=?UTF-8?B?SsOpcsO0bWUgTS4gQmVyZ2Vy?= <jeberger@free.fr> wrote:
> On 08/30/2015 07:38 PM, Alain wrote:
> > Ouch!
> > When you add a manual bounding to an union, that union is no longer
> > splitted. That cause it to take MUCH more time to render as each ray
> > that hit the manual bounding box now need to be tested against each and
>
> > every components of the union. It's similar in effect to uselessly usin
> g
> > merge instead of union.
> >
> > Also, as your manual bounding box is larger than the original union,
> > it's even slower.
> >
>     Have you tried it? ;) According to And, adding the manual bounding
> box decreases the render time from 2m04s to 2s...
>
>         Jerome
> --
> mailto:jeb### [at] freefr
> http://jeberger.free.fr
> Jabber: jeb### [at] jabberfr


That's true. And it's impressive.
Thanks for sharing the information.

-------------

Cheng Han Tsai

Post a reply to this message

"And" <49341109@ntnu.edu.tw> wrote:
> =?UTF-8?B?SsOpcsO0bWUgTS4gQmVyZ2Vy?= <jeberger@free.fr> wrote:
> > On 08/30/2015 07:38 PM, Alain wrote:
> > > Ouch!
> > > When you add a manual bounding to an union, that union is no longer
> > > splitted. That cause it to take MUCH more time to render as each ray
> > > that hit the manual bounding box now need to be tested against each and
> >
> > > every components of the union. It's similar in effect to uselessly usin
> > g
> > > merge instead of union.
> > >
> > > Also, as your manual bounding box is larger than the original union,
> > > it's even slower.
> > >
> >     Have you tried it? ;) According to And, adding the manual bounding
> > box decreases the render time from 2m04s to 2s...
> >
> >         Jerome
> > --
> > mailto:jeb### [at] freefr
> > http://jeberger.free.fr
> > Jabber: jeb### [at] jabberfr
>
>
> That's true. And it's impressive.
> Thanks for sharing the information.
>
> -------------
>
> Cheng Han Tsai

There may be a problem with the bounding on my floor tiles.  Each set of tiles
is parsed as a single prism, but there are a couple of unions and differences
that I could split and bound manually.

Regards,
A.D.B.

Post a reply to this message

On 08/31/2015 09:35 AM, Anthony D. Baye wrote:
> There may be a problem with the bounding on my floor tiles.  Each set o
f tiles
> is parsed as a single prism, but there are a couple of unions and diffe
rences
> that I could split and bound manually.
> 
    Note that AFAIK automatic bounding of unions is already optimal.
Manual bounding of differences can improve things, especially if you are
removing a large chunk of the original object, but finding the optimal
bounding box can be tricky. It's only for intersections that automatic
bounding is blatantly wrong and that manual bounding can very easily
improve things dramatically.

        Jerome
-- 
mailto:jeb### [at] freefr
http://jeberger.free.fr
Jabber: jeb### [at] jabberfr

Post a reply to this message

Attachments:
Download 'us-ascii' (1 KB)

"And" <49341109@ntnu.edu.tw> wrote:
> And here is a picture it works with radiosity feature on:

I forgot a matter. We should use a light_source that it assigns fade_distance 0
in POV-Ray 3.71 to get a realistic result.

Post a reply to this message

Attachments:
Download 'finish test2 6.png' (156 KB)

Preview of image 'finish test2 6.png'

=?UTF-8?B?SsOpcsO0bWUgTS4gQmVyZ2Vy?= <jeberger@free.fr> wrote:
> On 08/31/2015 09:35 AM, Anthony D. Baye wrote:
> > There may be a problem with the bounding on my floor tiles.  Each set o
> f tiles
> > is parsed as a single prism, but there are a couple of unions and diffe
> rences
> > that I could split and bound manually.
> >
>     Note that AFAIK automatic bounding of unions is already optimal.
> Manual bounding of differences can improve things, especially if you are
> removing a large chunk of the original object, but finding the optimal
> bounding box can be tricky. It's only for intersections that automatic
> bounding is blatantly wrong and that manual bounding can very easily
> improve things dramatically.
>
>         Jerome
> --
> mailto:jeb### [at] freefr
> http://jeberger.free.fr
> Jabber: jeb### [at] jabberfr

Okay ~.

