Loading

Sample menu:

Software Acceleration on Embedded GPUs


Abstract

Chips that contain embedded multicore processors and graphical processing units (GPU) will soon become a part of every smartphone and tablet. Even though there is a significant research done on GPU for personal computers there are a number of challenges that need to be overcome when GPUs are used as a part of the embedded system. Programs for smart-phones and tablets are not written with GPU acceleration in mind and they do not utilize GPU power. If the programs are rewritten to support GPUs, this needs to be done with care so that implementation does not result in any reduction of performance, in any circumstances, compared to the implementation on processors. In addition, programmers need to take care when program GPUs so that GPU compitations do not use resources that would otherwise be required by graphical applications.

YOUi Labs are working on using embedded multicores and GPUs to provide users with better graphical user interface and in general better user experience. YOUi Labs are interested in expanding their operation and providing a way to accelerate user programs that run on these embedded platforms. In order to support that, we will develop a set of libraries in this project that would allow users to address directly GPU resources and accelerate their programs. The libraries and additional software will be written in a way that GPU still provides high speed for graphical applications while supporting user applications at the same time.

Current Status and Goals

The overall goal of this project is to allow for automated mapping of legacy Java code on embedded hardware resources

Students
Yu Wang and Iype Joseph
Professors Involved in the Project
Miodrag Bolic, Amir Rajabzadeh
Funding
NSERC with support from YouI Labs