Comment #9 on issue 2097 by [email protected]: sha1 compiled with Emscripten runs 10X slower than with SpiderMonkey
http://code.google.com/p/v8/issues/detail?id=2097

Deoptimization with uint32 values is a long outstanding performance bug in V8:

How about the following simple fix: Treat the result of uint32 operations load from an Uint32Array as a double number to avoid deoptimization of half of the numbers.

As an optimization like vegorov@ suggested in comment #5, check if the result is only used in bit-ops (in which case it can be treated as signed int32)

--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev

Reply via email to