--- Comment #6 from siarhei dot siamashka at gmail dot com 2009-12-21
08:27 ---
Created an attachment (id=19356)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19356action=view)
return-address-prediction-bench.c
This looks like a really serious performance issue. Not just indirect
--- Comment #7 from siarhei dot siamashka at gmail dot com 2009-12-21
08:53 ---
(In reply to comment #4)
I would rather split the load out as a separate insn and allow it to be
scheduled separately.
A question just to clarify the status of this issue. Are you waiting for David
(or
--
jakub at gcc dot gnu dot org changed:
What|Removed |Added
Summary|Miscompilation of 2fish on |[4.4 Regression]
|s390
--- Comment #5 from abel at gcc dot gnu dot org 2009-12-21 11:38 ---
Me and Alexander will be looking at this together with the sel-sched ppc bugs.
--
abel at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #11 from mikpe at it dot uu dot se 2009-12-21 11:55 ---
(In reply to comment #10)
Fixed.
You forgot to add the test case in the 4.4 backport.
--
mikpe at it dot uu dot se changed:
What|Removed |Added
Compile the attached sources as follows:
g++ -std=gnu++0x -c -Wno-pmf-conversions file_reader.cpp
An Internal Compiler Error report will appear:
file_reader.cpp: In instantiation of
'std::unique_ptrbase::dynamic_dispatchvoid
(io::file_reader::*)(io::request)::entry [], base::release_deleter':
--- Comment #1 from piotr dot wyderski at gmail dot com 2009-12-21 12:41
---
Created an attachment (id=19357)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19357action=view)
The faulting code
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--- Comment #2 from paolo dot carlini at oracle dot com 2009-12-21 13:00
---
Any chance you can provide a smaller reproducer? Thanks.
--
paolo dot carlini at oracle dot com changed:
What|Removed |Added
Following testcase:
--cut here--
struct S2180
{
char t;
_Complex char u[2];
};
struct S2180 s2180;
int
main (void)
{
volatile struct S2180 x;
s2180.u[1] = 3 + 4i;
x.u[1] = s2180.u[1];
if (x.u[1] != s2180.u[1])
abort ();
return 0;
}
--cut here--
aborts for all optimization
--- Comment #1 from ubizjak at gmail dot com 2009-12-21 13:19 ---
Testsuite failure, so regression on all branches.
--
ubizjak at gmail dot com changed:
What|Removed |Added
--- Comment #3 from piotr dot wyderski at gmail dot com 2009-12-21 13:47
---
(In reply to comment #2)
Any chance you can provide a smaller reproducer? Thanks.
No, every simpler testcase based on the attached code I tried
to create compiles successfully. Perhaps here is the problem?
--- Comment #4 from paolo dot carlini at oracle dot com 2009-12-21 14:11
---
Certainly *is* a problem if we hope to debug the issue decently fast...
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--- Comment #5 from piotr dot wyderski at gmail dot com 2009-12-21 14:27
---
(In reply to comment #4)
Certainly *is* a problem if we hope to debug the issue decently fast...
I meant the cause of the problem is in its size, i.e.
there must be a critical mass of templates to trigger
--- Comment #6 from paolo dot carlini at oracle dot com 2009-12-21 14:34
---
Ok... Did you help yourself with delta? See:
http://gcc.gnu.org/wiki/A_guide_to_testcase_reduction
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--- Comment #20 from hjl at gcc dot gnu dot org 2009-12-21 15:32 ---
Subject: Bug 42345
Author: hjl
Date: Mon Dec 21 15:31:49 2009
New Revision: 155372
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=155372
Log:
Always pass --32 to ia32 Linux assembler.
2009-12-21 H.J. Lu
--- Comment #10 from jamborm at gcc dot gnu dot org 2009-12-21 15:35
---
So given the above, I still think the patch from comment 7 will help.
Because I don't have any access to a non-Linux platform I cannot try
it myself and thus can only rely on others to test it for me. I'll be
--- Comment #21 from hjl dot tools at gmail dot com 2009-12-21 16:06
---
Reopened
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
--- Comment #22 from hjl dot tools at gmail dot com 2009-12-21 16:07
---
Fixed.
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Severity|major |normal
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--- Comment #7 from paolo dot carlini at oracle dot com 2009-12-21 16:14
---
I'm reducing it.
