[LAD] opencl benchmark results

Hi, For those of you who are interested real world results for ffmpeg with opencl we have run some benchmarking tests across our cluster. At the end of last year there was a flurry of activity integrating opencl with ffmpeg. The work was undertaken by multicorewareinc who is also the official keeper of the code for AMD's proprietary version of ffmpeg. We have found that leveraging opencl with the latest development version of ffmpeg gives us upto 4566 DP-MFLOP on a single machine. Adding in the OpenCL component (GPUs) gives us around a 64% speed boost on AMD A4's. We got upto 400% increase on our intel machines (with alot of NVidia CUDA cores). We also noticed that essentially the same AMD/ATI card on a different bus (AGP-8 versus PCI-E) only picked up an extra 6% in speed. So clearly the recent work on the opencl integration with ffmpeg has provided some real world dividends. Just for the record our local cluster is clocking in around 20GFlops with the opencl code in the mix. We haven't added in the cloud servers though so that number is low. While we are not a not a threat to the big players, combining resources within the wider open source community creates a very serious competitor that shouldn't be ignored. Cheers -- Patrick Shirkey Boost Hardware Ltd