animation settings

[daros]

2k at 4 samples x pixel

https://drive.google.com/file/d/10493Tz ... drive_link

[daros]

2k at 4 samples x pixel, 10 seconds x frame

https://drive.google.com/file/d/10493Tz ... drive_link

Not so bad for a scene almost fully lighted by caustics

[daros]

Down to earth.
Scene with a lot of glass. Glass is hidden to GI.
https://drive.google.com/file/d/1uRlJcn ... drive_link

my luxcore settings:
path.pathdepth.total 256
path.pathdepth.diffuse 16
path.pathdepth.glossy 256
path.pathdepth.specular 256
film.noiseestimation.warmup 8
film.noiseestimation.step 32
sampler.sobol.adaptive.strength 0.95
sampler.sobol.bucketsize 1
sampler.sobol.tilesize 16
sampler.sobol.supersampling 1
sampler.sobol.overlapping 32
renderengine.type "BIDIRCPU"
sampler.type "METROPOLIS"
film.filter.type "NONE"
film.filter.width 1.5
lightstrategy.type "POWER"
filesaver.format "TXT"
path.clamping.variance.maxvalue 2000000
scene.epsilon.min 0.00001
scene.epsilon.max 0.1
path.albedospecular.type "REFLECT_TRANSMIT"
path.albedospecular.glossinessthreshold 0.05
lightstrategy.entry.radius 0
lightstrategy.entry.normalangle 10
lightstrategy.entry.maxpasses 1024
lightstrategy.entry.convergencethreshold 0.01
lightstrategy.entry.warmupsamples 12
lightstrategy.entry.volumes.enable 0
lightstrategy.lightthreshold 0.01
lightstrategy.targetcachehitratio 0.995
lightstrategy.maxdepth 4
lightstrategy.maxsamplescount 10000000
lightstrategy.persistent.file “”
path.photongi.photon.maxcount 80000000
path.photongi.photon.maxdepth 64
path.photongi.glossinessusagethreshold 0.049
path.photongi.indirect.enabled 0
path.photongi.indirect.maxsize 0.15
path.photongi.indirect.haltthreshold 0.02
path.photongi.indirect.lookup.radius 0
path.photongi.indirect.lookup.normalangle 10
path.photongi.indirect.usagethresholdscale 1
path.photongi.caustic.enabled 1
path.photongi.caustic.maxsize 5000000
path.photongi.caustic.lookup.radius 0.005
path.photongi.caustic.lookup.normalangle 10
path.photongi.caustic.updatespp 0
path.photongi.caustic.updatespp.radiusreduction 0.96
path.photongi.caustic.updatespp.minradius 0.005
path.hybridbackforward.enable 1
path.hybridbackforward.partition 0
path.hybridbackforward.glossinessthreshold 0.1
filesaver.renderengine.type "PATHCPU"

[epilectrolytics]

Good Job, the interior animation at 32sp looks great

About the issue with the complex glass scene I have no clue

[bartek_zgo]

I'm working with daros on animation issues. We are trying to activate photongi cache. We have found a very strange behavior when rendering frames on multiple machines.
1. If we start animation with photongi cache disabled the result is correct but the render is very slow.
2. If we enable photnongi cache without line "path.photongi.persistent.file = "photongi.pgi" the result is correct. The render is much faster. But when we render without presistent file, each machine creates it's own cache, so it takes time.
3. So we activate photonhi with persistent file. First we render one frame. Luxcore saves a photongi.pgi file. Than we pass that file to other machines and start render of other frames. We can see from log that all machines read the photongi.pgi file. Most of machines work as suspected and the render is correct and fast. Unfortunately some of machines (1 or 2 for for 15) creates broken frames. This does not depend on machine. If we repeat the render, machine that previously creates broken frame, creates correct one.

Some more details:
We are sure that photongi.pgi file is the same on all machines. I have added a line into code, that after load, we save the pgi file. The md5sum of that file is identical to original one.
In our case each machine renders more than one frame. So we load a scene into luxcore (including pgi file), render one frame, than edit scene (move camera) and render another frame. We have found that all frames made by one machine are correct or all broken.

