Comment #2 on issue 1101 by [email protected]: NaN is converted to 0xfff8000000000000 (-NaN)
http://code.google.com/p/v8/issues/detail?id=1101
This is a consequence of the fact that the OpenGL buffers expose the representation of floating point numbers, no? As I understand it (I haven't really followed this) in the version that is headed for standardization in ES Harmony the representations won't be exposed.
There are lots of different representations of NaN in the IEEE standard. V8 uses two of them and JSC apparently only uses one. There's no such thing as -NaN. Both the representations in the bug report are valid NaN representations.
-- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
