On Sat, Oct 25, 2014 at 12:43 PM, johannes hanika <hana...@gmail.com> wrote: > good, thanks for checking. > > we could replace that blocking raw read by a raw-metadata read, if there was > support for that in rawspeed (and if that would make any sense because it > would be any faster at all..).
Not right now and there are several issues with making that happen. While libraw parses by consuming input (fgetc and friends) rawspeed works by reading the whole file into RAM. We could mmap though. More importantly right now metadata and the bayer data are all returned in the same object and internally are produced by the same function. We'd have to do a pretty extensive change to fix that. I think it would make more sense to have an image switch logic that's something like this: - On switch very quickly put up one of our cached images as preview (the largest we have that fits) and grey out all the side panels (iop gui, history etc) - While in this state the filmstrip and the darkroom/lightroom/etc switcher should all still work - On a callback that says the raw has finished loading (or we got a MIP_F) actually do the rest of the process of setting up history, side panels, etc That way when paging through images quickly the visualization would work and on image change the visual feedback is immediate, but we don't have any race conditions between iops and raw loading. > how did you look at jpg? via film strip switching? it should not behave any > different to before the first 699 merge.. there should not be any difference > in reload_defaults() calls at all. I was switching back and forth with space/backspace. It now works fine for me with your latest commit. Cheers, Pedro ------------------------------------------------------------------------------ _______________________________________________ darktable-devel mailing list darktable-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/darktable-devel
