Divvying up the Work

If you want to split up the work of video encoding across multiple processors, you've got a bit of puzzle on your hands. There are two common approaches: divide the data into blocks and fully process each block with one processor; or divide the algorithm into processing blocks and assign each block to a processor. Neither approach prevents data starvation or excessive processor idle time, and neither approach scales well. Laurent Bonetto, Ram Natarajan, and Dr. R K Singh of Cradle Technologies describe an approach to the problem that tries to combine these two techniques for an optimized solution.

Posted by Kevin Carlson at 12:12 PM  Permalink