Okay.  This is my final attempt here.  I changed my email address, but for
some reason the message kept it.  No doubt my mistake.  If my email address
still stays, so does this message.  Very sorry to be making a mess of this
fine newsgroup.  Really though, I've slipped up here or there, but I don't
think I've received one piece of spam in about nine months with my new
email address.  So yes, in fact, it is worth it.  But here's the synopsis
of the original message:

Here's my attempt at the 'Visual Simulation of Smoke' paper (google it!).  I
have just about everything in place - vorticity confinement, a conjugate
gradient solver, the semi-lagrangian scheme with cubic interpolation, but
the obvious next step is to add a third dimension.  This shouldn't be too
difficult since everything is completely analogous.  In fact, this started
out as three dimensional, but it was too difficult and slow to debug, so I
reduced it to two.  The latest version adds a lot more detail, but the mpeg
compression still causes it to lose a fair amount.  Oh well, it will look
better in THREE DIMENSIONS!!  (read with thunderous voice).  Sorry, I'm
looking forward to a third dimension though.  I'll post the code when it's
in pretty good shape, but it's not too pretty since my c++ experience is
basically limited to what you see here.

As a reply to a reply to my previous message, sorry it's not in sdl, but 1)
that would take months in POV alone and 2) I don't have the capability to
patch POV.  I can't even get it to compile on my own.  Maybe some day.

 - Ricky

Post a reply to this message

Attachments:
Download 'fluid.mpg' (589 KB)

> Okay.  This is my final attempt here.  I changed my email address, but for
> some reason the message kept it.  No doubt my mistake.  If my email
address
> still stays, so does this message.

To keep you pacified: I've been posting to this newsgroup for years. Though
I've recently changed my email-adress due to SPAM, I believe that the
newsgroups here are pretty safe, mainly because it's not USENET and thus not
as widespread. Additionally, I believe that Websites are worse for
email-harvesting, I've made the mistake and had an email-adress of mine on
my website for three days, and right then I got 10 SPAM-Email, but it's
quiet again now. When I had my email-adress on these newsgroups by mistake,
nothing happened.

Regarding the animation: it does look fine and I sure hope you'll be
successful at a 3D-Version. How will you make it possible to have POV-Ray
interact with the scene? My initial thought (when I was thinking about
implementing that same paper) was to sample the simulation area and thus get
the boundaries of objects for the smoke to interact with. In effect, I just
"bake" the animation once by running the render without smoke and getting
the volumes which block the smoke, then run the simulation, and then render
the animation with the smoke. Since I haven't coded anything in that regard,
there's no proven method though.

Another advice: if you do release the application, how about making the
source-code available via email-request? I sure think some of us Povers
would like to have a look at it and see if they can improve it, or enhance
it with extra features...

Regards,
Tim

-- 
"Tim Nikias v2.0"
Homepage: <http://www.nolights.de>

Post a reply to this message

> To keep you pacified: ...

Well, that's good.  I just get paranoid about those things.  I guess there's
a reason people don't like spam.
> 
> My initial thought (when I was thinking about
> implementing that same paper) was to sample the simulation area and thus
> get the boundaries of objects for the smoke to interact with. In effect, I
> just "bake" the animation once by running the render without smoke and
> getting the volumes which block the smoke, then run the simulation, and
> then render the animation with the smoke.

I haven't really put a whole lot of thought into that yet.  That's not a bad
idea though.  For now, my approach is just to set up the simple primatives
in the c++ code and then have it write a pov include file that places
objects.  But I haven't really begun that stage yet, so I guess there's
plenty of time to think about it.  Ideally, it could be patched in kind of
like MechSim, but that's well outside of my ability.

> Another advice: if you do release the application, how about making the
> source-code available via email-request? I sure think some of us Povers
> would like to have a look at it and see if they can improve it, or enhance
> it with extra features...

Well, to be honest, there isn't really an 'application.'  It's just about
1000 lines of code.  It's set up as a big object though, so it runs with
only a few lines of setup and just 'Fluid1.frame();' to calculate one
frame.  So on the contrary, I will freely give away the source code (once
it's a little more complete) and maybe I'll have to send someone the
executable by email request.  But for now there's plenty to keep me busy: 
Re-implement boundary conditions (better), add a dimension, add support for
objects (static and dynamic), add an implicit solver for diffusion, add an
implicit solver for viscosity, add incomplete cholesky decomposition, and
maybe even make it user-friendly... maybe.  And I actually have time to do
these things.  I've forgotten what free time feels like.

 - Ricky

Post a reply to this message

By Jove, I think you've got it.

Post a reply to this message

Looks really nice. Are you thinking of an executable with a GUI/adjustable
parameters? That would be very interesting (and output file size should be
adjustable up to 1920 x 1080!).
Good luck with that.
Dennis
"Ricky Reusser" <no_### [at] nowherecom> wrote in message
news:40b9270d@news.povray.org...
>
> > To keep you pacified: ...
>
> Well, that's good.  I just get paranoid about those things.  I guess
there's
> a reason people don't like spam.
> >
> > My initial thought (when I was thinking about
> > implementing that same paper) was to sample the simulation area and thus
> > get the boundaries of objects for the smoke to interact with. In effect,
I
> > just "bake" the animation once by running the render without smoke and
> > getting the volumes which block the smoke, then run the simulation, and
> > then render the animation with the smoke.
>
> I haven't really put a whole lot of thought into that yet.  That's not a
bad
> idea though.  For now, my approach is just to set up the simple primatives
> in the c++ code and then have it write a pov include file that places
> objects.  But I haven't really begun that stage yet, so I guess there's
> plenty of time to think about it.  Ideally, it could be patched in kind of
> like MechSim, but that's well outside of my ability.
>
> > Another advice: if you do release the application, how about making the
> > source-code available via email-request? I sure think some of us Povers
> > would like to have a look at it and see if they can improve it, or
enhance
> > it with extra features...
>
> Well, to be honest, there isn't really an 'application.'  It's just about
> 1000 lines of code.  It's set up as a big object though, so it runs with
> only a few lines of setup and just 'Fluid1.frame();' to calculate one
> frame.  So on the contrary, I will freely give away the source code (once
> it's a little more complete) and maybe I'll have to send someone the
> executable by email request.  But for now there's plenty to keep me busy:
> Re-implement boundary conditions (better), add a dimension, add support
for
> objects (static and dynamic), add an implicit solver for diffusion, add an
> implicit solver for viscosity, add incomplete cholesky decomposition, and
> maybe even make it user-friendly... maybe.  And I actually have time to do
> these things.  I've forgotten what free time feels like.
>
>  - Ricky

Post a reply to this message

"Ricky Reusser" <no_### [at] nowherecom> wrote in message
news:40b8fbcd@news.povray.org...
>
> Here's my attempt at the 'Visual Simulation of Smoke' paper (google it!).
I
> have just about everything in place - vorticity confinement, a conjugate
> gradient solver, the semi-lagrangian scheme with cubic interpolation, but
> the obvious next step is to add a third dimension.

I take it that this is using semitransparency then? Your animations haven't
shown it against a background so I kept thinking of it as pigments going
from white (or gray) to black. Oh wait, that origin dot would be obscured by
darker parts if that were true.

It's a fascinating thing to see rendered from POV, even if not made from
within it entirely. People are going to want to see this interact with
objects, I know I sure do, like Tim was saying. Maybe could be used for
aerodynamics simulations, etc.!

Bob H.

Post a reply to this message