Comment #4 on issue 1750 by [email protected]: Improve performance of
mesh decompression benchmark
http://code.google.com/p/v8/issues/detail?id=1750
Non-ASCII should not be a problem.
I did some low-level profiling and I'm pretty sure that the performance
problem arises from the fact that the responseText string is not a "native"
V8 string, but rather a Webkit external string and we don't have inline
code to do charCodeAt from external strings. This means we go from JS
generated code to C++ for each call of charCodeAt.
This is definitely not ideal and should be fixed, but at first sight it is
not a trivial change because we don't have a direct pointer to the external
string data available in V8 land.
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
