Comment #33 on issue 3006 by [email protected]: Inaccurate sin/cos values
http://code.google.com/p/v8/issues/detail?id=3006
Sorry for the long delay, but I've been out sick for a while.
I now have a reference implementation for the sin, cos, and tan routines,
including accurate argument reduction for args up to about 2^19*pi/2.
The routines are at
https://x20web.corp.google.com/~rtoy/V8/kernel-trig.js. A test suite is
available at https://x20web.corp.google.com/~rtoy/V8/trig-tests.html. I
think there should be more tests, but I think all of the code paths are
covered by the existing tests.
The implementation is a very straightforward translation of fdlibm,
including really bad equivalents of __HI/__LO, but they're functionally
correct.
One thing is missing in the current implementation. There's a bit of
processing needed before calling k_rem_pio2 for the expensive Payne-Hanek
implementation. I haven't done the preprocessing part.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.
