Project

General

Profile

Bug #11772

Some IOP are CPU blended

Added by Paolo Astengo about 1 month ago. Updated 28 days ago.

Status:
Incomplete
Priority:
Low
Assignee:
-
Category:
OpenCL
Target version:
-
Start date:
10/18/2017
Due date:
% Done:

20%

Affected Version:
git master branch
System:
Ubuntu
bitness:
64-bit
hardware architecture:
amd64/x86

Description

I'm using a GTX960M with opencl enabled in "VERYFASTGPU" mode, no errors appears on cli, but some operations, like "demosaic", "tonemapping" and "equalizer" seem to be executed only in CPU mode (while they work n GPU mode for thumbnails).

[dev_pixelpipe] took 0.004 secs (0.019 CPU) initing base buffer [full]
[dev_pixelpipe] took 0.022 secs (0.042 CPU) processed `raw black/white point' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.006 secs (0.017 CPU) processed `white balance' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.007 secs (0.021 CPU) processed `highlight reconstruction' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.305 secs (1.560 CPU) processed `demosaic' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.243 secs (1.548 CPU) processed `tone mapping ' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.021 secs (0.019 CPU) processed `base curve' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.014 secs (0.001 CPU) processed `input color profile' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.078 secs (0.035 CPU) processed `shadows and highlights ' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.781 secs (5.089 CPU) processed `equalizer ' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.036 secs (0.027 CPU) processed `local contrast' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.015 secs (0.010 CPU) processed `color zones ' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.011 secs (0.005 CPU) processed `tone curve ' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.028 secs (0.003 CPU) processed `sharpen' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.282 secs (2.060 CPU) processed `output color profile' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.020 secs (0.014 CPU) processed `channel mixer' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.036 secs (0.070 CPU) processed `gamma' on CPU blended on CPU [full]

Do I'm doing something wrong?

dtout.txt Magnifier (134 KB) Paolo Astengo, 10/18/2017 03:51 PM

cltest.txt Magnifier (25.5 KB) Paolo Astengo, 10/18/2017 03:55 PM

dtest1.txt Magnifier (26.7 KB) Paolo Astengo, 10/21/2017 07:35 PM

20150424_0010.NEF (9.92 MB) Paolo Astengo, 10/21/2017 07:39 PM

20150424_0010_02.NEF.xmp (4.52 KB) Paolo Astengo, 10/21/2017 07:39 PM

dtest4.txt Magnifier (30.4 KB) Paolo Astengo, 10/21/2017 08:34 PM

dtest5.txt Magnifier (574 KB) Paolo Astengo, 10/21/2017 10:32 PM

nvidia.tx (1.68 KB) Paolo Astengo, 10/22/2017 12:00 AM

nvidia-smi.out (1.52 KB) Ulrich Pegelow, 10/22/2017 09:42 AM

nvidia-smi.out (1.59 KB) thokster ., 10/22/2017 10:19 AM

History

#1 Updated by Roman Lebedev about 1 month ago

  • % Done changed from 0 to 20
  • Status changed from New to Incomplete

Paolo Astengo wrote:

I'm using a GTX960M with opencl enabled in "VERYFASTGPU" mode, no errors appears on cli, but some operations, like "demosaic", "tonemapping" and "equalizer" seem to be executed only in CPU mode (while they work n GPU mode for thumbnails).

Please also pass -d dev and show the output of the same image export as this one \/
If the module was processed on CPU, or on GPU but with tiling, the blending is done on CPU.

[dev_pixelpipe] took 0.004 secs (0.019 CPU) initing base buffer [full]
[dev_pixelpipe] took 0.022 secs (0.042 CPU) processed `raw black/white point' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.006 secs (0.017 CPU) processed `white balance' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.007 secs (0.021 CPU) processed `highlight reconstruction' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.305 secs (1.560 CPU) processed `demosaic' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.243 secs (1.548 CPU) processed `tone mapping ' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.021 secs (0.019 CPU) processed `base curve' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.014 secs (0.001 CPU) processed `input color profile' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.078 secs (0.035 CPU) processed `shadows and highlights ' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.781 secs (5.089 CPU) processed `equalizer ' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.036 secs (0.027 CPU) processed `local contrast' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.015 secs (0.010 CPU) processed `color zones ' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.011 secs (0.005 CPU) processed `tone curve ' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.028 secs (0.003 CPU) processed `sharpen' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.282 secs (2.060 CPU) processed `output color profile' on CPU blended on CPU [full]
[dev_pixelpipe] took 0.020 secs (0.014 CPU) processed `channel mixer' on GPU blended on GPU [full]
[dev_pixelpipe] took 0.036 secs (0.070 CPU) processed `gamma' on CPU blended on CPU [full]

Do I'm doing something wrong?

