KAUST-NVIDIA Hackathon 2022

Accelerating Scientific Applications using GPUs

What is the KAUST-NVIDIA Hackathon?

As its name suggests, it’s an event organized by KAUST and NVIDIA aiming to accelerate scientific applications. The hackathon started as part of the KAUST-NVIDIA workshop in 2015. In 2022, Deep Imaging Group members participated in the 8th competition.

The Notorious Team

The notorious team of this year was formed by 3 amazing people: From left to right, Nick Luiken (Postdoc), Juan Romero (Ms. Student), and the writer of this post Miguel Corrales (PhD Student). If you want to know about everyone’s research, please visit the attached links.

Our approach

The problem to solve lies on Juan’s research . The main challenge is related to a segmentation step, where many independent bisection problems are needed to solve, which makes a perfect setting to optimize the running time performance using GPUs.

Results

After one day and a half, the team succesfully levaraged the capabilities of the algorithm. Roughly speaking, the team speed up the computation by 80 times. The code of the hackathon could be find in the followingrepository

Insights and comments

It was such a wonderful experience to participate in this event. Many thanks to the organizers from KAUST and NVIDIA, and of course, for all the support from the mentors who constantly help along this short journey. Luckily, the team was awarded the best team in terms of computing performance.

Illustrative experience

This was a post regarding the KAUST-NVIDIA Hackathon, here it is a illustrative summary related to our teamwork.

“It’s gonna be easy, probably just putting our code on numba and cuda”, that is what we thought