Changes between Version 14 and Version 15 of DataParallel/Dec2010Release


Ignore:
Timestamp:
Dec 14, 2010 1:24:59 AM (3 years ago)
Author:
chak
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • DataParallel/Dec2010Release

    v14 v15  
    2222Changes: 
    2323 * -fdph-par should be the default (and sensible error message if the dph package is not available) '''[MANUEL]''' 
    24  * -fvectorise should imply -Odph (if omitted, we get fatal errors) '''[MANUEL]''' 
    25    * rl notes, "We shouldn't! -fvectorise -O0 should work, I'll take a look. I should change -Odph to reflect what we currently need, though. Basically, it should be equivalent ot -O2 -fsimplifier-phases=3 -fsimplifier-iterations=20." 
     24 * -Odph should be equivalent to '-O2 -fsimplifier-phases=3 -fsimplifier-iterations=20' '''[MANUEL]''' 
    2625 * Move GHC.PArr into  the DPH libs.  (Needed for Haddock.) '''[MANUEL]''' 
    2726 * Find out if we still need the `NoSpecConstr` annotation and remove it if not '''[ROMAN]''' 
    2827 
    2928Bug fixes: 
    30  * Repa edge-detection is deadlocking with more than 2 threads [DONE] 
    31  * Fix the BH seg fault in DPH. Roman has found the problem [DONE] 
    32  * LLVM back end not working with DPH '''[BEN]''' 
     29 * Vectoriser needs to be adapted to Simon's recursive superclasses patch '''[ROMAN]''' 
     30 * The combination '-fvectorise -O0' should work '''[ROMAN]''' 
    3331 * Trying to vectorise the `DotP` example from the tutorial on the Haskell Wiki, `-fdph-seq` fails with (`-fdph-par` works fine) 
    3432{{{ 
     
    3634    Tycon not vectorised:  Data.Array.Parallel.Lifted.PArray.PArray 
    3735}}} 
     36 * LLVM back end not working with DPH (held up due to LLVM backend problems in the HEAD) '''[BEN]''' 
     37 * Repa edge-detection is deadlocking with more than 2 threads [DONE] 
     38 * Fix the BH seg fault in DPH. Roman has found the problem [DONE] 
    3839 
    3940 
     
    4344 * Repa works fast in parallel 
    4445  * MMult [OK, but about 20% slower than in 6.13; try with LLVM and w/o bounds checks] [BEN] 
    45   * Laplace '''[SLOW & DOESN'T SCALE]''' '''[BEN]''' 
     46  * Laplace (blocked on LLVM backend problems) '''[SLOW & DOESN'T SCALE]''' '''[BEN]''' 
    4647  * Blur [OK] 
    4748  * !EdgeDetect [OK] 
     
    5253  * Evens [OK (but more than 3 times slower than C)] 
    5354    * rl reckons this is due to GHC compiling modulo of powers of two inefficiently; c.f., #3065 (in `packByTags`) 
    54   * SMVM '''[SLOW (lack of fusion)]''' '''[BEN & ROMAN]''' 
     55  * SMVM (blocked on optimisation of lifted indexing) '''[BROKEN]''' '''[BEN & ROMAN]''' 
    5556 * Dynamically-nested DPH programs without user-defined datatypes should run correctly, but not necessarily fast 
    5657  * Quicksort '''[BROKEN (SpecConstr) & SLOW]''' '''[SIMON & BEN]'''