Windows 10 and Visual Studio 2015

A place to discuss everything related to Newton Dynamics.

Moderators: Sascha Willems, walaber

Windows 10 and Visual Studio 2015

Postby zak » Sat Jan 30, 2016 7:02 am

Hi,
my all physics soccer game is 90% ready now ( just a couple of months and i hope it will be the Newton killer app :twisted: )
Now i am on 2.12 and Visual Studio 2008 / Win32 but my intent is publishing the game for Windows 10 in Windows store as a Windows Universal App, using free Visual Studio 2015 Community Edition that has ALL the functionalities of professional version.

I'm going to update my pc to Windows 10 and my game from Win32 to WinRT (Windows Universal Platform) and from Visual Studio 2008 to Visual Studio 2015 CE .

Is Newton 3.14 compatible with Windows Store and Visual Studio 2015 ?

Thank you.
zak
 
Posts: 87
Joined: Mon Dec 06, 2004 9:30 am

Re: Windows 10 and Visual Studio 2015

Postby Julio Jerez » Sat Jan 30, 2016 10:24 am

I compiled newton 3.13 and 3.14 with VS 2015, but I have no tested window 10.
I do not see why it would not work, there is very lithe dependency between newton and the OS,
the only one I one think of is the mutt reading which is done via pthread.

3.12 has some bug and I lost everything since that was when goggled code stop code submission an the last check in has some bug, so I do no know what version is stable in for 3.11 to 3.13
3.13 is the most stable.

I recommend you fist get the zip version of newton 3.13 from gighub, and upgrade to that. once you get that going getting to 3.14 is straight forward and optional, there are bug fixes and new feature but, but it is still in progress.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: Windows 10 and Visual Studio 2015

Postby Julio Jerez » Wed Feb 03, 2016 5:46 pm

Oh I forget to mention, very important, if you are using rag dolls, and other articulated objects, after you convert to newton 3.13
then you can try newton 3.14, there are two new features.
1- skeleton style articulated joints, this make the joint rock solids, you object do no dismember, not withstanding numerical errors. All joint can be set that way, eh only restriction is that the Skelton has to satisfy the condition that each node has only one parent.
The feature is even more correct that what the exact solve use to be. Check out the articulated joint in the demos. there is a track vehicle that work in real time and the track are stable as long as not other body get wedged between the track and the tires.

2-aggregated, the feature let you group set of object that you will always be closed to each other
ex: a rag doll, and it will place a proxy on top of then that will be seen as a single node on the broad phase. This prevent the test of lots of bodies again lot of bodies example tow rage dual that the top aabb do not tock do not need to do checking on each of the children node.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: Windows 10 and Visual Studio 2015

Postby zak » Thu Feb 04, 2016 2:33 pm

OK, thank you.
I will keep you updated
zak
 
Posts: 87
Joined: Mon Dec 06, 2004 9:30 am

Re: Windows 10 and Visual Studio 2015

Postby XycsoscyX » Fri Feb 05, 2016 1:11 pm

I'm using VS2015 and Windows 10 without much issue, I still seem to run into the pthread error when compiling Newton though (3.14). I know a fix was checked in to define HAVE_STRUCT_TIMESPEC, but when I just try to build as is I still get compiler errors about it. Quick fix which I've been using was the below fix, but I haven't looked much more into it:

[code]
--- a/coreLibrary_300/source/pthreads.2/pthread.h
+++ b/coreLibrary_300/source/pthreads.2/pthread.h
@@ -118,6 +116,12 @@
+#ifndef HAVE_STRUCT_TIMESPEC
+ #if _MSC_VER >= 1400
+ #define HAVE_STRUCT_TIMESPEC
+ #endif
+#endif
[/code]

Other than that, I've had no issues with Windows 10 or VC2015 and Newton. :D
XycsoscyX
 
Posts: 12
Joined: Wed Dec 15, 2004 2:46 am
Location: WA, USA

Re: Windows 10 and Visual Studio 2015

Postby zak » Wed Oct 19, 2016 6:20 am

After a pause i can now work again on my soccer game.

Now i am on windows 10 and Visual Studio 2015 ( from windows Vista, Visual Studio 2008 and Newton 3.12)
Julio Jerez wrote:I recommend you fist get the zip version of newton 3.13 from gighub, and upgrade to that. once you get that going getting to 3.14 is straight forward and optional, there are bug fixes and new feature but, but it is still in progress.

I have compiled my program as Win32 with newton 3.13 and 3.14 and all is right.
I havent yet compiled as UWP (universal windows app for windows store).

But, i have noted a big difference from 3.12 to 3.13/3.14 in exact solver.
While in 3.12 exact solver was very precise, more than iterative, in 3.13/3.14 it has lost all his precision.

