darktable and AMD performance in benchmarks

reading here: Optimize Options (Using the GNU Compiler Collection (GCC))

i think -ftree-vectorize is just a meta switch enabling two tree vectorisation features that are already included in -O2, so i think explicitly adding this switch will not change anything. aurelien’s experiments seem to show that indeed after a lot of manual work (wrt alignment and pointer restriction etc) some auto vectorisation does happen already.