The goal of this project is to be a feature rich & performant renderer without any precomputed scene structures or light transport. The visual target is to get as close as possible to path-traced reference lighting. The performance target is to stay at a stable 6.94ms/144hz framerate at 1080p on a RTX 2080 TI. However, this is ultimately a hobby project & thus likely to remain perpetually a work in progress.
Ray Traced Global Illumination
- Screen space radiance cache (for ray traced radiance).
- Voxel cone traced world space radiance cache.
- Basic ReSTIR for diffuse hits.
- Recurrent blur denoiser.
- Volumetrics integration.
- Rough specular approximation from diffuse SH1.
- Async TLAS & BLAS builds.
Analytic Lights
- Spot, point & directional lights.
- Spherical area light estimation for all types.
- Poisson PCF shadow filtering.
- PCSS shadow filtering for directional lights.
- Bend screen space shadows.
- Directional shadow cascades.
- Clustered forward rendering.
Shading
- PBR BxDF
- Chan Diffuse term
- GGX normal distribution term.
- Smith GGX correlated visibility term.
- Hanrahan Krueger sub surface term.
- Clear coat support.
- Sheen support.
- PBR volumetric fog.
- Octahedron mapped GGX HDR IBL (used for rt & sky).
- Spherical harmonics convolution from from IBL (for volumetrics).
- GBuffers (normals, roughness, min/max/avg hierarchical depth).
Post Processing
- Temporal antialiasing.
- HDR bloom.
- Luminance histogram based auto exposure.
- Bokeh depth of field & auto focus.
- GT Uchimura tone mapping.
- Color grading.
- Film grain.
- Vignette.
Meshlet Rendering
- 16 byte vertex packing.
- unorm 16 xyz vertex position relative to submesh aabb.
- r4g4b4 color
- tangent sign 1 bit
- half2 texcoord
- 10r10b10g2a quaternion.
- directed acyclic graph lod levels.
- task shader cone culling.
- task shader frustum culling.
- global virtual buffers for all geometry resources.
- dynamically batched draw calls.
- material batching.
- all material properties in a single buffer.
- shader specific property layout & offset to property buffer.
Architecture
- Vulkan 1.3 rendering backend.
- Async compute & transfer queues.
- Multiple buffering.
- Automatic resource barrier management.
- Entity component system.
- Custom compressed mesh & shader formats.
- Asset hot reloading
- Console command interface.
Planned
- geometry lods.
- skinned geometry.
- GPU culling.
- SH probe based world space radiance cache.
- GPU particle engine.
- DX12 backend (unlikely... cant do async ray trace afaik).
- Serialized scene representation.
Required Vulkan & GLSL Features & Extensions
Vulkan Version 1.3
SPRIV Version 1.6
GLSL Extensions:
- GL_EXT_shader_explicit_arithmetic_types
- GL_EXT_nonuniform_qualifier
- GL_ARB_shader_viewport_layer_array
- GL_EXT_ray_tracing
- GL_KHR_shader_subgroup_ballot
- GL_EXT_control_flow_attributes
- GL_KHR_shader_subgroup_ballot
- GL_KHR_shader_subgroup_arithmetic
- GL_KHR_shader_subgroup_shuffle
- GL_KHR_shader_subgroup_vote
- GL_EXT_ray_tracing_position_fetch
- GL_EXT_mesh_shader
- GL_EXT_shader_image_load_formatted
Validation layers:
- VK_LAYER_KHRONOS_validation
Instance Extensions:
- VK_EXT_debug_utils
- VK_KHR_get_physical_device_properties2
Device Extensions:
- VK_KHR_surface
- VK_KHR_win32_surface
- VK_KHR_swapchain
- VK_KHR_deferred_host_operations
- VK_KHR_acceleration_structure
- VK_KHR_ray_tracing_pipeline
- VK_KHR_ray_query
- VK_EXT_conservative_rasterization
- VK_KHR_ray_tracing_position_fetch
- VK_EXT_mesh_shader
- VK_KHR_fragment_shading_rate
Physical Device Requirements:
- alphaToOne
- shaderImageGatherExtended
- sparseBinding
- sparseResidencyBuffer
- samplerAnisotropy
- multiViewport
- shaderSampledImageArrayDynamicIndexing
- shaderUniformBufferArrayDynamicIndexing
- shaderFloat64
- shaderInt16
- shaderInt64
- imageCubeArray
- fragmentStoresAndAtomics
- multiDrawIndirect
- shaderStorageImageReadWithoutFormat
- shaderStorageImageWriteWithoutFormat
- storageBuffer16BitAccess
- uniformAndStorageBuffer16BitAccess
- storagePushConstant16
- shaderUniformBufferArrayNonUniformIndexing
- shaderSampledImageArrayNonUniformIndexing
- runtimeDescriptorArray
- descriptorBindingVariableDescriptorCount
- descriptorBindingPartiallyBound
- scalarBlockLayout
- shaderFloat16
- shaderInt8
- shaderOutputViewportIndex
- shaderOutputLayer
- bufferDeviceAddress
- timelineSemaphore
- hostQueryReset
- accelerationStructure
- rayTracingPipeline
- rayQuery
- maintenance4
- rayTracingPositionFetch
- taskShader
- meshShader
- multiviewMeshShader
- primitiveFragmentShadingRateMeshShader
- pipelineFragmentShadingRate
Libraries & Dependencies
Warning The project has only been tested on Windows 10, support for systems or versions is untested.
Warning The project has only been tested on an NVIDIA RTX 2080 TI, support for other vendors or devices is untested.
Warning Building requires MSBuild and a C++17 standard compatible version of MSCV or clang-cl.
Warning Building only the PKRenderer project requires a compatible build of PKAssetTools.
Build using MSBuild CLI:
msbuild.exe PKRenderer.sln /property:Configuration=Release or
msbuild.exe PKRenderer.sln /property:Configuration=ClangRelease or
msbuild.exe PKRenderer.sln /property:Configuration=Debug or
msbuild.exe PKRenderer.sln /property:Configuration=ClangDebug or
through Visual Studio (you know how that goes).
Transfer to a more portable build system is planned but has a very low priority.