A Student Volunteer at SC25

I look forward with anticipation to SC every year, and this time was no different. It is an entire week dedicated to HPC, a place for academia and industry to meet; the event where, once a year, everyone in the HPC community gathers to discuss what they have been working on. New insights arise, new research collaborations begin, and opportunities are created. 

Student volunteer programme

This year, I was part of the Student Volunteer programme, an initiative of the conference that covers accommodation and (this year) travel costs in exchange for working at the conference to ensure it runs smoothly. It is an excellent opportunity to mingle with students from all over the world who are working on similar topics and to engage with the conference. 

As a tenured student volunteer, I can only recommend the experience to everyone working in HPC. This year around 400 students participated in the programme, which gives an indication of the scale of SC25, where over 15,000 people attended the technical programme. 

From Edinburgh to St Louis

I woke up early on 14 November in Edinburgh to catch my flight to St Louis, the city hosting the conference this year. After a long transatlantic journey, I was welcomed at my destination by a very hot sunny day – what a contrast with Scotland. But this was just an illusion, as the weather got increasingly cold during the conference! 

Student volunteers start their duties the day before the conference opens. We began with a briefing on how to navigate the convention centre (as we were the go-to people if attendees had any questions about the venue), as well as on security and emergencies. Once all the serious matters had been covered, there was a social event so the students could get to know each other. These are always fun and offer a great opportunity to make new friends and start potential collaborations! 

IA³ Session chair

The conference kicked off on Sunday with the workshop sessions. I was chair for the IA³ GPU graph and parallel algorithms session, a role I also fulfilled last year and one I find interesting as it provides a different point of view. We students focus mostly on our own research and are often unaware of how the events at which we present our work are organised, although they are key to ensuring research is communicated. After this, the programme became a myriad of events happening in parallel - schedules must be prepared in advance to avoid missing out on the talks you are interested in. 

LLVM Workshop talk

On Monday, I gave my first talk at the LLVM workshop, where I presented the paper “An MLIR pipeline for offloading Fortran to FPGAs via OpenMP.” 

The paper builds on previous work from my PhD and other work from my group that was presented at the same workshop last year. It proves the point that the composability of the MLIR infrastructure can be effectively leveraged to offload Fortran code to FPGAs without developing a whole new compiler from scratch – this is the approach taken in the state of the art. 

Not only do we demonstrate that it is possible to achieve this for Fortran, but we also lay the groundwork for doing the same with C/C++, and we provide pointers on how to achieve this with minimal technical effort. The presentation went smoothly, and I got some interest from Fortran enthusiasts. It is so rewarding when other researchers are interested in your work. 

Exhibition opening

In the evening, the exhibition opened, and this is where everyone makes use of their drink tickets provided with their registration and queues for food – there is space for relaxation at SC as well. For Nvidia fans, I can say that Jensen Huang made a surprise appearance at the exhibition and started signing autographs and taking pictures with attendees. It is not uncommon to see popular faces like his at SC (in case you need another reason to go). 

H²RC workshop

The programme went on until Friday with a succession of talks, Birds of a Feather sessions (where advances in particular topics are presented in a more informal manner, seeking discussion with the audience), and events at the exhibition. 

After a memorable reception on Thursday night at the St Louis Science Center, the conference was wrapped up with a workshop on Friday morning. H²RC, the workshop on reconfigurable computing, is where I presented the work “Programmer productivity and performance on AMD’s AI engines: offloading Fortran intrinsics via MLIR – a case study.” 

The paper describes the first steps of a methodology to develop a Fortran compiler for the AMD NPU. We focus on Fortran intrinsics, as the intention of the programmer is explicit and these are fundamental operations in scientific codes that need to be optimised for the architecture at hand. 

Benefits of collaboration

Once again, SC was a wonderful experience. It is a constant for me that every time I go to this conference, I come back with renewed motivation and new research insights. Whilst great work can be carried out individually, events such as SC remind you that you are able to impact the community, and that collaboration between researchers aligned to the same goal is what makes science progress. Until the next SC!

Author

Gabriel Rodríguez Canal, PhD student, EPCC

Follow Gabriel's work

Fortran HLS GitLab: https://gitlab.com/cerl/fortran-hls/
GitHub: https://github.com/gabrielrodcanal
LinkedIn: https://www.linkedin.com/in/gabriel-rodriguez-canal-572604109/
ORCID ID: https://orcid.org/0009-0005-0511-3922

PhDs at EPCC

You can read about PhD opportunities at EPCC and the work of our current students on our PhD Study site.