https://hgpu.org/?p=28609
Comparing Llama-2 and GPT-3 LLMs for HPC kernels generation