--- license: mit pretty_name: cakebench tags: - gpu-kernels - benchmark - triton - cuda --- # cakebench A kernel benchmark modelled on CAKE (arXiv 2608.12629). It scores a candidate kernel on 5 kernel families (21 shapes in the full GPU portfolio) by first checking correctness against an fp32 oracle on every shape and 3 input distributions (normal / heavy-tailed / tiny), then timing it (median GPU span, L2 flushed) against a torch baseline. **No model weights and no LLM-generated kernels are released here.** This repo holds the recorded scores of the five hand-written example kernels (plain torch) on one shared A100-SXM4-40GB, so you can compare your own harness runs. | task | correct shapes | geomean vs torch baseline | |---|---|---| | kmeans_assign | 3/3 | 1.69x | | knn_search | 3/3 | 1.13x | | tinygemm | 8/8 | 1.00x | | kda_decode | 4/4 | 1.03x | | kda_prefill | 3/3 | 1.06x | Shared box, noisy timings; the Cake compiler is not public, so this benches kernels, not the compiler; baselines are torch, not the paper's FlashML/FlashKDA/FlashInfer. - Code and usage: https://github.com/Bobliuuu/cakebench - Explainer video: https://github.com/Bobliuuu/cakebench/blob/main/video/cakebench_explainer.mp4 - Release: https://github.com/Bobliuuu/cakebench/releases/tag/v0.1.0