This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: RTL sharing checker update
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Jan Hubicka <jh at suse dot cz>
- Cc: gcc-patches at gcc dot gnu dot org, mark at codesourcery dot com
- Date: Thu, 5 Jul 2007 17:13:39 +0200
- Subject: Re: RTL sharing checker update
- References: <20070704120213.GA16492@kam.mff.cuni.cz>
> Mark,
> I re-run the testing on mainline with two additional fixes applied:
>
> http://gcc.gnu.org/ml/gcc-patches/2006-12/msg01441.html
> http://gcc.gnu.org/ml/gcc-patches/2007-06/msg02097.html
>
> The first is sort of hack and I don't think it is possible to fix the
> problem nicely without substantial rewrite of the pass that is desirable
> (using new dataflow) but I don't have time for it at the very moment.
> Since regrename is late pass, one option would be to first relax
> checking, but Kenny already mentioned me, that regrename is causing real
> ICEs even without the checker, so perhaps the two extra copy_rtxes are
> not bad way around.
>
> The other by Richard Standifold actually fix pretty important problem of
> new dataflow implementation.
>
> I tested on i686-linux, x86_64-linux, ia64-linux, ppc-linux and
> ppc64-linux. PPC64 linux bootstrap is currently broken, other
I've tested at least fortran-disabled PPC bootstrap. There are two
failures in testsuite:
gcc.c-torture/unsorted/buns.c, -O1 (internal compiler error)
gcc.c-torture/unsorted/buns.c, -O2 (internal compiler error)
gcc.c-torture/unsorted/buns.c, -O3 -fomit-frame-pointer (internal compiler +error)
gcc.c-torture/unsorted/buns.c, -O3 -g (internal compiler error)
gcc.c-torture/unsorted/buns.c, -Os (internal compiler error)
gcc.c-torture/unsorted/dead.c, -O1 (internal compiler error)
gcc.c-torture/unsorted/dead.c, -O2 (internal compiler error)
gcc.c-torture/unsorted/dead.c, -O3 -fomit-frame-pointer (internal compiler +error)
gcc.c-torture/unsorted/dead.c, -O3 -g (internal compiler error)
gcc.c-torture/unsorted/dead.c, -Os (internal compiler error)
I am going to test patch for this another side case.
I would expect that after enabling the testing by default we will notice
more such cases, this is also why I would like to have it enabled sooner
rather than later especially because I am leaving at 13th for 5 days
trip before the summit and would like to have some time before it to
have chance to fix the fallout on less common targets.
Honza