Speed-up Solving Linear Systems on Parallel Architectures via Aggregation of Clans (Dmitry Zaitsev, LVEE-2019) — различия между версиями
Материал из 0x1.tv
StasFomin (обсуждение | вклад) |
StasFomin (обсуждение | вклад) |
||
{{LinksSection}} * [ Talks page] <!-- <blockquote>[©]</blockquote> --> {{vklink|1435}} {{fblink|2423735191212821}} <references/> [[Категория:LVEE-2019]] [[Категория:Алгоритмы]] {{stats|disqus_comments=12|refresh_time=2020-01-22T20:21:5628T12:50:04.505485662802|vimeo_plays=24|youtube_comments=0|youtube_plays=12}}3}} |
Версия 09:50, 28 января 2020
- Докладчик
- Дмитрий Зайцев
Varying minimal clan size brings in certain imbalance when solving a linear (Diophantine) system on parallel architectures via composition of its clans using open source software PaAd. The problem is partially mended by dynamic scheduling of jobs. The present paper studies a task of preliminary balancing the clan size via their aggregation represented as a special case of graph partitioning. Due to complexity of the optimization criteria, taking into consideration both the number of equations and the number of variables on two stages of solution — solving a system for each clan and solving a composition system — simplified variants of the task are considered and solved using heuristic techniques: a fast bin packing with the first fit on a sorted array algorithm and a multi-objective graph partitioning with software package METIS. Obtained benchmarks show that aggregation of clans brings in an additional 3 times speed-up.
Содержание
Видео
Презентация
Thesis
Примечания и ссылки
- [ Talks page]
Plays:37 Comments:12