Rendering efficiency comparison between light and glow material

[1112]

In radiance, when using glow material to develop a light source, multiple rays should be send to find the light source. However, if use light material, only the center of the light source was sampled when no jitter was applied. That is to say, the rendering efficiecncy of light shall be higher than that of glow. But I got opposed results, the calculation time when using light material is much longer when using the same redering parameters.

Below are scene defination:
#glowWin/klems12L.rad (Light material)
void brightdata Klems_dist
4 noop ./glowWin/klems12.dat klems_full.cal kbin(0.0,1.0,0.0,0,0,1)
0
0

Klems_dist light winglow
0
0
3 1 1 1

winglow polygon win0
0
0
12
-3.867963 -3.191139 2.4
-3.867963 -3.191139 0.9
-5.497963 -3.191139 0.9
-5.497963 -3.191139 2.4
winglow polygon win1
0
0
12
-0.975422 -3.191139 2.4
-0.975422 -3.191139 0.9
-2.610422 -3.191139 0.9
-2.610422 -3.191139 2.4

#glowWin/klems12.rad (Glow material)
void brightdata Klems_dist
4 noop ./glowWin/klems12.dat klems_full.cal kbin(0.0,1.0,0.0,0,0,1)
0
0

Klems_dist glow winglow
0
0
4 1 1 1 0

winglow polygon win0
0
0
12
-3.867963 -3.191139 2.4
-3.867963 -3.191139 0.9
-5.497963 -3.191139 0.9
-5.497963 -3.191139 2.4
winglow polygon win1
0
0
12
-0.975422 -3.191139 2.4
-0.975422 -3.191139 0.9
-2.610422 -3.191139 0.9
-2.610422 -3.191139 2.4

Command Lines:
Light material
oconv -f glowWin/klems12.rad materials.rad Off_scene.rad > m.oct
rpict -vf views/view.vf -t 5 -aa 0.10 -ab 8 -ad 2000 -lw 0.0005 -x 800 -y 800 m.oct > m.hdr

Glow material
oconv -f glowWin/klems12L.rad materials.rad Off_scene.rad > mL.oct
rpict -vf views/view.vf -t 5 -aa 0.10 -ab 8 -ad 2000 -lw 0.0005 -x 800 -y 800 mL.oct > mL.hdr

Results:
Rendering Time (Light): 5m40.401s

Rendering Time(Glow): 3m0.233s

[Greg_Ward]

The rendering time with light sources will always be longer than without using the same parameters, because rays get sent to every (visible) light source from every intersected point in the scene. This is definitely worth the cost if the light sources are small and hard to find. If they are large in terms of solid angle from every important point in the scene, then you are often better off using the “glow” type as you discovered.

Also, the -ds setting affects whether light sources are subdivided based on their size. You can end up sending (many) more than a single ray to one light source.

Cheers,
-Greg