<div dir="ltr">adding Suyash,<div><br></div><div>I found the/a problem. Using ex56, which has a crappy decomposition, using one MPI process/GPU is much faster than using 8 (64 total). (I am looking at ex13 to see how much of this is due to the decomposition)</div><div>If you only use 8 processes it seems that all 8 are put on the first GPU, but adding -c8 seems to fix this.</div><div>Now the numbers are looking reasonable.</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 3:24 PM Barry Smith <<a href="mailto:bsmith@petsc.dev">bsmith@petsc.dev</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div style="overflow-wrap: break-word;"><div><br></div> For this, to start, someone can run <div><br></div><div><div style="margin:0px;font-stretch:normal;font-size:14px;line-height:normal;font-family:Menlo"><span style="font-variant-ligatures:no-common-ligatures">src/vec/vec/tutorials/performance.c </span></div><div style="margin:0px;font-stretch:normal;font-size:14px;line-height:normal;font-family:Menlo"><span style="font-variant-ligatures:no-common-ligatures"><br></span></div><div style="margin:0px;font-stretch:normal;line-height:normal"><span style="font-family:Helvetica;font-size:18px">and compare the performance to that in the technical report </span>Evaluation of PETSc on a Heterogeneous Architecture \\ the OLCF Summit System \\ Part I: Vector Node Performance. Google to find. One does not have to and shouldn't do an extensive study right now that compares everything, instead one should run a very small number of different size problems (make them big) and compare those sizes with what Summit gives. Note you will need to make sure that performance.c uses the Kokkos backend.</div><div style="margin:0px;font-stretch:normal;line-height:normal"><br></div><div style="margin:0px;font-stretch:normal;line-height:normal"> One hopes for better performance than Summit; if one gets tons worse we know something is very wrong somewhere. I'd love to see some comparisons.</div><div style="margin:0px;font-stretch:normal;line-height:normal"><br></div><div style="margin:0px;font-stretch:normal;line-height:normal"> Barry</div><div style="margin:0px;font-stretch:normal;line-height:normal"><br></div><div><br><blockquote type="cite"><div>On Jan 24, 2022, at 3:06 PM, Justin Chang <<a href="mailto:jychang48@gmail.com" target="_blank">jychang48@gmail.com</a>> wrote:</div><br><div><div dir="ltr">Also, do you guys have an OLCF liaison? That's actually your better bet if you do. <br><br>Performance issues with ROCm/Kokkos are pretty common in apps besides just PETSc. We have several teams actively working on rectifying this. However, I think performance issues can be quicker to identify if we had a more "official" and reproducible PETSc GPU benchmark, which I've already expressed to some folks in this thread, and as others already commented on the difficulty of such a task. Hopefully I will have more time soon to illustrate what I am thinking.</div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 1:57 PM Justin Chang <<a href="mailto:jychang48@gmail.com" target="_blank">jychang48@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="auto">My name has been called.</div><div dir="auto"><br></div><div dir="auto">Mark, if you're having issues with Crusher, please contact Veronica Vergara (<a href="mailto:vergaravg@ornl.gov" target="_blank">vergaravg@ornl.gov</a>). You can cc me (<a href="mailto:justin.chang@amd.com" target="_blank">justin.chang@amd.com</a>) in those emails</div></div><div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 1:49 PM Barry Smith <<a href="mailto:bsmith@petsc.dev" target="_blank">bsmith@petsc.dev</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div><br><div><br><blockquote type="cite"><div>On Jan 24, 2022, at 2:46 PM, Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> wrote:</div><br><div><div dir="ltr">Yea, CG/Jacobi is as close to a benchmark code as we could want. I could run this on one processor to get cleaner numbers.<div><br></div><div>Is there a designated ECP technical support contact?</div></div></div></blockquote><div><br></div> Mark, you've forgotten you work for DOE. There isn't a non-ECP technical support contact. </div><div><br></div><div> But if this is an AMD machine then maybe contact Matt's student Justin Chang?