KernelFoundry is introduced, an evolutionary framework that efficiently explores the space of GPU kernels through a meta-prompt evolution that co-evolves prompts with kernels to uncover task-specific optimization strategies, and a template-based parameter optimization approach to tune kernels to inputs and hardware.
Abstract
GPU kernel optimization challenges LLMs beyond standard coding tasks, as it requires an understanding of hardware architecture, parallel computing optimization strategies, and profiling outputs. However, most existing approaches leveraging LLMs for kernel generation apply standard prompting and feedback loops, considering hardware only through profiling feedback. We introduce KernelFoundry, an evolutionary framework that efficiently explores the space of GPU kernels through (1) MAP-Elites quality diversity search with kernel-specific behavioral dimensions to sustain exploration; (2) meta-prompt evolution that co-evolves prompts with kernels to uncover task-specific optimization strategies, and (3) a template-based parameter optimization approach to tune kernels to inputs and hardware. We evaluate this framework on Kernel-Bench, robust-kbench and custom tasks, generating SYCL kernels as a cross-platform GPU programming paradigm, and CUDA kernels for comparison to prior work. Our approach consistently outperforms the baseline methods and achieves an average speedup of 2.3 on KernelBench for SYCL. Moreover, KernelFoundry is implemented as a distributed framework with remote access to diverse hardware, allowing quick benchmarking and featuring a flexible user input layer to support kernel generation for a wide range of real use cases beyond benchmarking.
The established Kernel Tuner auto-tuning framework is rebuilt with Julia support, enabling systematic exploration of kernel configurations for hardware-agnostic GPU kernels targeting NVIDIA, AMD, Intel, and Apple GPUs.
Floris-Jan Willemsen, Evelyne Ringoot, Alan Edelman· Workshop Proceedings of the...· 0 citations
A hierarchical search-space planning framework for GPU kernel optimization that delivers stronger overall implementation validity, sample efficiency, and optimization performance than existing training-free methods, while remaining competitive with the training-based CUDA-L1 without additional model training is propose...
Jing-Hao Wang, Qiqi Gu, Chenpeng Wu et al.· 0 citations
KernelOPT is a multi-agent system that treats compiled models as structured artifacts and exclusively targets generated Triton sub-kernels using five profiling-guided LLM agents and achieves geometric mean speedups over PyTorch.
High-performance GPU kernels are essential to modern machine learning systems, yet automatically generating kernels that are both correct and efficient remains challenging. Existing LLM-based approaches face two major limitations: the scarcity of high-quality training data aligned with the model's current capabilities,...
PTXBench provides an auditable testbed for measuring and improving LLMs'ability to exploit evolving GPU architectures, and shows that architecture-specific PTX capability remains uneven.
Genghan Zhang, Yi-Xin Dong, Chengze Fan et al.· 1 citation
MorphX is a GPU multitasking system that embraces the runtime adjustability of kernels as the first principle, and transparently transforms traditional kernels into morphable ones at compile time, and leverages an in-place profiling mechanism to orchestrate fine-grained scheduling based on real-time resource contention...
Ting-Xu Ren, Ru-Wen Fan, Hao Guo et al.· Proceedings of the ACM SIGOP...· 1 citation
Related blog posts
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 13, 2026