[Bf-committers] BIG_MESH

[Chris Want]

> I know this was on the wish list just a long time! :)

Yep!

> The matter is not really whether or not, but more 'how'... just  
> changing the DNA_mesh_types.h file won't likely be sufficient, so for  
> that reason I hoped to see the actual patch.

Key word here is 'likely' ... I have been running with
BIG_MESH for a couple of weeks and haven't hit a single fault.
I have done little more than just change the MFace struct,
with some cuteness so that you can switch back and
forth between big and small mesh with different
compile flags. Since we do have this cuteness, a nice
strategy might be to add the support for big mesh as a
compile option like it is in tuhopuu, setting it to default
when we are reasonably convinced any problems have been
ironed out.

Here is the patch:

http://bebop.cns.ualberta.ca/~cwant/big_mesh.diff

Plus 2 files must be added to make the cuteness work:

http://projects.blender.org/viewcvs/viewcvs.cgi/*checkout*/tuhopuu2/source/blender/makesdna/DNA_facebig.h?rev=1.1&cvsroot=tuhopuu

http://projects.blender.org/viewcvs/viewcvs.cgi/*checkout*/tuhopuu2/source/blender/makesdna/DNA_facesmall.h?rev=1.1&cvsroot=tuhopuu


> - all over Blender code indices are assumed to be shorts. Was this  
> verified & fixed?

Not verified. Well ... semi-verified in the sense that I
haven't had any faults in the last few weeks. I have had no
problems loading old blends, and I have had no problems loading
blends created with this version in old versions of blender.

> - same for MAX #defines that go around.

#defines? I thought all numbers in blender were hard coded >=P
I just grep-ed for 65000 ...(which was hard coded in a few places.)

> - our importers/exports? python scripts?

Not checked. I would appreciate if somebody who runs a lot
of scripts could test with the big mesh tuhopuu.

> - some modes will fail faster now (VertPaint, WeightPaint, FaceSelect)  
> for systems using 16 bits display. The backbuffer selection method  
> stops with 65k faces here. Would be good to find a method to warn here.

I am very unfamiliar with this code ... I would appreciate it if
somebody with a big mesh tuhopuu could test this.

> - files will become larger, for 200k faces that means 2 Mb extra. Not a  
> *real* problem, we can assume that since 1995 the disk sizes have grown  
> some. :P

Well, minus the overhead of storing the 200k faces in 4 different
meshes.

Thanks!
Chris