[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 Jorn Wolfgang Rennecke amylaar at gcc dot gnu.org changed: What|Removed |Added CC||amylaar at gcc dot gnu.org --- Comment #33 from Jorn Wolfgang Rennecke amylaar at gcc dot gnu.org 2011-11-19 11:30:34 UTC --- There is also the case where no argument has been stored on the stack yet. That's quite common for ABIs that pass arguments in registers.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #32 from H.J. Lu hjl.tools at gmail dot com 2011-08-15 14:12:04 UTC --- Can we check 2 cases: 1. Indirect function call via a register (PR 50074). 2. Non of function arguments are pointers (PR 49179).
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #29 from rguenther at suse dot de rguenther at suse dot de 2011-07-12 08:35:18 UTC --- On Tue, 12 Jul 2011, ian at airs dot com wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 Ian Lance Taylor ian at airs dot com changed: What|Removed |Added CC||ian at airs dot com --- Comment #27 from Ian Lance Taylor ian at airs dot com 2011-07-12 04:37:01 UTC --- This patch seems too conservative and it appears that it will cause the compiler to miss all tailcalls with pointer arguments. It only matters if the register points to the incoming parameters, which can only happen in unusual cases. We should be able to determine that reliably. Janis, did you open a PR for the missing optimization? I didn't see one. The point is, of course, we have to determine it conservatively correct. The patch makes sure we do that. Richard.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519
--- Comment #30 from Eric Botcazou ebotcazou at gcc dot gnu.org 2011-07-12
08:42:40 UTC ---
This patch seems too conservative and it appears that it will cause the
compiler to miss all tailcalls with pointer arguments. It only matters if the
register points to the incoming parameters, which can only happen in unusual
cases. We should be able to determine that reliably.
This is a bit of an overstatement though, as
extern int foo (int, int *);
int i;
int bar (int a, int b)
{
int *p = i;
return foo (a, p);
}
will be caught. Even this
extern int foo (int, int);
int i;
int bar (int a, int b)
{
int *p = i;
return foo (a, *p);
}
will be caught thanks to alias analysis at the Tree level.
As for the reliably, this time it's a little optimistic IMO. :-) But we can
probably be less conservative with more work.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #31 from Janis Johnson janis at gcc dot gnu.org 2011-07-12 15:06:03 UTC --- I filed PR middle-end/49719 for the sibcall-1.c failure.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 Richard Guenther rguenth at gcc dot gnu.org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED --- Comment #26 from Richard Guenther rguenth at gcc dot gnu.org 2011-07-11 11:54:01 UTC --- Fixed. For the missed tailcall opt please open a separate bugreport.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 Ian Lance Taylor ian at airs dot com changed: What|Removed |Added CC||ian at airs dot com --- Comment #27 from Ian Lance Taylor ian at airs dot com 2011-07-12 04:37:01 UTC --- This patch seems too conservative and it appears that it will cause the compiler to miss all tailcalls with pointer arguments. It only matters if the register points to the incoming parameters, which can only happen in unusual cases. We should be able to determine that reliably. Janis, did you open a PR for the missing optimization? I didn't see one.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #28 from Ian Lance Taylor ian at airs dot com 2011-07-12 04:54:14 UTC --- *** Bug 49709 has been marked as a duplicate of this bug. ***
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #24 from hjl at gcc dot gnu.org hjl at gcc dot gnu.org 2011-07-08 13:12:05 UTC --- Author: hjl Date: Fri Jul 8 13:12:03 2011 New Revision: 176042 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=176042 Log: Fix PR middle-end/49519. gcc/ 2011-07-08 Kirill Yukhin kirill.yuk...@intel.com PR middle-end/49519 * calls.c (mem_overlaps_already_clobbered_arg_p): Additional check if address is stored in register. If so - give up. (check_sibcall_argument_overlap_1): Do not perform check of overlapping when it is call to address. gcc/tessuite/ 2011-07-08 Kirill Yukhin kirill.yuk...@intel.com PR middle-end/49519 * g++.dg/torture/pr49519.C: New test. Added: trunk/gcc/testsuite/g++.dg/torture/pr49519.C Modified: trunk/gcc/ChangeLog trunk/gcc/calls.c trunk/gcc/testsuite/ChangeLog