--
paolo dot carlini at oracle dot com changed:
What|Removed |Added
--- Comment #7 from jakub at gcc dot gnu dot org 2009-12-21 16:32 ---
Subject: Bug 42429
Author: jakub
Date: Mon Dec 21 16:32:44 2009
New Revision: 155374
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=155374
Log:
PR rtl-optimization/42429
* reload.c
--- Comment #8 from jakub at gcc dot gnu dot org 2009-12-21 16:34 ---
Subject: Bug 42429
Author: jakub
Date: Mon Dec 21 16:34:26 2009
New Revision: 155375
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=155375
Log:
PR rtl-optimization/42429
* reload.c
I have just updated to a new tango version.
The source for tango can be found in
http://www.dsource.org/projects/tango/wiki/Download
svn co can be done
svn co http://svn.dsource.org/projects/tango/tags/releases/0.99.8
The version report by svn update is revision 5270.
I build the tango library
--- Comment #8 from piotr dot wyderski at gmail dot com 2009-12-21 16:38
---
(In reply to comment #6)
A marvelous tool! I'm reducing it too,
staring at the proces with half-open mouth...
--
piotr dot wyderski at gmail dot com changed:
What|Removed
--
paolo dot carlini at oracle dot com changed:
What|Removed |Added
Severity|major |normal
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--- Comment #9 from paolo dot carlini at oracle dot com 2009-12-21 17:18
---
Note that a compiler built --disable-checking doesn't ICE, just errors out.
Thus, are you sure the code is valid, before anything else?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--
ramana at gcc dot gnu dot org changed:
What|Removed |Added
AssignedTo|unassigned at gcc dot gnu |ramana at gcc dot gnu dot
|dot org
--- Comment #10 from piotr dot wyderski at gmail dot com 2009-12-21 17:23
---
(In reply to comment #9)
Thus, are you sure the code is valid, before anything else?
It compiles and works on gcc version 4.5.0 20090604.
The current compiler is
Configured with: ../configure
--
paolo dot carlini at oracle dot com changed:
What|Removed |Added
Severity|major |normal
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42447
--- Comment #11 from piotr dot wyderski at gmail dot com 2009-12-21 17:38
---
An even more reduced testcase which ICEs. Delta is amazing...
I think I'll stop here.
// 8
namespace std
class tuple
{
};
templatestd::size_t __i,
--- Comment #2 from jamborm at gcc dot gnu dot org 2009-12-21 17:55 ---
This happens because IPA-CP (as opposed to indirect inlining) does not
perform parameter modification analysis but the associated streaming
(which is shared in between the two) code assumes it has been done. I
will
--- Comment #12 from paolo dot carlini at oracle dot com 2009-12-21 18:33
---
Ok, thanks. From your work I produced the below, which doesn't use any C++0x
feature and to me looks like a 4.5 Regression.
///
templatetypename _Tp, typename _Up
struct tuple { };
' CXXFLAGS='-march=x86-64
-mtune=generic -O2 -pipe' --enable-languages=c,c++,lto --no-create
--no-recursion
Thread model: posix
gcc version 4.5.0 20091221 (experimental) (GCC
--- Comment #1 from jacob dot benoit dot 1 at gmail dot com 2009-12-21
18:57 ---
Created an attachment (id=19358)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19358action=view)
Preprocessed C++ source triggering this ICE
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42450
--- Comment #20 from jacob dot benoit dot 1 at gmail dot com 2009-12-21
18:59 ---
Thank you, I can confirm that it's fixed here.
However I now hit another ICE, see bug #42450, but it seems unrelated.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42225
Command line:
g++ B.ii C.ii -fPIC -shared -flto
or
g++ B.ii C.ii -fPIC -shared -fwhopr
Tested revisions:
r155363 - warns
r155248 - warns
r155141 - warns
r154886 - warns
Output (r155363, r155248):
$ /mnt/svn/gcc-trunk/binary-155363-lto/bin/g++ B.ii C.ii -fPIC -shared -fwhopr
header.hpp:10:19:
--- Comment #1 from zsojka at seznam dot cz 2009-12-21 19:06 ---
Created an attachment (id=19359)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19359action=view)
first source file (preprocessed, reduced)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42451
--- Comment #2 from zsojka at seznam dot cz 2009-12-21 19:06 ---
Created an attachment (id=19360)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19360action=view)
second source file (preprocessed, reduced)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42451
--- Comment #2 from ubizjak at gmail dot com 2009-12-21 19:36 ---
This happens on !TARGET_BWX targets. Unaligned HImode access can cross widened
SImode boundaries. Following patch triggers on this invalid condition in
get_aligned_mem:
Index: alpha.c
--- Comment #14 from mdorey at bluearc dot com 2009-12-21 20:02 ---
so [implicit] conversion from NULL to int is OK.
