } | 216 | } |
| 178 | 217 | ||
| 179 | const auto primitive_topology = MaxwellToVK::PrimitiveTopology(device, ia.topology); | 218 | VkPipelineInputAssemblyStateCreateInfo input_assembly_ci; |
| 180 | const vk::PipelineInputAssemblyStateCreateInfo input_assembly_ci( | 219 | input_assembly_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO; |
| 181 | {}, primitive_topology, | 220 | input_assembly_ci.pNext = nullptr; |
| 182 | ia.primitive_restart_enable && SupportsPrimitiveRestart(primitive_topology)); | 221 | input_assembly_ci.flags = 0; |
| 183 | 222 | input_assembly_ci.topology = MaxwellToVK::PrimitiveTopology(device, ia.topology); | |
| 184 | const vk::PipelineTessellationStateCreateInfo tessellation_ci({}, ts.patch_control_points); | 223 | input_assembly_ci.primitiveRestartEnable = |
| 185 | 224 | ia.primitive_restart_enable && SupportsPrimitiveRestart(input_assembly_ci.topology); | |
| 186 | const vk::PipelineViewportStateCreateInfo viewport_ci({}, Maxwell::NumViewports, nullptr, | 225 | |
| 187 | Maxwell::NumViewports, nullptr); | 226 | VkPipelineTessellationStateCreateInfo tessellation_ci; |
| 188 | 227 | tessellation_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO; | |
| 189 | // TODO(Rodrigo): Find out what's the default register value for front face | 228 | tessellation_ci.pNext = nullptr; |
| 190 | const vk::PipelineRasterizationStateCreateInfo rasterizer_ci( | 229 | tessellation_ci.flags = 0; |
| 191 | {}, rs.depth_clamp_enable, false, vk::PolygonMode::eFill, | 230 | tessellation_ci.patchControlPoints = ts.patch_control_points; |
| 192 | rs.cull_enable ? MaxwellToVK::CullFace(rs.cull_face) : vk::CullModeFlagBits::eNone, | 231 | |
| 193 | MaxwellToVK::FrontFace(rs.front_face), rs.depth_bias_enable, 0.0f, 0.0f, 0.0f, 1.0f); | 232 | VkPipelineViewportStateCreateInfo viewport_ci; |
| 194 | 233 | viewport_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; | |
| 195 | const vk::PipelineMultisampleStateCreateInfo multisampling_ci( | 234 | viewport_ci.pNext = nullptr; |
| 196 | {}, vk::SampleCountFlagBits::e1, false, 0.0f, nullptr, false, false); | 235 | viewport_ci.flags = 0; |
| 197 | 236 | viewport_ci.viewportCount = Maxwell::NumViewports; | |
| 198 | const vk::CompareOp depth_test_compare = ds.depth_test_enable | 237 | viewport_ci.pViewports = nullptr; |
| 199 | ? MaxwellToVK::ComparisonOp(ds.depth_test_function) | 238 | viewport_ci.scissorCount = Maxwell::NumViewports; |
| 200 | : vk::CompareOp::eAlways; | 239 | viewport_ci.pScissors = nullptr; |
| 201 | 240 | ||
| 202 | const vk::PipelineDepthStencilStateCreateInfo depth_stencil_ci( | 241 | VkPipelineRasterizationStateCreateInfo rasterization_ci; |
| 203 | {}, ds.depth_test_enable, ds.depth_write_enable, depth_test_compare, ds.depth_bounds_enable, | 242 | rasterization_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; |
| 204 | ds.stencil_enable, GetStencilFaceState(ds.front_stencil), | 243 | rasterization_ci.pNext = nullptr; |
| 205 | GetStencilFaceState(ds.back_stencil), 0.0f, 0.0f); | 244 | rasterization_ci.flags = 0; |
| 206 | 245 | rasterization_ci.depthClampEnable = rs.depth_clamp_enable; | |
| 207 | std::array<vk::PipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments; | 246 | rasterization_ci.rasterizerDiscardEnable = VK_FALSE; |
| 247 | rasterization_ci.polygonMode = VK_POLYGON_MODE_FILL; | ||
| 248 | rasterization_ci.cullMode = | ||
| 249 | rs.cull_enable ? MaxwellToVK::CullFace(rs.cull_face) : VK_CULL_MODE_NONE; | ||
| 250 | rasterization_ci.frontFace = MaxwellToVK::FrontFace(rs.front_face); | ||
| 251 | rasterization_ci.depthBiasEnable = rs.depth_bias_enable; | ||
| 252 | rasterization_ci.depthBiasConstantFactor = 0.0f; | ||
| 253 | rasterization_ci.depthBiasClamp = 0.0f; | ||
| 254 | rasterization_ci.depthBiasSlopeFactor = 0.0f; | ||
| 255 | rasterization_ci.lineWidth = 1.0f; | ||
| 256 | |||
| 257 | VkPipelineMultisampleStateCreateInfo multisample_ci; | ||
| 258 | multisample_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; | ||
| 259 | multisample_ci.pNext = nullptr; | ||
| 260 | multisample_ci.flags = 0; | ||
| 261 | multisample_ci.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT; | ||
| 262 | multisample_ci.sampleShadingEnable = VK_FALSE; | ||
| 263 | multisample_ci.minSampleShading = 0.0f; | ||
| 264 | multisample_ci.pSampleMask = nullptr; | ||
| 265 | multisample_ci.alphaToCoverageEnable = VK_FALSE; | ||
| 266 | multisample_ci.alphaToOneEnable = VK_FALSE; | ||
| 267 | |||
| 268 | VkPipelineDepthStencilStateCreateInfo depth_stencil_ci; | ||
| 269 | depth_stencil_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO; | ||
| 270 | depth_stencil_ci.pNext = nullptr; | ||
| 271 | depth_stencil_ci.flags = 0; | ||
| 272 | depth_stencil_ci.depthTestEnable = ds.depth_test_enable; | ||
| 273 | depth_stencil_ci.depthWriteEnable = ds.depth_write_enable; | ||
| 274 | depth_stencil_ci.depthCompareOp = ds.depth_test_enable | ||
| 275 | ? MaxwellToVK::ComparisonOp(ds.depth_test_function) | ||
| 276 | : VK_COMPARE_OP_ALWAYS; | ||
| 277 | depth_stencil_ci.depthBoundsTestEnable = ds.depth_bounds_enable; | ||
| 278 | depth_stencil_ci.stencilTestEnable = ds.stencil_enable; | ||
| 279 | depth_stencil_ci.front = GetStencilFaceState(ds.front_stencil); | ||
| 280 | depth_stencil_ci.back = GetStencilFaceState(ds.back_stencil); | ||
| 281 | depth_stencil_ci.minDepthBounds = 0.0f; | ||
| 282 | depth_stencil_ci.maxDepthBounds = 0.0f; | ||
| 283 | |||
| 284 | std::array<VkPipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments; | ||
| 208 | const std::size_t num_attachments = | 285 | const std::size_t num_attachments = |
| 209 | std::min(cd.attachments_count, renderpass_params.color_attachments.size()); | 286 | std::min(cd.attachments_count, renderpass_params.color_attachments.size()); |
| 210 | for (std::size_t i = 0; i < num_attachments; ++i) { | 287 | for (std::size_t i = 0; i < num_attachments; ++i) { |
| 211 | constexpr std::array component_table{ | 288 | static constexpr std::array component_table = { |
| 212 | vk::ColorComponentFlagBits::eR, vk::ColorComponentFlagBits::eG, | 289 | VK_COLOR_COMPONENT_R_BIT, VK_COLOR_COMPONENT_G_BIT, VK_COLOR_COMPONENT_B_BIT, |
| 213 | vk::ColorComponentFlagBits::eB, vk::ColorComponentFlagBits::eA}; | 290 | VK_COLOR_COMPONENT_A_BIT}; |
| 214 | const auto& blend = cd.attachments[i]; | 291 | const auto& blend = cd.attachments[i]; |
| 215 | 292 | ||
| 216 | vk::ColorComponentFlags color_components{}; | 293 | VkColorComponentFlags color_components = 0; |
| 217 | for (std::size_t j = 0; j < component_table.size(); ++j) { | 294 | for (std::size_t j = 0; j < component_table.size(); ++j) { |
| 218 | if (blend.components[j]) | 295 | if (blend.components[j]) { |
| 219 | color_components |= component_table[j]; | 296 | color_components |= component_table[j]; |
| 297 | } | ||
| 220 | } | 298 | } |
| 221 | 299 | ||
| 222 | cb_attachments[i] = vk::PipelineColorBlendAttachmentState( | 300 | VkPipelineColorBlendAttachmentState& attachment = cb_attachments[i]; |
| 223 | blend.enable, MaxwellToVK::BlendFactor(blend.src_rgb_func), | 301 | attachment.blendEnable = blend.enable; |
| 224 | MaxwellToVK::BlendFactor(blend.dst_rgb_func), | 302 | attachment.srcColorBlendFactor = MaxwellToVK::BlendFactor(blend.src_rgb_func); |
| 225 | MaxwellToVK::BlendEquation(blend.rgb_equation), | 303 | attachment.dstColorBlendFactor = MaxwellToVK::BlendFactor(blend.dst_rgb_func); |
| 226 | MaxwellToVK::BlendFactor(blend.src_a_func), MaxwellToVK::BlendFactor(blend.dst_a_func), | 304 | attachment.colorBlendOp = MaxwellToVK::BlendEquation(blend.rgb_equation); |
| 227 | MaxwellToVK::BlendEquation(blend.a_equation), color_components); | 305 | attachment.srcAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.src_a_func); |
| 306 | attachment.dstAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.dst_a_func); | ||
| 307 | attachment.alphaBlendOp = MaxwellToVK::BlendEquation(blend.a_equation); | ||
| 308 | attachment.colorWriteMask = color_components; | ||
| 228 | } | 309 | } |
| 229 | const vk::PipelineColorBlendStateCreateInfo color_blending_ci({}, false, vk::LogicOp::eCopy, | 310 | |
| 230 | static_cast<u32>(num_attachments), | 311 | VkPipelineColorBlendStateCreateInfo color_blend_ci; |
| 231 | cb_attachments.data(), {}); | 312 | color_blend_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; |
| 232 | 313 | color_blend_ci.pNext = nullptr; | |
| 233 | constexpr std::array dynamic_states = { | 314 | color_blend_ci.flags = 0; |
| 234 | vk::DynamicState::eViewport, vk::DynamicState::eScissor, | 315 | color_blend_ci.logicOpEnable = VK_FALSE; |
| 235 | vk::DynamicState::eDepthBias, vk::DynamicState::eBlendConstants, | 316 | color_blend_ci.logicOp = VK_LOGIC_OP_COPY; |
| 236 | vk::DynamicState::eDepthBounds, vk::DynamicState::eStencilCompareMask, | 317 | color_blend_ci.attachmentCount = static_cast<u32>(num_attachments); |
| 237 | vk::DynamicState::eStencilWriteMask, vk::DynamicState::eStencilReference}; | 318 | color_blend_ci.pAttachments = cb_attachments.data(); |
| 238 | const vk::PipelineDynamicStateCreateInfo dynamic_state_ci( | 319 | std::memset(color_blend_ci.blendConstants, 0, sizeof(color_blend_ci.blendConstants)); |
| 239 | {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data()); | 320 | |
| 240 | 321 | static constexpr std::array dynamic_states = { | |
| 241 | vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; | 322 | VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR, |
| 323 | VK_DYNAMIC_STATE_DEPTH_BIAS, VK_DYNAMIC_STATE_BLEND_CONSTANTS, | ||
| 324 | VK_DYNAMIC_STATE_DEPTH_BOUNDS, VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK, | ||
| 325 | VK_DYNAMIC_STATE_STENCIL_WRITE_MASK, VK_DYNAMIC_STATE_STENCIL_REFERENCE}; | ||
| 326 | |||
| 327 | VkPipelineDynamicStateCreateInfo dynamic_state_ci; | ||
| 328 | dynamic_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO; | ||
| 329 | dynamic_state_ci.pNext = nullptr; | ||
| 330 | dynamic_state_ci.flags = 0; | ||
| 331 | dynamic_state_ci.dynamicStateCount = static_cast<u32>(dynamic_states.size()); | ||
| 332 | dynamic_state_ci.pDynamicStates = dynamic_states.data(); | ||
| 333 | |||
| 334 | VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; | ||
| 335 | subgroup_size_ci.sType = | ||
| 336 | VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT; | ||
| 337 | subgroup_size_ci.pNext = nullptr; | ||
| 242 | subgroup_size_ci.requiredSubgroupSize = GuestWarpSize; | 338 | subgroup_size_ci.requiredSubgroupSize = GuestWarpSize; |
| 243 | 339 | ||
| 244 | std::vector<vk::PipelineShaderStageCreateInfo> shader_stages; | 340 | std::vector<VkPipelineShaderStageCreateInfo> shader_stages; |
| 245 | std::size_t module_index = 0; | 341 | std::size_t module_index = 0; |
| 246 | for (std::size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) { | 342 | for (std::size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) { |
| 247 | if (!program[stage]) { | 343 | if (!program[stage]) { |
| 248 | continue; | 344 | continue; |
| 249 | } | 345 | } |
| 250 | const auto stage_enum = static_cast<Tegra::Engines::ShaderType>(stage); | 346 | VkPipelineShaderStageCreateInfo& stage_ci = shader_stages.emplace_back(); |
| 251 | const auto vk_stage = MaxwellToVK::ShaderStage(stage_enum); | 347 | stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; |
| 252 | auto& stage_ci = shader_stages.emplace_back(vk::PipelineShaderStageCreateFlags{}, vk_stage, | 348 | stage_ci.pNext = nullptr; |
| 253 | *modules[module_index++], "main", nullptr); | 349 | stage_ci.flags = 0; |
| 254 | if (program[stage]->entries.uses_warps && device.IsGuestWarpSizeSupported(vk_stage)) { | 350 | stage_ci.stage = MaxwellToVK::ShaderStage(static_cast<Tegra::Engines::ShaderType>(stage)); |
| 351 | stage_ci.module = *modules[module_index++]; | ||
| 352 | stage_ci.pName = "main"; | ||
| 353 | stage_ci.pSpecializationInfo = nullptr; | ||
| 354 | |||
| 355 | if (program[stage]->entries.uses_warps && device.IsGuestWarpSizeSupported(stage_ci.stage)) { | ||
| 255 | stage_ci.pNext = &subgroup_size_ci; | 356 | stage_ci.pNext = &subgroup_size_ci; |
| 256 | } | 357 | } |
| 257 | } | 358 | } |
| 258 | 359 | ||
| 259 | const vk::GraphicsPipelineCreateInfo create_info( | 360 | VkGraphicsPipelineCreateInfo ci; |
| 260 | {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input_ci, | 361 | ci.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; |
| 261 | &input_assembly_ci, &tessellation_ci, &viewport_ci, &rasterizer_ci, &multisampling_ci, | 362 | ci.pNext = nullptr; |
| 262 | &depth_stencil_ci, &color_blending_ci, &dynamic_state_ci, *layout, renderpass, 0, {}, 0); | 363 | ci.flags = 0; |
| 263 | 364 | ci.stageCount = static_cast<u32>(shader_stages.size()); | |
| 264 | const auto dev = device.GetLogical(); | 365 | ci.pStages = shader_stages.data(); |
| 265 | const auto& dld = device.GetDispatchLoader(); | 366 | ci.pVertexInputState = &vertex_input_ci; |
| 266 | return dev.createGraphicsPipelineUnique(nullptr, create_info, nullptr, dld); | 367 | ci.pInputAssemblyState = &input_assembly_ci; |
| 368 | ci.pTessellationState = &tessellation_ci; | ||
| 369 | ci.pViewportState = &viewport_ci; | ||
| 370 | ci.pRasterizationState = &rasterization_ci; | ||
| 371 | ci.pMultisampleState = &multisample_ci; | ||
| 372 | ci.pDepthStencilState = &depth_stencil_ci; | ||
| 373 | ci.pColorBlendState = &color_blend_ci; | ||
| 374 | ci.pDynamicState = &dynamic_state_ci; | ||
| 375 | ci.layout = *layout; | ||
| 376 | ci.renderPass = renderpass; | ||
| 377 | ci.subpass = 0; | ||
| 378 | ci.basePipelineHandle = nullptr; | ||
| 379 | ci.basePipelineIndex = 0; | ||
| 380 | return device.GetLogical().CreateGraphicsPipeline(ci); | ||
| 267 | } | 381 | } |
| 268 | 382 | ||
| 269 | } // namespace Vulkan | 383 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h index 4f5e4ea2d..7aba70960 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h | |||
| @@ -11,12 +11,12 @@ | |||
| 11 | #include <vector> | 11 | #include <vector> |
| 12 | 12 | ||
| 13 | #include "video_core/engines/maxwell_3d.h" | 13 | #include "video_core/engines/maxwell_3d.h" |
| 14 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 15 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" | 14 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" |
| 16 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" | 15 | #include "video_core/renderer_vulkan/vk_descriptor_pool.h" |
| 17 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" | 16 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" |
| 18 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 17 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 19 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" | 18 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" |
| 19 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 20 | 20 | ||
| 21 | namespace Vulkan { | 21 | namespace Vulkan { |
| 22 | 22 | ||
| @@ -39,52 +39,52 @@ public: | |||
| 39 | VKUpdateDescriptorQueue& update_descriptor_queue, | 39 | VKUpdateDescriptorQueue& update_descriptor_queue, |
| 40 | VKRenderPassCache& renderpass_cache, | 40 | VKRenderPassCache& renderpass_cache, |
| 41 | const GraphicsPipelineCacheKey& key, | 41 | const GraphicsPipelineCacheKey& key, |
| 42 | const std::vector<vk::DescriptorSetLayoutBinding>& bindings, | 42 | vk::Span<VkDescriptorSetLayoutBinding> bindings, |
| 43 | const SPIRVProgram& program); | 43 | const SPIRVProgram& program); |
| 44 | ~VKGraphicsPipeline(); | 44 | ~VKGraphicsPipeline(); |
| 45 | 45 | ||
| 46 | vk::DescriptorSet CommitDescriptorSet(); | 46 | VkDescriptorSet CommitDescriptorSet(); |
| 47 | 47 | ||
| 48 | vk::Pipeline GetHandle() const { | 48 | VkPipeline GetHandle() const { |
| 49 | return *pipeline; | 49 | return *pipeline; |
| 50 | } | 50 | } |
| 51 | 51 | ||
| 52 | vk::PipelineLayout GetLayout() const { | 52 | VkPipelineLayout GetLayout() const { |
| 53 | return *layout; | 53 | return *layout; |
| 54 | } | 54 | } |
| 55 | 55 | ||
| 56 | vk::RenderPass GetRenderPass() const { | 56 | VkRenderPass GetRenderPass() const { |
| 57 | return renderpass; | 57 | return renderpass; |
| 58 | } | 58 | } |
| 59 | 59 | ||
| 60 | private: | 60 | private: |
| 61 | UniqueDescriptorSetLayout CreateDescriptorSetLayout( | 61 | vk::DescriptorSetLayout CreateDescriptorSetLayout( |
| 62 | const std::vector<vk::DescriptorSetLayoutBinding>& bindings) const; | 62 | vk::Span<VkDescriptorSetLayoutBinding> bindings) const; |
| 63 | 63 | ||
| 64 | UniquePipelineLayout CreatePipelineLayout() const; | 64 | vk::PipelineLayout CreatePipelineLayout() const; |
| 65 | 65 | ||
| 66 | UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate( | 66 | vk::DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplate( |
| 67 | const SPIRVProgram& program) const; | 67 | const SPIRVProgram& program) const; |
| 68 | 68 | ||
| 69 | std::vector<UniqueShaderModule> CreateShaderModules(const SPIRVProgram& program) const; | 69 | std::vector<vk::ShaderModule> CreateShaderModules(const SPIRVProgram& program) const; |
| 70 | 70 | ||
| 71 | UniquePipeline CreatePipeline(const RenderPassParams& renderpass_params, | 71 | vk::Pipeline CreatePipeline(const RenderPassParams& renderpass_params, |
| 72 | const SPIRVProgram& program) const; | 72 | const SPIRVProgram& program) const; |
| 73 | 73 | ||
| 74 | const VKDevice& device; | 74 | const VKDevice& device; |
| 75 | VKScheduler& scheduler; | 75 | VKScheduler& scheduler; |
| 76 | const FixedPipelineState fixed_state; | 76 | const FixedPipelineState fixed_state; |
| 77 | const u64 hash; | 77 | const u64 hash; |
| 78 | 78 | ||
| 79 | UniqueDescriptorSetLayout descriptor_set_layout; | 79 | vk::DescriptorSetLayout descriptor_set_layout; |
| 80 | DescriptorAllocator descriptor_allocator; | 80 | DescriptorAllocator descriptor_allocator; |
| 81 | VKUpdateDescriptorQueue& update_descriptor_queue; | 81 | VKUpdateDescriptorQueue& update_descriptor_queue; |
| 82 | UniquePipelineLayout layout; | 82 | vk::PipelineLayout layout; |
| 83 | UniqueDescriptorUpdateTemplate descriptor_template; | 83 | vk::DescriptorUpdateTemplateKHR descriptor_template; |
| 84 | std::vector<UniqueShaderModule> modules; | 84 | std::vector<vk::ShaderModule> modules; |
| 85 | 85 | ||
| 86 | vk::RenderPass renderpass; | 86 | VkRenderPass renderpass; |
| 87 | UniquePipeline pipeline; | 87 | vk::Pipeline pipeline; |
| 88 | }; | 88 | }; |
| 89 | 89 | ||
| 90 | } // namespace Vulkan | 90 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_image.cpp b/src/video_core/renderer_vulkan/vk_image.cpp index 4bcbef959..9bceb3861 100644 --- a/src/video_core/renderer_vulkan/vk_image.cpp +++ b/src/video_core/renderer_vulkan/vk_image.cpp | |||
| @@ -6,22 +6,21 @@ | |||
| 6 | #include <vector> | 6 | #include <vector> |
| 7 | 7 | ||
| 8 | #include "common/assert.h" | 8 | #include "common/assert.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/vk_device.h" | 9 | #include "video_core/renderer_vulkan/vk_device.h" |
| 11 | #include "video_core/renderer_vulkan/vk_image.h" | 10 | #include "video_core/renderer_vulkan/vk_image.h" |
| 12 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 11 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| 16 | VKImage::VKImage(const VKDevice& device, VKScheduler& scheduler, | 16 | VKImage::VKImage(const VKDevice& device, VKScheduler& scheduler, const VkImageCreateInfo& image_ci, |
| 17 | const vk::ImageCreateInfo& image_ci, vk::ImageAspectFlags aspect_mask) | 17 | VkImageAspectFlags aspect_mask) |
| 18 | : device{device}, scheduler{scheduler}, format{image_ci.format}, aspect_mask{aspect_mask}, | 18 | : device{device}, scheduler{scheduler}, format{image_ci.format}, aspect_mask{aspect_mask}, |
| 19 | image_num_layers{image_ci.arrayLayers}, image_num_levels{image_ci.mipLevels} { | 19 | image_num_layers{image_ci.arrayLayers}, image_num_levels{image_ci.mipLevels} { |
| 20 | UNIMPLEMENTED_IF_MSG(image_ci.queueFamilyIndexCount != 0, | 20 | UNIMPLEMENTED_IF_MSG(image_ci.queueFamilyIndexCount != 0, |
| 21 | "Queue family tracking is not implemented"); | 21 | "Queue family tracking is not implemented"); |
| 22 | 22 | ||
| 23 | const auto dev = device.GetLogical(); | 23 | image = device.GetLogical().CreateImage(image_ci); |
| 24 | image = dev.createImageUnique(image_ci, nullptr, device.GetDispatchLoader()); | ||
| 25 | 24 | ||
| 26 | const u32 num_ranges = image_num_layers * image_num_levels; | 25 | const u32 num_ranges = image_num_layers * image_num_levels; |
| 27 | barriers.resize(num_ranges); | 26 | barriers.resize(num_ranges); |
| @@ -31,8 +30,8 @@ VKImage::VKImage(const VKDevice& device, VKScheduler& scheduler, | |||
| 31 | VKImage::~VKImage() = default; | 30 | VKImage::~VKImage() = default; |
| 32 | 31 | ||
| 33 | void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, | 32 | void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, |
| 34 | vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, | 33 | VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, |
| 35 | vk::ImageLayout new_layout) { | 34 | VkImageLayout new_layout) { |
| 36 | if (!HasChanged(base_layer, num_layers, base_level, num_levels, new_access, new_layout)) { | 35 | if (!HasChanged(base_layer, num_layers, base_level, num_levels, new_access, new_layout)) { |
| 37 | return; | 36 | return; |
| 38 | } | 37 | } |
| @@ -43,9 +42,21 @@ void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num | |||
| 43 | const u32 layer = base_layer + layer_it; | 42 | const u32 layer = base_layer + layer_it; |
| 44 | const u32 level = base_level + level_it; | 43 | const u32 level = base_level + level_it; |
| 45 | auto& state = GetSubrangeState(layer, level); | 44 | auto& state = GetSubrangeState(layer, level); |
| 46 | barriers[cursor] = vk::ImageMemoryBarrier( | 45 | auto& barrier = barriers[cursor]; |
| 47 | state.access, new_access, state.layout, new_layout, VK_QUEUE_FAMILY_IGNORED, | 46 | barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; |
| 48 | VK_QUEUE_FAMILY_IGNORED, *image, {aspect_mask, level, 1, layer, 1}); | 47 | barrier.pNext = nullptr; |
| 48 | barrier.srcAccessMask = state.access; | ||
| 49 | barrier.dstAccessMask = new_access; | ||
| 50 | barrier.oldLayout = state.layout; | ||
| 51 | barrier.newLayout = new_layout; | ||
| 52 | barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 53 | barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; | ||
| 54 | barrier.image = *image; | ||
| 55 | barrier.subresourceRange.aspectMask = aspect_mask; | ||
| 56 | barrier.subresourceRange.baseMipLevel = level; | ||
| 57 | barrier.subresourceRange.levelCount = 1; | ||
| 58 | barrier.subresourceRange.baseArrayLayer = layer; | ||
| 59 | barrier.subresourceRange.layerCount = 1; | ||
| 49 | state.access = new_access; | 60 | state.access = new_access; |
| 50 | state.layout = new_layout; | 61 | state.layout = new_layout; |
| 51 | } | 62 | } |
| @@ -53,16 +64,16 @@ void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num | |||
| 53 | 64 | ||
| 54 | scheduler.RequestOutsideRenderPassOperationContext(); | 65 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 55 | 66 | ||
| 56 | scheduler.Record([barriers = barriers, cursor](auto cmdbuf, auto& dld) { | 67 | scheduler.Record([barriers = barriers, cursor](vk::CommandBuffer cmdbuf) { |
| 57 | // TODO(Rodrigo): Implement a way to use the latest stage across subresources. | 68 | // TODO(Rodrigo): Implement a way to use the latest stage across subresources. |
| 58 | constexpr auto stage_stub = vk::PipelineStageFlagBits::eAllCommands; | 69 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, |
| 59 | cmdbuf.pipelineBarrier(stage_stub, stage_stub, {}, 0, nullptr, 0, nullptr, | 70 | VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, {}, {}, |
| 60 | static_cast<u32>(cursor), barriers.data(), dld); | 71 | vk::Span(barriers.data(), cursor)); |
| 61 | }); | 72 | }); |
| 62 | } | 73 | } |
| 63 | 74 | ||
| 64 | bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, | 75 | bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, |
| 65 | vk::AccessFlags new_access, vk::ImageLayout new_layout) noexcept { | 76 | VkAccessFlags new_access, VkImageLayout new_layout) noexcept { |
| 66 | const bool is_full_range = base_layer == 0 && num_layers == image_num_layers && | 77 | const bool is_full_range = base_layer == 0 && num_layers == image_num_layers && |
| 67 | base_level == 0 && num_levels == image_num_levels; | 78 | base_level == 0 && num_levels == image_num_levels; |
| 68 | if (!is_full_range) { | 79 | if (!is_full_range) { |
| @@ -91,11 +102,21 @@ bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num | |||
| 91 | 102 | ||
| 92 | void VKImage::CreatePresentView() { | 103 | void VKImage::CreatePresentView() { |
| 93 | // Image type has to be 2D to be presented. | 104 | // Image type has to be 2D to be presented. |
| 94 | const vk::ImageViewCreateInfo image_view_ci({}, *image, vk::ImageViewType::e2D, format, {}, | 105 | VkImageViewCreateInfo image_view_ci; |
| 95 | {aspect_mask, 0, 1, 0, 1}); | 106 | image_view_ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; |
| 96 | const auto dev = device.GetLogical(); | 107 | image_view_ci.pNext = nullptr; |
| 97 | const auto& dld = device.GetDispatchLoader(); | 108 | image_view_ci.flags = 0; |
| 98 | present_view = dev.createImageViewUnique(image_view_ci, nullptr, dld); | 109 | image_view_ci.image = *image; |
| 110 | image_view_ci.viewType = VK_IMAGE_VIEW_TYPE_2D; | ||
| 111 | image_view_ci.format = format; | ||
| 112 | image_view_ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY, | ||
| 113 | VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY}; | ||
| 114 | image_view_ci.subresourceRange.aspectMask = aspect_mask; | ||
| 115 | image_view_ci.subresourceRange.baseMipLevel = 0; | ||
| 116 | image_view_ci.subresourceRange.levelCount = 1; | ||
| 117 | image_view_ci.subresourceRange.baseArrayLayer = 0; | ||
| 118 | image_view_ci.subresourceRange.layerCount = 1; | ||
| 119 | present_view = device.GetLogical().CreateImageView(image_view_ci); | ||
| 99 | } | 120 | } |
| 100 | 121 | ||
| 101 | VKImage::SubrangeState& VKImage::GetSubrangeState(u32 layer, u32 level) noexcept { | 122 | VKImage::SubrangeState& VKImage::GetSubrangeState(u32 layer, u32 level) noexcept { |
diff --git a/src/video_core/renderer_vulkan/vk_image.h b/src/video_core/renderer_vulkan/vk_image.h index b78242512..b4d7229e5 100644 --- a/src/video_core/renderer_vulkan/vk_image.h +++ b/src/video_core/renderer_vulkan/vk_image.h | |||
| @@ -8,7 +8,7 @@ | |||
| 8 | #include <vector> | 8 | #include <vector> |
| 9 | 9 | ||
| 10 | #include "common/common_types.h" | 10 | #include "common/common_types.h" |
| 11 | #include "video_core/renderer_vulkan/declarations.h" | 11 | #include "video_core/renderer_vulkan/wrapper.h" |
| 12 | 12 | ||
| 13 | namespace Vulkan { | 13 | namespace Vulkan { |
| 14 | 14 | ||
| @@ -18,16 +18,16 @@ class VKScheduler; | |||
| 18 | class VKImage { | 18 | class VKImage { |
| 19 | public: | 19 | public: |
| 20 | explicit VKImage(const VKDevice& device, VKScheduler& scheduler, | 20 | explicit VKImage(const VKDevice& device, VKScheduler& scheduler, |
| 21 | const vk::ImageCreateInfo& image_ci, vk::ImageAspectFlags aspect_mask); | 21 | const VkImageCreateInfo& image_ci, VkImageAspectFlags aspect_mask); |
| 22 | ~VKImage(); | 22 | ~VKImage(); |
| 23 | 23 | ||
| 24 | /// Records in the passed command buffer an image transition and updates the state of the image. | 24 | /// Records in the passed command buffer an image transition and updates the state of the image. |
| 25 | void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, | 25 | void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, |
| 26 | vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, | 26 | VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, |
| 27 | vk::ImageLayout new_layout); | 27 | VkImageLayout new_layout); |
| 28 | 28 | ||
| 29 | /// Returns a view compatible with presentation, the image has to be 2D. | 29 | /// Returns a view compatible with presentation, the image has to be 2D. |
| 30 | vk::ImageView GetPresentView() { | 30 | VkImageView GetPresentView() { |
| 31 | if (!present_view) { | 31 | if (!present_view) { |
| 32 | CreatePresentView(); | 32 | CreatePresentView(); |
| 33 | } | 33 | } |
| @@ -35,28 +35,28 @@ public: | |||
| 35 | } | 35 | } |
| 36 | 36 | ||
| 37 | /// Returns the Vulkan image handler. | 37 | /// Returns the Vulkan image handler. |
| 38 | vk::Image GetHandle() const { | 38 | const vk::Image& GetHandle() const { |
| 39 | return *image; | 39 | return image; |
| 40 | } | 40 | } |
| 41 | 41 | ||
| 42 | /// Returns the Vulkan format for this image. | 42 | /// Returns the Vulkan format for this image. |
| 43 | vk::Format GetFormat() const { | 43 | VkFormat GetFormat() const { |
| 44 | return format; | 44 | return format; |
| 45 | } | 45 | } |
| 46 | 46 | ||
| 47 | /// Returns the Vulkan aspect mask. | 47 | /// Returns the Vulkan aspect mask. |
| 48 | vk::ImageAspectFlags GetAspectMask() const { | 48 | VkImageAspectFlags GetAspectMask() const { |
| 49 | return aspect_mask; | 49 | return aspect_mask; |
| 50 | } | 50 | } |
| 51 | 51 | ||
| 52 | private: | 52 | private: |
| 53 | struct SubrangeState final { | 53 | struct SubrangeState final { |
| 54 | vk::AccessFlags access{}; ///< Current access bits. | 54 | VkAccessFlags access = 0; ///< Current access bits. |
| 55 | vk::ImageLayout layout = vk::ImageLayout::eUndefined; ///< Current image layout. | 55 | VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED; ///< Current image layout. |
| 56 | }; | 56 | }; |
| 57 | 57 | ||
| 58 | bool HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, | 58 | bool HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, |
| 59 | vk::AccessFlags new_access, vk::ImageLayout new_layout) noexcept; | 59 | VkAccessFlags new_access, VkImageLayout new_layout) noexcept; |
| 60 | 60 | ||
| 61 | /// Creates a presentation view. | 61 | /// Creates a presentation view. |
| 62 | void CreatePresentView(); | 62 | void CreatePresentView(); |
| @@ -67,16 +67,16 @@ private: | |||
| 67 | const VKDevice& device; ///< Device handler. | 67 | const VKDevice& device; ///< Device handler. |
| 68 | VKScheduler& scheduler; ///< Device scheduler. | 68 | VKScheduler& scheduler; ///< Device scheduler. |
| 69 | 69 | ||
| 70 | const vk::Format format; ///< Vulkan format. | 70 | const VkFormat format; ///< Vulkan format. |
| 71 | const vk::ImageAspectFlags aspect_mask; ///< Vulkan aspect mask. | 71 | const VkImageAspectFlags aspect_mask; ///< Vulkan aspect mask. |
| 72 | const u32 image_num_layers; ///< Number of layers. | 72 | const u32 image_num_layers; ///< Number of layers. |
| 73 | const u32 image_num_levels; ///< Number of mipmap levels. | 73 | const u32 image_num_levels; ///< Number of mipmap levels. |
| 74 | 74 | ||
| 75 | UniqueImage image; ///< Image handle. | 75 | vk::Image image; ///< Image handle. |
| 76 | UniqueImageView present_view; ///< Image view compatible with presentation. | 76 | vk::ImageView present_view; ///< Image view compatible with presentation. |
| 77 | 77 | ||
| 78 | std::vector<vk::ImageMemoryBarrier> barriers; ///< Pool of barriers. | 78 | std::vector<VkImageMemoryBarrier> barriers; ///< Pool of barriers. |
| 79 | std::vector<SubrangeState> subrange_states; ///< Current subrange state. | 79 | std::vector<SubrangeState> subrange_states; ///< Current subrange state. |
| 80 | 80 | ||
| 81 | bool state_diverged = false; ///< True when subresources mismatch in layout. | 81 | bool state_diverged = false; ///< True when subresources mismatch in layout. |
| 82 | }; | 82 | }; |
diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.cpp b/src/video_core/renderer_vulkan/vk_memory_manager.cpp index 9cc9979d0..6a9e658bf 100644 --- a/src/video_core/renderer_vulkan/vk_memory_manager.cpp +++ b/src/video_core/renderer_vulkan/vk_memory_manager.cpp | |||
| @@ -11,9 +11,9 @@ | |||
| 11 | #include "common/assert.h" | 11 | #include "common/assert.h" |
| 12 | #include "common/common_types.h" | 12 | #include "common/common_types.h" |
| 13 | #include "common/logging/log.h" | 13 | #include "common/logging/log.h" |
| 14 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 15 | #include "video_core/renderer_vulkan/vk_device.h" | 14 | #include "video_core/renderer_vulkan/vk_device.h" |
