This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PowerPC64] Correct IBM long double frexpl
- From: Alan Modra <amodra at gmail dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Tue, 25 Mar 2014 23:54:13 +1030
- Subject: Re: [PowerPC64] Correct IBM long double frexpl
- Authentication-results: sourceware.org; auth=none
- References: <20140325023118 dot GG18201 at bubble dot grove dot modra dot org> <Pine dot LNX dot 4 dot 64 dot 1403251214090 dot 12472 at digraph dot polyomino dot org dot uk>
On Tue, Mar 25, 2014 at 12:16:51PM +0000, Joseph S. Myers wrote:
> On Tue, 25 Mar 2014, Alan Modra wrote:
>
> > Besides fixing the bugzilla, this also fixes corner-cases where the high
> > and low double differ greatly in magnitude, and handles a denormal
> > input without resorting to a fp rescale.
>
> Cases where the high and low double differ greatly in magnitude are bug
> 16619 (such test inputs can't be represented in libm-test.inc, but you
> should see improvements in test results for clog10l; it was large errors
> from clog10l that drew my attention to problems with such cases).
Yes, I forgot to mention the test improvement with the patch
submission. 16 clog10l tests now no longer fail, and 8 tests now have
only 1 ulp error.
--
Alan Modra
Australia Development Lab, IBM