Focus stacking performance 1.9.0.932 vs 1.9.1.952

[Freixas]

The comparison is between 1.9.0.932 (release) with hardware acceleration disabled (because it has a bug) and the 1.9.1.952 (beta) with hardware acceleration enabled (Radeon RX 580 series). The problem is that there is no performance difference.

I used a stack of 13 images. With the release version, no acceleration, it took 45 seconds. With the beta, accelerated, it took 43 seconds, then 46 seconds when I repeated.

In the post where I noted the bug in the release version, Walt Farrell suggested I report the beta performance (non)difference in this beta forum.

Edited February 11, 2021 by Freixas
Typo

[Chris B]

Hey Freixas,

Last week I asked if OpenCL should reduce the time in which a stack should be completed and I was told that in theory yes this should be possible, however for now due to the early implementation of OpenCL we want to focus on stability for stacking. I imagine speed improvements could come at a later date. 

[Freixas]

5 hours ago, Chris B said:

Hey Freixas,

Last week I asked if OpenCL should reduce the time in which a stack should be completed and I was told that in theory yes this should be possible, however for now due to the early implementation of OpenCL we want to focus on stability for stacking. I imagine speed improvements could come at a later date. 

Thanks, Chris. Actually, focus stacking in Affinity has always been (in my opinion) fast. I only posted the performance measurement because someone else asked me to.

I downloaded Helicon Focus because it (and Zerene Stacker) were most often mentioned in a macro group that I participate in. Affinity Photo seems as fast as Helicon and, despite the lack of any options to direct the stacking process, seems to produce as good or better results.There seems little awareness of this in the macro community. Helicon and Zerene are both more expensive the Affinity Photo and all they do is stacking.

Stability is of primary importance so that is the correct thing to prioritize.