pirateprincess wrote:
> I have been working with .POV scenes generated from PyMOL (PDB crystal
> structure files).  I am trying to tesselate a specific crystal structure
> image across a plane and have utilized 15-20 copies of a simple DNA double
> helix.  The .POV file is roughly 30 megabytes.  I am working on a quad-core
> G5 (Mac OS X 10.4.9, Pov-Ray 3.6.1 for Mac) with 4GB RAM and am receiving an
> error that reads "Parse Error: Out of memory.  Cannot allocate 116 bytes for
> finish.  File my_scene.pov line 665798"
> 
> Are there simple ways to fix this problem?  Is more information needed to
> help diagnose the problem / suggest a solution?

The 4 GB of RAM won't help a GUI application such as POV-Ray as those
applications cannot use more then 3 GB of memory each, regardless of how
much memory your system has.

To get over this limit, you need a 64 bit command line application, which
you would have to compile yourself based on the Unix source code of POV-Ray.
Apple includes all development tools you need in addition to the POV-Ray
source code (but the development tools are not installed by default).
However, it isn't trivial if you have never done this before.

    Thorsten, POV-Team

Post a reply to this message

"Christian Froeschlin" <chr### [at] chrfrde> schreef in bericht 
news:464ca294$1@news.povray.org...
>
> It would probably help to see an *excerpt* of your generated code
> with the bulk of the data removed so we see the object structure.
> AFAIK, if it is a mesh then additional copies should not take
> extra memory, but if its other objects that may be different.
>

Yes, I second that! So we can see if the structure is built up out of 
spheres and cylinders, or if it is built up out of mesh2 objects.

Thomas

Post a reply to this message

nemesis wrote:

> well, forget it:  CSG really turns povray into a memory hog no matter what
> you do.  It has gone over 200MB as well and I halted again.  It just took a
> lot more time to get there.  Important lesson here:  better few unions with
> lots of objects than lots of unions with fewer (but huge!) objects...

Well, the real killers seem to be the individual pigments on
your 300.000 objects. The stats on my system for your scene 1:

Largest  Alloc:             2400016 bytes
Peak memory used:         832856013 bytes

Applying the same pigment for all objects in the outermost union:

Largest  Alloc:             2400016 bytes
Peak memory used:         374455173 bytes

Actually, I usually prefer to use a blob object for rendering a
lot of spheres if I can get away with it, because it seems to render
a lot faster, especially with intersections such as water bubbles.
Blobbed version also helps memory (but yields domed cylinder caps):

Largest  Alloc:            60000016 bytes
Peak memory used:         152359044 bytes

I intentionally ignored the parse time: For the first version
most of the time was spent swapping, for the other versions the
times are similar and probably dominated by the SDL text
parser and not by efficency of data structures.

Post a reply to this message

"pirateprincess" <pir### [at] charternet> wrote:

> I only have the code for the DNA written out once, declared as an object as
> you have described below. Then to get the full picture, I have unions of
> various numbers of the helix declared as seperate objects. This has allowed
> me to work with only one or two of the helices at a time in order to get
> them the way I want them. But now that I am trying to view my final scene
> with all the copies, it is running out of memory.
>

Depending on the rest of the scene (shadows and reflections would make the
following technique more difficult) you could render the individual helices
using a transparent background (+UA command line option) and then combine
the images using a photo editor.

Mike S.

Post a reply to this message

Mike Sobers nous apporta ses lumieres en ce 2007/06/01 01:11:
> "pirateprincess" <pir### [at] charternet> wrote:
> 
>> I only have the code for the DNA written out once, declared as an object as
>> you have described below. Then to get the full picture, I have unions of
>> various numbers of the helix declared as seperate objects. This has allowed
>> me to work with only one or two of the helices at a time in order to get
>> them the way I want them. But now that I am trying to view my final scene
>> with all the copies, it is running out of memory.
>>
> 
> Depending on the rest of the scene (shadows and reflections would make the
> following technique more difficult) you could render the individual helices
> using a transparent background (+UA command line option) and then combine
> the images using a photo editor.
> 
> Mike S.
> 
> 
OR, you can do a final render using those as image_map on very thin boxes or 2 
triangles mesh or planes.

Alain

Post a reply to this message