POV-Ray : Newsgroups : povray.beta-test.binaries : Function / pattern issues. Povr supertoroid parametric. : Re: Function / pattern issues. Povr supertoroid parametric. Server Time
16 May 2024 17:01:49 EDT (-0400)
  Re: Function / pattern issues. Povr supertoroid parametric.  
From: William F Pokorny
Date: 21 Jun 2020 11:13:57
Message: <5eef7935$1@news.povray.org>
On 6/21/20 5:14 AM, jr wrote:
> hi,
> 
> William F Pokorny <ano### [at] anonymousorg> wrote:
>> ....
>> Dump the inbuilt benchmark. ...
> 
> popped into my head this morning.  :-)  occurs to me that not only is having
> this benchmark "nice", even if only for own information, it is also, in a sense,
> a convenient and built-in 'make check' equivalent.  so I think that unless
> there's an explicit 'make check' which exercises the build, the '--benchmark'
> (too) ought to be kept.
> 

There is today an explicit 'make check' which uses the advanced 
directory biscuit scene(1). I'm thinking about changing 'make check' to 
something much simpler like a sphere on a checkered plane. I don't think 
there is any reason for a more complete scene for the 'make check' 
functionality.

(1) - Running in a mangled way for my povr compiles at present.

The current inbuilt benchmark amounts to wrappers around the scene 
directory benchmark scene and ini files. Plus, it looks to me, 4 font 
files which exist in the include directory get wrapped too. These are 
the inbuilt fonts text{} can use. I think the whole idea is to let the 
benchmark types at whatever news publication run the benchmark without 
an actual install. The wrapped/c++ encapsulated stuff gets written out 
to files for the --benchmark. Not sure where written at the moment. I've 
looked at little at the encapsulation side, but not how it actually runs.

In any case, as done it amounts to a significant duplication of files. 
Nothing really against it, other than it being something more which must 
be maintained and that it takes around 700K (6.5% of povr today) of 
source code space.

Aside: The current benchmark scene could be made more representative - 
or perhaps their should be several benchmarks. Last I profiled the 
benchmark was, time wise, 65% or more noise(NG 1) and media.

Bill P.


Post a reply to this message

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