ice springs simulation performance
Sebastian Kowalski
im testing the ice springs, and i love em.
but simulating an "ice softbody" sphere with 86 points, is giving me 7 fps. and cpu is runing at 14 % on a 3 ghz dual quad.
ive went with the setup explained in the docs, getting neighbour points indices to fill up the connectionIndicies, guess thats the bottleneck.
could someone contribute some speed up tipps on that?
thanks
sebastian
Eric Thivierge / XSI Database
disc type object has 178 point and I got an average of 18 fps with
this setup. If you subdivide (add more real edges) it can slow to a
crawl. There are still some issues with my setup where parts cave in
and don't pop back out.
I'm using Julian Johnson's Neighbor Points Plus compound which allows
you to grab not only the neighbors but their neighbors as well which
gives a bit more stability to the simulation.
In the scene you see the light green dots are the neighbors that the
selected point is attached to and the maroon colored point across from
it is a point that it attaches to as well.
--------------------------------------------
Eric Thivierge
Technical Director
http://www.ethivierge.com
Eric Thivierge / XSI Database
attached unzipped so it didn't go through. The last mail had my reply
as quoted so you may have to view the quoted text if using gmail for
my reply.
Cheers,
--------------------------------------------
Eric Thivierge
Technical Director
http://www.ethivierge.com
Sebastian Kowalski
i guess the performance is weak, cause of the large arrays per point.
but it seems that it wont multithread at all, what makes all that
processing power senseless.(cant notice the physx bump either)
i hope we doing something wrong. im in need of a solid spring system.
sorry for my bad english..its monday morning.
-sebastian
Eric Thivierge / XSI Database
of the spring systems could help. If you can, I'd send your setup into
support and see if they can see any reason its not multi-threading.
Didn't notice that it wasn't myself.
--------------------------------------------
Eric Thivierge
Technical Director
http://www.ethivierge.com
Sebastian Kowalski
"Create meshless deformation simulations in ICE with the latest NVIDIA® PhysX® rigid body library. With new support for springs and dampers, you can more easily achieve a wide range of effects, such as jelly-like and plastic deformation. The new library also offers accelerated performance, enabling artists to focus on the creative process when working with dynamic simulations. Even faster performance is possible with the optional addition of an NVIDIA® CUDA enabled GPU."
quote form nvidia:
"Autodesk Softimage (built into Softimage, no download required)"
so actually, how could i get this to work? everything phsyx related is active, and i see no speed bumps
attached a test scene.
on my machine there is running only one core (of 8, 3 ghz. and a 4800 quadro card), getting with both mesh and particle deform 2.3 fps.
checked on winxp 64 and win 7 64, latest drivers.
Nassos Yiannopoulos
enable the "Show PhysX Visual Indicator".
what this does is print on the top left corner of every OpenGL viewport
the type of PhysX acceleration that is used (CPU/Hardware)
Although i haven't tried it with the built-in ICE nodes of 2011
(currently building my own PhysX ICE toolkit)
but it might give you a hint on what is happening
and BTW only Softbodies/Cloth/Fluids
and basic RBD are hardware accelerated,
the rest is done by the CPU
2010/4/13 Sebastian Kowalski <li...@sekow.com>:
Sebastian Kowalski
ive tested the nvidia physx sph demo, and it runs great. so physx runs
on my machine.
im getting headaches . . .
and why is my cpu performing so bad? is the simulate rigid bodies node
singlethreaded?
thanks
sebastian
Nassos Yiannopoulos
or you reached a limit after which it uses the CPU
or they forgot to enable Hardware solver switch
hardware softbodies are noticably faster
than cpu ones, though
2010/4/13 Sebastian Kowalski <li...@sekow.com>:
> arg, found it.
> and in soft i see physx > cpu . . . . .
> also on softbodies
Nassos Yiannopoulos
2010/4/13 Ciaran Moloney <moloney...@gmail.com>:
> Someone correct me if I'm wrong, but non-ICE physx hardware acceleration was
> never supported. Only software AFAIK. I'm not sure about 2011, though...
>
> On Tue, Apr 13, 2010 at 12:48 PM, Sebastian Kowalski <li...@sekow.com> wrote:
>>
>> "or they forgot to enable Hardware solver switch"
>>
>> that would be bad.
>>
>> it doesnt matter what im doing, ice or non ice, rigid or softbodies..its
>> always physx on cpu.
>> but the crawling cpu performance is what gives cause for some serious
>> concern. got to do a full character with some hundreds colliding instances
>> on it. hoped to use ice for it.
>>
>> -sebastian
Sebastian Kowalski
that would be bad.
it doesnt matter what im doing, ice or non ice, rigid or softbodies..its
always physx on cpu.
but the crawling cpu performance is what gives cause for some serious
concern. got to do a full character with some hundreds colliding
instances on it. hoped to use ice for it.
-sebastian
Am 13.04.2010 18:24, schrieb Nassos Yiannopoulos:
Sebastian Kowalski
and in soft i see physx > cpu . . . . .
also on softbodies
Am 13.04.2010 18:00, schrieb Sebastian Kowalski:
Ciaran Moloney
christian keller
christian keller
visual effects|direction
+49 0179 69 36 248
chr...@gmx.de
Am 14.04.2010 09:04, schrieb Sebastian Kowalski:
> jep the 2011 version
Sebastian Kowalski
Am 13.04.2010 19:15, schrieb Nassos Yiannopoulos: