LLM Code Generation

HPC-Bench: A Comprehensive Benchmark for High Performance Computing Codes
A benchmark for evaluating LLM-generated optimizations of HPC kernels with program-level correctness checks and performance measurements.
OpenReview
TileBench: A Controlled Benchmark for Performance Evaluation and Bottleneck Diagnosis of Tile-Based Programming Models
A controlled benchmark for comparing tile-based GPU programming models, diagnosing performance bottlenecks, and evaluating LLM-generated kernels.
PDF arXiv