On Thu, May 26, 2011 at 11:39:42AM -0400, Bruce Walker wrote: > At the time you were probably wondering what *possible* use there > could ever be for such an esoteric concept. ;-)
The first time I encountered FFTs was when they were used as an example of how to rearrange an algorithm for parallel processing (useful in achieving the first "F"). [Warning - seriously geeky talk ahead. You might want to skip it] Later on (some 25 years ago, now) I played with them in more detail when writing image filter kernels. While it may not seem impressive today with gigabytes of RAM and 1024-processor GPUs, being able to do an FFT of a 256x256 image was quite something at the time. I even managed to tighten up the inner loop on the Apollo DN10000 (a four- way parallel RISC machine) to achieve the maximum theoretical speed, with no stall cycles in the pipeline. No big deal, but it did get me bragging rights - the best anyone else managed was one I/O stall every 14 cycles because of register contention. -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net to UNSUBSCRIBE from the PDML, please visit the link directly above and follow the directions.