That's true where NULL is defined as 0 (or eg 0L), but that's not the only
permitted definition of NULL.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35669
--- Comment #5 from jakub at gcc dot gnu dot org 2009-12-21 20:17 ---
The problem is in the check_live_switch change.
As -march=native spec handling uses %march=native, SWITCH_IGNORE is set on the
option, and so the on the second and following input file check_live_switch
returns false.
--- Comment #3 from zsojka at seznam dot cz 2009-12-21 20:19 ---
I am also receiving ICE (with checking enabled) probably caused by this
incompability. I can provide reduced sources if needed, demonstrating that (it
would probably take several days of reducing).
The crash message in
--- Comment #13 from paolo dot carlini at oracle dot com 2009-12-21 20:29
---
A tad simpler:
templateint
void* get(int);
templatetypename
struct unique_ptr;
templatetypename _Tp
struct unique_ptr_Tp[]
{
typedef int __tuple_type;
void*
get() const
{ return
--- Comment #14 from hjl dot tools at gmail dot com 2009-12-21 20:54
---
It is caused by revision 151461:
http://gcc.gnu.org/ml/gcc-cvs/2009-09/msg00207.html
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
The options given when GCC was configured/built:
Using built-in specs.
Target: arm-none-eabi
Configured with: /scratch/julian/2009q3-respin-eabi-lite/src/gcc-4.4/configure
--build=i686-pc-linux-gnu --host=i686-min
gw32 --target=arm-none-eabi --enable-threads --disable-libmudflap
--disable-libssp
--- Comment #1 from cwang at rossvideo dot com 2009-12-21 21:27 ---
Actually, The value which r4 points to in the above function will be corrupted
when an interrupt occurs between address 11c6e2 and address 11c6e4 and the
interrupt service routine shares a same stack with the above
--- Comment #3 from d dot g dot gorbachev at gmail dot com 2009-12-21
21:42 ---
Created an attachment (id=19361)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19361action=view)
Small testcase
Run `gcc -flto -fuse-linker-plugin 1.c 2.c' (or just `gcc -flto
-fuse-linker-plugin
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Severity|major |normal
Component|c |target
--- Comment #4 from pinskia at gcc dot gnu dot org 2009-12-21 21:46 ---
*** Bug 42452 has been marked as a duplicate of this bug. ***
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #2 from pinskia at gcc dot gnu dot org 2009-12-21 21:46 ---
*** This bug has been marked as a duplicate of 38644 ***
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #2 from hjl dot tools at gmail dot com 2009-12-21 22:04 ---
It is caused by revision 151696:
http://gcc.gnu.org/ml/gcc-cvs/2009-09/msg00443.html
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
touch xxx.c
gcc -flto -fuse-linker-plugin xxx.c
collect2: ld terminated with signal 6 [Aborted]
ld: ../../gcc-4.5/lto-plugin/lto-plugin.c:325: write_resolution: Assertion
`syms' failed.
--
Summary: Assertion `syms' failed in lto-plugin
Product: gcc
Version:
--- Comment #1 from d dot g dot gorbachev at gmail dot com 2009-12-21
22:32 ---
Created an attachment (id=19362)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19362action=view)
Backtrace
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42453
--
reichelt at gcc dot gnu dot org changed:
What|Removed |Added
CC||reichelt at gcc dot gnu dot
|
When -ffunction-sections is turned on with -g2, gcc emits an entry into the
.debug_ranges section for the DW_TAG_compile_unit for the range (.Ltext,
.Letext) unconditionally. With -ffunction-sections, the .text section remains
unused, and this entry in .debug_ranges refers to a zero-length range.
--- Comment #1 from ccoutant at gcc dot gnu dot org 2009-12-21 23:13
---
Created an attachment (id=19363)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19363action=view)
Proposed fix for PR 42454
* dwarf2out.c (dwarf2out_finish): Don't output range for .text
--- Comment #6 from tkoenig at gcc dot gnu dot org 2009-12-21 23:14 ---
Test case committed to trund and 4.4 branch.
Closing.
