This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Fix ldbl-128ibm atan2l for x near 1
On 10/31/2012 07:10 PM, Joseph S. Myers wrote:
Bug 14610 is the ldbl-128ibm implementation of atan2l treating atan2l
(y, x) as atanl (y) if the high double of x is 1, without checking the
low double. This causes large ulps for various clog and clog10 tests.
This patch fixes the bug in the obvious way and adds an atan2 test for
it.
Tested x86_64 and x86 (to make sure no new ulps are needed there for
the new test), and powerpc (to confirm the bug is fixed; ulps for clog
and clog10 now look reasonable and there are no new ulps for atan2).
Thanks, this is fine.
It looks like Andreas's ulps patch
<http://sourceware.org/ml/libc-alpha/2012-10/msg00324.html>, or a
version updated after this patch, still needs to be committed (and
then there are still some other issues he noted that should be
resolved to get clean test results).
Let's get the patch in, ULPs updated are in general fine - I'll answer,
Andreas
--
Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126