Indigo 2.0.12

[suvakas]

Highly scattering unbiased SSS will always be slow.

No doubt you're right, but please acknowledge that highly scattering materials render way slower in the latest builds, at least when using bidirectional. I haven't fully tested the other tracing methods yet, but I think they are also slower with highly scattering materials. And I am referring to convergence, not sample rate.

In my experience, they converge faster.

Nope, Ono. It takes centuries compared to previous builds (before you did the speed up). It either was working incorrectly before or something is broken now.

[benn]

We'll take a look suvakas, can you give us a material that resolves slower?

[suvakas]

We'll take a look suvakas, can you give us a material that resolves slower?

Code: Select all

<?xml version="1.0" encoding="UTF-8"?>
<!--Generated using Maxigo version 2.0.12-beta -->

<scenedata>
	<medium>
		<name>slow_sss-medium</name>
		<precedence>2</precedence>
		<basic>
			<ior>1.53</ior>
			<cauchy_b_coeff>0.0</cauchy_b_coeff>
			<absorption_coefficient_spectrum>
			<rgb>
				<rgb>0.42 0.42 0.42</rgb>
				<gamma>1.0</gamma>
			</rgb>
			</absorption_coefficient_spectrum>
		<subsurface_scattering>
			<scattering_coefficient_spectrum>
			<rgb>
				<rgb>101.0 252.0 0.0</rgb>
				<gamma>1.0</gamma>
			</rgb>
			</scattering_coefficient_spectrum>
			<phase_function>
				<uniform/>
			</phase_function>
		</subsurface_scattering>
		</basic>
	</medium>
<material>
	<name>slow_sss</name>
	<specular>
	<transparent>true</transparent>
		<internal_medium_name>slow_sss-medium</internal_medium_name>
	</specular>
</material>
</scenedata>

[Zom-B]

Highly scattering unbiased SSS will always be slow.

Please just start a scene with a milk material Ono... !!!1!
the samples per second will drop down to a few hundred.... this is not slow, this is unusable... in this case drop SSS support!

Code: Select all

  <medium>
        <name>med_milk_regular</name>
        <precedence>10</precedence>
        <basic>
            <ior>1.35</ior>
            <cauchy_b_coeff>0.000000</cauchy_b_coeff>
            <absorption_coefficient_spectrum>
                <rgb>
                    <rgb>1.533 4.6 19.933</rgb>
                    <gamma>1.000000</gamma>
                </rgb>
            </absorption_coefficient_spectrum>
            <subsurface_scattering>
                <scattering_coefficient_spectrum>
                    <rgb>
                        <rgb>18205.267 20382.6 22369.8</rgb>
                        <gamma>1.000000</gamma>
                    </rgb>
                </scattering_coefficient_spectrum>
                <phase_function>
                    <henyey_greenstein>
                        <g_spectrum>
                            <rgb>
                                <rgb>0.75 0.714 0.681</rgb>
                                <gamma>1.000000</gamma>
                            </rgb>
                        </g_spectrum>
                    </henyey_greenstein>
                </phase_function>
            </subsurface_scattering>
        </basic>
    </medium>
    <material>
        <name>milk_regular</name>
        <specular>
            <internal_medium_name>med_milk_regular</internal_medium_name>
            <transparent>true</transparent>
        </specular>
    </material>

[suvakas]

So basically it goes slow when you use rgb instead of uniform. With uniform it is fast since the last speed up, but rgb is much more slower than before the speed up.

[Zom-B]

uniform is also veeeery slow if it is set up high!

[Zom-B]

Since reading the logoutput on the right of indigo is quite impossible if a lot (network related) of stuff is going on I tried to check out the log.txt just tos find out that NO network related data is saved there!
Please fix this!

A more convenient way to read the log would be first not to auto scroll down the log if the scroll bar is pressed!
maybe some stuff can also be reduced here or simplified.

Example:

Code: Select all

Receiving frame from 192.168.0.81:49872 :
	num layers: 1, num_samples: 2510000.000000, width: 772, height: 580
	Frame received. (lock wait time: 0.00 s, transfer time: 0.74 s)
	Frame received. (lock wait time: 0.02 s, transfer time: 0.59 s)

First I don't get it why there are two lines with frame received?!
The second linecould be reduced to something like:

Code: Select all

Layers: 1, num_samples: 2510k, w: 772, h: 580

Dropping the width/height info could be also a good point...
Also from the second line "num layers:" is pushed so far to the right that it isn't readable without scrolling...

[lycium]

uniform is also veeeery slow if it is set up high!

try rendering with normal (non-bidir) path tracing; i suspect what's happening is that the highly scattering medium is creating a great many path vertices, and bidir is O(N^2) with N path vertices where straight path tracing is O(N).