GPU cesgahack 18Appentra Solutions S.L., and CESGA, are delighted to announce the participants in the upcoming CESGAHACK: a hackathon taking place at CESGA, Santiago de Compostela on 5th-9th March 2018.

We received many fantastic applications for the hackathon, but sadly we have to limit the experience to just 6 teams. The successful teams cover a range of scientific applications and also a range of experiences.

The Participating Teams:


A previous participant of the the CESGAHACK hackathons, the team from the EDANYA group at the University of Málaga are developing a family of finite volume numerical solvers named as HySEA for the simulation of geophysical flows including tsunamis generated by earthquakes or landslides, river floodings, sediment transport, turbidity currents, ocean and coastal currents, etc.. A new code of this family is developing for the simulation of ocean currents in areas like the Strait of Gibraltar. The actual version of the code is written in C/C++ and comprises around 1000 lines of code.


The MADWAVE3 code is used for studying quantum dynamics of reactions A+BC –> AB+C. Currently parallelized using MPI and heavily reliant on 2D-FFTs. The team is currently working on developing code for performing quantum calculations for 4 atom reactions: a feat not currently possible in most codes in this domain. The results of this work are used to obtain reaction rates for astrophysics, where quantum effects are important because of the low temperatures (approx. 10K) and the high abundance of hydrogen atoms.

PORTA (POlarized Radiative TrAnsfer)

The PORTA code developed in our group is used for modeling the generation and transfer of polarized radiation (caused by scattering processes and the Hanle and Zeeman effects) in 3D models of stellar atmospheres. As part of the POLMAG project (financed with an European Research Council Advanced Grant) we plan to extend PORTA with what is called as Partial Redistribution (PRD). This increases tremendously the computational cost, so we are interested in porting the code (currently parallelized with MPI) to make use GPUs.


Fast Thermalization is a numerical General Relativity solver for Einstein’s equation in Anti de Sitter space. At approximately 100,000 lines of C code, this will be the largest code base that any group has brought to the CESGAHACK experience. The code is already parallelized using OpenMP and runs from live-demos on a laptop onto the MareNostrum supercomputer at Barcelona Supercomputing Centre with 48 cores per node.


The Simple Convolutional Neural Network Library aims to provide an easy to read and easy to use convolutional neural network library written in C that is able to understand digits from the MNIST dataset that is used extensively in training both image processing and machine learning tools. The code is approximately 1000 lines of C++ code, and currently takes just over 4 minutes to train the network on a i5-5300 computer. They hope to get this down to just 1 minute.


The I4Scheduling application, developed by a team from both Polytechnic Institute of Cávado and Ave, and Minho University, is a sequential C/C++ software application for simulating industrial production scheduling algorithms. The application explores

multiple scheduling algorithms that can be used to help industry optimize their manufacturing processes. The number of simulations is therefore key, but limited by the performance of the code. The parallelization of this code will open the simulation to multiple environment scenarios and enable a faster and more informed decision on production scheduling.

We are looking forward to working with all of these teams in March!

Full details of our hackathons are available  here.