I agree with you, crunching floating points is and always will be an
issue even for the most sophisticated processors.
The demand for "physically accurate" simulations will always be more
than what the cutting edge technology can offer.
On 11/1/2013 11:46 AM, Ponthieux, Joseph G. (LARC-E1A)[LITES] wrote:
I completely agree with this. However I would like to add that any
lack of performance regarding high object count is probably less an
issue with the 3D app than an inherent processing limitation within
most high end graphics systems.
We see this same behavior in other applications including very
expensive high end simulation software. At a basic level what most
apps appear to be doing at each iteration, frame change, camera
viewport change, other update, etc is cycling through each root level
object and checking for updates or mods that need to be applied per
the methodology within that application. One way you could think about
this is that it starts at the top of the explorer(outliner in maya)
and repetitively cycles through the explorer list over and over and
over applying updates as it goes and as needed.
We have scenes, simulations, and scenarios where we have processed
hundreds of thousands of aircraft for example and these kinds of
simulations can bring the most powerful computers to their knees even
though each individual object is as simple as possible. It's been my
personal opinion that the operating system and how it breaks things
down for processing between core and interface may also be part of the
issue. I've often wondered if it was an integer vs floating point
issue. Even more shocking was that we often saw the same results on
SGI and Windows simultaneously back when we were still using those
beasts.
One of our simulation software for example had a special flag that
turned off the interface refresh and we would get all the performance
back. They also provided us a special object container that acted like
a mini scenario holding hundreds of thousands of objects within a root
level multi-object. When using this special multi-object performance
never skipped a beat but getting access to or from the objects within
the root object was no longer as direct. The point being that if you
have unreasonably high object counts, you might be able to break your
scene down into a handful of Models with the objects distributed under
the Models. I've never tried this with Soft and don't know if this
would significantly affect Soft's performance, but it might be worth a
shot.
--
Joey Ponthieux
LaRC Information Technology Enhanced Services (LITES)
Mymic Technical Services
NASA Langley Research Center
__________________________________________________
Opinions stated here-in are strictly those of the author and do not
represent the opinions of NASA or any other party.
*From:*[email protected]
[mailto:[email protected]] *On Behalf Of *Alan
Fregtman
*Sent:* Friday, November 01, 2013 11:09 AM
*To:* XSI Mailing List
*Subject:* Re: Extracted meshes and performance
Historically Soft has dealt quite well with few objects with intense
topology much better than thousands of low-res objects.
--
--
ALOK
GANDHI
/ directeur technique senior- senior technical director
[email protected] <mailto:[email protected]>
T:
*450 430-0010 x225
F:
*450 430-0009
www.modusfx.com <http://www.modusfx.com>
---------------------------------------------------------------------------------------------
MODUS
FX
120 Rue Turgeon,
Sainte-Therese (Quebec) CANADA J7E 3J1
Follow us on
Facebook <http://www.facebook.com/ModusFX>
&
Twitter <https://twitter.com/Modusfx>
**
