GPU Programming

NKI Teaching Lab: An Accelerator-Free Toolchain for Learning Tile-Level AI Accelerator Programming
An accelerator-free teaching toolchain for learning tile-level AI accelerator programming.
Conference
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
publicationAACLGPU Programming

TileBench Accepted at AACL 2026

TileBench, our benchmark for performance evaluation and bottleneck diagnosis of tile-based programming models, has been accepted at AACL-IJCNLP 2026.

Read update