| 16 | #include "video_core/renderer_vulkan/vk_memory_manager.h" | 15 | #include "video_core/renderer_vulkan/vk_memory_manager.h" |
| 16 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 17 | 17 | ||
| 18 | namespace Vulkan { | 18 | namespace Vulkan { |
| 19 | 19 | ||
| @@ -30,17 +30,11 @@ u64 GetAllocationChunkSize(u64 required_size) { | |||
| 30 | class VKMemoryAllocation final { | 30 | class VKMemoryAllocation final { |
| 31 | public: | 31 | public: |
| 32 | explicit VKMemoryAllocation(const VKDevice& device, vk::DeviceMemory memory, | 32 | explicit VKMemoryAllocation(const VKDevice& device, vk::DeviceMemory memory, |
| 33 | vk::MemoryPropertyFlags properties, u64 allocation_size, u32 type) | 33 | VkMemoryPropertyFlags properties, u64 allocation_size, u32 type) |
| 34 | : device{device}, memory{memory}, properties{properties}, allocation_size{allocation_size}, | 34 | : device{device}, memory{std::move(memory)}, properties{properties}, |
| 35 | shifted_type{ShiftType(type)} {} | 35 | allocation_size{allocation_size}, shifted_type{ShiftType(type)} {} |
| 36 | |||
| 37 | ~VKMemoryAllocation() { | ||
| 38 | const auto dev = device.GetLogical(); | ||
| 39 | const auto& dld = device.GetDispatchLoader(); | ||
| 40 | dev.free(memory, nullptr, dld); | ||
| 41 | } | ||
| 42 | 36 | ||
| 43 | VKMemoryCommit Commit(vk::DeviceSize commit_size, vk::DeviceSize alignment) { | 37 | VKMemoryCommit Commit(VkDeviceSize commit_size, VkDeviceSize alignment) { |
| 44 | auto found = TryFindFreeSection(free_iterator, allocation_size, | 38 | auto found = TryFindFreeSection(free_iterator, allocation_size, |
| 45 | static_cast<u64>(commit_size), static_cast<u64>(alignment)); | 39 | static_cast<u64>(commit_size), static_cast<u64>(alignment)); |
| 46 | if (!found) { | 40 | if (!found) { |
| @@ -73,9 +67,8 @@ public: | |||
| 73 | } | 67 | } |
| 74 | 68 | ||
| 75 | /// Returns whether this allocation is compatible with the arguments. | 69 | /// Returns whether this allocation is compatible with the arguments. |
| 76 | bool IsCompatible(vk::MemoryPropertyFlags wanted_properties, u32 type_mask) const { | 70 | bool IsCompatible(VkMemoryPropertyFlags wanted_properties, u32 type_mask) const { |
| 77 | return (wanted_properties & properties) != vk::MemoryPropertyFlagBits(0) && | 71 | return (wanted_properties & properties) && (type_mask & shifted_type) != 0; |
| 78 | (type_mask & shifted_type) != 0; | ||
| 79 | } | 72 | } |
| 80 | 73 | ||
| 81 | private: | 74 | private: |
| @@ -111,11 +104,11 @@ private: | |||
| 111 | return std::nullopt; | 104 | return std::nullopt; |
| 112 | } | 105 | } |
| 113 | 106 | ||
| 114 | const VKDevice& device; ///< Vulkan device. | 107 | const VKDevice& device; ///< Vulkan device. |
| 115 | const vk::DeviceMemory memory; ///< Vulkan memory allocation handler. | 108 | const vk::DeviceMemory memory; ///< Vulkan memory allocation handler. |
| 116 | const vk::MemoryPropertyFlags properties; ///< Vulkan properties. | 109 | const VkMemoryPropertyFlags properties; ///< Vulkan properties. |
| 117 | const u64 allocation_size; ///< Size of this allocation. | 110 | const u64 allocation_size; ///< Size of this allocation. |
| 118 | const u32 shifted_type; ///< Stored Vulkan type of this allocation, shifted. | 111 | const u32 shifted_type; ///< Stored Vulkan type of this allocation, shifted. |
| 119 | 112 | ||
| 120 | /// Hints where the next free region is likely going to be. | 113 | /// Hints where the next free region is likely going to be. |
| 121 | u64 free_iterator{}; | 114 | u64 free_iterator{}; |
| @@ -125,22 +118,20 @@ private: | |||
| 125 | }; | 118 | }; |
| 126 | 119 | ||
| 127 | VKMemoryManager::VKMemoryManager(const VKDevice& device) | 120 | VKMemoryManager::VKMemoryManager(const VKDevice& device) |
| 128 | : device{device}, properties{device.GetPhysical().getMemoryProperties( | 121 | : device{device}, properties{device.GetPhysical().GetMemoryProperties()}, |
| 129 | device.GetDispatchLoader())}, | ||
| 130 | is_memory_unified{GetMemoryUnified(properties)} {} | 122 | is_memory_unified{GetMemoryUnified(properties)} {} |
| 131 | 123 | ||
| 132 | VKMemoryManager::~VKMemoryManager() = default; | 124 | VKMemoryManager::~VKMemoryManager() = default; |
| 133 | 125 | ||
| 134 | VKMemoryCommit VKMemoryManager::Commit(const vk::MemoryRequirements& requirements, | 126 | VKMemoryCommit VKMemoryManager::Commit(const VkMemoryRequirements& requirements, |
| 135 | bool host_visible) { | 127 | bool host_visible) { |
| 136 | const u64 chunk_size = GetAllocationChunkSize(requirements.size); | 128 | const u64 chunk_size = GetAllocationChunkSize(requirements.size); |
| 137 | 129 | ||
| 138 | // When a host visible commit is asked, search for host visible and coherent, otherwise search | 130 | // When a host visible commit is asked, search for host visible and coherent, otherwise search |
| 139 | // for a fast device local type. | 131 | // for a fast device local type. |
| 140 | const vk::MemoryPropertyFlags wanted_properties = | 132 | const VkMemoryPropertyFlags wanted_properties = |
| 141 | host_visible | 133 | host_visible ? VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
| 142 | ? vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | 134 | : VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT; |
| 143 | : vk::MemoryPropertyFlagBits::eDeviceLocal; | ||
| 144 | 135 | ||
| 145 | if (auto commit = TryAllocCommit(requirements, wanted_properties)) { | 136 | if (auto commit = TryAllocCommit(requirements, wanted_properties)) { |
| 146 | return commit; | 137 | return commit; |
| @@ -161,23 +152,19 @@ VKMemoryCommit VKMemoryManager::Commit(const vk::MemoryRequirements& requirement | |||
| 161 | return commit; | 152 | return commit; |
| 162 | } | 153 | } |
| 163 | 154 | ||
| 164 | VKMemoryCommit VKMemoryManager::Commit(vk::Buffer buffer, bool host_visible) { | 155 | VKMemoryCommit VKMemoryManager::Commit(const vk::Buffer& buffer, bool host_visible) { |
| 165 | const auto dev = device.GetLogical(); | 156 | auto commit = Commit(device.GetLogical().GetBufferMemoryRequirements(*buffer), host_visible); |
| 166 | const auto& dld = device.GetDispatchLoader(); | 157 | buffer.BindMemory(commit->GetMemory(), commit->GetOffset()); |
| 167 | auto commit = Commit(dev.getBufferMemoryRequirements(buffer, dld), host_visible); | ||
| 168 | dev.bindBufferMemory(buffer, commit->GetMemory(), commit->GetOffset(), dld); | ||
| 169 | return commit; | 158 | return commit; |
| 170 | } | 159 | } |
| 171 | 160 | ||
| 172 | VKMemoryCommit VKMemoryManager::Commit(vk::Image image, bool host_visible) { | 161 | VKMemoryCommit VKMemoryManager::Commit(const vk::Image& image, bool host_visible) { |
| 173 | const auto dev = device.GetLogical(); | 162 | auto commit = Commit(device.GetLogical().GetImageMemoryRequirements(*image), host_visible); |
| 174 | const auto& dld = device.GetDispatchLoader(); | 163 | image.BindMemory(commit->GetMemory(), commit->GetOffset()); |
| 175 | auto commit = Commit(dev.getImageMemoryRequirements(image, dld), host_visible); | ||
| 176 | dev.bindImageMemory(image, commit->GetMemory(), commit->GetOffset(), dld); | ||
| 177 | return commit; | 164 | return commit; |
| 178 | } | 165 | } |
| 179 | 166 | ||
| 180 | bool VKMemoryManager::AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 type_mask, | 167 | bool VKMemoryManager::AllocMemory(VkMemoryPropertyFlags wanted_properties, u32 type_mask, |
| 181 | u64 size) { | 168 | u64 size) { |
| 182 | const u32 type = [&] { | 169 | const u32 type = [&] { |
| 183 | for (u32 type_index = 0; type_index < properties.memoryTypeCount; ++type_index) { | 170 | for (u32 type_index = 0; type_index < properties.memoryTypeCount; ++type_index) { |
| @@ -191,24 +178,26 @@ bool VKMemoryManager::AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 | |||
| 191 | return 0U; | 178 | return 0U; |
| 192 | }(); | 179 | }(); |
| 193 | 180 | ||
| 194 | const auto dev = device.GetLogical(); | ||
| 195 | const auto& dld = device.GetDispatchLoader(); | ||
| 196 | |||
| 197 | // Try to allocate found type. | 181 | // Try to allocate found type. |
| 198 | const vk::MemoryAllocateInfo memory_ai(size, type); | 182 | VkMemoryAllocateInfo memory_ai; |
| 199 | vk::DeviceMemory memory; | 183 | memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; |
| 200 | if (const auto res = dev.allocateMemory(&memory_ai, nullptr, &memory, dld); | 184 | memory_ai.pNext = nullptr; |
| 201 | res != vk::Result::eSuccess) { | 185 | memory_ai.allocationSize = size; |
| 202 | LOG_CRITICAL(Render_Vulkan, "Device allocation failed with code {}!", vk::to_string(res)); | 186 | memory_ai.memoryTypeIndex = type; |
| 187 | |||
| 188 | vk::DeviceMemory memory = device.GetLogical().TryAllocateMemory(memory_ai); | ||
| 189 | if (!memory) { | ||
| 190 | LOG_CRITICAL(Render_Vulkan, "Device allocation failed!"); | ||
| 203 | return false; | 191 | return false; |
| 204 | } | 192 | } |
| 205 | allocations.push_back( | 193 | |
| 206 | std::make_unique<VKMemoryAllocation>(device, memory, wanted_properties, size, type)); | 194 | allocations.push_back(std::make_unique<VKMemoryAllocation>(device, std::move(memory), |
| 195 | wanted_properties, size, type)); | ||
| 207 | return true; | 196 | return true; |
| 208 | } | 197 | } |
| 209 | 198 | ||
| 210 | VKMemoryCommit VKMemoryManager::TryAllocCommit(const vk::MemoryRequirements& requirements, | 199 | VKMemoryCommit VKMemoryManager::TryAllocCommit(const VkMemoryRequirements& requirements, |
| 211 | vk::MemoryPropertyFlags wanted_properties) { | 200 | VkMemoryPropertyFlags wanted_properties) { |
| 212 | for (auto& allocation : allocations) { | 201 | for (auto& allocation : allocations) { |
| 213 | if (!allocation->IsCompatible(wanted_properties, requirements.memoryTypeBits)) { | 202 | if (!allocation->IsCompatible(wanted_properties, requirements.memoryTypeBits)) { |
| 214 | continue; | 203 | continue; |
| @@ -220,10 +209,9 @@ VKMemoryCommit VKMemoryManager::TryAllocCommit(const vk::MemoryRequirements& req | |||
| 220 | return {}; | 209 | return {}; |
| 221 | } | 210 | } |
| 222 | 211 | ||
| 223 | /*static*/ bool VKMemoryManager::GetMemoryUnified( | 212 | bool VKMemoryManager::GetMemoryUnified(const VkPhysicalDeviceMemoryProperties& properties) { |
| 224 | const vk::PhysicalDeviceMemoryProperties& properties) { | ||
| 225 | for (u32 heap_index = 0; heap_index < properties.memoryHeapCount; ++heap_index) { | 213 | for (u32 heap_index = 0; heap_index < properties.memoryHeapCount; ++heap_index) { |
| 226 | if (!(properties.memoryHeaps[heap_index].flags & vk::MemoryHeapFlagBits::eDeviceLocal)) { | 214 | if (!(properties.memoryHeaps[heap_index].flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT)) { |
| 227 | // Memory is considered unified when heaps are device local only. | 215 | // Memory is considered unified when heaps are device local only. |
| 228 | return false; | 216 | return false; |
| 229 | } | 217 | } |
| @@ -232,23 +220,19 @@ VKMemoryCommit VKMemoryManager::TryAllocCommit(const vk::MemoryRequirements& req | |||
| 232 | } | 220 | } |
| 233 | 221 | ||
| 234 | VKMemoryCommitImpl::VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation, | 222 | VKMemoryCommitImpl::VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation, |
| 235 | vk::DeviceMemory memory, u64 begin, u64 end) | 223 | const vk::DeviceMemory& memory, u64 begin, u64 end) |
| 236 | : device{device}, interval{begin, end}, memory{memory}, allocation{allocation} {} | 224 | : device{device}, memory{memory}, interval{begin, end}, allocation{allocation} {} |
| 237 | 225 | ||
| 238 | VKMemoryCommitImpl::~VKMemoryCommitImpl() { | 226 | VKMemoryCommitImpl::~VKMemoryCommitImpl() { |
| 239 | allocation->Free(this); | 227 | allocation->Free(this); |
| 240 | } | 228 | } |
| 241 | 229 | ||
| 242 | MemoryMap VKMemoryCommitImpl::Map(u64 size, u64 offset_) const { | 230 | MemoryMap VKMemoryCommitImpl::Map(u64 size, u64 offset_) const { |
| 243 | const auto dev = device.GetLogical(); | 231 | return MemoryMap{this, memory.Map(interval.first + offset_, size)}; |
| 244 | const auto address = reinterpret_cast<u8*>( | ||
| 245 | dev.mapMemory(memory, interval.first + offset_, size, {}, device.GetDispatchLoader())); | ||
| 246 | return MemoryMap{this, address}; | ||
| 247 | } | 232 | } |
| 248 | 233 | ||
| 249 | void VKMemoryCommitImpl::Unmap() const { | 234 | void VKMemoryCommitImpl::Unmap() const { |
| 250 | const auto dev = device.GetLogical(); | 235 | memory.Unmap(); |
| 251 | dev.unmapMemory(memory, device.GetDispatchLoader()); | ||
| 252 | } | 236 | } |
| 253 | 237 | ||
| 254 | MemoryMap VKMemoryCommitImpl::Map() const { | 238 | MemoryMap VKMemoryCommitImpl::Map() const { |
diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.h b/src/video_core/renderer_vulkan/vk_memory_manager.h index cd00bb91b..35ee54d30 100644 --- a/src/video_core/renderer_vulkan/vk_memory_manager.h +++ b/src/video_core/renderer_vulkan/vk_memory_manager.h | |||
| @@ -8,7 +8,7 @@ | |||
| 8 | #include <utility> | 8 | #include <utility> |
| 9 | #include <vector> | 9 | #include <vector> |
| 10 | #include "common/common_types.h" | 10 | #include "common/common_types.h" |
| 11 | #include "video_core/renderer_vulkan/declarations.h" | 11 | #include "video_core/renderer_vulkan/wrapper.h" |
| 12 | 12 | ||
| 13 | namespace Vulkan { | 13 | namespace Vulkan { |
| 14 | 14 | ||
| @@ -32,13 +32,13 @@ public: | |||
| 32 | * memory. When passing false, it will try to allocate device local memory. | 32 | * memory. When passing false, it will try to allocate device local memory. |
| 33 | * @returns A memory commit. | 33 | * @returns A memory commit. |
| 34 | */ | 34 | */ |
| 35 | VKMemoryCommit Commit(const vk::MemoryRequirements& reqs, bool host_visible); | 35 | VKMemoryCommit Commit(const VkMemoryRequirements& reqs, bool host_visible); |
| 36 | 36 | ||
| 37 | /// Commits memory required by the buffer and binds it. | 37 | /// Commits memory required by the buffer and binds it. |
| 38 | VKMemoryCommit Commit(vk::Buffer buffer, bool host_visible); | 38 | VKMemoryCommit Commit(const vk::Buffer& buffer, bool host_visible); |
| 39 | 39 | ||
| 40 | /// Commits memory required by the image and binds it. | 40 | /// Commits memory required by the image and binds it. |
| 41 | VKMemoryCommit Commit(vk::Image image, bool host_visible); | 41 | VKMemoryCommit Commit(const vk::Image& image, bool host_visible); |
| 42 | 42 | ||
| 43 | /// Returns true if the memory allocations are done always in host visible and coherent memory. | 43 | /// Returns true if the memory allocations are done always in host visible and coherent memory. |
| 44 | bool IsMemoryUnified() const { | 44 | bool IsMemoryUnified() const { |
| @@ -47,18 +47,18 @@ public: | |||
| 47 | 47 | ||
| 48 | private: | 48 | private: |
| 49 | /// Allocates a chunk of memory. | 49 | /// Allocates a chunk of memory. |
| 50 | bool AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 type_mask, u64 size); | 50 | bool AllocMemory(VkMemoryPropertyFlags wanted_properties, u32 type_mask, u64 size); |
| 51 | 51 | ||
| 52 | /// Tries to allocate a memory commit. | 52 | /// Tries to allocate a memory commit. |
| 53 | VKMemoryCommit TryAllocCommit(const vk::MemoryRequirements& requirements, | 53 | VKMemoryCommit TryAllocCommit(const VkMemoryRequirements& requirements, |
| 54 | vk::MemoryPropertyFlags wanted_properties); | 54 | VkMemoryPropertyFlags wanted_properties); |
| 55 | 55 | ||
| 56 | /// Returns true if the device uses an unified memory model. | 56 | /// Returns true if the device uses an unified memory model. |
| 57 | static bool GetMemoryUnified(const vk::PhysicalDeviceMemoryProperties& properties); | 57 | static bool GetMemoryUnified(const VkPhysicalDeviceMemoryProperties& properties); |
| 58 | 58 | ||
| 59 | const VKDevice& device; ///< Device handler. | 59 | const VKDevice& device; ///< Device handler. |
| 60 | const vk::PhysicalDeviceMemoryProperties properties; ///< Physical device properties. | 60 | const VkPhysicalDeviceMemoryProperties properties; ///< Physical device properties. |
| 61 | const bool is_memory_unified; ///< True if memory model is unified. | 61 | const bool is_memory_unified; ///< True if memory model is unified. |
| 62 | std::vector<std::unique_ptr<VKMemoryAllocation>> allocations; ///< Current allocations. | 62 | std::vector<std::unique_ptr<VKMemoryAllocation>> allocations; ///< Current allocations. |
| 63 | }; | 63 | }; |
| 64 | 64 | ||
| @@ -68,7 +68,7 @@ class VKMemoryCommitImpl final { | |||
| 68 | 68 | ||
| 69 | public: | 69 | public: |
| 70 | explicit VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation, | 70 | explicit VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation, |
| 71 | vk::DeviceMemory memory, u64 begin, u64 end); | 71 | const vk::DeviceMemory& memory, u64 begin, u64 end); |
| 72 | ~VKMemoryCommitImpl(); | 72 | ~VKMemoryCommitImpl(); |
| 73 | 73 | ||
| 74 | /// Maps a memory region and returns a pointer to it. | 74 | /// Maps a memory region and returns a pointer to it. |
| @@ -80,13 +80,13 @@ public: | |||
| 80 | MemoryMap Map() const; | 80 | MemoryMap Map() const; |
| 81 | 81 | ||
| 82 | /// Returns the Vulkan memory handler. | 82 | /// Returns the Vulkan memory handler. |
| 83 | vk::DeviceMemory GetMemory() const { | 83 | VkDeviceMemory GetMemory() const { |
| 84 | return memory; | 84 | return *memory; |
| 85 | } | 85 | } |
| 86 | 86 | ||
| 87 | /// Returns the start position of the commit relative to the allocation. | 87 | /// Returns the start position of the commit relative to the allocation. |
| 88 | vk::DeviceSize GetOffset() const { | 88 | VkDeviceSize GetOffset() const { |
| 89 | return static_cast<vk::DeviceSize>(interval.first); | 89 | return static_cast<VkDeviceSize>(interval.first); |
| 90 | } | 90 | } |
| 91 | 91 | ||
| 92 | private: | 92 | private: |
| @@ -94,8 +94,8 @@ private: | |||
| 94 | void Unmap() const; | 94 | void Unmap() const; |
| 95 | 95 | ||
| 96 | const VKDevice& device; ///< Vulkan device. | 96 | const VKDevice& device; ///< Vulkan device. |
| 97 | const vk::DeviceMemory& memory; ///< Vulkan device memory handler. | ||
| 97 | std::pair<u64, u64> interval{}; ///< Interval where the commit exists. | 98 | std::pair<u64, u64> interval{}; ///< Interval where the commit exists. |
| 98 | vk::DeviceMemory memory; ///< Vulkan device memory handler. | ||
| 99 | VKMemoryAllocation* allocation{}; ///< Pointer to the large memory allocation. | 99 | VKMemoryAllocation* allocation{}; ///< Pointer to the large memory allocation. |
| 100 | }; | 100 | }; |
| 101 | 101 | ||
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index c2a426aeb..90e3a8edd 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp | |||
| @@ -13,7 +13,6 @@ | |||
| 13 | #include "video_core/engines/kepler_compute.h" | 13 | #include "video_core/engines/kepler_compute.h" |
| 14 | #include "video_core/engines/maxwell_3d.h" | 14 | #include "video_core/engines/maxwell_3d.h" |
| 15 | #include "video_core/memory_manager.h" | 15 | #include "video_core/memory_manager.h" |
| 16 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 17 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" | 16 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" |
| 18 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 17 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 19 | #include "video_core/renderer_vulkan/vk_compute_pipeline.h" | 18 | #include "video_core/renderer_vulkan/vk_compute_pipeline.h" |
| @@ -26,6 +25,7 @@ | |||
| 26 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 25 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 27 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 26 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 28 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 27 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 28 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 29 | #include "video_core/shader/compiler_settings.h" | 29 | #include "video_core/shader/compiler_settings.h" |
| 30 | 30 | ||
| 31 | namespace Vulkan { | 31 | namespace Vulkan { |
| @@ -36,12 +36,11 @@ using Tegra::Engines::ShaderType; | |||
| 36 | 36 | ||
| 37 | namespace { | 37 | namespace { |
| 38 | 38 | ||
| 39 | // C++20's using enum | 39 | constexpr VkDescriptorType UNIFORM_BUFFER = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; |
| 40 | constexpr auto eUniformBuffer = vk::DescriptorType::eUniformBuffer; | 40 | constexpr VkDescriptorType STORAGE_BUFFER = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; |
| 41 | constexpr auto eStorageBuffer = vk::DescriptorType::eStorageBuffer; | 41 | constexpr VkDescriptorType UNIFORM_TEXEL_BUFFER = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER; |
| 42 | constexpr auto eUniformTexelBuffer = vk::DescriptorType::eUniformTexelBuffer; | 42 | constexpr VkDescriptorType COMBINED_IMAGE_SAMPLER = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; |
| 43 | constexpr auto eCombinedImageSampler = vk::DescriptorType::eCombinedImageSampler; | 43 | constexpr VkDescriptorType STORAGE_IMAGE = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE; |
| 44 | constexpr auto eStorageImage = vk::DescriptorType::eStorageImage; | ||
| 45 | 44 | ||
| 46 | constexpr VideoCommon::Shader::CompilerSettings compiler_settings{ | 45 | constexpr VideoCommon::Shader::CompilerSettings compiler_settings{ |
| 47 | VideoCommon::Shader::CompileDepth::FullDecompile}; | 46 | VideoCommon::Shader::CompileDepth::FullDecompile}; |
| @@ -126,32 +125,37 @@ ShaderType GetShaderType(Maxwell::ShaderProgram program) { | |||
| 126 | } | 125 | } |
| 127 | } | 126 | } |
| 128 | 127 | ||
| 129 | template <vk::DescriptorType descriptor_type, class Container> | 128 | template <VkDescriptorType descriptor_type, class Container> |
| 130 | void AddBindings(std::vector<vk::DescriptorSetLayoutBinding>& bindings, u32& binding, | 129 | void AddBindings(std::vector<VkDescriptorSetLayoutBinding>& bindings, u32& binding, |
| 131 | vk::ShaderStageFlags stage_flags, const Container& container) { | 130 | VkShaderStageFlags stage_flags, const Container& container) { |
| 132 | const u32 num_entries = static_cast<u32>(std::size(container)); | 131 | const u32 num_entries = static_cast<u32>(std::size(container)); |
| 133 | for (std::size_t i = 0; i < num_entries; ++i) { | 132 | for (std::size_t i = 0; i < num_entries; ++i) { |
| 134 | u32 count = 1; | 133 | u32 count = 1; |
| 135 | if constexpr (descriptor_type == eCombinedImageSampler) { | 134 | if constexpr (descriptor_type == VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER) { |
| 136 | // Combined image samplers can be arrayed. | 135 | // Combined image samplers can be arrayed. |
| 137 | count = container[i].Size(); | 136 | count = container[i].Size(); |
| 138 | } | 137 | } |
| 139 | bindings.emplace_back(binding++, descriptor_type, count, stage_flags, nullptr); | 138 | VkDescriptorSetLayoutBinding& entry = bindings.emplace_back(); |
| 139 | entry.binding = binding++; | ||
| 140 | entry.descriptorType = descriptor_type; | ||
| 141 | entry.descriptorCount = count; | ||
| 142 | entry.stageFlags = stage_flags; | ||
| 143 | entry.pImmutableSamplers = nullptr; | ||
| 140 | } | 144 | } |
| 141 | } | 145 | } |
| 142 | 146 | ||
| 143 | u32 FillDescriptorLayout(const ShaderEntries& entries, | 147 | u32 FillDescriptorLayout(const ShaderEntries& entries, |
| 144 | std::vector<vk::DescriptorSetLayoutBinding>& bindings, | 148 | std::vector<VkDescriptorSetLayoutBinding>& bindings, |
| 145 | Maxwell::ShaderProgram program_type, u32 base_binding) { | 149 | Maxwell::ShaderProgram program_type, u32 base_binding) { |
| 146 | const ShaderType stage = GetStageFromProgram(program_type); | 150 | const ShaderType stage = GetStageFromProgram(program_type); |
| 147 | const vk::ShaderStageFlags flags = MaxwellToVK::ShaderStage(stage); | 151 | const VkShaderStageFlags flags = MaxwellToVK::ShaderStage(stage); |
| 148 | 152 | ||
| 149 | u32 binding = base_binding; | 153 | u32 binding = base_binding; |
| 150 | AddBindings<eUniformBuffer>(bindings, binding, flags, entries.const_buffers); | 154 | AddBindings<UNIFORM_BUFFER>(bindings, binding, flags, entries.const_buffers); |
| 151 | AddBindings<eStorageBuffer>(bindings, binding, flags, entries.global_buffers); | 155 | AddBindings<STORAGE_BUFFER>(bindings, binding, flags, entries.global_buffers); |
| 152 | AddBindings<eUniformTexelBuffer>(bindings, binding, flags, entries.texel_buffers); | 156 | AddBindings<UNIFORM_TEXEL_BUFFER>(bindings, binding, flags, entries.texel_buffers); |
| 153 | AddBindings<eCombinedImageSampler>(bindings, binding, flags, entries.samplers); | 157 | AddBindings<COMBINED_IMAGE_SAMPLER>(bindings, binding, flags, entries.samplers); |
| 154 | AddBindings<eStorageImage>(bindings, binding, flags, entries.images); | 158 | AddBindings<STORAGE_IMAGE>(bindings, binding, flags, entries.images); |
| 155 | return binding; | 159 | return binding; |
| 156 | } | 160 | } |
| 157 | 161 | ||
| @@ -318,7 +322,7 @@ void VKPipelineCache::Unregister(const Shader& shader) { | |||
| 318 | RasterizerCache::Unregister(shader); | 322 | RasterizerCache::Unregister(shader); |
| 319 | } | 323 | } |
| 320 | 324 | ||
| 321 | std::pair<SPIRVProgram, std::vector<vk::DescriptorSetLayoutBinding>> | 325 | std::pair<SPIRVProgram, std::vector<VkDescriptorSetLayoutBinding>> |
| 322 | VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { | 326 | VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { |
| 323 | const auto& fixed_state = key.fixed_state; | 327 | const auto& fixed_state = key.fixed_state; |
| 324 | auto& memory_manager = system.GPU().MemoryManager(); | 328 | auto& memory_manager = system.GPU().MemoryManager(); |
| @@ -335,7 +339,7 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { | |||
| 335 | specialization.ndc_minus_one_to_one = fixed_state.rasterizer.ndc_minus_one_to_one; | 339 | specialization.ndc_minus_one_to_one = fixed_state.rasterizer.ndc_minus_one_to_one; |
| 336 | 340 | ||
| 337 | SPIRVProgram program; | 341 | SPIRVProgram program; |
| 338 | std::vector<vk::DescriptorSetLayoutBinding> bindings; | 342 | std::vector<VkDescriptorSetLayoutBinding> bindings; |
| 339 | 343 | ||
| 340 | for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) { | 344 | for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) { |
| 341 | const auto program_enum = static_cast<Maxwell::ShaderProgram>(index); | 345 | const auto program_enum = static_cast<Maxwell::ShaderProgram>(index); |
| @@ -371,32 +375,49 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { | |||
| 371 | return {std::move(program), std::move(bindings)}; | 375 | return {std::move(program), std::move(bindings)}; |
| 372 | } | 376 | } |
| 373 | 377 | ||
| 374 | template <vk::DescriptorType descriptor_type, class Container> | 378 | template <VkDescriptorType descriptor_type, class Container> |
| 375 | void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, u32& binding, | 379 | void AddEntry(std::vector<VkDescriptorUpdateTemplateEntry>& template_entries, u32& binding, |
| 376 | u32& offset, const Container& container) { | 380 | u32& offset, const Container& container) { |
| 377 | static constexpr u32 entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry)); | 381 | static constexpr u32 entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry)); |
| 378 | const u32 count = static_cast<u32>(std::size(container)); | 382 | const u32 count = static_cast<u32>(std::size(container)); |
| 379 | 383 | ||
| 380 | if constexpr (descriptor_type == eCombinedImageSampler) { | 384 | if constexpr (descriptor_type == COMBINED_IMAGE_SAMPLER) { |
| 381 | for (u32 i = 0; i < count; ++i) { | 385 | for (u32 i = 0; i < count; ++i) { |
| 382 | const u32 num_samplers = container[i].Size(); | 386 | const u32 num_samplers = container[i].Size(); |
| 383 | template_entries.emplace_back(binding, 0, num_samplers, descriptor_type, offset, | 387 | VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back(); |
| 384 | entry_size); | 388 | entry.dstBinding = binding; |
| 389 | entry.dstArrayElement = 0; | ||
| 390 | entry.descriptorCount = num_samplers; | ||
| 391 | entry.descriptorType = descriptor_type; | ||
| 392 | entry.offset = offset; | ||
| 393 | entry.stride = entry_size; | ||
| 394 | |||
| 385 | ++binding; | 395 | ++binding; |
| 386 | offset += num_samplers * entry_size; | 396 | offset += num_samplers * entry_size; |
| 387 | } | 397 | } |
| 388 | return; | 398 | return; |
| 389 | } | 399 | } |
| 390 | 400 | ||
| 391 | if constexpr (descriptor_type == eUniformTexelBuffer) { | 401 | if constexpr (descriptor_type == UNIFORM_TEXEL_BUFFER) { |
| 392 | // Nvidia has a bug where updating multiple uniform texels at once causes the driver to | 402 | // Nvidia has a bug where updating multiple uniform texels at once causes the driver to |
| 393 | // crash. | 403 | // crash. |
| 394 | for (u32 i = 0; i < count; ++i) { | 404 | for (u32 i = 0; i < count; ++i) { |
| 395 | template_entries.emplace_back(binding + i, 0, 1, descriptor_type, | 405 | VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back(); |
| 396 | offset + i * entry_size, entry_size); | 406 | entry.dstBinding = binding + i; |
| 407 | entry.dstArrayElement = 0; | ||
| 408 | entry.descriptorCount = 1; | ||
| 409 | entry.descriptorType = descriptor_type; | ||
| 410 | entry.offset = offset + i * entry_size; | ||
| 411 | entry.stride = entry_size; | ||
| 397 | } | 412 | } |
| 398 | } else if (count > 0) { | 413 | } else if (count > 0) { |
| 399 | template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size); | 414 | VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back(); |
| 415 | entry.dstBinding = binding; | ||
| 416 | entry.dstArrayElement = 0; | ||
| 417 | entry.descriptorCount = count; | ||
| 418 | entry.descriptorType = descriptor_type; | ||
| 419 | entry.offset = offset; | ||
| 420 | entry.stride = entry_size; | ||
| 400 | } | 421 | } |
| 401 | offset += count * entry_size; | 422 | offset += count * entry_size; |
| 402 | binding += count; | 423 | binding += count; |
| @@ -404,12 +425,12 @@ void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, | |||
| 404 | 425 | ||
| 405 | void FillDescriptorUpdateTemplateEntries( | 426 | void FillDescriptorUpdateTemplateEntries( |
| 406 | const ShaderEntries& entries, u32& binding, u32& offset, | 427 | const ShaderEntries& entries, u32& binding, u32& offset, |
| 407 | std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) { | 428 | std::vector<VkDescriptorUpdateTemplateEntryKHR>& template_entries) { |
| 408 | AddEntry<eUniformBuffer>(template_entries, offset, binding, entries.const_buffers); | 429 | AddEntry<UNIFORM_BUFFER>(template_entries, offset, binding, entries.const_buffers); |
| 409 | AddEntry<eStorageBuffer>(template_entries, offset, binding, entries.global_buffers); | 430 | AddEntry<STORAGE_BUFFER>(template_entries, offset, binding, entries.global_buffers); |
| 410 | AddEntry<eUniformTexelBuffer>(template_entries, offset, binding, entries.texel_buffers); | 431 | AddEntry<UNIFORM_TEXEL_BUFFER>(template_entries, offset, binding, entries.texel_buffers); |
| 411 | AddEntry<eCombinedImageSampler>(template_entries, offset, binding, entries.samplers); | 432 | AddEntry<COMBINED_IMAGE_SAMPLER>(template_entries, offset, binding, entries.samplers); |
| 412 | AddEntry<eStorageImage>(template_entries, offset, binding, entries.images); | 433 | AddEntry<STORAGE_IMAGE>(template_entries, offset, binding, entries.images); |
| 413 | } | 434 | } |
| 414 | 435 | ||
| 415 | } // namespace Vulkan | 436 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h index 27c01732f..7ccdb7083 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h | |||
| @@ -19,12 +19,12 @@ | |||
| 19 | #include "video_core/engines/const_buffer_engine_interface.h" | 19 | #include "video_core/engines/const_buffer_engine_interface.h" |
| 20 | #include "video_core/engines/maxwell_3d.h" | 20 | #include "video_core/engines/maxwell_3d.h" |
| 21 | #include "video_core/rasterizer_cache.h" | 21 | #include "video_core/rasterizer_cache.h" |
| 22 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 23 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" | 22 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" |
| 24 | #include "video_core/renderer_vulkan/vk_graphics_pipeline.h" | 23 | #include "video_core/renderer_vulkan/vk_graphics_pipeline.h" |
