We do everything we can to hide that latency - but grouping a number of huge layers, scaling them down, then moving them around is pretty much a perfect storm - so if you are repeatedly doing that, fast, at some point Photo will panic and start paging to disk.Īnyway, thanks again for helping to get this nailed down - we should now be back to 1.6-quality memory management, which hopefully, coupled with the improvements in the new 1.7 compositor, will yield a decent improvement for 1.7.2. By default, Pixave scan your photographs in JPEG format with the least level of compression, resulting in high quality images that are reasonably large in size. It's fair to expect Photo on iPad to page to disk at some point in pretty much any large project. Repeat this say 20 times quickly (so Photo doesn't have chance to optimise / deduplicate history in idle time) and you are using ~60GB of memory on a device which has (at most) 2GB of usable physical memory - on a 2GB/3GB device, only 1GB is usable.

You have 5 layers (500MB). You then scale them down by about 50% - Photo is lossless when scaling layers, so they balloon from 25MP per layer to 100MP per layer - 400MB. At this point, your 5 layers cost 2GB - not including the ~1GB undo history created by the initial steps. Worth saying though that there are still limits - your example of a 5000x5000 RGBA 8bit canvas - one layer is 25MP (100MB). Thanks for this thread - it has helped identify a significant problem which I am very pleased we have got fixed before the 1.7.2 release. Turns out that in builds since 1.7.0, we have, under certain circumstances, used only half the memory available to us on iPad. This causes disk thrashing and explains some of the terrible slowdowns you have been experiencing.

Pixave dropbox