Closed heckflosse closed 5 years ago
I already looked at the code and everything makes sense.
When i get home shortly I'll test it.
On my 2700X, the 1500px previews were 14% faster including demosaicing, and the full res renders were 22% faster not including demosaicing.
@CarVac
On my 2700X, the 1500px previews were 14% faster including demosaicing, and the full res renders were 22% faster not including demosaicing.
permamanantly small improvemts lead to large improvemets at long term :-)
Some remarks to the commits in this branch:
Fix xtrans premultiply
ist just a fix for xtrans files, which were not handled correctly beforeReduce memory usage. Also some small speedups
reduces memory usage and buffer ping-pong when bayer-ca-correction is used and adds some #pragma omp stuffDo not reallocate memory if matrix size does not change
reduces memory reallocations when adjusters are changed after loading an image and leads to some speedups in this case (as it avoids reallocation of large buffers of same size)Reduce buffer ping-pong for hl reconstruction settings 0 and 1 and for whitebalance
does what its topic sais and gives also speedups for this cases.Further speedups, mainly by reducing buffer ping-pong
does what its topic sais and gives also speedups for this cases.