i am having problems continuing a trace which uses radiosity.

I have specified Continue_Trace=on and +X10 in  povray.ini
and am using output option +FT
povray 3.5 for windows 
amdk6-2 400hz 256mb ram
rendering contains around a hundred area lights and the main light
source.  there is a lots of glass also. I am rendering at 1024 X 768
aa 0.03 with the default radiosity settings
when I restart the render the radiosity pretrace starts over from the
begining. it takes over two hours to do the pretrace alone.
it took over 17 hours to render 60 lines and I don't want to start
over until i can find a way to resume rendering if it is interupted.

can anyone help?

Richard

Post a reply to this message

On Mon, 29 Jul 2002 00:32:08 -0500, Richard Morehouse
<rmo### [at] coxnet> wrote:

>can anyone help?

You've taken a lightly wrong path, Richard. Not that it won't get you
where you want, it's just that you'll get there *very* slow.

Before you go on reading, please revise the radiosity section in the
manual, especially "always_sample", "save_file" and "load_file".

Now, if you want to have realistic render times with radiosity and
reflection or refraction, there is a trick most people do. Basically
the idea is to calculate radiosity first, then render
reflection/refraction. This is done in two steps. First, remove all
reflection, transparency and refraction from your objects, and render
the scene with:

always_sample on 
save_file "radiosity_file.dat"

This will calculate the radiosity contribution of all objects,
including reflective and refractive ones, without going through the
pain of including reflection and refraction in the render times. As a
bonus, this render can be done using lower image resolution settings
(no anti-aliasing and half the height and width).

Then, on the second stage, bring back your transparency, refraction
and reflection and use these settings:

always_sample off 
load_file "radiosity_file.dat" 
preview_start 1 preview_end 1

This will load the radiosity data already calculated and will not
calculate any more radiosity samples, which is the idea (it is very
slow with reflection and refraction). Render at normal resolution,
anti-aliasing and all.

This same trick is also applicable when you want to continue a trace
with radiosity - save the radiosity date in the first stage, then
start the real render using the second settings - you can continue
that render as often as you want to.

If I got anything wrong, someone please correct me before Richard
starts another month-long render..

Good luck!


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message

Peter Popov <pet### [at] vipbg> wrote:

>On Mon, 29 Jul 2002 00:32:08 -0500, Richard Morehouse
><rmo### [at] coxnet> wrote:
>
>>can anyone help?
>
>You've taken a lightly wrong path, Richard. Not that it won't get you
>where you want, it's just that you'll get there *very* slow.
>

>Before you go on reading, please revise the radiosity section in the
>manual, especially "always_sample", "save_file" and "load_file".
>
>Now, if you want to have realistic render times with radiosity and
>reflection or refraction, there is a trick most people do. Basically
>the idea is to calculate radiosity first, then render
>reflection/refraction. This is done in two steps. First, remove all
>reflection, transparency and refraction from your objects, and render
>the scene with:
>
>always_sample on 
>save_file "radiosity_file.dat"
>
I am still a bit of a noobie at this so i will have a lot of
questions. 

how would I change the glass objects?

how does colored glass affect this?

what about metal objects? I use quite a lot of the Polished_Chrome
texture

and water?

do i need to change any other finish statements? ie: finish {shiny}


>This will calculate the radiosity contribution of all objects,
>including reflective and refractive ones, without going through the
>pain of including reflection and refraction in the render times. As a
>bonus, this render can be done using lower image resolution settings
>(no anti-aliasing and half the height and width).
>
>Then, on the second stage, bring back your transparency, refraction
>and reflection and use these settings:
>
>always_sample off 
>load_file "radiosity_file.dat" 
>preview_start 1 preview_end 1
>
>This will load the radiosity data already calculated and will not
>calculate any more radiosity samples, which is the idea (it is very
>slow with reflection and refraction). Render at normal resolution,
>anti-aliasing and all.
so the first render is a throwaway just to get the radiosity data?
>
>This same trick is also applicable when you want to continue a trace
>with radiosity - save the radiosity date in the first stage, then
>start the real render using the second settings - you can continue
>that render as often as you want to.

could i use an ini file to do both renders one after the other?


