EPICURE meets Slovenia
16/05/2025

EPICURE: Empowering European Researchers with Optimised HPC

What is it like to use a Supercomputer?

 

For all those who have used supercomputers, you may have noticed that these powerful, expensive, very large computers can be tough. Really tough. You want to do research, obtain results, write papers or run your business, but trying to persuade these powerful, cutting-edge wonders of engineering to do what you want can be very frustrating. And what’s worse, the technology rushes ahead, even before you have finished reading the documentation.

 

 

What’s the difficulty?

 

For those who have persevered, doing something on a supercomputer starts to become possible. Enjoyable, even. But here is the difficult bit: getting your program, workflow or AI model to run really well on the big computer can be very challenging.  Why is this the case? HPC systems are complex, with many different types of processors and devices which must interact via multiple network connections, using disk spaces and memories with a variety of speeds and capacities which vary considerably. And then there is the software – exactly how does this ancient FORTRAN 77 code (that evolved rather than being designed) work? And not to mention the batch scheduler.

 

 

Getting a bit nervous … is code optimisation really necessary?

 

Your computer budget is limited, and there is a lot of competition for resources. With a fast, efficient, and optimised workflow, you can do more things in more detail for longer or run larger models. You can pull ahead of the opposition (and save money and energy).

 

 

Ok, I am convinced. Tell me more.

 

The EPICURE code optimisation teams can help. A diverse staff of HPC specialists from all over Europe, with experience in all the hardware architectures available on the continent.  Between them, they can program in MPI, OpenMP, CUDA, and HIP – if it is used in Europe, someone in the EPICURE staff will know it. Advanced profiling for pinpointing bottlenecks, experts in disk usage and data transfermulti-node scaling for AIporting your code to ARM or GPUs, and parallelising obscure FORTRAN– just ask! Finally, software does not exist without a reason, so don’t worry if your field is astrophysics, theoretical chemistry, CFD or Large Language Models since our teams come from diverse disciplines and are likely to understand what you wish to do.

 

 

Ok, where do I sign up?

 

Get in touch with us (https://epicure-hpc.eu/support-services/)  or check the website (https://epicure-hpc.eu/) for more details on the (easy) procedure.

 

 

Let the EPICURE code optimisation teams do the hard stuff so that you can do the interesting stuff!

 

 

 

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *