Overview
Developers around the world were challenged to create an innovative application that best takes advantage of the AMD Fusion architecture using the AMD APP SDK and OpenCL™ instructions.
One of the main features of the APU is that the GPU can be used to offload general compute tasks from the CPU with very low overhead since they are on the same die. So, applications that require a lot of data transfer between CPU and GPU really shine on the AMD Fusion Architecture.
We got submissions in a wide range of categories including image, video and signal processing, computer vision, simulations, financial modeling, machine learning, artificial intelligence, bioinformatics, gaming physics, and computer security.
The Winning Submissions
1st Place: Ghost Rider
Do you ever feel like the computer opponent has an advantage on first person shooter games? This application uses OpenCL for real-time image processing of images from a first person shooter game. The image processing performs feature extraction and matching so that the computer can use the same visual input that humans do when playing the game.
2nd Place: Real-time Video Effects with AMD & Kinect
This application creates smooth real-time video effects based on gesture recognition input from the Kinect box. One example is, the user can stand in front of the Kinect camera, reach their hand forward and make a tapping gesture. The video on the computer will show a real-time water rippling effect radiating from the tapped point.
3rd Place: Semi-supervised Learning-Based Method for Adaptive Shadow Detection
Typically, vision-based systems remove the background of an image in order to identify the objects of interest, however shadows can distort those objects. For accurate object shapes, it is important to remove shadows. This usually requires human input. This application takes a new approach and only requires human input for initial training, then adaptively learns where the shadows are and removes them.
4th Place: Numerical Simulation of an X-Ray Generator
This program does the computation and displays results of a numerical simulation of an X-Ray generator using an innovative merge of OpenCL and OpenGL.
Judging Criteria
The winners were chosen based on these criteria:
- Best use of AMD Fusion Architecture
- Novelty, creativity and innovation
- Algorithm and implementation efficiency and processor utilization
- Quality of software engineering
- Commercial and Open Source Potential
- Magnitude of effort
Quotes from the Competitors
Many of the competitors had never worked with OpenCL before. Here’s what some of them had to say:
“Honestly, I am new to OpenCL, except having a look at couple of CUDA programs before and little experience with OpenMP… It turned out to be not much difficult.”
“The videos and Webinar Series at AMD OpenCL Site are really helpful to get very quick understanding of OpenCL. After you got basic understanding of OpenCL, the best thing is to start playing with OpenCL”
“Thank you for ability to write my first OpenCL program with fun. Practical learning of OpenCL may be out of my plans a long time without this contest.”
“I started coding GPGPU on CUDA, but for now on CUDA will always be on the second place=)”
“Interesting that so many of us were brand new to OpenCL (I guess this is the aim of the contest from AMD's perspective too). Something about the competition format really motivates me to investigate and learn new stuff so I'm very grateful for the chance”
“I'm not new to OpenCL, but was very surprised how you all got so much out of it”