Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures最新文献_第2页

How to Manage High-Bandwidth Memory Automatically 如何自动管理高带宽内存

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400233

Rathish Das, Kunal Agrawal, M. A. Bender, Jonathan W. Berry, Benjamin Moseley, C. Phillips

{"title":"How to Manage High-Bandwidth Memory Automatically","authors":"Rathish Das, Kunal Agrawal, M. A. Bender, Jonathan W. Berry, Benjamin Moseley, C. Phillips","doi":"10.1145/3350755.3400233","DOIUrl":"https://doi.org/10.1145/3350755.3400233","url":null,"abstract":"This paper develops an algorithmic foundation for automated management of the multilevel-memory systems common to new supercomputers. In particular, the High-Bandwidth Memory (HBM) of these systems has a similar latency to that of DRAM and a smaller capacity, but it has much larger bandwidth. Systems equipped with HBM do not fit in classic memory-hierarchy models due to HBM's atypical characteristics. Unlike caches, which are generally managed automatically by the hardware, programmers of some current HBM-equipped supercomputers can choose to explicitly manage HBM themselves. This process is problem specific and resource intensive. Vendors offer this option because there is no consensus on how to automatically manage HBM to guarantee good performance, or whether this is even possible. In this paper, we give theoretical support for automatic HBM management by developing simple algorithms that can automatically control HBM and deliver good performance on multicore systems. HBM management is starkly different from traditional caching both in terms of optimization objectives and algorithm development. Since DRAM and HBM have similar latencies, minimizing HBM misses (provably) turns out not to be the right memory-management objective. Instead, we directly focus on minimizing makespan. In addition, while cache-management algorithms must focus on what pages to keep in cache; HBM management requires answering two questions: (1) which pages to keep in HBM and (2) how to use the limited bandwidth from HBM to DRAM. It turns out that the natural approach of using LRU for the first question and FCFS (First-Come-First-Serve) for the second question is provably bad. Instead, we provide a priority based approach that is simple, efficiently implementable and $O(1)$-competitive for makespan when all multicore threads are independent.","PeriodicalId":218208,"journal":{"name":"Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132724664","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 9

Multicore Paging Algorithms Cannot Be Competitive 多核分页算法不能竞争

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400270

Shahin Kamali, Helen Xu

引用次数: 3

Communication Lower Bounds of Convolutions in CNNs cnn中卷积的通信下界

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400267

Xiaoyang Zhang, Junmin Xiao, Guangming Tan

引用次数: 1

Contention Resolution with Message Deadlines 具有消息截止日期的争用解决

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400239

Kunal Agrawal, M. A. Bender, Jeremy T. Fineman, Seth Gilbert, Maxwell Young

{"title":"Contention Resolution with Message Deadlines","authors":"Kunal Agrawal, M. A. Bender, Jeremy T. Fineman, Seth Gilbert, Maxwell Young","doi":"10.1145/3350755.3400239","DOIUrl":"https://doi.org/10.1145/3350755.3400239","url":null,"abstract":"In the contention-resolution problem, multiple players contend for access to a shared resource. Contention resolution is used in wireless networks, where messages must be transmitted on a shared communication channel. When two or more messages are transmitted at the same time, a collision occurs, and none of the transmissions succeed. Much of the theoretical work on contention resolution has focused on efficiently resolving collisions in order to obtain throughput guarantees. However, in modern-day networks, not all traffic is treated equally. Instead, messages are often handled according to a notion of priority. While throughput remains an important metric, it fails to capture this increasingly-common scenario of traffic prioritization. Motivated by this concern, we design a contention-resolution algorithm where messages have delivery deadlines. Unit-length messages dynamically arrive over time, each with a corresponding delivery deadline that demarcates a window of time wherein the message must be transmitted successfully. We consider inputs that have a feasible schedule, even if message sizes increase by a constant factor. In this setting, we provide an algorithm which guarantees that each message succeeds by its deadline with high probability in its window size.","PeriodicalId":218208,"journal":{"name":"Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures","volume":"8 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123657866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 7