#2 Updated by Paolo Astengo about 1 month ago

Hi Roman, this is the output of the same image, same processing:

[dev] took 0.000 secs (0.000 CPU) to load the image.
[dev] took 0.000 secs (0.000 CPU) to load the image.
[dev] took 0.000 secs (0.000 CPU) to load the image.
[export] creating pixelpipe took 0.021 secs (0.059 CPU)
pixelpipe cacheline 0 used 0 by 18446744073709551615
pixelpipe cacheline 1 used 0 by 18446744073709551615
cache hit rate so far: -nan
[dev_pixelpipe] took 0.012 secs (0.029 CPU) initing base buffer [thumbnail]
[dev_pixelpipe] took 0.015 secs (0.008 CPU) processed `raw black/white point' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.012 secs (0.010 CPU) processed `white balance' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.007 secs (0.001 CPU) processed `highlight reconstruction' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.012 secs (0.022 CPU) processed `demosaic' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.008 secs (0.015 CPU) processed `tone mapping ' on CPU, blended on CPU [thumbnail]
[dev_pixelpipe] took 0.003 secs (0.000 CPU) processed `base curve' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.003 secs (0.000 CPU) processed `input color profile' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.004 secs (0.000 CPU) processed `shadows and highlights ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.009 secs (0.000 CPU) processed `equalizer ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.002 secs (0.001 CPU) processed `local contrast' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.002 secs (0.001 CPU) processed `color zones ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.005 secs (0.006 CPU) processed `tone curve ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.008 secs (0.003 CPU) processed `sharpen' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.001 secs (0.009 CPU) processed `output color profile' on CPU, blended on CPU [thumbnail]
[dev_pixelpipe] took 0.002 secs (0.001 CPU) processed `channel mixer' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.000 secs (0.001 CPU) processed `gamma' on CPU, blended on CPU [thumbnail]
[dev_process_thumbnail] pixel pipeline processing took 0.106 secs (0.106 CPU)
pixelpipe cacheline 0 used 0 by 18446744073709551615
pixelpipe cacheline 1 used 0 by 18446744073709551615
pixelpipe cacheline 2 used 0 by 18446744073709551615
pixelpipe cacheline 3 used 0 by 18446744073709551615
pixelpipe cacheline 4 used 0 by 18446744073709551615
cache hit rate so far: 0.000
[dev_pixelpipe] took 0.006 secs (0.024 CPU) initing base buffer [full]
[dev_pixelpipe] took 0.016 secs (0.011 CPU) processed `raw black/white point' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.005 secs (0.000 CPU) processed `white balance' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.013 secs (0.003 CPU) processed `highlight reconstruction' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.156 secs (0.128 CPU) processed `demosaic' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.326 secs (1.613 CPU) processed `tone mapping ' on CPU, blended on CPU [full]
[dev_pixelpipe] took 0.022 secs (0.020 CPU) processed `base curve' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.017 secs (0.000 CPU) processed `input color profile' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.075 secs (0.034 CPU) processed `shadows and highlights ' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.924 secs (4.990 CPU) processed `equalizer ' on CPU, blended on CPU [full]
[dev_pixelpipe] took 0.036 secs (0.023 CPU) processed `local contrast' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.015 secs (0.006 CPU) processed `color zones ' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.011 secs (0.002 CPU) processed `tone curve ' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.026 secs (0.013 CPU) processed `sharpen' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.287 secs (2.048 CPU) processed `output color profile' on CPU, blended on CPU [full]
[dev_pixelpipe] took 0.020 secs (0.019 CPU) processed `channel mixer' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.037 secs (0.063 CPU) processed `gamma' on CPU, blended on CPU [full]
[dev_process_image] pixel pipeline processing took 2.094 secs (9.104 CPU)
pixelpipe cacheline 0 used 0 by 18446744073709551615
pixelpipe cacheline 1 used 0 by 18446744073709551615
pixelpipe cacheline 2 used 0 by 18446744073709551615
pixelpipe cacheline 3 used 0 by 18446744073709551615
pixelpipe cacheline 4 used 0 by 18446744073709551615
cache hit rate so far: 0.000
[dev_pixelpipe] took 0.000 secs (0.000 CPU) initing base buffer [preview]
[dev_pixelpipe] took 0.003 secs (0.000 CPU) processed `raw black/white point' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.002 secs (0.000 CPU) processed `white balance' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.002 secs (0.000 CPU) processed `highlight reconstruction' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.005 secs (0.000 CPU) processed `demosaic' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.097 secs (0.385 CPU) processed `tone mapping ' on CPU, blended on CPU [preview]
[dev_pixelpipe] took 0.010 secs (0.001 CPU) processed `base curve' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.009 secs (0.000 CPU) processed `input color profile' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.022 secs (0.016 CPU) processed `shadows and highlights ' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.042 secs (0.007 CPU) processed `equalizer ' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.006 secs (0.005 CPU) processed `local contrast' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.011 secs (0.000 CPU) processed `color zones ' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.018 secs (0.024 CPU) processed `tone curve ' on GPU, collected histogram on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.006 secs (0.001 CPU) processed `sharpen' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.074 secs (0.491 CPU) processed `output color profile' on CPU, blended on CPU [preview]
[dev_pixelpipe] took 0.007 secs (0.006 CPU) processed `channel mixer' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.007 secs (0.011 CPU) processed `gamma' on CPU, blended on CPU [preview]
[dev_process_preview] pixel pipeline processing took 0.341 secs (0.974 CPU)

