On mercredi 19 août 2020 23:00:17 CEST Guillermo Rozas wrote: > > Am Mittwoch, den 19.08.2020, 14:40 -0400 schrieb Michael: > > > I thought compiling led to faster? > > > > This might be the case or not. Depending on your CPU. When you > > compile for your machine, the compiler can set all optimization > > options according to your system. A package maintiner however > > has to compile it in a way that it runs on all possible CPUs > > the package is meant for. This means it can not take best advantage > > of non standart CPU features you might have paid you CPU mmanufacturer > > for. A special built for your system can. > > That was the case before (although I don't know how much faster). However, > last time I asked I was told that darktable's binaries are now compiled for > different architectures [1], so this should be less important. So... YMMV ;) > > Regards, > Guillermo > > [1] https://discuss.pixls.us/t/darktable-3-0-1-compilation-issues/17084/4
A lot of the calculation-intensive work is done through openCL, and those parts are compiled at *runtime*. So getting the non-openCL part a few percent faster isn't going to have that much influence, the more as that part concerns user interaction. And user-interaction is boring for computers (they spend most of their time waiting...) And getting the fastest possible executable means picking the right compiler settings, not always that straigh-forward. Remco ____________________________________________________________________________ darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
