Citation Context

...ses1) nodes on the same level can potentially run in parallel;s2) nodes on different levels can also run in parallel.sOn a multithreaded architecture, latency hiding can be utilizedsin three ways [23]=-=[24]-=-[25]. First, two parallel memory accessesscan be issued consecutively, and thus the latency of the firstsmemory access can be partially hidden by that of the secondsmemory access. Second, the latency ...

Citation Context

...parallel by transformingsa global table access with an extra pointer based dereferences.s7) In general, automatically parallelizing a sequential applicationsis a hard problem with limited success [23]=-=[24]-=-. We believe thatsdomain knowledge could help lead to a viable solution insautomatically parallelizing a sequential application. Networksapplications are amenable to pipelined processing due to theirs...

Citation Context

...forts are therefore required to design parallel packetsclassification algorithms for the multi-cores, which normallysprovides either hardware-assisted multithreading [3][9] to hidesmemory latency [20]=-=[21]-=-[22] or cache memories to reducesmemory latency.sPermission to make digital or hard copies of all or part of this work for personal orsclassroom use is granted without fee provided that copies are not...

Citation Context

...thread on that processor. At most x tasks can be executed at the same time on a processor. Multi-threading is important for hiding communication latency and overlapping computation with communication =-=[24]-=-. 6.1. Finding a starting task In order to initiate computation at each processor, we need to know the set of starting tasks (i.e. the set of tasks that have no father). This is done by symbolically s...