#3 Updated by Roman Lebedev about 1 month ago

Paolo Astengo wrote:

Hi Roman, this is the output of the same image, same processing:

[dev] took 0.000 secs (0.000 CPU) to load the image.
[dev] took 0.000 secs (0.000 CPU) to load the image.
[dev] took 0.000 secs (0.000 CPU) to load the image.
[export] creating pixelpipe took 0.021 secs (0.059 CPU)
pixelpipe cacheline 0 used 0 by 18446744073709551615
pixelpipe cacheline 1 used 0 by 18446744073709551615
cache hit rate so far: -nan
[dev_pixelpipe] took 0.012 secs (0.029 CPU) initing base buffer [thumbnail]
[dev_pixelpipe] took 0.015 secs (0.008 CPU) processed `raw black/white point' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.012 secs (0.010 CPU) processed `white balance' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.007 secs (0.001 CPU) processed `highlight reconstruction' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.012 secs (0.022 CPU) processed `demosaic' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.008 secs (0.015 CPU) processed `tone mapping ' on CPU, blended on CPU [thumbnail]
[dev_pixelpipe] took 0.003 secs (0.000 CPU) processed `base curve' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.003 secs (0.000 CPU) processed `input color profile' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.004 secs (0.000 CPU) processed `shadows and highlights ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.009 secs (0.000 CPU) processed `equalizer ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.002 secs (0.001 CPU) processed `local contrast' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.002 secs (0.001 CPU) processed `color zones ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.005 secs (0.006 CPU) processed `tone curve ' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.008 secs (0.003 CPU) processed `sharpen' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.001 secs (0.009 CPU) processed `output color profile' on CPU, blended on CPU [thumbnail]
[dev_pixelpipe] took 0.002 secs (0.001 CPU) processed `channel mixer' on GPU, blended on GPU [thumbnail]
[dev_pixelpipe] took 0.000 secs (0.001 CPU) processed `gamma' on CPU, blended on CPU [thumbnail]
[dev_process_thumbnail] pixel pipeline processing took 0.106 secs (0.106 CPU)
pixelpipe cacheline 0 used 0 by 18446744073709551615
pixelpipe cacheline 1 used 0 by 18446744073709551615
pixelpipe cacheline 2 used 0 by 18446744073709551615
pixelpipe cacheline 3 used 0 by 18446744073709551615
pixelpipe cacheline 4 used 0 by 18446744073709551615
cache hit rate so far: 0.000
[dev_pixelpipe] took 0.006 secs (0.024 CPU) initing base buffer [full]
[dev_pixelpipe] took 0.016 secs (0.011 CPU) processed `raw black/white point' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.005 secs (0.000 CPU) processed `white balance' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.013 secs (0.003 CPU) processed `highlight reconstruction' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.156 secs (0.128 CPU) processed `demosaic' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.326 secs (1.613 CPU) processed `tone mapping ' on CPU, blended on CPU [full]
[dev_pixelpipe] took 0.022 secs (0.020 CPU) processed `base curve' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.017 secs (0.000 CPU) processed `input color profile' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.075 secs (0.034 CPU) processed `shadows and highlights ' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.924 secs (4.990 CPU) processed `equalizer ' on CPU, blended on CPU [full]
[dev_pixelpipe] took 0.036 secs (0.023 CPU) processed `local contrast' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.015 secs (0.006 CPU) processed `color zones ' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.011 secs (0.002 CPU) processed `tone curve ' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.026 secs (0.013 CPU) processed `sharpen' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.287 secs (2.048 CPU) processed `output color profile' on CPU, blended on CPU [full]