| 25 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" | 24 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" |
| 26 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 25 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 27 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" | 26 | #include "video_core/renderer_vulkan/vk_shader_decompiler.h" |
| 27 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 28 | #include "video_core/shader/registry.h" | 28 | #include "video_core/shader/registry.h" |
| 29 | #include "video_core/shader/shader_ir.h" | 29 | #include "video_core/shader/shader_ir.h" |
| 30 | #include "video_core/surface.h" | 30 | #include "video_core/surface.h" |
| @@ -172,7 +172,7 @@ protected: | |||
| 172 | void FlushObjectInner(const Shader& object) override {} | 172 | void FlushObjectInner(const Shader& object) override {} |
| 173 | 173 | ||
| 174 | private: | 174 | private: |
| 175 | std::pair<SPIRVProgram, std::vector<vk::DescriptorSetLayoutBinding>> DecompileShaders( | 175 | std::pair<SPIRVProgram, std::vector<VkDescriptorSetLayoutBinding>> DecompileShaders( |
| 176 | const GraphicsPipelineCacheKey& key); | 176 | const GraphicsPipelineCacheKey& key); |
| 177 | 177 | ||
| 178 | Core::System& system; | 178 | Core::System& system; |
| @@ -194,6 +194,6 @@ private: | |||
| 194 | 194 | ||
| 195 | void FillDescriptorUpdateTemplateEntries( | 195 | void FillDescriptorUpdateTemplateEntries( |
| 196 | const ShaderEntries& entries, u32& binding, u32& offset, | 196 | const ShaderEntries& entries, u32& binding, u32& offset, |
| 197 | std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries); | 197 | std::vector<VkDescriptorUpdateTemplateEntryKHR>& template_entries); |
| 198 | 198 | ||
| 199 | } // namespace Vulkan | 199 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_query_cache.cpp b/src/video_core/renderer_vulkan/vk_query_cache.cpp index ffbf60dda..0966c7ff7 100644 --- a/src/video_core/renderer_vulkan/vk_query_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_query_cache.cpp | |||
| @@ -8,19 +8,19 @@ | |||
| 8 | #include <utility> | 8 | #include <utility> |
| 9 | #include <vector> | 9 | #include <vector> |
| 10 | 10 | ||
| 11 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 12 | #include "video_core/renderer_vulkan/vk_device.h" | 11 | #include "video_core/renderer_vulkan/vk_device.h" |
| 13 | #include "video_core/renderer_vulkan/vk_query_cache.h" | 12 | #include "video_core/renderer_vulkan/vk_query_cache.h" |
| 14 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 13 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 15 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 14 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 15 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 16 | 16 | ||
| 17 | namespace Vulkan { | 17 | namespace Vulkan { |
| 18 | 18 | ||
| 19 | namespace { | 19 | namespace { |
| 20 | 20 | ||
| 21 | constexpr std::array QUERY_TARGETS = {vk::QueryType::eOcclusion}; | 21 | constexpr std::array QUERY_TARGETS = {VK_QUERY_TYPE_OCCLUSION}; |
| 22 | 22 | ||
| 23 | constexpr vk::QueryType GetTarget(VideoCore::QueryType type) { | 23 | constexpr VkQueryType GetTarget(VideoCore::QueryType type) { |
| 24 | return QUERY_TARGETS[static_cast<std::size_t>(type)]; | 24 | return QUERY_TARGETS[static_cast<std::size_t>(type)]; |
| 25 | } | 25 | } |
| 26 | 26 | ||
| @@ -35,29 +35,34 @@ void QueryPool::Initialize(const VKDevice& device_, VideoCore::QueryType type_) | |||
| 35 | type = type_; | 35 | type = type_; |
| 36 | } | 36 | } |
| 37 | 37 | ||
| 38 | std::pair<vk::QueryPool, std::uint32_t> QueryPool::Commit(VKFence& fence) { | 38 | std::pair<VkQueryPool, u32> QueryPool::Commit(VKFence& fence) { |
| 39 | std::size_t index; | 39 | std::size_t index; |
| 40 | do { | 40 | do { |
| 41 | index = CommitResource(fence); | 41 | index = CommitResource(fence); |
| 42 | } while (usage[index]); | 42 | } while (usage[index]); |
| 43 | usage[index] = true; | 43 | usage[index] = true; |
| 44 | 44 | ||
| 45 | return {*pools[index / GROW_STEP], static_cast<std::uint32_t>(index % GROW_STEP)}; | 45 | return {*pools[index / GROW_STEP], static_cast<u32>(index % GROW_STEP)}; |
| 46 | } | 46 | } |
| 47 | 47 | ||
| 48 | void QueryPool::Allocate(std::size_t begin, std::size_t end) { | 48 | void QueryPool::Allocate(std::size_t begin, std::size_t end) { |
| 49 | usage.resize(end); | 49 | usage.resize(end); |
| 50 | 50 | ||
| 51 | const auto dev = device->GetLogical(); | 51 | VkQueryPoolCreateInfo query_pool_ci; |
| 52 | const u32 size = static_cast<u32>(end - begin); | 52 | query_pool_ci.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO; |
| 53 | const vk::QueryPoolCreateInfo query_pool_ci({}, GetTarget(type), size, {}); | 53 | query_pool_ci.pNext = nullptr; |
| 54 | pools.push_back(dev.createQueryPoolUnique(query_pool_ci, nullptr, device->GetDispatchLoader())); | 54 | query_pool_ci.flags = 0; |
| 55 | query_pool_ci.queryType = GetTarget(type); | ||
| 56 | query_pool_ci.queryCount = static_cast<u32>(end - begin); | ||
| 57 | query_pool_ci.pipelineStatistics = 0; | ||
| 58 | pools.push_back(device->GetLogical().CreateQueryPool(query_pool_ci)); | ||
| 55 | } | 59 | } |
| 56 | 60 | ||
| 57 | void QueryPool::Reserve(std::pair<vk::QueryPool, std::uint32_t> query) { | 61 | void QueryPool::Reserve(std::pair<VkQueryPool, u32> query) { |
| 58 | const auto it = | 62 | const auto it = |
| 59 | std::find_if(std::begin(pools), std::end(pools), | 63 | std::find_if(pools.begin(), pools.end(), [query_pool = query.first](vk::QueryPool& pool) { |
| 60 | [query_pool = query.first](auto& pool) { return query_pool == *pool; }); | 64 | return query_pool == *pool; |
| 65 | }); | ||
| 61 | ASSERT(it != std::end(pools)); | 66 | ASSERT(it != std::end(pools)); |
| 62 | 67 | ||
| 63 | const std::ptrdiff_t pool_index = std::distance(std::begin(pools), it); | 68 | const std::ptrdiff_t pool_index = std::distance(std::begin(pools), it); |
| @@ -76,12 +81,11 @@ VKQueryCache::VKQueryCache(Core::System& system, VideoCore::RasterizerInterface& | |||
| 76 | 81 | ||
| 77 | VKQueryCache::~VKQueryCache() = default; | 82 | VKQueryCache::~VKQueryCache() = default; |
| 78 | 83 | ||
| 79 | std::pair<vk::QueryPool, std::uint32_t> VKQueryCache::AllocateQuery(VideoCore::QueryType type) { | 84 | std::pair<VkQueryPool, u32> VKQueryCache::AllocateQuery(VideoCore::QueryType type) { |
| 80 | return query_pools[static_cast<std::size_t>(type)].Commit(scheduler.GetFence()); | 85 | return query_pools[static_cast<std::size_t>(type)].Commit(scheduler.GetFence()); |
| 81 | } | 86 | } |
| 82 | 87 | ||
| 83 | void VKQueryCache::Reserve(VideoCore::QueryType type, | 88 | void VKQueryCache::Reserve(VideoCore::QueryType type, std::pair<VkQueryPool, u32> query) { |
| 84 | std::pair<vk::QueryPool, std::uint32_t> query) { | ||
| 85 | query_pools[static_cast<std::size_t>(type)].Reserve(query); | 89 | query_pools[static_cast<std::size_t>(type)].Reserve(query); |
| 86 | } | 90 | } |
| 87 | 91 | ||
| @@ -89,10 +93,10 @@ HostCounter::HostCounter(VKQueryCache& cache, std::shared_ptr<HostCounter> depen | |||
| 89 | VideoCore::QueryType type) | 93 | VideoCore::QueryType type) |
| 90 | : VideoCommon::HostCounterBase<VKQueryCache, HostCounter>{std::move(dependency)}, cache{cache}, | 94 | : VideoCommon::HostCounterBase<VKQueryCache, HostCounter>{std::move(dependency)}, cache{cache}, |
| 91 | type{type}, query{cache.AllocateQuery(type)}, ticks{cache.Scheduler().Ticks()} { | 95 | type{type}, query{cache.AllocateQuery(type)}, ticks{cache.Scheduler().Ticks()} { |
| 92 | const auto dev = cache.Device().GetLogical(); | 96 | const vk::Device* logical = &cache.Device().GetLogical(); |
| 93 | cache.Scheduler().Record([dev, query = query](vk::CommandBuffer cmdbuf, auto& dld) { | 97 | cache.Scheduler().Record([logical, query = query](vk::CommandBuffer cmdbuf) { |
| 94 | dev.resetQueryPoolEXT(query.first, query.second, 1, dld); | 98 | logical->ResetQueryPoolEXT(query.first, query.second, 1); |
| 95 | cmdbuf.beginQuery(query.first, query.second, vk::QueryControlFlagBits::ePrecise, dld); | 99 | cmdbuf.BeginQuery(query.first, query.second, VK_QUERY_CONTROL_PRECISE_BIT); |
| 96 | }); | 100 | }); |
| 97 | } | 101 | } |
| 98 | 102 | ||
| @@ -101,22 +105,16 @@ HostCounter::~HostCounter() { | |||
| 101 | } | 105 | } |
| 102 | 106 | ||
| 103 | void HostCounter::EndQuery() { | 107 | void HostCounter::EndQuery() { |
| 104 | cache.Scheduler().Record([query = query](auto cmdbuf, auto& dld) { | 108 | cache.Scheduler().Record( |
| 105 | cmdbuf.endQuery(query.first, query.second, dld); | 109 | [query = query](vk::CommandBuffer cmdbuf) { cmdbuf.EndQuery(query.first, query.second); }); |
| 106 | }); | ||
| 107 | } | 110 | } |
| 108 | 111 | ||
| 109 | u64 HostCounter::BlockingQuery() const { | 112 | u64 HostCounter::BlockingQuery() const { |
| 110 | if (ticks >= cache.Scheduler().Ticks()) { | 113 | if (ticks >= cache.Scheduler().Ticks()) { |
| 111 | cache.Scheduler().Flush(); | 114 | cache.Scheduler().Flush(); |
| 112 | } | 115 | } |
| 113 | 116 | return cache.Device().GetLogical().GetQueryResult<u64>( | |
| 114 | const auto dev = cache.Device().GetLogical(); | 117 | query.first, query.second, VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); |
| 115 | const auto& dld = cache.Device().GetDispatchLoader(); | ||
| 116 | u64 value; | ||
| 117 | dev.getQueryPoolResults(query.first, query.second, 1, sizeof(value), &value, sizeof(value), | ||
| 118 | vk::QueryResultFlagBits::e64 | vk::QueryResultFlagBits::eWait, dld); | ||
| 119 | return value; | ||
| 120 | } | 118 | } |
| 121 | 119 | ||
| 122 | } // namespace Vulkan | 120 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_query_cache.h b/src/video_core/renderer_vulkan/vk_query_cache.h index c3092ee96..b63784f4b 100644 --- a/src/video_core/renderer_vulkan/vk_query_cache.h +++ b/src/video_core/renderer_vulkan/vk_query_cache.h | |||
| @@ -12,8 +12,8 @@ | |||
| 12 | 12 | ||
| 13 | #include "common/common_types.h" | 13 | #include "common/common_types.h" |
| 14 | #include "video_core/query_cache.h" | 14 | #include "video_core/query_cache.h" |
| 15 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 16 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 15 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 16 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 17 | 17 | ||
| 18 | namespace VideoCore { | 18 | namespace VideoCore { |
| 19 | class RasterizerInterface; | 19 | class RasterizerInterface; |
| @@ -36,9 +36,9 @@ public: | |||
| 36 | 36 | ||
| 37 | void Initialize(const VKDevice& device, VideoCore::QueryType type); | 37 | void Initialize(const VKDevice& device, VideoCore::QueryType type); |
| 38 | 38 | ||
| 39 | std::pair<vk::QueryPool, std::uint32_t> Commit(VKFence& fence); | 39 | std::pair<VkQueryPool, u32> Commit(VKFence& fence); |
| 40 | 40 | ||
| 41 | void Reserve(std::pair<vk::QueryPool, std::uint32_t> query); | 41 | void Reserve(std::pair<VkQueryPool, u32> query); |
| 42 | 42 | ||
| 43 | protected: | 43 | protected: |
| 44 | void Allocate(std::size_t begin, std::size_t end) override; | 44 | void Allocate(std::size_t begin, std::size_t end) override; |
| @@ -49,7 +49,7 @@ private: | |||
| 49 | const VKDevice* device = nullptr; | 49 | const VKDevice* device = nullptr; |
| 50 | VideoCore::QueryType type = {}; | 50 | VideoCore::QueryType type = {}; |
| 51 | 51 | ||
| 52 | std::vector<UniqueQueryPool> pools; | 52 | std::vector<vk::QueryPool> pools; |
| 53 | std::vector<bool> usage; | 53 | std::vector<bool> usage; |
| 54 | }; | 54 | }; |
| 55 | 55 | ||
| @@ -61,9 +61,9 @@ public: | |||
| 61 | const VKDevice& device, VKScheduler& scheduler); | 61 | const VKDevice& device, VKScheduler& scheduler); |
| 62 | ~VKQueryCache(); | 62 | ~VKQueryCache(); |
| 63 | 63 | ||
| 64 | std::pair<vk::QueryPool, std::uint32_t> AllocateQuery(VideoCore::QueryType type); | 64 | std::pair<VkQueryPool, u32> AllocateQuery(VideoCore::QueryType type); |
| 65 | 65 | ||
| 66 | void Reserve(VideoCore::QueryType type, std::pair<vk::QueryPool, std::uint32_t> query); | 66 | void Reserve(VideoCore::QueryType type, std::pair<VkQueryPool, u32> query); |
| 67 | 67 | ||
| 68 | const VKDevice& Device() const noexcept { | 68 | const VKDevice& Device() const noexcept { |
| 69 | return device; | 69 | return device; |
| @@ -91,7 +91,7 @@ private: | |||
| 91 | 91 | ||
| 92 | VKQueryCache& cache; | 92 | VKQueryCache& cache; |
| 93 | const VideoCore::QueryType type; | 93 | const VideoCore::QueryType type; |
| 94 | const std::pair<vk::QueryPool, std::uint32_t> query; | 94 | const std::pair<VkQueryPool, u32> query; |
| 95 | const u64 ticks; | 95 | const u64 ticks; |
| 96 | }; | 96 | }; |
| 97 | 97 | ||
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 6b99cbbbc..33cbc0bb6 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp | |||
| @@ -19,7 +19,6 @@ | |||
| 19 | #include "core/memory.h" | 19 | #include "core/memory.h" |
| 20 | #include "video_core/engines/kepler_compute.h" | 20 | #include "video_core/engines/kepler_compute.h" |
| 21 | #include "video_core/engines/maxwell_3d.h" | 21 | #include "video_core/engines/maxwell_3d.h" |
| 22 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 23 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" | 22 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" |
| 24 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 23 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 25 | #include "video_core/renderer_vulkan/renderer_vulkan.h" | 24 | #include "video_core/renderer_vulkan/renderer_vulkan.h" |
| @@ -39,6 +38,7 @@ | |||
| 39 | #include "video_core/renderer_vulkan/vk_state_tracker.h" | 38 | #include "video_core/renderer_vulkan/vk_state_tracker.h" |
| 40 | #include "video_core/renderer_vulkan/vk_texture_cache.h" | 39 | #include "video_core/renderer_vulkan/vk_texture_cache.h" |
| 41 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 40 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 41 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 42 | 42 | ||
| 43 | namespace Vulkan { | 43 | namespace Vulkan { |
| 44 | 44 | ||
| @@ -60,32 +60,39 @@ namespace { | |||
| 60 | 60 | ||
| 61 | constexpr auto ComputeShaderIndex = static_cast<std::size_t>(Tegra::Engines::ShaderType::Compute); | 61 | constexpr auto ComputeShaderIndex = static_cast<std::size_t>(Tegra::Engines::ShaderType::Compute); |
| 62 | 62 | ||
| 63 | vk::Viewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::size_t index) { | 63 | VkViewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::size_t index) { |
| 64 | const auto& viewport = regs.viewport_transform[index]; | 64 | const auto& src = regs.viewport_transform[index]; |
| 65 | const float x = viewport.translate_x - viewport.scale_x; | 65 | VkViewport viewport; |
| 66 | const float y = viewport.translate_y - viewport.scale_y; | 66 | viewport.x = src.translate_x - src.scale_x; |
| 67 | const float width = viewport.scale_x * 2.0f; | 67 | viewport.y = src.translate_y - src.scale_y; |
| 68 | const float height = viewport.scale_y * 2.0f; | 68 | viewport.width = src.scale_x * 2.0f; |
| 69 | viewport.height = src.scale_y * 2.0f; | ||
| 69 | 70 | ||
| 70 | const float reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne; | 71 | const float reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne; |
| 71 | float near = viewport.translate_z - viewport.scale_z * reduce_z; | 72 | viewport.minDepth = src.translate_z - src.scale_z * reduce_z; |
| 72 | float far = viewport.translate_z + viewport.scale_z; | 73 | viewport.maxDepth = src.translate_z + src.scale_z; |
| 73 | if (!device.IsExtDepthRangeUnrestrictedSupported()) { | 74 | if (!device.IsExtDepthRangeUnrestrictedSupported()) { |
| 74 | near = std::clamp(near, 0.0f, 1.0f); | 75 | viewport.minDepth = std::clamp(viewport.minDepth, 0.0f, 1.0f); |
| 75 | far = std::clamp(far, 0.0f, 1.0f); | 76 | viewport.maxDepth = std::clamp(viewport.maxDepth, 0.0f, 1.0f); |
| 76 | } | 77 | } |
| 77 | 78 | return viewport; | |
| 78 | return vk::Viewport(x, y, width != 0 ? width : 1.0f, height != 0 ? height : 1.0f, near, far); | ||
| 79 | } | 79 | } |
| 80 | 80 | ||
| 81 | constexpr vk::Rect2D GetScissorState(const Maxwell& regs, std::size_t index) { | 81 | VkRect2D GetScissorState(const Maxwell& regs, std::size_t index) { |
| 82 | const auto& scissor = regs.scissor_test[index]; | 82 | const auto& src = regs.scissor_test[index]; |
| 83 | if (!scissor.enable) { | 83 | VkRect2D scissor; |
| 84 | return {{0, 0}, {INT32_MAX, INT32_MAX}}; | 84 | if (src.enable) { |
| 85 | scissor.offset.x = static_cast<s32>(src.min_x); | ||
| 86 | scissor.offset.y = static_cast<s32>(src.min_y); | ||
| 87 | scissor.extent.width = src.max_x - src.min_x; | ||
| 88 | scissor.extent.height = src.max_y - src.min_y; | ||
| 89 | } else { | ||
| 90 | scissor.offset.x = 0; | ||
| 91 | scissor.offset.y = 0; | ||
| 92 | scissor.extent.width = std::numeric_limits<s32>::max(); | ||
| 93 | scissor.extent.height = std::numeric_limits<s32>::max(); | ||
| 85 | } | 94 | } |
| 86 | const u32 width = scissor.max_x - scissor.min_x; | 95 | return scissor; |
| 87 | const u32 height = scissor.max_y - scissor.min_y; | ||
| 88 | return {{static_cast<s32>(scissor.min_x), static_cast<s32>(scissor.min_y)}, {width, height}}; | ||
| 89 | } | 96 | } |
| 90 | 97 | ||
| 91 | std::array<GPUVAddr, Maxwell::MaxShaderProgram> GetShaderAddresses( | 98 | std::array<GPUVAddr, Maxwell::MaxShaderProgram> GetShaderAddresses( |
| @@ -97,8 +104,8 @@ std::array<GPUVAddr, Maxwell::MaxShaderProgram> GetShaderAddresses( | |||
| 97 | return addresses; | 104 | return addresses; |
| 98 | } | 105 | } |
| 99 | 106 | ||
| 100 | void TransitionImages(const std::vector<ImageView>& views, vk::PipelineStageFlags pipeline_stage, | 107 | void TransitionImages(const std::vector<ImageView>& views, VkPipelineStageFlags pipeline_stage, |
| 101 | vk::AccessFlags access) { | 108 | VkAccessFlags access) { |
| 102 | for (auto& [view, layout] : views) { | 109 | for (auto& [view, layout] : views) { |
| 103 | view->Transition(*layout, pipeline_stage, access); | 110 | view->Transition(*layout, pipeline_stage, access); |
| 104 | } | 111 | } |
| @@ -127,13 +134,13 @@ Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry | |||
| 127 | 134 | ||
| 128 | class BufferBindings final { | 135 | class BufferBindings final { |
| 129 | public: | 136 | public: |
| 130 | void AddVertexBinding(const vk::Buffer* buffer, vk::DeviceSize offset) { | 137 | void AddVertexBinding(const VkBuffer* buffer, VkDeviceSize offset) { |
| 131 | vertex.buffer_ptrs[vertex.num_buffers] = buffer; | 138 | vertex.buffer_ptrs[vertex.num_buffers] = buffer; |
| 132 | vertex.offsets[vertex.num_buffers] = offset; | 139 | vertex.offsets[vertex.num_buffers] = offset; |
| 133 | ++vertex.num_buffers; | 140 | ++vertex.num_buffers; |
| 134 | } | 141 | } |
| 135 | 142 | ||
| 136 | void SetIndexBinding(const vk::Buffer* buffer, vk::DeviceSize offset, vk::IndexType type) { | 143 | void SetIndexBinding(const VkBuffer* buffer, VkDeviceSize offset, VkIndexType type) { |
| 137 | index.buffer = buffer; | 144 | index.buffer = buffer; |
| 138 | index.offset = offset; | 145 | index.offset = offset; |
| 139 | index.type = type; | 146 | index.type = type; |
| @@ -217,14 +224,14 @@ private: | |||
| 217 | // Some of these fields are intentionally left uninitialized to avoid initializing them twice. | 224 | // Some of these fields are intentionally left uninitialized to avoid initializing them twice. |
| 218 | struct { | 225 | struct { |
| 219 | std::size_t num_buffers = 0; | 226 | std::size_t num_buffers = 0; |
| 220 | std::array<const vk::Buffer*, Maxwell::NumVertexArrays> buffer_ptrs; | 227 | std::array<const VkBuffer*, Maxwell::NumVertexArrays> buffer_ptrs; |
| 221 | std::array<vk::DeviceSize, Maxwell::NumVertexArrays> offsets; | 228 | std::array<VkDeviceSize, Maxwell::NumVertexArrays> offsets; |
| 222 | } vertex; | 229 | } vertex; |
| 223 | 230 | ||
| 224 | struct { | 231 | struct { |
| 225 | const vk::Buffer* buffer = nullptr; | 232 | const VkBuffer* buffer = nullptr; |
| 226 | vk::DeviceSize offset; | 233 | VkDeviceSize offset; |
| 227 | vk::IndexType type; | 234 | VkIndexType type; |
| 228 | } index; | 235 | } index; |
| 229 | 236 | ||
| 230 | template <std::size_t N> | 237 | template <std::size_t N> |
| @@ -243,38 +250,35 @@ private: | |||
| 243 | return; | 250 | return; |
| 244 | } | 251 | } |
| 245 | 252 | ||
| 246 | std::array<vk::Buffer, N> buffers; | 253 | std::array<VkBuffer, N> buffers; |
| 247 | std::transform(vertex.buffer_ptrs.begin(), vertex.buffer_ptrs.begin() + N, buffers.begin(), | 254 | std::transform(vertex.buffer_ptrs.begin(), vertex.buffer_ptrs.begin() + N, buffers.begin(), |
| 248 | [](const auto ptr) { return *ptr; }); | 255 | [](const auto ptr) { return *ptr; }); |
| 249 | 256 | ||
| 250 | std::array<vk::DeviceSize, N> offsets; | 257 | std::array<VkDeviceSize, N> offsets; |
| 251 | std::copy(vertex.offsets.begin(), vertex.offsets.begin() + N, offsets.begin()); | 258 | std::copy(vertex.offsets.begin(), vertex.offsets.begin() + N, offsets.begin()); |
| 252 | 259 | ||
| 253 | if constexpr (is_indexed) { | 260 | if constexpr (is_indexed) { |
| 254 | // Indexed draw | 261 | // Indexed draw |
| 255 | scheduler.Record([buffers, offsets, index_buffer = *index.buffer, | 262 | scheduler.Record([buffers, offsets, index_buffer = *index.buffer, |
| 256 | index_offset = index.offset, | 263 | index_offset = index.offset, |
| 257 | index_type = index.type](auto cmdbuf, auto& dld) { | 264 | index_type = index.type](vk::CommandBuffer cmdbuf) { |
| 258 | cmdbuf.bindIndexBuffer(index_buffer, index_offset, index_type, dld); | 265 | cmdbuf.BindIndexBuffer(index_buffer, index_offset, index_type); |
| 259 | cmdbuf.bindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data(), | 266 | cmdbuf.BindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data()); |
| 260 | dld); | ||
| 261 | }); | 267 | }); |
| 262 | } else { | 268 | } else { |
| 263 | // Array draw | 269 | // Array draw |
| 264 | scheduler.Record([buffers, offsets](auto cmdbuf, auto& dld) { | 270 | scheduler.Record([buffers, offsets](vk::CommandBuffer cmdbuf) { |
| 265 | cmdbuf.bindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data(), | 271 | cmdbuf.BindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data()); |
| 266 | dld); | ||
| 267 | }); | 272 | }); |
| 268 | } | 273 | } |
| 269 | } | 274 | } |
| 270 | }; | 275 | }; |
| 271 | 276 | ||
| 272 | void RasterizerVulkan::DrawParameters::Draw(vk::CommandBuffer cmdbuf, | 277 | void RasterizerVulkan::DrawParameters::Draw(vk::CommandBuffer cmdbuf) const { |
| 273 | const vk::DispatchLoaderDynamic& dld) const { | ||
| 274 | if (is_indexed) { | 278 | if (is_indexed) { |
| 275 | cmdbuf.drawIndexed(num_vertices, num_instances, 0, base_vertex, base_instance, dld); | 279 | cmdbuf.DrawIndexed(num_vertices, num_instances, 0, base_vertex, base_instance); |
| 276 | } else { | 280 | } else { |
| 277 | cmdbuf.draw(num_vertices, num_instances, base_vertex, base_instance, dld); | 281 | cmdbuf.Draw(num_vertices, num_instances, base_vertex, base_instance); |
| 278 | } | 282 | } |
| 279 | } | 283 | } |
| 280 | 284 | ||
| @@ -337,7 +341,7 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) { | |||
| 337 | 341 | ||
| 338 | const auto renderpass = pipeline.GetRenderPass(); | 342 | const auto renderpass = pipeline.GetRenderPass(); |
| 339 | const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass); | 343 | const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass); |
| 340 | scheduler.RequestRenderpass({renderpass, framebuffer, {{0, 0}, render_area}, 0, nullptr}); | 344 | scheduler.RequestRenderpass(renderpass, framebuffer, render_area); |
| 341 | 345 | ||
| 342 | UpdateDynamicStates(); | 346 | UpdateDynamicStates(); |
| 343 | 347 | ||
| @@ -345,19 +349,19 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) { | |||
| 345 | 349 | ||
| 346 | if (device.IsNvDeviceDiagnosticCheckpoints()) { | 350 | if (device.IsNvDeviceDiagnosticCheckpoints()) { |
| 347 | scheduler.Record( | 351 | scheduler.Record( |
| 348 | [&pipeline](auto cmdbuf, auto& dld) { cmdbuf.setCheckpointNV(&pipeline, dld); }); | 352 | [&pipeline](vk::CommandBuffer cmdbuf) { cmdbuf.SetCheckpointNV(&pipeline); }); |
| 349 | } | 353 | } |
| 350 | 354 | ||
| 351 | BeginTransformFeedback(); | 355 | BeginTransformFeedback(); |
| 352 | 356 | ||
| 353 | const auto pipeline_layout = pipeline.GetLayout(); | 357 | const auto pipeline_layout = pipeline.GetLayout(); |
| 354 | const auto descriptor_set = pipeline.CommitDescriptorSet(); | 358 | const auto descriptor_set = pipeline.CommitDescriptorSet(); |
| 355 | scheduler.Record([pipeline_layout, descriptor_set, draw_params](auto cmdbuf, auto& dld) { | 359 | scheduler.Record([pipeline_layout, descriptor_set, draw_params](vk::CommandBuffer cmdbuf) { |
| 356 | if (descriptor_set) { | 360 | if (descriptor_set) { |
| 357 | cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, pipeline_layout, | 361 | cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline_layout, |
| 358 | DESCRIPTOR_SET, 1, &descriptor_set, 0, nullptr, dld); | 362 | DESCRIPTOR_SET, descriptor_set, {}); |
| 359 | } | 363 | } |
| 360 | draw_params.Draw(cmdbuf, dld); | 364 | draw_params.Draw(cmdbuf); |
| 361 | }); | 365 | }); |
| 362 | 366 | ||
| 363 | EndTransformFeedback(); | 367 | EndTransformFeedback(); |
| @@ -389,48 +393,54 @@ void RasterizerVulkan::Clear() { | |||
| 389 | DEBUG_ASSERT(texceptions.none()); | 393 | DEBUG_ASSERT(texceptions.none()); |
| 390 | SetupImageTransitions(0, color_attachments, zeta_attachment); | 394 | SetupImageTransitions(0, color_attachments, zeta_attachment); |
| 391 | 395 | ||
| 392 | const vk::RenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0)); | 396 | const VkRenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0)); |
| 393 | const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass); | 397 | const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass); |
| 394 | scheduler.RequestRenderpass({renderpass, framebuffer, {{0, 0}, render_area}, 0, nullptr}); | 398 | scheduler.RequestRenderpass(renderpass, framebuffer, render_area); |
| 395 | |||
| 396 | const auto& scissor = regs.scissor_test[0]; | ||
| 397 | const vk::Offset2D scissor_offset(scissor.min_x, scissor.min_y); | ||
| 398 | vk::Extent2D scissor_extent{scissor.max_x - scissor.min_x, scissor.max_y - scissor.min_y}; | ||
| 399 | scissor_extent.width = std::min(scissor_extent.width, render_area.width); | ||
| 400 | scissor_extent.height = std::min(scissor_extent.height, render_area.height); | ||
| 401 | 399 | ||
| 402 | const u32 layer = regs.clear_buffers.layer; | 400 | VkClearRect clear_rect; |
| 403 | const vk::ClearRect clear_rect({scissor_offset, scissor_extent}, layer, 1); | 401 | clear_rect.baseArrayLayer = regs.clear_buffers.layer; |
| 402 | clear_rect.layerCount = 1; | ||
| 403 | clear_rect.rect = GetScissorState(regs, 0); | ||
| 404 | clear_rect.rect.extent.width = std::min(clear_rect.rect.extent.width, render_area.width); | ||
| 405 | clear_rect.rect.extent.height = std::min(clear_rect.rect.extent.height, render_area.height); | ||
| 404 | 406 | ||
| 405 | if (use_color) { | 407 | if (use_color) { |
| 406 | const std::array clear_color = {regs.clear_color[0], regs.clear_color[1], | 408 | VkClearValue clear_value; |
| 407 | regs.clear_color[2], regs.clear_color[3]}; | 409 | std::memcpy(clear_value.color.float32, regs.clear_color, sizeof(regs.clear_color)); |
| 408 | const vk::ClearValue clear_value{clear_color}; | 410 | |
| 409 | const u32 color_attachment = regs.clear_buffers.RT; | 411 | const u32 color_attachment = regs.clear_buffers.RT; |
| 410 | scheduler.Record([color_attachment, clear_value, clear_rect](auto cmdbuf, auto& dld) { | 412 | scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf) { |
| 411 | const vk::ClearAttachment attachment(vk::ImageAspectFlagBits::eColor, color_attachment, | 413 | VkClearAttachment attachment; |
| 412 | clear_value); | 414 | attachment.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; |
| 413 | cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld); | 415 | attachment.colorAttachment = color_attachment; |
| 416 | attachment.clearValue = clear_value; | ||
| 417 | cmdbuf.ClearAttachments(attachment, clear_rect); | ||
| 414 | }); | 418 | }); |
| 415 | } | 419 | } |
| 416 | 420 | ||
| 417 | if (!use_depth && !use_stencil) { | 421 | if (!use_depth && !use_stencil) { |
| 418 | return; | 422 | return; |
| 419 | } | 423 | } |
| 420 | vk::ImageAspectFlags aspect_flags; | 424 | VkImageAspectFlags aspect_flags = 0; |
| 421 | if (use_depth) { | 425 | if (use_depth) { |
| 422 | aspect_flags |= vk::ImageAspectFlagBits::eDepth; | 426 | aspect_flags |= VK_IMAGE_ASPECT_DEPTH_BIT; |
| 423 | } | 427 | } |
| 424 | if (use_stencil) { | 428 | if (use_stencil) { |
| 425 | aspect_flags |= vk::ImageAspectFlagBits::eStencil; | 429 | aspect_flags |= VK_IMAGE_ASPECT_STENCIL_BIT; |
| 426 | } | 430 | } |
| 427 | 431 | ||
| 428 | scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil, | 432 | scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil, |
| 429 | clear_rect, aspect_flags](auto cmdbuf, auto& dld) { | 433 | clear_rect, aspect_flags](vk::CommandBuffer cmdbuf) { |
| 430 | const vk::ClearDepthStencilValue clear_zeta(clear_depth, clear_stencil); | 434 | VkClearValue clear_value; |
| 431 | const vk::ClearValue clear_value{clear_zeta}; | 435 | clear_value.depthStencil.depth = clear_depth; |
| 432 | const vk::ClearAttachment attachment(aspect_flags, 0, clear_value); | 436 | clear_value.depthStencil.stencil = clear_stencil; |
| 433 | cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld); | 437 | |
| 438 | VkClearAttachment attachment; | ||
| 439 | attachment.aspectMask = aspect_flags; | ||
| 440 | attachment.colorAttachment = 0; | ||
| 441 | attachment.clearValue.depthStencil.depth = clear_depth; | ||
| 442 | attachment.clearValue.depthStencil.stencil = clear_stencil; | ||
| 443 | cmdbuf.ClearAttachments(attachment, clear_rect); | ||
| 434 | }); | 444 | }); |
| 435 | } | 445 | } |
| 436 | 446 | ||
| @@ -463,24 +473,24 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) { | |||
| 463 | 473 | ||
| 464 | buffer_cache.Unmap(); | 474 | buffer_cache.Unmap(); |
| 465 | 475 | ||
| 466 | TransitionImages(sampled_views, vk::PipelineStageFlagBits::eComputeShader, | 476 | TransitionImages(sampled_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, |
| 467 | vk::AccessFlagBits::eShaderRead); | 477 | VK_ACCESS_SHADER_READ_BIT); |
| 468 | TransitionImages(image_views, vk::PipelineStageFlagBits::eComputeShader, | 478 | TransitionImages(image_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, |
| 469 | vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite); | 479 | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT); |
| 470 | 480 | ||
| 471 | if (device.IsNvDeviceDiagnosticCheckpoints()) { | 481 | if (device.IsNvDeviceDiagnosticCheckpoints()) { |
| 472 | scheduler.Record( | 482 | scheduler.Record( |
| 473 | [&pipeline](auto cmdbuf, auto& dld) { cmdbuf.setCheckpointNV(nullptr, dld); }); | 483 | [&pipeline](vk::CommandBuffer cmdbuf) { cmdbuf.SetCheckpointNV(nullptr); }); |
| 474 | } | 484 | } |
| 475 | 485 | ||
| 476 | scheduler.Record([grid_x = launch_desc.grid_dim_x, grid_y = launch_desc.grid_dim_y, | 486 | scheduler.Record([grid_x = launch_desc.grid_dim_x, grid_y = launch_desc.grid_dim_y, |
| 477 | grid_z = launch_desc.grid_dim_z, pipeline_handle = pipeline.GetHandle(), | 487 | grid_z = launch_desc.grid_dim_z, pipeline_handle = pipeline.GetHandle(), |
| 478 | layout = pipeline.GetLayout(), | 488 | layout = pipeline.GetLayout(), |
| 479 | descriptor_set = pipeline.CommitDescriptorSet()](auto cmdbuf, auto& dld) { | 489 | descriptor_set = pipeline.CommitDescriptorSet()](vk::CommandBuffer cmdbuf) { |
| 480 | cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline_handle, dld); | 490 | cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline_handle); |
| 481 | cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, DESCRIPTOR_SET, 1, | 491 | cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, DESCRIPTOR_SET, |
| 482 | &descriptor_set, 0, nullptr, dld); | 492 | descriptor_set, {}); |
