Hello Georg,
Georg Wünsch wrote:
> Yes I am concerned about precision. Of course rather with transformations
> and other background calculations than with the geometry data.
ok. Do you have particular application(s) in mind where you currently
encounter limitations with float precision or is this more a general
interest question -- just being curious what you are working on ;)
> But I am just
> wondering where the journey is going in terms of graphics hardware. I would
> expect the graphics hardware to be operating on 64Bits with all data in the
> long run, too - please correct me if I am wrong. I would also expect the use
> of floats on hardware that is (or will be) designed for doubles to become a
> bottle neck in the future.
I don't have any particular insights into where graphics hardware is
heading, so this is just my guess on things:
I would expect any (compute performance) penalty for using doubles to
disappear in the future, however I'm not convinced that using doubles
will become the "default". My impression is that the gaming industry
does not and has no intention of using doubles and I think that is
pretty much the only industry the hardware vendors really cater to these
days.
So while the whole GPU compute drive puts a demand to properly support
doubles without penalty on the hardware vendors I don't see it become
the default for *graphics* in the near future.
> E.g. for an application, which's core has 64 Bit precision and OpenSG is
> used to do the rendering, takes mainly a lot of transformations and
> sometimes new mesh data.
>
> Every time a 64Bit precision transformation is copied to the 32Bit OpenSG
> system and later to the graphics hardware, everything needs to be converted
> to float. Will this lead to performance losses at the point where simulation
> data crosses from 64 to 32 bit?
OpenSG 2 can be compiled to use doubles for its matrix stack and only
convert to float when passing transforms to OpenGL, it could probably be
taught to us glLoadMatrixd instead to actually upload double data to
OpenGL, but as far as I understand it that would only move the
conversion into the driver.
> Can OpenGL handle mesh data in floats transform them with transformations
> given in doubles?
AFAIK yes, well sort of at least ;)
You can set matrices with double entries, but they get converted to
float internally and I don't think there is a way to choose the
precision of OpenGL internal computations.
Cheers,
Carsten
------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit. See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
Opensg-users mailing list
Opensg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensg-users
