The ia32 version has a case for integer32, that claims to also handle smis.
I'm
not sure that's correct, though.
http://codereview.chromium.org/2834028/diff/1/3
File src/x64/codegen-x64.cc (right):
http://codereview.chromium.org/2834028/diff/1/3#newcode5349
src/x64/codegen-x64.cc:5349: dest->false_target()->Branch(equal);
I doubt it will happen in practice that we convert a value to boolean
that we *know* is a double.
I don't know whether we would want to do something for the known string
case. Again, I don't expect known string operation results to be used as
booleans much.
Maybe we should add a counter for the other special cases and see if
they are hit.
http://codereview.chromium.org/2834028/show
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
