| name | kernelevole-agentic |
| title | KernelEvolve: Scaling Agentic Kernel Coding for AI Accelerators |
| version | 0.0.2 |
| engine | skillxiv-v0.0.2-claude-opus-4.6 |
| license | MIT |
| url | https://arxiv.org/abs/2512.23236 |
| keywords | ["kernel-optimization","agents","automated-synthesis","hardware-aware"] |
| description | Automate compute kernel optimization via agentic AI and retrieval-augmented prompting. Unified context-aware transformation function, hardware-specific constraint KB, self-improving state machine exploring kernel variants—achieving 1.25-17× speedups on production recommendation workloads in hours vs weeks of manual effort. |
Overview
Autonomous kernel optimization across heterogeneous hardware via agentic synthesis.
Core Technique
Context-Aware Transformation:
kernel = universal_operator.transform(
operator_spec,
hardware_context=gpu_type,
optimization_history=prior_attempts
)
Retrieval-Augmented Prompting:
Hardware-specific constraints guide generation.
Performance
- 1.25-17× speedup on production workloads
- Hours vs weeks of manual optimization
References
- Context-aware kernel generation
- Hardware constraint knowledge bases
- Self-improving kernel optimization
