PDA

View Full Version : Is the flex engine taking over completly from the bullet engine?



prometheus
03-25-2017, 11:01 AM
Just wondering..

We have seen some nice stuff recently with flex engine, and I reckon there is cloth and soft body and hard body showing up, all supported to take advantage of nvidia cards.
Now ...did the Lighwave team opted for the wrong phys engine? maybe it was the better at that current time when they left the old hardFX.
So what do you guys think will happen here, or should happen.
Should they start over again and drop bullet, or improve that if it could be compared to the flex stuff?...or will we be left with two engines where we have excellent performance and features in one of them but not in the other, like you have better rigid bodies, cloth etc..but it still hasnīt got bone dynamics..while bullet has etc?

I havenīt really read up on the base tech behind them both, maybe I should..but I am interested to hear from those more versed in bullet source code and the flex engine.

Michael

darkChief
03-25-2017, 11:47 AM
Whichever direction they decide to go, SDK access pretty pretty please.

erikals
03-25-2017, 12:26 PM
don't care much.

the best thing for NT to do Next would be to add good
Water fluids / Smoke fluids

only then would your training be complete.

136349

Oedo 808
03-25-2017, 01:31 PM
Whichever direction they decide to go, SDK access pretty pretty please.

The vehicle may change but the destination remains the same. If that is true then SDK access should be forthcoming. On the other hand gimping the SDK access for third parties then whinging about the architecture as an excuse may have become all too comfortable.

prometheus
03-25-2017, 01:57 PM
don't care much.

the best thing for NT to do Next would be to add good
Water fluids / Smoke fluids

only then would your training be complete.

136349

Many of us seem to want that natively donīt we:D
If it because getting a simultanous development alongside lightwave...so there wonīt be any time displacement between lightwave versions and what third party developers has for update releases in time?
Or may it be that we think it would be cheaper?
Or may it be that we think such water / smoke fluids would be better integrated with the other physics engines?

As far as the main concern I have with the bullet system and any upcoming different engine PHysx/flex, it is wether or not the lightwave team continues to work on bullet, they have something really interesting with all the newer motors, bone dynamics etc, but if the deforming body bullet system is considerably slower and less realistic than a flex cloth engine, then you would be sort of stranded with no use of the bullet bone dynamics, if it doesnīt play along with flex cloth etc, just a scene case.

Michael

erikals
03-25-2017, 02:09 PM
it's not impossible that they will jump on the NVidia Volkswagon, as the bullet coders seems a bit halted?

i'm very happy we are finally getting fluids in LW though, it's nice to have several to choose from

- Hurley Up
- Deep Rising FX
- RealFlow
- Houdini Indie

wingzeta
03-25-2017, 02:41 PM
Flex doesn't work on the Mac at present, so I would imagine it is not a viable option for a cross platform application like LW. Hopefully we will get improved physics and particle performance for real time feedback, one way or another, in the not too distant future. Good fire, smoke and fluids are going to be required functions of any 3D app in the near future. I would guess everyone who doesn't already have those things native, is working on it, to keep up.

darkChief
03-25-2017, 04:00 PM
I will try to give a non bias opinion.

Not familiar with what technology powers Bullet, possibly a spring solver. But I heard they implemented a few.position based dynamics in there.

The Bullet community seems active, they have regular releases, but they don't have big headlining features like Physx/Flex. There doesn't seem to be a focus, or vision where their going. They have GPU support (for a few years now, even before Physx) but they don't seem interested in making it a priority, so it's limited from what I've heard. I wonder who runs the community, seems random people contribute the big features.

I personally like there API, doesn't take much to get going, but Nvidia has better documentation.

Nvidia is a company, and their goal is to make money. Probably why they haven't supported in latest Opencl specification. So Cuda is the fastest compute language at the moment. They know their audience, even though they acknowledge the digital content creation industry Physx and Flex are targeted at the gaming market. This is intentional, they fund half the technology papers I've been reading. Physx 2 is drastically different from Physx 3, I bet a lot of people had to rewrite a lot of code to reintegrate it back into their products. And they deprecate stuff without consultation. However the guys who code Physx & Flex are really cool, no matter what position they hold, they are always willing to help out.

I will avoid the topic of which is better:D

prometheus
03-25-2017, 04:18 PM
Flex doesn't work on the Mac at present, so I would imagine it is not a viable option for a cross platform application like LW. Hopefully we will get improved physics and particle performance for real time feedback, one way or another, in the not too distant future. Good fire, smoke and fluids are going to be required functions of any 3D app in the near future. I would guess everyone who doesn't already have those things native, is working on it, to keep up.

Never pick a mac for 3d work....(sorry for that98~)

Not sure One could "guess" everyoe who doesnīt have those thingsz natively is working on it, if turbulenceFD is working great with lightwave, My guess would be the lw team even this very day hasnīt got any person with the core team to work on such thing, and since we got two upcomers writing some fluid liquid stuff, my guess is that they simple save the time and resources on that and let the third party developers do their thing....but itīs just guessing.

m.d.
03-26-2017, 12:13 AM
I think LW should focus on a unified solver data flow SDK, and nodal access..... meaning the solver itself is not the focus...and the data can be passed from 1 solver to the next...be it native or third party.
That way FLEX can talk to bullet etc etc. Takes the heavy lifting off Newtek for writing dynamics, and allows simpler integration for developers....and beyond that, artists should get nodal access to the data coming to and from the various solvers.

vncnt
03-26-2017, 01:46 AM
You canīt interface every engine unless you define a global standard for that interface. This would also reduce the problems that will come with depricated software and hardware. Software that relies on specific hardware is even worse (+ expensive for both developer and user).

