Problem with supersample=1 command line

[Former user]

The flame file has a supersample option set to 1.

When trying to render, command freezes on 100%. High GPU usage. Forcing to supersample 1 on command line gives the same problem.

C:\Program Files (x86)\Electric Sheep HD>%APPDATA%\Fractorium\emberanimate --in "C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame" --format=jpg --opencl --sp --isaac_seed=fractorium --verbose --progress --supersample=1
Using single precision.
Using OpenCL to render.
Platform: NVIDIA Corporation NVIDIA CUDA OpenCL 1.2 CUDA 11.0.208
Device: NVIDIA Corporation GeForce RTX 2060
Parsing ember file C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame
Finished parsing.
Time = 1 / 1,800 / 1
Stage = iterating, progress = 100%, eta = 0.00s

Setting to supersample 2, command works fine

C:\Program Files (x86)\Electric Sheep HD>%APPDATA%\Fractorium\emberanimate --in "C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame" --format=jpg --opencl --sp --isaac_seed=fractorium --verbose --progress --supersample=1
Using single precision.
Using OpenCL to render.
Platform: NVIDIA Corporation NVIDIA CUDA OpenCL 1.2 CUDA 11.0.208
Device: NVIDIA Corporation GeForce RTX 2060
Parsing ember file C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame
Finished parsing.
Time = 1 / 1,800 / 1
Stage = iterating, progress = 100%, eta = 0.00s^C
C:\Program Files (x86)\Electric Sheep HD>
C:\Program Files (x86)\Electric Sheep HD>
C:\Program Files (x86)\Electric Sheep HD>%APPDATA%\Fractorium\emberanimate --in "C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame" --format=jpg --opencl --sp --isaac_seed=fractorium --verbose --progress --supersample=2
Using single precision.
Using OpenCL to render.
Platform: NVIDIA Corporation NVIDIA CUDA OpenCL 1.2 CUDA 11.0.208
Device: NVIDIA Corporation GeForce RTX 2060
Parsing ember file C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame
Finished parsing.
Time = 1 / 1,800 / 1
Stage = iterating, progress = 100%, eta = 0.00s
Stage = filtering, progress = 100%, eta = 0.00s
Iters ran/requested: 8,493,465,600 / 8,493,465,600 (100.00%)
Render time: 13.07s
Pure iter time: 12.61s
Iters/sec: 673,755,051
Time = 2 / 1,800 / 1
Writing C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\0000.jpg

My openclinfo

C:\Program Files (x86)\Electric Sheep HD>%APPDATA%\Fractorium\emberanimate --openclinfo
OpenCL Info:
Platform 0: NVIDIA Corporation NVIDIA CUDA OpenCL 1.2 CUDA 11.0.208
Device 0: NVIDIA Corporation GeForce RTX 2060
CL_DEVICE_OPENCL_C_VERSION: OpenCL C 1.2
CL_DEVICE_LOCAL_MEM_SIZE: 49,152
CL_DEVICE_LOCAL_MEM_TYPE: 1
CL_DEVICE_MAX_COMPUTE_UNITS: 30
CL_DEVICE_MAX_READ_IMAGE_ARGS: 256
CL_DEVICE_MAX_WRITE_IMAGE_ARGS: 32
CL_DEVICE_MAX_MEM_ALLOC_SIZE: 1,610,612,736
CL_DEVICE_ADDRESS_BITS: 64
CL_DEVICE_GLOBAL_MEM_CACHE_TYPE: 2
CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE: 128
CL_DEVICE_GLOBAL_MEM_CACHE_SIZE: 983,040
CL_DEVICE_GLOBAL_MEM_SIZE: 6,442,450,944
CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 65,536
CL_DEVICE_MAX_CONSTANT_ARGS: 9
CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS: 3
CL_DEVICE_MAX_WORK_GROUP_SIZE: 1,024
CL_DEVICE_MAX_WORK_ITEM_SIZES: 1,024, 1,024, 64

Im using version WIN-1.0.0.20

[Matt Feemster]

Thank you for reporting this. Your second command line still has --supersample=1, I assume that’s a typo?

Can you please attach the flames that are causing this problem?

Also, can you tell me if this happens with all flames or just this specific one?

Can you please login before posting bugs here, it helps me keep track of who is posting what. As of now, it only shows you as “Former user”.

Thanks.

[Rafael Gassner]

Hi,

Sorry, i repeated the command on the second line with supersample=1 goes to 100% when it freezes, but right below, with supersample=2, it works.

The flame can be downloaded here:

https://drive.google.com/file/d/1I1ic2yqUdFMR4kOJaO9f9YsxNrjL3xk2/view?usp=sharing

As long as i have tried, it freezes with any flame from the electricsheep project. I have been able to render the default examples that come with fractorium with no problems though.

