On macOS, the fastest keyboard repeat rate is thirty repeats per second. The next notch down is ten repeats per second. If you're testing by holding down the right arrow key, then the rate at which you can view images will be governed at a minimum by that rate. There can be other limiting factors like disk speed, CPU speed, memory bandwidth, etc. that can slow things down further. For instance, browsing on a fast internal SSD will always outpace browsing on a network drive.
The crux of the problem with PM6 and macOS Ventura is that it was taking a long time to physically draw the image on the window. For our tests, we precache the images by having a limited set of photos (in my test it was 365, 8 megapixel images) and then test the speed of browsing.
We have some timing tests that we can run on our internal builds that take the user out of the equation. Here are some example timings for build 6645 and 6675, with all images pre-cached which is at this point, timing mostly the 'draw speed'. In my tests I am hiding the thumbnail gallery, hiding the histogram, hiding the info text, and not having the contact sheet scroll along with the preview. Each of those features has some overhead. The window is as large as I can make it on a 4k display.
Mac OS Mojave, Mac Pro 2012 128GB RAM, build 6645: 10.969 previews per second
Mac OS Mojave, Mac Pro 2012 128GB RAM, build 6675: 36.008 previews per second
Mac OS Ventura, Mac Mini M1 16GB RAM, build 6675: 58.445 previews per second
I don't have a timing for build 6645 on my Mac Mini, unfortunately. Though with the above timings, you can see that the Rosetta2 emulated build of 6675 is nearly twice as fast as my Intel CPU system with no emulation.
Perhaps a better test is to see how much lag there is to go between two images (back and forth) with you alternating key presses of the right and left arrow keys. Is it responsive? Or does it lag? If there's no lag with this test and there are lags elsewhere then it's not a draw speed issue. The issue lies somewhere else and would have to be investigated.
A faster key repeat rate would be a better test from your end, given that you don't have access to our timing tests.
HTH,
-Kirk
Hi Kirk
Thanks for the detail of your testing, I think I’m in the fortunate position that the degraded performance of build 6645 and earlier builds on Ventura does not have a significant impact on my day to day use of PM with my particular hardware setup, particularly when using the external 4K monitor. I can see that build 6675 performs better on the laptop screen, but I don’t think I can stress the system enough with keyboard inputs to really see much difference on the 4K screen.
I tried my test again with the key repeat rate on "fast", just looking at the performance of the preview window this time. On the laptop screen build 6675 is significantly faster, 22 images per second as opposed to 5 images per second on build 6645. On the 4K screen the difference is much less, 23 images per second vs 19 images per second.
I also looked for lag by alternating presses on the left and right arrow keys as quickly as possible. On the laptop screen I can make build 6645 lag quite a bit, but build 6675 is fine: on the 4K screen I get no lag from either version, I just can’t press the keys fast enough to see any lag.
My conclusion is that build 6675 definitely has improved performance, but my system is performant enough that it has to be really stressed to show up the deficiencies in build 6645.
Thank you for all your efforts in addressing this issue, I’ll look forward to a formally released build with the changes included.
