@article{10.1145/3046678,
title = {LD: Low-Overhead GPU Race Detection Without Access Monitoring},
author = {Pengcheng Li and Xiaoyu Hu and Dong Chen and Jacob Brock and Hao Luo and Eddy Z Zhang and Chen Ding},
url = {https://doi.org/10.1145/3046678},
doi = {10.1145/3046678},
issn = {1544-3566},
year = {2017},
date = {2017-01-01},
journal = {ACM Transaction on Architecture and Code Optimization (TACO 2017)},
volume = {14},
number = {1},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
abstract = {Data race detection has become an important problem in GPU programming. Previous designs of CPU race-checking tools are mainly task parallel and incur high overhead on GPUs due to access instrumentation, especially when monitoring many thousands of threads routinely used by GPU programs.This article presents a novel data-parallel solution designed and optimized for the GPU architecture. It includes compiler support and a set of runtime techniques. It uses value-based checking, which detects the races reported in previous work, finds new races, and supports race-free deterministic GPU execution. More important, race checking is massively data parallel and does not introduce divergent branching or atomic synchronization. Its slowdown is less than 5 texttimes for over half of the tests and 10 texttimes on average, which is orders of magnitude more efficient than the cuda-memcheck tool by Nvidia and the methods that use fine-grained access instrumentation.},
keywords = {GPU race detection, Instrumentation-free, Value-based checking},
pubstate = {published},
tppubtype = {article}
}