PDA

View Full Version : Lagoa Multiphysics :: Technology Focus


ds
09-15-2010, 02:15 AM
The Inventor of the Lagoa Multiphysics Technology demonstrates his new solver ahead of the Autodesk announcement that the technology will be part of Softimage 2011.5.

Click the image to find out more.

http://features.cgsociety.org/stories/2010_09/lagoa/lagoa_plug.jpg (http://features.cgsociety.org/story_custom.php?story_id=5866)

TyroneMaddams
09-15-2010, 08:23 AM
There is so much potential application of this, i'm excited just reading about it

- Ty

MikeRhone
09-15-2010, 08:25 AM
Great article! I am glad to see some major steps forward in cloth/fluids in the last year or so. That looks like some of the the most impressive commercially availble sims I've seen yet. I'm so glad to see Thiago didn't keep this proprietary!

ximage
09-15-2010, 08:36 AM
Congratz to Thiago.

its good to see this is getting some love on CG talk guys!!

looking forward to having a play with this. :)

Sweeney

jezsmith
09-15-2010, 04:57 PM
nice to see this made it here.
the first effect in the video is very interesting.

looking forward to seeing more :)

ViCoX
09-15-2010, 05:07 PM
............
How does that relate to anything, really? : ) Theres finished work section you know..

Katachi
09-15-2010, 08:18 PM
Itīs an inspiring implementation that Thiago has there. I think he did a great job.

But it would be interesting to know simulation times for real world complexity scenes or simulation times at all. The teaser shows interactive rates but this only shows some hundreds or a few thousands of particles at max. For some things this may be appropriate but in vfx production you really never use a hundred particles only so itīs not telling me very much actually.

The implementation seems very cool, but it would be nice to know how millions of particles work or interact and what the performance says in such a case. Otherwise nice article. I enjoyed reading it.

ThE_JacO
09-15-2010, 11:58 PM
Insofar everybody who used it/tested it agrees that it's one of the fastest and most scalable solutions around for SPH style work.
It's seen production use in a couple places well before it even made it into Softimage's beta (predators in hybrid etc.)

Katachi
09-16-2010, 08:36 AM
Insofar everybody who used it/tested it agrees that it's one of the fastest and most scalable solutions around for SPH style work.
It's seen production use in a couple places well before it even made it into Softimage's beta (predators in hybrid etc.)

thx first of all for answering. I do not doubt its scalability or production capability (I also read the article and others before ;) ), also I donīt want to make any point about if thatīs good or bad (if things can be done with it, then the simulation times are secondary but they are not unimportant of course), but "fastest, best, greatest, longest, slowest"...these are all subjective and still not tell me anything Iīm afraid. They just reflect the users opinion (which is fine and surely a good indicator donīt get me wrong). I simply have a more technical interest.

Fastest compared to what? I am really talking about explicit times. That is something that has not been talked about at all. The teaser is around for quite some time, the technology...well, if you are into the sph topic you knew before what the work was mostly based on. My interest lies therefore in the implementation and this would first of all be the performance times.

How long do for example 10 million particles take to simulate 100 frames including some static obstacles?

The article is revisited, so maybe they include some statistical facts. At least that would be great and I think if it is as fast as people say (and as itīs a particle-spring-sph model it should have a fast "base speed") itīd be a pusher. On the contrary, if nothing is said about it at all, it makes me wonder why.

Stellios
09-16-2010, 08:04 PM
thx first of all for answering. I do not doubt its scalability or production capability (I also read the article and others before ;) ), also I donīt want to make any point about if thatīs good or bad (if things can be done with it, then the simulation times are secondary but they are not unimportant of course), but "fastest, best, greatest, longest, slowest"...these are all subjective and still not tell me anything Iīm afraid. They just reflect the users opinion (which is fine and surely a good indicator donīt get me wrong). I simply have a more technical interest.

Fastest compared to what? I am really talking about explicit times. That is something that has not been talked about at all. The teaser is around for quite some time, the technology...well, if you are into the sph topic you knew before what the work was mostly based on. My interest lies therefore in the implementation and this would first of all be the performance times.

How long do for example 10 million particles take to simulate 100 frames including some static obstacles?

