Updates:
Owner: [email protected]
Comment #11 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
Stealing this... I've uploaded a tentative implementation of an approach I
have outlined in comment number #5.
https://chromiumcodereview.appspot.com/10778029/
I did not go as far as propagating uint32 uses through an additions, but my
approach still handles them (though in a less efficient manner: addition
just gets double type feedback and I support conversion of uint32 to
double).
With this change performance of sha1.c.js is greatly improved:
vegorov@magus ~/src/v8 % out/ia32.release/d8
sha1.c.js [uint32@20255]
SHA1=15dd99a1991e0b3826fede3deffc1feba42278e6
SHA1 computation time in milliseconds: 163
vegorov@magus ~/src/v8 % out/ia32.release/d8 --noopt-safe-uint32-operations
sha1.c.js [uint32@20255]
SHA1=15dd99a1991e0b3826fede3deffc1feba42278e6
SHA1 computation time in milliseconds: 2428
This looks like a factor of 14 to me.
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
