This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Update powerpc libm ULPs
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Andreas Schwab <schwab at linux-m68k dot org>
- Cc: <libc-alpha at sourceware dot org>, Ryan Arnold <rsa at us dot ibm dot com>
- Date: Thu, 11 Oct 2012 12:53:18 +0000
- Subject: Re: [PATCH] Update powerpc libm ULPs
- References: <m2obk9ky66.fsf@igel.home>
On Thu, 11 Oct 2012, Andreas Schwab wrote:
> Failure: Real part of: clog10 (0x1.00000000000000123456789abcp0 + 0x1.23456789p-1000 i) == 4.285899851347756186652871946325962330640e-19 + 4.611541215247321502041995872887317363241e-302 i: Exception "Underflow" set
> Failure: Test: Imaginary part of: clog10 (0x1.00000000000000123456789abcp0 + 0x1.23456789p-1000 i) == 4.285899851347756186652871946325962330640e-19 + 4.611541215247321502041995872887317363241e-302 i
> Result:
> is: 4.61154121524732158678e-302 0x1.f9fd5fef6543bd860600p-1002
> should be: 4.61154121524732158678e-302 0x1.f9fd5fef6543bd620b00p-1002
> difference: 4.55083866384372191841e-320 0x0.00000000023fb0000000p-1022
> ulp : 9211.0000
> max.ulp : 0.0000
The clog/clog10 failures for long double are bug 14610 (a bug in atan2l).
I don't know about the others, but ldbl-128ibm fma is pretty broken anyway
and it's probably reasonable to disable affected fma tests for it, if
there aren't easy fixes, until we have Bruno Haible's gnulib
implementation (bug 13304) in glibc to serve as a generic fallback (both
for ldbl-128ibm, and for architectures without exceptions / rounding
modes).
--
Joseph S. Myers
joseph@codesourcery.com