Meta describes more than 1,500 recommendation models running across NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators. A missing or slow kernel can force an operator back to CPUs, add network crossings, or block a model launch. KernelEvolve treats that combinatorial maintenance problem as a measured search rather than a one-shot code-generation prompt[1].

Generate, execute, select, remember

The system represents optimization as a graph of candidate states. Agents retrieve documentation and production code, generate Triton, CuTe DSL, or lower-level implementations, compile and run them, check correctness, measure fitness, and expand promising branches. A persistent knowledge base carries hardware rules that a general model could not have learned, including proprietary accelerator constraints.

KernelEvolve’s optimization loop. Retrieved hardware knowledge guides code generation; candidates are compiled, checked, benchmarked, and selected into the next search state. Evaluation covered all 480 combinations of 160 operators and three platforms, plus 250 KernelBench problems. Original figure created for this article.

KernelEvolve passed all 250 KernelBench problems and all 480 combinations of 160 ATen operators across three platforms. Reported production speedups range from 1.25× to 17× over PyTorch baselines, while development cycles fall from weeks to hours. The paper also reports shape-specific regressions, which is why candidate acceptance uses production distributions rather than one benchmark input.

What we take from it

Agentic coding becomes infrastructure only after execution closes the loop. Retrieval supplies private hardware knowledge, tests reject incorrect code, performance measurements replace stylistic preference, and persistent state survives long campaigns. The headline speedups are workload-specific; the more important result is an operational process that can maintain coverage across changing chips without trusting generated code before it runs.

Source and attribution

This digest uses original wording and a new deterministic figure. It reproduces no paper prose, code, figure, or table. Meta provides the full technical report publicly; the ISCA proceedings version is © IEEE 2026.