Skip to content

Latest commit

 

History

History
61 lines (42 loc) · 13 KB

README.md

File metadata and controls

61 lines (42 loc) · 13 KB
Raw

vku

Linux Windows

A set of Vulkan utilities based on RAII idiom and Modern C++ features.

Features

  • Based on C++20 module and (optionally) C++23 standard library module: just single line import vku; will cover all. Module support is mandatory! (No header file provided.)
  • Can be easily integrated to your existing project by vcpkg.
  • RAII handle for buffer and image with Vulkan Memory Allocator (VMA), with rich set of information.
  • Bootstrapping for physical device, device and VMA allocator generation with fully customizable configurations and compile time safety.
    • Automatic physical device selection based on queue family and extension requirements.
    • Show accepted/rejected physical devices with your requirements.
  • Templated descriptor set layout, compile-time descriptor write type inferring and pool size estimation.
  • Managing multiple attachment image and views with MSAA and VK_KHR_dynamic_rendering support.
  • Combining runtime shader compilation, shader loading and pipeline creation to single function.
  • Executing hierarchical commands across the multiple queues with most efficient way using VK_KHR_timeline_semaphore, by runtime dependency graph solver.
  • Forwarding your Vulkan code logics by "inlining" the structure initialization.
  • Useful utility functions including debugging support.
  • And more...

Projects that uses vku

I wrote several Vulkan projects using vku.

Project Description
Vulkan glTF Viewer screenshot
Vulkan glTF Viewer		 Blazingly fast Vulkan glTF viewer. Fully bindless, GPU driven, async compute/transfer support.
Detailed features
  • Fully bindless (descriptor set only updated in the model loading time), vertex pulling based, and GPU driven indirect rendering. All scene nodes can be rendered with up to 8 draw calls regardless of their material properties and complexity.
  • PBR (physical based rendering) + IBL (image based lighting), support asynchronous runtime spherical harmonics and prefiltered map generation using only compute shader.
  • Can render opaque, alpha tested (using Alpha to Coverage) and translucent (using Weighted Blended OIT) materials.
  • Directly memcpy the glTF buffer memory into GPU using multi-thread, dedicated transfer queue without pre-processing, makes faster resource loading time.
  • Unlimited texture coordinate indices count (can render a primitive that has arbitrary TEXCOORD_<i>) using buffer device address approach.
  • Pixel-perfect mouse picking, outline rendering and transformation gizmo for scene nodes.
  • Can mainpulate the node visibility based on hierarchy.
vk-deferred screenshot
vk-deferred		 Application that demonstrates the deferred rendering with 1500 lights.
vk-weighted-blended screenshot
vk-weighted-blended		 Rendering 25 opaque and 100 translucent objects using Weighted Blended OIT.

Getting Started

Build Steps

vku can be used with CMake find_package, FetchContent, CPM.cmake or vcpkg (using overlay ports). See Using vku page for the details.

Note

vku can only be used with overlay ports when using vcpkg for now. I'm currently planning for make vku to be available with vcpkg official ports.

Tutorials

I wrote step-by-step tutorials of building some Vulkan applications using vku for best understanding. For each tutorial, it first shows how Vulkan-Hpp (with RAII binding) does, then which portion of the code can be simplified using vku. I really recommend you to read them.

Tutorial Description
Tutorial 1
Hello Triangle		 It demonstrates the most simplest Vulkan application with graphics operation, rendering a colored triangle onto image, and additionally persist the image into the file.

You'll learn:
  • Bootstrapping physical device, device and VMA allocator creation with vku::Gpu<QueueFamilies, Queues>.
  • Creating RAII buffer (vku::MappedBuffer) and image (vku::AllocatedImage).
  • Creating graphics pipeline with predefined states (vku::getDefaultGraphicsPipelineCreateInfo).
  • Allocate command buffer, record commands, then submit it at once (vku::executeSingleCommand)
Tutorial 2
Hello Triangle Window (In Progress)		 It renders the triangle in a resizable GLFW window.

You'll learn:
  • Customizing vku::Gpu queue family selection configuration.
  • Use vku::AttachmentGroup to manage multiple swapchain images and image views into single object.
  • Structuring your Vulkan project by separating the objects into the single-responsibility classes.
  • Implementing Frames in flight.
Tutorial 3
Rotating Box (In Progress)		 It renders a textured rotating box with 4x MSAA.

You'll learn:
  • Use vku::MsaaAttachmentGroup to manage the whole color/resolve/depth/swapchain images at once.
  • Creating typed descriptor set layout with vku::DescriptorSet<vk::DescriptorType...>.
  • Estimating the required descriptor pool size from descriptor sets automatically using vku::PoolSize.
  • Use inferred descriptor write type with vku::DescriptorSet<DescriptorSetLayout>.

Tests

It uses GitHub Runner to test the build availability with Clang (Linux) or MSVC (Windows). See workflow files for details.

Also, some parts of codes have the test codes to validate their intended behaviors. You can set VKU_ENABLE_TEST CMake variable as ON in configuration time to enable the test build.

License

This project is licensed under the MIT License - see the LICENSE file for details.