Changeset d0fcc82
- Timestamp:
- Jul 29, 2022, 10:49:52 AM (2 years ago)
- Branches:
- ADT, ast-experimental, master, pthread-emulation
- Children:
- b507dcd
- Parents:
- c4c8571
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
doc/theses/thierry_delisle_PhD/thesis/text/front.tex
rc4c8571 rd0fcc82 123 123 \begin{center}\textbf{Abstract}\end{center} 124 124 125 This is the abstract. 126 127 Vulputate minim vel consequat praesent at vel iusto et, ex delenit, esse euismod luptatum augue ut sit et eu vel augue autem feugiat, quis ad dolore. Nulla vel, laoreet lobortis te commodo elit qui aliquam enim ex iriure ea ullamcorper nostrud lorem, lorem laoreet eu ex ut vel in zzril wisi quis. Nisl in autem praesent dignissim, sit vel aliquam at te, vero dolor molestie consequat. 128 129 Tation iriure sed wisi feugait odio dolore illum duis in accumsan velit illum consequat consequat ipsum molestie duis duis ut ullamcorper. Duis exerci odio blandit vero dolore eros odio amet et nisl in nostrud consequat iusto eum suscipit autem vero. Iusto dolore exerci, ut erat ex, magna in facilisis duis amet feugait augue accumsan zzril delenit aliquip dignissim at. Nisl molestie nibh, vulputate feugait nibh luptatum ea delenit nostrud dolore minim veniam odio volutpat delenit nulla accumsan eum vero ullamcorper eum. Augue velit veniam, dolor, exerci ea feugiat nulla molestie, veniam nonummy nulla dolore tincidunt, consectetuer dolore nulla ipsum commodo. 130 131 At nostrud lorem, lorem laoreet eu ex ut vel in zzril wisi. Suscipit consequat in autem praesent dignissim, sit vel aliquam at te, vero dolor molestie consequat eros tation facilisi diam dolor. Odio luptatum dolor in facilisis et facilisi et adipiscing suscipit eu iusto praesent enim, euismod consectetuer feugait duis. Odio veniam et iriure ad qui nonummy aliquip at qui augue quis vel diam, nulla. Autem exerci tation iusto, hendrerit et, tation esse consequat ut velit te dignissim eu esse eros facilisis lobortis, lobortis hendrerit esse dignissim nisl. Nibh nulla minim vel consequat praesent at vel iusto et, ex delenit, esse euismod luptatum. 132 133 Ut eum vero ullamcorper eum ad velit veniam, dolor, exerci ea feugiat nulla molestie, veniam nonummy nulla. Elit tincidunt, consectetuer dolore nulla ipsum commodo, ut, at qui blandit suscipit accumsan feugiat vel praesent. In dolor, ea elit suscipit nisl blandit hendrerit zzril. Sit enim, et dolore blandit illum enim duis feugiat velit consequat iriure sed wisi feugait odio dolore illum duis. Et accumsan velit illum consequat consequat ipsum molestie duis duis ut ullamcorper nulla exerci odio blandit vero dolore eros odio amet et. 134 135 In augue quis vel diam, nulla dolore exerci tation iusto, hendrerit et, tation esse consequat ut velit. Duis dignissim eu esse eros facilisis lobortis, lobortis hendrerit esse dignissim nisl illum nulla minim vel consequat praesent at vel iusto et, ex delenit, esse euismod. Nulla augue ut sit et eu vel augue autem feugiat, quis ad dolore te vel, laoreet lobortis te commodo elit qui aliquam enim ex iriure. Ut ullamcorper nostrud lorem, lorem laoreet eu ex ut vel in zzril wisi quis consequat in autem praesent dignissim, sit vel. Dolore at te, vero dolor molestie consequat eros tation facilisi diam. Feugait augue luptatum dolor in facilisis et facilisi et adipiscing suscipit eu iusto praesent enim, euismod consectetuer feugait duis vulputate veniam et. 136 137 Ad eros odio amet et nisl in nostrud consequat iusto eum suscipit autem vero enim dolore exerci, ut. Esse ex, magna in facilisis duis amet feugait augue accumsan zzril. Lobortis aliquip dignissim at, in molestie nibh, vulputate feugait nibh luptatum ea delenit nostrud dolore minim veniam odio. Euismod delenit nulla accumsan eum vero ullamcorper eum ad velit veniam. Quis, exerci ea feugiat nulla molestie, veniam nonummy nulla. Elit tincidunt, consectetuer dolore nulla ipsum commodo, ut, at qui blandit suscipit accumsan feugiat vel praesent. 138 139 Dolor zzril wisi quis consequat in autem praesent dignissim, sit vel aliquam at te, vero. Duis molestie consequat eros tation facilisi diam dolor augue. Dolore dolor in facilisis et facilisi et adipiscing suscipit eu iusto praesent enim, euismod consectetuer feugait duis vulputate. 125 User-Level threading (M:N) is gaining popularity over kernel-level threading (1:1) in many programming languages. 126 The user-level approach is often a better mechanism to express complex concurrent applications by efficiently running 10,000+ threads on multi-core systems. 127 Indeed, over-partitioning into small work-units significantly eases load balancing while providing user threads for each unit of work offers greater freedom to the programmer. 128 To manage these high levels of concurrency, the underlying runtime must efficiently schedule many user threads across a few kernel threads; 129 which begs of the question of how many kernel threads are needed and when should the need be re-evaliated. 130 Furthermore, the scheduler must prevent kernel threads from blocking, otherwise user-thread parallelism drops, and put idle kernel-threads to sleep to avoid wasted resources. 131 Finally, the scheduling system must provide fairness to prevent a user thread from monopolizing a kernel thread; 132 otherwise other user threads can experience short/long term starvation or kernel threads can deadlock waiting for events to occur. 133 134 This thesis analyses multiple scheduler systems, where each system attempts to fulfill the necessary requirements for user-level threading. 135 The predominant technique for manage high levels of concurrency is sharding the ready-queue with one queue per kernel-threads and using some form of work stealing/sharing to dynamically rebalance workload shifts. 136 Fairness can be handled through preemption or ad-hoc solutions, which leads to coarse-grained fairness and pathological cases. 137 Preventing kernel blocking is accomplish by transforming kernel locks and I/O operations into user-level operations that do not block the kernel thread or spin up new kernel threads to manage the blocking. 138 139 After selecting specific approaches to these scheduling issues, a complete implementation was created and tested in the \CFA (C-for-all) runtime system. 140 \CFA is a modern extension of C using user-level threading as its fundamental threading model. 141 As one of its primary goals, \CFA aims to offer increased safety and productivity without sacrificing performance. 142 The new scheduler achieves this goal by demonstrating equivalent performance to work-stealing schedulers while offering better fairness. 143 This is achieved through several optimization that successfully eliminate the cost of the additional fairness, some of these optimization relying on interesting hardware optimizations present on most modern cpus. 144 This work also includes support for user-level \io, allowing programmers to have many more user-threads blocking on \io operations than there are \glspl{kthrd}. 145 The implementation is based on @io_uring@, a recent addition to the Linux kernel, and achieves the same performance and fairness. 146 To complete the picture, the idle sleep mechanism that goes along is presented. 147 148 140 149 141 150 \cleardoublepage
Note: See TracChangeset
for help on using the changeset viewer.