High-Performance Code Generation Using LLMs: Tensor Reduction

A research project examining how well large language models can generate high-performance CUDA kernels for tensor reduction workloads on NVIDIA A100 GPUs.

Studies LLM-generated CUDA kernels for tensor reduction tasks.
Focuses on high-performance code generation behavior on NVIDIA A100 GPUs.

The retained artifact is the project paper. Public code visibility is subject to course syllabus guidelines.