--
tkoenig at gcc dot gnu dot org changed:
What|Removed |Added
--
ccoutant at gcc dot gnu dot org changed:
What|Removed |Added
Status|UNCONFIRMED |ASSIGNED
Ever Confirmed|0 |1
Last
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 00:01
---
Subject: Bug 36903
Author: hutchinsonandy
Date: Tue Dec 22 00:00:50 2009
New Revision: 155382
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=155382
Log:
2009-12-21 Andy Hutchinson
Command line:
g++ -fcompare-debug -O2 testcase.ii
Tested revisions:
r155363 - crash
r155256 - crash
r155141 - crash
r154886 - crash
r154830 - OK
r153685 - OK
4.4 - doesn't know -fcompare-debug
Output:
$ /mnt/svn/gcc-trunk/binary-155363-lto/bin/g++ -fcompare-debug testcase.ii -O2
g++:
--- Comment #1 from zsojka at seznam dot cz 2009-12-22 00:07 ---
Created an attachment (id=19364)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19364action=view)
original testcase
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42455
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 00:16
---
Fixed 4.5. AVR wont vectorize with word size of 1 byte!
--
hutchinsonandy at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #2 from zsojka at seznam dot cz 2009-12-22 00:10 ---
Created an attachment (id=19365)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19365action=view)
somewhat reduced testcase
Reducing the testcase is quite complicated, removing seemingly irrelevant lines
causes the
On Linux/ia32, revision 155377 caused:
FAIL: 29_atomics/atomic_flag/clear/1.cc (test for excess errors)
FAIL: 29_atomics/atomic_flag/test_and_set/explicit.cc (test for excess errors)
FAIL: 29_atomics/atomic_flag/test_and_set/implicit.cc (test for excess errors)
--
Summary: [4.5
--- Comment #3 from hjl dot tools at gmail dot com 2009-12-22 01:52 ---
It is caused by revision 154869:
http://gcc.gnu.org/ml/gcc-cvs/2009-12/msg00013.html
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
--- Comment #1 from danglin at gcc dot gnu dot org 2009-12-22 03:09 ---
Compiles successfully without -fbounds-check on hppa-unknown-linux-gnu.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42169
AVR will fail to build with (default) LTO enabled.
LTO becomes enabled when a working libelf implementation is found.
Mike Stein brought this to my attention and I have been able to replicate it
now I have loaded libelf.
The undefined symbol `builtin_define_std' is defined in c-cppbuiltin.c. It
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 03:20
---
Created an attachment (id=19366)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19366action=view)
Make log file
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
--- Comment #2 from pinskia at gcc dot gnu dot org 2009-12-22 03:29 ---
A way to fix this is to have avr-c.c like most other targets have. And declare
avr_cpu_cpp_builtins in there. Note this change broke other languages besides
C/C++/Objective-C/Objective-C++ so this is a regression
--- Comment #1 from hjl dot tools at gmail dot com 2009-12-22 03:36 ---
It may also have caused:
UNRESOLVED: attribute_plugin.c compilation, -I.
-I/export/gnu/import/svn/gcc-test/src-trunk/gcc/testsuite
-I/export/gnu/import/svn/gcc-test/src-trunk/gcc/testsuite/../../gcc
--- Comment #2 from bkoz at gcc dot gnu dot org 2009-12-22 04:20 ---
HJ what are the excess errors that you see (from libstdc++.log file?)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42456
--- Comment #23 from bkoz at gcc dot gnu dot org 2009-12-22 04:33 ---
thanks HJ
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42345
--- Comment #3 from hjl dot tools at gmail dot com 2009-12-22 05:56 ---
I got
/export/gnu/import/svn/gcc-test/src-trunk/libstdc++-v3/testsuite/29_atomics/atomic_flag/clear/1.cc:28:
undefined reference to
`std::__atomic0::atomic_flag::clear(std::memory_order)'^M
--- Comment #4 from hjl dot tools at gmail dot com 2009-12-22 06:03 ---
(In reply to comment #1)
It may also have caused:
UNRESOLVED: attribute_plugin.c compilation, -I.
-I/export/gnu/import/svn/gcc-test/src-trunk/gcc/testsuite
I have picked the following lines from file gcc-4.5-b20091203.pot:
#: c-common.c:7053 c-common.c:7145
#, gcc-internal-format
msgid can%'t set %qE attribute after definition
#: c-decl.c:1533
#, gcc-internal-format
msgid a parameter list with an ellipsis can%'t match an empty parameter name
list
75 matches
Mail list logo
