There is none. Does more tiles per thread increase performance or reduce it?
Presumably that is the reason there is a log check box but how to use it?
That depends on your machine and your raw files.
You can measure it by enabling the measure checkbox and running RT from command line (to get the measures).
Good way to measure ist to put a file 7 times into the queue and take the median of the 7 measures. Then repeat the procedure for different Tiles per thread
I measured up to 20% speedup for RCD demosaic on my machine when switching between 1 and 6 tiles per thread
Yes that was the one I meant. I just thought it might be useful to have something in documentation saying that more tiles per thread might increase performance. (It is not obvious because if each thread had less work to do then things might be faster.)
Also your tip on how to use it (and saying that once you have set it then you can uncheck the measure box would be helpful.
@RichardRegal I just tested again. At the time I made this optimizations, there was a very clearly measurable speedup for tiles per thread > 1. My new tests (with gcc 10.x) show not much speedup (if at all), probably because I’m testing with a native build where gcc inserts good prefetch instructions for my machine. I will test now with clang builds and also generic x86_64 builds to check whether we will keep this stuff or not. Will take a day or two though.