<div dir="ltr">On Fri, Jul 24, 2015 at 1:58 PM, Ian Hinder <span dir="ltr"><<a href="mailto:ian.hinder@aei.mpg.de" target="_blank">ian.hinder@aei.mpg.de</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div style="word-wrap:break-word"><br><div><span class=""><div>On 24 Jul 2015, at 19:42, Erik Schnetter <<a href="mailto:schnetter@cct.lsu.edu" target="_blank">schnetter@cct.lsu.edu</a>> wrote:</div><br><blockquote type="cite"><div dir="ltr">On Fri, Jul 24, 2015 at 1:39 PM, Ian Hinder <span dir="ltr"><<a href="mailto:ian.hinder@aei.mpg.de" target="_blank">ian.hinder@aei.mpg.de</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div style="word-wrap:break-word"><span><br><div><div>On 24 Jul 2015, at 19:15, Erik Schnetter <<a href="mailto:schnetter@cct.lsu.edu" target="_blank">schnetter@cct.lsu.edu</a>> wrote:</div><br><blockquote type="cite"><div dir="ltr">On Fri, Jul 24, 2015 at 11:57 AM, Ian Hinder <span dir="ltr"><<a href="mailto:ian.hinder@aei.mpg.de" target="_blank">ian.hinder@aei.mpg.de</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="word-wrap:break-word"><br><div><span><div>On 8 Jul 2015, at 16:53, Ian Hinder <<a href="mailto:ian.hinder@aei.mpg.de" target="_blank">ian.hinder@aei.mpg.de</a>> wrote:</div><br><blockquote type="cite"><div style="word-wrap:break-word"><br><div><div>On 8 Jul 2015, at 15:14, Erik Schnetter <<a href="mailto:schnetter@cct.lsu.edu" target="_blank">schnetter@cct.lsu.edu</a>> wrote:</div><br><blockquote type="cite"><div dir="ltr">I added a second benchmark, using a Thornburg04 patch system, 8th order finite differencing, and 4th order patch interpolation. The results are<div><br></div><div><div style="margin:0px;font-size:10px;font-family:Menlo">original: 8.53935e-06 sec</div><div style="margin:0px;font-size:10px;font-family:Menlo">rewrite: 8.55188e-06 sec</div><div style="margin:0px;font-size:10px;font-family:Menlo"><br></div><div style="margin:0px;font-size:10px;font-family:Menlo"><span style="font-family:arial,sans-serif;font-size:small">this time with 1 thread per MPI process, since that was most efficient in both cases. Most of the time is spent in inter-patch interpolation, which is much more expensive than in a "regular" case since this benchmark is run on a single node and hence with very small grids.</span><br></div><div style="margin:0px;font-size:10px;font-family:Menlo"><span style="font-family:arial,sans-serif;font-size:small"><br></span></div><div style="margin:0px;font-size:10px;font-family:Menlo"><span style="font-family:arial,sans-serif;font-size:small">With these numbers under our belt, can we merge the rewrite branch?</span></div></div></div></blockquote><div><br></div><div>The "jacobian" benchmark that I gave you was still a pure kernel benchmark, involving no interpatch interpolation. It just measured the speed of the RHSs when Jacobians were included. I would also not use a single-threaded benchmark with very small grid sizes; this might have been fastest in this artificial case, but in practice I don't think we would use that configuration. The benchmark you have now run seems to be more of a "complete system" benchmark, which is useful, but different.</div><div><br></div><div>I think it is important that the kernel itself has not gotten slower, even if the kernel is not currently a major contributor to runtime. We specifically split out the advection derivatives because they made the code with 8th order and Jacobians a fair bit slower. I would just like to see that this is not still the case with the new version, which has changed the way this is handled.</div></div></div></blockquote><div><br></div></span><div>I have now run my benchmarks on both the original and the rewritten McLachlan. I seem to find that the ML_BSSN_* functions in</div><div>Evolve/CallEvol/CCTK_EVOL/CallFunction/thorns, excluding the constraint calculations, are between 11% and 15% slower with the rewrite branch, depending on the details of the evolution. See attached plot. This is on Datura with quite old CPUs (Intel Xeon CPU X5650 2.67GHz).</div></div></div></blockquote><div><br></div><div>What exactly do you measure -- which bins or routines? Does this involve communication? Are you using thorn Dissipation?</div></div></div></div></blockquote></div><div><br></div></span><div>I take all the timers in Evolve/CallEvol/CCTK_EVOL/CallFunction/thorns that start with ML_BSSN_ and eliminate the ones containing "constraints" (case insensitive). This is running on two processes, one node, 6 threads per node. Threads are correctly bound to cores. There is ghostzone exchange between the processes, so yes, there is communication in the ML_BSSN_SelectBCs SYNC calls, but it is node-local.</div></div></blockquote><div><br></div><div>Can you include thorn Dissipation in the "before" case, and use McLachlan's dissipation in the "after" case?</div></div></div></div></blockquote><div><br></div></span><div>There is no dissipation in either case.</div><div><br></div><div>The output data is in</div><div><br></div><div><span style="white-space:pre-wrap"> </span><a href="http://git.barrywardell.net/?p=McLachlanBenchmarks.git;h=refs/runs/orig/20150724-174334" target="_blank">http://git.barrywardell.net/?p=McLachlanBenchmarks.git;h=refs/runs/orig/20150724-174334</a></div><div><span style="white-space:pre-wrap"> </span><a href="http://git.barrywardell.net/?p=McLachlanBenchmarks.git;h=refs/runs/rewrite/20150724-170542" target="_blank">http://git.barrywardell.net/?p=McLachlanBenchmarks.git;h=refs/runs/rewrite/20150724-170542</a></div><div><br></div><div>including the parameter files.</div><div><br></div><div>Actually, what I said before was wrong; the timers I am using are under "thorns", not "syncs", so even the node-local communication should not be counted.</div></div></div></blockquote><div><br></div><div>McLachlan has not been optimized for runs without dissipation. If you this this is important, then we can introduce a special case. I expect this to improve performance. However, running BSSN without dissipation is not what one would do in production, so I didn't investigate this case.</div><div><br></div><div>-erik</div></div><div><br></div>-- <br><div class="gmail_signature">Erik Schnetter <<a href="mailto:schnetter@cct.lsu.edu" target="_blank">schnetter@cct.lsu.edu</a>><br><a href="http://www.perimeterinstitute.ca/personal/eschnetter/" target="_blank">http://www.perimeterinstitute.ca/personal/eschnetter/</a></div>
</div></div>