The MediaBench server is now back up, and the files should all be correct.
Our apologies for the down time.


An extended length video input (30 second; 100x longer than original) will be posted next week.


Welcome to MediaBench!

As many of you may know, the second version of the MediaBench has been in the works for some time. Since we first decided to form an organization for the continued development and refinement of MediaBench, we have selected and developed data sets for a number of the benchmarks. More importantly though, we have continued to develop and refine the mission of this organization.

Our early goals for MediaBench, as presented at MASES 2004, were for an organization for regularly (yearly, or even quarterly) adjusting the benchmark suite and input data sets to reflect both the current state of the media industry and the recommendations of the researchers using the benchmarks. However, after reflecting upon Skadron et al.'s article [1] entitled "Challenges in Computer Architecture Evaluation", which appeared in IEEE Computer in August 2003, we realized that we needed to broaden our scope, and to provide a "forum" for benchmark development. In particular, two key points raised in the article include:

and

The major point in the first paragraph is the concept of a "forum". What we envisioned initially provided only for a small forum of invited individuals, so we intend to expand the scope to provide a web-based forum for users to communicate with each other on a regular basis. And if or when sufficient interest develops for an in-person forum for MediaBench development, we shall take the requisite steps to organize a special conference or sub-conference. Note: If anyone has any advice on how to easily create a web-based forum, please let me know...

The second paragraph both re-affirms the need for MediaBench as an open-source reasearch-oriented benchmark suite, but also argues the need for allowing researchers to improve upon the benchmarks. We have always considered that a desirable trait and have been targeting this goal, but Skadron et al.'s article now makes it an greater target, particularly considering that as we improve upon an algorithm, we continue to come closer to an (ideal) industry version of the application. Since industry rarely ever supplies the source code for an application, this is the next best way to attain applications with the characteristics of real industry solutions.

In summary, the MediaBench Consortium is making it its mission to provide a forum for developing and refining benchmark suites for multimedia systems. As such, we are depending upon researchers worldwide to assist in recommending and expanding the scope of representative benchmarks, advising on input data sets, improving the existing algorithms, porting the algorithms to available simulation and compilation environments, standardizing methods for performance analysis, and in general maximizing the utility and representativeness of the benchmark suite with respect to the multimedia industry. In short, the success of MediaBench is in your hands.

- Jason Fritts
Director of the MediaBench Consortium

[1] Kevin Skadron, Margaret Martonosi, David I. August, Mark D. Hill, David J. Lilja, and Vijay S. Pai, "Challenges in Computer Architecture Evaluation," IEEE Computer, 36(8):30-36, August 2003.