CESGAHACK 4 will once again bring together scientists and application developers from around the world to meet for five days, aiming to accelerate their applications.
Learn how to use HPC and parallel computing while developing your own software.
About
Our goal is to help you in doing more science by speeding up your application runtime and by spending less time coding. If your development team has little or no experience of parallel programming, then CESGAHACK4 is for you.
During the week the event is structured to ensure that the majority of the time is spent working on on your own code, with the assistance of expert mentors. Participants will get access to the Finis Terrae II supercomputer at CESGA.
We will also use Parallelware Tools Suite to help you identify the parallel patterns in your code, and learn how to implement those patterns efficiently with OpenMP and/or OpenACC.
FAQ
DO I NEED PREVIOUS EXPERIENCE IN parallel PROGRAMMING?
No, you do not need any experience in parallel programming. Expert mentors will be on hand to guide you through the week and by using Parallelware Trainer you will quickly be able to identify real opportunities to parallelize your application. Our goal is for participants to spend 95% of the time in the hackathon working on their own code. Our mentors will help you to use OpenACC for use on GPUs, and/or OpenMP for multithreading and SIMD. This will include understanding what the best options are for your particular code and whether hybrid parallel programming is useful to you.
WHAT TEAMS ARE WE LOOKING FOR?
To ensure everyone is given the time needed during the event, we ask that each team has no more than 2 developers. As our goal is to help build parallel programming knowledge we are looking for teams with little or no parallel programming expertise, but where at least one of the team has significant experience with the code base that will be worked on during the event. The ideal application is one that is written in C/C++, which has hit a limitation that can be addressed with parallel programming. For example, your code might currently need runtimes that are too long to complete the science you wish to understand, or you might need to have access to more memory so that you can use larger problem sets.
WHAT IS THE SELECTION PROCESS?
Our experienced team of mentors start working with potential attendees as soon as the call for participants is opened. We work at finding candidate teams that will benefit from attending the hackathon, identifying their needs, and working with them to provide a useful application. If you have questions about whether the hackathon is for you then get in touch!
WHAT HAPPENS AFTER THE HACKATHON?
We like to stay in touch! Our goal is to establish a collaborative relationship with your team, and where possible to create success stories that can be published during the months following the hackathon. We want to increase the impact of your work!
WHAT DOES IT COST?
Participation is FREE. Complimentary tea, coffee and lunches are provided as part of the event.
WILL THERE BE PRIZES?
The jury will evaluate the participants and the winners will take a NVIDIA GeForce GTX 1080Ti! (sponsored by NVIDIA).
registration deadline
February 28, 2019
Agenda
25-29 MARCH
09:00 – 09:30 Registration
09:30 – 10:00 Opening Session
09:30 – 10:30 Welcome from the teams (Presentations of team’s projects)
10:30 – 13:30 Hands-on session: Get started & Find your hotspots.
- The parallelization process: Milestones & Progress Logs
- Including completion of knowledge survey
- Correctness & performance
- Profiling
- Using the Git Version Control System (on demand)
- Using Finis Terrae II (on demand)
13:30 – 14:30 Lunch
14:30 – 15:00 Session 1: Parallel patterns
15:00 – 18:00 Hands-on session: Identify the patterns of your hotspots
09:00 – 09:45 Session 1: Introducing parallelism with OpenACC and OpenMP
- An introduction to OpenACC and GPUs (on demand)
- Accelerating with OpenACC (on demand)
- OpenACC directives (on demand)
- Minimizing data transfers: optimize data regions (on demand)
- Introducing parallelism with OpenMP (on demand)
09:45 – 13:30 Hands-on session: Go from sequential to parallel with your hotspots.
13:30 – 14:30 Lunch
14:30 – 18:00 Hands-on session: Go from sequential to parallel with your hotspots.
09:00 – 13:30 Hands-on session
13:30 – 14:30 Lunch
14:30 – 15:30 Hands-on session
- Including preparation of 10 minute presentations updating on your progress.
15:30 – 18:00 Presentations of team’s progress.
20:30 – 22:30 Networking dinner (Cost not included: please bring personal payment method if you wish to participate)
09:00 – 13:30 Hands-on session
13:30 – 14:30 Lunch
14:30 – 18:00 Hands-on session
09:00 – 11:00 Hands-on session: Preparation of final presentations
- Including completion of event survey and knowledge survey
- Highlight your achievements and design the future roadmap
11:00 – 13:00 Presentations of team’s achievements and future roadmap
13:00 – 13:30 Presentations from staff on progress made with Q&A
13:30 – 14:00 Close the event
Explore our past events
CESGAHACK17 / CESGAHACK18 / CESGAHACK3
Read the post about experiences of the event
CESGAHACK17 / CESGAHACK18 / CESGAHACK3 / CESGAHACK4
Testimonials
“Reaching this level of work on our own it would have taken weeks of work and here we have achieved it in a couple of days.“
CESGAHACK17 participant
“We got further insight over hybrid MPI/OpenMP programming. This was achieved in a warm atmosphere with very fast and nice answers by mentors”
CESGAHACK3 participant.
“This is a very good opportunity to get away from normal working life and focus for five days; otherwise, it would take us much longer at home.”
CESGAHACK 18
“I have been able to increase the speed of the application noticeably, from 40 minutes to just 200-300 seconds.”
CESGAHACK17 participant
“The experience was very positive. I have learnt many things I must be aware of for my future programming.”
CESGAHACK3 participant
“We leave Santiago very, very happy, taking a code 47.79 times faster than we brought, thanks to CesgaHack18. “
CESGAHACK18 participant