Am 12.05.2009 21:32, Xueming Shen schrieb:
Updated the webrev with some "final tweak".
http://cr.openjdk.java.net/~sherman/6831794_6229811/webrev
(1)change the MSB to < 0x80 as Ulf suggested, hope it faster, but it
at lease saved one field from the class file:-)
(2)tweak the encoder code a little to encode supplementary character
only after BMP character encoding failed,
as commented in the code, it boots the BMP encoding about 10% on the
price of supplementary character, but
something worth doing.
I think, this was good idea !
(3)tweak the initC2B code a little to make it a little faster (to
avoid the n * segSize)
Also good :-)
Hope this is really a final one:-) the performance data looks good
enough for me, yesterday and today's tweak
seems to be helpful, thanks Ulf!
Could you share the new performance data ?
After we are done with this one, please keep doing the real "ibm
charset rewrite":-)
Tomorrow..... :-)
-Ulf
