Hello again! This blog has been dormant for a while because I’ve been super busy with my day job and side job (only so many hours in the day, etc). But! I have been working with a few great communicators to create some cool content, so I figured it’d be good to post the links here so they’re available in one place. I’ve also given a few talks recently, not all of which were recorded. I’m pretty proud of them though (and they were quite well received by the audience), so I thought it could be good to turn my notes and slides into blog posts to make them more widely available. Those aren’t quite ready yet, but I have a folder full of drafts so keep an eye out over the next few months.

Anyway, that’s enough pre-amble; here’s the content!

SC23: Women’s History Month Feature

I was recently interviewed by Cristin Merritt as part of a series profiling women in HPC. It was great to be a part of and Cristin was a delight to work with, and it gave me a chance to express some long-held but seldom shared opinions on equity and diversity initiatives in large organisations. Cristin’s put together a fantastic series of profiles with some really cool women, so check out the rest of the series once you’ve read mine: https://sc23.supercomputing.org/category/i-am-hpc/.

Cosmos: Simulating Molecule Flow - the Future of Desalination

My boss Debra and I were recently interviewed by the excellent Ellen Phiddian from Cosmos Magazine, talking about our group’s work on doing large-scale simulations of molecular fluid flow. Non-equilibrium statistical mechanics is an extremely important field of classical physics, so when I first started at AIBN I was astonished to learn how much there still is to discover about the fundamental physics of systems far from thermodynamic equilibrium. I think we did a good job of condensing down the important physics (although there was one subtle error at the beginning of the video that doesn’t really take away from the message, no need to point it out if you spot it). There’s a video on YouTube and a longer version of the interview available as a podcast.

Pawsey: Enter the flow state — what the atomic world could teach us about the future of batteries and more

Lastly, here’s another piece on the fluid flow project. This time we were interviewed by the Pawsey Centre about our work to develop massively-parallel algorithms for molecular fluid flow.

This is distinct from the better-known computational fluid dynamics (CFD): CFD deals with continuous fluids, we’re interested in modelling the motion of individual molecules and how that might affect physical or chemical reactions. These kinds of processes are really interesting because we can potentially use flow to drive self-assembly of carbon nanostructures like graphene sheets with fewer byproducts than traditional chemical manufacturing processes (here’s a cool experimental paper if you want to know more: https://doi.org/10.1016/j.ensm.2022.09.026).

Fluid flow is somewhat tricky to correctly model at the molecular level and there’s lots of subtle traps you can fall into when translating it into molecular dynamics code, so we’re aiming to make the One True Implementation and to write it so we can take advantage of modern, heterogeneous HPC architectures. Ideally, we’ll end up with some code that is robust, really fast and open-source. Isn’t that the dream?

Anyway it’s a pretty good (if somewhat informal) overview of what I/we’ve been up to lately.

Some other cool links

Not strictly “Emily content”, but I’ve read some really cool stuff this last week and I thought it’d be fun to chuck them in at the end of this link round-up.

The Open Source Way

First, I’ve been reading a really great guidebook on community management for open source software, put together by some people from Red Hat and other major FOSS projects. Community management (and the closely related problem of managing a distributed development team) are necessary prerequisites for sustainable open-source development, so it’s something we in the computational sciences should really be paying attention to. I’ll have more to say on this in another post, but in the meantime, this is an excellent primer on community management (even some of its prescriptions may not fit communities outside of the commercial open-source space).

Metal Benchmarks, by Philip Turner

And second, I’ve been messing around with a spare M1 MacBook that we had lying around the office to see whether or not the M1 GPU could be a viable compute platform for our work. Benchmarks like this one are promising, especially when considering its impressive performance-to-price ratio (very important to a small, grant-funded research group).

Unfortunately, Apple being Apple, there’s a lot of murky, proprietary stuff that makes it hard to get a view into what the M1 is actually doing at a low level. Luckily, philipturner published an awesome deep dive into the M1 microarchitecture, complete with a suite of benchmarks that you can run to get a feel for its performance characteristics. This has made my job so much easier (since I’m not starting from scratch) and it’s also just super interesting stuff. Check it out!

Conclusion

That’s all for this week! Hope you enjoy it and I’ll see you next time. Keep an eye out for more posts, and if you have any feedback, you can contact me at emily [at] atomwitch [dot] net.