In my joint stress test, iterative 3.13/3.14 has the same values as 3.12, while exact has a very soft behavior.
In 3.13/3.14 exact solver is much less accurate than iterative.

I think something went wrong from 3.12 to 3.13 in exact solver.
zak
 
Posts: 87
Joined: Mon Dec 06, 2004 9:30 am

Re: Windows 10 and Visual Studio 2015

Postby Julio Jerez » Wed Oct 19, 2016 10:11 am

there has been lot of changes in 3.13 and 3.14 that has no reflect to the exact solver.
I am working in something in an update the will make the exact solve practically unnecessary.

basically the method will make that any contraction made of joint will always has exact solution and the contact will use the current iteration.
Is started that about 9 moth ago and has to stop for personal reason, but I resume recently again.

then I will add a new version of the exact solver that will use Lemkel or Danzig direct method instead of the, Zdenek Dostal gradient descend algorithm, that was use in 3.12 an lower.
This solver is already in by and I use for some functionality but is no use as a general solve yet.

Anyway after few more days when I complete the Solver for articulation that will be the best of all, because the solver for articulation is always exact, and faster that all other solvers.

so what will get is fast and strong joints 100% of the time, and constraint violations will be projected to collisions. but since collision are not a major problem I think this will be the best of all world.

mean time do no use the exact solver.

may I ask what are you doing that required exact solver, maybe we can work on testing the newer solver.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: Windows 10 and Visual Studio 2015

Postby zak » Wed Oct 19, 2016 11:42 am

wow, a new exact solver and a super articulation solver. I can not wait to try them.
I am developing a soccer game for ten years. It is based deeply on physics. Players are ragdolls that move their legs and synchronize the movement so as to actually hit the ball. No default animation.
No "magic magnetic forces" moves the ball when a player approaches as in actual games. It is always free and will only move if the player's foot hits. It's a crazy project, but step by step is coming to reality.
For the player to move the legs (for example, shortening or lengthening the pace when necessary) in order to hit the ball I have developed sophisticated algorithms. The values returned by these algorithms must be executed by the joints of the ragdoll. If these are not accurate the foot will not hit the ball at the point and at the right time. In this ten years i have developed several physics tricks to "help" ragdoll to moves in correct manner but solver precision remain fundamental. Until now I used exact but I got good results with iterative also. I had not yet decided which one to use in the final product ( waiting for new solvers).

So i have great experience of exact solver and i have noted immediately that it no longer works well in 3.13.
Julio Jerez wrote:there has been lot of changes in 3.13 and 3.14 that has no reflect to the exact solver.
But even if it will be replaced and as it should work as usual would be interesting to understand why it does not because there may be a bug somewhere that could also influence your work on the new solver. Substantially the defect consists of an extremely soft behavior that involves disarticulation whereas before it was solid as a rock.
zak
 
Posts: 87
Joined: Mon Dec 06, 2004 9:30 am

Re: Windows 10 and Visual Studio 2015

Postby Julio Jerez » Wed Oct 19, 2016 1:54 pm

In this ten years i have developed several physics tricks to "help" ragdoll to moves in correct manner but solver precision remain fundamental

oh I see, I believe this articulation solver is what you need. this solve will calculator the correct forces.
It will still have the normal precision problems characteristic of that reduced coordinate system solvers, but it will always produce the correct forces.
part of that is already in by I never has the time for completed, I will this weekend hopefully.

on the al gradient descend solver, not the best to do is to replace it, debugging is will be a waste of time since we know that is too slow for complex systems.

in fact there replacement solver is in te engione for a long time now the function name is
Code: Select all
template <class T>
bool dgSolveDantzigLCP(dgInt32 size, T* const matrix, T* const x, T* const b, T* const low, T* const high)


and there is a version of the dMath library named
Code: Select all
template <class T>
bool dSolveDantzigLCP(int size, T* const matrix, T* const x, T* const b, T* const low, T* const high)


what I do is that I use then to as debug tool to measure how accurate the iterative solver is, it does have real real time value.
But like said after I am done wit the new functionality I will replace the gradient descend solve with the danzig because there are some app that do not simulate large scenes and could use direct solve for very high accuracy. again My goal is that the iterative solution is good enough that the exact is no needed.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: Windows 10 and Visual Studio 2015

Postby Julio Jerez » Wed Oct 19, 2016 2:00 pm

here is something you can start using right away, look at the demos
../applications\demosSandbox\sdkDemos\demos\StandardJoints.cpp

they use the new solver, there are more advanced demos but that's the introduction to how to get started with NewtonSkeletonContainerCreate interface.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles


Return to General Discussion

Who is online

Users browsing this forum: No registered users and 18 guests

cron