Using a .KRA image made with 4.0 Pre-Alpha 1, the time it takes to filter it is almost instant, but with the latest Pre-Alpha, the performance regression is so severe that people who use non-destructive editing+instanced layers will no longer be able to work with later version of 4.0 on some images. For instances, the attached file. It is a small image taking 20.6 of memory space. In my 32 GB RAM computer, it should work nearly like an instance, and yet the filtering takes forever.
What attached file?
Created attachment 108766 [details]
Wolthera mentioned to test the file without fill layer, but in a newly created document, fill layer seem to be the source of the performance regression. In that document, newly created color adjustment filter layer or Power/Slope/Output Filter Layer sometimes do cause performance regression. I believe the issues lies within fill layer, color adjustment, and the power/slope/output filter layer.
Yes, I can confirm a performance regression. Opening the image in 4.0-prealpha.1 is fine; in git master, a20c53eea0c1548502b78a73879488a09a27844c, it starts an endless round of recalculations.
Okay, I think this bug report can be closed as it's invalid, but I do think this bug report reveals a big issue that needs to be addressed at some point. If someone works with a bazillion filter layers/clone layers, then it can be difficult to track them, and to resolve this issue, there needs to be some way to reveal how the document works via a network to track those filters/clones easier.
If the developers believe that there needs to be some way to track those layers down easier, I'll make the bug report if any developers asks me to do so if necessary, and let me know on the decision on that.
Please disregard the above. That was meant for another report which has to do with a problem with many clone layers.
*** This bug has been marked as a duplicate of bug 407062 ***