‌

Thanks for the help

‌

[Rafael Gassner]

I´m not sure if this information helps, but i am able to render when using CPU

C:\Program Files (x86)\Electric Sheep HD>%APPDATA%\Fractorium\emberanimate --in "C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame" --format=jpg  --sp --isaac_seed=fractorium --verbose --progress
Using single precision.
Using 6 automatically detected threads.
Parsing ember file C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.flame
Finished parsing.
Time = 1 / 1,800 / 1
Stage = iterating, progress = 100%, eta = 0.00s
Stage = filtering, progress = 100%, eta = 0.00s
Iters ran/requested: 8,493,465,600 / 8,493,465,600 (100.00%)
Bad values: 2
Render time: 4m 18.96s
Pure iter time: 4m 18.58s
Iters/sec: 32,846,574
Time = 2 / 1,800 / 1
Writing C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\0000.jpg

‌

[Matt Feemster]

I need a few more tests from you before I start to spend time on this, because I am extremely busy with other things, including a new open source project.

‌

The bug appears to be happening when rendering a sequence in emberanimate. Can you try to just render the first few frames in emberrender and see if the same problem happens?

‌

Also, I wonder if the file size is a problem. In theory it shouldn’t be, but we need to check just to be sure. Can you cut the file down to maybe 10 frames or less, and then feed it to emberanimate again? I wonder if the 1800 frame file is causing a problem?

‌

Thanks.

[Rafael Gassner]

Thank you for the guidance

emberrender shows me the same problem even with a 2 frames file. When using CPU files are rendered.

C:\Program Files (x86)\Electric Sheep HD>%APPDATA%\Fractorium\emberrender --in "C:\Program Files (x86)\Electric Sheep HD\branches\G244_W3840_H2160_Q1024_KFVGM5cUGPU\animated_genomes\electricsheep.244.00619.12967_electricsheep.244.00619.12967.short.flame" --format=jpg --opencl --sp --isaac_seed=fractorium --verbose --progress
Using single precision.
Using OpenCL to render.
Platform: NVIDIA Corporation NVIDIA CUDA OpenCL 1.2 CUDA 11.0.228
Device: NVIDIA Corporation GeForce RTX 2060

Flame = 1/2
Stage = iterating, progress = 100%, eta = 0.00s

‌

I have tried with 10 frames, and even with 2 frames in a file with emberanimate, and it freezes in the same part of the process.

Please let me know if there are any other tests that i can help with.

[Matt Feemster]

Thank you, this helps narrow it down some. Now perform one more test please:

‌

Load the file in fractorium, using the same settings: opencl, single precision, supersample 1 or 2

‌

Does the image render in the editor? If so, move onto the final render dialog. Do a final render there using the same settings.

‌

Let me know if you still see the same freezing problem there.

‌

[Rafael Gassner]

Hello Sir,

When i open the file in Fractorium, the sequence preview is presented normally. The Xml Render Values also loads Supersample 1.

‌

‌

While rendering with supersample 2 it works, but with supersample 1 it freezes right after the Iteration process, using opencl and single precision on both.

Not sure if this info helps, but when i press the Render button it shows me the supersample 2 value and Double precision already selected as default. Image Below.

Fyi, I have already uninstalled fractorium and removed the .ini file and other files that were left, before reinstalling it.

[Matt Feemster]

I’ve tested this with the example flame both inside of Fractorium and on the command line with EmberRender and can’t reproduce it. I am running on a Dell XPS laptop with an Nvidia 1050 card in it. No problems on my end.

I wonder if it’s something to do with the newer cards that is causing your problem?

As a workaround, use SS 2.

In fact, this is generally what you should use for your final renders anyway as it tends to produce smoother lines, without much of a performance penalty.

The only time it costs you is if your DE filter max width is > 0. Supersampling scales the size of the filter, so it’ll slow it down some. However, filtering is usually only a small part of the render time, so it doesn’t impact overall time by very much.

Sorry I couldn’t help more, but it’s hard for me to diagnose problems on hardware I don’t have.

[Matt Feemster]

One more note, I think you are confused on this:

The Xml Render Values also loads Supersample 1.

Those fields specify what values are used when saving the flame parameters out to an Xml file. You can think of them as overrides for those fields. The idea being that you might always want to save your files with a certain value for supersample, quality and temporal sample (that last one being used only for animation).

The values in that tab have nothing to do with what is used when rendering within fractorium. For those values, you’ll want to see the options dialog and the main gui for interactive rendering, and the final render dialog for doing final renders.

Hope that helps.

‌

[Rafael Gassner]

Thank you Sir,

The workaround works just fine for rendering, and i´ve been able to produce my own videos. By the way, congratulations on the software, it is awesome, and the GPU support makes it super fast when compared to CPU!

Unfortunately i cannot change the parameters for the electricsheep project (https://github.com/kochd/electricsheep-hd-client) since it is collaborative and it would result in frames that wouldn´t match to produce a smooth video.

I will try to find someone with equivalent hardware and see if i can reproduce the scenario.

Best regards,

[Matt Feemster]

Changing the supersample shouldn’t make much of a difference in the output. They should still match. To be honest, kochd should be running the renders with an SS of 2 anyway. You should contact him.

Thanks for the kind words. As you can see, it’s hard to get it working perfectly on everyone’s hardware. I think the new nvidia cards probably have some issues.

[Matt Feemster]

Could not reproduce. Use SS 2 as a workaround.

[Matt Feemster]

Hi Rafael, I’m not sure if you are around, but another user reported this same issue. We’ve resolved it and it’ll be fixed in the next release.

[Rafael Gassner]

I do am, and using fractorium daily . Thank you sir!