I think it would be more logical to enhance the Bullet development by donating code for hardware acceleration and code for fluids/smoke.

I think it would be sensible to focus on a modern key handling system that enables advanced manual animation editing first.
If other methods fail we need to be able to manually tweak on all available levels, in a single interface, as fast as hell.

For some, manual animation is the least sexy subject.
But physics engines don't rely on Lightwave for rendering. Many other programs can do that. If they can code physics simulation I'm sure they can enhance their software to render as well. Piece of cake --> Lightwave out of the game.
Advanced pose and motion manipulation, from different sources, would add something meaningful for (new) Lightwave customers.

Without a strong and flexible system architecture any support for other sw/hw physics engines is reduced to a pimple on your beautiful system.
A necessity that will lose its relevance over time, faster than you may think.

erikals
03-26-2017, 03:00 AM
Software that relies on specific hardware is even worse
yes, maybe we are going back to a Silicone Graphics state...   :)

MichaelT
03-26-2017, 04:05 AM
Flex is a lightweight solution to physics.. It will do the job nicely in many applications (particularly in games, where I do most of my things) but for accurate renderings you would still need traditional solutions. That said, I'm having fun with playing around with their source myself. I know I said I'm going to do something for LW, but I haven't gotten around to it yet. I'm not very versed in LW SDK either, and being faced with learning that (more than I already have) over other more pressing things.. have pushed doing that ahead. Luckily it does run on both AMD & nVidia. As for Mac, it seems to be more about Apple shenanigans. I don't have a Mac myself.. but there is no technical reason for it not working on that platform. That makes me think that it is in fact something that Apple requires that is preventing it. Probably something neither AMD, nor nVidia is willing to deal with yet. Speculation I know.. but since I don't see a technical reason for it not working there...

ianr
03-26-2017, 05:53 AM
I will try to give a non bias opinion.

Not familiar with what technology powers Bullet, possibly a spring solver. But I heard they implemented a few.position based dynamics in there.

The Bullet community seems active, they have regular releases, but they don't have big headlining features like Physx/Flex. There doesn't seem to be a focus, or vision where their going. They have GPU support (for a few years now, even before Physx) but they don't seem interested in making it a priority, so it's limited from what I've heard. I wonder who runs the community, seems random people contribute the big features.

I personally like there API, doesn't take much to get going, but Nvidia has better documentation.

Nvidia is a company, and their goal is to make money. Probably why they haven't supported in latest Opencl specification. So Cuda is the fastest compute language at the moment. They know their audience, even though they acknowledge the digital content creation industry Physx and Flex are targeted at the gaming market. This is intentional, they fund half the technology papers I've been reading. Physx 2 is drastically different from Physx 3, I bet a lot of people had to rewrite a lot of code to reintegrate it back into their products. And they deprecate stuff without consultation. However the guys who code Physx & Flex are really cool, no matter what position they hold, they are always willing to help out.

I will avoid the topic of which is better:D



Erwin Coumans for BULLET on Github

darkChief
03-26-2017, 06:35 AM
Thanks for sharing.

Greenlaw
03-26-2017, 11:35 PM
Personally, I don't care which physics engine LW3DG or third-parties go with so long as it's easy to use and gets the job done in a reasonable time (in other words, as quickly as possible.)

FWIW, I've been using LightWave's Bullet for physical dynamics in feature films, commercials and TV shows ever since it was introduced in LW 11, and the tool has always come through for me. I started using it at Rhythm & Hues in video game cinematics, and my '2015 Asylum' demo reel (http://forums.newtek.com/showthread.php?153202-D-R-Greenlaw-My-The-Asylum-VFX-Demo-Reel) I posted the other day has Bullet all over it, for everything from rigid body destruction, environmental fx, bone and cloth dynamics. I've found Bullet to be easy to use, fast, and reasonably controllable even under the tightest deadline. I've even use Bullet successfully for hair dynamics without much fussing. Today, I use Bullet routinely in personal projects (in the Brudders music video, it's used for wind blown leaves and grass, as well as hair fx,) and I occasinally use it for fx in shows at DreamWorks. I would love to see LW3DG continue supporting and developing LW Bullet, but even in its current form, Bullet is an incredibly powerful tool. It's one of many reasons I continue to use LightWave in personal and professional work.

Now, if I can combine Bullet with either or both of the two new liquid plugins for LightWave (like I can with Turbulence FD,) that would be perfect for me since I already know Bullet very well. But I'm totally fine if another physics engine is brought into the mix as long as it works as well as or better than what we have now.

Norka
03-27-2017, 06:41 AM
It seems that in March 2013, there was talk of Bullet (rigid) being 100% GPU accelerated in version 3.x, which was presumably in beta at that point, and release was imminent.... That was four years ago, and Bullet 2.86 is the latest, as of January. I can see on their GitHub that they have steady releases etc. I'm sure it's constantly being developed, but it is just a little perplexing how there is no full GPU acceleration yet...

vncnt
03-27-2017, 10:09 PM
I would not describe it as "perplexing" because I do not pay them.

Norka
03-28-2017, 06:23 AM
Sorry. It is not like this affects me in any profound way. I just figured something must have thrown a monkey wrench into your GPU acceleration part of development. And I only recently became interested in all this, out of a desire to have everything be accelerated by GPUs, where possible. No biggie. Take your good old time. UP is looking pretty promising, and may be able to speed up some things, as Steve adds more. :-)

jwiede
03-29-2017, 03:01 AM
yes, maybe we are going back to a Silicone Graphics state...   :)

Are you really citing SGI as an example of a company using software to lock customers into hardware?

cherishjoo
06-04-2017, 02:26 AM
Is the problem solved?