I tried using photons in a scene recently and was suprised at how long it
took to render it in 3.5 RC5.  I'm posting in this group in case I'm missing
some subtle change in the photon syntax. The following scene at 320x240, AA
0.3 takes 19 seconds to render in MegaPOV 0.5 and 3 minutes 37 seconds to
render in RC5.

/* #version Unofficial MegaPOV 0.5; */
global_settings {
   photons {
     count 100000
     autostop 0
     }
 }

#declare WATERMAT =
material {

texture {
        pigment {color rgb <1, 1, 1> filter .9}
        finish {
        ambient 0
        }
        normal {bozo scale 1 bump_size 5}
}
interior

        ior 1.33
        }
}

disc {<0, 0, 0>, <0, 1, 0>, 5
material {WATERMAT}

photons{
  target
  refraction on
  collect off      /* 3.5 */
  /*ignore_photons*/ /* mega */
}
}

camera {location <5, 15, -25> look_at -2*y angle 27}

light_source {<0, 0, -100> color 1*<1, 1, 1>
rotate 55*x rotate -170*y
}

difference {
cylinder {<0, -2, 0>, <0, .6, 0>, 6.5
pigment {color rgb <1, 1, 1>}
}

union

cylinder {<0, -3, 0>, <0, 2, 0>, 5}
cylinder {<0, -3, 0>, <0, 2, 0>, 6 inverse}
pigment {color rgb <1, 1, 1>}
}
finish {ambient 0}
}

plane {y, -2
pigment {color rgb <1, 1, 1>}
finish {ambient 0 diffuse .5}
}

Post a reply to this message

"Mike Hough" <Ama### [at] aolcom> wrote in message
news:3cead5cf$1@news.povray.org...
> I tried using photons in a scene recently and was suprised at how long it
> took to render it in 3.5 RC5.  I'm posting in this group in case I'm
missing
> some subtle change in the photon syntax. The following scene at 320x240,
AA
> 0.3 takes 19 seconds to render in MegaPOV 0.5 and 3 minutes 37 seconds to
> render in RC5.

That's interesting. I hadn't made any such comparisons.
Reposting your script below because it lacked a couple braces and maybe
slightly easier for people to try that way.
But also, you hadn't put a photon block into the light source... so I did
and turned reflection off since there isn't any reflective objects present.
Doing a quick 160x120 without AA I had the following results:

Only refraction on (photons {} in light)
Parse = 0s
Photons=3s
Trace=80s

Reflection and refraction (or no photons{} in light)
Parse = 0s
Photons=3s
Trace=82s

Not exactly anything significant. However, I tried again with a
        photons {
                collect off
                }
added to only the union of that cylinder pair which is differenced from the
remainder cylinder and the render time increased 8 fold.
I stopped there because I couldn't figure the logic in your CSG creation.
:-)

________________

/* #version Unofficial MegaPOV 0.5; */
global_settings {
   photons {
     count 100000
     autostop 0
     }
 }

#declare WATERMAT =
material {

texture {
        pigment {color rgb <1, 1, 1> filter .9}
        finish {
        ambient 0
        }
        normal {bozo scale 1 bump_size 5}
}
interior
        {
        ior 1.33
        }
}

disc {<0, 0, 0>, <0, 1, 0>, 5
material {WATERMAT}

photons{
  target
  refraction on
  collect off      /* 3.5 */
  /*ignore_photons*/ /* mega */
}
}

camera {location <5, 15, -25> look_at -2*y angle 27}

light_source {<0, 0, -100> color 1*<1, 1, 1>
rotate 55*x rotate -170*y
        photons {
                reflection off
                refraction on
                }
}

difference {
cylinder {<0, -2, 0>, <0, .6, 0>, 6.5
pigment {color rgb <1, 1, 1>}
}

union
        {
cylinder {<0, -3, 0>, <0, 2, 0>, 5}
cylinder {<0, -3, 0>, <0, 2, 0>, 6 inverse}
pigment {color rgb <1, 1, 1>}
}
finish {ambient 0}
}

plane {y, -2
pigment {color rgb <1, 1, 1>}
finish {ambient 0 diffuse .5}
}

Post a reply to this message

> Not exactly anything significant. However, I tried again with a
>         photons {
>                 collect off
>                 }
> added to only the union of that cylinder pair which is differenced from
the
> remainder cylinder and the render time increased 8 fold.

That's not very encouraging.  Have you tried it with MegaPOV?  It rips right
through the scene when I try it here.

> I stopped there because I couldn't figure the logic in your CSG creation.
> :-)

I had to think about it for a moment myself.  It makes sense when you give
the two parts different textures.
The union cuts away the inside and outside of the first cylinder.

Post a reply to this message

Just occured to me that you meant the render time decreased (sped up).  I
tried the file with the added anti-collection code in the difference.  I
get:

MegaPOV - 18s
RC5 - 23s

Removing photons I get (at 640x480):

MegaPOV - 8s
RC5         - 5s

So RC is almost twice as fast without photons.  At that resolution with the
original code it's:

MegaPOV - 34s
RC5          - 12m 2s

Over 20x slower

Go figure...

