What was the topic of your project?
The topic of my MSc dissertation is the GPU implementation of iterative solvers in numerical weather predicting models (NWP).
Which external company/organisation did you work with?
It was co-supervised by Xu Guo from EPCC and Eike Mueller from the University of Bath, who currently is involved in the UK Met Office project GungHo (Globally Uniform Next Generation Highly Optimized), which is a project for developing a new dynamical core suitable for next generation weather and climate simulation. In their project, researches currently are basically focused on massive parallel CPUs while my dissertation is like an extension to their project, which is focused on the same topic but with GPU implementations. Thus my dissertation becomes an alternative or complement to their large project.
How successful was the project?
The success of my dissertation project is based on both fully understanding the behaviour of GPU devices and of characteristics of the dynamical core governing PDEs and its iterative solvers. To be more precise, the aim of my project is to find an optimal solution which is suitable for GPU to solve such an enormous amount unknown equation with such characteristics.
What was the most challenging part of the project?
The above two aspects were the main challenges of my project, as well as the parts where I got the most fun. My supervisors gave me the most helpful guides and heuristic suggestions during the period of my dissertation. Xu is the expert in parallel computing and GPU implementation, while Eike is a mathematician. With their nice help while I’m facing these challenges, the result of my dissertation is quite exiting.
How did undertaking this project with an external company/organisation benefit you?
According to my experience last year, I can say that there are many benefits of getting involved in an external project.
First, the external people have a very clear idea about the project including methods, aims and so on, with the expertise in the field needed. So my questions can be settled very quickly and precisely.
Second, the external supervisors, who may have very different background with computing, bring a lot of knowledge in their fields; in my case, Eike brings me a lot of knowledge in numerical algorithms and climate dynamics which are very interesting to me.
The last benefit which I think is the most important benefit to me, is one can have more opportunities of participating in various activities of scientific research, for example I got a session in Bath numerical analysis seminar to present my results, and the paper about the topic of my dissertation is also about to be published. These various experiences are the most vital things for a ‘newcomer’ in scientific research field.