PS. James graciously sent me a copy of his code privately, and I can verify that it is doing what he claims. This looks like a very impressive result, and an important advance for the complexity theory of FFT algorithms: breaking the 36-year-old record for the minimal flop count in power-of-two sizes, from 1160 to 1152 flops for n=64. I hope he publishes his algorithm in the near future, and encourage him to post his code (gzipped it is < 10k). PPS. I should caution other readers that a small change in the flop count alone is unlikely to significantly impact FFT speed, which is dominated these days by cache and pipeline utilization. But the question of minimal arithmetic complexity is still important on theoretical grounds.
Large FFT --- NOT radix-2
Started by ●January 1, 2004
Reply by ●January 23, 20042004-01-23
Reply by ●January 23, 20042004-01-23
"Steven G. Johnson" <stevenj@alum.mit.edu> wrote in message news:bus6sq$8ib$1@news.fas.harvard.edu...> I hope > he publishes his algorithm in the near future, and encourage him to post > his code (gzipped it is < 10k).Perhaps if you have a little corner on your website reserved for miscellaneous stuff, you could place it there and post a link. Seems to me to be the preferred method for communicating largish chunks of data in newsgroups, if it's not too much trouble from your viewpoint. -- write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, & 6.0134700243160014d-154/),(/'x'/)); end
Reply by ●January 23, 20042004-01-23
James Van Buskirk wrote:> "Steven G. Johnson" <stevenj@alum.mit.edu> wrote in message > news:bus6sq$8ib$1@news.fas.harvard.edu... > > >> I hope >>he publishes his algorithm in the near future, and encourage him to post >>his code (gzipped it is < 10k). > > > Perhaps if you have a little corner on your website reserved for > miscellaneous stuff, you could place it there and post a link. Seems > to me to be the preferred method for communicating largish chunks > of data in newsgroups, if it's not too much trouble from your > viewpoint.I'll be happy to host it. Even with my slow dial-up connection, it won't be much trouble. Check out http://users.rcn.com/jyavins to see if it would be appropriate. It could replace Jeff Briden's simulation that is old hat by now, have a line of its own, or be anonymous like http://users.rcn.com/jyavins/walala-lena.jpg. Jerry P.S. If you look at the site, please tell me what you think of my new favicon. It's a reproduction of the stamp I put on the jewelry I make. -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●January 24, 20042004-01-24
"Jerry Avins" <jya@ieee.org> wrote in message news:4011d483$0$12745$61fed72c@news.rcn.com...> I'll be happy to host it. Even with my slow dial-up connection, it won't > be much trouble. Check out http://users.rcn.com/jyavins to see if it > would be appropriate. It could replace Jeff Briden's simulation that is > old hat by now, have a line of its own, or be anonymous like > http://users.rcn.com/jyavins/walala-lena.jpg.Thanks for the offer; cool website. I finally got around to figuring out how to create a personal web page through my ISP. http://home.comcast.net/~kmbtib/index.html I uploaded the files. If there are any questions, post them here or to comp.lang.fortran, or decode my email address... -- write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, & 6.0134700243160014d-154/),(/'x'/)); end
Reply by ●January 24, 20042004-01-24
James Van Buskirk wrote:> ... cool website. I finally got around to figuring > out how to create a personal web page through my ISP. > > http://home.comcast.net/~kmbtib/index.html > > I uploaded the files. If there are any questions, post them here > or to comp.lang.fortran, or decode my email address...Nifty code! My site started as a way to show my face to a a stranger I had agreed to meet at the airport. The rest grew by accretion. CuteFTP and CuteHTML made it easy. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●January 25, 20042004-01-25
"Jerry Avins" <jya@ieee.org> wrote in message news:4012c8ef$0$7333$61fed72c@news.rcn.com...> James Van Buskirk wrote:> > ... cool website. I finally got around to figuring > > out how to create a personal web page through my ISP.> > http://home.comcast.net/~kmbtib/index.html> > I uploaded the files. If there are any questions, post them here > > or to comp.lang.fortran, or decode my email address...> Nifty code! My site started as a way to show my face to a a stranger I > had agreed to meet at the airport. The rest grew by accretion. CuteFTP > and CuteHTML made it easy.Well, I got back from my walk in the park and concluded that maybe I should put some descriptions on my website of what the files are and also, to be topical to the original thread (now that the subject line has changed, of course :) ) I zipped up my code for all prime power orders <= 64, so it includes the factors 25 and 49, although as we have seen, this old code is now known not to be optimal. -- write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, & 6.0134700243160014d-154/),(/'x'/)); end
