ilia3101
Dual ISO single frame computation speedup
Parallelize Dual ISO's demosaic + parallelize interpolation. Dual ISO computation for a single frame takes ~28% less time (tested on 2 of my PCs, depending on CPU speed and core number).
I tested full rendering (+ DNG rendering) with dozens of MLV videos I made, and the result is the exact same, but faster. If someone is interested in testing the changes themselves, it would be great.
Fixes in part #86 (I will send another PR in a different area that speeds DNG rendering further soon, if testing goes well)
I tried this out now. I don't find the reason why, but here this needs >200% of time to what we had before (42sec vs. 20sec). So it is half speed. I checked it several times. Do you have any idea what happens?! I tested on a MacBook Air M1, LLVM16 (by brew), Qt5.15.10. The result looks okay. CPU usage is near 100%.
I don't find the reason why, but here this needs >200% of time to what we had before (42sec vs. 20sec).
It's weird that it takes that much. Looks like thread overhead for me. In what mode were you rendering in? Cinema DNG?
I rendered into AVFoundation, so this is the Apple hardware encoder, and I get full processing CPU power. I almost never use cDNG - I use it just for testing the App.