{"responseHeader":{"status":0,"QTime":6,"params":{"q":"{!q.op=AND}id:\"99919\"","hl":"true","hl.simple.post":"","hl.fragsize":"5000","fq":"!embargo_tdt:[NOW TO *]","hl.fl":"ocr_t","hl.method":"unified","wt":"json","hl.simple.pre":""}},"response":{"numFound":1,"start":0,"docs":[{"file_name_t":"Gu-Parallel_Algorithms.pdf","thumb_s":"/c9/9b/c99be502e90940279093463b72a7dfa31d18a022.jpg","oldid_t":"compsci 8032","setname_s":"ir_computersa","restricted_i":0,"format_t":"application/pdf","modified_tdt":"2016-05-25T00:00:00Z","file_s":"/df/43/df43e4840346baac5e07780994e3f11f8983c6db.pdf","title_t":"Page 271","ocr_t":"253 ent performance requirements. Sections 6.3.2 and 6.4.5 give some I/0 discussion. Several better designs based on input buffer memory, programmable I/ 0 networks and special-purpose constraint generators are given in [68). 7.4 Simulation and Performance Analysis The CLP1 architectural model was simulated using the CLP simulator (Section 7.1.3.1). Various problem instances in terms of different network structures, relational constraints and varying problem sizes were analyzed. Statistics accumulated and analyzed in this chapter were based on Assumption 6.6 that each machine instruction contains only one micro instruction. 7.4.1 Efficiency and Performance Improvement Real algorithm runs and simulation results provide many insights and interesting results. Among these, speedup and performance improvements are of the highest interest. Table 7.17 presents performance improvement in terms of the speedup for the sequential CLP algorithm (without m search branches). It was observed that a steady speedup for different problem instances can be achieved. The speedup figures appear almost the same in the case that none of the solutions (or empty solutions) were found. Speedup figures listed in Table 7.17 actually represent the worst case minimum speedup that one can obtain on a parallel CLP1 machine. Maximum speedup for a problem instance is achieved when the workload of the problem is uniformly distributed within all m search branches. Table 7.18 lists the maximum speedup figures for such problem instances (with even load distribution). For a problem instance whose workload distribution is uneven within m search branches, its maximum execution time to find all solutions is determined by the longest search time","id":99919,"created_tdt":"2016-05-25T00:00:00Z","parent_i":99969,"_version_":1679953745592451073}]},"highlighting":{"99919":{"ocr_t":[]}}}