--- Comment #31 from rguenth at gcc dot gnu dot org 2010-05-04 14:02
---
Fixed!
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
Status|ASSIGNE
--- Comment #30 from rguenth at gcc dot gnu dot org 2010-05-04 13:12
---
Subject: Bug 43879
Author: rguenth
Date: Tue May 4 13:12:02 2010
New Revision: 159026
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159026
Log:
2010-05-04 Richard Guenther
PR tree-optimizatio
--- Comment #29 from rguenth at gcc dot gnu dot org 2010-05-04 11:16
---
Indeed. Many thanks for these testcases! I have a fix in testing.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #28 from zsojka at seznam dot cz 2010-05-03 23:45 ---
Thank you, Steven and Richard! It's nice to see the testcases are useful and
appreciated.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #27 from steven at gcc dot gnu dot org 2010-05-03 16:56 ---
Zdenek, has anyone told you how amazing your contribution is here? Wow!
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #26 from zsojka at seznam dot cz 2010-05-03 15:02 ---
Created an attachment (id=20547)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20547&action=view)
next testcase
Reduced from libstdc++-v3/testsuite/23_containers/bitset/operations/1.cc
I am happy those testcases he
--- Comment #25 from zsojka at seznam dot cz 2010-05-03 14:56 ---
Created an attachment (id=20546)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20546&action=view)
next testcase, second part will follow
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #24 from rguenth at gcc dot gnu dot org 2010-05-02 18:12
---
Thanks for the testcase! The issue is now fixed and bootstrap with IPA-PTA
enabled succeeds and the testresults are the same as without (with -fipa-pta
enabled).
There are still the same libstdc++ miscompiles and
--- Comment #23 from rguenth at gcc dot gnu dot org 2010-05-02 18:11
---
Subject: Bug 43879
Author: rguenth
Date: Sun May 2 18:10:53 2010
New Revision: 158977
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158977
Log:
2010-05-02 Richard Guenther
PR tree-optimizatio
--- Comment #22 from zsojka at seznam dot cz 2010-05-02 13:47 ---
Of course the overwritten register is r14:
mov QWORD PTR [rsp-16], r14 #,
mov DWORD PTR [r15], ebx# *this_index_24(D), index
mov r14, QWORD PTR [rsp+72] #,
--
http://gcc.gnu.org
--- Comment #21 from zsojka at seznam dot cz 2010-05-02 13:45 ---
Created an attachment (id=20534)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20534&action=view)
reduced testcase, from lto-section-out.c
Miscompiled function is lto-section-out.c:lto_output_fn_decl_index()
lto_out
--- Comment #20 from rguenth at gcc dot gnu dot org 2010-04-30 18:53
---
Subject: Bug 43879
Author: rguenth
Date: Fri Apr 30 18:52:44 2010
New Revision: 158945
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158945
Log:
2010-04-30 Richard Guenther
PR tree-optimizatio
--- Comment #19 from rguenth at gcc dot gnu dot org 2010-04-30 17:12
---
(In reply to comment #18)
> > As far as I understand this use statement causes GFortran to read in the
> > bytecode from the .mod file and combine the files again to a single
> > translation unit. So this split is
--- Comment #18 from dominiq at lps dot ens dot fr 2010-04-30 15:53 ---
> As far as I understand this use statement causes GFortran to read in the
> bytecode from the .mod file and combine the files again to a single
> translation unit. So this split is in fact not a split.
>From my ve
--- Comment #17 from jv244 at cam dot ac dot uk 2010-04-30 15:26 ---
in this case the Fortran bits depend on PR40011
--
jv244 at cam dot ac dot uk changed:
What|Removed |Added
--- Comment #16 from rguenth at gcc dot gnu dot org 2010-04-30 15:17
---
(In reply to comment #15)
> > ... without -fwhole-file. But all IPA optimizations rely on that, so it's
> > critical that -fwhole-file is enabled.
>
> -fwhole-file works only at the file level, i.e., if proc_ptr_
--- Comment #15 from dominiq at lps dot ens dot fr 2010-04-30 14:51 ---
> ... without -fwhole-file. But all IPA optimizations rely on that, so it's
> critical that -fwhole-file is enabled.
-fwhole-file works only at the file level, i.e., if proc_ptr_comp_12.f90 is
split as:
[macbook]
--- Comment #14 from rguenth at gcc dot gnu dot org 2010-04-30 13:57
---
Indeed all proc_ptr_* execution FAILs are fixed with -fwhole-program. The
frontend does not properly unify decls for example in proc_ptr_comp_12.f90
for
testObj%test => returnMat
vs.
function returnMat( a,
--- Comment #13 from rguenth at gcc dot gnu dot org 2010-04-30 10:06
---
I can now successfully bootstrap with -O2 -fipa-pta but lto1 seems to be
miscompiled (all LTO tests ICE).
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #12 from rguenth at gcc dot gnu dot org 2010-04-30 09:00
---
With the last patch we are down to
=== libstdc++ tests ===
Running target unix/-fipa-pta/
FAIL: 23_containers/bitset/operations/1.cc execution test
FAIL: 23_containers/bitset/to_ullong/1.cc execu
--- Comment #11 from rguenth at gcc dot gnu dot org 2010-04-30 08:23
---
Subject: Bug 43879
Author: rguenth
Date: Fri Apr 30 08:22:15 2010
New Revision: 158924
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158924
Log:
2010-04-30 Richard Guenther
PR tree-optimizatio
--- Comment #10 from rguenth at gcc dot gnu dot org 2010-04-29 15:37
---
And the bug is:
struct TBL tbl = { foo };
but:
Generating constraints for global initializers
...
tbl = NONLOCAL
tbl = &NULL
ouch. I have a patch.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #9 from rguenth at gcc dot gnu dot org 2010-04-29 15:29 ---
Wow, thanks for the testcase!
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #8 from zsojka at seznam dot cz 2010-04-28 15:12 ---
Created an attachment (id=20508)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20508&action=view)
second part of reduced testcase
These testcases demonstrate what happens in cfgrtl.c, why is
cfg_layout_merge_blocks m
--- Comment #7 from zsojka at seznam dot cz 2010-04-28 15:09 ---
Created an attachment (id=20507)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20507&action=view)
first part of reduced testcase
second part to follow
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43879
--- Comment #6 from rguenth at gcc dot gnu dot org 2010-04-28 12:53 ---
Real miscompiles prevailing:
=== libstdc++ tests ===
Running target unix/-fipa-pta/
FAIL: 20_util/shared_ptr/thread/default_weaktoshared.cc execution test
FAIL: 23_containers/bitset/cons/1.cc execu
--- Comment #5 from rguenth at gcc dot gnu dot org 2010-04-28 11:51 ---
Subject: Bug 43879
Author: rguenth
Date: Wed Apr 28 11:51:31 2010
New Revision: 158825
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158825
Log:
2010-04-28 Richard Guenther
PR tree-optimization/
--- Comment #4 from rguenth at gcc dot gnu dot org 2010-04-24 20:04 ---
(In reply to comment #2)
> Bootstrap fails too with -fipa-pta, (at least) cfgrtl.o is miscompiled.
Yep, I know. If you can isolate the miscompile that would be great
(I am concentrating on completing IPA PTA as tim
--- Comment #3 from rguenth at gcc dot gnu dot org 2010-04-24 18:55 ---
Well, -fipa-pta is no longer a no-op and I wouldn't call the existing (known)
bugs a regression. I have some pending patches to fix issues.
--
rguenth at gcc dot gnu dot org changed:
What|Removed
29 matches
Mail list logo