Denoising crashes on cpu after rendering a few tiles (2.79b) #54466

New Issue

Benjamin · 2018-03-30T13:27:15+02:00

Benjamin commented

2018-03-30 13:27:15 +02:00

System Information
Windows 10 Pro 32 bit
Intel Core 2 duo
radeon hd 3400

Blender Version
Broken: (version 2.79, hash: 5bd8ac9, 32 bit), (version 2.79b, hash: f4dc9f9d68, 32 bit)

Short description of error
If I try to render any scene with cycles on my cpu with denoising enabled, Blender crashes after rendering a few tiles. My taskmanager doesn't show any critical RAM usage. Here is the output log (with the debug option): blender_denoise_debug_output.txt
It throws an error which seems to be related to RAM, but in the log Blender doesn't seems to reserve much RAM, it only needs a few mb, and I have hundreds of mb of free ram

Exact steps for others to reproduce the error
Just take the factory startup file, change the renderer to cycles, change the tile size for example to 16 x 16, enable denoising without tweaking something, and hit f12. I'll attach the file anyways: crashScene.blend

**System Information** Windows 10 Pro 32 bit Intel Core 2 duo radeon hd 3400 **Blender Version** Broken: (version 2.79, hash: 5bd8ac9, 32 bit), (version 2.79b, hash: f4dc9f9d68b, 32 bit) **Short description of error** If I try to render any scene with cycles on my cpu with denoising enabled, Blender crashes after rendering a few tiles. My taskmanager doesn't show any critical RAM usage. Here is the output log (with the debug option): [blender_denoise_debug_output.txt](https://archive.blender.org/developer/F2549036/blender_denoise_debug_output.txt) It throws an error which seems to be related to RAM, but in the log Blender doesn't seems to reserve much RAM, it only needs a few mb, and I have hundreds of mb of free ram **Exact steps for others to reproduce the error** Just take the factory startup file, change the renderer to cycles, change the tile size for example to 16 x 16, enable denoising without tweaking something, and hit f12. I'll attach the file anyways: [crashScene.blend](https://archive.blender.org/developer/F2549082/crashScene.blend)

blender-admin commented

2018-03-30 13:27:15 +02:00

#55910 was marked as duplicate of this issue

Benjamin commented

2018-03-30 13:27:16 +02:00

Added subscriber: @BenjaminMuller

Brecht Van Lommel commented

2018-04-01 07:31:17 +02:00

Added subscribers: @LukasStockner, @brecht

Brecht Van Lommel commented

2018-04-01 07:31:17 +02:00

I guess this happens on 32 bit only, we don't get a lot of testing there. Can you attach the output of Help > System Info? So we can see which instructions are supported by your CPU.

@LukasStockner, can you look into this? Otherwise I'll check it.

I guess this happens on 32 bit only, we don't get a lot of testing there. Can you attach the output of Help > System Info? So we can see which instructions are supported by your CPU. @LukasStockner, can you look into this? Otherwise I'll check it.

Benjamin commented

2018-04-01 12:11:35 +02:00

You mean msinfo32.exe, right? (At least on my 32 bit system)

Brecht Van Lommel commented

2018-04-01 12:47:22 +02:00

I mean inside Blender, open the Help menu and save the System Info from there.

Lukas Stockner commented

2018-04-01 13:04:51 +02:00

@brecht I can reproduce the crash using the 32-bit Windows build of 2.79b with Wine, I'll reboot to Windows and take a look.

Benjamin commented

2018-04-01 13:19:36 +02:00

Here is the system info file:
system-info.txt

Here is the system info file: [system-info.txt](https://archive.blender.org/developer/F2565228/system-info.txt)

Brecht Van Lommel commented

2018-04-01 14:05:26 +02:00

@LukasStockner, if it's a SIMD issue we can disable it for 32 bit for the filter kernels, we already do that for GCC/Clang since #36316.

Lukas Stockner commented

2018-04-03 18:44:49 +02:00

Okay, so it turns out that this is caused by one of the pointers in the NLM kernels not being aligned, but it's actually already fixed in the current master.

So, I'm not sure whether there's a point in fixing this for 2.79 considering that there probably won't be another bugfix release?

Okay, so it turns out that this is caused by one of the pointers in the NLM kernels not being aligned, but it's actually already fixed in the current master. So, I'm not sure whether there's a point in fixing this for 2.79 considering that there probably won't be another bugfix release?

Benjamin commented

2018-04-04 12:18:22 +02:00

Ok, thanks! Then I will just build it by myself.