Post a reply to this message

Mike Hough wrote:

> Just occured to me that you meant the render time decreased (sped up).  I
> tried the file with the added anti-collection code in the difference.

I am a bit taken aback by the silence on this issue. Anyone can try this for
themselves. Run the file "optics.pov" from the scenes/advanced folder under 3.5
RC5 and watch your system slow to a crawl. Now try it under MegaPOV 0.7. The
file screams by in a small fraction of the time. The output quality will show a
bit brighter media in RC5, but that is the only difference, other than the
exponential time increase.

Post a reply to this message

"Dave Dunn" <poi### [at] aolcom> wrote in message
news:3CE### [at] aolcom...

>
> I am a bit taken aback by the silence on this issue. Anyone can try this
for
> themselves. Run the file "optics.pov" from the scenes/advanced folder
under 3.5
> RC5 and watch your system slow to a crawl. Now try it under MegaPOV 0.7.
The
> file screams by in a small fraction of the time. The output quality will
show a
> bit brighter media in RC5, but that is the only difference, other than the
> exponential time increase.
>

Testing....

Confirmed (2 minor changes to scene file for MP7).

win95 p2 233 (iirc)
optics.pov under 3.5 RC5 MSVC
30m30s
optics.pov under MP7
5m46s

Post a reply to this message

"Tom Melly" <tom### [at] tomandlucouk> wrote in message
news:3cf08542@news.povray.org...
>
> Confirmed (2 minor changes to scene file for MP7).
>
> win95 p2 233 (iirc)
> optics.pov under 3.5 RC5 MSVC
> 30m30s
> optics.pov under MP7
> 5m46s

Yeah gee, that's a major slowdown. Admittedly, I was skeptical when Mike
first told of it but that decrease in rendering time is extremely obvious
and obviously not witnessed by one person.

Guessing (wild guess at that) perhaps one of the more recent 'automated'
parts Nathan had worked on caused this. Or could it be that it goes back to
an earlier time? I hadn't noticed any change in render time regarding
photons or photons + media; but last time I was running MegaPOV for any
reason on here was probably about 4 months ago, before a reformat and OS
reinstall. POV 3.1 and MegaPOV 0.7 are still on CD-R.

bob h

Post a reply to this message

In article <3cead5cf$1@news.povray.org> , "Mike Hough" <Ama### [at] aolcom>
wrote:

> I tried using photons in a scene recently and was suprised at how long it
> took to render it in 3.5 RC5.  I'm posting in this group in case I'm missing
> some subtle change in the photon syntax. The following scene at 320x240, AA
> 0.3 takes 19 seconds to render in MegaPOV 0.5 and 3 minutes 37 seconds to
> render in RC5.

POV-Ray 3.5 is *not* MegaPOV, in particular not such an old version of
MegaPOV.  You cannot just copy you scene one to one an expect it to work
exactly the same way.  There might be a million things wrong in your scene and
none has to do with POV-Ray 3.5.  Or you (accidentally) enabled any of a
number of other features in your scene that did not exists or were not enabled
by default in MegaPOV versions or POV-Ray 3.1.


    Thorsten

____________________________________________________
Thorsten Froehlich
e-mail: mac### [at] povrayorg

I am a member of the POV-Ray Team.
Visit POV-Ray on the web: http://mac.povray.org

Post a reply to this message

Thorsten Froehlich wrote:

> >POV-Ray 3.5 is *not* MegaPOV, in particular not such an old version of
> >MegaPOV.  You cannot just copy you scene one to one an expect it to work
> >exactly the same way.  There might be a million things wrong in your scene and
> >none has to do with POV-Ray 3.5.  Or you (accidentally) enabled any of a
> >number of other features in your scene that did not exists or were not enabled
> >by default in MegaPOV versions or POV-Ray 3.1.

I am trying to resist saying "Yeah, MegaPOV's photons are 20 times faster," or
some such other sarcastic remark, because I haven't done enough testing. And
indeed, if Mike's scene had been the only one causing this condition, using only
Megapov 0.5, I'd say you have a point there; however, the slowdown has been
confirmed using a file distributed in the scenes/advanced folder of RC 5
(optics.pov), using the final version of MegaPOV (0.7). It may be, as you say,
some "accidental" enabling of features. But it could also be a serious problem
with the speed of photon mapping in the current release. Dismissing experimental
evidence out of had without at least running a few tests is hardly scientific.

Post a reply to this message

"Thorsten Froehlich" <tho### [at] trfde> wrote in message
news:3cf0a045@news.povray.org...

> POV-Ray 3.5 is *not* MegaPOV, in particular not such an old version of
> MegaPOV.  You cannot just copy you scene one to one an expect it to work
> exactly the same way.  There might be a million things wrong in your scene
and
> none has to do with POV-Ray 3.5.  Or you (accidentally) enabled any of a
> number of other features in your scene that did not exists or were not
enabled
> by default in MegaPOV versions or POV-Ray 3.1.

... which is presumably why this discussion is taking place in general
rather than beta-test. I do think it deserves further investigation from
someone who knows what they're doing (i.e. not me).

Post a reply to this message