PDA

View Full Version : Lightwave 11 questions



larkis
08-31-2012, 03:28 PM
I'm a long time user of lightwave who has not upgraded to version 11 due to various factors that made the app not fit within the pipeline I'm working with. There is an upcoming project where I need to explode the interior of a house in a pretty realistic way. I'm thinking of upgrading my lightwave to 11 because of it's bullet support but wonder how bullet functions in LW in non demo video reality. Is it useful for actual production ? Mainly:

1) Does it handle a lot of geometry that is not instanced ?
2) Where is the dynamics cache stored ?
3) Can solving be resumed after a crash ?
4) Is the result exportable via a method that can handle a lot of geometry ?
5) What is the speed of the solver compared to other bullet implementations ?

Thanks for any answers in advance.

dwburman
08-31-2012, 09:03 PM
Probably the best way to get the answers you need is to download the 30 trial and try it out.

To answer the questions I know the answer to: A cache file is made in one of the subfolders in your content directory. I don't think it'll recover from a crash, but I haven't tried that. I know it will not pick up where it left off if you extend the timeline beyond the initial solve.

Hopefully someone with some more production experience with it will chime in here.

jwiede
09-01-2012, 01:08 AM
1) Does it handle a lot of geometry that is not instanced ?
Yes and no. It can handle a fair amount of basic shapes, but if they're all complex meshes, you'll likely have painfully long sim times. Also, the meshes need to have "real dimensions", meaning segments need actual thickness, etc. so you'll need to build things accordingly. If you can keep to fairly basic shapes (boxes, spheres, capsules, pyramids/prisms, etc.) you should be okay, but avoid complex and esp. concave meshes as much as possible, esp. in regions where they'll be constantly part of the spatial-sectioning evaluations.

To be clear, that's a general Bullet limitation across implementations, though, it isn't specific to LW's implementation, most implementations bog down quickly as the number of complex meshes involved rises.

2) Where is the dynamics cache stored ?
Separate folder in content dir, IIRC.

3) Can solving be resumed after a crash ?
Not really, it more or less always needs to do a complete run from whatever starting state is given. As mentioned, this kicks in even if just extending the existing range. This is definitely an area where the LW implementation needs some improvement, allowing precise setting of ranges to be considered, and capturing specific frame sim state to reuse as "start state" for subsequent sims would add significantly more ability to control sim times/scope.

(edit) See Cageman's comments below, his test showed simply extending range didn't require recompute from start. YMMV.

4) Is the result exportable via a method that can handle a lot of geometry ?
Not quite sure what you're asking here, are you asking whether results can be baked to an exportable format like MDD?

5) What is the speed of the solver compared to other bullet implementations ?
It's not the fastest Bullet implementation I've used, but not the slowest either, I'd say (based solely on anecdotal observation, mind you) that it's pretty average/typical in solving speed among Bullet implementations. There really isn't a tremendous amount of variation in performance across the Bullet implementations out there, really, that I've noticed anyway.

Thanks for any answers in advance.
No problem, you're welcome, and please ask any further questions that come to mind.

I do feel obligated to point out that the LW11 implementation is pretty lacking in constraints versus most other Bullet implementations I've used. If you need much in the way of constraints (motors, connection types, springs, forces, etc.) you'll probably not be happy with it. If you can get by without constraints mostly gravity or key-motion-driven, though, you should be in pretty good shape. I strongly recommend playing with the demo and trying out some usage examples of the type you care about, to see whether the current implementation supports everything you need.

Hope this helps!

Cageman
09-01-2012, 03:34 AM
Not really, it more or less always needs to do a complete run from whatever starting state is given. As mentioned, this kicks in even if just extending the existing range.

Extending the desired simulation, lets say it stops at frame 100 (timeline is set to it) and you realize that you need another 50 frames, extending the timeline does not require a recomputation of the simulation up to frame 100. In the example I wrote, when you hit play, it will only simulate the new frames.

The only reason it woud require a resimulation of the entire thing is if you are changing dynmaics parameters.

In essence, one can start a Bullet sim, save the scene (which also saves the Dynamics cache into the Dynamics folder of your content dir), close LW and open it again and continue the simulation from where it stopped.

jwiede
09-01-2012, 04:13 AM
Hmm, I wrote what I did because my experiences (which matched up with dwburman's) were that when extending timeline it started over, but perhaps something's changed since I (and presumably, he) saw that behavior in LW? Cageman, were you using LW11.0.3 for your test? I know the test I'm thinking of was in the 11.0 timeframe.

I've got the older versions here, if I get a chance this weekend I'll test a bit to see if the behavior changed between 11.0 and 11.0.3 (and/or if there's a potential Win vs Mac difference here, as well).

dwburman
09-01-2012, 05:59 AM
It's been a while since I looked at it too, and it may have been that I just rendered more frames than had been calculated.