| 483 | cmdbuf.dispatch(grid_x, grid_y, grid_z, dld); | 493 | cmdbuf.Dispatch(grid_x, grid_y, grid_z); |
| 484 | }); | 494 | }); |
| 485 | } | 495 | } |
| 486 | 496 | ||
| @@ -625,13 +635,13 @@ bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachmen | |||
| 625 | continue; | 635 | continue; |
| 626 | } | 636 | } |
| 627 | overlap = true; | 637 | overlap = true; |
| 628 | *layout = vk::ImageLayout::eGeneral; | 638 | *layout = VK_IMAGE_LAYOUT_GENERAL; |
| 629 | } | 639 | } |
| 630 | return overlap; | 640 | return overlap; |
| 631 | } | 641 | } |
| 632 | 642 | ||
| 633 | std::tuple<vk::Framebuffer, vk::Extent2D> RasterizerVulkan::ConfigureFramebuffers( | 643 | std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers( |
| 634 | vk::RenderPass renderpass) { | 644 | VkRenderPass renderpass) { |
| 635 | FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(), | 645 | FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(), |
| 636 | std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()}; | 646 | std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()}; |
| 637 | 647 | ||
| @@ -658,15 +668,20 @@ std::tuple<vk::Framebuffer, vk::Extent2D> RasterizerVulkan::ConfigureFramebuffer | |||
| 658 | const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key); | 668 | const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key); |
| 659 | auto& framebuffer = fbentry->second; | 669 | auto& framebuffer = fbentry->second; |
| 660 | if (is_cache_miss) { | 670 | if (is_cache_miss) { |
| 661 | const vk::FramebufferCreateInfo framebuffer_ci( | 671 | VkFramebufferCreateInfo framebuffer_ci; |
| 662 | {}, key.renderpass, static_cast<u32>(key.views.size()), key.views.data(), key.width, | 672 | framebuffer_ci.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; |
| 663 | key.height, key.layers); | 673 | framebuffer_ci.pNext = nullptr; |
| 664 | const auto dev = device.GetLogical(); | 674 | framebuffer_ci.flags = 0; |
| 665 | const auto& dld = device.GetDispatchLoader(); | 675 | framebuffer_ci.renderPass = key.renderpass; |
| 666 | framebuffer = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld); | 676 | framebuffer_ci.attachmentCount = static_cast<u32>(key.views.size()); |
| 667 | } | 677 | framebuffer_ci.pAttachments = key.views.data(); |
| 668 | 678 | framebuffer_ci.width = key.width; | |
| 669 | return {*framebuffer, vk::Extent2D{key.width, key.height}}; | 679 | framebuffer_ci.height = key.height; |
| 680 | framebuffer_ci.layers = key.layers; | ||
| 681 | framebuffer = device.GetLogical().CreateFramebuffer(framebuffer_ci); | ||
| 682 | } | ||
| 683 | |||
| 684 | return {*framebuffer, VkExtent2D{key.width, key.height}}; | ||
| 670 | } | 685 | } |
| 671 | 686 | ||
| 672 | RasterizerVulkan::DrawParameters RasterizerVulkan::SetupGeometry(FixedPipelineState& fixed_state, | 687 | RasterizerVulkan::DrawParameters RasterizerVulkan::SetupGeometry(FixedPipelineState& fixed_state, |
| @@ -714,10 +729,9 @@ void RasterizerVulkan::SetupShaderDescriptors( | |||
| 714 | void RasterizerVulkan::SetupImageTransitions( | 729 | void RasterizerVulkan::SetupImageTransitions( |
| 715 | Texceptions texceptions, const std::array<View, Maxwell::NumRenderTargets>& color_attachments, | 730 | Texceptions texceptions, const std::array<View, Maxwell::NumRenderTargets>& color_attachments, |
| 716 | const View& zeta_attachment) { | 731 | const View& zeta_attachment) { |
| 717 | TransitionImages(sampled_views, vk::PipelineStageFlagBits::eAllGraphics, | 732 | TransitionImages(sampled_views, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, VK_ACCESS_SHADER_READ_BIT); |
| 718 | vk::AccessFlagBits::eShaderRead); | 733 | TransitionImages(image_views, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, |
| 719 | TransitionImages(image_views, vk::PipelineStageFlagBits::eAllGraphics, | 734 | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT); |
| 720 | vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite); | ||
| 721 | 735 | ||
| 722 | for (std::size_t rt = 0; rt < std::size(color_attachments); ++rt) { | 736 | for (std::size_t rt = 0; rt < std::size(color_attachments); ++rt) { |
| 723 | const auto color_attachment = color_attachments[rt]; | 737 | const auto color_attachment = color_attachments[rt]; |
| @@ -725,19 +739,19 @@ void RasterizerVulkan::SetupImageTransitions( | |||
| 725 | continue; | 739 | continue; |
| 726 | } | 740 | } |
| 727 | const auto image_layout = | 741 | const auto image_layout = |
| 728 | texceptions[rt] ? vk::ImageLayout::eGeneral : vk::ImageLayout::eColorAttachmentOptimal; | 742 | texceptions[rt] ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; |
| 729 | color_attachment->Transition( | 743 | color_attachment->Transition(image_layout, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, |
| 730 | image_layout, vk::PipelineStageFlagBits::eColorAttachmentOutput, | 744 | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | |
| 731 | vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite); | 745 | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT); |
| 732 | } | 746 | } |
| 733 | 747 | ||
| 734 | if (zeta_attachment != nullptr) { | 748 | if (zeta_attachment != nullptr) { |
| 735 | const auto image_layout = texceptions[ZETA_TEXCEPTION_INDEX] | 749 | const auto image_layout = texceptions[ZETA_TEXCEPTION_INDEX] |
| 736 | ? vk::ImageLayout::eGeneral | 750 | ? VK_IMAGE_LAYOUT_GENERAL |
| 737 | : vk::ImageLayout::eDepthStencilAttachmentOptimal; | 751 | : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; |
| 738 | zeta_attachment->Transition(image_layout, vk::PipelineStageFlagBits::eLateFragmentTests, | 752 | zeta_attachment->Transition(image_layout, VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, |
| 739 | vk::AccessFlagBits::eDepthStencilAttachmentRead | | 753 | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | |
| 740 | vk::AccessFlagBits::eDepthStencilAttachmentWrite); | 754 | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT); |
| 741 | } | 755 | } |
| 742 | } | 756 | } |
| 743 | 757 | ||
| @@ -773,9 +787,9 @@ void RasterizerVulkan::BeginTransformFeedback() { | |||
| 773 | const std::size_t size = binding.buffer_size; | 787 | const std::size_t size = binding.buffer_size; |
| 774 | const auto [buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true); | 788 | const auto [buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true); |
| 775 | 789 | ||
| 776 | scheduler.Record([buffer = *buffer, offset = offset, size](auto cmdbuf, auto& dld) { | 790 | scheduler.Record([buffer = *buffer, offset = offset, size](vk::CommandBuffer cmdbuf) { |
| 777 | cmdbuf.bindTransformFeedbackBuffersEXT(0, {buffer}, {offset}, {size}, dld); | 791 | cmdbuf.BindTransformFeedbackBuffersEXT(0, 1, &buffer, &offset, &size); |
| 778 | cmdbuf.beginTransformFeedbackEXT(0, {}, {}, dld); | 792 | cmdbuf.BeginTransformFeedbackEXT(0, 0, nullptr, nullptr); |
| 779 | }); | 793 | }); |
| 780 | } | 794 | } |
| 781 | 795 | ||
| @@ -786,7 +800,7 @@ void RasterizerVulkan::EndTransformFeedback() { | |||
| 786 | } | 800 | } |
| 787 | 801 | ||
| 788 | scheduler.Record( | 802 | scheduler.Record( |
| 789 | [](auto cmdbuf, auto& dld) { cmdbuf.endTransformFeedbackEXT(0, {}, {}, dld); }); | 803 | [](vk::CommandBuffer cmdbuf) { cmdbuf.EndTransformFeedbackEXT(0, 0, nullptr, nullptr); }); |
| 790 | } | 804 | } |
| 791 | 805 | ||
| 792 | void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex_input, | 806 | void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex_input, |
| @@ -837,7 +851,7 @@ void RasterizerVulkan::SetupIndexBuffer(BufferBindings& buffer_bindings, DrawPar | |||
| 837 | } else { | 851 | } else { |
| 838 | const auto [buffer, offset] = | 852 | const auto [buffer, offset] = |
| 839 | quad_array_pass.Assemble(params.num_vertices, params.base_vertex); | 853 | quad_array_pass.Assemble(params.num_vertices, params.base_vertex); |
| 840 | buffer_bindings.SetIndexBinding(&buffer, offset, vk::IndexType::eUint32); | 854 | buffer_bindings.SetIndexBinding(buffer, offset, VK_INDEX_TYPE_UINT32); |
| 841 | params.base_vertex = 0; | 855 | params.base_vertex = 0; |
| 842 | params.num_vertices = params.num_vertices * 6 / 4; | 856 | params.num_vertices = params.num_vertices * 6 / 4; |
| 843 | params.is_indexed = true; | 857 | params.is_indexed = true; |
| @@ -1022,7 +1036,7 @@ void RasterizerVulkan::SetupTexture(const Tegra::Texture::FullTextureInfo& textu | |||
| 1022 | update_descriptor_queue.AddSampledImage(sampler, image_view); | 1036 | update_descriptor_queue.AddSampledImage(sampler, image_view); |
| 1023 | 1037 | ||
| 1024 | const auto image_layout = update_descriptor_queue.GetLastImageLayout(); | 1038 | const auto image_layout = update_descriptor_queue.GetLastImageLayout(); |
| 1025 | *image_layout = vk::ImageLayout::eShaderReadOnlyOptimal; | 1039 | *image_layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; |
| 1026 | sampled_views.push_back(ImageView{std::move(view), image_layout}); | 1040 | sampled_views.push_back(ImageView{std::move(view), image_layout}); |
| 1027 | } | 1041 | } |
| 1028 | 1042 | ||
| @@ -1039,7 +1053,7 @@ void RasterizerVulkan::SetupImage(const Tegra::Texture::TICEntry& tic, const Ima | |||
| 1039 | update_descriptor_queue.AddImage(image_view); | 1053 | update_descriptor_queue.AddImage(image_view); |
| 1040 | 1054 | ||
| 1041 | const auto image_layout = update_descriptor_queue.GetLastImageLayout(); | 1055 | const auto image_layout = update_descriptor_queue.GetLastImageLayout(); |
| 1042 | *image_layout = vk::ImageLayout::eGeneral; | 1056 | *image_layout = VK_IMAGE_LAYOUT_GENERAL; |
| 1043 | image_views.push_back(ImageView{std::move(view), image_layout}); | 1057 | image_views.push_back(ImageView{std::move(view), image_layout}); |
| 1044 | } | 1058 | } |
| 1045 | 1059 | ||
| @@ -1056,9 +1070,7 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg | |||
| 1056 | GetViewportState(device, regs, 10), GetViewportState(device, regs, 11), | 1070 | GetViewportState(device, regs, 10), GetViewportState(device, regs, 11), |
| 1057 | GetViewportState(device, regs, 12), GetViewportState(device, regs, 13), | 1071 | GetViewportState(device, regs, 12), GetViewportState(device, regs, 13), |
| 1058 | GetViewportState(device, regs, 14), GetViewportState(device, regs, 15)}; | 1072 | GetViewportState(device, regs, 14), GetViewportState(device, regs, 15)}; |
| 1059 | scheduler.Record([viewports](auto cmdbuf, auto& dld) { | 1073 | scheduler.Record([viewports](vk::CommandBuffer cmdbuf) { cmdbuf.SetViewport(0, viewports); }); |
| 1060 | cmdbuf.setViewport(0, static_cast<u32>(viewports.size()), viewports.data(), dld); | ||
| 1061 | }); | ||
| 1062 | } | 1074 | } |
| 1063 | 1075 | ||
| 1064 | void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs) { | 1076 | void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs) { |
| @@ -1072,9 +1084,7 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs | |||
| 1072 | GetScissorState(regs, 9), GetScissorState(regs, 10), GetScissorState(regs, 11), | 1084 | GetScissorState(regs, 9), GetScissorState(regs, 10), GetScissorState(regs, 11), |
| 1073 | GetScissorState(regs, 12), GetScissorState(regs, 13), GetScissorState(regs, 14), | 1085 | GetScissorState(regs, 12), GetScissorState(regs, 13), GetScissorState(regs, 14), |
| 1074 | GetScissorState(regs, 15)}; | 1086 | GetScissorState(regs, 15)}; |
| 1075 | scheduler.Record([scissors](auto cmdbuf, auto& dld) { | 1087 | scheduler.Record([scissors](vk::CommandBuffer cmdbuf) { cmdbuf.SetScissor(0, scissors); }); |
| 1076 | cmdbuf.setScissor(0, static_cast<u32>(scissors.size()), scissors.data(), dld); | ||
| 1077 | }); | ||
| 1078 | } | 1088 | } |
| 1079 | 1089 | ||
| 1080 | void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) { | 1090 | void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) { |
| @@ -1082,8 +1092,8 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) { | |||
| 1082 | return; | 1092 | return; |
| 1083 | } | 1093 | } |
| 1084 | scheduler.Record([constant = regs.polygon_offset_units, clamp = regs.polygon_offset_clamp, | 1094 | scheduler.Record([constant = regs.polygon_offset_units, clamp = regs.polygon_offset_clamp, |
| 1085 | factor = regs.polygon_offset_factor](auto cmdbuf, auto& dld) { | 1095 | factor = regs.polygon_offset_factor](vk::CommandBuffer cmdbuf) { |
| 1086 | cmdbuf.setDepthBias(constant, clamp, factor / 2.0f, dld); | 1096 | cmdbuf.SetDepthBias(constant, clamp, factor / 2.0f); |
| 1087 | }); | 1097 | }); |
| 1088 | } | 1098 | } |
| 1089 | 1099 | ||
| @@ -1093,9 +1103,8 @@ void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& reg | |||
| 1093 | } | 1103 | } |
| 1094 | const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b, | 1104 | const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b, |
| 1095 | regs.blend_color.a}; | 1105 | regs.blend_color.a}; |
| 1096 | scheduler.Record([blend_color](auto cmdbuf, auto& dld) { | 1106 | scheduler.Record( |
| 1097 | cmdbuf.setBlendConstants(blend_color.data(), dld); | 1107 | [blend_color](vk::CommandBuffer cmdbuf) { cmdbuf.SetBlendConstants(blend_color.data()); }); |
| 1098 | }); | ||
| 1099 | } | 1108 | } |
| 1100 | 1109 | ||
| 1101 | void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) { | 1110 | void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) { |
| @@ -1103,7 +1112,7 @@ void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) | |||
| 1103 | return; | 1112 | return; |
| 1104 | } | 1113 | } |
| 1105 | scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]]( | 1114 | scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]]( |
| 1106 | auto cmdbuf, auto& dld) { cmdbuf.setDepthBounds(min, max, dld); }); | 1115 | vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); }); |
| 1107 | } | 1116 | } |
| 1108 | 1117 | ||
| 1109 | void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) { | 1118 | void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) { |
| @@ -1116,24 +1125,24 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) | |||
| 1116 | [front_ref = regs.stencil_front_func_ref, front_write_mask = regs.stencil_front_mask, | 1125 | [front_ref = regs.stencil_front_func_ref, front_write_mask = regs.stencil_front_mask, |
| 1117 | front_test_mask = regs.stencil_front_func_mask, back_ref = regs.stencil_back_func_ref, | 1126 | front_test_mask = regs.stencil_front_func_mask, back_ref = regs.stencil_back_func_ref, |
| 1118 | back_write_mask = regs.stencil_back_mask, | 1127 | back_write_mask = regs.stencil_back_mask, |
| 1119 | back_test_mask = regs.stencil_back_func_mask](auto cmdbuf, auto& dld) { | 1128 | back_test_mask = regs.stencil_back_func_mask](vk::CommandBuffer cmdbuf) { |
| 1120 | // Front face | 1129 | // Front face |
| 1121 | cmdbuf.setStencilReference(vk::StencilFaceFlagBits::eFront, front_ref, dld); | 1130 | cmdbuf.SetStencilReference(VK_STENCIL_FACE_FRONT_BIT, front_ref); |
| 1122 | cmdbuf.setStencilWriteMask(vk::StencilFaceFlagBits::eFront, front_write_mask, dld); | 1131 | cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_FRONT_BIT, front_write_mask); |
| 1123 | cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eFront, front_test_mask, dld); | 1132 | cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_FRONT_BIT, front_test_mask); |
| 1124 | 1133 | ||
| 1125 | // Back face | 1134 | // Back face |
| 1126 | cmdbuf.setStencilReference(vk::StencilFaceFlagBits::eBack, back_ref, dld); | 1135 | cmdbuf.SetStencilReference(VK_STENCIL_FACE_BACK_BIT, back_ref); |
| 1127 | cmdbuf.setStencilWriteMask(vk::StencilFaceFlagBits::eBack, back_write_mask, dld); | 1136 | cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_BACK_BIT, back_write_mask); |
| 1128 | cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eBack, back_test_mask, dld); | 1137 | cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_BACK_BIT, back_test_mask); |
| 1129 | }); | 1138 | }); |
| 1130 | } else { | 1139 | } else { |
| 1131 | // Front face defines both faces | 1140 | // Front face defines both faces |
| 1132 | scheduler.Record([ref = regs.stencil_back_func_ref, write_mask = regs.stencil_back_mask, | 1141 | scheduler.Record([ref = regs.stencil_back_func_ref, write_mask = regs.stencil_back_mask, |
| 1133 | test_mask = regs.stencil_back_func_mask](auto cmdbuf, auto& dld) { | 1142 | test_mask = regs.stencil_back_func_mask](vk::CommandBuffer cmdbuf) { |
| 1134 | cmdbuf.setStencilReference(vk::StencilFaceFlagBits::eFrontAndBack, ref, dld); | 1143 | cmdbuf.SetStencilReference(VK_STENCIL_FACE_FRONT_AND_BACK, ref); |
| 1135 | cmdbuf.setStencilWriteMask(vk::StencilFaceFlagBits::eFrontAndBack, write_mask, dld); | 1144 | cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_FRONT_AND_BACK, write_mask); |
| 1136 | cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eFrontAndBack, test_mask, dld); | 1145 | cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_FRONT_AND_BACK, test_mask); |
| 1137 | }); | 1146 | }); |
| 1138 | } | 1147 | } |
| 1139 | } | 1148 | } |
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index f642dde76..46037860a 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h | |||
| @@ -17,7 +17,6 @@ | |||
| 17 | #include "video_core/memory_manager.h" | 17 | #include "video_core/memory_manager.h" |
| 18 | #include "video_core/rasterizer_accelerated.h" | 18 | #include "video_core/rasterizer_accelerated.h" |
| 19 | #include "video_core/rasterizer_interface.h" | 19 | #include "video_core/rasterizer_interface.h" |
| 20 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 21 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" | 20 | #include "video_core/renderer_vulkan/fixed_pipeline_state.h" |
| 22 | #include "video_core/renderer_vulkan/vk_buffer_cache.h" | 21 | #include "video_core/renderer_vulkan/vk_buffer_cache.h" |
| 23 | #include "video_core/renderer_vulkan/vk_compute_pass.h" | 22 | #include "video_core/renderer_vulkan/vk_compute_pass.h" |
| @@ -32,6 +31,7 @@ | |||
| 32 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" | 31 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" |
| 33 | #include "video_core/renderer_vulkan/vk_texture_cache.h" | 32 | #include "video_core/renderer_vulkan/vk_texture_cache.h" |
| 34 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 33 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 34 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 35 | 35 | ||
| 36 | namespace Core { | 36 | namespace Core { |
| 37 | class System; | 37 | class System; |
| @@ -49,11 +49,10 @@ namespace Vulkan { | |||
| 49 | 49 | ||
| 50 | struct VKScreenInfo; | 50 | struct VKScreenInfo; |
| 51 | 51 | ||
| 52 | using ImageViewsPack = | 52 | using ImageViewsPack = boost::container::static_vector<VkImageView, Maxwell::NumRenderTargets + 1>; |
| 53 | boost::container::static_vector<vk::ImageView, Maxwell::NumRenderTargets + 1>; | ||
| 54 | 53 | ||
| 55 | struct FramebufferCacheKey { | 54 | struct FramebufferCacheKey { |
| 56 | vk::RenderPass renderpass{}; | 55 | VkRenderPass renderpass{}; |
| 57 | u32 width = 0; | 56 | u32 width = 0; |
| 58 | u32 height = 0; | 57 | u32 height = 0; |
| 59 | u32 layers = 0; | 58 | u32 layers = 0; |
| @@ -101,7 +100,7 @@ class BufferBindings; | |||
| 101 | 100 | ||
| 102 | struct ImageView { | 101 | struct ImageView { |
| 103 | View view; | 102 | View view; |
| 104 | vk::ImageLayout* layout = nullptr; | 103 | VkImageLayout* layout = nullptr; |
| 105 | }; | 104 | }; |
| 106 | 105 | ||
| 107 | class RasterizerVulkan final : public VideoCore::RasterizerAccelerated { | 106 | class RasterizerVulkan final : public VideoCore::RasterizerAccelerated { |
| @@ -137,7 +136,7 @@ public: | |||
| 137 | 136 | ||
| 138 | private: | 137 | private: |
| 139 | struct DrawParameters { | 138 | struct DrawParameters { |
| 140 | void Draw(vk::CommandBuffer cmdbuf, const vk::DispatchLoaderDynamic& dld) const; | 139 | void Draw(vk::CommandBuffer cmdbuf) const; |
| 141 | 140 | ||
| 142 | u32 base_instance = 0; | 141 | u32 base_instance = 0; |
| 143 | u32 num_instances = 0; | 142 | u32 num_instances = 0; |
| @@ -154,7 +153,7 @@ private: | |||
| 154 | 153 | ||
| 155 | Texceptions UpdateAttachments(); | 154 | Texceptions UpdateAttachments(); |
| 156 | 155 | ||
| 157 | std::tuple<vk::Framebuffer, vk::Extent2D> ConfigureFramebuffers(vk::RenderPass renderpass); | 156 | std::tuple<VkFramebuffer, VkExtent2D> ConfigureFramebuffers(VkRenderPass renderpass); |
| 158 | 157 | ||
| 159 | /// Setups geometry buffers and state. | 158 | /// Setups geometry buffers and state. |
| 160 | DrawParameters SetupGeometry(FixedPipelineState& fixed_state, BufferBindings& buffer_bindings, | 159 | DrawParameters SetupGeometry(FixedPipelineState& fixed_state, BufferBindings& buffer_bindings, |
| @@ -272,7 +271,7 @@ private: | |||
| 272 | u32 draw_counter = 0; | 271 | u32 draw_counter = 0; |
| 273 | 272 | ||
| 274 | // TODO(Rodrigo): Invalidate on image destruction | 273 | // TODO(Rodrigo): Invalidate on image destruction |
| 275 | std::unordered_map<FramebufferCacheKey, UniqueFramebuffer> framebuffer_cache; | 274 | std::unordered_map<FramebufferCacheKey, vk::Framebuffer> framebuffer_cache; |
| 276 | }; | 275 | }; |
| 277 | 276 | ||
| 278 | } // namespace Vulkan | 277 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp b/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp index 93f5d7ba0..4e5286a69 100644 --- a/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp | |||
| @@ -6,10 +6,10 @@ | |||
| 6 | #include <vector> | 6 | #include <vector> |
| 7 | 7 | ||
| 8 | #include "video_core/engines/maxwell_3d.h" | 8 | #include "video_core/engines/maxwell_3d.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 9 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 11 | #include "video_core/renderer_vulkan/vk_device.h" | 10 | #include "video_core/renderer_vulkan/vk_device.h" |
| 12 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" | 11 | #include "video_core/renderer_vulkan/vk_renderpass_cache.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -17,7 +17,7 @@ VKRenderPassCache::VKRenderPassCache(const VKDevice& device) : device{device} {} | |||
| 17 | 17 | ||
| 18 | VKRenderPassCache::~VKRenderPassCache() = default; | 18 | VKRenderPassCache::~VKRenderPassCache() = default; |
| 19 | 19 | ||
| 20 | vk::RenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) { | 20 | VkRenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) { |
| 21 | const auto [pair, is_cache_miss] = cache.try_emplace(params); | 21 | const auto [pair, is_cache_miss] = cache.try_emplace(params); |
| 22 | auto& entry = pair->second; | 22 | auto& entry = pair->second; |
| 23 | if (is_cache_miss) { | 23 | if (is_cache_miss) { |
| @@ -26,9 +26,9 @@ vk::RenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) | |||
| 26 | return *entry; | 26 | return *entry; |
| 27 | } | 27 | } |
| 28 | 28 | ||
| 29 | UniqueRenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const { | 29 | vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const { |
| 30 | std::vector<vk::AttachmentDescription> descriptors; | 30 | std::vector<VkAttachmentDescription> descriptors; |
| 31 | std::vector<vk::AttachmentReference> color_references; | 31 | std::vector<VkAttachmentReference> color_references; |
| 32 | 32 | ||
| 33 | for (std::size_t rt = 0; rt < params.color_attachments.size(); ++rt) { | 33 | for (std::size_t rt = 0; rt < params.color_attachments.size(); ++rt) { |
| 34 | const auto attachment = params.color_attachments[rt]; | 34 | const auto attachment = params.color_attachments[rt]; |
| @@ -39,16 +39,25 @@ UniqueRenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& par | |||
| 39 | 39 | ||
| 40 | // TODO(Rodrigo): Add eMayAlias when it's needed. | 40 | // TODO(Rodrigo): Add eMayAlias when it's needed. |
| 41 | const auto color_layout = attachment.is_texception | 41 | const auto color_layout = attachment.is_texception |
| 42 | ? vk::ImageLayout::eGeneral | 42 | ? VK_IMAGE_LAYOUT_GENERAL |
| 43 | : vk::ImageLayout::eColorAttachmentOptimal; | 43 | : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; |
| 44 | descriptors.emplace_back(vk::AttachmentDescriptionFlagBits::eMayAlias, format.format, | 44 | VkAttachmentDescription& descriptor = descriptors.emplace_back(); |
| 45 | vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eLoad, | 45 | descriptor.flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT; |
| 46 | vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eDontCare, | 46 | descriptor.format = format.format; |
| 47 | vk::AttachmentStoreOp::eDontCare, color_layout, color_layout); | 47 | descriptor.samples = VK_SAMPLE_COUNT_1_BIT; |
| 48 | color_references.emplace_back(static_cast<u32>(rt), color_layout); | 48 | descriptor.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; |
| 49 | descriptor.storeOp = VK_ATTACHMENT_STORE_OP_STORE; | ||
| 50 | descriptor.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; | ||
| 51 | descriptor.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; | ||
| 52 | descriptor.initialLayout = color_layout; | ||
| 53 | descriptor.finalLayout = color_layout; | ||
| 54 | |||
| 55 | VkAttachmentReference& reference = color_references.emplace_back(); | ||
| 56 | reference.attachment = static_cast<u32>(rt); | ||
| 57 | reference.layout = color_layout; | ||
| 49 | } | 58 | } |
| 50 | 59 | ||
| 51 | vk::AttachmentReference zeta_attachment_ref; | 60 | VkAttachmentReference zeta_attachment_ref; |
| 52 | if (params.has_zeta) { | 61 | if (params.has_zeta) { |
| 53 | const auto format = | 62 | const auto format = |
| 54 | MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.zeta_pixel_format); | 63 | MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.zeta_pixel_format); |
| @@ -56,45 +65,68 @@ UniqueRenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& par | |||
| 56 | static_cast<u32>(params.zeta_pixel_format)); | 65 | static_cast<u32>(params.zeta_pixel_format)); |
| 57 | 66 | ||
| 58 | const auto zeta_layout = params.zeta_texception | 67 | const auto zeta_layout = params.zeta_texception |
| 59 | ? vk::ImageLayout::eGeneral | 68 | ? VK_IMAGE_LAYOUT_GENERAL |
| 60 | : vk::ImageLayout::eDepthStencilAttachmentOptimal; | 69 | : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; |
| 61 | descriptors.emplace_back(vk::AttachmentDescriptionFlags{}, format.format, | 70 | VkAttachmentDescription& descriptor = descriptors.emplace_back(); |
| 62 | vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eLoad, | 71 | descriptor.flags = 0; |
| 63 | vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eLoad, | 72 | descriptor.format = format.format; |
| 64 | vk::AttachmentStoreOp::eStore, zeta_layout, zeta_layout); | 73 | descriptor.samples = VK_SAMPLE_COUNT_1_BIT; |
| 65 | zeta_attachment_ref = | 74 | descriptor.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; |
| 66 | vk::AttachmentReference(static_cast<u32>(params.color_attachments.size()), zeta_layout); | 75 | descriptor.storeOp = VK_ATTACHMENT_STORE_OP_STORE; |
| 76 | descriptor.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD; | ||
| 77 | descriptor.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE; | ||
| 78 | descriptor.initialLayout = zeta_layout; | ||
| 79 | descriptor.finalLayout = zeta_layout; | ||
| 80 | |||
| 81 | zeta_attachment_ref.attachment = static_cast<u32>(params.color_attachments.size()); | ||
| 82 | zeta_attachment_ref.layout = zeta_layout; | ||
| 67 | } | 83 | } |
| 68 | 84 | ||
| 69 | const vk::SubpassDescription subpass_description( | 85 | VkSubpassDescription subpass_description; |
| 70 | {}, vk::PipelineBindPoint::eGraphics, 0, nullptr, static_cast<u32>(color_references.size()), | 86 | subpass_description.flags = 0; |
| 71 | color_references.data(), nullptr, params.has_zeta ? &zeta_attachment_ref : nullptr, 0, | 87 | subpass_description.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; |
| 72 | nullptr); | 88 | subpass_description.inputAttachmentCount = 0; |
| 73 | 89 | subpass_description.pInputAttachments = nullptr; | |
| 74 | vk::AccessFlags access; | 90 | subpass_description.colorAttachmentCount = static_cast<u32>(color_references.size()); |
| 75 | vk::PipelineStageFlags stage; | 91 | subpass_description.pColorAttachments = color_references.data(); |
| 92 | subpass_description.pResolveAttachments = nullptr; | ||
| 93 | subpass_description.pDepthStencilAttachment = params.has_zeta ? &zeta_attachment_ref : nullptr; | ||
| 94 | subpass_description.preserveAttachmentCount = 0; | ||
| 95 | subpass_description.pPreserveAttachments = nullptr; | ||
| 96 | |||
| 97 | VkAccessFlags access = 0; | ||
| 98 | VkPipelineStageFlags stage = 0; | ||
| 76 | if (!color_references.empty()) { | 99 | if (!color_references.empty()) { |
| 77 | access |= | 100 | access |= VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; |
| 78 | vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite; | 101 | stage |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; |
| 79 | stage |= vk::PipelineStageFlagBits::eColorAttachmentOutput; | ||
| 80 | } | 102 | } |
| 81 | 103 | ||
| 82 | if (params.has_zeta) { | 104 | if (params.has_zeta) { |
| 83 | access |= vk::AccessFlagBits::eDepthStencilAttachmentRead | | 105 | access |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | |
| 84 | vk::AccessFlagBits::eDepthStencilAttachmentWrite; | 106 | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT; |
| 85 | stage |= vk::PipelineStageFlagBits::eLateFragmentTests; | 107 | stage |= VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; |
| 86 | } | 108 | } |
| 87 | 109 | ||
| 88 | const vk::SubpassDependency subpass_dependency(VK_SUBPASS_EXTERNAL, 0, stage, stage, {}, access, | 110 | VkSubpassDependency subpass_dependency; |
| 89 | {}); | 111 | subpass_dependency.srcSubpass = VK_SUBPASS_EXTERNAL; |
| 90 | 112 | subpass_dependency.dstSubpass = 0; | |
| 91 | const vk::RenderPassCreateInfo create_info({}, static_cast<u32>(descriptors.size()), | 113 | subpass_dependency.srcStageMask = stage; |
| 92 | descriptors.data(), 1, &subpass_description, 1, | 114 | subpass_dependency.dstStageMask = stage; |
| 93 | &subpass_dependency); | 115 | subpass_dependency.srcAccessMask = 0; |
| 94 | 116 | subpass_dependency.dstAccessMask = access; | |
| 95 | const auto dev = device.GetLogical(); | 117 | subpass_dependency.dependencyFlags = 0; |
| 96 | const auto& dld = device.GetDispatchLoader(); | 118 | |
| 97 | return dev.createRenderPassUnique(create_info, nullptr, dld); | 119 | VkRenderPassCreateInfo ci; |
| 120 | ci.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; | ||
| 121 | ci.pNext = nullptr; | ||
| 122 | ci.flags = 0; | ||
| 123 | ci.attachmentCount = static_cast<u32>(descriptors.size()); | ||
| 124 | ci.pAttachments = descriptors.data(); | ||
| 125 | ci.subpassCount = 1; | ||
| 126 | ci.pSubpasses = &subpass_description; | ||
| 127 | ci.dependencyCount = 1; | ||
| 128 | ci.pDependencies = &subpass_dependency; | ||
| 129 | return device.GetLogical().CreateRenderPass(ci); | ||
| 98 | } | 130 | } |
| 99 | 131 | ||
| 100 | } // namespace Vulkan | 132 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_renderpass_cache.h b/src/video_core/renderer_vulkan/vk_renderpass_cache.h index b49b2db48..921b6efb5 100644 --- a/src/video_core/renderer_vulkan/vk_renderpass_cache.h +++ b/src/video_core/renderer_vulkan/vk_renderpass_cache.h | |||
| @@ -12,7 +12,7 @@ | |||
| 12 | #include <boost/functional/hash.hpp> | 12 | #include <boost/functional/hash.hpp> |
| 13 | 13 | ||
| 14 | #include "video_core/engines/maxwell_3d.h" | 14 | #include "video_core/engines/maxwell_3d.h" |
| 15 | #include "video_core/renderer_vulkan/declarations.h" | 15 | #include "video_core/renderer_vulkan/wrapper.h" |
| 16 | #include "video_core/surface.h" | 16 | #include "video_core/surface.h" |
| 17 | 17 | ||
| 18 | namespace Vulkan { | 18 | namespace Vulkan { |
| @@ -85,13 +85,13 @@ public: | |||
| 85 | explicit VKRenderPassCache(const VKDevice& device); | 85 | explicit VKRenderPassCache(const VKDevice& device); |
| 86 | ~VKRenderPassCache(); | 86 | ~VKRenderPassCache(); |
| 87 | 87 | ||
| 88 | vk::RenderPass GetRenderPass(const RenderPassParams& params); | 88 | VkRenderPass GetRenderPass(const RenderPassParams& params); |
| 89 | 89 | ||
| 90 | private: | 90 | private: |
| 91 | UniqueRenderPass CreateRenderPass(const RenderPassParams& params) const; | 91 | vk::RenderPass CreateRenderPass(const RenderPassParams& params) const; |
