POV-Ray : Newsgroups : povray.newusers : Should Parsing really take that long? How to speed it up? : Re: Should Parsing really take that long? How to speed it up? Server Time
3 Oct 2023 07:17:41 EDT (-0400)
  Re: Should Parsing really take that long? How to speed it up?  
From: clipka
Date: 4 Dec 2015 11:56:47
Message: <5661c5cf$1@news.povray.org>
Am 04.12.2015 um 14:59 schrieb ILM:
> Hi there,
> first of all: some background informations. I not a designer or graphics artist.
> I am working with an simulation software that has an option to export the
> simulation as a video. Therefor it creates 25 pictures per second and saves them
> as .pov.
> For me Pov-Ray is a blackbox. I put in the .pov files and get out an .jpg.
> My problem is, that scaling up the cpu count of my render machine the pure
> render time decreases. BUT parsing still takes "forever". In my case parsing
> takes about 10 minutes per picture while the render process takes less than a
> minute.
> Is this a normal phenomen?

Parsing times of 10 minutes are not unheard of; it all depends on how
complex the .pov file is.

I guess that the best way to speed up parsing times is:

(1) Put the .pov files on the fastest drive you have available
(preferably an SSD) and render them from there.

(2) Run multiple renders in parallel. You will need to make sure that
"Options -> Keep Single Instance" is turned off, and then open multiple
POV-Ray windows. (Keep an eye on memory consumption!)

Also, make sure that input images don't have an excessively high
resolution, and that the geometry you're rendering does not have an
excessively high level of detail.

That said, the piece of software you're using to generate the .pov files
may be doing a poor job at creating good, fast-parsing .pov files.

Most notably, POV-Ray provides two different ways of representing
meshes: The older "mesh" format, in which the vertex coordinates and
normal vectors of each triangle are specified directly as explicit
coordinates; and the newer "mesh2" format, in which the vertices and
normal vectors are instead specified indirectly via indices into a table
of the actual vertex and normal vector coordinates. The latter parses
noticeably faster, especially for large meshes -- not only because the
notation is considerably more compact, but also because it is closer to
the data structure POV-Ray uses internally, and therefore some
preprocessing steps can be skipped.

Also, if multiple objects use the same texture properties, maybe the
generating software explicitly writes those properties over and over
again, instead of defining those properties only once and then just
referencing them wherever they are used -- which not only speeds up
parsing time, but can also reduce memory consumption during the render.
This is especially true if the texture properties include input image
files like image maps or bump maps.

> While parsing the console all over shows the following message:
> "Possible Parse Error: PNG input image default gamma handling has changes for
> pre POV-Ray 3.7 scenes; results may diffe from original intention. see the
> documentation for more details."
> Would it maybe help even help to stop logging this error to the console every
> time? If yes how can i do this? Or how can I solve the error itself?

This is unlikely to have any major impact on parsing time: It is output
at most once per PNG input file read from disk, and reading a PNG file
typically takes considerably more time than writing the log message to
the console.

However, if you really get spammed by this message even though you only
have a comparatively small number of input images, it would be an
indication that there's room for optimization in your software that is
generating the .pov files.

Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.