A vulkan rendering engine written in C++ 23.

• A fiber-based cooperative task system
• Render graph based vulkan abstraction
• Retained mode GUI
• Two-pass occlusion culling with hierarchical depth buffer
• Compressed pack file format
• Very WIP scripting and shading languages

• bc7enc_rdo (for vpak tool; vendored)
• freetype
• harfbuzz
• libpng (for vpak tool)
• liburing
• meshoptimizer (for vpak tool)
• shaderc (for offline shader compilation step)
• xcb
• xkbcommon
• xxHash
• zstd

• Vulkan 1.3
• VK_EXT_descriptor_buffer
• VK_EXT_shader_atomic_float
• VK_EXT_shader_atomic_float2

Vull uses cmake as its build system. Either make or ninja will additionally be needed. Windows is not currently supported. Alternatively, vull can be built in a reproducible docker environment (see below).

emerge -n \
 app-arch/zstd \
 dev-libs/xxhash \
 dev-util/cmake \
 dev-util/ninja \
 media-libs/freetype \
 media-libs/harfbuzz \
 media-libs/libpng \
 media-libs/shaderc \
 media-libs/vulkan-layers \
 media-libs/vulkan-loader \
 sys-libs/liburing \
 x11-libs/libxcb \
 x11-libs/libxkbcommon

Note that meshoptimizer is always linked to via FetchContent, and that bc7enc_rdo is vendored. The vulkan validation layers (vulkan-layers) are not required but are extremely useful for development and testing.

To configure vull, use one of the available presets:

cmake . --preset release-gcc -GNinja

¹: Requires the graphics component to be enabled

²: Requires the vpak tool to be enabled, as well as the graphics, physics, ui components, and at least one window component.

cmake --build build

If VULL_BUILD_TESTS is enabled, tests can be run with

cmake --build build --target test

If doxygen is available, documentation can be built with

cmake --build build --target vull-docs

docker run --rm -u $(id -u) -v $(pwd):/src ghcr.io/yeetari/vull:master cmake \
 --preset release-gcc \
 -B/src/build \
 -GNinja \
 /src
docker run --rm -u $(id -u) -v $(pwd):/src ghcr.io/yeetari/vull:master cmake --build /src/build

Vull has support for fuzzing with AFLplusplus. Fuzz harnesses are located in tools/fuzz. To start fuzzing, use the fuzz cmake preset.

Input corpuses can be reduced with tools/reduce_corpus.bash, as well as any found crashes. A full workflow (using the script lexer fuzzer as an example) might look like:

cmake --preset fuzz
cd build-fuzz
cmake --build .
mkdir test-cases
# Add some test cases
../tools/reduce_corpus.bash test-cases fuzz-corpus ./tools/fuzz/lexer-fuzz
afl-fuzz -i fuzz-corpus -o fuzz-output -- ./tools/fuzz/lexer-fuzz

If any crashes are found, the crash directory can be reduced and inspected:

../tools/reduce_corpus.bash fuzz-output/default/crashes reduced-crashes ./tools/fuzz/lexer-fuzz