GraphChi: Large-Scale Graph Computation on Just a PC | USENIX2012 OSDI
Parallel Sliding Windows (PSW)
关键:以interval为单元的BSP(Bulk-Synchronous Parallel);设计存储layout,使得获取一个interval中所有顶点的邻居只需要 顺序 访问全图磁盘...
MSBFSThe More the Merrier: Efficient Multi-Source Graph TraversalVLDB 2014
基准
bit op
ANP(这个)
tuningcache line
heuristic maximum sharing way of grouping BFSs
reason:
defintions
unit network单位网络:每个顶点要么只有一条出边且容量为1,要么只有一条入边且容量为1
unit capacity network单位容量网络:所有边容量=1
reductions of problem versionundirected->directed
https://stackoverflow.com/questions/297416...