Changes between Version 101 and Version 102 of DataParallel/BenchmarkStatus


Ignore:
Timestamp:
Dec 13, 2010 1:34:31 AM (3 years ago)
Author:
benl
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • DataParallel/BenchmarkStatus

    v101 v102  
    1515 * !QuickHull: vectorised.par.N1 version is 6x slower than the immutable Data.Vector version in absolute terms. 
    1616 * !QuickSort: vectorised.seq version doesn't compile due to a blow-up in !SpecConstr. 
    17  * SMVM: runs 1000x slower than the C program with small matrix sizes. For 1000x1000 with 10% fill ratio it takes about 1s and allocates 400M memory, while the C program is instantaneous. For larger sizes it dies with OOM. 
    18  * !BarnesHut: builds but runs very slowly. The immediate problem is that some dictionaries are recursive, their methods don't inlined, so fusion doesn't work. 
     17 * SMVM: Fusion doesn't wokr. Vectorised program 1000x slower than the C version with small matrix sizes. For 1000x1000 with 10% fill ratio it takes about 1s and allocates 400M memory, while the C program is instantaneous. For larger sizes it dies with OOM. 
     18 * !BarnesHut: Builds but runs very slowly. The immediate problem is that some dictionaries are recursive, their methods don't inlined, so fusion doesn't work. Version with -fllvm takes 30 min to compile, suspect complexity problems in LLVM mangler. 
    1919 
    2020= !ToDo =