Loading ...
Sorry, an error occurred while loading the content.

Please Benchmark the Multi-threaded version of Freecell Solver

Expand Messages
  • Shlomi Fish
    Hi all! I recently created a thread-enabled version of Freecell Solver, that uses several solver instances in separate threads. I d like to get some indication
    Message 1 of 3 , May 15, 2009
    • 0 Attachment
      Hi all!

      I recently created a thread-enabled version of Freecell Solver, that uses
      several solver instances in separate threads. I'd like to get some indication
      of the performance of the solver with different number of threads on different
      machines. The instructions can be found below.

      On my Pentium 2.4 GHz with 2.5 GB of RAM and 1 GB of swap, I'm getting the
      following results: (in seconds - lower is better)

      {{{{{{{{{{{{
      DUMPS/dump001:139.560144901276
      DUMPS/dump002:115.644785881042
      DUMPS/dump003:114.645025968552
      DUMPS/dump004:118.50590801239
      DUMPS/dump005:117.550628900528
      DUMPS/dump006:119.174820899963
      }}}}}}}}}}}}

      So the optimum is 3 threads.

      Anyway here are the instructions:

      {{{{{{{{{{{
      Requirements: perl-5.8.x or above, CMake 2.6.3, gcc, and a working pthreads-
      devel library.

      1. Download freecell-solver-2.27.2 from:

      http://www.shlomifish.org/Files/files/code/fcs/

      2. Unpack and cd to the freecell-solver-* directory.

      3. Run "./configure -l tt".

      4. Type "make" to build everything.

      5. Run "bash scripts/time-threads-num.bash $MAX_NUM_THREADS" . The default
      for $MAX_NUM_THREADS is "6", where 7 completely hogs my system. If you have
      many CPUs and/or cores and are confident that raising it will not overload
      your
      system, you can raise it.

      This stage better be done when the computer is not loaded.

      6. Wait.

      7. Type "perl scripts/time-fcs.pl DUMPS/*" and copy-and-paste the results to
      me with specifications of your comptuer that are as detailed as possible.
      }}}}}}}}}}}}}

      Any help would be appreciated.

      Regards,

      Shlomi Fish

      --
      -----------------------------------------------------------------
      Shlomi Fish http://www.shlomifish.org/
      Original Riddles - http://www.shlomifish.org/puzzles/

      God gave us two eyes and ten fingers so we will type five times as much as we
      read.
    • Shlomi Fish
      ... Hi all! I should note that the link below now only contains an archive of version 2.27.3 which is more up-to-date and fixes some problems. I should also
      Message 2 of 3 , May 15, 2009
      • 0 Attachment
        On Friday 15 May 2009 19:08:25 Shlomi Fish wrote:
        > Hi all!
        >
        > I recently created a thread-enabled version of Freecell Solver, that uses
        > several solver instances in separate threads. I'd like to get some
        > indication of the performance of the solver with different number of
        > threads on different machines. The instructions can be found below.
        >

        Hi all!

        I should note that the link below now only contains an archive of version
        2.27.3 which is more up-to-date and fixes some problems. I should also note
        that if you're using most versions of Ubuntu, then the CMake that ships with
        them (2.6.2) won't be adequate for building the solver, and you'll need to
        compile a latest CMake from source.

        In any case, I already got one result from a volunteer that ran it on an Intel
        Core 2 @ 2.4 GHz with 3GB of RAM and 8.8GB of swap:

        {{{
        DUMPS/dump001:86.4667541980743
        DUMPS/dump002:47.5844478607178
        DUMPS/dump003:48.3457579612732
        DUMPS/dump004:51.4420628547668
        DUMPS/dump005:48.824597120285
        DUMPS/dump006:48.0212950706482
        }}}

        So it seems that there the optimal number of threads is 2, and that the code
        runs 2.4 times faster than on my machine. Thanks to that volunteer!

        Regards,

        Shlomi Fish

        > On my Pentium 2.4 GHz with 2.5 GB of RAM and 1 GB of swap, I'm getting the
        > following results: (in seconds - lower is better)
        >
        > {{{{{{{{{{{{
        > DUMPS/dump001:139.560144901276
        > DUMPS/dump002:115.644785881042
        > DUMPS/dump003:114.645025968552
        > DUMPS/dump004:118.50590801239
        > DUMPS/dump005:117.550628900528
        > DUMPS/dump006:119.174820899963
        > }}}}}}}}}}}}
        >
        > So the optimum is 3 threads.
        >
        > Anyway here are the instructions:
        >
        > {{{{{{{{{{{
        > Requirements: perl-5.8.x or above, CMake 2.6.3, gcc, and a working
        > pthreads- devel library.
        >
        > 1. Download freecell-solver-2.27.2 from:
        >
        > http://www.shlomifish.org/Files/files/code/fcs/
        >
        > 2. Unpack and cd to the freecell-solver-* directory.
        >
        > 3. Run "./configure -l tt".
        >
        > 4. Type "make" to build everything.
        >
        > 5. Run "bash scripts/time-threads-num.bash $MAX_NUM_THREADS" . The default
        > for $MAX_NUM_THREADS is "6", where 7 completely hogs my system. If you have
        > many CPUs and/or cores and are confident that raising it will not overload
        > your
        > system, you can raise it.
        >
        > This stage better be done when the computer is not loaded.
        >
        > 6. Wait.
        >
        > 7. Type "perl scripts/time-fcs.pl DUMPS/*" and copy-and-paste the results
        > to me with specifications of your comptuer that are as detailed as
        > possible. }}}}}}}}}}}}}
        >
        > Any help would be appreciated.
        >
        > Regards,
        >
        > Shlomi Fish

        --
        -----------------------------------------------------------------
        Shlomi Fish http://www.shlomifish.org/
        Original Riddles - http://www.shlomifish.org/puzzles/

        God gave us two eyes and ten fingers so we will type five times as much as we
        read.
      • Dotan Cohen
        ... Do you still need this benchmarked? -- Dotan Cohen http://what-is-what.com http://gibberish.co.il
        Message 3 of 3 , May 21, 2009
        • 0 Attachment
          > I recently created a thread-enabled version of Freecell Solver, that uses
          > several solver instances in separate threads. I'd like to get some indication
          > of the performance of the solver with different number of threads on different
          > machines. The instructions can be found below.
          >

          Do you still need this benchmarked?

          --
          Dotan Cohen

          http://what-is-what.com
          http://gibberish.co.il
        Your message has been successfully submitted and would be delivered to recipients shortly.