[Bug target/109929] New: profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[Bug target/109929] New: profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[xry111 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

            Bug ID: 109929
           Summary: profiledbootstrap failure on aarch64-linux-gnu with
                    graphite optimization
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: xry111 at gcc dot gnu.org
  Target Milestone: ---

Configure GCC trunk with:

../gcc/configure --prefix=$HOME/gcc-trunk --enable-languages=c,c++
--with-system-zlib --disable-fixincludes --disable-werror

Note that --disable-werror is needed to work around PR100416.

And make it with:

make STAGE1_CFLAGS="-O2 -pipe" BOOT_CFLAGS="-O2 -fgraphite-identity -pipe"
profiledbootstrap

Then some error messages shows up during stagetrain, like:

* genhooks: No place specified to document hook TARGET_ASM_OPEN_PAREN
* build/genmodes: ../../gcc/gcc/machmode.def:222: no mode "TI"

etc.

I guess stageprofile compiler is miscompiled, and it miscompiled genhooks and
genmodes then.  Not sure how to debug the issue.

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I think nobody paid attention to graphite the last years - I suggest to
try simpler things like non-profiled bootstrap or just running the
testsuite with -fgraphite-identity added.

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[xry111 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

--- Comment #2 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #1)
> I think nobody paid attention to graphite the last years - I suggest to
> try simpler things like non-profiled bootstrap or just running the
> testsuite with -fgraphite-identity added.

A non-profiled bootstrap works.

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[xry111 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

--- Comment #3 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
Hmm, it seems no longer happening with current trunk.

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
let's close this then

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[xry111 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

--- Comment #5 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
The first good commit is:

commit aa2ad77a9b3994fb679e5295d9570f6f8696540d
Author: Szabolcs Nagy <szabolcs.nagy@arm.com>
Date:   Tue May 9 11:07:05 2023 +0100

    aarch64: Do not force a stack frame for EH returns

but I cannot see how it's related to the issue...

And with the parent of this commit, the bootstrap fails in stagefeedback where
build/genhooks -d ../../gcc/gcc/doc/tm.texi.in > tmp-tm.texi segfaults.  I'll
do another round of bisection to see when the stagetrain error was "fixed". 
(Just for better understanding the issue.)

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[xry111 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

--- Comment #6 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
The first commit deferring the failure to stagefeedback is:

commit 575858508090b18dcbc176db285c9f55227ca4c0
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Tue Oct 17 23:46:33 2023 +0100

    aarch64: Use vecs to store register save order

[Bug target/109929] profiledbootstrap failure on aarch64-linux-gnu with graphite optimization

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109929

--- Comment #7 from Richard Sandiford <rsandifo at gcc dot gnu.org> ---
Hmm, yeah, like you say, neither of those commits should have made a different
to whether bootstrap works.  I guess the problem is just latent now.