Permalink
Commits on Dec 15, 2011
  1. bump to 0.1.0.2

    simonmar committed Dec 15, 2011
  2. bump deepseq dependency

    simonmar committed Dec 15, 2011
Commits on Apr 7, 2011
  1. bump to 0.1.0.1

    simonmar committed Apr 7, 2011
  2. add categories

    simonmar committed Apr 7, 2011
  3. default GHC is 'ghc'

    simonmar committed Apr 7, 2011
Commits on Mar 24, 2011
  1. various cleanups

    simonmar committed Mar 24, 2011
  2. Added a simple logging/timing framework and used it in the disjoint_w…

    rrnewton committed Mar 24, 2011
    …orking_sets benchmark.
    
    Try somethnig like this:
      time ./examples/stream/disjoint_working_sets_pipeline.exe monad 4 1024 10 50 +RTS -N4
    And you should see each of the kernels executing consistently on the same thread.
  3. Got rid of the lazy list conversion in the disjoint_working_sets benc…

    rrnewton committed Mar 23, 2011
    …hmark. Uses streamFold instead.
Commits on Mar 22, 2011
  1. Added a serial scheduler, ParELision, that works only for a subclass …

    rrnewton committed Mar 22, 2011
    …of problems. Initial performance surprisingly bad!
Commits on Mar 21, 2011
  1. Made the benchmark.sh script a bit more flexible and reactived the be…

    rrnewton committed Mar 21, 2011
    …nchmarks in subdirs.
  2. Added a simple benchmarking framework since simons isnt available. On…

    rrnewton committed Mar 21, 2011
    …e goal is to help avoid regression. I still have a couple small changes to try out in the scheduler.
  3. Added a sparks/strategies based version of disjoint_working_sets_pipe…

    rrnewton committed Mar 21, 2011
    …line. The monad-par one seems better so far.
Commits on Mar 20, 2011
  1. Added yielding to Par.hs, which enables the stream processing example…

    rrnewton committed Mar 20, 2011
    …, filterbank, to run on one thread. There's still no way to put backpressure on a source node though, which means that the FFT example in this checkin will space leak over time.
Commits on Mar 19, 2011
  1. Intermediate checkin. Refactored filterbank stream ops to have a nice…

    rrnewton committed Mar 19, 2011
    …r interface. But in exchange an assumption is made about the forking structure. The streamops each contain one fork. My strategy for timing with measureRate doesn't work because the unsafePerformIO to measure time only happens once. About to move that into the IO monad.
Commits on Mar 18, 2011
  1. Added nofib benchmark "coins". The performance is comparable to Versi…

    cchen15 committed Mar 18, 2011
    …on 7
    
    on Harpertown windows.
  2. Exposed IList from Openlist as well. Began a simple streaming benchma…

    rrnewton committed Mar 18, 2011
    …rk. Currently unfinished.
Commits on Mar 17, 2011
  1. Added parfib benchmark.

    rrnewton committed Mar 17, 2011
Commits on Mar 16, 2011
  1. Experimenting with a parallel for implementation. Using it for an unf…

    rrnewton committed Mar 16, 2011
    …inished binomial lattice benchmark.