More than 80 attendees participated in the two WORKSHOPS “PARALLEL COMPUTING AND GPU COMPUTING” organized by SIE in collaboration with NDIVIA.
On May 5, the Hotel Villa de Alcobendas in Madrid was the setting for the course “PARALLEL COMPUTING AND GPU COMPUTING”, in Barcelona we were the next day, May 6, at the Hotel Villa Emilia.
The use of GPU's for scientific calculation
In both events the interest shown by the researchers was enormous and we immediately had to open a waiting list. Such was the demand that we enabled the broadcast of the event on our YouTube channel. At the end of this post you can play the full video of the event, or if you want you can visit the European Computer Systems Youtube channel.
Photograph of one of the slides of the presentation in which you can see information about Pascal GPU featuring NVLINK and Stacked Memory
Scientific computing based on GPUs, specifically CUDA, is growing more and more, due to the cost savings that this implies, as well as the lower space and consumption that it allows.
Summary of Carlo Nardone's presentation
Sistemas Informáticos Europeos is putting a lot of interest in these solutions, and for us it was a privilege to have Carlo Nardone, Senior Solution Architect of NVIDIA, speaker of the courses, who offered us an intense and very useful program.
In this photo we can see Carlo Nardone senior technical specialist at NVIDIA
SIE representatives attended Raúl Díaz, Project Manager, David Ramírez, Systems Manager and Belén Moreno, from the Marketing department. Rafael Lainez and Francisco Gómez from AZKEN and Javier Perez from IECISA also attended the event. During lunch the attendees were able to exchange experiences from their different departments, and in Madrid we celebrated 25 years of SIE with a cake.
Here is a complete list of the documents that have been used in Carlo Nardone's presentation:
SIS presentation
Introduction to Parallel and GPU computing
Introduction to GPU programming
CUDA programming model
Introduction to libraries and the CUDA ecosystem
CUDA Streams approach for optimization (kernels and memory transfers overlap
Some features introduced with CUDA 6
New features with CUDA 7 and beyond
Introduction to OpenACC
Advanced features of OpenACC
Very detailed prisoner on GPU optimization techniques:
Introduction to CUDA profiling tools
CUDA features for multi-GPU programming
Thrust library for C++:
PyCUDA open source Python module:
Numba Pro Python for GPUs:
R on GPU:
ArrayFire GPU library:
HPC Admintech meeting 2016
The parallel computing Workshop also in Barcelona
The event was repeated in Barcelona on May 6, where we were able to meet many researchers interested in Carlo Nardone's presentation. You can follow the full video of the first Workshop, with a presentation by Carlo Nardone, in the following video.
In Barcelona to round off the day, Carlo Cardone was able to cheer on Juventus in the broadcast of the match against Madrid and enjoy a tour of Barcelona.