1. 19 Feb, 2020 1 commit
  2. 18 Feb, 2020 9 commits
  3. 13 Feb, 2020 1 commit
  4. 03 Feb, 2020 1 commit
  5. 25 Jan, 2020 1 commit
    • Sergey Torokhov's avatar
      cltostring.py: force using utf8 · 7f851dc7
      Sergey Torokhov authored
      In some cases [1] (maybe due to user locale ussies) the system env codepage
      could differ from utf8. As some of .cl files contains non-ascii symbold
      e.g authora names the cltostring.py script then fails to convert .cl file
      to header file with the error message like:
      "UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 783:
      ordinal not in range(128)"
      This patch resolves such issues by forcing utf8 codepage while reading .cl.
      The resulting cltostring.py script is compatible both python2.7 and python3.x.
      When python2.7 support will be dropped the addition of
      'from __future__ import unicode_literals', 'import io'
      and 'io.' could be removed as not neccessary for python3.x.
      [1] https://bugs.gentoo.org/705286
  6. 17 Jan, 2020 2 commits
  7. 16 Jan, 2020 2 commits
    • Ell's avatar
      crop: don't clip output to input bounding box · 1298d41c
      Ell authored
      Let the bounding box of gegl:crop be exactly the cropped rect,
      rather than its intersection with the input's bounding box.  The
      latter used to work around a bug when the output of gegl:crop was
      used in-place, which should now be fixed.
    • Ell's avatar
      operation: only do in-place processing if ROI fits in input abyss · dd642566
      Ell authored
      In gegl_can_do_inplace_processing(), check if the ROI fits in the
      input buffer's abyss, rather than its extent, since we're
      discarding written data outside the abyss.
  8. 12 Jan, 2020 1 commit
  9. 09 Jan, 2020 1 commit
    • Ell's avatar
      buffer: in swap backend, only compress current tile when stalling · bb394b04
      Ell authored
      In the swap backend, when attempting to push a tile to a full
      queue, only compress the current tile before waiting for free queue
      space, instead of searching the rest of the queue for additional
      tiles to compress.  When the queue is very long, the recurring
      linear search itself can end up taking a significant amount of
      time, during which the queue must be locked, blocking both other
      threads attempting to push a tile simultaneously, as well as the
      swap-writer thread attempting to free queue space.
      This partially reverts commit
  10. 07 Jan, 2020 1 commit
  11. 06 Jan, 2020 2 commits
    • Marco Ciampa's avatar
      Updated Italian translation · 05f31371
      Marco Ciampa authored
    • Ell's avatar
      process: destroy node cache, instead of clearing, when new bounding box is smaller · caae2071
      Ell authored
      When a node's bounding box changes, if the old tile-aligned
      bounding box is not contained within the new tile-aligned bounding
      box, destroy the node's cache, instead of clearing the now-out-of-
      bounds region.  The problem with the latter is that the node's
      bounding box can be arbitrarily large, even without actually having
      been rendered; if the bounding box then becomes much smaller,
      clearing the difference can take a long time, even if the cache
      doesn't have any actual tiles in this region.  Destroying the
      cache, on the other hand, is much faster.
  12. 23 Dec, 2019 1 commit
  13. 21 Dec, 2019 1 commit
  14. 20 Dec, 2019 2 commits
  15. 08 Dec, 2019 1 commit
  16. 30 Nov, 2019 4 commits
  17. 21 Nov, 2019 1 commit
  18. 18 Nov, 2019 2 commits
  19. 17 Nov, 2019 1 commit
  20. 11 Nov, 2019 1 commit
    • Ell's avatar
      gegl-parallel: avoid extraneous threads when distributing range/area · 4873b164
      Ell authored
      In gegl_parallel_distribute_{range,area}(), make sure not to use
      more threads than there are pixels to process (or, in the case of
      gegl_parallel_distribute_area(), rows/columns to process), so that
      each thread has at least some work to do.  Since thread cost is now
      calculated dynamically, low thread-cost values could result in too
      many threads being used, some of which would process an empty area.
      This could lead to bugs, on top of being inefficient.
  21. 31 Oct, 2019 1 commit
  22. 28 Oct, 2019 2 commits
  23. 27 Oct, 2019 1 commit
    • Christoph Reiter's avatar
      meson: fix cpu detection · 6bcf95fd
      Christoph Reiter authored
      Use host_machine.cpu_family() instead of cpu(). Only the former
      provides a stable/defined set of values.
      Also don't error out on some arches for no good reason.
      This fixes the build on 32bit Windows.