[dev_pixelpipe] took 0.020 secs (0.019 CPU) processed `channel mixer' on GPU, blended on GPU [full]
[dev_pixelpipe] took 0.037 secs (0.063 CPU) processed `gamma' on CPU, blended on CPU [full]
[dev_process_image] pixel pipeline processing took 2.094 secs (9.104 CPU)
pixelpipe cacheline 0 used 0 by 18446744073709551615
pixelpipe cacheline 1 used 0 by 18446744073709551615
pixelpipe cacheline 2 used 0 by 18446744073709551615
pixelpipe cacheline 3 used 0 by 18446744073709551615
pixelpipe cacheline 4 used 0 by 18446744073709551615
cache hit rate so far: 0.000
[dev_pixelpipe] took 0.000 secs (0.000 CPU) initing base buffer [preview]
[dev_pixelpipe] took 0.003 secs (0.000 CPU) processed `raw black/white point' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.002 secs (0.000 CPU) processed `white balance' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.002 secs (0.000 CPU) processed `highlight reconstruction' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.005 secs (0.000 CPU) processed `demosaic' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.097 secs (0.385 CPU) processed `tone mapping ' on CPU, blended on CPU [preview]
[dev_pixelpipe] took 0.010 secs (0.001 CPU) processed `base curve' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.009 secs (0.000 CPU) processed `input color profile' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.022 secs (0.016 CPU) processed `shadows and highlights ' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.042 secs (0.007 CPU) processed `equalizer ' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.006 secs (0.005 CPU) processed `local contrast' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.011 secs (0.000 CPU) processed `color zones ' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.018 secs (0.024 CPU) processed `tone curve ' on GPU, collected histogram on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.006 secs (0.001 CPU) processed `sharpen' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.074 secs (0.491 CPU) processed `output color profile' on CPU, blended on CPU [preview]
[dev_pixelpipe] took 0.007 secs (0.006 CPU) processed `channel mixer' on GPU, blended on GPU [preview]
[dev_pixelpipe] took 0.007 secs (0.011 CPU) processed `gamma' on CPU, blended on CPU [preview]
[dev_process_preview] pixel pipeline processing took 0.341 secs (0.974 CPU)

Here, only the equalizer is processed on CPU for export.
I would guess there isn't enough GPU memory.
Please show the output of $ darktable-cltest

#4 Updated by Paolo Astengo about 1 month ago

Just to be sure to include any possible information, I'm attaching the output for -d all

#5 Updated by Paolo Astengo about 1 month ago

Here is the output

#6 Updated by Roman Lebedev about 1 month ago

Paolo Astengo wrote:

Here is the output

[opencl_init] device 0 `GeForce GTX 960M' allows GPU memory allocations of up to 500MB

Well, that is the problem :/
This is most likely Closed: Invalid

#7 Updated by Paolo Astengo about 1 month ago

I guessed 2GB card could be enough, sorry if I didn't check the max_alloc...

Do you know any way to tell the GPU to see some more space?

#8 Updated by Ulrich Pegelow 29 days ago

There can be several reasons why a module is not processed on the GPU. In most cases this is because there is no OpenCL kernel for that specific action. E.g. if you use Amaze demosaic: that method is only implemented for the CPU; for thumbnails, preview window and zoomed-out full view darktable uses a more simple demosaic algorithm with OpenCL support. Therefore GPU yes/no depends on the pixelpipe.

