PhysX Physics Engine

[sobek]

We could think of PhysX SDK as DirectX SDK in Graphic Engine!

 

Why on gods green earth would ED want to do that? You keep coming back with this idea every month or so just to get told that it doesn't work, so i'm starting to wonder what your point is... :)

[Witchking]

They can't use the Physx SDK without trademarking it Physx and limiting it to Nvidia cards. Besides...exclusivity...is just not something they can afford with such a niche market.

 

I think the priority comes down to usage of multi core/multi threads in the next engine. That will perhaps greatly enhance performance and quality of DCS.

 

Sobek, I understand ED is capable of making their own physics system (considering the amazing FMs ED has already made). But, why not license things like Havok for effects? Would tremendously help them. Havok probably has a huge library of existing templates/examples and is highly customizable.. and very very resource efficient.

 

Like one time ED was considering Speedtree. I wonder what happened to that. Would definitely make terrain more like ARMA and pretty detailed.

[EtherealN]

I'm not a coder, but regarding Havok: it's geared for "shooter" style engines.

 

Where, exactly, would you see it applied in DCS?

[OttoPus]

Ballistics and particle effects, just to name two GE area that use physics calculation.

 

Physics are physics, I think that "style" is not a word suitable for such matter.

[Bushmanni]

Havok, Physx and such are geared for visual effects simulations, they aren't built for producing physically accurate results like you need with ballistic calculations. There's multiple "styles" or ways you can solve the math problems in physics numerically and they all have their pros and cons. While the laws are universal, the way that they are solved in different situations isn't.

[Witchking]

I'm not a coder, but regarding Havok: it's geared for "shooter" style engines.

 

Where, exactly, would you see it applied in DCS?

 

rigid body collisions. Projectile impact. Fragmentation and rag doll physics.

[sobek]

rigid body collisions. Projectile impact. Fragmentation and rag doll physics.

 

The latter two are simply out of scope for a flightsim at this time, you won't be simulating any of that at the unit counts that DCS sports. Unless you like to measure frame time in hours.

[SkateZilla]

EDPE already exists :p

[danilop]

Havok, Physx and such are geared for visual effects simulations, they aren't built for producing physically accurate results like you need with ballistic calculations....

 

+1

 

It seems that people are not aware of the complexity of the mathematical model when you want to simulate advanced aerodynamics, rocket and jet engines to AFM standard. Radar? Sensors?

 

You don't do it with physX or any other consumer technology meant for simulating environment in FPS games, but with serious scientific math libraries running on the cutting edge processors.

 

So, as already mentioned, the only reasonable and sane way to increase "physics" performance within DCS is to fully optimize code for multi core / multi threaded processors.

 

That's urgent.

 

Icing on the cake, although very unlikely, would be support for serious scientific computing ability on the new Nvidia and AMD processors which bring unprecedented parallel processing power to consumer computers.

[c0ff]

Why don't you suggest to incorporate Bullet?

 

I can tell you why. Because Havok and PhysX are commercial products which need a hype to sell.

[c0ff]

Icing on the cake, although very unlikely, would be support for serious scientific computing ability on the new Nvidia and AMD processors which bring unprecedented parallel processing power to consumer computers.

The problem is slow readback is still not solved.

[c0ff]

So, as already mentioned, the only reasonable and sane way to increase "physics" performance within DCS is to fully optimize code for multi core / multi threaded processors.

 

Yes. Unfortunately it requires rearchitecturing the whole thing from the ground up.

[Witchking]

Yes. Unfortunately it requires rearchitecturing the whole thing from the ground up.

 

Yeah....but isn't that worth it? All the fps issues and too many objects limit will be "alleviated" at least by a fair margin. Why not do this now as edge is still being developed than postpone it again.

[SkateZilla]

EDGE isnt just a terrain engine, it's a render engine for everything. so...... take it for what it's worth.

 

It would be beneficial to move some stuff to the 2nd thread, cause AFAIK it's EVERYTHING then SOUND on it's own thread.

[OttoPus]

Yes. Unfortunately it requires rearchitecturing the whole thing from the ground up.

 

Which is what you are doing with EDGE right?

[SkateZilla]

Which is what you are doing with EDGE right?

 

The entire base code is not the same thing as a Image Renderer

[Darkwolf]

I don't see where that engine could be relevant in a full scale simulation. That would be loss of CPU usage for the simulation itself.

 

If naval area would be develloped, I would prefer real water sound propagation simulation such as Dangerous waters, than a fancy water. Fancy water is always cool but accurate torpedo and sensors sonar simulation would just be the most important thing to simulate

[Witchking]

Yeah... Considering edge is going to be the umbrella renderer for terrain and everything else...why not leave the simulation/physics core on first thread as before and focus the newer development on the next cores/threads. Atlesst this way the rest of the stuff is taken over by others and the process becomes efficient. Its not too early to restrict dcs to a dual/quad core processor.

[c0ff]

The nearest target we are moving to is decoupling of graphics and simulation into separate threads.

This is quite possible with the current codebase, without a major rewrite.

[SkateZilla]

that would be a good thing actually.

[OttoPus]

The nearest target we are moving to is decoupling of graphics and simulation into separate threads.

This is quite possible with the current codebase, without a major rewrite.

 

This is a very good news. :thumbup:

 

Thank you.

[Witchking]

[jmod]

First, I really appreciate that ED team read our opinions in the forums!

 

Havok and PhysX are commercial products which need a hype to sell.

Like DirectX, PhysX is free for Windows platform.

The PhysX SDK is provided to developers for free for both commercial and non-commercial use on Windows.

 

Unfortunately it requires rearchitecturing the whole thing from the ground up.

It's a one time effort, and finally, by utilizing DirectX 11 for EDGE and PhysX 3.0 for EDPE, we can get the performance of ARMA 3!

 

Don't forget, some 3rd parties are going to implement some tanks with DCS: A-10C fidelity!

[c0ff]

Like DirectX, PhysX is free for Windows platform.

Oh, I see, if Bullet were BulletX, it'd get your approval, the same as OdeX would.

OpenGLX would feel some love as well.

 

EDIT: PhysX is free to sell NVidia cards.

 

 

It's a one time effort

 

If I had a dime for every time I heard that...

Edited March 6, 2013 by c0ff

[danilop]

The nearest target we are moving to is decoupling of graphics and simulation into separate threads.

This is quite possible with the current codebase, without a major rewrite.

 

Wow, that should boost performance considerably! :thumbup:

 

So basically we will have three-threaded application? Sound, Graphics, Simulation?