11th community OpenFOAM Workshop: finally managed to attend one of these! Part 2
Posted July 3, 2016 at 13:43 by wyldckat
This is the second part of a long blog post, see the first part here.
Continuing about the 29th:
Continuing about the 29th:
- Next was Robin Knowles presentation, with a pretty awesome presentation on the first season of 12 pod-casts he's done with CFD experts, although mostly oriented to the business side of things. He told us a bit about himself and his career; then he gave the highlights of the podcasts.
- One of the highlights that stuck with me was the one about Patrick Hanley of Hanley Innovations, where Patrick pretty much hand-crafts CFD binaries for Windows to have no extra bits in the binaries, going to the point where the biggest download available is a 10MB application.
- Another thing I just remembered is about UX: User Experience. GUI's may be nice looking, have all features or just a few, but it's the user's experience with the software that matters. I can't remember exactly if this was on one of the podcasts or if it was one of the lessons at the end of the first season.
- Afterwards was the coffee break, had a talk with Henrik Rüsche again, about limitations in OpenFOAM for GPU support and some of the foam-extend features that are going to be developed for the next iterations.
- Then I managed to meet and talk with a few more people from Engys.
- Ended up going with Paolo Geremia to the session track that started with "Introducing A New Client-Server Framework For Large CFD Models", with is part of their latest developments in HELYX.
- Pretty advanced tech, based on VTK and some hints at features that ParaView also provides from VTK.
- This to say that they are now able to leverage raw CPU and GPU power from servers, making it possible to post-process almost linearly with pretty big cases, as many cores and GPU cards you can invest on the hardware!
- In other words: post-processing in HELYX is no longer bound to workstations!
- The next presentation was "A Generic App Gui For OpenFOAM® And Variants" from the people at ICON that I met the day before (Jacques Papper to be precise).
- The title was updated to something less generic (I was one of the people who complained title seemed... too generic
). I can't remember the exact updated title, but it was inclined to be an app store, than just an app. - They essentially have a new - and free to use - infrastructure online where we can provide Python template scripts, along with JSON description files, where template-macros are used in the code to have a pre-processor in their software provide the GUI for simply filling out the entries that need to be adjusted.
- This makes it a lot easier to have snippets of code and user interface that assist in setting up the cases.
- The templates are shared on their platform between those registered on the website.
- It's a pretty neat tech, provided for free! (At least it's the idea I got from the presentation, or maybe I'm mixing up with something I talked the day before with them) And it's also part of their on-site products as well, therefore customers can also use the same templates on-site.
- The title was updated to something less generic (I was one of the people who complained title seemed... too generic
- Next was "An Update On The Integration Of ADIOS Into OpenFOAM For Handling Disk I/O".
- It's still something like version 0.2, but it's pretty powerful already! The next big step is to give it support for post-processing directly in ParaView.
- It's done and tested with OpenFOAM+, which includes a few features that are needed for this integration to make things run more optimally.
- I can't remember the exact numbers, but it was something like 240 cores (subdomains) writing time snapshots for a transient multiphase simulation, would take over 3000s over NFS, while using ADIOS would reduce it to 600s or 1200s!? I can't remember exactly. But the idea was that when using 24 cores (subdomains), the simulation would take around 1300s to do the same time steps.
- This is one of the topics of interest to me, because I've worked on this kind of stuff in the last couple of years of my graduation course at IST (Lisbon, Portugal, 2005-2006).
- At the end of that presentation I managed to talk with Fred Mendonça and we had a great+nice talk over lunch about OpenFOAM, OpenFOAM+, foam-extend, community, forum and so on.
- Most of the things we talked about, were presented the next day (Thursday) at 8h30 and later on one of the SIG meetings. More on that later below (or on part 3?).
- After lunch... I can't remember exactly, but I talked with several people, some to sort out questions in using OpenFOAM. And I ended up not being able to see the presentation "Open Source CFD Methodology, Research & Industrial Applications", by Torbjörn Larsson, Creo Dynamics, which I wanted to see.
- Then I was able to see most of the presentations at the "Multiphase Flows I" track.
- I think I missed the presentation "Solving Numerical Stability Problems In Multiphase Flows With Large Interphase Heat Transfer", given by Luiz Fernando L. R. Silva.
- The presentation "Two-Way Coupled Euler-Euler Simulation Of Drifting Snow" by Ziad Boutanios was the one that stuck with me, because the level of complexity that was used to perform the simulations... and well, it's about snow
- I had to grade a couple of presentations here, because the best student presentation of the workshop was also going to be awarded as well and I was just contributing a bit to the grading.
- And at the coffee break, Robin Knowles and I started talking with each other and before we knew it, we had been standing and talking for over an hour about all sorts of things. I ended up not attending the track "Multiphase Flows II".
- Then, back to the hotel, had a quick bath and went back to the venue, where we started a 15-20 minute walk to where the banquet was going to be, namely in the Palace of the Dukes of Braganza.
- At around 19h30, we were greeted by a drums band named Bohemia, composed of students of the University of Minho and they were in a formation of a Nabla: https://twitter.com/hrvoje_jasak/sta...53067133460480 - they did a performance for around 20-30 minutes, I think.
- Afterwards, we went into the palace, were we had a few drinks and appetizers in the foyer... mmm, it's not a foyer, it's an open space area that is inside the palace. Had a few nice chats and learned a few more things about GPU implementations from Daniel Combest (works at Engys), the main developer of the cufflink-library project: https://github.com/chegdan/cufflink-library
- Then the long dinner, we had soup, fish plate, meat plate and desert, with around 30-45 minutes between each, intertwined with live music sessions of Fado, a portuguese music genre. Sat at a table with a few of the people from the workshop committee. Had a few chats with Håkan Nilsson who was next to me; at one point we had a few brainstorming ideas on how to index the reports of his students onto openfoamwiki.net.
- Before I knew it, it was past midnight. Had a few chats with two fellow engineers about a couple of issues they were having with OpenFOAM and foam-extend.
- I left at... I don't know, sometime after 01h00. Got a bit lost on the way back to the hotel, my phone was having a hard time with GPS triangulation and I luckily found one of the engineers I had talked with a few minutes before (from the previous bullet point); he had a physical map and a better understanding of where we were and off I went in the right direction! GPS was back online after a few minutes and I managed to not get lost again.
- Took me a long time to fall asleep, specially because I had tried to build OpenFOAM-dev when I got back to the hotel, but stumbled over a snag while trying to build GCC 4.8.5 on CentOS 6.6, so I ended up giving up and fell asleep sometime after 03h00. It was the shortest night of sleep I had in the past few weeks.
Total Comments 0
