Old Synopsis: 3.4: optimization problem with ix86 fast prologue New Synopsis: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC State-Changed-From-To: open->analyzed State-Changed-By: steven State-Changed-When: Wed Mar 26 08:22:58 2003 State-Changed-Why: Confirmed # gcc-3.3 10213.cc -O2 -fPIC # a.out 8048470 8048470 # gcc-3.4 10213.cc -O2 -fPIC # a.out 8048480 8048421 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10213
The following reply was made to PR target/10213; it has been noted by GNATS. From: Dara Hazeghi <dhazeghi@yahoo.com> To: jh@suse.cz, gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, snyder@fnal.gov Cc: Subject: Re: target/10213: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC Date: Wed, 14 May 2003 17:29:42 -0700 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit- trail&database=gcc&pr=10213 Hello, with gcc 3.2, 3.3 branch and mainline cvs (20030511), I get the same output on the first and second lines with this testcase (when compiled with -O2 -fPIC). ie 3.4 gives: 402341ec 402341ec It seems like the bug has been fixed. Is that true? Thanks, Dara
The following reply was made to PR target/10213; it has been noted by GNATS. From: snyder <snyder@fnal.gov> To: Dara Hazeghi <dhazeghi@yahoo.com> Cc: jh@suse.cz, gcc-gnats@gcc.gnu.org Subject: Re: target/10213: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC Date: 14 May 2003 20:00:00 -0500 >It seems like the bug has been fixed. Is that true? Thanks, Yes, it was fixed on the mainline, i think by turning off the fast prologues. (I had thought that the PR was closed too; maybe not..) sss
Synopsis: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC State-Changed-From-To: analyzed->feedback State-Changed-By: bangerth State-Changed-When: Thu May 15 04:41:54 2003 State-Changed-Why: So, Scott, what's your verdict: should it be closed? What's the state on the 3.3 branch? Other question: if it's "fixed" by just disabling something (is this what you mean by "switching off") -- will we see the problem happening again if this code patch is switched on again by someone? Or was the code that produced the wrong code removed altogether? If this isn't the case, then I'd say we should keep the report, but lacking technical insight I leave this up to you. Thanks W. http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10213
The following reply was made to PR target/10213; it has been noted by GNATS. From: Jan Hubicka <jh@suse.cz> To: bangerth@dealii.org, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org, jh@suse.cz, nobody@gcc.gnu.org, snyder@fnal.gov, gcc-gnats@gcc.gnu.org Cc: Subject: Re: target/10213: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC Date: Sun, 18 May 2003 20:40:46 +0200 > Synopsis: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC > > State-Changed-From-To: analyzed->feedback > State-Changed-By: bangerth > State-Changed-When: Thu May 15 04:41:54 2003 > State-Changed-Why: > So, Scott, what's your verdict: should it be closed? What's > the state on the 3.3 branch? > > Other question: if it's "fixed" by just disabling something > (is this what you mean by "switching off") -- will we see > the problem happening again if this code patch is switched > on again by someone? Or was the code that produced the > wrong code removed altogether? If this isn't the case, then > I'd say we should keep the report, but lacking technical > insight I leave this up to you. Can you point me to the patch? I was off for a while and missed the thread. In case the fast prologues was disabled, I would like to fix the real problem instead. Honza
The following reply was made to PR target/10213; it has been noted by GNATS. From: snyder <snyder@fnal.gov> To: Jan Hubicka <jh@suse.cz> Cc: bangerth@dealii.org, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org, jh@suse.cz, nobody@gcc.gnu.org, gcc-gnats@gcc.gnu.org Subject: Re: target/10213: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC Date: 18 May 2003 23:32:34 -0500 hi - >Jan Hubicka writes: >Can you point me to the patch? I was off for a while and missed the >thread. In case the fast prologues was disabled, I would like to fix >the real problem instead. I think this is the patch that i was remembering: Thu Apr 3 00:18:49 CEST 2003 Jan Hubicka <jh@suse.cz> * i386.c (override_options): Disable red zone by default on i386. (compute_frame_layout, ix86_force_to_memory, ix86_free_from_memory): Do not test TARGET_64BIT together with TARGET_RED_ZONE I admit i never looked at the actual code, though, to be sure of what it was doing. When i saw that appear in the changelog, i thought that it was a reaction to the bug that i had reported... sss
Synopsis: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC State-Changed-From-To: feedback->closed State-Changed-By: bangerth State-Changed-When: Mon May 19 13:54:27 2003 State-Changed-Why: Jan says so. http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10213
The following reply was made to PR target/10213; it has been noted by GNATS. From: Jan Hubicka <jh@suse.cz> To: snyder <snyder@fnal.gov> Cc: Jan Hubicka <jh@suse.cz>, bangerth@dealii.org, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org, nobody@gcc.gnu.org, gcc-gnats@gcc.gnu.org Subject: Re: target/10213: [3.4 regression] optimization problem with ix86 fast prologue with -O2 -fPIC Date: Mon, 19 May 2003 15:51:43 +0200 > > hi - > > >Jan Hubicka writes: > >Can you point me to the patch? I was off for a while and missed the > >thread. In case the fast prologues was disabled, I would like to fix > >the real problem instead. > > I think this is the patch that i was remembering: > > Thu Apr 3 00:18:49 CEST 2003 Jan Hubicka <jh@suse.cz> > > * i386.c (override_options): Disable red zone by default on i386. > (compute_frame_layout, ix86_force_to_memory, ix86_free_from_memory): > Do not test TARGET_64BIT together with TARGET_RED_ZONE > > > I admit i never looked at the actual code, though, to be sure > of what it was doing. When i saw that appear in the changelog, > i thought that it was a reaction to the bug that i had reported... I see, that patch disable red zone on i386, where one can't use it but not for x86-64, so the code path is not dead. The bugreport is ineed solved and can be closed. Thanks! Honza > > sss