Hey I agree, when we are comparing accuracy RK4 has it's merits. But right now I am visualizing the means, not hurrying towards the ends (convergence). I want *small* step sizes so we get that nice 60fps and less look, while holding true to actual physics. This library is meant to be built into websites, games... maybe voxel-engine<https://github.com/maxogden/voxel-engine>? Of course with algorithms like RK4 we can converge faster with larger steps, but it is still more expensive per *small* *dt *then a simple forward Euler. Don't get me wrong, I have coded all of these in C for homework over the years... I just want physics based waves, ripples and diffusion. Right now I don't see a better approach than the least computationally expensive (and solving them with simple convolution operators is easier).... which is all pde-engine<https://github.com/bpostlethwaite/pde-engine.git>is offering. That said, until I get some benchmarks up this is all hypothesis. By the way, if you want to prove me wrong, hack on the code and show me your Science skillz yo! ;)
---ben On Friday, 25 January 2013 23:15:06 UTC-8, jesusabdullah wrote: > > > Actually, I think it's likely to be (much) faster for the same accuracy, > since you can use a bigger step size. > > Yes, in order to compare their speeds you need to compare on the same > accuracy scale and not on the same step size. Forward Euler requires a > comparatively small step size to get anywhere near the accuracy of > RK4. :) > > > But of course, it's a little bit more programming. > > A little bit, but you should be able to find "the equations" on wikipedia, > no? > > --Josh > > On Sat, Jan 26, 2013 at 12:10 AM, Harald Hanche-Olsen > <[email protected] <javascript:>> wrote: > > [Postlethwaite <[email protected] <javascript:>> (2013-01-25 17:29:52 > UTC)] > > > >> > The forward Euler step is employed for simplicity and speed > >> > Please please please use RK4 instead of foward Euler, it's way way > >> > more accurate. https://en.wikipedia.org/wiki/RK4 > >> > >> > >> Ha, of course I should be using RK4! But it's slower [...] > > > > Actually, I think it's likely to be (much) faster for the same > > accuracy, since you can use a bigger step size. But of course, it's a > > little bit more programming. > > > >> I mean, you have to understand that for many problems in scientific > >> > computing, the matrices are large enough that you really want to bust > >> > out C and fortran libraries. > > > > Indeed. But scientific computing needs vary a lot, from the stuff you > > do with a calculator and the back of an envelope, to the really hard > > problems that require supercomputers. And as hardware (and javascript) > > gets faster, the boundaries keeps moving, and you can do more with > > simple means as time passes. > > > > - Harald > > > > -- > > -- > > Job Board: http://jobs.nodejs.org/ > > Posting guidelines: > https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines > > You received this message because you are subscribed to the Google > > Groups "nodejs" group. > > To post to this group, send email to [email protected]<javascript:> > > To unsubscribe from this group, send email to > > [email protected] <javascript:> > > For more options, visit this group at > > http://groups.google.com/group/nodejs?hl=en?hl=en > > > > > > > -- -- Job Board: http://jobs.nodejs.org/ Posting guidelines: https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines You received this message because you are subscribed to the Google Groups "nodejs" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/nodejs?hl=en?hl=en