The Append Memory Model: Why BlockDAGs Excel Blockchains 附加内存模型:blockdag优于区块链的原因

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400272

D. Melnyk, Roger Wattenhofer

{"title":"The Append Memory Model: Why BlockDAGs Excel Blockchains","authors":"D. Melnyk, Roger Wattenhofer","doi":"10.1145/3350755.3400272","DOIUrl":"https://doi.org/10.1145/3350755.3400272","url":null,"abstract":"This paper presents a novel shared memory model that simplifies the analysis of consensus on a Chain and a DAG. In this new model, referred to as the append memory model, nodes are allowed to write new values to the unordered memory, but not to overwrite already existing values. We show that although this model differs from the standard shared memory model with n shared read-write registers, many known results from the shared memory model still hold in the append memory model: It is, for example, impossible to establish consensus on n nodes with one crash failure if the nodes in the system are asynchronous. We also consider the append memory model in a synchronous setting with Byzantine failures. For this case, we show that Byzantine agreement cannot be solved in less than t+1 rounds, where t is the number of Byzantine nodes in the system. Assuming a probabilistic access restriction to the append memory, we compare the Byzantine agreement protocols on the Chain and the DAG. We show that the DAG structure achieves an almost optimal resilience (close to t<n/2) in contrast to the Chain structure that can tolerate less than t<n/(1+l*(n-t)) Byzantine nodes, where l is the rate at which the nodes access the memory.","PeriodicalId":218208,"journal":{"name":"Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124069431","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 3

A Closer Look at Quantum Distributed Consensus 量子分布式共识研究

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400215

W. Golab, Hao Hao Tan

引用次数: 1

Towards Lockfree Persistent Homology 迈向无锁持久同源

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400244

D. Morozov, Arnur Nigmetov

引用次数: 11

Unconditional Lower Bounds for Adaptive Massively Parallel Computation 自适应大规模并行计算的无条件下界

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400230

M. Charikar, Weiyun Ma, Li-Yang Tan

{"title":"Unconditional Lower Bounds for Adaptive Massively Parallel Computation","authors":"M. Charikar, Weiyun Ma, Li-Yang Tan","doi":"10.1145/3350755.3400230","DOIUrl":"https://doi.org/10.1145/3350755.3400230","url":null,"abstract":"We consider unconditional lower bounds in the Adaptive Massively Parallel Computation (AMPC) model introduced by Behnezhad et al. (SPAA 19), which is an adaptive variant of the Massively Parallel Computation (MPC) model. Our first contribution is an optimal lower bound on the round complexity of distinguishing whether an input graph is a cycle of length n or two cycles of length n/2. This problem, 1v2-CIRCLE, has emerged as a central problem in the study of modern massively parallel computation. We prove that any AMPC algorithm for the 1v2-CIRCLE problem with I/O capacity O(nε) per machine requires Ω(1/ε) rounds, matching the upper bound of Behnezhad et al. More generally, we develop an abstract framework for proving lower bounds in the AMPC model using techniques from Boolean function complexity. Our framework extends the polynomial method of Roughgarden et al. (JACM 18) for the MPC model in two dimensions: [o] Adaptivity, where machines can write to and adaptively read from shared memory throughout the execution of the computation. [o] Promise problems, where the algorithm only has to succeed on certain inputs. These inputs may have special structure that is of particular interest, or they may be representative of hard instances of the overall problem.","PeriodicalId":218208,"journal":{"name":"Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures","volume":"6 10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127399802","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 8

On the Hardness of Massively Parallel Computation 关于大规模并行计算的硬度

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400223

Kai-Min Chung, Kuan-Yi Ho, Xiaorui Sun

{"title":"On the Hardness of Massively Parallel Computation","authors":"Kai-Min Chung, Kuan-Yi Ho, Xiaorui Sun","doi":"10.1145/3350755.3400223","DOIUrl":"https://doi.org/10.1145/3350755.3400223","url":null,"abstract":"We investigate whether there are inherent limits of parallelization in the (randomized) massively parallel computation (MPC) model by comparing it with the (sequential) RAM model. As our main result, we show the existence of hard functions that are essentially not parallelizable in the MPC model. Based on the widely-used random oracle methodology in cryptography with a cryptographic hash function h:{0,1}n → {0,1}n computable in time th, we show that there exists a function that can be computed in time O(T · th) and space S by a RAM algorithm, but any MPC algorithm with local memory size s < S/c for some c>1 requires at least ~Ω(T) rounds to compute the function, even in the average case, for a wide range of parameters n ≤ S ≤ T ≤ 2n1/4. Our result is almost optimal in the sense that by taking T to be much larger than th,e.g., T to be sub-exponential in th, to compute the function, the round complexity of any MPC algorithm with small local memory size is asymptotically the same (up to a polylogarithmic factor) as the time complexity of the RAM algorithm. Our result is obtained by adapting the so-called compression argument from the data structure lower bounds and cryptography literature to the context of massively parallel computation.","PeriodicalId":218208,"journal":{"name":"Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures","volume":"408 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126685533","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 1

Giving Future(s) to Transactional Memory 为事务性内存赋予未来

Proceedings of the 32nd ACM Symposium on Parallelism in Algorithms and Architectures Pub Date : 2020-07-06 DOI: 10.1145/3350755.3400220

Jingna Zeng, Seif Haridi, S. Issa, P. Romano, L. Rodrigues

引用次数: 0