Artefact, calculation limits or mistake ?

phweyland · April 3, 2018, 1:50am

Hi,
It happens with some images to have some (dark) defects on the bottom (and only on the bottom).
Changing the size of the display changes the defect (as attached images).
On the exported image I’ve still some defects, just a bit different.
The upper left preview seems not to be affected.
For base curve I use Nikon D750 plus fusion (per default settings 1 and 1).
Has anyone experienced this ?
Any idea about the cause … and remedy ?

Windows 10 + darktable 2.4.2 (but I’ve seen similar effects with 2.4.1 and another plugin but I don’t remember which one, lowpass maybe ?)

Thanks

phweyland · April 3, 2018, 1:38pm

Found a work around. If I uncheck setting activate OpenCL support the defect goes off…
Let me know if more information or tests are needed to refine the issue.

darix · April 3, 2018, 1:42pm

you could show the output of darktable-cltest

phweyland · April 3, 2018, 2:24pm

darktable-cltest.txt (30.7 KB)

darix · April 3, 2018, 2:26pm

I am amazed it even tried to use opencl on your machine. In the linux version openCL is blocked on Intel GPUs.

phweyland · April 3, 2018, 3:26pm

Another point linked to activate OpenCL support in my case: in export selected, the style is not applied when activate OpenCL support is set.
When exporting images I’ve 5 to 10 % of images presenting this kind of effect, which disappears when I export them again (this time with or without activate OpenCL support):

(crop of the output image)

Pascal_Obry · April 3, 2018, 3:40pm

OpenCL should not be used with Intel card as not supported.

phweyland · April 3, 2018, 4:11pm

OK then ! Thanks for your quick answers !