This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Minimum floating-point requirements
- From: James Cloos <cloos at jhcloos dot com>
- To: libc-alpha at sourceware dot org
- Cc: Rich Felker <dalias at aerifal dot cx>, "Joseph S. Myers" <joseph at codesourcery dot com>, Carlos O'Donell <carlos at redhat dot com>
- Date: Thu, 30 Jan 2014 15:56:38 -0500
- Subject: Re: Minimum floating-point requirements
- Authentication-results: sourceware.org; auth=none
- Copyright: Copyright 2013 James Cloos
- Openpgp: ED7DAEA6; url=http://jhcloos.com/public_key/0xED7DAEA6.asc
- Openpgp-fingerprint: E9E9 F828 61A4 6EA9 0F2B 63E7 997A 9F17 ED7D AEA6
- References: <Pine dot LNX dot 4 dot 64 dot 1401290320320 dot 2449 at digraph dot polyomino dot org dot uk> <52E9D943 dot 7010908 at redhat dot com> <Pine dot LNX dot 4 dot 64 dot 1401301425160 dot 3467 at digraph dot polyomino dot org dot uk> <20140130170012 dot GV24286 at brightrain dot aerifal dot cx>
>>>>> "RF" == Rich Felker <dalias@aerifal.cx> writes:
RF> it's there to allow software to get access to higher precision
RF> supported by the hardware, not for software emulation of
RF> higher-precision types, and especially not poor-quality ones like
RF> IBM long double that break basic floating point semantics.
OTOH, on platforms where there is hardware support (or assist) for
double-double, such as on IBM's hardware, there often is a need to
support it for interaction with other languages and applications
which expect double-double on said platforms.
For easier debugging, it might be useful to have a generic double-
double test suite, which works w/o hardware assist (probably using
libqd, as used by gcc) so that one can develop and test the algos
without full-time access to such hardware.
Such a double-double development and testing suite need not be part of
glibc, but should have a license which ensures that code developed there
can be incorporated as needed in to glibc.
-JimC
--
James Cloos <cloos@jhcloos.com> OpenPGP: 1024D/ED7DAEA6