The article is revisited, so maybe they include some statistical facts. At least that would be great and I think if it is as fast as people say (and as itīs a particle-spring-sph model it should have a fast "base speed") itīd be a pusher. On the contrary, if nothing is said about it at all, it makes me wonder why.

It is all very exciting. But I agree, either there are benchmarks or its heresay. :)

MCook
09-16-2010, 11:37 PM
Shame, story still not up again?

Really wanted to read it.

Aeonsfall
09-17-2010, 07:27 PM
when is it going to online again?

circusboy
09-20-2010, 05:50 PM
No benchmarks mind you.
But you can get some sense of user interaction with these feature preview movies and they were recorded 'live'. Look for the Lagoa stuff.
http://www.the-area.com/blogs/marks/introducing_softimage_2011_5_with_lagoa

And just to turn the tables-are there performance benchmarks listed for DPit yet? :twisted:

KidderD
09-20-2010, 06:15 PM
There kind of is benchmarks for Dpit, and they are impressive.

circusboy
09-20-2010, 06:31 PM
"impressive" isn't a link.
I looked on the site-I haven't found anything yet...

Katachi
09-20-2010, 06:56 PM
thanks. I have seen all of those videos and I only see very low particle counts, at max maybe 2000-4000 thousand (in the wine scene, otherwise itīs only a bunch or some hundreds). Thatīs not really saying anything about complex scenes Iīm afraid. I am talking of ranges of millions.

There are no simulation benchmarks on the dpit website but I have posted simulation times in the cinema 4d subforum here on cgtalk (10 million particles, coast scnene of trailer, 1:40m/frame). But itīs maybe not appropriate to start talking about my own software when this is about lagoa. I really was interested in its simulation times, just out of curiosity. :)

thx guys

ViCoX
09-20-2010, 07:11 PM
thanks. I have seen all of those videos and I only see very low particle counts, at max maybe 2000-4000 thousand (in the wine scene, otherwise itīs only a bunch or some hundreds). Thatīs not really saying anything about complex scenes Iīm afraid. I am talking of ranges of millions.

There are no simulation benchmarks on the dpit website but I have posted simulation times in the cinema 4d subforum here on cgtalk (10 million particles, coast scnene of trailer, 1:40m/frame). But itīs maybe not appropriate to start talking about my own software when this is about lagoa. I really was interested in its simulation times, just out of curiosity. :)

thx guys
dpit looks nice! What it would take to convert you to ICEholic!?
"Dude, just try it once"
"One time is ok" "Its not addicting"
"The first try is free, just download the trial"

and soon you`ll be attending ICEholic gatherings with Thiago and Helge! Buahah! : )

circusboy
09-20-2010, 07:43 PM
dpit looks nice! What it would take to convert you to ICEholic!?
"Dude, just try it once"
"One time is ok" "Its not addicting"
"The first try is free, just download the trial"

and soon you`ll be attending ICEholic gatherings with Thiago and Helge! Buahah! : )
Us softimage users would love you for it! Best way for you to get into the autodesk market is to make plugins for autodesk products too. :applause:

Katachi
09-21-2010, 08:47 AM
hehe, well, maybe I will have a look but it would take some serious interest (and me liking ICE programming capabilities) to make me port dpit.

ViCoX
09-21-2010, 10:26 AM
hehe, well, maybe I will have a look but it would take some serious interest (and me liking ICE programming capabilities) to make me port dpit.
You`ll definetly will appreciate ICE. : )
Check out Thiagos(Lagoa author) ICE tutorials from cmivfx (http://www.cmivfx.com/store/Browse.aspx).
They will give you good headstart and probably ingnite your intrest.

Hirazi
09-21-2010, 11:39 AM
Well, appreciating and using ICE is not the same as programming nice ready solutions for it.
I'd take a look at the SDK documentation before anything else ;)
(And that last remark is a slight exaggeration, obviously)

Katachi
09-22-2010, 10:45 AM
thank you, but honestly, I would never buy anything from cmivfx. I have made very bad experiences with them (they got a copy of DPIT some years ago as they wanted to make training material for it, well, guess if they did...), so my trust and support for them is gone.

