Indigo 2.0.10

[Whaat]

Hi whaat, it seems that Indigo 2.0.10 is reacting with your scene as if the lighting condition were particularily tricky IIRC.

Raise the blackbody gain and set the camera ISO value to something reasonable to compensate, hope this works :)

Thanks CTZn but I am sure this is an issue with the new scattering code. I just ran a bunch of tests. Indigo's performance gets worse and worse with higher scattering values. The scattering values for the liquid in this scene are between 7000-8000 (using RGB spectrum). When I lower them to even 1000, the performance is more normal. As I keep increasing them, the performance gets worse and worse. I was getting about 24,000 samples/s with 2.0.9 and now getting less than 500 samples/s with 2.0.10.

Another issue that seems to be new in 2.0.10 is that when I stop a render and press the 'open scene' button, Indigo can take sometimes up to a minute to display the 'open scene' dialog.

[pixie]

When I lower them to even 1000, the performance is more normal. As I keep increasing them, the performance gets worse and worse. I was getting about 24,000 samples/s with 2.0.9 and now getting less than 500 samples/s with 2.0.10.

I've done a rough scene and it seems to be the case, something along happens when no light reaches the scene...

[CTZn]

You replied quickly whaat, I came back to append the note that it was just a thought from me and was untested. I answered quickly too in first instance, I should have tagged this as a thought or have kept it for me.

[SzLaszlo]

Hi,


SSS is faster ?

Sample scene :
Indigo 1.1.18 > 126 s/px.
indigo 2.0.9. > 132 s/px.
Indigo 2.0.10. > 113 s/px.

Sys: Ubuntu Linux 64bit

[CTZn]

SzLaszlo, enforce scattering_coefficient_spectrum uniformity as mentioned in the first post and see...

[SzLaszlo]

SzLaszlo, enforce scattering_coefficient_spectrum uniformity as mentioned in the first post and see...

I try this mat :

<medium>
<name>Glass</name>
<precedence>10</precedence>
<basic>
<ior>1.01</ior>
<cauchy_b_coeff>0.0003</cauchy_b_coeff>
<absorption_coefficient_spectrum>
<rgb>
<rgb>1 0.44747 0</rgb>
<gamma>2.2</gamma>
</rgb>
</absorption_coefficient_spectrum>
<subsurface_scattering>
<scattering_coefficient_spectrum>
<uniform>
<value>10</value>
</uniform>
</scattering_coefficient_spectrum>
<phase_function>
<uniform/>
</phase_function>
</subsurface_scattering>
</basic>
</medium>
<material>
<name>Glass</name>
<specular>
<transparent>true</transparent>
<internal_medium_name>Glass</internal_medium_name>
</specular>
</material>


,and 1.1.18 > 38.7 s/px.
- 2.0.10. > 17.75 s/px.

[Zom-B]

the samples per second speed don't has to be measure here!
Maybe the number is lower, but the sample quality is higher! (or at least it should be!)

render some images for the same time and compare the visual output!

[benn]

The samples/sec may be slower but the SSS material should resolve faster.

[Whaat]

The samples/sec may be slower but the SSS material should resolve faster.

Well, there are definite issues between 2.0.9 and 2.0.10. Whatever Ono changed to optimize scattering for uniform spectra also has affected non-uniform spectra. I rendered the exact same scene with 2.0.9 and 2.0.10. There is no question that the noise is far greater in 2.0.10 for the same amount of time.

Also, the 2.0.10 images looks darker even though I have the same tonemapping settings (maybe it's just darker because it's noiser...)

[Zom-B]

Also, the 2.0.10 images looks darker even though I have the same tonemapping settings (maybe it's just darker because it's noiser...)

I can confirm this, camera tonemapping a least seems a little darker!

Another Bug I found is the missing sky if you render with Extra Atmospheric!

[Whaat]

Also, the 2.0.10 images looks darker even though I have the same tonemapping settings (maybe it's just darker because it's noiser...)

I can confirm this, camera tonemapping a least seems a little darker!

My images were done with camera tonemapping.

[Whaat]

Another Bug I found is the missing sky if you render with Extra Atmospheric!

If extra_atmospheric is true, then the skylight is computed as if it was outside the atmosphere.
This means that the sun spectrum is not attenuated by atmospheric scattering, and the sky will be black, since there is no atmospheric scattering.

[pixie]

I'm getting some mesh troubles, first I though it to be bump related but after taking it away the persist...

[Zom-B]

Seems like the testscenes "€.igs" and "unicode_€_test.igs" cause an error... so the unicode support seems not to work!?

[SzLaszlo]

the samples per second speed don't has to be measure here!
.........

ok ,ok , thanks ,sorry :)