canmove, Confirmed users
1,448
edits
L20n/Firefox/Performance: Difference between revisions
Jump to navigation
Jump to search
m
→Hypothesis 6: DOM translation is slow
(→Hypothesis 6: DOM translation is slow: microoptimizations result in microwins) |
|||
| Line 103: | Line 103: | ||
** for-of loops in hot paths might be expensive, | ** for-of loops in hot paths might be expensive, | ||
** detect null values in the parser. | ** detect null values in the parser. | ||
** I tested all three together and I was able to get down from 20ms to 16ms on my machine. try also shows [https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=47cb962768ac&newProject=try&newRevision=2133bc69c4c5&framework=1&showOnlyImportant=0 small perf wins] but they're too small to reason about reliably | ** I tested all three together and I was able to get down from 20ms to 16ms on my machine. try also shows [https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=47cb962768ac&newProject=try&newRevision=2133bc69c4c5&framework=1&showOnlyImportant=0 small perf wins] but they're too small to reason about reliably (-8ms on tpaint looks pretty accurate though assuming try has slower machinces) | ||
* if node iteration is fast (H2) maybe we could make one qSA for each known Localization rather than make just one and then group nodes by l10n? | * if node iteration is fast (H2) maybe we could make one qSA for each known Localization rather than make just one and then group nodes by l10n? | ||