From Newton 1.53 to Bullet to Newton 3.1

[pitsikoko]

hey guys

more than 5 years ago i attempted to create a vehicle simulation using Newton 1.53 : viewtopic.php?f=14&t=3903
at the time Newton was closed source, i decided to switch to Bullet, then busy at work and the last 3 years on a degree i only worked on it from time to time, especially in the summer.

anyway, today it's n almost complete vehicle simulation framework with network tools to tweak in realtime that is almost ready to make a much more professional demo.
the reason for this post is that 2 days ago i switched back to Newton 3.1 and i wanted to share some feedback about it.

first i would like to say that you have done some AMAZING JOB, it's a really top physics lib which i'm happy to compare to bullet now (i only use the rigid body part)

performance : can't say really i haven't ran test with a huge amount of rigid bodies / constraints. but i'm not so much interested into having 10000 dynamics objects at the same time, i find this competition to have more and more triangles draw with more and more textures and more and more dynamic objects in games stupid. better to have quality instead isn't it ? that said it seems it can simulate a reasonable amount of rigid bodies at the same time, and more than enough for my needs.
quality : the quality of the collision response and constraint solving is much better than Bullet. i can run at 60 Hz and still have a high quality stacking and constraint solving. plus the response itself is so much more convincing visually (the way the bodies bounce, slide, roll...). this is really impressive, i had forgotten how good a simulation running with Newton look.

few comments about Newton3.1:
- there is not enough doc about the engine. it was enough for me because I used Newton in the past but for a new user it must be very difficult. There used to be an API ref in .chm format which was really useful, with some great information inside.
- few things don't work such as the exact solver (option 0). maybe it would be good to have comments in the code or a file coming with the release describing the issues.
- i used only the C API. I didn't use the dClasses, it didn't seem complete to me plus it's using half the C api half the C++ classes. i actually wrote a c++ interface to use it with Ogre.

i'm going to stick with it it's really a good engine. and hopefully showcase a demo soon.

BRAVO! (and thanks again..)

[Julio Jerez]

Yes the Newton library has being the victim of a conspiracy orchestrated by a bunch of self appointed experts like:
Mr Adrian Boeing from http://www.adrianboeing.com/contact.html, Ms Kenneth Bodin Holmlund from VRlab/HPC2N, Umeå University, Sweden
And some of the sheer leader and ex employee of Havok and Math engine that after losing their jobs with those companies has made Open source version of those companies engine engines.

Basically the strategy is flood the intended with lot of misinformation to proven casual user form even trying the Newton.
This attacks are so ferocious that some of the people has even contacted employers of mine to try to get me fired from my work so that can implement open source physics library.
Basically the adoption of these libraries is no base of the merit, it is base of Networking, and that goes for all Physic libraries with the exception of Newton.
They offer lot of perks and free stuff to companies that endorse those engines, and the decision is made politically by PR, Mangers, and middle to high executives, never made not based on the merit of the technology.

The other day I have an email from video, asking me what they have to do to get Newton engine integrated into a performance tool named PEEL.
There were trying to integrate another version of Newton into that tool and somehow they could not get any results, I have to sign an internet NDA so I cannot publish the results, or the copy of the tool.
So I spend about a week integration Newton 3.09 into that tool. There say they will publish the result If I author then and I say yes of course. In their mine they thought that Physx vs Newton was going to be a Blow out.
There was a blow out right, but not what they were expected, Newton beat Physx 3.3 in all tests. Newton was the faster, more accurate and more stable, an d the most feature completed (not including the particle and sift bodies why was not part of test)
Basically in end up for the rethonk what to do to Physx to match newton perrmance, and acuracy.
After soem for an dback email, the test the send me another Email telling me that unfortunately they were no going to be able to publish the result.
It made be really mad and over reacted and posted one of thier user use forum. And the n I have to react the comment because of legality.
Here are some of the comments they sent me in private.

#1 : Newton seems to be a lot more efficient with compound shapes than PhysX. This is especially noticeable in test #59 and #60
#2 : Newton handles interpenetration much better than PhysX.
#3 : Newton seems to have a much more aggressive sleep algorithm than PhysX; as can be seen in large box stack tests where PhysX falls over but Newton puts them to sleep fairly quickly. This isn't necessarily a good or bad thing; it's usually just a tuning parameter.
#4 : All of the demos involving constraints where there are long chains or, especially the 'nets', the behavior of Newton looks *much* better. The performance is lower, but that's almost for sure simply because the solver-iteration count on the PhysX constraint is too low. I'm sure if we increased the solver-iteration count on the PhysX constraint to make the behavior match more closely to the quality of the behavior in Newton the performance would be much more close.
#5 : The ragdoll demo looks better in PhysX than in Newton. This may be just because you haven't fully hooked up all of the constraint properties yet? The PhysX constraints have a lot of spring/softness to them that tend to make ragdolls look a little nicer.
#6 : In general Newton seems to be very performance competitive with PhysX until you start hitting extreme situations of massive numbers of dynamic objects all in a giant pile. I'm going to assume this is what you meant by an artificial demo. I agree this isn't necessarily that big of a deal.
As we discussed earlier, you have to be careful about drawing too many conclusions too quickly, since many differences could be due to iteration counts or other kinds of tuning parameters.
These are just some things that I have observed so far.
You are putting a lot of work into getting Newton to work with PEEL; I think we should discuss a plan to present the data when we are finished. Perhaps something on either your website or my blog. Something which is diplomatic and complimentary.
In my personal view, Newton is a very impressive high quality physics engine with excellent behavior and performance. The fact that it is even competitive with commercial physics engines that have had dozens of engineers working on them for years is really an incredible testament to your talent and dedication.
I think you have done an amazing job, and we should communicate that to the public as well.
...

But those will never be known because I do have the resurges that companies like Nvidea, Intel, and AMD spend maintaining “lost leaders” product because they think it will sell hardware.
Their strategy is to ignore me or anyone else that could be a real competitor on merits alone.
They promise they will make a seprate report, whe they will mantion Netwon, bu I will not hold my breath on that, knwo those people they wuill
do wh athe Crackpot ex Havok Guy did to me with collada stuff.

[Julio Jerez]

for some of these reason I am starting serieus of Plugin for some of the popular open suorce game engines.

I took Ogre engine as the pilot project but there will be more integrations.
it will be noce it you jsut take a look at what I have so far

viewtopic.php?f=2&t=7662

these plug in encapulate late Newton on a true C++ api dNetwon
and yes not all the C API is exposed but since the classes alowe access to teh Netwon objects, it is posible to have access to the complete C API.
Please take a look, maybe we can join effort rather that work separate.
if you sync for SVN you cna see were teh wraoer is at now.

[speps]

I started working on a wrapper for Unity3D, it's working against Newton 1.53 because I needed proper vehicle physics AND a C API. Getting a good vehicle C API for 3.XX would be great so I can finish my Unity3D plugin

[alezy]

Please get the late update from SVN, google does not allow more stable upload anymore.