[Bug libc/11422] sin(x) (actually probably all trig) is inaccurate for large x

["simon dot fenney at imgtec dot com" <sourceware-bugzilla at sourceware dot org>]

------- Additional Comments From simon dot fenney at imgtec dot com  2010-03-25 11:48 -------
Dear Ulrich, 
For my own sanity, I created a vintage SuSE 8 system and now see that sin was
indeed broken back then. I apologise for the assertion that it had been working
on older systems.

However, I don't believe x86-64 is an option as the code has to run on Atom
platforms. 

The thing that is effectively broken is the "range reduction" (i.e. mapping X to
the equivalent value between 0 and Pi/2 (or perhaps Pi/4) which is, in effect, a
(multi-word precision) integer operation. Can't the 32-bit builds use the same
code base that the x86-64 systems use (which presumably avoid using the dubious
fsin instruction on non trivial values)?

Regards
Simon

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11422

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.