Concerning the equalizer. That module has full OpenCL support. Right now I can't tell why the GPU is not used. Even if GPU memory is tight, darktable should still use the GPU with tiling. Please run an export job of your image with 'darktable -d opencl'.

#9 Updated by Paolo Astengo 29 days ago

Hi Ulrich,

I'm attaching the trace, the raw and the sidecar, thanks

#10 Updated by Ulrich Pegelow 29 days ago

Oh, you have set opencl_memory_headroom to zero in darktablerc. That's not good. Your GPU will require some of the GPU memory for display purposes and internal data. This should be reflected in the a.m. config parameter. If you set this to zero darktable will assume that it may allocate all available GPU memory and this will make allocations fail in memory hungry IOPs (equalizer in your case).

You should restore the default setting of 300 or even go to 400.

#11 Updated by Paolo Astengo 29 days ago

I'm sorry, but it does not work.

#12 Updated by Ulrich Pegelow 29 days ago

Do you have other programs in background that require high loads of GPU memory? Try to close them when running darktable. Else you can increase opencl_memory_headroom in steps of 100.

#13 Updated by Paolo Astengo 29 days ago

I've just tried to increase the headroom in steps of 100 up to 1100, without results. All the programs are closed and this is the output of "free" command":

free -h
total used free shared buff/cache available
Mem: 15G 3,6G 8,6G 119M 3,3G 11G
Swap: 15G 0B 15G

#14 Updated by Paolo Astengo 29 days ago

Changing the memory headroom setting to 2400 (!), tiling starts working as you can see in the attached file.

2400 seems to me something huge, indeed...

#15 Updated by Ulrich Pegelow 29 days ago

Yep, that's an unreasonably high value. Doesn't make sense. There seems to be some other issue. BTW the 'free' command is of no help here. Please try 'nvidia-smi'.

#16 Updated by Paolo Astengo 29 days ago

nvidia-smi
Sat Oct 21 22:58:24 2017
--------------------------------------------------------------------------- | NVIDIA-SMI 387.12 Driver Version: 387.12 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 GeForce GTX 960M Off | 00000000:01:00.0 Off | N/A | | N/A 50C P0 N/A / N/A | 1025MiB / 2002MiB | 4% Default |
---------------------------------------------------+----------------------+

--------------------------------------------------------------------------- | Processes: GPU Memory | | GPU PID Type Process name Usage | |=============================================================================| | 0 1687 G /usr/lib/xorg/Xorg 71MiB | | 0 1772 G /usr/bin/gnome-shell 48MiB | | 0 2082 G /usr/lib/xorg/Xorg 391MiB | | 0 2252 G /usr/bin/gnome-shell 469MiB | | 0 3418 G ...-token=B3CE4FD90A6BA8BE76103039C54EF459 39MiB | | 0 17632 G evolution 1MiB |
---------------------------------------------------------------------------

#17 Updated by Paolo Astengo 29 days ago

here is the output

#18 Updated by Ulrich Pegelow 29 days ago

here is the output

This indicates an astonishly high GPU memory consumption of your system. For comparison I have attached the corresponding output here on my OpenSUSE 42.2. If you can't make your system use less GPU memory then your options are limited. Even if you set opencl_memory_headroom to 1100, which is what your output suggest, you can't be sure that the peak memory usage of your system might be higher.

#19 Updated by thokster . 29 days ago

Attached my nvidia-smi output for comparison.
Configured opencl_memory_headroom to 900 and I try to use no other applications during export.
Maybe there is nothing wrong with the system of Paolo, maybe it's just gnome-shell.

#20 Updated by Ulrich Pegelow 29 days ago

Could it be that gnome-shell makes heavy use of screen effects?

#21 Updated by thokster . 28 days ago

What's the difference between gnome-shell and a screen effect? ;-)

#22 Updated by Ulrich Pegelow 28 days ago

I was thinking of something like 3D effects - nice but potentially memory hungry.

#23 Updated by thokster . 28 days ago

If I deactivate animation I don't see any significant change in memory usage.

Also available in: Atom PDF