{"title":"Implementing parallel shortest-paths algorithms","authors":"M. Papaefthymiou, Joseph Rodrigue","doi":"10.1090/dimacs/030/04","DOIUrl":"https://doi.org/10.1090/dimacs/030/04","url":null,"abstract":"We have implemented two parallel versions of the Bellman-Ford-Moore algorithm for the single-source shortest-paths problem. Our software is written in C and runs on the CM-5 parallel supercomputer using CMMD communication primitives. We have compared the performance of our parallel programs with a serial implementation of the Bellman-Ford-Moore algorithm developed by Cherkassky, Goldberg and Radzik. In our experiments we used 50 randomly generated graphs with up to 2 15 vertices and 2 21 edges. Our fastest parallel program always achieved speedups greater than 1 when the average degree of the graphs exceeded 2 5 for the smaller and 2 3 for the larger inputs in our test suite. On dense graphs, we achieved speedups of up to 7.8 on a CM-5 with 32 processors.","PeriodicalId":123907,"journal":{"name":"Parallel Algorithms","volume":"193 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129404177","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}
{"title":"Connected components algorithms for mesh-connected parallel computers","authors":"S. Goddard, Subodh Kumar, J. Prins","doi":"10.1090/dimacs/030/03","DOIUrl":"https://doi.org/10.1090/dimacs/030/03","url":null,"abstract":"We present a new CREW PRAM algorithm for nding connected components. For a graphG with n vertices andm edges, algorithmA0 requires at mostO(logn) parallel steps and performsO((n+m) logn) work in the worst case. The advantage our algorithm has over others in the literature is that it can be adapted to a 2-D mesh-connected communication model in which all CREW operations are replaced byO(logn) parallel row and column operations without increasing the time complexity. We present the mapping of A0 to a mesh-connected computer and describe two implementations, A1 and A2. Algorithm A1, which uses an adjacency matrix to represent the graph, performs O(n logn) work. Hence, it only achieveswork e ciency on dense graphs. The second implementation,A2, uses a sparse representation of the adjacency matrix and again performs O(logn) row and column operations but reduces the work to O((m + n) logn) on all graphs. We report MasPar MP-1 performance gures for implementations of the algorithms described. The implementations are exercised on a variety of parametrically generated graphs, di ering in structure and connectivity. These graphs are generated externally and read in as input for the algorithms, permitting comparison of di erent implementations on identical graphs.","PeriodicalId":123907,"journal":{"name":"Parallel Algorithms","volume":"48 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116650039","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}
V. Cung, S. Dowaji, B. L. Cun, T. Mautor, C. Roucairol
{"title":"Concurrent data structures and load balancing strategies for parallel branch-and-bound/A* algorithms","authors":"V. Cung, S. Dowaji, B. L. Cun, T. Mautor, C. Roucairol","doi":"10.1090/dimacs/030/09","DOIUrl":"https://doi.org/10.1090/dimacs/030/09","url":null,"abstract":"","PeriodicalId":123907,"journal":{"name":"Parallel Algorithms","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131586370","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}
{"title":"A practical comparison of N-body algorithms","authors":"G. Blelloch, G. Narlikar","doi":"10.1090/dimacs/030/06","DOIUrl":"https://doi.org/10.1090/dimacs/030/06","url":null,"abstract":"This work compares three algorithms for the three dimensional N-body problem, the Barnes-Hut algorithm, Greengard's Fast Multipole Method(FMM), and the Parallel Multi-pole Tree Algorithm (PMTA) to determine which of the algorithms performs best in practice. Although FMM has a better asymptotic running time (O(N) instead of O(N log N) for uniform distributions), the algorithm is more complicated and it is not immediately clear above what values of N it performs better in practice. We studied the dependence of accuracy on the variable parameters , p and , and then compared the oating point operation counts of the three algorithms at similar levels of accuracy, for both charged and uncharged random distributions. At a high level of accuracy (RMS-error 10 ?5), the FMM did the least number of operations for N > 10 4 , assuming both charged and uncharged distributions of points. At a lower level of accuracy, (RMS-error 10 ?3) for uncharged distributions, the FMM did not outperform Barnes-Hut even for N > 10 8. For charged distributions of particles, both the FMM and PMTA were comparable at low accuracy. The algorithms were implemented in the parallel language Nesl.","PeriodicalId":123907,"journal":{"name":"Parallel Algorithms","volume":"159 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116265203","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}
A. Krishnamurthy, S. Lumetta, D. Culler, K. Yelick
{"title":"Connected components on distributed memory machines","authors":"A. Krishnamurthy, S. Lumetta, D. Culler, K. Yelick","doi":"10.1090/dimacs/030/01","DOIUrl":"https://doi.org/10.1090/dimacs/030/01","url":null,"abstract":"In this paper, we describe an implementation of the connected components algorithm on a distributed memory machine. A direct implementation of the PRAM algorithm results in an ineecient implementation due to the huge number of remote accesses generated by the algorithm. Instead, we use a hybrid algorithm that invokes the sequential algorithm as a local preprocessing phase before entering a global phase, which is a modiied version of the PRAM algorithm. We use the Split-C language, which provides the abstraction of a global address space, for building the distributed graph data structure. We obtain speedups in the order of 20 on a 32 processor CM5 for certain kinds of graphs.","PeriodicalId":123907,"journal":{"name":"Parallel Algorithms","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130966268","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}
{"title":"Parallel implementation of algorithms for finding connected components in graphs","authors":"T. Hsu, V. Ramachandran, N. Dean","doi":"10.1090/dimacs/030/02","DOIUrl":"https://doi.org/10.1090/dimacs/030/02","url":null,"abstract":"In this paper, we describe our implementation of several parallel graph algorithms for nding connected components. Our implementation, with virtual processing, is on a 16,384-processor MasPar MP-1 using the language MPL. We present extensive test data on our code. In our previous projects [21, 22, 23], we reported the implementation of an extensible parallel graph algorithms library. We developed general implementation and ne-tuning techniques without expending too much e ort on optimizing each individual routine. We also handled the issue of implementing virtual processing. In this paper, we describe several algorithms and ne-tuning techniques that we developed for the problem of nding connected components in parallel; many of the ne-tuning techniques are of general interest, and should be applicable to code for other problems. We present data on the execution time and memory usage of our various implementations. 1991 Mathematics Subject Classi cation. Primary 68-04; Secondary 05-04, 05C85, 68Q22.","PeriodicalId":123907,"journal":{"name":"Parallel Algorithms","volume":"1943 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129237515","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}