Nicolas Alvarez <nic### [at] gmailisthebestcom> wrote:
> How big are the input files needed to render a single frame? (that is,
> the output of the simulation, input to pov)

The generated datas by the simulation tooks 1.6 Go on my hard disk for the 95
frames.

Regards,
Fidos.

Post a reply to this message

"Paean" <Pae### [at] gmailcom> wrote:
> amazing.
>
>
> Hi, fidos, Would you like tell something on the simulation techniques you used
> in this video? Did u use partilce methods?
> Thank you.
>

I use the particule level set method.

The algorithm is very simple : The particules are populated and advected with
the fluid as describe by Douglas Enright (paper "A Hybrid Particle Level Set
Method for Improved Interface Capturing"). When a particule escaped from the
fluid, it is animated as a free particule (no influence of the speed of the
fluid). When it touch the fluid, it is removed.

The particules rendered in PovRay are the excaped particules. The surface of the
liquid is rendered with the levelset after polygonisation.

Regards,
Fidos.

Post a reply to this message

> Nicolas Alvarez <nic### [at] gmailisthebestcom> wrote:
>> How big are the input files needed to render a single frame? (that is,
>> the output of the simulation, input to pov)
> 
> The generated datas by the simulation tooks 1.6 Go on my hard disk for the 95
> frames.

Aw well... then I don't think I can offer you my renderfarm. That's too 
big...

Post a reply to this message

Fidos,

I have been a fan of your animations since about your 10th post. Your use of
solvers in the animations are incredible.

I am an engineering graduate who did CFD in uni, unfortunately I have lost most
of my notes and don't have any data on implicit solvers, they are notoriously
hard to come by on the internet. I need a one dimensional solver with
compressible properties in order to solve for fluid flow in compressed air
lines in real time. I am working on a ring main type airline with multiple take
off points, and want to justify reducing the size of pipe for the next
installation.

Code in matlab or POV code for a simple implicit solver would be good to get me
started. Who knows, I might be able to animate the results and post them here
in the future.

James

"fidos" <fid### [at] wanadoofr> wrote:
> Hello,
>
> With the same simulation results of my last post in the group, I used PovRay 3.7
> beta23 to try to do a multi CPU rendering on a Intel Q6600.
>
> PovRay 3.7 is very impressive in speed. Each frame render in around 20 min
> (there is 95 frames in total).
>
> The video is MPEG1 encoded.
>
> http://pagesperso-orange.fr/fidos/
>
> Any comments are wellcome.
>
> Regards,
> Fidos.

Post a reply to this message

"fidos" <fid### [at] wanadoofr> wrote:
> Hello,
>
> With the same simulation results of my last post in the group, I used PovRay 3.7
> beta23 to try to do a multi CPU rendering on a Intel Q6600.

Beautiful!  If you moved the camera along a spline, added some mood lighting,
and hid the source behind something -- I know you can't really move it out of
the rectangle -- this would be even more impressive.

To ask the obvious, isn't animation-rendering embarrassingly parallel?  Can't
you just render four frames at a time?  I just finished parsing and rendering
about 80+ GB of data on a Q6600 last night.  Just split up the files into three
directories. and ran them separately.

Wish someone would assign me a project like this.  Then I could justify spending
time on it.  Apparently Roe schemes and discontinuous Galerkin are more
interesting.  So I'm just milking last year's CFD data for all it's worth
instead.  Oh well.  Some day.

Random thoughts.  But very nice.

 - Ricky

Post a reply to this message

> "fidos" <fid### [at] wanadoofr> wrote:
>> Hello,
>>
>> With the same simulation results of my last post in the group, I used PovRay 3.7
>> beta23 to try to do a multi CPU rendering on a Intel Q6600.
> 
> Beautiful!  If you moved the camera along a spline, added some mood lighting,
> and hid the source behind something -- I know you can't really move it out of
> the rectangle -- this would be even more impressive.
> 
> To ask the obvious, isn't animation-rendering embarrassingly parallel?  Can't
> you just render four frames at a time?  I just finished parsing and rendering
> about 80+ GB of data on a Q6600 last night.  Just split up the files into three
> directories. and ran them separately.

I offered him my renderfarm, but his input data is way too big to 
transfer over Internet.

Post a reply to this message

Nicolas Alvarez <nic### [at] gmailisthebestcom> wrote:

> I offered him my renderfarm, but his input data is way too big to
> transfer over Internet.

That's unfortunate because his simulations always leave me wanting that extra
few seconds.  My solution for now is that I think I will set up a script:
Crunch-parse-render-repeat.  Something like a screensaver.  Right now I run out
of space long before I run out of time.  It took ~10-15 hours to accumulate 80GB
of data, although that's the full double-precision velocity field.  I'd be
willing to run something on my home computer, but at 384 MB RAM and 867 MHz, I
don't know how far it would get.  Perhaps if the renderfarm were used for both
calculating and rendering, only an occasional restart file and images would
need to be transfered.  Not that I'm offering your resources, of course.
Excellent work though.  It'd be a shame for it to go unrendered.

 - Ricky

Post a reply to this message

wow!

hey, that place is from JVP's scenes, isn't it?  flooded... :)

so, a core 2 quad, huh?  That explains you interest in implementing path tracing
in povray and the relatively fast render times of the tests.  If I had this much
power in my hands, I too would like to see the CPU sweating -- all 4 cores! :)

Post a reply to this message

> To ask the obvious, isn't animation-rendering embarrassingly parallel?  Can't
> you just render four frames at a time?  I just finished parsing and rendering
> about 80+ GB of data on a Q6600 last night.  Just split up the files into three
> directories. and ran them separately.
When I can't use PovRay 3.7, I do the same : 4 instance of MegaPov running at
the same time (but in the same directory). I divide the frame range in 4 and
start the 4 renderings at 4 different positions.

Fidos

Post a reply to this message

"nemesis" <nam### [at] gmailcom> wrote:
> hey, that place is from JVP's scenes, isn't it?  flooded... :)
You are right, I gave credit in this post :
http://news.povray.org/web.4764e52fc1d3aa61423caea0%40news.povray.org

> so, a core 2 quad, huh?  That explains you interest in implementing path tracing
> in povray and the relatively fast render times of the tests.  If I had this much
> power in my hands, I too would like to see the CPU sweating -- all 4 cores! :)
It's clear that without 4 cores, I wouldn't start to play with path tracing !

Post a reply to this message