2019 Scientific Conference on Network, Power Systems and Computing
Multi-thread Weak Deterministic Method Based on Cache Optimization
Kaiyu Wang, Zhenzhou Ji
Aiming at the problem of multi-thread uncertainty and the large system overhead of the existing deterministic implementation, this paper proposed to enter from the cache level. Research on weak deterministic methods for row optimization. First, for the problem that the LRU replacement algorithm does not work as well for multithreading as LIRS. The replacement strategy is optimized, followed by the weak deterministic method of ignoring data competition, and the deterministic order is obtained for the thread synchronization point. The competition that occurs with the card's synchronous operation is improved using a deterministic approach. We use of isolated thread memory in data competition, to change the thread into a lightweight process, and use experiments to prove the feasibility of the method.
Component, Cache replacement, Multithreading, LIRS, Weak Determinism
Cite this paper:
Kaiyu Wang, Zhenzhou Ji, Multi-thread Weak Deterministic Method Based on Cache Optimization. 2019 Scientific Conference on Network, Power Systems and Computing (NPSC 2019), 2019: 63-66. DOI: https://doi.org/10.33969/EECS.V3.015.