Compute

SkillMedia

Work on CUDA, OpenCL, GLSL compute, GL/CUDA or GL/OpenCL interop, GPU data ownership, kernels, argument packing, and compute-driven rendering.

Available today. Use it from your connected AI after setup.

No other account needed.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Compute skill

What this skill tells your AI

The instructions your AI receives, as published by node-3d/node-3d in .agents/skills/compute/SKILL.md and read by ahel’s review.

GPU compute architecture, resource ownership, synchronization, and compute examples.

Identify the backend and where simulation/render state lives.
Keep GPU data resident where practical; avoid per-frame CPU round trips.
Make GL/compute ownership transitions explicit and exception-safe.
Keep JS responsible for orchestration and kernels for data-parallel work.
Keep kernel argument layout explicit; validate handles, contexts, lifetimes, and dynamic pointer updates.
Give examples explicit runtime/device requirements and a bounded automation path where practical.
Choose texture ping-pong for browser-style GPGPU or shared GL VBOs for compute/render interop; keep dynamic state stride and typed representation explicit.

Load .agents/references/gpu-interop.md for GL/CUDA/OpenCL ownership, synchronization, resource handles, or argument packing.

Load .agents/references/boids-workflow.md when extending or diagnosing the canonical packages/core/examples/core/boids implementations.

Use $examples/$tests as secondary skills only when modifying their artifacts.

Advanced