Changes between Version 89 and Version 90 of DataParallel/BenchmarkStatus


Ignore:
Timestamp:
Dec 7, 2010 7:21:56 AM (3 years ago)
Author:
benl
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • DataParallel/BenchmarkStatus

    v89 v90  
    1111[[br]] 
    1212= Summary = 
     13 * Running debug threaded programs with heap profiling triggers an assertion in the RTS. Running them without debugging is a segfault. 
    1314 * !QuickHull: vectorised.par.N1 version is 6x slower than the immutable Data.Vector version in absolute terms. 
    1415 * !QuickSort: vectorised.seq version doesn't compile due to a blow-up in !SpecConstr. 
    15  * SMVM: appears to have rotted since the change to Data.Vector. Doesn't appear to read the input files properly. 
     16 * 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. 
    1617 * !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. 
    1718 
     
    135136 Multiplies a dense vector with a sparse matrix represented in the ''compressed sparse row format (CSR).'' [[br]] 
    136137 
    137  '''Status:''' Runs on 1000x1000 matrices with 1% fill ratio, but about 1000x slower than the C program. Dies with OOM for 2000x2000. Segfaults with 10000x10000. 
     138 '''Status:''' Runs on 1000x1000 matrices with 10% fill ratio, but about 1000x slower than the C program. Dies with OOM for 2000x2000. Segfaults with 10000x10000. 
    138139 
    139140