Is there any other material for this available? I donīt need training material, all I need is a basic introduction to the concept of ICE and a reference documentation. I can figure this out myself otherwise (never have problems with that). And is the SDK accessible from within the demo? Is there a demo at all? Is there a programmers support forum or something similar? Where do I get the docs?

thx

P.S.: Darn, and now we turned this into an offtopic anyway. Sorry!

Hirazi
09-22-2010, 11:00 AM
The documentation & the SDK documentation are freely available here (http://tinyurl.com/39xavy8) and the Mod Tool 7.5 (http://tinyurl.com/2v2t93z) which includes ICE, up-to-date to version 7.5, seems a good way to get your feet wet.

ViCoX
09-22-2010, 11:25 AM
and.. There is demo available for SI. : )
you should also post&subscribe all the questions to softimage mailing list (http://groups.google.com/group/xsi_list), there everybody are very active answering, both SDK and app. specific problems.
Its much more active than any xsi specific forum ect.
Its not really offtopic, were talking about ICE so its kinda related. : )
Have fun!

Thats pretty suprising to hear about cmivfx. Too bad they didn`t keep their promise. : /
I`m sure your technical knowledge is more than enough to learn it without cmivfx help, tho! : )

Katachi
09-22-2010, 11:40 AM
thx both. I hope the download works this time. Last time I tried to download something from Autodesk I needed to install a strange downloader and that didnīt work and I had to seek the forums to get a real link and...argh, it was a PITA really. Why having it easy when you can make it hard. ;) Iīll check everything out.

circusboy
09-22-2010, 01:09 PM
Go go go! :D
There is definetly a hole to fill for Aerodynamic fluid sims *with* the shaders to match.
Yours look very nice I must say.

DaveA
09-27-2010, 10:19 AM
hehe, well, maybe I will have a look but it would take some serious interest (and me liking ICE programming capabilities) to make me port dpit.

Don't do it Samir....just say NO!

- A C4D and future DPIT 1.5 user.

CiaranM
09-27-2010, 10:23 AM
The documentation & the SDK documentation are freely available here (http://tinyurl.com/39xavy8) and the Mod Tool 7.5 (http://tinyurl.com/2v2t93z) which includes ICE, up-to-date to version 7.5, seems a good way to get your feet wet.

I think this is quite an old release as they've added some nice ICE related stuff to the SDK since then. The latest demo (http://usa.autodesk.com/adsk/servlet/pc/index?id=13571168&siteID=123112) might be best, if you can manage on only 30-days.

ViCoX
09-27-2010, 10:29 AM
Don't do it Samir....just say NO!

- A C4D and future DPIT 1.5 user.
Never say NO to drugs ; )

paulselhi
09-27-2010, 11:37 AM
Great to see this thread is staying on topic

TwinSnakes
09-27-2010, 07:52 PM
So am I getting this right? They released "some" of their ICE nodes to Autodesk and kept the rest for a separate product SlipStreamVX?

bhouston
09-27-2010, 07:58 PM
So am I getting this right? They released "some" of their ICE nodes to Autodesk and kept the rest for a separate product SlipStreamVX?
Sort of. Lagoa Multiphysics, which was created by Thiago Costa independently, was licensed to Autodesk. SlipstreamVX is a separate ICE simulation project (which focused on simulating different physical phenomena) that resulted from collaboration between Exocortex Technologies (my company) and Thiago Costa.

If you want to try out SlipstreamVX and its GPU-accelerated point renderer (which is similar to a previous project of mine, Krakatoa) you can download a free trial here:

http://www.exocortex.com/simulation/slipstreamvx

-ben

Hirazi
09-28-2010, 03:45 PM
I think this is quite an old release as they've added some nice ICE related stuff to the SDK since then. The latest demo (http://usa.autodesk.com/adsk/servlet/pc/index?id=13571168&siteID=123112) might be best, if you can manage on only 30-days.

Very true, but my idea to recommend the Mod Tool was merely a suggestion to "get your feet wet", to find out if ICE even works for you, without the 30 days restriction of the demo version. That demo would have been the next step after you "get your feet wet": "diving in", so to speak... ;)

TheGrak
09-28-2010, 04:17 PM
The multiphysics teaser is very cool! :)

Fluckrat
10-13-2010, 03:36 PM
I've asked in a few places and heard nothing, so here goes one last try...

