This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: AArch64 patches for glibc master.
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Carlos O'Donell <carlos at redhat dot com>
- Cc: Marcus Shawcroft <marcus dot shawcroft at linaro dot org>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Mon, 9 Dec 2013 23:25:53 +0000
- Subject: Re: AArch64 patches for glibc master.
- Authentication-results: sourceware.org; auth=none
- References: <52A60A93 dot 1050402 at redhat dot com> <Pine dot LNX dot 4 dot 64 dot 1312091852010 dot 17984 at digraph dot polyomino dot org dot uk> <52A61590 dot 9060303 at redhat dot com> <Pine dot LNX dot 4 dot 64 dot 1312092136450 dot 22134 at digraph dot polyomino dot org dot uk> <52A64BF5 dot 1070404 at redhat dot com>
On Mon, 9 Dec 2013, Carlos O'Donell wrote:
> Is this requirement because IEC 60559 requires everything to be correctly
> rounded, therefore if C11 defines sqrt in terms of the IEC 60559 operation
> then the function itself must be correctly rounding as required by the
> standard (IEEE Std 754-2008)?
Yes.
I believe TS 18661-4, providing C bindings to additional functions from
IEEE 754-2008, will distinguish between the (optional) correctly-rounding
functions bound to the "Recommended correctly rounded functions" in IEEE
754-2008 Table 9.1, and functions providing approximations that need not
be correctly rounded (so, for example, exp will remain as a function
providing an approximation, crexp will be reserved for a correctly
rounding version; rsqrt will be added for an approximation to reciprocal
square root and crrsqrt will be reserved for a correctly rounding
version). (No public draft of parts 4 or 5 of this TS is yet available,
but there's an old summary of plans at
<http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1582.pdf>.)
--
Joseph S. Myers
joseph@codesourcery.com