Deprecating IA64 and selective scheduling

[Frank Scheiner]

On 10.12.25 14:28, Richard Biener wrote:
> On Wed, Dec 10, 2025 at 3:07 AM Andrew Pinski via Gcc <gcc@gcc.gnu.org> wrote:
>> Why can't you keep the port maintained out of tree like you do already
>> for glibc and the kernel?
>> I am not seeing why it needs to be upstream because right now it is
>> blocking cleanups and there is no way for anyone outside of your group
>> to be able to test it.
> 
> I suppose that the target blocks cleanups means maintaining it out-of-tree
> is not easily possible.

Also I think the GCC actually benefits from the ia64 port. A good
example are bugs 111425 ([1]) and 116627 ([2]):

I filed [1] in September 2023, but the problem was actually detected
even earlier than that by Debian kernel builds for ia64 (sadly no
longer accessible). The same issue ([2]) was then detected another time
for armv6l, but only about a year later. It was fixed in a month's
time. Actually Andrew noticed the relation first.

[1]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425

[2]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116627

> IMO as there is interest in having the Itanium port in-tree and work is done
> when required (like enabling LRA) I see no reason for deprecating or removing
> the port.  There are other unmaintained ports and frontends and IMO this
> looks like a double-standard to me.
> 
> I do agree that posting testresults from trunk semi-regularly is an important
> sign of activity in addition to the bootstrap work folks are doing (but not
> so much visible).

Yeah, that visibility needs to be improved. I fully agree.

Cheers,
Frank

> Thanks,
> Richard.