| 92 | 92 | ||
| 93 | const VKDevice& device; | 93 | const VKDevice& device; |
| 94 | std::unordered_map<RenderPassParams, UniqueRenderPass> cache; | 94 | std::unordered_map<RenderPassParams, vk::RenderPass> cache; |
| 95 | }; | 95 | }; |
| 96 | 96 | ||
| 97 | } // namespace Vulkan | 97 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.cpp b/src/video_core/renderer_vulkan/vk_resource_manager.cpp index 525b4bb46..dc06f545a 100644 --- a/src/video_core/renderer_vulkan/vk_resource_manager.cpp +++ b/src/video_core/renderer_vulkan/vk_resource_manager.cpp | |||
| @@ -6,83 +6,83 @@ | |||
| 6 | #include <optional> | 6 | #include <optional> |
| 7 | #include "common/assert.h" | 7 | #include "common/assert.h" |
| 8 | #include "common/logging/log.h" | 8 | #include "common/logging/log.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 10 | #include "video_core/renderer_vulkan/vk_device.h" | 9 | #include "video_core/renderer_vulkan/vk_device.h" |
| 11 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 10 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 11 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 12 | 12 | ||
| 13 | namespace Vulkan { | 13 | namespace Vulkan { |
| 14 | 14 | ||
| 15 | namespace { | ||
| 16 | |||
| 15 | // TODO(Rodrigo): Fine tune these numbers. | 17 | // TODO(Rodrigo): Fine tune these numbers. |
| 16 | constexpr std::size_t COMMAND_BUFFER_POOL_SIZE = 0x1000; | 18 | constexpr std::size_t COMMAND_BUFFER_POOL_SIZE = 0x1000; |
| 17 | constexpr std::size_t FENCES_GROW_STEP = 0x40; | 19 | constexpr std::size_t FENCES_GROW_STEP = 0x40; |
| 18 | 20 | ||
| 21 | VkFenceCreateInfo BuildFenceCreateInfo() { | ||
| 22 | VkFenceCreateInfo fence_ci; | ||
| 23 | fence_ci.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO; | ||
| 24 | fence_ci.pNext = nullptr; | ||
| 25 | fence_ci.flags = 0; | ||
| 26 | return fence_ci; | ||
| 27 | } | ||
| 28 | |||
| 29 | } // Anonymous namespace | ||
| 30 | |||
| 19 | class CommandBufferPool final : public VKFencedPool { | 31 | class CommandBufferPool final : public VKFencedPool { |
| 20 | public: | 32 | public: |
| 21 | CommandBufferPool(const VKDevice& device) | 33 | CommandBufferPool(const VKDevice& device) |
| 22 | : VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {} | 34 | : VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {} |
| 23 | 35 | ||
| 24 | void Allocate(std::size_t begin, std::size_t end) override { | 36 | void Allocate(std::size_t begin, std::size_t end) override { |
| 25 | const auto dev = device.GetLogical(); | ||
| 26 | const auto& dld = device.GetDispatchLoader(); | ||
| 27 | const u32 graphics_family = device.GetGraphicsFamily(); | ||
| 28 | |||
| 29 | auto pool = std::make_unique<Pool>(); | ||
| 30 | |||
| 31 | // Command buffers are going to be commited, recorded, executed every single usage cycle. | 37 | // Command buffers are going to be commited, recorded, executed every single usage cycle. |
| 32 | // They are also going to be reseted when commited. | 38 | // They are also going to be reseted when commited. |
| 33 | const auto pool_flags = vk::CommandPoolCreateFlagBits::eTransient | | 39 | VkCommandPoolCreateInfo command_pool_ci; |
| 34 | vk::CommandPoolCreateFlagBits::eResetCommandBuffer; | 40 | command_pool_ci.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO; |
| 35 | const vk::CommandPoolCreateInfo cmdbuf_pool_ci(pool_flags, graphics_family); | 41 | command_pool_ci.pNext = nullptr; |
| 36 | pool->handle = dev.createCommandPoolUnique(cmdbuf_pool_ci, nullptr, dld); | 42 | command_pool_ci.flags = |
| 37 | 43 | VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT; | |
| 38 | const vk::CommandBufferAllocateInfo cmdbuf_ai(*pool->handle, | 44 | command_pool_ci.queueFamilyIndex = device.GetGraphicsFamily(); |
| 39 | vk::CommandBufferLevel::ePrimary, | 45 | |
| 40 | static_cast<u32>(COMMAND_BUFFER_POOL_SIZE)); | 46 | Pool& pool = pools.emplace_back(); |
| 41 | pool->cmdbufs = | 47 | pool.handle = device.GetLogical().CreateCommandPool(command_pool_ci); |
| 42 | dev.allocateCommandBuffersUnique<std::allocator<UniqueCommandBuffer>>(cmdbuf_ai, dld); | 48 | pool.cmdbufs = pool.handle.Allocate(COMMAND_BUFFER_POOL_SIZE); |
| 43 | |||
| 44 | pools.push_back(std::move(pool)); | ||
| 45 | } | 49 | } |
| 46 | 50 | ||
| 47 | vk::CommandBuffer Commit(VKFence& fence) { | 51 | VkCommandBuffer Commit(VKFence& fence) { |
| 48 | const std::size_t index = CommitResource(fence); | 52 | const std::size_t index = CommitResource(fence); |
| 49 | const auto pool_index = index / COMMAND_BUFFER_POOL_SIZE; | 53 | const auto pool_index = index / COMMAND_BUFFER_POOL_SIZE; |
| 50 | const auto sub_index = index % COMMAND_BUFFER_POOL_SIZE; | 54 | const auto sub_index = index % COMMAND_BUFFER_POOL_SIZE; |
| 51 | return *pools[pool_index]->cmdbufs[sub_index]; | 55 | return pools[pool_index].cmdbufs[sub_index]; |
| 52 | } | 56 | } |
| 53 | 57 | ||
| 54 | private: | 58 | private: |
| 55 | struct Pool { | 59 | struct Pool { |
| 56 | UniqueCommandPool handle; | 60 | vk::CommandPool handle; |
| 57 | std::vector<UniqueCommandBuffer> cmdbufs; | 61 | vk::CommandBuffers cmdbufs; |
| 58 | }; | 62 | }; |
| 59 | 63 | ||
| 60 | const VKDevice& device; | 64 | const VKDevice& device; |
| 61 | 65 | std::vector<Pool> pools; | |
| 62 | std::vector<std::unique_ptr<Pool>> pools; | ||
| 63 | }; | 66 | }; |
| 64 | 67 | ||
| 65 | VKResource::VKResource() = default; | 68 | VKResource::VKResource() = default; |
| 66 | 69 | ||
| 67 | VKResource::~VKResource() = default; | 70 | VKResource::~VKResource() = default; |
| 68 | 71 | ||
| 69 | VKFence::VKFence(const VKDevice& device, UniqueFence handle) | 72 | VKFence::VKFence(const VKDevice& device) |
| 70 | : device{device}, handle{std::move(handle)} {} | 73 | : device{device}, handle{device.GetLogical().CreateFence(BuildFenceCreateInfo())} {} |
| 71 | 74 | ||
| 72 | VKFence::~VKFence() = default; | 75 | VKFence::~VKFence() = default; |
| 73 | 76 | ||
| 74 | void VKFence::Wait() { | 77 | void VKFence::Wait() { |
| 75 | static constexpr u64 timeout = std::numeric_limits<u64>::max(); | 78 | switch (const VkResult result = handle.Wait()) { |
| 76 | const auto dev = device.GetLogical(); | 79 | case VK_SUCCESS: |
| 77 | const auto& dld = device.GetDispatchLoader(); | ||
| 78 | switch (const auto result = dev.waitForFences(1, &*handle, true, timeout, dld)) { | ||
| 79 | case vk::Result::eSuccess: | ||
| 80 | return; | 80 | return; |
| 81 | case vk::Result::eErrorDeviceLost: | 81 | case VK_ERROR_DEVICE_LOST: |
| 82 | device.ReportLoss(); | 82 | device.ReportLoss(); |
| 83 | [[fallthrough]]; | 83 | [[fallthrough]]; |
| 84 | default: | 84 | default: |
| 85 | vk::throwResultException(result, "vk::waitForFences"); | 85 | throw vk::Exception(result); |
| 86 | } | 86 | } |
| 87 | } | 87 | } |
| 88 | 88 | ||
| @@ -107,13 +107,11 @@ bool VKFence::Tick(bool gpu_wait, bool owner_wait) { | |||
| 107 | return false; | 107 | return false; |
| 108 | } | 108 | } |
| 109 | 109 | ||
| 110 | const auto dev = device.GetLogical(); | ||
| 111 | const auto& dld = device.GetDispatchLoader(); | ||
| 112 | if (gpu_wait) { | 110 | if (gpu_wait) { |
| 113 | // Wait for the fence if it has been requested. | 111 | // Wait for the fence if it has been requested. |
| 114 | dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld); | 112 | (void)handle.Wait(); |
| 115 | } else { | 113 | } else { |
| 116 | if (dev.getFenceStatus(*handle, dld) != vk::Result::eSuccess) { | 114 | if (handle.GetStatus() != VK_SUCCESS) { |
| 117 | // Vulkan fence is not ready, not much it can do here | 115 | // Vulkan fence is not ready, not much it can do here |
| 118 | return false; | 116 | return false; |
| 119 | } | 117 | } |
| @@ -126,7 +124,7 @@ bool VKFence::Tick(bool gpu_wait, bool owner_wait) { | |||
| 126 | protected_resources.clear(); | 124 | protected_resources.clear(); |
| 127 | 125 | ||
| 128 | // Prepare fence for reusage. | 126 | // Prepare fence for reusage. |
| 129 | dev.resetFences({*handle}, dld); | 127 | handle.Reset(); |
| 130 | is_used = false; | 128 | is_used = false; |
| 131 | return true; | 129 | return true; |
| 132 | } | 130 | } |
| @@ -299,21 +297,16 @@ VKFence& VKResourceManager::CommitFence() { | |||
| 299 | return *found_fence; | 297 | return *found_fence; |
| 300 | } | 298 | } |
| 301 | 299 | ||
| 302 | vk::CommandBuffer VKResourceManager::CommitCommandBuffer(VKFence& fence) { | 300 | VkCommandBuffer VKResourceManager::CommitCommandBuffer(VKFence& fence) { |
| 303 | return command_buffer_pool->Commit(fence); | 301 | return command_buffer_pool->Commit(fence); |
| 304 | } | 302 | } |
| 305 | 303 | ||
| 306 | void VKResourceManager::GrowFences(std::size_t new_fences_count) { | 304 | void VKResourceManager::GrowFences(std::size_t new_fences_count) { |
| 307 | const auto dev = device.GetLogical(); | ||
| 308 | const auto& dld = device.GetDispatchLoader(); | ||
| 309 | const vk::FenceCreateInfo fence_ci; | ||
| 310 | |||
| 311 | const std::size_t previous_size = fences.size(); | 305 | const std::size_t previous_size = fences.size(); |
| 312 | fences.resize(previous_size + new_fences_count); | 306 | fences.resize(previous_size + new_fences_count); |
| 313 | 307 | ||
| 314 | std::generate(fences.begin() + previous_size, fences.end(), [&]() { | 308 | std::generate(fences.begin() + previous_size, fences.end(), |
| 315 | return std::make_unique<VKFence>(device, dev.createFenceUnique(fence_ci, nullptr, dld)); | 309 | [this] { return std::make_unique<VKFence>(device); }); |
| 316 | }); | ||
| 317 | } | 310 | } |
| 318 | 311 | ||
| 319 | } // namespace Vulkan | 312 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.h b/src/video_core/renderer_vulkan/vk_resource_manager.h index d4cbc95a5..f683d2276 100644 --- a/src/video_core/renderer_vulkan/vk_resource_manager.h +++ b/src/video_core/renderer_vulkan/vk_resource_manager.h | |||
| @@ -7,7 +7,7 @@ | |||
| 7 | #include <cstddef> | 7 | #include <cstddef> |
| 8 | #include <memory> | 8 | #include <memory> |
| 9 | #include <vector> | 9 | #include <vector> |
| 10 | #include "video_core/renderer_vulkan/declarations.h" | 10 | #include "video_core/renderer_vulkan/wrapper.h" |
| 11 | 11 | ||
| 12 | namespace Vulkan { | 12 | namespace Vulkan { |
| 13 | 13 | ||
| @@ -42,7 +42,7 @@ class VKFence { | |||
| 42 | friend class VKResourceManager; | 42 | friend class VKResourceManager; |
| 43 | 43 | ||
| 44 | public: | 44 | public: |
| 45 | explicit VKFence(const VKDevice& device, UniqueFence handle); | 45 | explicit VKFence(const VKDevice& device); |
| 46 | ~VKFence(); | 46 | ~VKFence(); |
| 47 | 47 | ||
| 48 | /** | 48 | /** |
| @@ -69,7 +69,7 @@ public: | |||
| 69 | void RedirectProtection(VKResource* old_resource, VKResource* new_resource) noexcept; | 69 | void RedirectProtection(VKResource* old_resource, VKResource* new_resource) noexcept; |
| 70 | 70 | ||
| 71 | /// Retreives the fence. | 71 | /// Retreives the fence. |
| 72 | operator vk::Fence() const { | 72 | operator VkFence() const { |
| 73 | return *handle; | 73 | return *handle; |
| 74 | } | 74 | } |
| 75 | 75 | ||
| @@ -87,7 +87,7 @@ private: | |||
| 87 | bool Tick(bool gpu_wait, bool owner_wait); | 87 | bool Tick(bool gpu_wait, bool owner_wait); |
| 88 | 88 | ||
| 89 | const VKDevice& device; ///< Device handler | 89 | const VKDevice& device; ///< Device handler |
| 90 | UniqueFence handle; ///< Vulkan fence | 90 | vk::Fence handle; ///< Vulkan fence |
| 91 | std::vector<VKResource*> protected_resources; ///< List of resources protected by this fence | 91 | std::vector<VKResource*> protected_resources; ///< List of resources protected by this fence |
| 92 | bool is_owned = false; ///< The fence has been commited but not released yet. | 92 | bool is_owned = false; ///< The fence has been commited but not released yet. |
| 93 | bool is_used = false; ///< The fence has been commited but it has not been checked to be free. | 93 | bool is_used = false; ///< The fence has been commited but it has not been checked to be free. |
| @@ -181,7 +181,7 @@ public: | |||
| 181 | VKFence& CommitFence(); | 181 | VKFence& CommitFence(); |
| 182 | 182 | ||
| 183 | /// Commits an unused command buffer and protects it with a fence. | 183 | /// Commits an unused command buffer and protects it with a fence. |
| 184 | vk::CommandBuffer CommitCommandBuffer(VKFence& fence); | 184 | VkCommandBuffer CommitCommandBuffer(VKFence& fence); |
| 185 | 185 | ||
| 186 | private: | 186 | private: |
| 187 | /// Allocates new fences. | 187 | /// Allocates new fences. |
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp index 204b7c39c..07bbcf520 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp | |||
| @@ -7,64 +7,64 @@ | |||
| 7 | #include <unordered_map> | 7 | #include <unordered_map> |
| 8 | 8 | ||
| 9 | #include "common/assert.h" | 9 | #include "common/assert.h" |
| 10 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 11 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 10 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 12 | #include "video_core/renderer_vulkan/vk_sampler_cache.h" | 11 | #include "video_core/renderer_vulkan/vk_sampler_cache.h" |
| 12 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 13 | #include "video_core/textures/texture.h" | 13 | #include "video_core/textures/texture.h" |
| 14 | 14 | ||
| 15 | namespace Vulkan { | 15 | namespace Vulkan { |
| 16 | 16 | ||
| 17 | static std::optional<vk::BorderColor> TryConvertBorderColor(std::array<float, 4> color) { | 17 | namespace { |
| 18 | |||
| 19 | VkBorderColor ConvertBorderColor(std::array<float, 4> color) { | ||
| 18 | // TODO(Rodrigo): Manage integer border colors | 20 | // TODO(Rodrigo): Manage integer border colors |
| 19 | if (color == std::array<float, 4>{0, 0, 0, 0}) { | 21 | if (color == std::array<float, 4>{0, 0, 0, 0}) { |
| 20 | return vk::BorderColor::eFloatTransparentBlack; | 22 | return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK; |
| 21 | } else if (color == std::array<float, 4>{0, 0, 0, 1}) { | 23 | } else if (color == std::array<float, 4>{0, 0, 0, 1}) { |
| 22 | return vk::BorderColor::eFloatOpaqueBlack; | 24 | return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK; |
| 23 | } else if (color == std::array<float, 4>{1, 1, 1, 1}) { | 25 | } else if (color == std::array<float, 4>{1, 1, 1, 1}) { |
| 24 | return vk::BorderColor::eFloatOpaqueWhite; | 26 | return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE; |
| 27 | } | ||
| 28 | if (color[0] + color[1] + color[2] > 1.35f) { | ||
| 29 | // If color elements are brighter than roughly 0.5 average, use white border | ||
| 30 | return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE; | ||
| 31 | } else if (color[3] > 0.5f) { | ||
| 32 | return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK; | ||
| 25 | } else { | 33 | } else { |
| 26 | if (color[0] + color[1] + color[2] > 1.35f) { | 34 | return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK; |
| 27 | // If color elements are brighter than roughly 0.5 average, use white border | ||
| 28 | return vk::BorderColor::eFloatOpaqueWhite; | ||
| 29 | } | ||
| 30 | if (color[3] > 0.5f) { | ||
| 31 | return vk::BorderColor::eFloatOpaqueBlack; | ||
| 32 | } | ||
| 33 | return vk::BorderColor::eFloatTransparentBlack; | ||
| 34 | } | 35 | } |
| 35 | } | 36 | } |
| 36 | 37 | ||
| 38 | } // Anonymous namespace | ||
| 39 | |||
| 37 | VKSamplerCache::VKSamplerCache(const VKDevice& device) : device{device} {} | 40 | VKSamplerCache::VKSamplerCache(const VKDevice& device) : device{device} {} |
| 38 | 41 | ||
| 39 | VKSamplerCache::~VKSamplerCache() = default; | 42 | VKSamplerCache::~VKSamplerCache() = default; |
| 40 | 43 | ||
| 41 | UniqueSampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) const { | 44 | vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) const { |
| 42 | const float max_anisotropy{tsc.GetMaxAnisotropy()}; | 45 | VkSamplerCreateInfo ci; |
| 43 | const bool has_anisotropy{max_anisotropy > 1.0f}; | 46 | ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO; |
| 44 | 47 | ci.pNext = nullptr; | |
| 45 | const auto border_color{tsc.GetBorderColor()}; | 48 | ci.flags = 0; |
| 46 | const auto vk_border_color{TryConvertBorderColor(border_color)}; | 49 | ci.magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter); |
| 47 | 50 | ci.minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter); | |
| 48 | constexpr bool unnormalized_coords{false}; | 51 | ci.mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter); |
| 49 | 52 | ci.addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter); | |
| 50 | const vk::SamplerCreateInfo sampler_ci( | 53 | ci.addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter); |
| 51 | {}, MaxwellToVK::Sampler::Filter(tsc.mag_filter), | 54 | ci.addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter); |
| 52 | MaxwellToVK::Sampler::Filter(tsc.min_filter), | 55 | ci.mipLodBias = tsc.GetLodBias(); |
| 53 | MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter), | 56 | ci.anisotropyEnable = tsc.GetMaxAnisotropy() > 1.0f ? VK_TRUE : VK_FALSE; |
| 54 | MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter), | 57 | ci.maxAnisotropy = tsc.GetMaxAnisotropy(); |
| 55 | MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter), | 58 | ci.compareEnable = tsc.depth_compare_enabled; |
| 56 | MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter), tsc.GetLodBias(), | 59 | ci.compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func); |
| 57 | has_anisotropy, max_anisotropy, tsc.depth_compare_enabled, | 60 | ci.minLod = tsc.GetMinLod(); |
| 58 | MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func), tsc.GetMinLod(), | 61 | ci.maxLod = tsc.GetMaxLod(); |
| 59 | tsc.GetMaxLod(), vk_border_color.value_or(vk::BorderColor::eFloatTransparentBlack), | 62 | ci.borderColor = ConvertBorderColor(tsc.GetBorderColor()); |
| 60 | unnormalized_coords); | 63 | ci.unnormalizedCoordinates = VK_FALSE; |
| 61 | 64 | return device.GetLogical().CreateSampler(ci); | |
| 62 | const auto& dld{device.GetDispatchLoader()}; | ||
| 63 | const auto dev{device.GetLogical()}; | ||
| 64 | return dev.createSamplerUnique(sampler_ci, nullptr, dld); | ||
| 65 | } | 65 | } |
| 66 | 66 | ||
| 67 | vk::Sampler VKSamplerCache::ToSamplerType(const UniqueSampler& sampler) const { | 67 | VkSampler VKSamplerCache::ToSamplerType(const vk::Sampler& sampler) const { |
| 68 | return *sampler; | 68 | return *sampler; |
| 69 | } | 69 | } |
| 70 | 70 | ||
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.h b/src/video_core/renderer_vulkan/vk_sampler_cache.h index 1f73b716b..a33d1c0ee 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.h +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.h | |||
| @@ -4,7 +4,7 @@ | |||
| 4 | 4 | ||
| 5 | #pragma once | 5 | #pragma once |
| 6 | 6 | ||
| 7 | #include "video_core/renderer_vulkan/declarations.h" | 7 | #include "video_core/renderer_vulkan/wrapper.h" |
| 8 | #include "video_core/sampler_cache.h" | 8 | #include "video_core/sampler_cache.h" |
| 9 | #include "video_core/textures/texture.h" | 9 | #include "video_core/textures/texture.h" |
| 10 | 10 | ||
| @@ -12,15 +12,15 @@ namespace Vulkan { | |||
| 12 | 12 | ||
| 13 | class VKDevice; | 13 | class VKDevice; |
| 14 | 14 | ||
| 15 | class VKSamplerCache final : public VideoCommon::SamplerCache<vk::Sampler, UniqueSampler> { | 15 | class VKSamplerCache final : public VideoCommon::SamplerCache<VkSampler, vk::Sampler> { |
| 16 | public: | 16 | public: |
| 17 | explicit VKSamplerCache(const VKDevice& device); | 17 | explicit VKSamplerCache(const VKDevice& device); |
| 18 | ~VKSamplerCache(); | 18 | ~VKSamplerCache(); |
| 19 | 19 | ||
| 20 | protected: | 20 | protected: |
| 21 | UniqueSampler CreateSampler(const Tegra::Texture::TSCEntry& tsc) const override; | 21 | vk::Sampler CreateSampler(const Tegra::Texture::TSCEntry& tsc) const override; |
| 22 | 22 | ||
| 23 | vk::Sampler ToSamplerType(const UniqueSampler& sampler) const override; | 23 | VkSampler ToSamplerType(const vk::Sampler& sampler) const override; |
| 24 | 24 | ||
| 25 | private: | 25 | private: |
| 26 | const VKDevice& device; | 26 | const VKDevice& device; |
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp index b61d4fe63..900f551b3 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.cpp +++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp | |||
| @@ -10,23 +10,22 @@ | |||
| 10 | 10 | ||
| 11 | #include "common/assert.h" | 11 | #include "common/assert.h" |
| 12 | #include "common/microprofile.h" | 12 | #include "common/microprofile.h" |
| 13 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 14 | #include "video_core/renderer_vulkan/vk_device.h" | 13 | #include "video_core/renderer_vulkan/vk_device.h" |
| 15 | #include "video_core/renderer_vulkan/vk_query_cache.h" | 14 | #include "video_core/renderer_vulkan/vk_query_cache.h" |
| 16 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 15 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 17 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 16 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 18 | #include "video_core/renderer_vulkan/vk_state_tracker.h" | 17 | #include "video_core/renderer_vulkan/vk_state_tracker.h" |
| 18 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 19 | 19 | ||
| 20 | namespace Vulkan { | 20 | namespace Vulkan { |
| 21 | 21 | ||
| 22 | MICROPROFILE_DECLARE(Vulkan_WaitForWorker); | 22 | MICROPROFILE_DECLARE(Vulkan_WaitForWorker); |
| 23 | 23 | ||
| 24 | void VKScheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf, | 24 | void VKScheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf) { |
| 25 | const vk::DispatchLoaderDynamic& dld) { | ||
| 26 | auto command = first; | 25 | auto command = first; |
| 27 | while (command != nullptr) { | 26 | while (command != nullptr) { |
| 28 | auto next = command->GetNext(); | 27 | auto next = command->GetNext(); |
| 29 | command->Execute(cmdbuf, dld); | 28 | command->Execute(cmdbuf); |
| 30 | command->~Command(); | 29 | command->~Command(); |
| 31 | command = next; | 30 | command = next; |
| 32 | } | 31 | } |
| @@ -51,7 +50,7 @@ VKScheduler::~VKScheduler() { | |||
| 51 | worker_thread.join(); | 50 | worker_thread.join(); |
| 52 | } | 51 | } |
| 53 | 52 | ||
| 54 | void VKScheduler::Flush(bool release_fence, vk::Semaphore semaphore) { | 53 | void VKScheduler::Flush(bool release_fence, VkSemaphore semaphore) { |
| 55 | SubmitExecution(semaphore); | 54 | SubmitExecution(semaphore); |
| 56 | if (release_fence) { | 55 | if (release_fence) { |
| 57 | current_fence->Release(); | 56 | current_fence->Release(); |
| @@ -59,7 +58,7 @@ void VKScheduler::Flush(bool release_fence, vk::Semaphore semaphore) { | |||
| 59 | AllocateNewContext(); | 58 | AllocateNewContext(); |
| 60 | } | 59 | } |
| 61 | 60 | ||
| 62 | void VKScheduler::Finish(bool release_fence, vk::Semaphore semaphore) { | 61 | void VKScheduler::Finish(bool release_fence, VkSemaphore semaphore) { |
| 63 | SubmitExecution(semaphore); | 62 | SubmitExecution(semaphore); |
| 64 | current_fence->Wait(); | 63 | current_fence->Wait(); |
| 65 | if (release_fence) { | 64 | if (release_fence) { |
| @@ -89,17 +88,34 @@ void VKScheduler::DispatchWork() { | |||
| 89 | AcquireNewChunk(); | 88 | AcquireNewChunk(); |
| 90 | } | 89 | } |
| 91 | 90 | ||
| 92 | void VKScheduler::RequestRenderpass(const vk::RenderPassBeginInfo& renderpass_bi) { | 91 | void VKScheduler::RequestRenderpass(VkRenderPass renderpass, VkFramebuffer framebuffer, |
| 93 | if (state.renderpass && renderpass_bi == *state.renderpass) { | 92 | VkExtent2D render_area) { |
| 93 | if (renderpass == state.renderpass && framebuffer == state.framebuffer && | ||
| 94 | render_area.width == state.render_area.width && | ||
| 95 | render_area.height == state.render_area.height) { | ||
| 94 | return; | 96 | return; |
| 95 | } | 97 | } |
| 96 | const bool end_renderpass = state.renderpass.has_value(); | 98 | const bool end_renderpass = state.renderpass != nullptr; |
| 97 | state.renderpass = renderpass_bi; | 99 | state.renderpass = renderpass; |
| 98 | Record([renderpass_bi, end_renderpass](auto cmdbuf, auto& dld) { | 100 | state.framebuffer = framebuffer; |
| 101 | state.render_area = render_area; | ||
| 102 | |||
| 103 | VkRenderPassBeginInfo renderpass_bi; | ||
| 104 | renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; | ||
| 105 | renderpass_bi.pNext = nullptr; | ||
| 106 | renderpass_bi.renderPass = renderpass; | ||
| 107 | renderpass_bi.framebuffer = framebuffer; | ||
| 108 | renderpass_bi.renderArea.offset.x = 0; | ||
| 109 | renderpass_bi.renderArea.offset.y = 0; | ||
| 110 | renderpass_bi.renderArea.extent = render_area; | ||
| 111 | renderpass_bi.clearValueCount = 0; | ||
| 112 | renderpass_bi.pClearValues = nullptr; | ||
| 113 | |||
| 114 | Record([renderpass_bi, end_renderpass](vk::CommandBuffer cmdbuf) { | ||
| 99 | if (end_renderpass) { | 115 | if (end_renderpass) { |
| 100 | cmdbuf.endRenderPass(dld); | 116 | cmdbuf.EndRenderPass(); |
| 101 | } | 117 | } |
| 102 | cmdbuf.beginRenderPass(renderpass_bi, vk::SubpassContents::eInline, dld); | 118 | cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE); |
| 103 | }); | 119 | }); |
| 104 | } | 120 | } |
| 105 | 121 | ||
| @@ -107,13 +123,13 @@ void VKScheduler::RequestOutsideRenderPassOperationContext() { | |||
| 107 | EndRenderPass(); | 123 | EndRenderPass(); |
| 108 | } | 124 | } |
| 109 | 125 | ||
| 110 | void VKScheduler::BindGraphicsPipeline(vk::Pipeline pipeline) { | 126 | void VKScheduler::BindGraphicsPipeline(VkPipeline pipeline) { |
| 111 | if (state.graphics_pipeline == pipeline) { | 127 | if (state.graphics_pipeline == pipeline) { |
| 112 | return; | 128 | return; |
| 113 | } | 129 | } |
| 114 | state.graphics_pipeline = pipeline; | 130 | state.graphics_pipeline = pipeline; |
| 115 | Record([pipeline](auto cmdbuf, auto& dld) { | 131 | Record([pipeline](vk::CommandBuffer cmdbuf) { |
| 116 | cmdbuf.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline, dld); | 132 | cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); |
| 117 | }); | 133 | }); |
| 118 | } | 134 | } |
| 119 | 135 | ||
| @@ -126,37 +142,50 @@ void VKScheduler::WorkerThread() { | |||
| 126 | } | 142 | } |
| 127 | auto extracted_chunk = std::move(chunk_queue.Front()); | 143 | auto extracted_chunk = std::move(chunk_queue.Front()); |
| 128 | chunk_queue.Pop(); | 144 | chunk_queue.Pop(); |
| 129 | extracted_chunk->ExecuteAll(current_cmdbuf, device.GetDispatchLoader()); | 145 | extracted_chunk->ExecuteAll(current_cmdbuf); |
| 130 | chunk_reserve.Push(std::move(extracted_chunk)); | 146 | chunk_reserve.Push(std::move(extracted_chunk)); |
| 131 | } while (!quit); | 147 | } while (!quit); |
| 132 | } | 148 | } |
| 133 | 149 | ||
| 134 | void VKScheduler::SubmitExecution(vk::Semaphore semaphore) { | 150 | void VKScheduler::SubmitExecution(VkSemaphore semaphore) { |
| 135 | EndPendingOperations(); | 151 | EndPendingOperations(); |
| 136 | InvalidateState(); | 152 | InvalidateState(); |
| 137 | WaitWorker(); | 153 | WaitWorker(); |
| 138 | 154 | ||
| 139 | std::unique_lock lock{mutex}; | 155 | std::unique_lock lock{mutex}; |
| 140 | 156 | ||
| 141 | const auto queue = device.GetGraphicsQueue(); | 157 | current_cmdbuf.End(); |
| 142 | const auto& dld = device.GetDispatchLoader(); | ||
| 143 | current_cmdbuf.end(dld); | ||
| 144 | 158 | ||
| 145 | const vk::SubmitInfo submit_info(0, nullptr, nullptr, 1, ¤t_cmdbuf, semaphore ? 1U : 0U, | 159 | VkSubmitInfo submit_info; |
| 146 | &semaphore); | 160 | submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; |
| 147 | queue.submit({submit_info}, static_cast<vk::Fence>(*current_fence), dld); | 161 | submit_info.pNext = nullptr; |
| 162 | submit_info.waitSemaphoreCount = 0; | ||
| 163 | submit_info.pWaitSemaphores = nullptr; | ||
| 164 | submit_info.pWaitDstStageMask = nullptr; | ||
| 165 | submit_info.commandBufferCount = 1; | ||
| 166 | submit_info.pCommandBuffers = current_cmdbuf.address(); | ||
| 167 | submit_info.signalSemaphoreCount = semaphore ? 1 : 0; | ||
| 168 | submit_info.pSignalSemaphores = &semaphore; | ||
| 169 | device.GetGraphicsQueue().Submit(submit_info, *current_fence); | ||
| 148 | } | 170 | } |
| 149 | 171 | ||
| 150 | void VKScheduler::AllocateNewContext() { | 172 | void VKScheduler::AllocateNewContext() { |
| 151 | ++ticks; | 173 | ++ticks; |
| 152 | 174 | ||
| 175 | VkCommandBufferBeginInfo cmdbuf_bi; | ||
| 176 | cmdbuf_bi.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; | ||
| 177 | cmdbuf_bi.pNext = nullptr; | ||
| 178 | cmdbuf_bi.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT; | ||
| 179 | cmdbuf_bi.pInheritanceInfo = nullptr; | ||
| 180 | |||
| 153 | std::unique_lock lock{mutex}; | 181 | std::unique_lock lock{mutex}; |
| 154 | current_fence = next_fence; | 182 | current_fence = next_fence; |
| 155 | next_fence = &resource_manager.CommitFence(); | 183 | next_fence = &resource_manager.CommitFence(); |
| 156 | 184 | ||
| 157 | current_cmdbuf = resource_manager.CommitCommandBuffer(*current_fence); | 185 | current_cmdbuf = vk::CommandBuffer(resource_manager.CommitCommandBuffer(*current_fence), |
| 158 | current_cmdbuf.begin({vk::CommandBufferUsageFlagBits::eOneTimeSubmit}, | 186 | device.GetDispatchLoader()); |
| 159 | device.GetDispatchLoader()); | 187 | current_cmdbuf.Begin(cmdbuf_bi); |
| 188 | |||
| 160 | // Enable counters once again. These are disabled when a command buffer is finished. | 189 | // Enable counters once again. These are disabled when a command buffer is finished. |
| 161 | if (query_cache) { | 190 | if (query_cache) { |
| 162 | query_cache->UpdateCounters(); | 191 | query_cache->UpdateCounters(); |
| @@ -177,8 +206,8 @@ void VKScheduler::EndRenderPass() { | |||
| 177 | if (!state.renderpass) { | 206 | if (!state.renderpass) { |
| 178 | return; | 207 | return; |
| 179 | } | 208 | } |
| 180 | state.renderpass = std::nullopt; | 209 | state.renderpass = nullptr; |
| 181 | Record([](auto cmdbuf, auto& dld) { cmdbuf.endRenderPass(dld); }); | 210 | Record([](vk::CommandBuffer cmdbuf) { cmdbuf.EndRenderPass(); }); |
| 182 | } | 211 | } |
| 183 | 212 | ||
| 184 | void VKScheduler::AcquireNewChunk() { | 213 | void VKScheduler::AcquireNewChunk() { |
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h index c7cc291c3..82a8adc69 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.h +++ b/src/video_core/renderer_vulkan/vk_scheduler.h | |||
| @@ -13,7 +13,7 @@ | |||
| 13 | #include <utility> | 13 | #include <utility> |
| 14 | #include "common/common_types.h" | 14 | #include "common/common_types.h" |
| 15 | #include "common/threadsafe_queue.h" | 15 | #include "common/threadsafe_queue.h" |
| 16 | #include "video_core/renderer_vulkan/declarations.h" | 16 | #include "video_core/renderer_vulkan/wrapper.h" |
| 17 | 17 | ||
| 18 | namespace Vulkan { | 18 | namespace Vulkan { |
| 19 | 19 | ||
| @@ -49,10 +49,10 @@ public: | |||
| 49 | ~VKScheduler(); | 49 | ~VKScheduler(); |
| 50 | 50 | ||
| 51 | /// Sends the current execution context to the GPU. | 51 | /// Sends the current execution context to the GPU. |
| 52 | void Flush(bool release_fence = true, vk::Semaphore semaphore = nullptr); | 52 | void Flush(bool release_fence = true, VkSemaphore semaphore = nullptr); |
| 53 | 53 | ||
| 54 | /// Sends the current execution context to the GPU and waits for it to complete. | 54 | /// Sends the current execution context to the GPU and waits for it to complete. |
| 55 | void Finish(bool release_fence = true, vk::Semaphore semaphore = nullptr); | 55 | void Finish(bool release_fence = true, VkSemaphore semaphore = nullptr); |
| 56 | 56 | ||
| 57 | /// Waits for the worker thread to finish executing everything. After this function returns it's | 57 | /// Waits for the worker thread to finish executing everything. After this function returns it's |
| 58 | /// safe to touch worker resources. | 58 | /// safe to touch worker resources. |
| @@ -62,14 +62,15 @@ public: | |||
| 62 | void DispatchWork(); | 62 | void DispatchWork(); |
| 63 | 63 | ||
| 64 | /// Requests to begin a renderpass. | 64 | /// Requests to begin a renderpass. |
| 65 | void RequestRenderpass(const vk::RenderPassBeginInfo& renderpass_bi); | 65 | void RequestRenderpass(VkRenderPass renderpass, VkFramebuffer framebuffer, |
| 66 | VkExtent2D render_area); | ||
