repa not running in parallel?
Hi, I'm new to haskell and I tried to reproduce the perfomance values from the paper "Regular, Shape-polymorphic, Parallel Arrays in Haskell". I modified the repa-mmult example from the repa-example package to use the mmultP from Data.Array.Repa.Algorithms.Matrix. Then I compile it with "ghc -threaded -fglasgow-exts -O Main.hs" but running it with two large matrices : "./Main -random 4096 4096 -random 4096 4096" I can see that only one of my 4 cpus is utilized. What am I doing wrong? Note: The unmodified repa-example also does not run in parallel. Thx, Alex
Try adding +RTS -N at the end of your ./Main command line. * Alexander Herz <alexander.herz@mytum.de> [2013-11-07 10:40:59+0100]
Hi,
I'm new to haskell and I tried to reproduce the perfomance values from the paper "Regular, Shape-polymorphic, Parallel Arrays in Haskell".
I modified the repa-mmult example from the repa-example package to use the mmultP from Data.Array.Repa.Algorithms.Matrix.
Then I compile it with "ghc -threaded -fglasgow-exts -O Main.hs" but running it with two large matrices :
"./Main -random 4096 4096 -random 4096 4096" I can see that only one of my 4 cpus is utilized.
What am I doing wrong?
Note: The unmodified repa-example also does not run in parallel.
Thx, Alex _______________________________________________ Haskell mailing list Haskell@haskell.org http://www.haskell.org/mailman/listinfo/haskell
I think you need to specify how many cores to use when you call your ./Main program. Something like "./Main -random 4096 4096 -random 4096 4096 +RTS -N4" (given that you have 4 cores). You'll also need to add -rtsopts when you're compiling your code with ghc. On Thu, Nov 7, 2013 at 1:40 AM, Alexander Herz <alexander.herz@mytum.de>wrote:
Hi,
I'm new to haskell and I tried to reproduce the perfomance values from the paper "Regular, Shape-polymorphic, Parallel Arrays in Haskell".
I modified the repa-mmult example from the repa-example package to use the mmultP from Data.Array.Repa.Algorithms.Matrix.
Then I compile it with "ghc -threaded -fglasgow-exts -O Main.hs" but running it with two large matrices :
"./Main -random 4096 4096 -random 4096 4096" I can see that only one of my 4 cpus is utilized.
What am I doing wrong?
Note: The unmodified repa-example also does not run in parallel.
Thx, Alex _______________________________________________ Haskell mailing list Haskell@haskell.org http://www.haskell.org/mailman/listinfo/haskell
Problem solved! Sorry for using the wrong list. Alex On 11/07/2013 10:40 AM, Alexander Herz wrote:
Hi,
I'm new to haskell and I tried to reproduce the perfomance values from the paper "Regular, Shape-polymorphic, Parallel Arrays in Haskell".
I modified the repa-mmult example from the repa-example package to use the mmultP from Data.Array.Repa.Algorithms.Matrix.
Then I compile it with "ghc -threaded -fglasgow-exts -O Main.hs" but running it with two large matrices :
"./Main -random 4096 4096 -random 4096 4096" I can see that only one of my 4 cpus is utilized.
What am I doing wrong?
Note: The unmodified repa-example also does not run in parallel.
Thx, Alex _______________________________________________ Haskell mailing list Haskell@haskell.org http://www.haskell.org/mailman/listinfo/haskell
Hi Alexander, On 07/11/13 09:40, Alexander Herz wrote:
Hi,
I'm new to haskell and I tried to reproduce the perfomance values from the paper "Regular, Shape-polymorphic, Parallel Arrays in Haskell".
I modified the repa-mmult example from the repa-example package to use the mmultP from Data.Array.Repa.Algorithms.Matrix.
Then I compile it with "ghc -threaded -fglasgow-exts -O Main.hs" but running it with two large matrices :
"./Main -random 4096 4096 -random 4096 4096" I can see that only one of my 4 cpus is utilized.
What am I doing wrong?
Probably, not running with ./Main +RTS -N -RTS ... http://www.haskell.org/ghc/docs/7.6.3/html/users_guide/using-smp.html#parall...
Note: The unmodified repa-example also does not run in parallel.
Thx, Alex
Claude -- http://mathr.co.uk
participants (4)
-
Alexander Herz -
Claude Heiland-Allen -
Roman Cheplyaka -
Tikhon Jelvis