bad allocation and workaround

[Alvaro L]

hi there long time not seen

My modest computer is throwing bad allocation during alignment of 2200 photos even though is not chewing out all the available RAM memory. But at least two times I been able to work around this issue.

I keep throtling dow setting, particularly key point  / tie point count till I can complete an alignment, then within the same Agisoft session I repeat alignments crackin up setting till the desired key point / tie point combo, with no problem.

It seems that in this workaround Agisoft memcaching is able to find from previous alignments the missing bytes is not able to find with a direct alignment, which makes me think about the compilator / compilation choices Agisoft developers are making to compile.

I know the competition is hard but maybe there is a compilation issue to improve on with safer memcache choices instead of performance. Cheers

[olihar]

You are definitely running out of RAM.

[Alvaro L]

I dont think so.

With the workaround described I can align even with higher key point / tie point settings than with a direct alignment with a bad allocation event. RAM monitor patters also differ a lot. Also the alignment settings at which Agisoft stop twrowing bad allocation are usually very low.

There is few scenarios at which some like this could happen, one of them is compilation time memcaching choices.

[Alexey Pasumansky]

Hello Alvaro,

Can you share the full processing log corresponding to failed Align Photos operation and the screenshots of the Align Photos dialog with the parameters used and of GPU Preferences tab?