Correct frame:

Broken frame:

Code: Select all

path.pathdepth.total = "256"
path.pathdepth.diffuse = "16"
path.pathdepth.glossy = "256"
path.pathdepth.specular = "256"
film.noiseestimation.warmup = "8"
film.noiseestimation.step = "32"
sampler.sobol.adaptive.strength = "0.95"
sampler.sobol.bucketsize = "1"
sampler.sobol.tilesize = "16"
sampler.sobol.supersampling = "1"
sampler.sobol.overlapping = "32"
renderengine.type = "BIDIRCPU"
sampler.type = "METROPOLIS"
film.filter.type = "NONE"
film.filter.width = "1.5"
lightstrategy.type = "POWER"
filesaver.format = "TXT"
path.clamping.variance.maxvalue = "25"
scene.epsilon.min = "0.00001"
scene.epsilon.max = "0.1"
path.albedospecular.type = "REFLECT_TRANSMIT"
path.albedospecular.glossinessthreshold = "0.05"
lightstrategy.entry.radius = "0"
lightstrategy.entry.normalangle = "10"
lightstrategy.entry.maxpasses = "1024"
lightstrategy.entry.convergencethreshold = "0.01"
lightstrategy.entry.warmupsamples = "12"
lightstrategy.entry.volumes.enable = "0"
lightstrategy.lightthreshold = "0.01"
lightstrategy.targetcachehitratio = "0.995"
lightstrategy.maxdepth = "4"
lightstrategy.maxsamplescount = "10000000"
lightstrategy.persistent.file = ""
path.photongi.photon.maxcount = "80000000"
path.photongi.photon.maxdepth = "64"
path.photongi.glossinessusagethreshold = "0.049"
path.photongi.indirect.enabled = "0"
path.photongi.indirect.maxsize = "0.15"
path.photongi.indirect.haltthreshold = "0.02"
path.photongi.indirect.lookup.radius = "0"
path.photongi.indirect.lookup.normalangle = "10"
path.photongi.indirect.usagethresholdscale = "1"
path.photongi.caustic.enabled = "1"
path.photongi.caustic.maxsize = "5000000"
path.photongi.caustic.lookup.radius = "0.005"
path.photongi.caustic.lookup.normalangle = "10"
path.photongi.caustic.updatespp = "0"
path.photongi.caustic.updatespp.radiusreduction = "0.96"
path.photongi.caustic.updatespp.minradius = "0.005"
path.photongi.persistent.file = "/mnt/z/MATADOR/AnimOutput/FA_1098_01_80958.fxs/anim_1829/photongi.pgi"
path.hybridbackforward.enable = "1"
path.hybridbackforward.partition = "0"
path.hybridbackforward.glossinessthreshold = "0.1"
filesaver.renderengine.type = "PATHCPU"
renderengine.seed = 1
film.width = 640
film.height = 360
batch.halttime = 60000
batch.haltspp = 4
film.imagepipelines.0.0.type = "NOP"
film.imagepipelines.0.1.type = "INTEL_OIDN"
film.imagepipelines.0.1.prefilter.enable = "1"
film.imagepipelines.0.2.type = "TONEMAP_LUXLINEAR"
film.imagepipelines.0.2.sensitivity = "100"
film.imagepipelines.0.2.exposure = 0.8
film.imagepipelines.0.2.fstop = 4
film.outputs.0.type = "RGB_IMAGEPIPELINE"
film.outputs.0.filename = "/home/node/matadorNode/tmp/0/out.png"
film.outputs.0.index = 0
film.outputs.1.type = "ALBEDO"
film.outputs.1.filename = "/home/node/matadorNode/tmp/0/out_albedo.exr"
film.outputs.2.type = "AVG_SHADING_NORMAL"
film.outputs.2.filename = "/home/node/matadorNode/tmp/0/out_avgShadingNormal.exr"
scene.file = "scene.scn"

[bartek_zgo]

I simplified the problem to one machine:
Render first frame and make luxcore save the pgi file.
while(noAllFramesRendered)
{
start render of next frame
make luxcore load pgi file
after frame is finished close luxcore
}

Some frames are correct some are not.
So it looks that loading pgi file from disk is not working in a stable way. In my case about 30% of frames is broken.

[epilectrolytics]

about 30% of frames is broken.

You should open an issue on GitHub about this bug.

[CodeHD]

Hi,

I have done a brief search on the forum and in open Issues.
I could find an old topic by Sharlybg where Dade commented that it might be due to the periodic update:
viewtopic.php?p=30140#p30140

However, the config you posted indicates you have set it to "0", i.e. disabled. Shouldn't matter that it is enclosed in Quotation marks (i.e. passed as a string rather than an int) according to quick testing.

Would be nice indeed if you can open A GitHub issue, and if possible provide a test scene that shows this problem.
Which would be my next question anyways: Does it happen on every scene? Did you only just notice this? Is persistent cache not a feature you are often using anyways? Or could it be something introduced with version 2.10?

[FarbigeWelt]

Or could it be something introduced with version 2.10?

No. This behavior is quite old. I guess it never got fixed. If one rendered a messed up frame there was a good chance the second time the same frame rendered okay.

[epilectrolytics]

No. This behavior is quite old. I guess it never got fixed.

Some older report here.
I have not encountered this for a long time now.