| 66 | 67 | ||
| 67 | /// Requests the current executino context to be able to execute operations only allowed outside | 68 | /// Requests the current executino context to be able to execute operations only allowed outside |
| 68 | /// of a renderpass. | 69 | /// of a renderpass. |
| 69 | void RequestOutsideRenderPassOperationContext(); | 70 | void RequestOutsideRenderPassOperationContext(); |
| 70 | 71 | ||
| 71 | /// Binds a pipeline to the current execution context. | 72 | /// Binds a pipeline to the current execution context. |
| 72 | void BindGraphicsPipeline(vk::Pipeline pipeline); | 73 | void BindGraphicsPipeline(VkPipeline pipeline); |
| 73 | 74 | ||
| 74 | /// Assigns the query cache. | 75 | /// Assigns the query cache. |
| 75 | void SetQueryCache(VKQueryCache& query_cache_) { | 76 | void SetQueryCache(VKQueryCache& query_cache_) { |
| @@ -101,8 +102,7 @@ private: | |||
| 101 | public: | 102 | public: |
| 102 | virtual ~Command() = default; | 103 | virtual ~Command() = default; |
| 103 | 104 | ||
| 104 | virtual void Execute(vk::CommandBuffer cmdbuf, | 105 | virtual void Execute(vk::CommandBuffer cmdbuf) const = 0; |
| 105 | const vk::DispatchLoaderDynamic& dld) const = 0; | ||
| 106 | 106 | ||
| 107 | Command* GetNext() const { | 107 | Command* GetNext() const { |
| 108 | return next; | 108 | return next; |
| @@ -125,9 +125,8 @@ private: | |||
| 125 | TypedCommand(TypedCommand&&) = delete; | 125 | TypedCommand(TypedCommand&&) = delete; |
| 126 | TypedCommand& operator=(TypedCommand&&) = delete; | 126 | TypedCommand& operator=(TypedCommand&&) = delete; |
| 127 | 127 | ||
| 128 | void Execute(vk::CommandBuffer cmdbuf, | 128 | void Execute(vk::CommandBuffer cmdbuf) const override { |
| 129 | const vk::DispatchLoaderDynamic& dld) const override { | 129 | command(cmdbuf); |
| 130 | command(cmdbuf, dld); | ||
| 131 | } | 130 | } |
| 132 | 131 | ||
| 133 | private: | 132 | private: |
| @@ -136,7 +135,7 @@ private: | |||
| 136 | 135 | ||
| 137 | class CommandChunk final { | 136 | class CommandChunk final { |
| 138 | public: | 137 | public: |
| 139 | void ExecuteAll(vk::CommandBuffer cmdbuf, const vk::DispatchLoaderDynamic& dld); | 138 | void ExecuteAll(vk::CommandBuffer cmdbuf); |
| 140 | 139 | ||
| 141 | template <typename T> | 140 | template <typename T> |
| 142 | bool Record(T& command) { | 141 | bool Record(T& command) { |
| @@ -175,7 +174,7 @@ private: | |||
| 175 | 174 | ||
| 176 | void WorkerThread(); | 175 | void WorkerThread(); |
| 177 | 176 | ||
| 178 | void SubmitExecution(vk::Semaphore semaphore); | 177 | void SubmitExecution(VkSemaphore semaphore); |
| 179 | 178 | ||
| 180 | void AllocateNewContext(); | 179 | void AllocateNewContext(); |
| 181 | 180 | ||
| @@ -198,8 +197,10 @@ private: | |||
| 198 | VKFence* next_fence = nullptr; | 197 | VKFence* next_fence = nullptr; |
| 199 | 198 | ||
| 200 | struct State { | 199 | struct State { |
| 201 | std::optional<vk::RenderPassBeginInfo> renderpass; | 200 | VkRenderPass renderpass = nullptr; |
| 202 | vk::Pipeline graphics_pipeline; | 201 | VkFramebuffer framebuffer = nullptr; |
| 202 | VkExtent2D render_area = {0, 0}; | ||
| 203 | VkPipeline graphics_pipeline = nullptr; | ||
| 203 | } state; | 204 | } state; |
| 204 | 205 | ||
| 205 | std::unique_ptr<CommandChunk> chunk; | 206 | std::unique_ptr<CommandChunk> chunk; |
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp index b9f9e2714..62e4ca488 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp | |||
| @@ -801,7 +801,7 @@ private: | |||
| 801 | if (IsOutputAttributeArray()) { | 801 | if (IsOutputAttributeArray()) { |
| 802 | const u32 num = GetNumOutputVertices(); | 802 | const u32 num = GetNumOutputVertices(); |
| 803 | type = TypeArray(type, Constant(t_uint, num)); | 803 | type = TypeArray(type, Constant(t_uint, num)); |
| 804 | if (device.GetDriverID() != vk::DriverIdKHR::eIntelProprietaryWindows) { | 804 | if (device.GetDriverID() != VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS_KHR) { |
| 805 | // Intel's proprietary driver fails to setup defaults for arrayed output | 805 | // Intel's proprietary driver fails to setup defaults for arrayed output |
| 806 | // attributes. | 806 | // attributes. |
| 807 | varying_default = ConstantComposite(type, std::vector(num, varying_default)); | 807 | varying_default = ConstantComposite(type, std::vector(num, varying_default)); |
diff --git a/src/video_core/renderer_vulkan/vk_shader_util.cpp b/src/video_core/renderer_vulkan/vk_shader_util.cpp index b97c4cb3d..784839327 100644 --- a/src/video_core/renderer_vulkan/vk_shader_util.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_util.cpp | |||
| @@ -8,27 +8,25 @@ | |||
| 8 | #include "common/alignment.h" | 8 | #include "common/alignment.h" |
| 9 | #include "common/assert.h" | 9 | #include "common/assert.h" |
| 10 | #include "common/common_types.h" | 10 | #include "common/common_types.h" |
| 11 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 12 | #include "video_core/renderer_vulkan/vk_device.h" | 11 | #include "video_core/renderer_vulkan/vk_device.h" |
| 13 | #include "video_core/renderer_vulkan/vk_shader_util.h" | 12 | #include "video_core/renderer_vulkan/vk_shader_util.h" |
| 13 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 14 | 14 | ||
| 15 | namespace Vulkan { | 15 | namespace Vulkan { |
| 16 | 16 | ||
| 17 | UniqueShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data) { | 17 | vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data) { |
| 18 | // Avoid undefined behavior by copying to a staging allocation | 18 | // Avoid undefined behavior by copying to a staging allocation |
| 19 | ASSERT(code_size % sizeof(u32) == 0); | 19 | ASSERT(code_size % sizeof(u32) == 0); |
| 20 | const auto data = std::make_unique<u32[]>(code_size / sizeof(u32)); | 20 | const auto data = std::make_unique<u32[]>(code_size / sizeof(u32)); |
| 21 | std::memcpy(data.get(), code_data, code_size); | 21 | std::memcpy(data.get(), code_data, code_size); |
| 22 | 22 | ||
| 23 | const auto dev = device.GetLogical(); | 23 | VkShaderModuleCreateInfo ci; |
| 24 | const auto& dld = device.GetDispatchLoader(); | 24 | ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; |
| 25 | const vk::ShaderModuleCreateInfo shader_ci({}, code_size, data.get()); | 25 | ci.pNext = nullptr; |
| 26 | vk::ShaderModule shader_module; | 26 | ci.flags = 0; |
| 27 | if (dev.createShaderModule(&shader_ci, nullptr, &shader_module, dld) != vk::Result::eSuccess) { | 27 | ci.codeSize = code_size; |
| 28 | UNREACHABLE_MSG("Shader module failed to build!"); | 28 | ci.pCode = data.get(); |
| 29 | } | 29 | return device.GetLogical().CreateShaderModule(ci); |
| 30 | |||
| 31 | return UniqueShaderModule(shader_module, vk::ObjectDestroy(dev, nullptr, dld)); | ||
| 32 | } | 30 | } |
| 33 | 31 | ||
| 34 | } // namespace Vulkan | 32 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_shader_util.h b/src/video_core/renderer_vulkan/vk_shader_util.h index c06d65970..be38d6697 100644 --- a/src/video_core/renderer_vulkan/vk_shader_util.h +++ b/src/video_core/renderer_vulkan/vk_shader_util.h | |||
| @@ -6,12 +6,12 @@ | |||
| 6 | 6 | ||
| 7 | #include <vector> | 7 | #include <vector> |
| 8 | #include "common/common_types.h" | 8 | #include "common/common_types.h" |
| 9 | #include "video_core/renderer_vulkan/declarations.h" | 9 | #include "video_core/renderer_vulkan/wrapper.h" |
| 10 | 10 | ||
| 11 | namespace Vulkan { | 11 | namespace Vulkan { |
| 12 | 12 | ||
| 13 | class VKDevice; | 13 | class VKDevice; |
| 14 | 14 | ||
| 15 | UniqueShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data); | 15 | vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data); |
| 16 | 16 | ||
| 17 | } // namespace Vulkan | 17 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp index 374959f82..94d954d7a 100644 --- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp | |||
| @@ -13,6 +13,7 @@ | |||
| 13 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 13 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 14 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 14 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 15 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" | 15 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" |
| 16 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 16 | 17 | ||
| 17 | namespace Vulkan { | 18 | namespace Vulkan { |
| 18 | 19 | ||
| @@ -71,17 +72,23 @@ VKBuffer* VKStagingBufferPool::TryGetReservedBuffer(std::size_t size, bool host_ | |||
| 71 | } | 72 | } |
| 72 | 73 | ||
| 73 | VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_visible) { | 74 | VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_visible) { |
| 74 | const auto usage = | ||
| 75 | vk::BufferUsageFlagBits::eTransferSrc | vk::BufferUsageFlagBits::eTransferDst | | ||
| 76 | vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer | | ||
| 77 | vk::BufferUsageFlagBits::eIndexBuffer; | ||
| 78 | const u32 log2 = Common::Log2Ceil64(size); | 75 | const u32 log2 = Common::Log2Ceil64(size); |
| 79 | const vk::BufferCreateInfo buffer_ci({}, 1ULL << log2, usage, vk::SharingMode::eExclusive, 0, | 76 | |
| 80 | nullptr); | 77 | VkBufferCreateInfo ci; |
| 81 | const auto dev = device.GetLogical(); | 78 | ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; |
| 79 | ci.pNext = nullptr; | ||
| 80 | ci.flags = 0; | ||
| 81 | ci.size = 1ULL << log2; | ||
| 82 | ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | | ||
| 83 | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | | ||
| 84 | VK_BUFFER_USAGE_INDEX_BUFFER_BIT; | ||
| 85 | ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; | ||
| 86 | ci.queueFamilyIndexCount = 0; | ||
| 87 | ci.pQueueFamilyIndices = nullptr; | ||
| 88 | |||
| 82 | auto buffer = std::make_unique<VKBuffer>(); | 89 | auto buffer = std::make_unique<VKBuffer>(); |
| 83 | buffer->handle = dev.createBufferUnique(buffer_ci, nullptr, device.GetDispatchLoader()); | 90 | buffer->handle = device.GetLogical().CreateBuffer(ci); |
| 84 | buffer->commit = memory_manager.Commit(*buffer->handle, host_visible); | 91 | buffer->commit = memory_manager.Commit(buffer->handle, host_visible); |
| 85 | 92 | ||
| 86 | auto& entries = GetCache(host_visible)[log2].entries; | 93 | auto& entries = GetCache(host_visible)[log2].entries; |
| 87 | return *entries.emplace_back(std::move(buffer), scheduler.GetFence(), epoch).buffer; | 94 | return *entries.emplace_back(std::move(buffer), scheduler.GetFence(), epoch).buffer; |
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h index 4d9488f49..a0840ff8c 100644 --- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h +++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h | |||
| @@ -11,9 +11,9 @@ | |||
| 11 | 11 | ||
| 12 | #include "common/common_types.h" | 12 | #include "common/common_types.h" |
| 13 | 13 | ||
| 14 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 15 | #include "video_core/renderer_vulkan/vk_memory_manager.h" | 14 | #include "video_core/renderer_vulkan/vk_memory_manager.h" |
| 16 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 15 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 16 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 17 | 17 | ||
| 18 | namespace Vulkan { | 18 | namespace Vulkan { |
| 19 | 19 | ||
| @@ -22,7 +22,7 @@ class VKFenceWatch; | |||
| 22 | class VKScheduler; | 22 | class VKScheduler; |
| 23 | 23 | ||
| 24 | struct VKBuffer final { | 24 | struct VKBuffer final { |
| 25 | UniqueBuffer handle; | 25 | vk::Buffer handle; |
| 26 | VKMemoryCommit commit; | 26 | VKMemoryCommit commit; |
| 27 | }; | 27 | }; |
| 28 | 28 | ||
diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp index d48d3b44c..38a93a01a 100644 --- a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp +++ b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp | |||
| @@ -9,11 +9,11 @@ | |||
| 9 | 9 | ||
| 10 | #include "common/alignment.h" | 10 | #include "common/alignment.h" |
| 11 | #include "common/assert.h" | 11 | #include "common/assert.h" |
| 12 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 13 | #include "video_core/renderer_vulkan/vk_device.h" | 12 | #include "video_core/renderer_vulkan/vk_device.h" |
| 14 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 13 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 15 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 14 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 16 | #include "video_core/renderer_vulkan/vk_stream_buffer.h" | 15 | #include "video_core/renderer_vulkan/vk_stream_buffer.h" |
| 16 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 17 | 17 | ||
| 18 | namespace Vulkan { | 18 | namespace Vulkan { |
| 19 | 19 | ||
| @@ -25,8 +25,8 @@ constexpr u64 WATCHES_RESERVE_CHUNK = 0x1000; | |||
| 25 | constexpr u64 STREAM_BUFFER_SIZE = 256 * 1024 * 1024; | 25 | constexpr u64 STREAM_BUFFER_SIZE = 256 * 1024 * 1024; |
| 26 | 26 | ||
| 27 | std::optional<u32> FindMemoryType(const VKDevice& device, u32 filter, | 27 | std::optional<u32> FindMemoryType(const VKDevice& device, u32 filter, |
| 28 | vk::MemoryPropertyFlags wanted) { | 28 | VkMemoryPropertyFlags wanted) { |
| 29 | const auto properties = device.GetPhysical().getMemoryProperties(device.GetDispatchLoader()); | 29 | const auto properties = device.GetPhysical().GetMemoryProperties(); |
| 30 | for (u32 i = 0; i < properties.memoryTypeCount; i++) { | 30 | for (u32 i = 0; i < properties.memoryTypeCount; i++) { |
| 31 | if (!(filter & (1 << i))) { | 31 | if (!(filter & (1 << i))) { |
| 32 | continue; | 32 | continue; |
| @@ -35,13 +35,13 @@ std::optional<u32> FindMemoryType(const VKDevice& device, u32 filter, | |||
| 35 | return i; | 35 | return i; |
| 36 | } | 36 | } |
| 37 | } | 37 | } |
| 38 | return {}; | 38 | return std::nullopt; |
| 39 | } | 39 | } |
| 40 | 40 | ||
| 41 | } // Anonymous namespace | 41 | } // Anonymous namespace |
| 42 | 42 | ||
| 43 | VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, | 43 | VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, |
| 44 | vk::BufferUsageFlags usage) | 44 | VkBufferUsageFlags usage) |
| 45 | : device{device}, scheduler{scheduler} { | 45 | : device{device}, scheduler{scheduler} { |
| 46 | CreateBuffers(usage); | 46 | CreateBuffers(usage); |
| 47 | ReserveWatches(current_watches, WATCHES_INITIAL_RESERVE); | 47 | ReserveWatches(current_watches, WATCHES_INITIAL_RESERVE); |
| @@ -78,17 +78,13 @@ std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) { | |||
| 78 | invalidated = true; | 78 | invalidated = true; |
| 79 | } | 79 | } |
| 80 | 80 | ||
| 81 | const auto dev = device.GetLogical(); | 81 | return {memory.Map(offset, size), offset, invalidated}; |
| 82 | const auto& dld = device.GetDispatchLoader(); | ||
| 83 | const auto pointer = reinterpret_cast<u8*>(dev.mapMemory(*memory, offset, size, {}, dld)); | ||
| 84 | return {pointer, offset, invalidated}; | ||
| 85 | } | 82 | } |
| 86 | 83 | ||
| 87 | void VKStreamBuffer::Unmap(u64 size) { | 84 | void VKStreamBuffer::Unmap(u64 size) { |
| 88 | ASSERT_MSG(size <= mapped_size, "Reserved size is too small"); | 85 | ASSERT_MSG(size <= mapped_size, "Reserved size is too small"); |
| 89 | 86 | ||
| 90 | const auto dev = device.GetLogical(); | 87 | memory.Unmap(); |
| 91 | dev.unmapMemory(*memory, device.GetDispatchLoader()); | ||
| 92 | 88 | ||
| 93 | offset += size; | 89 | offset += size; |
| 94 | 90 | ||
| @@ -101,30 +97,42 @@ void VKStreamBuffer::Unmap(u64 size) { | |||
| 101 | watch.fence.Watch(scheduler.GetFence()); | 97 | watch.fence.Watch(scheduler.GetFence()); |
| 102 | } | 98 | } |
| 103 | 99 | ||
| 104 | void VKStreamBuffer::CreateBuffers(vk::BufferUsageFlags usage) { | 100 | void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) { |
| 105 | const vk::BufferCreateInfo buffer_ci({}, STREAM_BUFFER_SIZE, usage, vk::SharingMode::eExclusive, | 101 | VkBufferCreateInfo buffer_ci; |
| 106 | 0, nullptr); | 102 | buffer_ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; |
| 107 | const auto dev = device.GetLogical(); | 103 | buffer_ci.pNext = nullptr; |
| 108 | const auto& dld = device.GetDispatchLoader(); | 104 | buffer_ci.flags = 0; |
| 109 | buffer = dev.createBufferUnique(buffer_ci, nullptr, dld); | 105 | buffer_ci.size = STREAM_BUFFER_SIZE; |
| 106 | buffer_ci.usage = usage; | ||
| 107 | buffer_ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; | ||
| 108 | buffer_ci.queueFamilyIndexCount = 0; | ||
| 109 | buffer_ci.pQueueFamilyIndices = nullptr; | ||
| 110 | |||
| 111 | const auto& dev = device.GetLogical(); | ||
| 112 | buffer = dev.CreateBuffer(buffer_ci); | ||
| 110 | 113 | ||
| 111 | const auto requirements = dev.getBufferMemoryRequirements(*buffer, dld); | 114 | const auto& dld = device.GetDispatchLoader(); |
| 115 | const auto requirements = dev.GetBufferMemoryRequirements(*buffer); | ||
| 112 | // Prefer device local host visible allocations (this should hit AMD's pinned memory). | 116 | // Prefer device local host visible allocations (this should hit AMD's pinned memory). |
| 113 | auto type = FindMemoryType(device, requirements.memoryTypeBits, | 117 | auto type = |
| 114 | vk::MemoryPropertyFlagBits::eHostVisible | | 118 | FindMemoryType(device, requirements.memoryTypeBits, |
| 115 | vk::MemoryPropertyFlagBits::eHostCoherent | | 119 | VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT | |
| 116 | vk::MemoryPropertyFlagBits::eDeviceLocal); | 120 | VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT); |
| 117 | if (!type) { | 121 | if (!type) { |
| 118 | // Otherwise search for a host visible allocation. | 122 | // Otherwise search for a host visible allocation. |
| 119 | type = FindMemoryType(device, requirements.memoryTypeBits, | 123 | type = FindMemoryType(device, requirements.memoryTypeBits, |
| 120 | vk::MemoryPropertyFlagBits::eHostVisible | | 124 | VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | |
| 121 | vk::MemoryPropertyFlagBits::eHostCoherent); | 125 | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT); |
| 122 | ASSERT_MSG(type, "No host visible and coherent memory type found"); | 126 | ASSERT_MSG(type, "No host visible and coherent memory type found"); |
| 123 | } | 127 | } |
| 124 | const vk::MemoryAllocateInfo alloc_ci(requirements.size, *type); | 128 | VkMemoryAllocateInfo memory_ai; |
| 125 | memory = dev.allocateMemoryUnique(alloc_ci, nullptr, dld); | 129 | memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; |
| 126 | 130 | memory_ai.pNext = nullptr; | |
| 127 | dev.bindBufferMemory(*buffer, *memory, 0, dld); | 131 | memory_ai.allocationSize = requirements.size; |
| 132 | memory_ai.memoryTypeIndex = *type; | ||
| 133 | |||
| 134 | memory = dev.AllocateMemory(memory_ai); | ||
| 135 | buffer.BindMemory(*memory, 0); | ||
| 128 | } | 136 | } |
| 129 | 137 | ||
| 130 | void VKStreamBuffer::ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size) { | 138 | void VKStreamBuffer::ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size) { |
diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.h b/src/video_core/renderer_vulkan/vk_stream_buffer.h index 187c0c612..58ce8b973 100644 --- a/src/video_core/renderer_vulkan/vk_stream_buffer.h +++ b/src/video_core/renderer_vulkan/vk_stream_buffer.h | |||
| @@ -9,7 +9,7 @@ | |||
| 9 | #include <vector> | 9 | #include <vector> |
| 10 | 10 | ||
| 11 | #include "common/common_types.h" | 11 | #include "common/common_types.h" |
| 12 | #include "video_core/renderer_vulkan/declarations.h" | 12 | #include "video_core/renderer_vulkan/wrapper.h" |
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -21,7 +21,7 @@ class VKScheduler; | |||
| 21 | class VKStreamBuffer final { | 21 | class VKStreamBuffer final { |
| 22 | public: | 22 | public: |
| 23 | explicit VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, | 23 | explicit VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, |
| 24 | vk::BufferUsageFlags usage); | 24 | VkBufferUsageFlags usage); |
| 25 | ~VKStreamBuffer(); | 25 | ~VKStreamBuffer(); |
| 26 | 26 | ||
| 27 | /** | 27 | /** |
| @@ -35,7 +35,7 @@ public: | |||
| 35 | /// Ensures that "size" bytes of memory are available to the GPU, potentially recording a copy. | 35 | /// Ensures that "size" bytes of memory are available to the GPU, potentially recording a copy. |
| 36 | void Unmap(u64 size); | 36 | void Unmap(u64 size); |
| 37 | 37 | ||
| 38 | vk::Buffer GetHandle() const { | 38 | VkBuffer GetHandle() const { |
| 39 | return *buffer; | 39 | return *buffer; |
| 40 | } | 40 | } |
| 41 | 41 | ||
| @@ -46,20 +46,18 @@ private: | |||
| 46 | }; | 46 | }; |
| 47 | 47 | ||
| 48 | /// Creates Vulkan buffer handles committing the required the required memory. | 48 | /// Creates Vulkan buffer handles committing the required the required memory. |
| 49 | void CreateBuffers(vk::BufferUsageFlags usage); | 49 | void CreateBuffers(VkBufferUsageFlags usage); |
| 50 | 50 | ||
| 51 | /// Increases the amount of watches available. | 51 | /// Increases the amount of watches available. |
| 52 | void ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size); | 52 | void ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size); |
| 53 | 53 | ||
| 54 | void WaitPendingOperations(u64 requested_upper_bound); | 54 | void WaitPendingOperations(u64 requested_upper_bound); |
| 55 | 55 | ||
| 56 | const VKDevice& device; ///< Vulkan device manager. | 56 | const VKDevice& device; ///< Vulkan device manager. |
| 57 | VKScheduler& scheduler; ///< Command scheduler. | 57 | VKScheduler& scheduler; ///< Command scheduler. |
| 58 | const vk::AccessFlags access; ///< Access usage of this stream buffer. | ||
| 59 | const vk::PipelineStageFlags pipeline_stage; ///< Pipeline usage of this stream buffer. | ||
| 60 | 58 | ||
| 61 | UniqueBuffer buffer; ///< Mapped buffer. | 59 | vk::Buffer buffer; ///< Mapped buffer. |
| 62 | UniqueDeviceMemory memory; ///< Memory allocation. | 60 | vk::DeviceMemory memory; ///< Memory allocation. |
| 63 | 61 | ||
| 64 | u64 offset{}; ///< Buffer iterator. | 62 | u64 offset{}; ///< Buffer iterator. |
| 65 | u64 mapped_size{}; ///< Size reserved for the current copy. | 63 | u64 mapped_size{}; ///< Size reserved for the current copy. |
diff --git a/src/video_core/renderer_vulkan/vk_swapchain.cpp b/src/video_core/renderer_vulkan/vk_swapchain.cpp index 9e73fa9cd..bffd8f32a 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.cpp +++ b/src/video_core/renderer_vulkan/vk_swapchain.cpp | |||
| @@ -11,69 +11,64 @@ | |||
| 11 | #include "common/logging/log.h" | 11 | #include "common/logging/log.h" |
| 12 | #include "core/core.h" | 12 | #include "core/core.h" |
| 13 | #include "core/frontend/framebuffer_layout.h" | 13 | #include "core/frontend/framebuffer_layout.h" |
| 14 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 15 | #include "video_core/renderer_vulkan/vk_device.h" | 14 | #include "video_core/renderer_vulkan/vk_device.h" |
| 16 | #include "video_core/renderer_vulkan/vk_resource_manager.h" | 15 | #include "video_core/renderer_vulkan/vk_resource_manager.h" |
| 17 | #include "video_core/renderer_vulkan/vk_swapchain.h" | 16 | #include "video_core/renderer_vulkan/vk_swapchain.h" |
| 17 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 18 | 18 | ||
| 19 | namespace Vulkan { | 19 | namespace Vulkan { |
| 20 | 20 | ||
| 21 | namespace { | 21 | namespace { |
| 22 | 22 | ||
| 23 | vk::SurfaceFormatKHR ChooseSwapSurfaceFormat(const std::vector<vk::SurfaceFormatKHR>& formats, | 23 | VkSurfaceFormatKHR ChooseSwapSurfaceFormat(vk::Span<VkSurfaceFormatKHR> formats, bool srgb) { |
| 24 | bool srgb) { | 24 | if (formats.size() == 1 && formats[0].format == VK_FORMAT_UNDEFINED) { |
| 25 | if (formats.size() == 1 && formats[0].format == vk::Format::eUndefined) { | 25 | VkSurfaceFormatKHR format; |
| 26 | vk::SurfaceFormatKHR format; | 26 | format.format = VK_FORMAT_B8G8R8A8_UNORM; |
| 27 | format.format = vk::Format::eB8G8R8A8Unorm; | 27 | format.colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; |
| 28 | format.colorSpace = vk::ColorSpaceKHR::eSrgbNonlinear; | ||
| 29 | return format; | 28 | return format; |
| 30 | } | 29 | } |
| 31 | const auto& found = std::find_if(formats.begin(), formats.end(), [srgb](const auto& format) { | 30 | const auto& found = std::find_if(formats.begin(), formats.end(), [srgb](const auto& format) { |
| 32 | const auto request_format = srgb ? vk::Format::eB8G8R8A8Srgb : vk::Format::eB8G8R8A8Unorm; | 31 | const auto request_format = srgb ? VK_FORMAT_B8G8R8A8_SRGB : VK_FORMAT_B8G8R8A8_UNORM; |
| 33 | return format.format == request_format && | 32 | return format.format == request_format && |
| 34 | format.colorSpace == vk::ColorSpaceKHR::eSrgbNonlinear; | 33 | format.colorSpace == VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; |
| 35 | }); | 34 | }); |
| 36 | return found != formats.end() ? *found : formats[0]; | 35 | return found != formats.end() ? *found : formats[0]; |
| 37 | } | 36 | } |
| 38 | 37 | ||
| 39 | vk::PresentModeKHR ChooseSwapPresentMode(const std::vector<vk::PresentModeKHR>& modes) { | 38 | VkPresentModeKHR ChooseSwapPresentMode(vk::Span<VkPresentModeKHR> modes) { |
| 40 | // Mailbox doesn't lock the application like fifo (vsync), prefer it | 39 | // Mailbox doesn't lock the application like fifo (vsync), prefer it |
| 41 | const auto& found = std::find_if(modes.begin(), modes.end(), [](const auto& mode) { | 40 | const auto found = std::find(modes.begin(), modes.end(), VK_PRESENT_MODE_MAILBOX_KHR); |
| 42 | return mode == vk::PresentModeKHR::eMailbox; | 41 | return found != modes.end() ? *found : VK_PRESENT_MODE_FIFO_KHR; |
| 43 | }); | ||
| 44 | return found != modes.end() ? *found : vk::PresentModeKHR::eFifo; | ||
| 45 | } | 42 | } |
| 46 | 43 | ||
| 47 | vk::Extent2D ChooseSwapExtent(const vk::SurfaceCapabilitiesKHR& capabilities, u32 width, | 44 | VkExtent2D ChooseSwapExtent(const VkSurfaceCapabilitiesKHR& capabilities, u32 width, u32 height) { |
| 48 | u32 height) { | ||
| 49 | constexpr auto undefined_size{std::numeric_limits<u32>::max()}; | 45 | constexpr auto undefined_size{std::numeric_limits<u32>::max()}; |
| 50 | if (capabilities.currentExtent.width != undefined_size) { | 46 | if (capabilities.currentExtent.width != undefined_size) { |
| 51 | return capabilities.currentExtent; | 47 | return capabilities.currentExtent; |
| 52 | } | 48 | } |
| 53 | vk::Extent2D extent = {width, height}; | 49 | VkExtent2D extent; |
| 54 | extent.width = std::max(capabilities.minImageExtent.width, | 50 | extent.width = std::max(capabilities.minImageExtent.width, |
| 55 | std::min(capabilities.maxImageExtent.width, extent.width)); | 51 | std::min(capabilities.maxImageExtent.width, width)); |
| 56 | extent.height = std::max(capabilities.minImageExtent.height, | 52 | extent.height = std::max(capabilities.minImageExtent.height, |
| 57 | std::min(capabilities.maxImageExtent.height, extent.height)); | 53 | std::min(capabilities.maxImageExtent.height, height)); |
| 58 | return extent; | 54 | return extent; |
| 59 | } | 55 | } |
| 60 | 56 | ||
| 61 | } // Anonymous namespace | 57 | } // Anonymous namespace |
| 62 | 58 | ||
| 63 | VKSwapchain::VKSwapchain(vk::SurfaceKHR surface, const VKDevice& device) | 59 | VKSwapchain::VKSwapchain(VkSurfaceKHR surface, const VKDevice& device) |
| 64 | : surface{surface}, device{device} {} | 60 | : surface{surface}, device{device} {} |
| 65 | 61 | ||
| 66 | VKSwapchain::~VKSwapchain() = default; | 62 | VKSwapchain::~VKSwapchain() = default; |
| 67 | 63 | ||
| 68 | void VKSwapchain::Create(u32 width, u32 height, bool srgb) { | 64 | void VKSwapchain::Create(u32 width, u32 height, bool srgb) { |
| 69 | const auto& dld = device.GetDispatchLoader(); | ||
| 70 | const auto physical_device = device.GetPhysical(); | 65 | const auto physical_device = device.GetPhysical(); |
| 71 | const auto capabilities{physical_device.getSurfaceCapabilitiesKHR(surface, dld)}; | 66 | const auto capabilities{physical_device.GetSurfaceCapabilitiesKHR(surface)}; |
| 72 | if (capabilities.maxImageExtent.width == 0 || capabilities.maxImageExtent.height == 0) { | 67 | if (capabilities.maxImageExtent.width == 0 || capabilities.maxImageExtent.height == 0) { |
| 73 | return; | 68 | return; |
| 74 | } | 69 | } |
| 75 | 70 | ||
| 76 | device.GetLogical().waitIdle(dld); | 71 | device.GetLogical().WaitIdle(); |
| 77 | Destroy(); | 72 | Destroy(); |
| 78 | 73 | ||
| 79 | CreateSwapchain(capabilities, width, height, srgb); | 74 | CreateSwapchain(capabilities, width, height, srgb); |
| @@ -84,10 +79,8 @@ void VKSwapchain::Create(u32 width, u32 height, bool srgb) { | |||
| 84 | } | 79 | } |
| 85 | 80 | ||
| 86 | void VKSwapchain::AcquireNextImage() { | 81 | void VKSwapchain::AcquireNextImage() { |
| 87 | const auto dev{device.GetLogical()}; | 82 | device.GetLogical().AcquireNextImageKHR(*swapchain, std::numeric_limits<u64>::max(), |
| 88 | const auto& dld{device.GetDispatchLoader()}; | 83 | *present_semaphores[frame_index], {}, &image_index); |
| 89 | dev.acquireNextImageKHR(*swapchain, std::numeric_limits<u64>::max(), | ||
| 90 | *present_semaphores[frame_index], {}, &image_index, dld); | ||
| 91 | 84 | ||
| 92 | if (auto& fence = fences[image_index]; fence) { | 85 | if (auto& fence = fences[image_index]; fence) { |
| 93 | fence->Wait(); | 86 | fence->Wait(); |
| @@ -96,29 +89,37 @@ void VKSwapchain::AcquireNextImage() { | |||
| 96 | } | 89 | } |
| 97 | } | 90 | } |
| 98 | 91 | ||
| 99 | bool VKSwapchain::Present(vk::Semaphore render_semaphore, VKFence& fence) { | 92 | bool VKSwapchain::Present(VkSemaphore render_semaphore, VKFence& fence) { |
| 100 | const vk::Semaphore present_semaphore{*present_semaphores[frame_index]}; | 93 | const VkSemaphore present_semaphore{*present_semaphores[frame_index]}; |
| 101 | const std::array<vk::Semaphore, 2> semaphores{present_semaphore, render_semaphore}; | 94 | const std::array<VkSemaphore, 2> semaphores{present_semaphore, render_semaphore}; |
| 102 | const u32 wait_semaphore_count{render_semaphore ? 2U : 1U}; | ||
| 103 | const auto& dld{device.GetDispatchLoader()}; | ||
| 104 | const auto present_queue{device.GetPresentQueue()}; | 95 | const auto present_queue{device.GetPresentQueue()}; |
| 105 | bool recreated = false; | 96 | bool recreated = false; |
| 106 | 97 | ||
| 107 | const vk::PresentInfoKHR present_info(wait_semaphore_count, semaphores.data(), 1, | 98 | VkPresentInfoKHR present_info; |
| 108 | &swapchain.get(), &image_index, {}); | 99 | present_info.sType = VK_STRUCTURE_TYPE_PRESENT_INFO_KHR; |
| 109 | switch (const auto result = present_queue.presentKHR(&present_info, dld); result) { | 100 | present_info.pNext = nullptr; |
| 110 | case vk::Result::eSuccess: | 101 | present_info.waitSemaphoreCount = render_semaphore ? 2U : 1U; |
| 102 | present_info.pWaitSemaphores = semaphores.data(); | ||
| 103 | present_info.swapchainCount = 1; | ||
| 104 | present_info.pSwapchains = swapchain.address(); | ||
| 105 | present_info.pImageIndices = &image_index; | ||
| 106 | present_info.pResults = nullptr; | ||
| 107 | |||
| 108 | switch (const VkResult result = present_queue.Present(present_info)) { | ||
| 109 | case VK_SUCCESS: | ||
| 110 | break; | ||
| 111 | case VK_SUBOPTIMAL_KHR: | ||
| 112 | LOG_DEBUG(Render_Vulkan, "Suboptimal swapchain"); | ||
| 111 | break; | 113 | break; |
| 112 | case vk::Result::eErrorOutOfDateKHR: | 114 | case VK_ERROR_OUT_OF_DATE_KHR: |
| 113 | if (current_width > 0 && current_height > 0) { | 115 | if (current_width > 0 && current_height > 0) { |
| 114 | Create(current_width, current_height, current_srgb); | 116 | Create(current_width, current_height, current_srgb); |
| 115 | recreated = true; | 117 | recreated = true; |
| 116 | } | 118 | } |
| 117 | break; | 119 | break; |
| 118 | default: | 120 | default: |
| 119 | LOG_CRITICAL(Render_Vulkan, "Vulkan failed to present swapchain due to {}!", | 121 | LOG_CRITICAL(Render_Vulkan, "Failed to present with error {}", vk::ToString(result)); |
| 120 | vk::to_string(result)); | 122 | break; |
| 121 | UNREACHABLE(); | ||
| 122 | } | 123 | } |
| 123 | 124 | ||
| 124 | ASSERT(fences[image_index] == nullptr); | 125 | ASSERT(fences[image_index] == nullptr); |
| @@ -132,74 +133,92 @@ bool VKSwapchain::HasFramebufferChanged(const Layout::FramebufferLayout& framebu | |||
