[Webkit-unassigned] [Bug 64262] Small speed up, which switches some virtual functions to inline ones.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=64262





--- Comment #30 from Benjamin Poulain <benjamin at webkit.org>  2011-08-01 08:43:33 PST ---
I figured we might as weel get more data so I ran our "cycler" benchmark on my phone :)

Without patch:
        benchmark: cycler_one_qnetworkaccessmanager::load       http://www.facebook.com/
                mean: 309 msecs +/- 32 msecs, +/- 10.356 %
                avg:  318 msecs
                        407 311 309 310 308 309 305 311 307 309 
        benchmark: cycler_one_qnetworkaccessmanager::load       http://www.twitter.com/
                mean: 514 msecs +/- 6 msecs, +/- 1.16732 %
                avg:  512 msecs
                        517 514 507 504 509 506 514 511 519 521 
        benchmark: cycler_one_qnetworkaccessmanager::load       http://www.youtube.com/
                mean: 1930 msecs +/- 3 msecs, +/- 0.15544 %
                avg:  1930 msecs
                        1924 1931 1929 1928 1929 1927 1934 1937 1930 1931 
        benchmark: cycler_one_qnetworkaccessmanager::load       https://twitter.com/?lang=en&logged_out=1#!/search/webkit
                mean: 30 msecs +/- 11 msecs, +/- 36.6667 %
                avg:  23 msecs
                        14 31 15 33 16 30 15 31 15 32 


With patch: 
        benchmark: cycler_one_qnetworkaccessmanager::load       http://www.facebook.com/
                mean: 310 msecs +/- 31 msecs, +/- 10 %
                avg:  321 msecs
                        401 309 305 310 307 309 305 320 329 316 
        benchmark: cycler_one_qnetworkaccessmanager::load       http://www.twitter.com/
                mean: 511 msecs +/- 4 msecs, +/- 0.782779 %
                avg:  510 msecs
                        514 511 504 505 506 509 513 511 515 513 
        benchmark: cycler_one_qnetworkaccessmanager::load       http://www.youtube.com/
                mean: 1899 msecs +/- 4 msecs, +/- 0.210637 %
                avg:  1898 msecs
                        1889 1899 1900 1901 1896 1898 1898 1903 1895 1905 
        benchmark: cycler_one_qnetworkaccessmanager::load       https://twitter.com/?lang=en&logged_out=1#!/search/webkit
                mean: 31 msecs +/- 11 msecs, +/- 35.4839 %
                avg:  23 msecs
                        15 31 15 31 15 31 16 31 15 31 

So basically no difference.
I really doubt this patch improves loading performance. Maybe you should try to write a test page that keep changing the CSS values of the Elements to demonstrate how much time can be taken by CSSValue?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.