This is the mail archive of the
libc-ports@sources.redhat.com
mailing list for the libc-ports project.
Re: Conditionally falling back to the .c code or a different .S
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: David Gilbert <david dot gilbert at linaro dot org>
- Cc: libc-ports at sourceware dot org
- Date: Thu, 3 Mar 2011 23:46:14 +0000 (UTC)
- Subject: Re: Conditionally falling back to the .c code or a different .S
- References: <AANLkTin0ZXuNmp3Tv3PbsMvK7XgBH1hTaL6d=dYAB+VJ@mail.gmail.com>
My inclination is to use the usual Implies approach (so that an armv8
directory implies the armv7 directory). I think the directory should best
be armv7-a as that's the value that will work with -march= for a GNU/Linux
target, and the only v7 variant supported for such a target (directories
for older versions - if some functions only need v6, say - could end up
being more complicated). That way it can work with --with-cpu as well as
any autodetection you implement.
Please make sure any functions you add have been tested (with the glibc
testsuite) for both big and little endian, as it's quite easy for string
functions to have endian-dependent bugs.
--
Joseph S. Myers
joseph@codesourcery.com