| 132 | return framebuffer.width != current_width || framebuffer.height != current_height; | 133 | return framebuffer.width != current_width || framebuffer.height != current_height; |
| 133 | } | 134 | } |
| 134 | 135 | ||
| 135 | void VKSwapchain::CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities, u32 width, | 136 | void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities, u32 width, |
| 136 | u32 height, bool srgb) { | 137 | u32 height, bool srgb) { |
| 137 | const auto& dld{device.GetDispatchLoader()}; | ||
| 138 | const auto physical_device{device.GetPhysical()}; | 138 | const auto physical_device{device.GetPhysical()}; |
| 139 | const auto formats{physical_device.getSurfaceFormatsKHR(surface, dld)}; | 139 | const auto formats{physical_device.GetSurfaceFormatsKHR(surface)}; |
| 140 | const auto present_modes{physical_device.getSurfacePresentModesKHR(surface, dld)}; | 140 | const auto present_modes{physical_device.GetSurfacePresentModesKHR(surface)}; |
| 141 | 141 | ||
| 142 | const vk::SurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats, srgb)}; | 142 | const VkSurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats, srgb)}; |
| 143 | const vk::PresentModeKHR present_mode{ChooseSwapPresentMode(present_modes)}; | 143 | const VkPresentModeKHR present_mode{ChooseSwapPresentMode(present_modes)}; |
| 144 | 144 | ||
| 145 | u32 requested_image_count{capabilities.minImageCount + 1}; | 145 | u32 requested_image_count{capabilities.minImageCount + 1}; |
| 146 | if (capabilities.maxImageCount > 0 && requested_image_count > capabilities.maxImageCount) { | 146 | if (capabilities.maxImageCount > 0 && requested_image_count > capabilities.maxImageCount) { |
| 147 | requested_image_count = capabilities.maxImageCount; | 147 | requested_image_count = capabilities.maxImageCount; |
| 148 | } | 148 | } |
| 149 | 149 | ||
| 150 | vk::SwapchainCreateInfoKHR swapchain_ci( | 150 | VkSwapchainCreateInfoKHR swapchain_ci; |
| 151 | {}, surface, requested_image_count, surface_format.format, surface_format.colorSpace, {}, 1, | 151 | swapchain_ci.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR; |
| 152 | vk::ImageUsageFlagBits::eColorAttachment, {}, {}, {}, capabilities.currentTransform, | 152 | swapchain_ci.pNext = nullptr; |
| 153 | vk::CompositeAlphaFlagBitsKHR::eOpaque, present_mode, false, {}); | 153 | swapchain_ci.flags = 0; |
| 154 | swapchain_ci.surface = surface; | ||
| 155 | swapchain_ci.minImageCount = requested_image_count; | ||
| 156 | swapchain_ci.imageFormat = surface_format.format; | ||
| 157 | swapchain_ci.imageColorSpace = surface_format.colorSpace; | ||
| 158 | swapchain_ci.imageArrayLayers = 1; | ||
| 159 | swapchain_ci.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; | ||
| 160 | swapchain_ci.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; | ||
| 161 | swapchain_ci.queueFamilyIndexCount = 0; | ||
| 162 | swapchain_ci.pQueueFamilyIndices = nullptr; | ||
| 163 | swapchain_ci.preTransform = capabilities.currentTransform; | ||
| 164 | swapchain_ci.compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR; | ||
| 165 | swapchain_ci.presentMode = present_mode; | ||
| 166 | swapchain_ci.clipped = VK_FALSE; | ||
| 167 | swapchain_ci.oldSwapchain = nullptr; | ||
| 154 | 168 | ||
| 155 | const u32 graphics_family{device.GetGraphicsFamily()}; | 169 | const u32 graphics_family{device.GetGraphicsFamily()}; |
| 156 | const u32 present_family{device.GetPresentFamily()}; | 170 | const u32 present_family{device.GetPresentFamily()}; |
| 157 | const std::array<u32, 2> queue_indices{graphics_family, present_family}; | 171 | const std::array<u32, 2> queue_indices{graphics_family, present_family}; |
| 158 | if (graphics_family != present_family) { | 172 | if (graphics_family != present_family) { |
| 159 | swapchain_ci.imageSharingMode = vk::SharingMode::eConcurrent; | 173 | swapchain_ci.imageSharingMode = VK_SHARING_MODE_CONCURRENT; |
| 160 | swapchain_ci.queueFamilyIndexCount = static_cast<u32>(queue_indices.size()); | 174 | swapchain_ci.queueFamilyIndexCount = static_cast<u32>(queue_indices.size()); |
| 161 | swapchain_ci.pQueueFamilyIndices = queue_indices.data(); | 175 | swapchain_ci.pQueueFamilyIndices = queue_indices.data(); |
| 162 | } else { | 176 | } else { |
| 163 | swapchain_ci.imageSharingMode = vk::SharingMode::eExclusive; | 177 | swapchain_ci.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; |
| 164 | } | 178 | } |
| 165 | 179 | ||
| 166 | // Request the size again to reduce the possibility of a TOCTOU race condition. | 180 | // Request the size again to reduce the possibility of a TOCTOU race condition. |
| 167 | const auto updated_capabilities = physical_device.getSurfaceCapabilitiesKHR(surface, dld); | 181 | const auto updated_capabilities = physical_device.GetSurfaceCapabilitiesKHR(surface); |
| 168 | swapchain_ci.imageExtent = ChooseSwapExtent(updated_capabilities, width, height); | 182 | swapchain_ci.imageExtent = ChooseSwapExtent(updated_capabilities, width, height); |
| 169 | // Don't add code within this and the swapchain creation. | 183 | // Don't add code within this and the swapchain creation. |
| 170 | const auto dev{device.GetLogical()}; | 184 | swapchain = device.GetLogical().CreateSwapchainKHR(swapchain_ci); |
| 171 | swapchain = dev.createSwapchainKHRUnique(swapchain_ci, nullptr, dld); | ||
| 172 | 185 | ||
| 173 | extent = swapchain_ci.imageExtent; | 186 | extent = swapchain_ci.imageExtent; |
| 174 | current_width = extent.width; | 187 | current_width = extent.width; |
| 175 | current_height = extent.height; | 188 | current_height = extent.height; |
| 176 | current_srgb = srgb; | 189 | current_srgb = srgb; |
| 177 | 190 | ||
| 178 | images = dev.getSwapchainImagesKHR(*swapchain, dld); | 191 | images = swapchain.GetImages(); |
| 179 | image_count = static_cast<u32>(images.size()); | 192 | image_count = static_cast<u32>(images.size()); |
| 180 | image_format = surface_format.format; | 193 | image_format = surface_format.format; |
| 181 | } | 194 | } |
| 182 | 195 | ||
| 183 | void VKSwapchain::CreateSemaphores() { | 196 | void VKSwapchain::CreateSemaphores() { |
| 184 | const auto dev{device.GetLogical()}; | ||
| 185 | const auto& dld{device.GetDispatchLoader()}; | ||
| 186 | |||
| 187 | present_semaphores.resize(image_count); | 197 | present_semaphores.resize(image_count); |
| 188 | for (std::size_t i = 0; i < image_count; i++) { | 198 | std::generate(present_semaphores.begin(), present_semaphores.end(), |
| 189 | present_semaphores[i] = dev.createSemaphoreUnique({}, nullptr, dld); | 199 | [this] { return device.GetLogical().CreateSemaphore(); }); |
| 190 | } | ||
| 191 | } | 200 | } |
| 192 | 201 | ||
| 193 | void VKSwapchain::CreateImageViews() { | 202 | void VKSwapchain::CreateImageViews() { |
| 194 | const auto dev{device.GetLogical()}; | 203 | VkImageViewCreateInfo ci; |
| 195 | const auto& dld{device.GetDispatchLoader()}; | 204 | ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; |
| 205 | ci.pNext = nullptr; | ||
| 206 | ci.flags = 0; | ||
| 207 | // ci.image | ||
| 208 | ci.viewType = VK_IMAGE_VIEW_TYPE_2D; | ||
| 209 | ci.format = image_format; | ||
| 210 | ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY, | ||
| 211 | VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY}; | ||
| 212 | ci.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; | ||
| 213 | ci.subresourceRange.baseMipLevel = 0; | ||
| 214 | ci.subresourceRange.levelCount = 1; | ||
| 215 | ci.subresourceRange.baseArrayLayer = 0; | ||
| 216 | ci.subresourceRange.layerCount = 1; | ||
| 196 | 217 | ||
| 197 | image_views.resize(image_count); | 218 | image_views.resize(image_count); |
| 198 | for (std::size_t i = 0; i < image_count; i++) { | 219 | for (std::size_t i = 0; i < image_count; i++) { |
| 199 | const vk::ImageViewCreateInfo image_view_ci({}, images[i], vk::ImageViewType::e2D, | 220 | ci.image = images[i]; |
| 200 | image_format, {}, | 221 | image_views[i] = device.GetLogical().CreateImageView(ci); |
| 201 | {vk::ImageAspectFlagBits::eColor, 0, 1, 0, 1}); | ||
| 202 | image_views[i] = dev.createImageViewUnique(image_view_ci, nullptr, dld); | ||
| 203 | } | 222 | } |
| 204 | } | 223 | } |
| 205 | 224 | ||
diff --git a/src/video_core/renderer_vulkan/vk_swapchain.h b/src/video_core/renderer_vulkan/vk_swapchain.h index 2f3b2ccd5..a35d61345 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.h +++ b/src/video_core/renderer_vulkan/vk_swapchain.h | |||
| @@ -7,7 +7,7 @@ | |||
| 7 | #include <vector> | 7 | #include <vector> |
| 8 | 8 | ||
| 9 | #include "common/common_types.h" | 9 | #include "common/common_types.h" |
| 10 | #include "video_core/renderer_vulkan/declarations.h" | 10 | #include "video_core/renderer_vulkan/wrapper.h" |
| 11 | 11 | ||
| 12 | namespace Layout { | 12 | namespace Layout { |
| 13 | struct FramebufferLayout; | 13 | struct FramebufferLayout; |
| @@ -20,7 +20,7 @@ class VKFence; | |||
| 20 | 20 | ||
| 21 | class VKSwapchain { | 21 | class VKSwapchain { |
| 22 | public: | 22 | public: |
| 23 | explicit VKSwapchain(vk::SurfaceKHR surface, const VKDevice& device); | 23 | explicit VKSwapchain(VkSurfaceKHR surface, const VKDevice& device); |
| 24 | ~VKSwapchain(); | 24 | ~VKSwapchain(); |
| 25 | 25 | ||
| 26 | /// Creates (or recreates) the swapchain with a given size. | 26 | /// Creates (or recreates) the swapchain with a given size. |
| @@ -31,12 +31,12 @@ public: | |||
| 31 | 31 | ||
| 32 | /// Presents the rendered image to the swapchain. Returns true when the swapchains had to be | 32 | /// Presents the rendered image to the swapchain. Returns true when the swapchains had to be |
| 33 | /// recreated. Takes responsability for the ownership of fence. | 33 | /// recreated. Takes responsability for the ownership of fence. |
| 34 | bool Present(vk::Semaphore render_semaphore, VKFence& fence); | 34 | bool Present(VkSemaphore render_semaphore, VKFence& fence); |
| 35 | 35 | ||
| 36 | /// Returns true when the framebuffer layout has changed. | 36 | /// Returns true when the framebuffer layout has changed. |
| 37 | bool HasFramebufferChanged(const Layout::FramebufferLayout& framebuffer) const; | 37 | bool HasFramebufferChanged(const Layout::FramebufferLayout& framebuffer) const; |
| 38 | 38 | ||
| 39 | const vk::Extent2D& GetSize() const { | 39 | VkExtent2D GetSize() const { |
| 40 | return extent; | 40 | return extent; |
| 41 | } | 41 | } |
| 42 | 42 | ||
| @@ -48,15 +48,15 @@ public: | |||
| 48 | return image_index; | 48 | return image_index; |
| 49 | } | 49 | } |
| 50 | 50 | ||
| 51 | vk::Image GetImageIndex(std::size_t index) const { | 51 | VkImage GetImageIndex(std::size_t index) const { |
| 52 | return images[index]; | 52 | return images[index]; |
| 53 | } | 53 | } |
| 54 | 54 | ||
| 55 | vk::ImageView GetImageViewIndex(std::size_t index) const { | 55 | VkImageView GetImageViewIndex(std::size_t index) const { |
| 56 | return *image_views[index]; | 56 | return *image_views[index]; |
| 57 | } | 57 | } |
| 58 | 58 | ||
| 59 | vk::Format GetImageFormat() const { | 59 | VkFormat GetImageFormat() const { |
| 60 | return image_format; | 60 | return image_format; |
| 61 | } | 61 | } |
| 62 | 62 | ||
| @@ -65,30 +65,30 @@ public: | |||
| 65 | } | 65 | } |
| 66 | 66 | ||
| 67 | private: | 67 | private: |
| 68 | void CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities, u32 width, u32 height, | 68 | void CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities, u32 width, u32 height, |
| 69 | bool srgb); | 69 | bool srgb); |
| 70 | void CreateSemaphores(); | 70 | void CreateSemaphores(); |
| 71 | void CreateImageViews(); | 71 | void CreateImageViews(); |
| 72 | 72 | ||
| 73 | void Destroy(); | 73 | void Destroy(); |
| 74 | 74 | ||
| 75 | const vk::SurfaceKHR surface; | 75 | const VkSurfaceKHR surface; |
| 76 | const VKDevice& device; | 76 | const VKDevice& device; |
| 77 | 77 | ||
| 78 | UniqueSwapchainKHR swapchain; | 78 | vk::SwapchainKHR swapchain; |
| 79 | 79 | ||
| 80 | std::size_t image_count{}; | 80 | std::size_t image_count{}; |
| 81 | std::vector<vk::Image> images; | 81 | std::vector<VkImage> images; |
| 82 | std::vector<UniqueImageView> image_views; | 82 | std::vector<vk::ImageView> image_views; |
| 83 | std::vector<UniqueFramebuffer> framebuffers; | 83 | std::vector<vk::Framebuffer> framebuffers; |
| 84 | std::vector<VKFence*> fences; | 84 | std::vector<VKFence*> fences; |
| 85 | std::vector<UniqueSemaphore> present_semaphores; | 85 | std::vector<vk::Semaphore> present_semaphores; |
| 86 | 86 | ||
| 87 | u32 image_index{}; | 87 | u32 image_index{}; |
| 88 | u32 frame_index{}; | 88 | u32 frame_index{}; |
| 89 | 89 | ||
| 90 | vk::Format image_format{}; | 90 | VkFormat image_format{}; |
| 91 | vk::Extent2D extent{}; | 91 | VkExtent2D extent{}; |
| 92 | 92 | ||
| 93 | u32 current_width{}; | 93 | u32 current_width{}; |
| 94 | u32 current_height{}; | 94 | u32 current_height{}; |
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 5b9b39670..de4c23120 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp | |||
| @@ -17,7 +17,6 @@ | |||
| 17 | #include "core/memory.h" | 17 | #include "core/memory.h" |
| 18 | #include "video_core/engines/maxwell_3d.h" | 18 | #include "video_core/engines/maxwell_3d.h" |
| 19 | #include "video_core/morton.h" | 19 | #include "video_core/morton.h" |
| 20 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 21 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" | 20 | #include "video_core/renderer_vulkan/maxwell_to_vk.h" |
| 22 | #include "video_core/renderer_vulkan/vk_device.h" | 21 | #include "video_core/renderer_vulkan/vk_device.h" |
| 23 | #include "video_core/renderer_vulkan/vk_memory_manager.h" | 22 | #include "video_core/renderer_vulkan/vk_memory_manager.h" |
| @@ -25,6 +24,7 @@ | |||
| 25 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 24 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 26 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" | 25 | #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" |
| 27 | #include "video_core/renderer_vulkan/vk_texture_cache.h" | 26 | #include "video_core/renderer_vulkan/vk_texture_cache.h" |
| 27 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 28 | #include "video_core/surface.h" | 28 | #include "video_core/surface.h" |
| 29 | #include "video_core/textures/convert.h" | 29 | #include "video_core/textures/convert.h" |
| 30 | 30 | ||
| @@ -39,18 +39,18 @@ using VideoCore::Surface::SurfaceTarget; | |||
| 39 | 39 | ||
| 40 | namespace { | 40 | namespace { |
| 41 | 41 | ||
| 42 | vk::ImageType SurfaceTargetToImage(SurfaceTarget target) { | 42 | VkImageType SurfaceTargetToImage(SurfaceTarget target) { |
| 43 | switch (target) { | 43 | switch (target) { |
| 44 | case SurfaceTarget::Texture1D: | 44 | case SurfaceTarget::Texture1D: |
| 45 | case SurfaceTarget::Texture1DArray: | 45 | case SurfaceTarget::Texture1DArray: |
| 46 | return vk::ImageType::e1D; | 46 | return VK_IMAGE_TYPE_1D; |
| 47 | case SurfaceTarget::Texture2D: | 47 | case SurfaceTarget::Texture2D: |
| 48 | case SurfaceTarget::Texture2DArray: | 48 | case SurfaceTarget::Texture2DArray: |
| 49 | case SurfaceTarget::TextureCubemap: | 49 | case SurfaceTarget::TextureCubemap: |
| 50 | case SurfaceTarget::TextureCubeArray: | 50 | case SurfaceTarget::TextureCubeArray: |
| 51 | return vk::ImageType::e2D; | 51 | return VK_IMAGE_TYPE_2D; |
| 52 | case SurfaceTarget::Texture3D: | 52 | case SurfaceTarget::Texture3D: |
| 53 | return vk::ImageType::e3D; | 53 | return VK_IMAGE_TYPE_3D; |
| 54 | case SurfaceTarget::TextureBuffer: | 54 | case SurfaceTarget::TextureBuffer: |
| 55 | UNREACHABLE(); | 55 | UNREACHABLE(); |
| 56 | return {}; | 56 | return {}; |
| @@ -59,35 +59,35 @@ vk::ImageType SurfaceTargetToImage(SurfaceTarget target) { | |||
| 59 | return {}; | 59 | return {}; |
| 60 | } | 60 | } |
| 61 | 61 | ||
| 62 | vk::ImageAspectFlags PixelFormatToImageAspect(PixelFormat pixel_format) { | 62 | VkImageAspectFlags PixelFormatToImageAspect(PixelFormat pixel_format) { |
| 63 | if (pixel_format < PixelFormat::MaxColorFormat) { | 63 | if (pixel_format < PixelFormat::MaxColorFormat) { |
| 64 | return vk::ImageAspectFlagBits::eColor; | 64 | return VK_IMAGE_ASPECT_COLOR_BIT; |
| 65 | } else if (pixel_format < PixelFormat::MaxDepthFormat) { | 65 | } else if (pixel_format < PixelFormat::MaxDepthFormat) { |
| 66 | return vk::ImageAspectFlagBits::eDepth; | 66 | return VK_IMAGE_ASPECT_DEPTH_BIT; |
| 67 | } else if (pixel_format < PixelFormat::MaxDepthStencilFormat) { | 67 | } else if (pixel_format < PixelFormat::MaxDepthStencilFormat) { |
| 68 | return vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil; | 68 | return VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; |
| 69 | } else { | 69 | } else { |
| 70 | UNREACHABLE_MSG("Invalid pixel format={}", static_cast<u32>(pixel_format)); | 70 | UNREACHABLE_MSG("Invalid pixel format={}", static_cast<int>(pixel_format)); |
| 71 | return vk::ImageAspectFlagBits::eColor; | 71 | return VK_IMAGE_ASPECT_COLOR_BIT; |
| 72 | } | 72 | } |
| 73 | } | 73 | } |
| 74 | 74 | ||
| 75 | vk::ImageViewType GetImageViewType(SurfaceTarget target) { | 75 | VkImageViewType GetImageViewType(SurfaceTarget target) { |
| 76 | switch (target) { | 76 | switch (target) { |
| 77 | case SurfaceTarget::Texture1D: | 77 | case SurfaceTarget::Texture1D: |
| 78 | return vk::ImageViewType::e1D; | 78 | return VK_IMAGE_VIEW_TYPE_1D; |
| 79 | case SurfaceTarget::Texture2D: | 79 | case SurfaceTarget::Texture2D: |
| 80 | return vk::ImageViewType::e2D; | 80 | return VK_IMAGE_VIEW_TYPE_2D; |
| 81 | case SurfaceTarget::Texture3D: | 81 | case SurfaceTarget::Texture3D: |
| 82 | return vk::ImageViewType::e3D; | 82 | return VK_IMAGE_VIEW_TYPE_3D; |
| 83 | case SurfaceTarget::Texture1DArray: | 83 | case SurfaceTarget::Texture1DArray: |
| 84 | return vk::ImageViewType::e1DArray; | 84 | return VK_IMAGE_VIEW_TYPE_1D_ARRAY; |
| 85 | case SurfaceTarget::Texture2DArray: | 85 | case SurfaceTarget::Texture2DArray: |
| 86 | return vk::ImageViewType::e2DArray; | 86 | return VK_IMAGE_VIEW_TYPE_2D_ARRAY; |
| 87 | case SurfaceTarget::TextureCubemap: | 87 | case SurfaceTarget::TextureCubemap: |
| 88 | return vk::ImageViewType::eCube; | 88 | return VK_IMAGE_VIEW_TYPE_CUBE; |
| 89 | case SurfaceTarget::TextureCubeArray: | 89 | case SurfaceTarget::TextureCubeArray: |
| 90 | return vk::ImageViewType::eCubeArray; | 90 | return VK_IMAGE_VIEW_TYPE_CUBE_ARRAY; |
| 91 | case SurfaceTarget::TextureBuffer: | 91 | case SurfaceTarget::TextureBuffer: |
| 92 | break; | 92 | break; |
| 93 | } | 93 | } |
| @@ -95,73 +95,88 @@ vk::ImageViewType GetImageViewType(SurfaceTarget target) { | |||
| 95 | return {}; | 95 | return {}; |
| 96 | } | 96 | } |
| 97 | 97 | ||
| 98 | UniqueBuffer CreateBuffer(const VKDevice& device, const SurfaceParams& params, | 98 | vk::Buffer CreateBuffer(const VKDevice& device, const SurfaceParams& params, |
| 99 | std::size_t host_memory_size) { | 99 | std::size_t host_memory_size) { |
| 100 | // TODO(Rodrigo): Move texture buffer creation to the buffer cache | 100 | // TODO(Rodrigo): Move texture buffer creation to the buffer cache |
| 101 | const vk::BufferCreateInfo buffer_ci({}, host_memory_size, | 101 | VkBufferCreateInfo ci; |
| 102 | vk::BufferUsageFlagBits::eUniformTexelBuffer | | 102 | ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; |
| 103 | vk::BufferUsageFlagBits::eTransferSrc | | 103 | ci.pNext = nullptr; |
| 104 | vk::BufferUsageFlagBits::eTransferDst, | 104 | ci.flags = 0; |
| 105 | vk::SharingMode::eExclusive, 0, nullptr); | 105 | ci.size = static_cast<VkDeviceSize>(host_memory_size); |
| 106 | const auto dev = device.GetLogical(); | 106 | ci.usage = VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | |
| 107 | const auto& dld = device.GetDispatchLoader(); | 107 | VK_BUFFER_USAGE_TRANSFER_DST_BIT; |
| 108 | return dev.createBufferUnique(buffer_ci, nullptr, dld); | 108 | ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; |
| 109 | ci.queueFamilyIndexCount = 0; | ||
| 110 | ci.pQueueFamilyIndices = nullptr; | ||
| 111 | return device.GetLogical().CreateBuffer(ci); | ||
| 109 | } | 112 | } |
| 110 | 113 | ||
| 111 | vk::BufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device, | 114 | VkBufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device, |
| 112 | const SurfaceParams& params, | 115 | const SurfaceParams& params, VkBuffer buffer, |
| 113 | vk::Buffer buffer, | 116 | std::size_t host_memory_size) { |
| 114 | std::size_t host_memory_size) { | ||
| 115 | ASSERT(params.IsBuffer()); | 117 | ASSERT(params.IsBuffer()); |
| 116 | 118 | ||
| 117 | const auto format = | 119 | VkBufferViewCreateInfo ci; |
| 118 | MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format; | 120 | ci.sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO; |
| 119 | return vk::BufferViewCreateInfo({}, buffer, format, 0, host_memory_size); | 121 | ci.pNext = nullptr; |
| 122 | ci.flags = 0; | ||
| 123 | ci.buffer = buffer; | ||
| 124 | ci.format = MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format; | ||
| 125 | ci.offset = 0; | ||
| 126 | ci.range = static_cast<VkDeviceSize>(host_memory_size); | ||
| 127 | return ci; | ||
| 120 | } | 128 | } |
| 121 | 129 | ||
| 122 | vk::ImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) { | 130 | VkImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) { |
| 123 | constexpr auto sample_count = vk::SampleCountFlagBits::e1; | ||
| 124 | constexpr auto tiling = vk::ImageTiling::eOptimal; | ||
| 125 | |||
| 126 | ASSERT(!params.IsBuffer()); | 131 | ASSERT(!params.IsBuffer()); |
| 127 | 132 | ||
| 128 | const auto [format, attachable, storage] = | 133 | const auto [format, attachable, storage] = |
| 129 | MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.pixel_format); | 134 | MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.pixel_format); |
| 130 | 135 | ||
| 131 | auto image_usage = vk::ImageUsageFlagBits::eSampled | vk::ImageUsageFlagBits::eTransferDst | | 136 | VkImageCreateInfo ci; |
| 132 | vk::ImageUsageFlagBits::eTransferSrc; | 137 | ci.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO; |
| 138 | ci.pNext = nullptr; | ||
| 139 | ci.flags = 0; | ||
| 140 | ci.imageType = SurfaceTargetToImage(params.target); | ||
| 141 | ci.format = format; | ||
| 142 | ci.mipLevels = params.num_levels; | ||
| 143 | ci.arrayLayers = static_cast<u32>(params.GetNumLayers()); | ||
| 144 | ci.samples = VK_SAMPLE_COUNT_1_BIT; | ||
| 145 | ci.tiling = VK_IMAGE_TILING_OPTIMAL; | ||
| 146 | ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; | ||
| 147 | ci.queueFamilyIndexCount = 0; | ||
| 148 | ci.pQueueFamilyIndices = nullptr; | ||
| 149 | ci.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; | ||
| 150 | |||
| 151 | ci.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT | | ||
| 152 | VK_IMAGE_USAGE_TRANSFER_SRC_BIT; | ||
| 133 | if (attachable) { | 153 | if (attachable) { |
| 134 | image_usage |= params.IsPixelFormatZeta() ? vk::ImageUsageFlagBits::eDepthStencilAttachment | 154 | ci.usage |= params.IsPixelFormatZeta() ? VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT |
| 135 | : vk::ImageUsageFlagBits::eColorAttachment; | 155 | : VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; |
| 136 | } | 156 | } |
| 137 | if (storage) { | 157 | if (storage) { |
| 138 | image_usage |= vk::ImageUsageFlagBits::eStorage; | 158 | ci.usage |= VK_IMAGE_USAGE_STORAGE_BIT; |
| 139 | } | 159 | } |
| 140 | 160 | ||
| 141 | vk::ImageCreateFlags flags; | ||
| 142 | vk::Extent3D extent; | ||
| 143 | switch (params.target) { | 161 | switch (params.target) { |
| 144 | case SurfaceTarget::TextureCubemap: | 162 | case SurfaceTarget::TextureCubemap: |
| 145 | case SurfaceTarget::TextureCubeArray: | 163 | case SurfaceTarget::TextureCubeArray: |
| 146 | flags |= vk::ImageCreateFlagBits::eCubeCompatible; | 164 | ci.flags |= VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT; |
| 147 | [[fallthrough]]; | 165 | [[fallthrough]]; |
| 148 | case SurfaceTarget::Texture1D: | 166 | case SurfaceTarget::Texture1D: |
| 149 | case SurfaceTarget::Texture1DArray: | 167 | case SurfaceTarget::Texture1DArray: |
| 150 | case SurfaceTarget::Texture2D: | 168 | case SurfaceTarget::Texture2D: |
| 151 | case SurfaceTarget::Texture2DArray: | 169 | case SurfaceTarget::Texture2DArray: |
| 152 | extent = vk::Extent3D(params.width, params.height, 1); | 170 | ci.extent = {params.width, params.height, 1}; |
| 153 | break; | 171 | break; |
| 154 | case SurfaceTarget::Texture3D: | 172 | case SurfaceTarget::Texture3D: |
| 155 | extent = vk::Extent3D(params.width, params.height, params.depth); | 173 | ci.extent = {params.width, params.height, params.depth}; |
| 156 | break; | 174 | break; |
| 157 | case SurfaceTarget::TextureBuffer: | 175 | case SurfaceTarget::TextureBuffer: |
| 158 | UNREACHABLE(); | 176 | UNREACHABLE(); |
| 159 | } | 177 | } |
| 160 | 178 | ||
| 161 | return vk::ImageCreateInfo(flags, SurfaceTargetToImage(params.target), format, extent, | 179 | return ci; |
| 162 | params.num_levels, static_cast<u32>(params.GetNumLayers()), | ||
| 163 | sample_count, tiling, image_usage, vk::SharingMode::eExclusive, 0, | ||
| 164 | nullptr, vk::ImageLayout::eUndefined); | ||
| 165 | } | 180 | } |
| 166 | 181 | ||
| 167 | } // Anonymous namespace | 182 | } // Anonymous namespace |
| @@ -175,15 +190,13 @@ CachedSurface::CachedSurface(Core::System& system, const VKDevice& device, | |||
| 175 | memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{staging_pool} { | 190 | memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{staging_pool} { |
| 176 | if (params.IsBuffer()) { | 191 | if (params.IsBuffer()) { |
| 177 | buffer = CreateBuffer(device, params, host_memory_size); | 192 | buffer = CreateBuffer(device, params, host_memory_size); |
| 178 | commit = memory_manager.Commit(*buffer, false); | 193 | commit = memory_manager.Commit(buffer, false); |
| 179 | 194 | ||
| 180 | const auto buffer_view_ci = | 195 | const auto buffer_view_ci = |
| 181 | GenerateBufferViewCreateInfo(device, params, *buffer, host_memory_size); | 196 | GenerateBufferViewCreateInfo(device, params, *buffer, host_memory_size); |
| 182 | format = buffer_view_ci.format; | 197 | format = buffer_view_ci.format; |
| 183 | 198 | ||
| 184 | const auto dev = device.GetLogical(); | 199 | buffer_view = device.GetLogical().CreateBufferView(buffer_view_ci); |
| 185 | const auto& dld = device.GetDispatchLoader(); | ||
| 186 | buffer_view = dev.createBufferViewUnique(buffer_view_ci, nullptr, dld); | ||
| 187 | } else { | 200 | } else { |
| 188 | const auto image_ci = GenerateImageCreateInfo(device, params); | 201 | const auto image_ci = GenerateImageCreateInfo(device, params); |
| 189 | format = image_ci.format; | 202 | format = image_ci.format; |
| @@ -221,16 +234,15 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) { | |||
| 221 | // We can't copy images to buffers inside a renderpass | 234 | // We can't copy images to buffers inside a renderpass |
| 222 | scheduler.RequestOutsideRenderPassOperationContext(); | 235 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 223 | 236 | ||
| 224 | FullTransition(vk::PipelineStageFlagBits::eTransfer, vk::AccessFlagBits::eTransferRead, | 237 | FullTransition(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, |
| 225 | vk::ImageLayout::eTransferSrcOptimal); | 238 | VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL); |
| 226 | 239 | ||
| 227 | const auto& buffer = staging_pool.GetUnusedBuffer(host_memory_size, true); | 240 | const auto& buffer = staging_pool.GetUnusedBuffer(host_memory_size, true); |
| 228 | // TODO(Rodrigo): Do this in a single copy | 241 | // TODO(Rodrigo): Do this in a single copy |
| 229 | for (u32 level = 0; level < params.num_levels; ++level) { | 242 | for (u32 level = 0; level < params.num_levels; ++level) { |
| 230 | scheduler.Record([image = image->GetHandle(), buffer = *buffer.handle, | 243 | scheduler.Record([image = *image->GetHandle(), buffer = *buffer.handle, |
| 231 | copy = GetBufferImageCopy(level)](auto cmdbuf, auto& dld) { | 244 | copy = GetBufferImageCopy(level)](vk::CommandBuffer cmdbuf) { |
| 232 | cmdbuf.copyImageToBuffer(image, vk::ImageLayout::eTransferSrcOptimal, buffer, {copy}, | 245 | cmdbuf.CopyImageToBuffer(image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, buffer, copy); |
| 233 | dld); | ||
| 234 | }); | 246 | }); |
| 235 | } | 247 | } |
| 236 | scheduler.Finish(); | 248 | scheduler.Finish(); |
| @@ -257,15 +269,27 @@ void CachedSurface::UploadBuffer(const std::vector<u8>& staging_buffer) { | |||
| 257 | std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size); | 269 | std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size); |
| 258 | 270 | ||
| 259 | scheduler.Record([src_buffer = *src_buffer.handle, dst_buffer = *buffer, | 271 | scheduler.Record([src_buffer = *src_buffer.handle, dst_buffer = *buffer, |
| 260 | size = host_memory_size](auto cmdbuf, auto& dld) { | 272 | size = host_memory_size](vk::CommandBuffer cmdbuf) { |
| 261 | const vk::BufferCopy copy(0, 0, size); | 273 | VkBufferCopy copy; |
| 262 | cmdbuf.copyBuffer(src_buffer, dst_buffer, {copy}, dld); | 274 | copy.srcOffset = 0; |
| 263 | 275 | copy.dstOffset = 0; | |
| 264 | cmdbuf.pipelineBarrier( | 276 | copy.size = size; |
| 265 | vk::PipelineStageFlagBits::eTransfer, vk::PipelineStageFlagBits::eVertexShader, {}, {}, | 277 | cmdbuf.CopyBuffer(src_buffer, dst_buffer, copy); |
| 266 | {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, | 278 | |
| 267 | vk::AccessFlagBits::eShaderRead, 0, 0, dst_buffer, 0, size)}, | 279 | VkBufferMemoryBarrier barrier; |
| 268 | {}, dld); | 280 | barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; |
| 281 | barrier.pNext = nullptr; | ||
| 282 | barrier.srcAccessMask = VK_PIPELINE_STAGE_TRANSFER_BIT; | ||
| 283 | barrier.dstAccessMask = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT; | ||
| 284 | barrier.srcQueueFamilyIndex = VK_ACCESS_TRANSFER_WRITE_BIT; | ||
| 285 | barrier.dstQueueFamilyIndex = VK_ACCESS_SHADER_READ_BIT; | ||
| 286 | barrier.srcQueueFamilyIndex = 0; | ||
| 287 | barrier.dstQueueFamilyIndex = 0; | ||
| 288 | barrier.buffer = dst_buffer; | ||
| 289 | barrier.offset = 0; | ||
| 290 | barrier.size = size; | ||
| 291 | cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT, | ||
| 292 | 0, {}, barrier, {}); | ||
| 269 | }); | 293 | }); |
| 270 | } | 294 | } |
| 271 | 295 | ||
| @@ -273,43 +297,49 @@ void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) { | |||
| 273 | const auto& src_buffer = staging_pool.GetUnusedBuffer(host_memory_size, true); | 297 | const auto& src_buffer = staging_pool.GetUnusedBuffer(host_memory_size, true); |
| 274 | std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size); | 298 | std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size); |
| 275 | 299 | ||
| 276 | FullTransition(vk::PipelineStageFlagBits::eTransfer, vk::AccessFlagBits::eTransferWrite, | 300 | FullTransition(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, |
| 277 | vk::ImageLayout::eTransferDstOptimal); | 301 | VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL); |
| 278 | 302 | ||
| 279 | for (u32 level = 0; level < params.num_levels; ++level) { | 303 | for (u32 level = 0; level < params.num_levels; ++level) { |
| 280 | vk::BufferImageCopy copy = GetBufferImageCopy(level); | 304 | const VkBufferImageCopy copy = GetBufferImageCopy(level); |
| 281 | if (image->GetAspectMask() == | 305 | if (image->GetAspectMask() == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) { |
| 282 | (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil)) { | 306 | scheduler.Record([buffer = *src_buffer.handle, image = *image->GetHandle(), |
| 283 | vk::BufferImageCopy depth = copy; | 307 | copy](vk::CommandBuffer cmdbuf) { |
| 284 | vk::BufferImageCopy stencil = copy; | 308 | std::array<VkBufferImageCopy, 2> copies = {copy, copy}; |
| 285 | depth.imageSubresource.aspectMask = vk::ImageAspectFlagBits::eDepth; | 309 | copies[0].imageSubresource.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT; |
| 286 | stencil.imageSubresource.aspectMask = vk::ImageAspectFlagBits::eStencil; | 310 | copies[1].imageSubresource.aspectMask = VK_IMAGE_ASPECT_STENCIL_BIT; |
| 287 | scheduler.Record([buffer = *src_buffer.handle, image = image->GetHandle(), depth, | 311 | cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, |