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 Janis Johnson janis at gcc dot gnu.org changed: What|Removed |Added CC||janis at gcc dot gnu.org --- Comment #25 from Janis Johnson janis at gcc dot gnu.org 2011-07-08 23:36:25 UTC --- Test gcc.target/arm/sibcall-1.c starts failing with r176042 for arm-none-linux-gnueabi configured with defaults for arch and cpu. The failure can be reproduced with a simple cross cc1.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 Richard Guenther rguenth at gcc dot gnu.org changed: What|Removed |Added Keywords||wrong-code Target|i686|i?86-*-* Component|c++ |middle-end --- Comment #15 from Richard Guenther rguenth at gcc dot gnu.org 2011-07-06 10:30:26 UTC --- The testcase reproduces the problem for me on trunk already with -m32 -O -foptimize-sibling-calls, so it doesn't seem to be IPA SRA related.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #16 from Yukhin Kirill kirill.yukhin at intel dot com 2011-07-06 10:35:03 UTC --- Yes. This is because expander prepares arguments like this: ... (insn 6 5 7 2 (parallel [ (set (reg:SI 64) (plus:SI (reg/f:SI 53 virtual-incoming-args) (const_int 4 [0x4]))) (clobber (reg:CC 17 flags)) ]) 2.cc:103 -1 (nil)) (insn 7 6 8 2 (set (reg:SI 65) (mem/c:SI (plus:SI (reg:SI 64) (const_int 12 [0xc])) [0 MEM[(int )t + 12]+0 S4 A32])) 2.cc:103 -1 (nil)) (insn 8 7 9 2 (set (mem:SI (plus:SI (reg/f:SI 53 virtual-incoming-args) (const_int 12 [0xc])) [0 S4 A32]) (reg:SI 65)) 2.cc:103 -1 (nil)) ... So, calls.c/mem_overlaps_already_clobbered_arg_p unable to determine, that data come from stack, so it returns true and enables tailcall.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #17 from Richard Guenther rguenth at gcc dot gnu.org 2011-07-06 10:37:57 UTC --- I suspect that the tailcalling code does not expect TER to happen when expanding the call arguments? With -fno-tree-ter the issue also isn't appearant.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #18 from Richard Guenther rguenth at gcc dot gnu.org 2011-07-06 10:49:24 UTC --- Hm, the argument setup for the tailcall clobbers the incoming stack argument. And without TER we do not overlap argument setup with the loads from the incoming stack argument.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #19 from Yukhin Kirill kirill.yukhin at intel dot com 2011-07-06 11:49:34 UTC --- Created attachment 24701 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=24701 Patch to make tailcall check more conservative Attached patch adds another check for clobbered stack area. If address comes from a register - we have no idea about destination address. That means we must act in conservative way - address possibly overlaps with stack area of interest.
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #20 from Yukhin Kirill kirill.yukhin at intel dot com 2011-07-06 11:50:51 UTC --- With patch attached both tescase and 447.dealII passing
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #21 from rguenther at suse dot de rguenther at suse dot de 2011-07-06 11:53:56 UTC --- On Wed, 6 Jul 2011, kirill.yukhin at intel dot com wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #19 from Yukhin Kirill kirill.yukhin at intel dot com 2011-07-06 11:49:34 UTC --- Created attachment 24701 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=24701 Patch to make tailcall check more conservative Attached patch adds another check for clobbered stack area. If address comes from a register - we have no idea about destination address. That means we must act in conservative way - address possibly overlaps with stack area of interest. That looks reasonable. Can you bootstrap test this fix and post it to gcc-patches?
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #22 from Yukhin Kirill kirill.yukhin at intel dot com 2011-07-06 11:57:21 UTC --- (In reply to comment #21) On Wed, 6 Jul 2011, kirill.yukhin at intel dot com wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 --- Comment #19 from Yukhin Kirill kirill.yukhin at intel dot com 2011-07-06 11:49:34 UTC --- Created attachment 24701 [details] -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=24701 Patch to make tailcall check more conservative Attached patch adds another check for clobbered stack area. If address comes from a register - we have no idea about destination address. That means we must act in conservative way - address possibly overlaps with stack area of interest. That looks reasonable. Can you bootstrap test this fix and post it to gcc-patches? Already in progress :)
[Bug middle-end/49519] [4.7 Regression] Revision 175272 miscompiled 447.dealII in SPEC CPU 2006
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519 H.J. Lu hjl.tools at gmail dot com changed: What|Removed |Added URL||http://gcc.gnu.org/ml/gcc-p ||atches/2011-07/msg00359.htm ||l --- Comment #23 from H.J. Lu hjl.tools at gmail dot com 2011-07-06 19:37:39 UTC --- A patch is posted at http://gcc.gnu.org/ml/gcc-patches/2011-07/msg00359.html
