summaryrefslogtreecommitdiff
path: root/test/README
diff options
context:
space:
mode:
authorJohn Mark Bell <jmb@netsurf-browser.org>2009-08-29 23:38:32 +0000
committerJohn Mark Bell <jmb@netsurf-browser.org>2009-08-29 23:38:32 +0000
commit4a9d61f75baf66c6c6fed350477d2a48ff3814e6 (patch)
treeadfa70ae44667bccee241e49ce4ac237c42715d5 /test/README
parent969324103791a79a0c8202958acd5f8bbf993406 (diff)
downloadlibcss-4a9d61f75baf66c6c6fed350477d2a48ff3814e6.tar.gz
libcss-4a9d61f75baf66c6c6fed350477d2a48ff3814e6.tar.bz2
1) Allocate css_style bytecode in 16 byte chunks
2) Cache unused css_style objects with chunks of size 16, 32, 48, and 64 bytes The above should reduce heap churn somewhat. Further improvements are possible: 1) Make the property parsers write the parsed values direct into the output bytecode, instead of into a temporary object which is then merged into the output. 2) Perform similar caching for css_rule and selector objects. 3) Shrink-wrap finalised output styles rather than leaving them oversized. 4) Perform measurement to determine the optimal chunk sizes (power-of-2 makes maths simple and 16 is plucked from thin air) and cache bucket count/sizes. svn path=/trunk/libcss/; revision=9502
Diffstat (limited to 'test/README')
0 files changed, 0 insertions, 0 deletions