Hi Jake, I could provide more details about the HPXCL, espacially the CUDA support there. We could do a Skype meeting or a chat in Google Hangouts.
Best, Patrick Am 07/06/2016 um 11:35 PM schrieb Hemstad, Jacob: > Hello, > > > My name is Jake Hemstad, I'm a High Performance Computing researcher at > Sandia National Laboratories while working on my PhD through the > University of Minnesota. > > > I work with Mike Heroux at Sandia and we are currently working on > creating detailed (2-3 pages) summaries targeted towards the HPC > community of many of the modern task-parallel runtimes. > > > I'm currently summarizing HPX (your HPX, which we call HPX-3) and I've > included my current draft of the HPX summary and I'm hoping you can > point out what I've got wrong and the important bits I've left out. HPX > has definitely been the neatest runtime I've summarized thus far and I'd > like to get it right. > > > I was also hoping to pick the brains of one of you over at LSU to answer > a number of questions I have had difficulty finding answers to in the > documentation. What I would really enjoy is a phone call with someone > who is willing to chat with me for a while and put up with a peppering > of questions, but I'll try and provide a sampling of what I'm still > trying to figure out: > > > * What is the state of GPU support in HPX? > o I've seen "percolation" mentioned several times in both the > ParalleX and HPX description in relation to accelerators, but > have yet to find any concrete explanation of what it is. > o I've read your blog post here > > <http://stellar.cct.lsu.edu/2016/02/c-and-the-heterogeneous-challenge/> and > poster here > <http://stellar.cct.lsu.edu/pubs/stumpf_poster_mandelbrot.pdf> and > I'm still hoping for more details. I've also looked through your > HPXCL repo <https://github.com/STEllAR-GROUP/hpxcl> and I > haven't been able to find a detailed overview of what HPXCL does > or how it works. > * Threads are first class objects (i.e. have a globally accessible > name in the AGAS)? > o This is a key point in several of the ParalleX and HPX papers, > but I don't see how this is realized in HPX-3. > o In fact, given this talk > <https://www.youtube.com/watch?v=4OCUEgSNIAY> by Hartmut, I'm > surprised that threads would be first class objects in HPX-3. > o Is this simply a remnant of ParalleX that you've since moved > away from? > * I need some help to more concretely understand executors > > Thank you for any and all help you'd be willing to provide. > > Cheers, > Jake > > > > > _______________________________________________ > hpx-users mailing list > [email protected] > https://mail.cct.lsu.edu/mailman/listinfo/hpx-users > -- Patrick Diehl Institut für Numerische Simulation Rheinische Friedrich-Wilhelms Universität Bonn Wegelerstrasse 4 , Room 0.001 D-53115 Bonn Germany Tel : +49 228 733846 Mail: [email protected]
signature.asc
Description: OpenPGP digital signature
_______________________________________________ hpx-users mailing list [email protected] https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
