Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Andrew Pinski]

-fgnu-tm support has not been improved since GCC 5 or earlier. It is
not even supported with LTO. Does it make sense to deprecate the
support for GCC 14 and remove it in GCC 15?

Thanks,
Andrew Pinski

Re: Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Florian Weimer]

* Andrew Pinski via Gcc:

> -fgnu-tm support has not been improved since GCC 5 or earlier. It is
> not even supported with LTO. Does it make sense to deprecate the
> support for GCC 14 and remove it in GCC 15?

Is this the stuff around libitm and that adds _ITM_registerTMCloneTable
and _ITM_deregisterTMCloneTable symbol references to *all* binaries
(whether they use transactional memory or not)?

Thanks,
Florian

Re: Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Andrew Pinski]

On Sun, Dec 17, 2023 at 8:26 AM Florian Weimer <fweimer@redhat.com> wrote:
>
> * Andrew Pinski via Gcc:
>
> > -fgnu-tm support has not been improved since GCC 5 or earlier. It is
> > not even supported with LTO. Does it make sense to deprecate the
> > support for GCC 14 and remove it in GCC 15?
>
> Is this the stuff around libitm and that adds _ITM_registerTMCloneTable
> and _ITM_deregisterTMCloneTable symbol references to *all* binaries
> (whether they use transactional memory or not)?

Yes. and the front-end support for it.

Thanks,
Andrew

>
> Thanks,
> Florian
>

Re: Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Eric Gallager]

On Sat, Dec 16, 2023 at 3:16 PM Andrew Pinski via Gcc <gcc@gcc.gnu.org> wrote:
>
> -fgnu-tm support has not been improved since GCC 5 or earlier. It is
> not even supported with LTO. Does it make sense to deprecate the
> support for GCC 14 and remove it in GCC 15?
>
> Thanks,
> Andrew Pinski

Personally, since GCC is in stage 3 now, I would push that schedule
back a release and move deprecation to GCC 15, and then only remove it
for GCC 16 if no one objects, but then again I don't actually use
-fgnu-tm myself, so I wouldn't be too upset if the faster schedule is
chosen instead.
Eric Gallager

Re: Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Andrew Pinski]

On Sun, Dec 17, 2023 at 1:20 PM Eric Gallager <egall@gwmail.gwu.edu> wrote:
>
> On Sat, Dec 16, 2023 at 3:16 PM Andrew Pinski via Gcc <gcc@gcc.gnu.org> wrote:
> >
> > -fgnu-tm support has not been improved since GCC 5 or earlier. It is
> > not even supported with LTO. Does it make sense to deprecate the
> > support for GCC 14 and remove it in GCC 15?
> >
> > Thanks,
> > Andrew Pinski
>
> Personally, since GCC is in stage 3 now, I would push that schedule
> back a release and move deprecation to GCC 15, and then only remove it
> for GCC 16 if no one objects, but then again I don't actually use
> -fgnu-tm myself, so I wouldn't be too upset if the faster schedule is
> chosen instead.

Considering -fgnu-tm has been broken for LTO ever since LTO was
introduced, and broken with -fsanitize=undefined and broken with many
code that might use internal functions (known since 2015), I suspect
nobody is using this option in production nor even trying it out. If
this was stage1, I might even just recommend removing the support. But
deprecating it during stage 3 seems like a fair compromise.

> Eric Gallager

Re: Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Richard Biener]

On Mon, Dec 18, 2023 at 2:35 AM Andrew Pinski via Gcc <gcc@gcc.gnu.org> wrote:
>
> On Sun, Dec 17, 2023 at 1:20 PM Eric Gallager <egall@gwmail.gwu.edu> wrote:
> >
> > On Sat, Dec 16, 2023 at 3:16 PM Andrew Pinski via Gcc <gcc@gcc.gnu.org> wrote:
> > >
> > > -fgnu-tm support has not been improved since GCC 5 or earlier. It is
> > > not even supported with LTO. Does it make sense to deprecate the
> > > support for GCC 14 and remove it in GCC 15?
> > >
> > > Thanks,
> > > Andrew Pinski
> >
> > Personally, since GCC is in stage 3 now, I would push that schedule
> > back a release and move deprecation to GCC 15, and then only remove it
> > for GCC 16 if no one objects, but then again I don't actually use
> > -fgnu-tm myself, so I wouldn't be too upset if the faster schedule is
> > chosen instead.
>
> Considering -fgnu-tm has been broken for LTO ever since LTO was
> introduced, and broken with -fsanitize=undefined and broken with many
> code that might use internal functions (known since 2015), I suspect
> nobody is using this option in production nor even trying it out. If
> this was stage1, I might even just recommend removing the support. But
> deprecating it during stage 3 seems like a fair compromise.

Btw, I'm OK with deprecating it for GCC 14.  Can you please propose a
patch for changes.html and add a diagnostic message when -fgnu-tm is used
(disabled with -Wno-deprecated)?

Thanks,
Richard.

> > Eric Gallager

Re: Deprecating -fgnu-tm support for GCC 14 and removing it for GCC 15

[Jason Merrill]

[-- Attachment #1: Type: text/plain, Size: 1986 bytes --]

On Mon, Dec 18, 2023 at 3:04 AM Richard Biener via Gcc <gcc@gcc.gnu.org>
wrote:

> On Mon, Dec 18, 2023 at 2:35 AM Andrew Pinski via Gcc <gcc@gcc.gnu.org>
> wrote:
> >
> > On Sun, Dec 17, 2023 at 1:20 PM Eric Gallager <egall@gwmail.gwu.edu>
> wrote:
> > >
> > > On Sat, Dec 16, 2023 at 3:16 PM Andrew Pinski via Gcc <gcc@gcc.gnu.org>
> wrote:
> > > >
> > > > -fgnu-tm support has not been improved since GCC 5 or earlier. It is
> > > > not even supported with LTO. Does it make sense to deprecate the
> > > > support for GCC 14 and remove it in GCC 15?
> > > >
> > > > Thanks,
> > > > Andrew Pinski
> > >
> > > Personally, since GCC is in stage 3 now, I would push that schedule
> > > back a release and move deprecation to GCC 15, and then only remove it
> > > for GCC 16 if no one objects, but then again I don't actually use
> > > -fgnu-tm myself, so I wouldn't be too upset if the faster schedule is
> > > chosen instead.
> >
> > Considering -fgnu-tm has been broken for LTO ever since LTO was
> > introduced, and broken with -fsanitize=undefined and broken with many
> > code that might use internal functions (known since 2015), I suspect
> > nobody is using this option in production nor even trying it out. If
> > this was stage1, I might even just recommend removing the support. But
> > deprecating it during stage 3 seems like a fair compromise.
>
> Btw, I'm OK with deprecating it for GCC 14.  Can you please propose a
> patch for changes.html and add a diagnostic message when -fgnu-tm is used
> (disabled with -Wno-deprecated)?
>

Deprecation makes sense to me.

But keep in mind that transactional memory is still the subject of research
and standardization efforts, though the current proposal (wg21.link/n4923)
is significantly simpler than the earlier TS that GCC implemented.  I don't
know how much of the current implementation would carry over, but I'd be
cautious about tearing everything out just yet.

Jason