Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
bolha.us is one of the many independent Mastodon servers you can use to participate in the fediverse.
We're a Brazilian IT Community. We love IT/DevOps/Cloud, but we also love to talk about life, the universe, and more. | Nós somos uma comunidade de TI Brasileira, gostamos de Dev/DevOps/Cloud e mais!

Administered by:

Server stats:

254
active users

bolha.us: AboutProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.7

#blas

0 posts0 participants0 posts today
Christos Argyropoulos<p>Question for the <a href="https://mast.hpc.social/tags/rstats" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>rstats</span></a> crowd. Do you disable hyperthreads when you run analyses in R with a multithreaded version of <a href="https://mast.hpc.social/tags/blas" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>blas</span></a> e.g. <a href="https://mast.hpc.social/tags/openblas" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>openblas</span></a> <a href="https://mast.hpc.social/tags/mkl" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>mkl</span></a> etc ?</p>
Christos Argyropoulos MD, PhD, FASN 🇺🇸<p>Question for the <a class="hashtag" href="https://bsky.app/search?q=%23rstats" rel="nofollow noopener noreferrer" target="_blank">#rstats</a> crowd. Do you disable hyperthreads when you run analyses in R with a multithreaded version of <a class="hashtag" href="https://bsky.app/search?q=%23blas" rel="nofollow noopener noreferrer" target="_blank">#blas</a> e.g. <a class="hashtag" href="https://bsky.app/search?q=%23openblas" rel="nofollow noopener noreferrer" target="_blank">#openblas</a> <a class="hashtag" href="https://bsky.app/search?q=%23mkl" rel="nofollow noopener noreferrer" target="_blank">#mkl</a> etc ?</p>
Christos Argyropoulos<p>5 of these methods can leverage multithreaded (MT) <a href="https://mast.hpc.social/tags/BLAS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>BLAS</span></a> with a sweet spot ~ 6 threads for the 40% of the time spent in MT regions. E5-2697 has 36/72 (physical/logical) cores, so the avg case scenario is one in which 0.4x3x6 cores +2 (serial methods) tie up ~ 9.2 cores ~13% of the 72 logical cores. So far the back of envelope calculation, i.e. if I run 5 out of the 2100 design points in parallel, I will stay within 15% of resource use is holding rather well! <a href="https://mast.hpc.social/tags/benchmarking" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>benchmarking</span></a> <a href="https://mast.hpc.social/tags/hpc" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>hpc</span></a> <a href="https://mast.hpc.social/tags/rstats" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>rstats</span></a></p>
Christos Argyropoulos MD, PhD<p>Multiple cores to the rescue as I am using a custom D-optimal design to benchmark memory/CPU utilization of 7 alternative implementations of frailty models for big data from <a href="https://mstdn.science/tags/EHRs" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>EHRs</span></a>. By limiting the number of models that are run simultaneously in <a href="https://mstdn.science/tags/rstats" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>rstats</span></a> to use &lt; 30% of CPU one can treat concurrent runs as independent when evaluating the sweet spot of <a href="https://mstdn.science/tags/BLAS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>BLAS</span></a> threads (6) for these methods</p>
Christos Argyropoulos<p>Multiple cores to the rescue as I am using a custom D-optimal design to benchmark memory/CPU utilization of 7 alternative implementations of frailty models for big data from <a href="https://mast.hpc.social/tags/EHRs" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>EHRs</span></a>. By limiting the number of models that are run simultaneously in <a href="https://mast.hpc.social/tags/rstats" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>rstats</span></a> to use &lt; 30% of CPU one can treat concurrent runs as independent when evaluating the sweet spot of <a href="https://mast.hpc.social/tags/BLAS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>BLAS</span></a> threads (6) for these methods</p>
FCLC<p>Simple question: what is your *default* BLAS package? <br><a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>HPC</span></a> <a href="https://mast.hpc.social/tags/BLAS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>BLAS</span></a></p>
FCLC<p>Time for an <a href="https://mast.hpc.social/tags/introduction" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>introduction</span></a>! <br>I'm a young Canuck with interests/experience in <a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>HPC</span></a>, <a href="https://mast.hpc.social/tags/Linux" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Linux</span></a>, <a href="https://mast.hpc.social/tags/BLAS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>BLAS</span></a>, <a href="https://mast.hpc.social/tags/SYCL" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>SYCL</span></a>, <a href="https://mast.hpc.social/tags/C" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>C</span></a>, <a href="https://mast.hpc.social/tags/AVX512" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AVX512</span></a>, <a href="https://mast.hpc.social/tags/Rust" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Rust</span></a>, heterogeneous compute &amp; other such things. </p><p>Currently my personal projects are bringing <a href="https://mast.hpc.social/tags/FP16" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>FP16</span></a> to the <a href="https://mast.hpc.social/tags/OpenBLAS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>OpenBLAS</span></a> library, working to standardize what Complex domain BLAS FP16 kernels/implementations should look like, and making sure <a href="https://mast.hpc.social/tags/SYCL" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>SYCL</span></a> is available everywhere. </p><p>I also write every now and again. Here's the tail of AVX512 FP16 on Alderlake <br><a href="https://gist.github.com/FCLC/56e4b3f4a4d98cfd274d1430fabb9458" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">gist.github.com/FCLC/56e4b3f4a</span><span class="invisible">4d98cfd274d1430fabb9458</span></a></p>
ExploreLive feeds
About