En Jämförande Studie av Parallella Söktekniker i en Schackmotor : Young Brothers Wait Concept eller Lazy SMP?

An important aspect to consider when creating a chess engine is the choice of parallel search algorithm. Two of the most common approaches are Young Brothers Wait Concept (YBWC) and Lazy Symmetric Multiprocessing (SMP). During the research process, it became apparent that while quite a few papers have been written about Chess Engines (CEs) with one of the two algorithms, very little exists regarding comparing parallel search techniques of CEs head- to-head. This project aims to determine how YBWC and Lazy SMP compare in terms of scalability, speedup, and playing strength. To accomplish this, the popular open source chess engine Crafty was modified to use Lazy SMP instead of YBWC. Both versions were then tasked with searching for the best move from 30 different positions 50 times each to determine the speedup from increasing thread-counts. They also played 30 matches against four engine opponents of different strengths to calculate the Elo rating of both versions of Crafty at different thread-counts. The findings of this project indicate that neither algorithm consistently outperforms the other. While each algorithm excels in specific scenarios, their strengths and weaknesses depend on resource availability. These results suggest that YBWC performs better than Lazy SMP when utilising a large amount of computer resources. Conversely, Lazy SMP outperforms YBWC when operating with limited resources. YBWC’s structured approach to organising resource utilisation gives it an advantage in resource-rich environments, but it becomes a drawback when resources are scarce. On the other hand, Lazy SMP’s more chaotic, direct approach works well in low- resource situations, but becomes harder to manage as resource availability increases. The lack of previous research could make the findings valuable for other chess engine developers when selecting chess algorithms, particularly when deciding between Lazy SMP and YBWC.