BTW: I kept commenting on that closed PR. I know, :-) I said so, because I was indirectly pushing for this topic with my wish of detachable side panels. in my first quick attempt bisect ask me to remove src/external/rawspeed/ ? tricky as it could(!) be the culprit and I always would need to move it in and out during testing.Īu contraire, if its preview2 it will be easier for me to fix it.defining good/bad will become challenging, as I seem have to always recompile and test.
o.k., I found gitschooldude bisect tutorial and is pretty easy to understand.I will learn about bisect (unfortunately that will take time), saw a very first tutorial, so I get an idea.I expected a kind of roll-back or move forward tool.Git bisec is an utility that check-out a commit, you try it, report back and check-out another based on your feedback. I did a quicky try without recording the -d perf but -disable-opencl actually gives me the same user experience, where dt-2.6.2 is snappier than dt 2.7 (the lag feels the same) I did so donno, in which timezone you are in, but good morning :-) I hope that split is appreciated and could be implemented? I would like to open a new topic on for casual talk, awareness and share experiences, while keeping this issue here for the developers and solution oriented topics. darktable 2.7.0+1087~g7fe35d67d (found it time ago already).Platform (please complete the following information): It might be, we find either something totally different (anything else might influence the UI) or a hardware related issue. ?), that just now stated in #2064, for him, the 2.7 is way more responsive than 2.6.2 I don't want to misapply (correct word acc. (see section Expected behavior)īeing not slower, "laggier" than 2.6.2, ideally even faster, more responsive than 2.6.2 The lag I "feel" (until screen gets updated), seems actually even harder than the values I found here. in the WB-Chart I have marked some obvious findings, where in the other sheets, for "efficiency" reasons, merely marked the most obvious oneįrom there you can see, that dt-2.7 (at least here) responds way slower in total, than 2.6.2, while the single actions, they seem to be faster.In coloumn "B" I placed a formular like =B12-B11 in order to have the used time per step.In coloumn "A" I put a formular like =D11-$D$11 in order to eliminate the offset and have a relative time.All those I have done and concluded in attached.repeat 1-5 with tonecurve Gamma 0.5 and Zoom from 200% to fit2screen.Change WB from 2000 to 25000 (copy the relevant lines from log).(pixelpipe merely has sharpen, rotation and basecurve).(play raw, red tulips by pphoto CC-BY-NC 4.0) double preview triggered when zooming in #2064.High CPU load even opencl is working #2404.
This bug report, from my point of view (and I might be wrong), tries to comprehend: I think it is a bit complex and even I tried to be narrow, the report got a bit longer (also there is always a risk, I do make mistakes in English)
Please read this entirely, before you draw your conclusion. official 2.6.2, which makes it feel very sticky on my side. According to my observation, we do have a more than noticeable performance issue for darkroom in git-master vs.
0 kommentar(er)