Post a reply to this message

Am 31.08.2015 um 22:11 schrieb Jérôme M. Berger:
> On 08/31/2015 09:35 AM, Anthony D. Baye wrote:
>> There may be a problem with the bounding on my floor tiles.  Each set of tiles
>> is parsed as a single prism, but there are a couple of unions and differences
>> that I could split and bound manually.
>>
>     Note that AFAIK automatic bounding of unions is already optimal.

Strictly speaking, that depends on whether the bounding of its elements
is optimal.

Post a reply to this message

Am 01.09.2015 um 05:27 schrieb And:
> "And" <49341109@ntnu.edu.tw> wrote:
>> And here is a picture it works with radiosity feature on:
> 
> I forgot a matter. We should use a light_source that it assigns fade_distance 0
> in POV-Ray 3.71 to get a realistic result.

I would recommend using a non-zero fade_distance, set to be equal to the
corresponding object's radius. That more closely mimicks the
characteristics of a non-point light source.

Obviously you'll have to adjust the light source's brightness accordingly.

Post a reply to this message

On 2015-09-02 09:49 AM (-4), clipka wrote:
> I would recommend using a non-zero fade_distance, set to be equal to the
> corresponding object's radius. That more closely mimicks the
> characteristics of a non-point light source.

Is this true even for area_illumination with circular orient?

Post a reply to this message

Am 02.09.2015 um 20:59 schrieb Cousin Ricky:
> On 2015-09-02 09:49 AM (-4), clipka wrote:
>> I would recommend using a non-zero fade_distance, set to be equal to the
>> corresponding object's radius. That more closely mimicks the
>> characteristics of a non-point light source.
> 
> Is this true even for area_illumination with circular orient?

Hum - that's a good question I've never thought about. But by extension
of what I said, if you model a single non-point light source as an array
of smaller light sources, then those smaller light sources'
fade_distance should be set equal to their radius, so for a highly
subdivided area_illumination light source the fade_distance should
indeed probably be close to zero.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 02.09.2015 um 20:59 schrieb Cousin Ricky:
> > On 2015-09-02 09:49 AM (-4), clipka wrote:
> >> I would recommend using a non-zero fade_distance, set to be equal to the
> >> corresponding object's radius. That more closely mimicks the
> >> characteristics of a non-point light source.
> >
> > Is this true even for area_illumination with circular orient?
>
> Hum - that's a good question I've never thought about. But by extension
> of what I said, if you model a single non-point light source as an array
> of smaller light sources, then those smaller light sources'
> fade_distance should be set equal to their radius, so for a highly
> subdivided area_illumination light source the fade_distance should
> indeed probably be close to zero.


That's true. If you would like to simulation an area/surface light, every
element light source of it should obey the inverse square law. So it should set
the fade_distance 0 when the area light resolution is enough.

I've researched this and I would like to share it. In fact, I even know the
formula between the light source color and the emission surface color. If you
have any lambertian emission surface that emit light and you do not want to use
the radiosity, you can put the corresponding light sources to simulate it(just
like the area light). The formula between them is:
(emission_surface_color) = (light_source_color) * pi * resolution /
(surface_area)

Notice that it must set spotlight and correct radius, falloff, and tightness in
the lightsource.

You can implement it like this:


//---------------the code start---------------------

#declare disc_radius = 1;
#declare disc_color = rgb<1,1,1>;
#declare resolution = 100;
#declare emission_area = pi*disc_radius*disc_radius;


#declare looks_like_obj =
disc{
<0,0,0>, z, disc_radius,
    texture{
        pigment{color disc_color}
        finish{ambient 0 diffuse 0 emission 1}
    }
    interior_texture{
        pigment{color rgb<0,0,0>}
        finish{ambient 0 diffuse 0 emission 0}
    }
no_shadow
no_radiosity
}




#declare light_seed = seed(13);
#declare light_source_obj =
union{
    #for(i, 0, resolution-1)
        #declare this_r = sqrt(1/2+i)/sqrt(resolution);
        #declare this_theta = rand(light_seed)*2*pi;
        #declare this_point = <this_r*cos(this_theta), this_r*sin(this_theta),
0>* disc_radius;
        light_source{
        <0,0,0>
        color disc_color*emission_area/resolution/pi
        spotlight
        radius 90
        falloff 90
        tightness 1
        point_at z
        fade_power 2
        fade_distance 0
        }
    #end
}




union{
    object{light_source_obj}
    object{looks_like_obj}
}

Post a reply to this message