Hi, I wanted to get some feedback on a fontconfig patch I've been working on to speed up FcFontMatch(). (I noticed it taking a good chunk of time during Mozilla page layout). The idea is to optimize the case where we have one or more fonts that are an exact match for the family given in the pattern. All of the patterns in a fontset are put into a hash table that is keyed on the family. Of course, since there will be more than one pattern for each family (slant/weight variations), the hash entries are linked lists of patterns. Assuming that the pattern passed in to FcFontMatch doesn't specify any higher-priority values than family, the hash table lookup can quickly reduce the number of patterns you need to compare against, usually to just a handful. If there's no exact match, it falls back to the existing matching code. The biggest question for me is whether this will always produce the same results as the slower matching code. I think that it will, but any suggestions would be great, particularly regarding testing. Thanks. -- -Brian Ryner bryner@xxxxxxxxxxxxxx -------------- next part -------------- A non-text attachment was scrubbed... Name: fontconfig-20040822.diff Type: text/x-patch Size: 11000 bytes Desc: not available Url : http://freedesktop.org/pipermail/fontconfig/attachments/20040823/55d0b35b/fontconfig-20040822.bin