>
>If I got anything wrong, someone please correct me before Richard
>starts another month-long render..
my last month long render was done on a much slower machine.  100mhz
pentium 32mb ram. vs 400mhz k6-2 w 256mb ram. big difference. 

in addition to sharing the machine with my wife I have a mischievious
toddler who likes to turn the computer off when I am not looking. and
we have alot of power outages due to weather this time of year so
anything helps :)

after my first post i started the render again without area lights.
this is making a big difference in render time. 

is the difference in quality significant enough with area lights to
make it worthwhile. ?

>
>Good luck!

Thank you Peter

Richard

Post a reply to this message

On Mon, 29 Jul 2002 09:58:43 -0500, Richard Morehouse
<rmo### [at] coxnet> wrote:

>I am still a bit of a noobie at this so i will have a lot of
>questions. 

That's what these newsgroups are for, after all.

>how would I change the glass objects?

Removing the transparency should suffice. If you have reflection,
remove that as well.

>how does colored glass affect this?

Most people use a high ambient and low diffuse for glasses. When you
remove the transparency, the colored glass will color its surroundings
- maybe not very realistically, maybe vice versa, you'll have to try
and see if you like it :)

>what about metal objects? I use quite a lot of the Polished_Chrome
>texture

Well, use the same texture but without reflection. Read on.

>and water?

Same as glass.

>do i need to change any other finish statements? ie: finish {shiny}

Anything that has reflection and/or transparency - I think Shiny in
particular does not, but you should check.

>could i use an ini file to do both renders one after the other?

Of course :)

Here is a rough draft of what you should do - I am writing this
directly in the newsreader so I can't test it at the moment but it
should give you an idea. Render as an animation with two frames - that
is, command line switches +kfi0 +kff1 

// The first frame is the pretrace, the second is the trace
#if (clock<0.5)
  #declare Pretrace = true;
#else
  #declare Pretrace = false;
#end

global_settings {          
  radiosity {              
    count 100
    nearest_count 5        
    error_bound 1.8        
    recursion_limit 3      
    low_error_factor .5    
    gray_threshold 0.0     
    minimum_reuse 0.015    
    brightness 1           
    adc_bailout 0.01/2     

// Now, if we are pretracing, use high quality settings and save a
// nice, dense radiosity file

#if (Pretrace)
    pretrace_start 0.08
    pretrace_end   0.01
    always_sample on
    save_file "provide_file_name_here"

// Otherwise, load the file and do not calculate additional samples

#else
    pretrace_start 1
    pretrace_end   1
    always_sample off
    load_file "provide_file_name_here"
#end

  }                        
}

// Here is an example of how to make a glass material
// Shows how to switch reflection and transparency on/off

#declare MyGlassMaterial = material {
  texture {
    pigment {
      color rgb <0.95,0.975,1>

// Only add transparency if not pretracing

#if (!Pretrace)
      filter 0.95
#end
    }
    finish {
      ambient 0.8 diffuse 0.2 brilliance 0.8
      specular 0.4 roughness 0.0005
      phong 0.2 phong_size 7.5

// Only add reflection if not pretracing

#if (!Pretrace)
      reflection 0.075
#end
    }
  }

  interior { ior 1.765 }
}

>after my first post i started the render again without area lights.
>this is making a big difference in render time. 

>is the difference in quality significant enough with area lights to
>make it worthwhile. ?

Try without radiosity first and see whether area lights are
contributing enough to your scene to pay the price.

I hope you get a more clearer idea of what I tried to explain in my
last post. My code probably may not run the first time but it should
at least tip you in the right direction.


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message

Thank you. your example helps a great deal.
I'll let you know how it turns out.

Richard

Post a reply to this message