<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Mon, Aug 10, 2015 at 7:47 AM, Leoni, Massimiliano <span dir="ltr"><<a href="mailto:Massimiliano.Leoni@rolls-royce.com" target="_blank">Massimiliano.Leoni@rolls-royce.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">> -----Original Message-----<br>
> From: Karl Rupp [mailto:<a href="mailto:rupp@iue.tuwien.ac.at">rupp@iue.tuwien.ac.at</a>]<br>
> Sent: 10 August 2015 11:54<br>
> To: Leoni, Massimiliano<br>
> Cc: <a href="mailto:slepc-maint@upv.es">slepc-maint@upv.es</a>; <a href="mailto:petsc-dev@mcs.anl.gov">petsc-dev@mcs.anl.gov</a><br>
> Subject: Re: [petsc-dev] [GPU - slepc] Hands-on exercise 4 (SVD) not working<br>
> with GPU and default configurations<br>
<br>
> The use of aijcusp instead of a dense matrix type certainly adds to the issue.<br>
I know, but I couldn't find a dense gpu type in the petsc manual, please correct me if there is any.<br>
<br>
> Please send the output of -log_summary so that we can see where most<br>
> time is spent.<br>
I am unable to do that as somehow I am having no output when I use that option. I also tried to explicitly call PetscLogView but still nothing is printed out.<br>
If I try with one of the slepc examples, I get the output.<br>
Why is this happening? If I run my code with -info or -log_trace I see their output, only -log_summary is shy!<br></blockquote><div><br></div><div>Try calling PetscLogBegin() after PetscInitialize(). We have now put in an error if this is not initialized correctly.</div><div><br></div><div>I agree with Karl that not much speedup can be expected with GPUs. This is the fault of dishonest marketing. None</div><div>of the computations in PETSc are limited by the computation rate, rather they are limited by memory bandwidth. The</div><div>bandwidth is at best 2-3x better, and less for modern CPUs. The dense SVD can be better than this, but you are</div><div>eventually limited by offload times and memory latency. The story of 100x, or even 10x, speedups is just a fraud.</div><div><br></div><div> Thanks,</div><div><br></div><div> Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
> If you have good (recent) CPUs in dual-socket configuration, it's more than<br>
> unlikely that you will gain anything beyond ~2x with an optimized GPU setup.<br>
> Even that ~2x may only be possible with heavily tweaking the current SVD-<br>
> implementation in SLEPc, of which I don't know the details.<br>
I used Xeon processors from 2010, just like the GPUs.<br>
This is not good news, as my supervisor is really optimist about using GPUs and getting high speed-ups!<br>
Anyway, at the moment my gpu version is several times slower than the cpu version, so even a 2x would be a win now :D<br>
<br>
<br>
Massimiliano<br>
The data contained in, or attached to, this e-mail, may contain confidential information. If you have received it in error you should notify the sender immediately by reply e-mail, delete the message from your system and contact <a href="tel:%2B44%20%280%29%201332%20622800" value="+441332622800">+44 (0) 1332 622800</a>(Security Operations Centre) if you need assistance. Please do not copy it for any purpose, or disclose its contents to any other person.<br>
<br>
An e-mail response to this address may be subject to interception or monitoring for operational reasons or for lawful business practices.<br>
<br>
(c) 2015 Rolls-Royce plc<br>
<br>
Registered office: 62 Buckingham Gate, London SW1E 6AT Company number: 1003142. Registered in England.<br>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div class="gmail_signature">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</div>
</div></div>