What you mean to say Jim is that at many occasions most software is using O ( n ^ 2 ) for example to sort polygons instead of doing it O ( n log n ) as things are more than fast enough, or they use a real slow C++ way of inserting objects, which is easy to program yet of course loses you another factor 100 as it's real slow allocating and deallocating objects.
No problem for most scenes. I did not put in money to remove a bunch of O ( n ^ 2 ) problems either in the 3d engine. It's fast enough realtime! Still it's faster there for the same high quality high polygon animations than the software that gets actually used onto the cpu cores (without much of a SIMD optimization yet obviously). Know what you speak about if you speak about graphics design... Someone who knows something about 'algorithms' and knows how to spell that word - you can really speedup things there. Vincent On Nov 7, 2012, at 4:08 PM, Lux, Jim (337C) wrote: > >> >> Not quite. Articles typically mention that the time to render a >> frame >> of film has remainted pretty constant over the last decade, any speed >> increases in the hardware being used to execute more complex >> algorithms. The move from 24fps to 48fps (ie The Hobbitt) isn't >> going >> to help matters either. >> > > I suspect that the "render time" requirement is driven by workflow, > rather > than horsepower. That is, it needs to be "fast enough", and they > either > buy more hardware to make it faster or accept lower quality, > depending on > their budget and cash position. > > Budgets probably are done (at the inception of the project) in > terms of "# > of bodies working for # of months" and the manager gets to trade > dollars > on hardware vs dollars on bodies, but the delivery date is essentially > fixed. > > Like drag racing, it's "how fast can you afford to go" > > >> > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin > Computing > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
