Montreal Pango Summit
drove over from Ottawa yesterday and we spent about 10 hours straight optimizing Pango even more.
Billy have started looking into computing log_attrs lazily
. We didn't know about the open bug though.
I committed my mini-fribidi and opentype patches, then produced a more efficient mirroring table for glib using my tools
. Used the tools to make a better gunichar->script table
for Pango too.
Last night did some profiling and found that calling pango_fc_font_lock_face() and pango_fc_font_unlock_face() from static functions is causing several redundant PANGO_IS_FC_FONT checks, that show up in the profile (around 1%). So ripped them off.
Today I committed more bits: Federico's quark and John Rice's G_MODULE_BIND_LAZY mini patches.
All in all, very productive so far.
To be continued...