<div dir="ltr">It could be also that you're hitting thread migration across cores, thus missing out on the clock ramp up. I saw quite a few issues with that on the p8 running benchmarks a while back. If that's the case then you can try to pin your benchmark to a single cpu core (or a set of them if youd like) by running it under "numactl -C <CORE_NUMBER> ./benchmark"<br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Oct 28, 2018 at 9:23 AM Jeffrey Walton via cfarm-users <<a href="mailto:cfarm-users@lists.tetaneutral.net">cfarm-users@lists.tetaneutral.net</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">On Sun, Oct 28, 2018 at 8:42 AM Andy Polyakov via cfarm-users<br>
<<a href="mailto:cfarm-users@lists.tetaneutral.net" target="_blank">cfarm-users@lists.tetaneutral.net</a>> wrote:<br>
><br>
> > The initial benchmarks are kind of flat when using 3.8 GHz as the<br>
> > frequency. I think the problem is, we are not working the machine hard<br>
> > enough so the cpu's are reluctant to move from a low energy state.<br>
><br>
> I'd say it's more likely because POWER9 appears to be "allergic" to<br>
> mixtures of vector and scalar instructions. And since you are likely to<br>
> reference memory you will always have scalar instructions at least to<br>
> calculate effective addresses. Normalized[!] difference to POWER8 can be<br>
> anywhere from "little" to a "lot". Example of "little" can be ~15% in<br>
> SHA512(*) and VSX Chacha20. Example of "lot" is ~50% for pre-VSX<br>
> Chacha20 implementation where one interleaves scalar and vector in more<br>
> or less equal proportion. Though on the other hand pure scalar code is<br>
> normally faster...<br>
<br>
Thanks Andy.<br>
<br>
That's what we are seeing. AES and SHA slowed down, and the ChaChaR<br>
sped-up (even the SIMD version of ChaCha benefited).<br>
<br>
Jeff<br>
_______________________________________________<br>
cfarm-users mailing list<br>
<a href="mailto:cfarm-users@lists.tetaneutral.net" target="_blank">cfarm-users@lists.tetaneutral.net</a><br>
<a href="https://lists.tetaneutral.net/listinfo/cfarm-users" rel="noreferrer" target="_blank">https://lists.tetaneutral.net/listinfo/cfarm-users</a><br>
</blockquote></div>