GPU-Kit is in concept phase. Contributions should preserve the architecture before expanding the scope.
- Do not expose foreign graphics stack concepts as required public GPU-Kit concepts.
- Keep OpenGL, Vulkan, SDL and Mesa as research, backend or compatibility material.
- Do not put BMessage in the rendering hot path.
- Prefer command buffers, shared memory and explicit fences.
- Keep capabilities explicit.
- Avoid hidden fallbacks.
- Document every major design decision as an ADR.
- Start small.
Ask:
Does this make the native Haiku model clearer?
Does it keep apps away from direct GPU access?
Does it preserve the control/data plane split?
Does it avoid accidental Mesa/OpenGL/Vulkan leakage?
Can it be tested with null or software backend?
During early phases, useful contributions include:
documentation reviews
terminology improvements
API sketch criticism
Haiku-specific integration notes
software backend experiments
trace/debug tooling
research notes
small demos
When studying another project, add a note under docs/research/ using this structure:
# Research note: <topic>
## Source
## What problem does it solve?
## Useful concepts
## Concepts to reject
## Haiku translation
## Open questionsUse direct language. Prefer small APIs. Prefer explicit constraints. Do not hide uncertainty.
GPU-Kit should read like Haiku: clear, native, responsive.