This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: ABI / dynamic list now on wiki
- From: Chris Metcalf <cmetcalf at tilera dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>
- Cc: <libc-alpha at sourceware dot org>, Richard Henderson <rth at redhat dot com>, Carlos O'Donell <carlos at redhat dot com>, Andreas Schwab <schwab at linux-m68k dot org>, Andreas Krebbel <Andreas dot Krebbel at de dot ibm dot com>, "David S. Miller" <davem at davemloft dot net>, David Holsgrove <david dot holsgrove at xilinx dot com>
- Date: Thu, 23 Jan 2014 13:51:06 -0500
- Subject: Re: ABI / dynamic list now on wiki
- Authentication-results: sourceware.org; auth=none
- References: <Pine dot LNX dot 4 dot 64 dot 1401230114430 dot 16981 at digraph dot polyomino dot org dot uk> <52E14305 dot 2030705 at tilera dot com> <Pine dot LNX dot 4 dot 64 dot 1401231639040 dot 31884 at digraph dot polyomino dot org dot uk>
On 1/23/2014 11:44 AM, Joseph S. Myers wrote:
> On Thu, 23 Jan 2014, Chris Metcalf wrote:
>
>> On 1/22/2014 8:34 PM, Joseph S. Myers wrote:
>>> https://sourceware.org/glibc/wiki/ABIList
>> There is no actual software stack support for BE tilepro, and as that
>> chip architecture is in maintenance mode, we at Tilera are not planning
>> to do this work. It is possible in principle but would require both
>> kernel and libc work to make it happen. Is it worth calling this out as
>> an actual option, given all those caveats?
> No, I'd suggest removing the BE entries for tilepro
Thanks, done.
> (and probably also
> tilegx - in fact I don't see any BE support for that in GCC either).
There is BE support for tilegx in binutils, gcc, linux, and glibc, and we do some limited regression testing on this configuration every night, so at least I know what we ship customers isn't too broken. I'm pretty confident support should be in the open-source gcc, though I'm not that close to the community return process for gcc.
> Having a bits/endian.h that allows for different endiannesses is confusing
> when you don't actually have support for those endiannesses.
Yes, I could move ports/sysdeps/tile/bits/endian.h down to the tilegx subdirectory, and put an unconditional __LITTLE_ENDIAN version in the tilepro subdirectory, but that does kind of feel like churn, so I'm inclined not to. Or do you think it does make more sense that way?
--
Chris Metcalf, Tilera Corp.
http://www.tilera.com