The judges were really impressed by the incredible effort from across the community that saw the participants get 31 codes working on Graviton2-based #HPC clusters that previously only built and ran on other, mostly x86-based, lifeforms. 2/17
This lifts the water level for the whole Arm-HPC community. 3/17
The teams contributed loads of updates to @spackpm recipes, and built @reframehpc scripts so we can keep running these codes in CI/CD to make sure they only get better over time. 4/17
Automation will help us build a map of the performance space so we don't need to navigate by trial and error. That's a pretty new approach for a novel architecture in HPC. 5/17
The teams near the top of the table were all insanely good. But since A-HUG said there would be a prize (Apple M1 MacBook Pros hpc.news/macbook for the winning team), _there can be only one_ ... 6/17
The first prize in the AWS/Arm Summer Cloud #HPC#ahugHackathon goes to team ”DogeCoinToTheMoon” - a group of MSc students from @EdinburghUni. The judges were super impressed by not only their depth of analysis, but their breadth of coverage, too. 7/17
Check out just *one* of DogeCoin's many performance reports: hpc.news/kripke 8/17
However: the sheer strength of the SECOND PLACE team gave the panel a moment of pause. Team Wolfpack from @NCState in Raleigh really dived deep in their optimization studies, doing investigations into the code to produce some huge performance improvements. 9/17
Finally, Team Iman - one individual from @nyuniversity working all hours of the day and night from his apartment in Brooklyn, quite incredibly beat ALL the other teams to take THIRD PLACE. It's a pretty heroic achievement. 11/17
Not only that, but he contributed thoroughly to check-in discussions, *set the bar* for others with his Laghos study (hpc.news/laghos) and helped other people solve problems. He also ate the best pizza, because it was real NYC pizza. 12/17
As sponsors, @arm and @awscloud dug into our piggy banks so @ArmHPCUserGroup could award some extra prizes for 2nd place and 3rd place groups, too. 13/17
They'll all receive some Apple M1-based iPad Pros (apple.com/ipad-pro/), with our thanks, and the appreciation of the whole community. 14/17
This event contributed a lot of new working #HPC codes to the whole community. We're so thankful that we were able to be part of it, and to support it. 15/17
Would not have been possible without our army of volunteers who formed our mentors team for the #ahughackathon, because they helped - and inspired - all the teams to learn more, go further and try harder. 16/17
There's loads more to come from this event, so stay tuned over the coming weeks for more results, outcomes and outputs from this incredible community effort. We all had a barrel of fun, too. 17/17
A lot of folks have asked about what we used to build out the infrastructure for the hackathon. 1/22
I'm going to lean on @OllyPerksHPC and @CQnib and others who worked really closely to look after both the orchestration of the clusters themselves as well as the "insides" of the clusters (packages, compilers, libraries etc). 2/22
Firstly, we extensively used @awscloud#ParallelCluster, which takes spec files that more or less say - "gimme a cluster with, lemme see ... I think I want Slurm today, and up to 16 compute nodes made from Graviton2, I also want EFA for doing fast MPI stuff, oh, and a… 3/22
When we (@awscloud) started building EFA, our fast network fabric for HPC, I was skeptical whether we'd be able to run the really hard "latency-sensitive" codes like weather simulations or molecular dynamics. Boy was I wrong. Turns out: we rock at these codes. 1/16
We learned, tho, that single-packet latency is really distracting when trying to predict code performance on an HPC cluster.
Don't misunderstand me tho: it's not irrelevant. 2/16
It's just that so many of us in HPC ignored the real goal which is for MPI ranks on different machines to exchange chunks of data quickly - we mistook single-packet transit times as a proxy for that. 3/16