</div></div><div><div><br></div><div><br></div><div><br><blockquote type="cite"><div><div dir="ltr"><div><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 2:18 PM Barry Smith <<a href="mailto:bsmith@petsc.dev" target="_blank">bsmith@petsc.dev</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div><div><br></div> I think you should contact the crusher ECP technical support team and tell them you are getting dismel performance and ask if you should expect better. Don't waste time flogging a dead horse. <br><div><br><blockquote type="cite"><div>On Jan 24, 2022, at 2:16 PM, Matthew Knepley <<a href="mailto:knepley@gmail.com" target="_blank">knepley@gmail.com</a>> wrote:</div><br><div><div dir="ltr"><div dir="ltr">On Mon, Jan 24, 2022 at 2:11 PM Junchao Zhang <<a href="mailto:junchao.zhang@gmail.com" target="_blank">junchao.zhang@gmail.com</a>> wrote:<br></div><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 12:55 PM Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 1:38 PM Junchao Zhang <<a href="mailto:junchao.zhang@gmail.com" target="_blank">junchao.zhang@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">Mark, I think you can benchmark individual vector operations, and once we get reasonable profiling results, we can move to solvers etc.</div></blockquote><div><br></div><div>Can you suggest a code to run or are you suggesting making a vector benchmark code?</div></div></div></blockquote><div>Make a vector benchmark code, testing vector operations that would be used in your solver.</div><div>Also, we can run MatMult() to see if the profiling result is reasonable.</div><div>Only once we get some solid results on basic operations, it is useful to run big codes.</div></div></div></blockquote><div><br></div><div>So we have to make another throw-away code? Why not just look at the vector ops in Mark's actual code?</div><div><br></div><div> Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div><br><div><div><div dir="ltr"><div dir="ltr">--Junchao Zhang</div></div></div><br></div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 12:09 PM Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 24, 2022 at 12:44 PM Barry Smith <<a href="mailto:bsmith@petsc.dev" target="_blank">bsmith@petsc.dev</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div><div><br></div> Here except for VecNorm the GPU is used effectively in that most of the time is time is spent doing real work on the GPU<div><br></div><div><div>VecNorm 402 1.0 4.4100e-01 6.1 1.69e+09 1.0 0.0e+00 0.0e+00 4.0e+02 0 1 0 0 20 9 1 0 0 33 30230 225393 0 0.00e+00 0 0.00e+00 100</div><div><br></div><div>Even the dots are very effective, only the VecNorm flop rate over the full time is much much lower than the vecdot. Which is somehow due to the use of the GPU or CPU MPI in the allreduce?</div></div></div></blockquote><div><br></div><div>The VecNorm GPU rate is relatively high on Crusher and the CPU rate is about the same as the other vec ops. I don't know what to make of that.</div><div><br></div><div>But Crusher is clearly not crushing it. </div><div><br></div><div>Junchao: Perhaps we should ask Kokkos if they have any experience with Crusher that they can share. They could very well find some low level magic.</div><div><br></div><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div><div><div><br></div><div><br></div><div><br><blockquote type="cite"><div>On Jan 24, 2022, at 12:14 PM, Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> wrote:</div><br><div><div dir="ltr"><div dir="ltr"><br></div><div dir="ltr"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
Mark, can we compare with Spock?<br></blockquote><div><br></div><div> Looks much better. This puts two processes/GPU because there are only 4.</div></div></div>
</div>
<span id="gmail-m_-2479378921784158436gmail-m_6588895530181966739gmail-m_7911931152491619875m_-793860302615684860gmail-m_1602764271510560966gmail-m_7034942330692579726gmail-m_557496224917787162gmail-m_6873512096861806548gmail-m_-3802472944769362989gmail-m_2338317291810935402cid:f_kysy6if70"><jac_out_001_kokkos_Spock_6_1_notpl.txt></span></div></blockquote></div><br></div></div></blockquote></div></div>
</blockquote></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div><br clear="all"><div><br></div>-- <br><div dir="ltr"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div>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><br></div><div><a href="http://www.cse.buffalo.edu/~knepley/" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br></div></div></div></div></div></div></div></div>
</div></blockquote></div><br></div></blockquote></div>
</div></blockquote></div><br></div></blockquote></div></div>
</blockquote></div>
</div></blockquote></div><br></div></div></blockquote></div>