This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Fix dbl-64 hypot spurious underflows (bug 16314)
- From: Andreas Jaeger <aj at suse dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>, libc-alpha at sourceware dot org
- Date: Tue, 17 Dec 2013 10:17:49 +0100
- Subject: Re: Fix dbl-64 hypot spurious underflows (bug 16314)
- Authentication-results: sourceware.org; auth=none
- References: <Pine dot LNX dot 4 dot 64 dot 1312162224490 dot 29265 at digraph dot polyomino dot org dot uk>
On 12/16/2013 11:26 PM, Joseph S. Myers wrote:
> This patch, relative to a tree with
> <https://sourceware.org/ml/libc-alpha/2013-12/msg00597.html> applied,
> fixes bug 16314, spurious underflows from dbl-64 hypot for input
> values a bit above 0x1p-500. In these cases, squaring the inputs does
> not result in underflow, but squaring values constructed as low parts
> of inputs does result in underflow; the fix is simply to adjust the
> threshold for scaling up inputs from 0x1p-500 to 0x1p-450 (the same as
> used for ldbl-128ibm). I don't think any of the other hypot
> implementations have this particular problem.
ok,
Thanks,
Andreas
--
Andreas Jaeger aj@{suse.com,opensuse.org} Twitter: 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