Will this article ever be restored? Anyone know?

dwigfor
10-14-2010, 06:34 AM
Works for me. Seems to be back online.

oktawu
10-14-2010, 08:39 AM
i find it funny that the article was substantially changed.
even funnier is that the whole part where thiago said
he was looking forward to implementing the tech in other apps
got scrapped. wonder why....:)

Katachi
10-14-2010, 10:25 AM
And still not a single word about runtimes, although itīs even already released. Well, I have the feeling I shouldnīt wait for it.

oddforce
10-14-2010, 12:09 PM
Congratulations, Thiago. It is very much fun to work with it. Softimage is so ridiculously underrated (or rather under-represented I should say) - I hope this helps change that a little

PiotrekM
10-14-2010, 02:08 PM
And still not a single word about runtimes, although itīs even already released. Well, I have the feeling I shouldnīt wait for it.

I think you must give people some time to test Lagoa.

//I'm running simple fluid test with 3 million particles, 300 deforming rigid bodies acting on it and it's simming stable at 15 minutes/frame / 30 substeps.

milanS
10-14-2010, 03:05 PM
yeah, Lagoa is awesome and very inspiring to me!

A few weeks ago I created a custom SPH fluid plugin for SideFx Houdini: http://vimeo.com/15536503
aprox. Release: December 2010, of course FREE :bounce:

It is running on the GPU through OpenCL. Lagoa runs only on the CPU (if I'm right), so there is huge potential for future versions. Congratulations and many awesome simulations to XSI community.

ThE_JacO
10-14-2010, 11:11 PM
And still not a single word about runtimes, although itīs even already released. Well, I have the feeling I shouldnīt wait for it.
Not everybody (in fact nobody I know of in any FX department) is that obsessed with pure numbers that are close to impossible to compare and contextualize anyway. User experience is all here.

Why don't you just wait for a demo and try it yourself, since I suspect even if people posted something you would spend countless posts arguing how those numbers were obtained and in what context :)

Katachi
10-15-2010, 09:07 AM
Not everybody (in fact nobody I know of in any FX department) is that obsessed with pure numbers

Well, I donīt know if anybody asks this in a production environement but for a researcher in the cg fluid dynamics field, statistics of the usual scenarios (double-dam breaking simulation would be a good example) is the most normal thing in life.

that are close to impossible to compare and contextualize anyway. User experience is all here.

Of course they are comparable. If you know the amount of particles and the system specs then you can compare it just like that. No interpretation or anything necessary or contextualisation. I am not sure what you mean or where you derive this argument from. You may not be fully able to compare the results, this is absolutely down to the algorithm but thatīs about it. I donīt see a reason one cannot compare the standard fluid scenes in terms of performance.

Why don't you just wait for a demo and try it yourself, since I suspect even if people posted something you would spend countless posts arguing how those numbers were obtained and in what context :)

As I wrote before I am interested in its speed. Itīs been talked about a lot and the results look great (maybe you should read the whole thread beforehand?) and please stop getting sarcastically harsh. You donīt know me so please stop pretending I would react like this or like that. Either you answer constructively or you leave it alone but donīt judge my possible personal reaction. you donīt know me.

And if you had read the thread before you would have seen that I am not the kind of person who wants to argue but simply shows interest in the technology behind it, without any judgement. If you have read the thread then I must conclude that you are stirring on purpose and then this is discussion is over. I want to discuss constructively not personally.

noouch
10-15-2010, 12:37 PM
Those granular examples look real pretty, as do all the other solvers.

My biggest worry would be rendering the stuff. Changing topology has been a tough nut to crack in situations like that, or at least not many people have presented truly production-ready solutions. And although meshing functionality was shown, I'm not too sure how well it scales to millions of particles. I'm itching for a demo version :)

circusboy
10-15-2010, 01:43 PM
I'm itching for a demo version :)
Might have to wait until v2012 then. Since its a subscription advantage release-those with subscription get it anyway. So they don't need to 'demo' it.

CGTalk Moderation
10-15-2010, 01:43 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.