Newton 2.0x Archemedia Open Beta

[winstrol]

i did solve that somehow, but...

I am doin racing game, and when I use 4 cars in the scene I have about 4 FPS

Can this lib be optimized somehow for the race only?

and BTW, when can we expect wheel steering ?

[Julio Jerez]

For those following this thread, I am reporting that although I made good progress I did no cover as much as I wanted and still do no have a final SDK demo,
But for is something something for those following the thread.

http://www.newtondynamics.com/downloads/LevelDestruction.wmv
There is more work to do but I hope this gives an idea of the final version of this feature.

[thedmd]

I just look very good Julio. Excelent work.

[JernejL]

i did solve that somehow, but...

I am doin racing game, and when I use 4 cars in the scene I have about 4 FPS

Can this lib be optimized somehow for the race only?

and BTW, when can we expect wheel steering ?

In what kind of setting, running on what kind of hardware? are you sure it's newton using so much resources to drop your game to 4 fps?

[Julio Jerez]

Oh wow I was debug why when I hi t the building in one place it cause a cascade reaction the destroyed almost the entire rest of the building.

What I found is unexpected but very nice.
When a piece breaks, I get stoked in the other pieces, just like in real life.
This pieces generate a tremendous amp8ng of pressure trying to get out, but the keep getting stocked until the force outweigh the braking force of the other pieces and they break too, this keep happing until the y get wiggle room to move.
This is an amassing results that I was no expecting but it is very nice,
Although I do not think people will like or appreciate this level of realism

I made a new video
http://www.newtondynamics.com/downloads/LevelDestruction1.wmv

I am going to make a Debrief Material to reduce the friction among the debrie and the breakable pieces so that they can get loose quicker.

I do not like that solution because it makes the debris to slidy,
I was not expecting I have to tweak this because it was too good; by it is very nice to know.

I believe I have a better solution, than makin a debree materail,
Maybe scaling the collision shapes by a small amount, to allow some breading room to debieee to operate operate and not create those tremendous internal forces.
I am still down founded by this result.

[Julio Jerez]

Ok I implement the Scaling and now I get a much better behavior.
Now I have to add the code that detach large pieces as compound aggregates that can also break when hitting other pieces and that will be all for the first pass of this functionality.

http://www.newtondynamics.com/downloads/LevelDestruction2.wmv

[Dave Gravel]

Very nice.
Good work and thanks.

[Stucuk]

Nice. Btw instead of trying to guess where the piller/whatever is when your trying to hit it, wouldn't it be easier to create a object that starts life where the camera is and flys forward based on the direction of the camera(If you don't want more than 1, you could just move the original block to the camera's origin and re-fire it). That way you just need to aim at where ever you want a block/whatever to hit the building and press a button.

[Julio Jerez]

yes that will be cool, but I am hoping the aprentices has some fun doing that kind of demos in thier projects.
for now I can use the time finding the last bugs I still have here and there.

It would be cool though to see objets jitting a building and blowing pieces at the impact.

[Van]

Great job! Nice improvements. We are migrating over from 1.53 to 2.x in our test environment.

Question: Is the TreeCollision still a stationary object? Is it still not a movable object?

Suggestion: add new function NewtonBodySetScale( NewtonBody, dFloat* Vector3[0] ) and NewtonBodyGetScale()

It would be nice to be able to set a scale factor to a NewtonBody or NewtonCollision. This is so that it can be changed during runtime. The scale should default to (1, 1, 1). All zeros would be invalid therefore the library should probably clamp it to 0.001 or something.

Example Case Scenario:
In our application we have asteroids. All our asteroids have a pre-defined mesh. What we do is scale the asteroid based on their contents. That is, the more ore an asteroid has the larger the asteroid appears (i.e. we "scale" it). If an asteroid has 100 units of ore, then the asteroid is scaled to (100, 100, 100). As the asteroid is mined and ore is removed by the player, the asteroid shrinks by the number of units removed. The scale is reduced and thus the asteroid appears to "shrink". This works fine in the rendering scene manager - the SceneNode is scaled accordingly. However, this presents a problem in Newton. There is no scale factor so we have to delete the NewtonBody and recreate it. We use the TreeCollision for asteroids so there is some significant overhead to tearing down and recreating this NewtonBody - enough to cause "flicker" in the game.

We believe having a scale factor would be a valuable asset the newton library. Of coarse the internal math would have to include Scale but this would really add a lot of power to Newton.

[Julio Jerez]

collision tree are static yes, Collision trees have a huge limitation to make then dynamics,

Howvere station with 2.09 compound collision can handle unlimited number of convex objects just liek collsiion tree can handle unlimited number of faces.
if you can create your model from a set of convex pieces then you can make dynamics solid artrary shapes.

In the destrution demo the mesh is now made of 9000+ pieces, and the effect is neglegible, in fact collision trees and compound collision are equally efficent in Netwon,
you can make levels of convex shape of 10's of touasnd of pices. maybe you can use that.
I will add a convex aproximation to help make solid meshes from meshes, but that will be for the next, next Beta.

Scale? there is the Scale modifier shape that gives you the full scale control. non uniform or uniform scale plus movement of the relative position of the shape.
Is not that enought for the job?

[Van]

collision tree are static yes, Collision trees have a huge limitation to make then dynamics,

Howvere station with 2.09 compound collision can handle unlimited number of convex objects just liek collsiion tree can handle unlimited number of faces.
if you can create your model from a set of convex pieces then you can make dynamics solid artrary shapes.

In the destrution demo the mesh is now made of 9000+ pieces, and the effect is neglegible, in fact collision trees and compound collision are equally efficent in Netwon,
you can make levels of convex shape of 10's of touasnd of pices. maybe you can use that.
I will add a convex aproximation to help make solid meshes from meshes, but that will be for the next, next Beta.

Ok, we'll look into that.

Scale? there is the Scale modifier shape that gives you the full scale control. non uniform or uniform scale plus movement of the relative position of the shape. Is not that enought for the job?

What is the name of the function? I'm searching the Newton.H file and don't see anything with "scale" in it that appears appropriate.

[JernejL]

What you are looking for is NewtonCreateConvexHullModifier and NewtonConvexHullModifierSetMatrix.

This is the only proper way to scale bodies in newton, but it works only with convex hulls.

[Van]

What you are looking for is NewtonCreateConvexHullModifier and NewtonConvexHullModifierSetMatrix.

This is the only proper way to scale bodies in newton, but it works only with convex hulls.

Thank you for your help.

However, convex hulls will not work for us. Convex hulls are to granular. We need to use the tree collision.

[Marc]

Those links in the first post to Newton 2.11 link to files called Newton 2.08. Is that by accident or intentional? Anyways, there is this file http://www.newtondynamics.com/downloads/NewtonWin-2.11.rar