Graphics Processing Units (GPUs) accelerate many modern computationally intensive workloads, and achieving high performance on them depends heavily on sophisticated GPU compiler optimizations. However, even small changes in GPU compiler optimization heuristics or code-generation decisions can silently introduce severe...
Wai-Kin Wong, Dong-Wei Xiao, Ping Ke et al.· Proceedings of the 32nd ACM...· 0 citations
The StateLens framework is presented, a framework that employs Large Language Models (LLM) to automate the discovery of deep internal states and significantly outperforms state-of-the-art fuzzers and uncovers 68 new bugs.
Wai-Kin Wong, Dong-Wei Xiao, Anthony Cheuk Tung Lai et al.· Proceedings of the ACM SIGOP...· 0 citations
FIONA is presented, an offline optimizer that selectively ternarizes weights within a given packing layout based on the estimated effect of ternary conversion on the model's performance, and fits lower-degree replacements under a cumulative accuracy budget, reducing multiplicative depth and bootstrapping.
Yi-Teng Peng, Zhibo Liu, Dong-Wei Xiao et al.· 0 citations
A neuro-symbolic program synthesis approach for batch-editing collision meshes is presented, which synthesizes a reusable program that captures the editing intent and applies it to non-demonstration meshes.