Changes between Version 40 and Version 41 of DataParallel/BenchmarkStatus


Ignore:
Timestamp:
Mar 9, 2009 1:51:00 PM (6 years ago)
Author:
chak
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • DataParallel/BenchmarkStatus

    v40 v41  
    104104
    105105On this machine, "SMVM primitives" & "SMVM, vectorised" also have a quirk from 2 to 4 threads.  This re-enforces the suspicion that this is a scheduling problem.
     106
     107=== Summary ===
     108
     109The speedup relative to a sequential C program for !SumSq, DotP, and SMVM on both architectures is illustrated by [http://www.cse.unsw.edu.au/~chak/project/dph/benchmarks/speedup-mar09.png this graph.]  In all cases, the data parallel Haskell program outperforms the sequential C program by a large margin on 8 cores.