Friends, I am pleased to announce the release of the Accelerate 0.15 family of packages. Accelerate defines an embedded language of array computations for high performance computing in Haskell. Computations on multi-dimensional, regular arrays are expressed in the form of parameterised collective operations, such as maps, reductions, and permutations. These computations may then be online compiled and executed on a range of architectures, such as GPUs. This release brings mainly bug fixes and performance improvements. The following packages are available on Hackage: accelerate The language definition and reference implementation accelerate-cuda A high performance parallel backend targeting NVIDIA GPUs accelerate-io Fast conversion between Accelerate arrays and other formats, including ‘vector’ and ‘repa’. accelerate-fft Discrete Fourier transforms, backed by CUFFT where available accelerate-examples Computational kernels and applications showcasing Accelerate The code can be found on GitHub, where you can also submit issues: https://github.com/AccelerateHS https://github.com/AccelerateHS/accelerate/issues Finally, there is also a mailing list that can be used for both use and development discussions: accelerate-haskell@googlegroups.com http://groups.google.com/group/accelerate-haskell -The Accelerate Team
Excellent! Now here comes my number 1 feature request for the next version: IO/ST arrays/computations in accelerate, so that I can implement stuff that can work on a 1GB device array with interleaved IO, without having to download and-re-upload that array every time. Keep up the good work!
On 15-09-2014 11:08, Niklas Hambüchen wrote:
Excellent!
Now here comes my number 1 feature request for the next version:
IO/ST arrays/computations in accelerate, so that I can implement stuff that can work on a 1GB device array with interleaved IO, without having to download and-re-upload that array every time.
Do you mean using mmap'd memory? Cheers! -- Felipe.
[Sorry for duplicates, but I first had the wrong sender address.]
We always very gladly accept pull requests! :)
In any case, please add feature requests as tickets to the GitHub issue tracker (so they don’t get lost):
https://github.com/AccelerateHS/accelerate/issues
We are trying our best to add requested functionality, but our resources are also very limited — hence, any contributions are most welcome.
Manuel
Niklas Hambüchen
Excellent!
Now here comes my number 1 feature request for the next version:
IO/ST arrays/computations in accelerate, so that I can implement stuff that can work on a 1GB device array with interleaved IO, without having to download and-re-upload that array every time.
Keep up the good work! _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
Hey Trevor, Congrats for the new release. Tried to get back to my CUBLAS binding work recently, but it was a bit hard as the accelerate repos were transitionning to getting ready for 0.15. Just a quick question: is there any plan to make what's in http://www.cse.unsw.edu.au/~tmcdonell/papers/acc-multidev-icfp2014-sub.pdf land in accelerate any time soon? On Mon, Sep 15, 2014 at 4:04 PM, Trevor L. McDonell < tmcdonell@cse.unsw.edu.au> wrote:
Friends,
I am pleased to announce the release of the Accelerate 0.15 family of packages.
Accelerate defines an embedded language of array computations for high performance computing in Haskell. Computations on multi-dimensional, regular arrays are expressed in the form of parameterised collective operations, such as maps, reductions, and permutations. These computations may then be online compiled and executed on a range of architectures, such as GPUs.
This release brings mainly bug fixes and performance improvements.
The following packages are available on Hackage:
accelerate http://hackage.haskell.org/package/accelerate The language definition and reference implementation accelerate-cuda http://hackage.haskell.org/package/accelerate-cuda A high performance parallel backend targeting NVIDIA GPUs accelerate-io http://hackage.haskell.org/package/accelerate-io Fast conversion between Accelerate arrays and other formats, including ‘vector’ and ‘repa’. accelerate-fft http://hackage.haskell.org/package/accelerate-fft Discrete Fourier transforms, backed by CUFFT where available accelerate-examples http://hackage.haskell.org/package/accelerate-examples Computational kernels and applications showcasing Accelerate
The code can be found on GitHub, where you can also submit issues:
https://github.com/AccelerateHS https://github.com/AccelerateHS/accelerate/issues
Finally, there is also a mailing list that can be used for both use and development discussions:
accelerate-haskell@googlegroups.com http://groups.google.com/group/accelerate-haskell
-The Accelerate Team
-- You received this message because you are subscribed to the Google Groups "parallel-haskell" group. To unsubscribe from this group and stop receiving emails from it, send an email to parallel-haskell+unsubscribe@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
-- Alp Mestanogullari
On 15 Sep 2014, at 10:10 am, Alp Mestanogullari
Just a quick question: is there any plan to make what's in http://www.cse.unsw.edu.au/~tmcdonell/papers/acc-multidev-icfp2014-sub.pdf land in accelerate any time soon?
Yes, we have plans to revive and fully bake this work in the near future. Stay tuned! (: -Trev
participants (5)
-
Alp Mestanogullari -
Felipe Lessa -
Manuel M T Chakravarty -
Niklas Hambüchen -
Trevor L. McDonell