I’m master student of computer science. About 1 - 1,5 year ago I started learning about GPGPU, OpenCL, parallel programming etc. I like it, I’m interested in it, I want to do this.
Recently I defended my Bachelor’s degree thesis. Basically I adjust algorithm in a way to execute it in parallel and implement it (console application, C++11, Boost). I used OpenCL (with C++ wrapper) to perform calculations on GPU. The results were quite good. I can describe it in details if you like.
Now I’m searching for a master’s thesis topic related to OpenCL, GPGPU, mobile computations, etc. Unfortunately I cant get any interesting proposition from lecturers I know and I don’t to limit myself with thesis that is not about what I really like or like the most, so I’m looking outside my university.
Perhaps someone could give me some suggestions about topic or where to find someone who might help me.
Maybe there is a need for some research, benchmarks regarding memory, especially now with Shared Virtual Memory. Maybe there is an idea to use capabilities introduced/enhanced in OpenCL 2.0 (pipes, dynamic parallelism, etc.) to make a next step in optimization some algorithms. Maybe there is something you/your company/research team would like to check/solve/research/develop about GPGPU, mobile GPU acceleration, or there is a need to rewrite algorithm to OpenCL (although it’s master’s thesis it cant be to simple).
As I said I will be grateful for any help, suggestions, comments 
Thanks a lot in advance & Best Regards.
How far into the degree are you? The reason I ask is because the search for the topic is actually a lengthy process and it’s supposed to be part of your work during a master’s degree. When you apply for a master’s you usually select a broad research area - in your case that would be parallel computing. After you’ve completed the course requirement you would then spend 1-2 semester doing what’s called a literature review, which is basically reading a ton of research papers. During this phase you learn about the major achievements and the current hot topics which should help you narrow down your topic. For example, I started with an interest in natural language processing. During the literature review I considered text summarization, question-answering and finally landed on sentiment analysis. I then focused my readings on that topic and narrowed it down even further to aspect-based sentiment analysis.
Your advisor should be helping you through all of this as well. Have you gotten any input from him?
Well. I’m from Poland. I had 3,5 year of Bachelor’s degree studies and recently I started my Master’s degree studies and I have 3 semesters including current one. What you’re saying sound great, but here on master studies the procedure sadly look like this: get the advisor (find by yourself or there’s a list with topic/research area), he gives you your ms degree topic (more or less), do it, be happy. Well I couldn’t find anyone with topic which would interest me (well I have one but in other area - heuristic). I’m planing to defend my thesis in June\July 2016 (worst case scenario: September 2016). I’m thinking about spending the rest of this semester and summer break for reviewing/reading/etc and next 2 semesters for doing.
So I don’t have advisor and currently I’m searching topic on my own. If it doesnt work out I’ll try heuristic alg.
I suggest DNN - Deep Neural Network. This is a hot topic these days, and there are various accelerated applications which start to appear - here’s an example: http://on-demand.gputechconf.com/gtc/2014/presentations/S4651-deep-learning-meets-heterogeneous-computing.pdf
NVIDIA recently released cuDNN (http://devblogs.nvidia.com/parallelforall/accelerate-machine-learning-cudnn-deep-neural-network-library/) - Implementing this functionality in OpenCL and comparing between the two might be a good subject for a thesis.