On Thu, Apr 9, 2009 at 11:10 AM, Farshid Mossaiby <span dir="ltr"><<a href="mailto:mossaiby@yahoo.com">mossaiby@yahoo.com</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>
Hi,<br>
<br>
May I know how can I take a look at current PETSc-GPU?</blockquote><div><br>Nothing is ready yet.<br><br> Matt<br> </div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>
--- On Thu, 4/9/09, Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>> wrote:<br>
<br>
> From: Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>><br>
> Subject: Re: PETSc acceleration on novel architectures<br>
> To: "For users of the development version of PETSc" <<a href="mailto:petsc-dev@mcs.anl.gov">petsc-dev@mcs.anl.gov</a>><br>
> Date: Thursday, April 9, 2009, 7:38 PM<br>
<div><div></div><div class="h5">> On Wed, Apr 8, 2009 at 10:29 PM, Ahmed El Zein<br>
> <<a href="mailto:ahmed@azein.com">ahmed@azein.com</a>> wrote:<br>
><br>
> > On Wed, 2009-04-08 at 10:52 -0500, Matthew Knepley<br>
> wrote:<br>
> > > On Tue, Apr 7, 2009 at 9:51 PM, Ahmed El Zein<br>
> <<a href="mailto:ahmed@azein.com">ahmed@azein.com</a>> wrote:<br>
> > > On Tue, 2009-04-07 at 10:39 -0500,<br>
> Matthew Knepley wrote:<br>
> > > > On Tue, Apr 7, 2009 at 9:07 AM,<br>
> Stephen Ball<br>
> > > > <<a href="mailto:Stephen.R.Ball@awe.co.uk">Stephen.R.Ball@awe.co.uk</a>><br>
> wrote:<br>
> > > > Hi<br>
> > > ><br>
> > > > We are keen to start<br>
> investigating whether or not<br>
> > > PETSc is<br>
> > > > suitable for<br>
> > > > acceleration on novel<br>
> architectures like GPUs, Cell<br>
> > > > processors, etc.<br>
> > > ><br>
> > > > I would very much like to<br>
> get your opinions on this.<br>
> > > ><br>
> > > > Do you think such an<br>
> endeavour is at all feasible<br>
> > > with PETSc?<br>
> > > > If so,<br>
> > > > what areas of PETSc do you<br>
> think our efforts would<br>
> > > best be<br>
> > > > spent?<br>
> > > ><br>
> > > > Should we for example focus<br>
> on matrix operations, or<br>
> > > on<br>
> > > > specific<br>
> > > > preconditioners or solvers?<br>
> Where would be a good<br>
> > > place to<br>
> > > > start?<br>
> > > ><br>
> > > > Can you suggest some<br>
> specific routines/functions in<br>
> > > PETSc that<br>
> > > > are<br>
> > > > potential candidates for<br>
> acceleration?<br>
> > > ><br>
> > > > We are actually already working on<br>
> this, and I plan on<br>
> > > having a<br>
> > > > PETSc-GPU<br>
> > > > come out at the end of the year.<br>
> > ><br>
> > ><br>
> > > What language are you using? I would have<br>
> thought that OpenCL<br>
> > > would be<br>
> > > the best solution. Maybe even rewriting<br>
> the whole of PETSc in<br>
> > > OpenCL,<br>
> > > targeting both multicore CPUs and a few<br>
> novel architectures at<br>
> > > the same<br>
> > > time. AMD and NVIDIA are both going to<br>
> support OpenCL for<br>
> > > their GPUs and<br>
> > > I believe that it will be supported on<br>
> Intel's larrabee and<br>
> > > the Cell<br>
> > > processor.<br>
> > ><br>
> > > That is probably a mistake. OpenCL is not mature<br>
> and only a few<br>
> > > operations in<br>
> > > PETSc would really benefit.<br>
> > I would still be interested in what language you are<br>
> using for PETSc-GPU<br>
> > and what PETSc-GPU is?<br>
><br>
><br>
> I would like to use OpenCL, but CUDA is it right now.<br>
><br>
> Matt<br>
><br>
><br>
> ><br>
> > Ahmed<br>
> > ><br>
> > > Matt<br>
> > ><br>
> > ><br>
> > > Ahmed<br>
> > ><br>
> > ><br>
> > > > Therefore, I suggest working on PCs<br>
> that<br>
> > > > are specific to your problems.<br>
> People are already doing good<br>
> > > work on<br>
> > > > sparse<br>
> > > > matrices in general, and solvers<br>
> will see no speedup at all,<br>
> > > since<br>
> > > > they are all<br>
> > > > logic.<br>
> > > ><br>
> > > > Matt<br>
> > > ><br>
> > > ><br>
> > > > Regards<br>
> > > ><br>
> > > > Stephen R. Ball<br>
> > > > Advanced Technologies<br>
> > > > HPC<br>
> > > > DRAS<br>
> > > > Rm: G17<br>
> > > > Bldg: E1.1<br>
> > > > AWE(A)<br>
> > > > Aldermaston<br>
> > > > Reading<br>
> > > > Berkshire<br>
> > > > ENGLAND<br>
> > > > RG7 4PR<br>
> > > > Tel: +44 (0)118 982 4528<br>
> > > > e-mail:<br>
> <a href="mailto:stephen.r.ball@awe.co.uk">stephen.r.ball@awe.co.uk</a><br>
> > > ><br>
> > > > This e-mail and any<br>
> attachments may contain<br>
> > > confidential and<br>
> > > > privileged information. If<br>
> you are not the intended<br>
> > > recipient,<br>
> > > > please notify the sender<br>
> immediately by return<br>
> > > e-mail, delete<br>
> > > > this<br>
> > > > e-mail and destroy any<br>
> copies. Any dissemination or<br>
> > > use of<br>
> > > > this<br>
> > > > information by a person<br>
> other than the intended<br>
> > > recipient is<br>
> > > > unauthorized and may be<br>
> illegal.<br>
> > > ><br>
> > > ><br>
> > > ><br>
> > > > --<br>
> > > > What most experimenters take for<br>
> granted before they begin<br>
> > > their<br>
> > > > experiments is infinitely more<br>
> interesting than any results<br>
> > > to which<br>
> > > > their experiments lead.<br>
> > > > -- Norbert Wiener<br>
> > ><br>
> > ><br>
> > ><br>
> > ><br>
> > ><br>
> > > --<br>
> > > What most experimenters take for granted before<br>
> they begin their<br>
> > > experiments is infinitely more interesting than<br>
> any results to which<br>
> > > their experiments lead.<br>
> > > -- Norbert Wiener<br>
> ><br>
> ><br>
><br>
><br>
> --<br>
> What most experimenters take for granted before they begin<br>
> their experiments<br>
> is infinitely more interesting than any results to which<br>
> their experiments<br>
> lead.<br>
> -- Norbert Wiener<br>
<br>
<br>
<br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br>What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>
-- Norbert Wiener<br>