back in march 1996, while roaming the early internet in search of mathematics, i rather quickly came across GIMPS. at the time i was working in advanced development on philips speech recognition and had access to 52 workstation computers. whenever they were not needed for speech recognition tests, i was allowed to use them with the explicit approval of my boss, who was secretly hoping for some results too :-) a colleague had implemented a rather sophisticated queuing system for this, so we could make good use of every free cpu cycle.
i believe i was about the eighth person (!) to contribute to GIMPS at all, and with my 52 cpus i even made it into the worldwide top 10 contributors for a while. after moving from speech recognition to astronomy in 2008, i no longer had access to that kind of computing power, but i still contributed the occasional factorization to help eliminate candidates, though.
today, more than 3 million gpus and cpus are involved, and GIMPS has long since set numerous world records in computational mathematics. as far as i know, george woltman's fft implementation is still hard to beat.
| useful links |
|---|
| american mathematical society |
| gimps |
| matlab |
| narrow admissible tuples |