| 288 | stencil](auto cmdbuf, auto& dld) { | 312 | copies); |
| 289 | cmdbuf.copyBufferToImage(buffer, image, vk::ImageLayout::eTransferDstOptimal, | ||
| 290 | {depth, stencil}, dld); | ||
| 291 | }); | 313 | }); |
| 292 | } else { | 314 | } else { |
| 293 | scheduler.Record([buffer = *src_buffer.handle, image = image->GetHandle(), | 315 | scheduler.Record([buffer = *src_buffer.handle, image = *image->GetHandle(), |
| 294 | copy](auto cmdbuf, auto& dld) { | 316 | copy](vk::CommandBuffer cmdbuf) { |
| 295 | cmdbuf.copyBufferToImage(buffer, image, vk::ImageLayout::eTransferDstOptimal, | 317 | cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy); |
| 296 | {copy}, dld); | ||
| 297 | }); | 318 | }); |
| 298 | } | 319 | } |
| 299 | } | 320 | } |
| 300 | } | 321 | } |
| 301 | 322 | ||
| 302 | vk::BufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const { | 323 | VkBufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const { |
| 303 | const u32 vk_depth = params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1; | 324 | VkBufferImageCopy copy; |
| 304 | const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted); | 325 | copy.bufferOffset = params.GetHostMipmapLevelOffset(level, is_converted); |
| 305 | 326 | copy.bufferRowLength = 0; | |
| 306 | return vk::BufferImageCopy( | 327 | copy.bufferImageHeight = 0; |
| 307 | mip_offset, 0, 0, | 328 | copy.imageSubresource.aspectMask = image->GetAspectMask(); |
| 308 | {image->GetAspectMask(), level, 0, static_cast<u32>(params.GetNumLayers())}, {0, 0, 0}, | 329 | copy.imageSubresource.mipLevel = level; |
| 309 | {params.GetMipWidth(level), params.GetMipHeight(level), vk_depth}); | 330 | copy.imageSubresource.baseArrayLayer = 0; |
| 331 | copy.imageSubresource.layerCount = static_cast<u32>(params.GetNumLayers()); | ||
| 332 | copy.imageOffset.x = 0; | ||
| 333 | copy.imageOffset.y = 0; | ||
| 334 | copy.imageOffset.z = 0; | ||
| 335 | copy.imageExtent.width = params.GetMipWidth(level); | ||
| 336 | copy.imageExtent.height = params.GetMipHeight(level); | ||
| 337 | copy.imageExtent.depth = | ||
| 338 | params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1; | ||
| 339 | return copy; | ||
| 310 | } | 340 | } |
| 311 | 341 | ||
| 312 | vk::ImageSubresourceRange CachedSurface::GetImageSubresourceRange() const { | 342 | VkImageSubresourceRange CachedSurface::GetImageSubresourceRange() const { |
| 313 | return {image->GetAspectMask(), 0, params.num_levels, 0, | 343 | return {image->GetAspectMask(), 0, params.num_levels, 0, |
| 314 | static_cast<u32>(params.GetNumLayers())}; | 344 | static_cast<u32>(params.GetNumLayers())}; |
| 315 | } | 345 | } |
| @@ -321,12 +351,12 @@ CachedSurfaceView::CachedSurfaceView(const VKDevice& device, CachedSurface& surf | |||
| 321 | aspect_mask{surface.GetAspectMask()}, device{device}, surface{surface}, | 351 | aspect_mask{surface.GetAspectMask()}, device{device}, surface{surface}, |
| 322 | base_layer{params.base_layer}, num_layers{params.num_layers}, base_level{params.base_level}, | 352 | base_layer{params.base_layer}, num_layers{params.num_layers}, base_level{params.base_level}, |
| 323 | num_levels{params.num_levels}, image_view_type{image ? GetImageViewType(params.target) | 353 | num_levels{params.num_levels}, image_view_type{image ? GetImageViewType(params.target) |
| 324 | : vk::ImageViewType{}} {} | 354 | : VK_IMAGE_VIEW_TYPE_1D} {} |
| 325 | 355 | ||
| 326 | CachedSurfaceView::~CachedSurfaceView() = default; | 356 | CachedSurfaceView::~CachedSurfaceView() = default; |
| 327 | 357 | ||
| 328 | vk::ImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y_source, | 358 | VkImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y_source, |
| 329 | SwizzleSource z_source, SwizzleSource w_source) { | 359 | SwizzleSource z_source, SwizzleSource w_source) { |
| 330 | const u32 swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source); | 360 | const u32 swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source); |
| 331 | if (last_image_view && last_swizzle == swizzle) { | 361 | if (last_image_view && last_swizzle == swizzle) { |
| 332 | return last_image_view; | 362 | return last_image_view; |
| @@ -351,37 +381,45 @@ vk::ImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource | |||
| 351 | 381 | ||
| 352 | // Games can sample depth or stencil values on textures. This is decided by the swizzle value on | 382 | // Games can sample depth or stencil values on textures. This is decided by the swizzle value on |
| 353 | // hardware. To emulate this on Vulkan we specify it in the aspect. | 383 | // hardware. To emulate this on Vulkan we specify it in the aspect. |
| 354 | vk::ImageAspectFlags aspect = aspect_mask; | 384 | VkImageAspectFlags aspect = aspect_mask; |
| 355 | if (aspect == (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil)) { | 385 | if (aspect == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) { |
| 356 | UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G); | 386 | UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G); |
| 357 | const bool is_first = x_source == SwizzleSource::R; | 387 | const bool is_first = x_source == SwizzleSource::R; |
| 358 | switch (params.pixel_format) { | 388 | switch (params.pixel_format) { |
| 359 | case VideoCore::Surface::PixelFormat::Z24S8: | 389 | case VideoCore::Surface::PixelFormat::Z24S8: |
| 360 | case VideoCore::Surface::PixelFormat::Z32FS8: | 390 | case VideoCore::Surface::PixelFormat::Z32FS8: |
| 361 | aspect = is_first ? vk::ImageAspectFlagBits::eDepth : vk::ImageAspectFlagBits::eStencil; | 391 | aspect = is_first ? VK_IMAGE_ASPECT_DEPTH_BIT : VK_IMAGE_ASPECT_STENCIL_BIT; |
| 362 | break; | 392 | break; |
| 363 | case VideoCore::Surface::PixelFormat::S8Z24: | 393 | case VideoCore::Surface::PixelFormat::S8Z24: |
| 364 | aspect = is_first ? vk::ImageAspectFlagBits::eStencil : vk::ImageAspectFlagBits::eDepth; | 394 | aspect = is_first ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT; |
| 365 | break; | 395 | break; |
| 366 | default: | 396 | default: |
| 367 | aspect = vk::ImageAspectFlagBits::eDepth; | 397 | aspect = VK_IMAGE_ASPECT_DEPTH_BIT; |
| 368 | UNIMPLEMENTED(); | 398 | UNIMPLEMENTED(); |
| 369 | } | 399 | } |
| 370 | 400 | ||
| 371 | // Vulkan doesn't seem to understand swizzling of a depth stencil image, use identity | 401 | // Vulkan doesn't seem to understand swizzling of a depth stencil image, use identity |
| 372 | swizzle_x = vk::ComponentSwizzle::eR; | 402 | swizzle_x = VK_COMPONENT_SWIZZLE_R; |
| 373 | swizzle_y = vk::ComponentSwizzle::eG; | 403 | swizzle_y = VK_COMPONENT_SWIZZLE_G; |
| 374 | swizzle_z = vk::ComponentSwizzle::eB; | 404 | swizzle_z = VK_COMPONENT_SWIZZLE_B; |
| 375 | swizzle_w = vk::ComponentSwizzle::eA; | 405 | swizzle_w = VK_COMPONENT_SWIZZLE_A; |
| 376 | } | 406 | } |
| 377 | 407 | ||
| 378 | const vk::ImageViewCreateInfo image_view_ci( | 408 | VkImageViewCreateInfo ci; |
| 379 | {}, surface.GetImageHandle(), image_view_type, surface.GetImage().GetFormat(), | 409 | ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; |
| 380 | {swizzle_x, swizzle_y, swizzle_z, swizzle_w}, | 410 | ci.pNext = nullptr; |
| 381 | {aspect, base_level, num_levels, base_layer, num_layers}); | 411 | ci.flags = 0; |
| 412 | ci.image = surface.GetImageHandle(); | ||
| 413 | ci.viewType = image_view_type; | ||
| 414 | ci.format = surface.GetImage().GetFormat(); | ||
| 415 | ci.components = {swizzle_x, swizzle_y, swizzle_z, swizzle_w}; | ||
| 416 | ci.subresourceRange.aspectMask = aspect; | ||
| 417 | ci.subresourceRange.baseMipLevel = base_level; | ||
| 418 | ci.subresourceRange.levelCount = num_levels; | ||
| 419 | ci.subresourceRange.baseArrayLayer = base_layer; | ||
| 420 | ci.subresourceRange.layerCount = num_layers; | ||
| 421 | image_view = device.GetLogical().CreateImageView(ci); | ||
| 382 | 422 | ||
| 383 | const auto dev = device.GetLogical(); | ||
| 384 | image_view = dev.createImageViewUnique(image_view_ci, nullptr, device.GetDispatchLoader()); | ||
| 385 | return last_image_view = *image_view; | 423 | return last_image_view = *image_view; |
| 386 | } | 424 | } |
| 387 | 425 | ||
| @@ -418,25 +456,36 @@ void VKTextureCache::ImageCopy(Surface& src_surface, Surface& dst_surface, | |||
| 418 | scheduler.RequestOutsideRenderPassOperationContext(); | 456 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 419 | 457 | ||
| 420 | src_surface->Transition(copy_params.source_z, copy_params.depth, copy_params.source_level, 1, | 458 | src_surface->Transition(copy_params.source_z, copy_params.depth, copy_params.source_level, 1, |
| 421 | vk::PipelineStageFlagBits::eTransfer, vk::AccessFlagBits::eTransferRead, | 459 | VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, |
| 422 | vk::ImageLayout::eTransferSrcOptimal); | 460 | VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL); |
| 423 | dst_surface->Transition( | 461 | dst_surface->Transition(dst_base_layer, num_layers, copy_params.dest_level, 1, |
| 424 | dst_base_layer, num_layers, copy_params.dest_level, 1, vk::PipelineStageFlagBits::eTransfer, | 462 | VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, |
| 425 | vk::AccessFlagBits::eTransferWrite, vk::ImageLayout::eTransferDstOptimal); | 463 | VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL); |
| 426 | 464 | ||
| 427 | const vk::ImageSubresourceLayers src_subresource( | 465 | VkImageCopy copy; |
| 428 | src_surface->GetAspectMask(), copy_params.source_level, copy_params.source_z, num_layers); | 466 | copy.srcSubresource.aspectMask = src_surface->GetAspectMask(); |
| 429 | const vk::ImageSubresourceLayers dst_subresource( | 467 | copy.srcSubresource.mipLevel = copy_params.source_level; |
| 430 | dst_surface->GetAspectMask(), copy_params.dest_level, dst_base_layer, num_layers); | 468 | copy.srcSubresource.baseArrayLayer = copy_params.source_z; |
| 431 | const vk::Offset3D src_offset(copy_params.source_x, copy_params.source_y, 0); | 469 | copy.srcSubresource.layerCount = num_layers; |
| 432 | const vk::Offset3D dst_offset(copy_params.dest_x, copy_params.dest_y, dst_offset_z); | 470 | copy.srcOffset.x = copy_params.source_x; |
| 433 | const vk::Extent3D extent(copy_params.width, copy_params.height, extent_z); | 471 | copy.srcOffset.y = copy_params.source_y; |
| 434 | const vk::ImageCopy copy(src_subresource, src_offset, dst_subresource, dst_offset, extent); | 472 | copy.srcOffset.z = 0; |
| 435 | const vk::Image src_image = src_surface->GetImageHandle(); | 473 | copy.dstSubresource.aspectMask = dst_surface->GetAspectMask(); |
| 436 | const vk::Image dst_image = dst_surface->GetImageHandle(); | 474 | copy.dstSubresource.mipLevel = copy_params.dest_level; |
| 437 | scheduler.Record([src_image, dst_image, copy](auto cmdbuf, auto& dld) { | 475 | copy.dstSubresource.baseArrayLayer = dst_base_layer; |
| 438 | cmdbuf.copyImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image, | 476 | copy.dstSubresource.layerCount = num_layers; |
| 439 | vk::ImageLayout::eTransferDstOptimal, {copy}, dld); | 477 | copy.dstOffset.x = copy_params.dest_x; |
| 478 | copy.dstOffset.y = copy_params.dest_y; | ||
| 479 | copy.dstOffset.z = dst_offset_z; | ||
| 480 | copy.extent.width = copy_params.width; | ||
| 481 | copy.extent.height = copy_params.height; | ||
| 482 | copy.extent.depth = extent_z; | ||
| 483 | |||
| 484 | const VkImage src_image = src_surface->GetImageHandle(); | ||
| 485 | const VkImage dst_image = dst_surface->GetImageHandle(); | ||
| 486 | scheduler.Record([src_image, dst_image, copy](vk::CommandBuffer cmdbuf) { | ||
| 487 | cmdbuf.CopyImage(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image, | ||
| 488 | VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy); | ||
| 440 | }); | 489 | }); |
| 441 | } | 490 | } |
| 442 | 491 | ||
| @@ -445,25 +494,34 @@ void VKTextureCache::ImageBlit(View& src_view, View& dst_view, | |||
| 445 | // We can't blit inside a renderpass | 494 | // We can't blit inside a renderpass |
| 446 | scheduler.RequestOutsideRenderPassOperationContext(); | 495 | scheduler.RequestOutsideRenderPassOperationContext(); |
| 447 | 496 | ||
| 448 | src_view->Transition(vk::ImageLayout::eTransferSrcOptimal, vk::PipelineStageFlagBits::eTransfer, | 497 | src_view->Transition(VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, VK_PIPELINE_STAGE_TRANSFER_BIT, |
| 449 | vk::AccessFlagBits::eTransferRead); | 498 | VK_ACCESS_TRANSFER_READ_BIT); |
| 450 | dst_view->Transition(vk::ImageLayout::eTransferDstOptimal, vk::PipelineStageFlagBits::eTransfer, | 499 | dst_view->Transition(VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_PIPELINE_STAGE_TRANSFER_BIT, |
| 451 | vk::AccessFlagBits::eTransferWrite); | 500 | VK_ACCESS_TRANSFER_WRITE_BIT); |
| 452 | 501 | ||
| 453 | const auto& cfg = copy_config; | 502 | VkImageBlit blit; |
| 454 | const auto src_top_left = vk::Offset3D(cfg.src_rect.left, cfg.src_rect.top, 0); | 503 | blit.srcSubresource = src_view->GetImageSubresourceLayers(); |
| 455 | const auto src_bot_right = vk::Offset3D(cfg.src_rect.right, cfg.src_rect.bottom, 1); | 504 | blit.srcOffsets[0].x = copy_config.src_rect.left; |
| 456 | const auto dst_top_left = vk::Offset3D(cfg.dst_rect.left, cfg.dst_rect.top, 0); | 505 | blit.srcOffsets[0].y = copy_config.src_rect.top; |
| 457 | const auto dst_bot_right = vk::Offset3D(cfg.dst_rect.right, cfg.dst_rect.bottom, 1); | 506 | blit.srcOffsets[0].z = 0; |
| 458 | const vk::ImageBlit blit(src_view->GetImageSubresourceLayers(), {src_top_left, src_bot_right}, | 507 | blit.srcOffsets[1].x = copy_config.src_rect.right; |
| 459 | dst_view->GetImageSubresourceLayers(), {dst_top_left, dst_bot_right}); | 508 | blit.srcOffsets[1].y = copy_config.src_rect.bottom; |
| 509 | blit.srcOffsets[1].z = 1; | ||
| 510 | blit.dstSubresource = dst_view->GetImageSubresourceLayers(); | ||
| 511 | blit.dstOffsets[0].x = copy_config.dst_rect.left; | ||
| 512 | blit.dstOffsets[0].y = copy_config.dst_rect.top; | ||
| 513 | blit.dstOffsets[0].z = 0; | ||
| 514 | blit.dstOffsets[1].x = copy_config.dst_rect.right; | ||
| 515 | blit.dstOffsets[1].y = copy_config.dst_rect.bottom; | ||
| 516 | blit.dstOffsets[1].z = 1; | ||
| 517 | |||
| 460 | const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear; | 518 | const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear; |
| 461 | 519 | ||
| 462 | scheduler.Record([src_image = src_view->GetImage(), dst_image = dst_view->GetImage(), blit, | 520 | scheduler.Record([src_image = src_view->GetImage(), dst_image = dst_view->GetImage(), blit, |
| 463 | is_linear](auto cmdbuf, auto& dld) { | 521 | is_linear](vk::CommandBuffer cmdbuf) { |
| 464 | cmdbuf.blitImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image, | 522 | cmdbuf.BlitImage(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image, |
| 465 | vk::ImageLayout::eTransferDstOptimal, {blit}, | 523 | VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, blit, |
| 466 | is_linear ? vk::Filter::eLinear : vk::Filter::eNearest, dld); | 524 | is_linear ? VK_FILTER_LINEAR : VK_FILTER_NEAREST); |
| 467 | }); | 525 | }); |
| 468 | } | 526 | } |
| 469 | 527 | ||
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 22e3d34de..115595f28 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h | |||
| @@ -13,10 +13,10 @@ | |||
| 13 | #include "common/math_util.h" | 13 | #include "common/math_util.h" |
| 14 | #include "video_core/gpu.h" | 14 | #include "video_core/gpu.h" |
| 15 | #include "video_core/rasterizer_cache.h" | 15 | #include "video_core/rasterizer_cache.h" |
| 16 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 17 | #include "video_core/renderer_vulkan/vk_image.h" | 16 | #include "video_core/renderer_vulkan/vk_image.h" |
| 18 | #include "video_core/renderer_vulkan/vk_memory_manager.h" | 17 | #include "video_core/renderer_vulkan/vk_memory_manager.h" |
| 19 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 18 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 19 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 20 | #include "video_core/texture_cache/surface_base.h" | 20 | #include "video_core/texture_cache/surface_base.h" |
| 21 | #include "video_core/texture_cache/texture_cache.h" | 21 | #include "video_core/texture_cache/texture_cache.h" |
| 22 | #include "video_core/textures/decoders.h" | 22 | #include "video_core/textures/decoders.h" |
| @@ -60,15 +60,15 @@ public: | |||
| 60 | void UploadTexture(const std::vector<u8>& staging_buffer) override; | 60 | void UploadTexture(const std::vector<u8>& staging_buffer) override; |
| 61 | void DownloadTexture(std::vector<u8>& staging_buffer) override; | 61 | void DownloadTexture(std::vector<u8>& staging_buffer) override; |
| 62 | 62 | ||
| 63 | void FullTransition(vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, | 63 | void FullTransition(VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, |
| 64 | vk::ImageLayout new_layout) { | 64 | VkImageLayout new_layout) { |
| 65 | image->Transition(0, static_cast<u32>(params.GetNumLayers()), 0, params.num_levels, | 65 | image->Transition(0, static_cast<u32>(params.GetNumLayers()), 0, params.num_levels, |
| 66 | new_stage_mask, new_access, new_layout); | 66 | new_stage_mask, new_access, new_layout); |
| 67 | } | 67 | } |
| 68 | 68 | ||
| 69 | void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, | 69 | void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, |
| 70 | vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, | 70 | VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, |
| 71 | vk::ImageLayout new_layout) { | 71 | VkImageLayout new_layout) { |
| 72 | image->Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask, | 72 | image->Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask, |
| 73 | new_access, new_layout); | 73 | new_access, new_layout); |
| 74 | } | 74 | } |
| @@ -81,15 +81,15 @@ public: | |||
| 81 | return *image; | 81 | return *image; |
| 82 | } | 82 | } |
| 83 | 83 | ||
| 84 | vk::Image GetImageHandle() const { | 84 | VkImage GetImageHandle() const { |
| 85 | return image->GetHandle(); | 85 | return *image->GetHandle(); |
| 86 | } | 86 | } |
| 87 | 87 | ||
| 88 | vk::ImageAspectFlags GetAspectMask() const { | 88 | VkImageAspectFlags GetAspectMask() const { |
| 89 | return image->GetAspectMask(); | 89 | return image->GetAspectMask(); |
| 90 | } | 90 | } |
| 91 | 91 | ||
| 92 | vk::BufferView GetBufferViewHandle() const { | 92 | VkBufferView GetBufferViewHandle() const { |
| 93 | return *buffer_view; | 93 | return *buffer_view; |
| 94 | } | 94 | } |
| 95 | 95 | ||
| @@ -104,9 +104,9 @@ private: | |||
| 104 | 104 | ||
| 105 | void UploadImage(const std::vector<u8>& staging_buffer); | 105 | void UploadImage(const std::vector<u8>& staging_buffer); |
| 106 | 106 | ||
| 107 | vk::BufferImageCopy GetBufferImageCopy(u32 level) const; | 107 | VkBufferImageCopy GetBufferImageCopy(u32 level) const; |
| 108 | 108 | ||
| 109 | vk::ImageSubresourceRange GetImageSubresourceRange() const; | 109 | VkImageSubresourceRange GetImageSubresourceRange() const; |
| 110 | 110 | ||
| 111 | Core::System& system; | 111 | Core::System& system; |
| 112 | const VKDevice& device; | 112 | const VKDevice& device; |
| @@ -116,11 +116,11 @@ private: | |||
| 116 | VKStagingBufferPool& staging_pool; | 116 | VKStagingBufferPool& staging_pool; |
| 117 | 117 | ||
| 118 | std::optional<VKImage> image; | 118 | std::optional<VKImage> image; |
| 119 | UniqueBuffer buffer; | 119 | vk::Buffer buffer; |
| 120 | UniqueBufferView buffer_view; | 120 | vk::BufferView buffer_view; |
| 121 | VKMemoryCommit commit; | 121 | VKMemoryCommit commit; |
| 122 | 122 | ||
| 123 | vk::Format format; | 123 | VkFormat format = VK_FORMAT_UNDEFINED; |
| 124 | }; | 124 | }; |
| 125 | 125 | ||
| 126 | class CachedSurfaceView final : public VideoCommon::ViewBase { | 126 | class CachedSurfaceView final : public VideoCommon::ViewBase { |
| @@ -129,16 +129,16 @@ public: | |||
| 129 | const ViewParams& params, bool is_proxy); | 129 | const ViewParams& params, bool is_proxy); |
| 130 | ~CachedSurfaceView(); | 130 | ~CachedSurfaceView(); |
| 131 | 131 | ||
| 132 | vk::ImageView GetHandle(Tegra::Texture::SwizzleSource x_source, | 132 | VkImageView GetHandle(Tegra::Texture::SwizzleSource x_source, |
| 133 | Tegra::Texture::SwizzleSource y_source, | 133 | Tegra::Texture::SwizzleSource y_source, |
| 134 | Tegra::Texture::SwizzleSource z_source, | 134 | Tegra::Texture::SwizzleSource z_source, |
| 135 | Tegra::Texture::SwizzleSource w_source); | 135 | Tegra::Texture::SwizzleSource w_source); |
| 136 | 136 | ||
| 137 | bool IsSameSurface(const CachedSurfaceView& rhs) const { | 137 | bool IsSameSurface(const CachedSurfaceView& rhs) const { |
| 138 | return &surface == &rhs.surface; | 138 | return &surface == &rhs.surface; |
| 139 | } | 139 | } |
| 140 | 140 | ||
| 141 | vk::ImageView GetHandle() { | 141 | VkImageView GetHandle() { |
| 142 | return GetHandle(Tegra::Texture::SwizzleSource::R, Tegra::Texture::SwizzleSource::G, | 142 | return GetHandle(Tegra::Texture::SwizzleSource::R, Tegra::Texture::SwizzleSource::G, |
| 143 | Tegra::Texture::SwizzleSource::B, Tegra::Texture::SwizzleSource::A); | 143 | Tegra::Texture::SwizzleSource::B, Tegra::Texture::SwizzleSource::A); |
| 144 | } | 144 | } |
| @@ -159,24 +159,24 @@ public: | |||
| 159 | return buffer_view; | 159 | return buffer_view; |
| 160 | } | 160 | } |
| 161 | 161 | ||
| 162 | vk::Image GetImage() const { | 162 | VkImage GetImage() const { |
| 163 | return image; | 163 | return image; |
| 164 | } | 164 | } |
| 165 | 165 | ||
| 166 | vk::BufferView GetBufferView() const { | 166 | VkBufferView GetBufferView() const { |
| 167 | return buffer_view; | 167 | return buffer_view; |
| 168 | } | 168 | } |
| 169 | 169 | ||
| 170 | vk::ImageSubresourceRange GetImageSubresourceRange() const { | 170 | VkImageSubresourceRange GetImageSubresourceRange() const { |
| 171 | return {aspect_mask, base_level, num_levels, base_layer, num_layers}; | 171 | return {aspect_mask, base_level, num_levels, base_layer, num_layers}; |
| 172 | } | 172 | } |
| 173 | 173 | ||
| 174 | vk::ImageSubresourceLayers GetImageSubresourceLayers() const { | 174 | VkImageSubresourceLayers GetImageSubresourceLayers() const { |
| 175 | return {surface.GetAspectMask(), base_level, base_layer, num_layers}; | 175 | return {surface.GetAspectMask(), base_level, base_layer, num_layers}; |
| 176 | } | 176 | } |
| 177 | 177 | ||
| 178 | void Transition(vk::ImageLayout new_layout, vk::PipelineStageFlags new_stage_mask, | 178 | void Transition(VkImageLayout new_layout, VkPipelineStageFlags new_stage_mask, |
| 179 | vk::AccessFlags new_access) const { | 179 | VkAccessFlags new_access) const { |
| 180 | surface.Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask, | 180 | surface.Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask, |
| 181 | new_access, new_layout); | 181 | new_access, new_layout); |
| 182 | } | 182 | } |
| @@ -196,9 +196,9 @@ private: | |||
| 196 | 196 | ||
| 197 | // Store a copy of these values to avoid double dereference when reading them | 197 | // Store a copy of these values to avoid double dereference when reading them |
| 198 | const SurfaceParams params; | 198 | const SurfaceParams params; |
| 199 | const vk::Image image; | 199 | const VkImage image; |
| 200 | const vk::BufferView buffer_view; | 200 | const VkBufferView buffer_view; |
| 201 | const vk::ImageAspectFlags aspect_mask; | 201 | const VkImageAspectFlags aspect_mask; |
| 202 | 202 | ||
| 203 | const VKDevice& device; | 203 | const VKDevice& device; |
| 204 | CachedSurface& surface; | 204 | CachedSurface& surface; |
| @@ -206,12 +206,12 @@ private: | |||
| 206 | const u32 num_layers; | 206 | const u32 num_layers; |
| 207 | const u32 base_level; | 207 | const u32 base_level; |
| 208 | const u32 num_levels; | 208 | const u32 num_levels; |
| 209 | const vk::ImageViewType image_view_type; | 209 | const VkImageViewType image_view_type; |
| 210 | 210 | ||
| 211 | vk::ImageView last_image_view; | 211 | VkImageView last_image_view = nullptr; |
| 212 | u32 last_swizzle{}; | 212 | u32 last_swizzle = 0; |
| 213 | 213 | ||
| 214 | std::unordered_map<u32, UniqueImageView> view_cache; | 214 | std::unordered_map<u32, vk::ImageView> view_cache; |
| 215 | }; | 215 | }; |
| 216 | 216 | ||
| 217 | class VKTextureCache final : public TextureCacheBase { | 217 | class VKTextureCache final : public TextureCacheBase { |
diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp index 0e577b9ff..4bfec0077 100644 --- a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp +++ b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp | |||
| @@ -7,10 +7,10 @@ | |||
| 7 | 7 | ||
| 8 | #include "common/assert.h" | 8 | #include "common/assert.h" |
| 9 | #include "common/logging/log.h" | 9 | #include "common/logging/log.h" |
| 10 | #include "video_core/renderer_vulkan/declarations.h" | ||
| 11 | #include "video_core/renderer_vulkan/vk_device.h" | 10 | #include "video_core/renderer_vulkan/vk_device.h" |
| 12 | #include "video_core/renderer_vulkan/vk_scheduler.h" | 11 | #include "video_core/renderer_vulkan/vk_scheduler.h" |
| 13 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" | 12 | #include "video_core/renderer_vulkan/vk_update_descriptor.h" |
| 13 | #include "video_core/renderer_vulkan/wrapper.h" | ||
| 14 | 14 | ||
| 15 | namespace Vulkan { | 15 | namespace Vulkan { |
| 16 | 16 | ||
| @@ -27,8 +27,8 @@ void VKUpdateDescriptorQueue::Acquire() { | |||
| 27 | entries.clear(); | 27 | entries.clear(); |
| 28 | } | 28 | } |
| 29 | 29 | ||
| 30 | void VKUpdateDescriptorQueue::Send(vk::DescriptorUpdateTemplate update_template, | 30 | void VKUpdateDescriptorQueue::Send(VkDescriptorUpdateTemplateKHR update_template, |
| 31 | vk::DescriptorSet set) { | 31 | VkDescriptorSet set) { |
| 32 | if (payload.size() + entries.size() >= payload.max_size()) { | 32 | if (payload.size() + entries.size() >= payload.max_size()) { |
| 33 | LOG_WARNING(Render_Vulkan, "Payload overflow, waiting for worker thread"); | 33 | LOG_WARNING(Render_Vulkan, "Payload overflow, waiting for worker thread"); |
| 34 | scheduler.WaitWorker(); | 34 | scheduler.WaitWorker(); |
| @@ -37,21 +37,21 @@ void VKUpdateDescriptorQueue::Send(vk::DescriptorUpdateTemplate update_template, | |||
| 37 | 37 | ||
| 38 | const auto payload_start = payload.data() + payload.size(); | 38 | const auto payload_start = payload.data() + payload.size(); |
| 39 | for (const auto& entry : entries) { | 39 | for (const auto& entry : entries) { |
| 40 | if (const auto image = std::get_if<vk::DescriptorImageInfo>(&entry)) { | 40 | if (const auto image = std::get_if<VkDescriptorImageInfo>(&entry)) { |
| 41 | payload.push_back(*image); | 41 | payload.push_back(*image); |
| 42 | } else if (const auto buffer = std::get_if<Buffer>(&entry)) { | 42 | } else if (const auto buffer = std::get_if<Buffer>(&entry)) { |
| 43 | payload.emplace_back(*buffer->buffer, buffer->offset, buffer->size); | 43 | payload.emplace_back(*buffer->buffer, buffer->offset, buffer->size); |
| 44 | } else if (const auto texel = std::get_if<vk::BufferView>(&entry)) { | 44 | } else if (const auto texel = std::get_if<VkBufferView>(&entry)) { |
| 45 | payload.push_back(*texel); | 45 | payload.push_back(*texel); |
| 46 | } else { | 46 | } else { |
| 47 | UNREACHABLE(); | 47 | UNREACHABLE(); |
| 48 | } | 48 | } |
| 49 | } | 49 | } |
| 50 | 50 | ||
| 51 | scheduler.Record([dev = device.GetLogical(), payload_start, set, | 51 | scheduler.Record( |
| 52 | update_template]([[maybe_unused]] auto cmdbuf, auto& dld) { | 52 | [payload_start, set, update_template, logical = &device.GetLogical()](vk::CommandBuffer) { |
| 53 | dev.updateDescriptorSetWithTemplate(set, update_template, payload_start, dld); | 53 | logical->UpdateDescriptorSet(set, update_template, payload_start); |
| 54 | }); | 54 | }); |
| 55 | } | 55 | } |
| 56 | 56 | ||
| 57 | } // namespace Vulkan | 57 | } // namespace Vulkan |
diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.h b/src/video_core/renderer_vulkan/vk_update_descriptor.h index 8c825aa29..a9e3d5dba 100644 --- a/src/video_core/renderer_vulkan/vk_update_descriptor.h +++ b/src/video_core/renderer_vulkan/vk_update_descriptor.h | |||
| @@ -9,7 +9,7 @@ | |||
| 9 | #include <boost/container/static_vector.hpp> | 9 | #include <boost/container/static_vector.hpp> |
| 10 | 10 | ||
| 11 | #include "common/common_types.h" | 11 | #include "common/common_types.h" |
| 12 | #include "video_core/renderer_vulkan/declarations.h" | 12 | #include "video_core/renderer_vulkan/wrapper.h" |
| 13 | 13 | ||
| 14 | namespace Vulkan { | 14 | namespace Vulkan { |
| 15 | 15 | ||
| @@ -20,18 +20,18 @@ class DescriptorUpdateEntry { | |||
| 20 | public: | 20 | public: |
| 21 | explicit DescriptorUpdateEntry() : image{} {} | 21 | explicit DescriptorUpdateEntry() : image{} {} |
| 22 | 22 | ||
| 23 | DescriptorUpdateEntry(vk::DescriptorImageInfo image) : image{image} {} | 23 | DescriptorUpdateEntry(VkDescriptorImageInfo image) : image{image} {} |
| 24 | 24 | ||
| 25 | DescriptorUpdateEntry(vk::Buffer buffer, vk::DeviceSize offset, vk::DeviceSize size) | 25 | DescriptorUpdateEntry(VkBuffer buffer, VkDeviceSize offset, VkDeviceSize size) |
| 26 | : buffer{buffer, offset, size} {} | 26 | : buffer{buffer, offset, size} {} |
| 27 | 27 | ||
| 28 | DescriptorUpdateEntry(vk::BufferView texel_buffer) : texel_buffer{texel_buffer} {} | 28 | DescriptorUpdateEntry(VkBufferView texel_buffer) : texel_buffer{texel_buffer} {} |
| 29 | 29 | ||
| 30 | private: | 30 | private: |
| 31 | union { | 31 | union { |
| 32 | vk::DescriptorImageInfo image; | 32 | VkDescriptorImageInfo image; |
| 33 | vk::DescriptorBufferInfo buffer; | 33 | VkDescriptorBufferInfo buffer; |
| 34 | vk::BufferView texel_buffer; | 34 | VkBufferView texel_buffer; |
| 35 | }; | 35 | }; |
| 36 | }; | 36 | }; |
| 37 | 37 | ||
| @@ -44,37 +44,35 @@ public: | |||
| 44 | 44 | ||
| 45 | void Acquire(); | 45 | void Acquire(); |
| 46 | 46 | ||
| 47 | void Send(vk::DescriptorUpdateTemplate update_template, vk::DescriptorSet set); | 47 | void Send(VkDescriptorUpdateTemplateKHR update_template, VkDescriptorSet set); |
| 48 | 48 | ||
| 49 | void AddSampledImage(vk::Sampler sampler, vk::ImageView image_view) { | 49 | void AddSampledImage(VkSampler sampler, VkImageView image_view) { |
| 50 | entries.emplace_back(vk::DescriptorImageInfo{sampler, image_view, {}}); | 50 | entries.emplace_back(VkDescriptorImageInfo{sampler, image_view, {}}); |
| 51 | } | 51 | } |
| 52 | 52 | ||
| 53 | void AddImage(vk::ImageView image_view) { | 53 | void AddImage(VkImageView image_view) { |
| 54 | entries.emplace_back(vk::DescriptorImageInfo{{}, image_view, {}}); | 54 | entries.emplace_back(VkDescriptorImageInfo{{}, image_view, {}}); |
| 55 | } | 55 | } |
| 56 | 56 | ||
| 57 | void AddBuffer(const vk::Buffer* buffer, u64 offset, std::size_t size) { | 57 | void AddBuffer(const VkBuffer* buffer, u64 offset, std::size_t size) { |
| 58 | entries.push_back(Buffer{buffer, offset, size}); | 58 | entries.push_back(Buffer{buffer, offset, size}); |
| 59 | } | 59 | } |
| 60 | 60 | ||
| 61 | void AddTexelBuffer(vk::BufferView texel_buffer) { | 61 | void AddTexelBuffer(VkBufferView texel_buffer) { |
| 62 | entries.emplace_back(texel_buffer); | 62 | entries.emplace_back(texel_buffer); |
| 63 | } | 63 | } |
| 64 | 64 | ||
| 65 | vk::ImageLayout* GetLastImageLayout() { | 65 | VkImageLayout* GetLastImageLayout() { |
| 66 | return &std::get<vk::DescriptorImageInfo>(entries.back()).imageLayout; | 66 | return &std::get<VkDescriptorImageInfo>(entries.back()).imageLayout; |
| 67 | } | 67 | } |
| 68 | 68 | ||
| 69 | private: | 69 | private: |
| 70 | struct Buffer { | 70 | struct Buffer { |
| 71 | const vk::Buffer* buffer{}; | 71 | const VkBuffer* buffer = nullptr; |
| 72 | u64 offset{}; | 72 | u64 offset = 0; |
| 73 | std::size_t size{}; | 73 | std::size_t size = 0; |
| 74 | }; | 74 | }; |
| 75 | using Variant = std::variant<vk::DescriptorImageInfo, Buffer, vk::BufferView>; | 75 | using Variant = std::variant<VkDescriptorImageInfo, Buffer, VkBufferView>; |
| 76 | // Old gcc versions don't consider this trivially copyable. | ||
| 77 | // static_assert(std::is_trivially_copyable_v<Variant>); | ||
| 78 | 76 | ||
| 79 | const VKDevice& device; | 77 | const VKDevice& device; |
| 80 | VKScheduler& scheduler; | 78 | VKScheduler& scheduler; |