Closed: 4.6.4 b1: More sluggish?

[Hayo Baan]

I can't really quantify this, but 4.6.4 b1 seems a lot more sluggish to me than 4.6.3. For instance, I seem to get the spinning beach ball much more often than before. Am I seeing ghosts here, or is this really something that has changed?

Note: running 4.6.4 b1 on a Mac with OS X 10.6.3.

[Kirk Baker]

I can't really quantify this, but 4.6.4 b1 seems a lot more sluggish to me than 4.6.3. For instance, I seem to get the spinning beach ball much more often than before. Am I seeing ghosts here, or is this really something that has changed?

Note: running 4.6.4 b1 on a Mac with OS X 10.6.3.

I don't know.  Under what circumstances are you seeing a slowdown?  If you run the Activity Monitor do you see any unusual resource usage by Photo Mechanic?

-Kirk

[Hayo Baan]

As I said, I can't really quantify the feeling of sluggishness and as the periods of "waiting" are pretty brief (1-2 seconds, where previously the actions seemed instantaneous), any measuring of anomalies in performance is quite hard. So far I did see PM spike cpu usage to ±100% just after opening a contact sheet with about 170 images in it, but I think this is quite normal.

I'll monitor things a bit more and also try to find occasions where I can "reproduce" sluggishness.

I could still be seeing ghosts of course (and perhaps the recent OS update to 10.6.3 is also influencing stuff).

Cheers,
     Hayo

[Kirk Baker]

Hayo,

As I said, I can't really quantify the feeling of sluggishness and as the periods of "waiting" are pretty brief (1-2 seconds, where previously the actions seemed instantaneous), any measuring of anomalies in performance is quite hard. So far I did see PM spike cpu usage to ±100% just after opening a contact sheet with about 170 images in it, but I think this is quite normal.

I'll monitor things a bit more and also try to find occasions where I can "reproduce" sluggishness.

I could still be seeing ghosts of course (and perhaps the recent OS update to 10.6.3 is also influencing stuff).

One change that was made was that when various parts of Photo Mechanic are used and settings are changed, the Preferences file is written out.  This can take a second or so.  Prior to making this change, it was possible for settings to be lost if Photo Mechanic were to crash before being shut down completely.

So perhaps that is the issue, though Preferences should not be saved when a Contact Sheet is opened.

-Kirk

[Hayo Baan]

Kirk,

I investigated some more and 4.6.4 b1 seems more sluggish when I perform a "complex" (e.g., one involving exif data) sort on a contact sheet. This seems to slow down the whole of PM quite a bit, and in fact it seems to block every other action I want to perform within PM (e.g., switch to a different contact sheet), at least for a while (until the sort is done???). To my knowledge this didn't use to happen before. Has anything changed in this respect?

Thanks,
    Hayo

[Kirk Baker]

Hayo,

I investigated some more and 4.6.4 b1 seems more sluggish when I perform a "complex" (e.g., one involving exif data) sort on a contact sheet. This seems to slow down the whole of PM quite a bit, and in fact it seems to block every other action I want to perform within PM (e.g., switch to a different contact sheet), at least for a while (until the sort is done???). To my knowledge this didn't use to happen before. Has anything changed in this respect?

No, not to my knowledge.  Sorting is always done on the main event thread (it really can't be done on a background thread) and this behavior hasn't changed since it was implemented.

What are your sort parameters?  If you run an older version and do the same thing, does it take less time to perform the sorting?

-Kirk

[Hayo Baan]

I investigated some more and 4.6.4 b1 seems more sluggish when I perform a "complex" (e.g., one involving exif data) sort on a contact sheet. This seems to slow down the whole of PM quite a bit, and in fact it seems to block every other action I want to perform within PM (e.g., switch to a different contact sheet), at least for a while (until the sort is done???). To my knowledge this didn't use to happen before. Has anything changed in this respect?

No, not to my knowledge.  Sorting is always done on the main event thread (it really can't be done on a background thread) and this behavior hasn't changed since it was implemented.

What are your sort parameters?  If you run an older version and do the same thing, does it take less time to perform the sorting?

Kirk, the sort in question was eg. on colour class (I think), but it could also have been selecting "untagged" (chosen right after opening the sheet).

Weird thing though is that I just opened a contact sheet with only 15 files (ok, 5 of those are really big tifs; >300MB each) and I got the beach ball for about half a second. I then tried the same thing with 4.6.3: no beach ball. However, retrying with 4.6.4.b1, yielded no beach ball either so I can't reproduce it properly. (Note: I have PM clean the cache when closing so this can't be it).

Things get stranger still: when I try more things and get the beach ball, I don't get the beach ball when I try it again after quitting and restarting PM…

My guess therefore now is that I am seeing some internal OS X caching behaviour that may have changed in 10.6.3, and not a PM problem per se, and that the beach balls and resulting "sluggishness" come from there. Pity I can't verify this by going back to 10.6.2…

Does this sound reasonable to you (do you perhaps know of changes in 10.6.3. in this respect, I don't…)?

Thanks,
     Hayo

[Hayo Baan]

Hmm, I noticed something else as well. Photomechanic only seems to really use one CPU on my system (max CPU around 100%). I could be mistaken, but I seem to remember it would use more cores when loading a contact sheet. I have verified with 4.6.3 and 4.6.4 b2 though and both spike only at around 100%

So either something has changed before 4.6.3 in this respect, or something strange in combination with OS X 10.6.3 is happening, or I am seeing ghosts… 

Puzzled,
    Hayo