[Bf-committers] Re: [Bf-blender-cvs] CVS commit: blender/source/blender/blenkernel/intern exotic.c blender/source/blender/makesdna DNA_mesh_types.h blender/source/blender/makesdna/intern makesdna.c blender/source/blender/src editmesh.c editobject.c

[Douglas Bischoff]

On Dec 16, 2003, at 09:02, Ton Roosendaal wrote:

> This excluding shadowbuffers, texturemaps, particles, etc! When we 
> move to a true 64 bits version of Blender (which is already possible 
> on SGI for ages) we can easily go to:

This puts me in mind of 3 questions I've had for awhile:

1) What, specifically, is necessary to enable 64 bit advantages on 
Blender? Given that Mac G5's are 64 bit machines, can that 
addressability be simply "turned on" and away it goes?
2) Multiprocessor support: can or does Blender recognize the existence 
of a 2nd or more processor? If so, what does it do with it? I suspect 
it doesn't, and that it really couldn't do much with it unless perhaps 
specific parts of the renderer were "farmed" to different processors 
then recombined... would it really be worth it? Being non-threaded (I 
think?) it seems unlikely that multiprocessor support is in our future.
3) Vector math: PowerPC chips, and I think some other architectures, 
support Vector math on dedicated subprocessors and/or optimized 
routines (AltiVec, et. al.). Is this at all useful to us? Can vector 
math be applied to any parts of Blender to enable this specialized 
functionality where it exists? I confess that I don't know what it is 
useful FOR (aside from... uh... vectors...) but is there any thought to 
taking advantage of this? If there is no vector optimization available 
on a particular machine I presume it would simply run the same... yes?

Just some thoughts from out in left field,

-Bischofftep