--- Comment #5 from hjl dot tools at gmail dot com 2009-02-14 23:38 ---
This latent bug is triggered by the part of revision 143157:
---
Index: i386.c
===
--- i386.c (revision 143154)
+++ i386.c (working copy)
@@
--
ktietz at gcc dot gnu dot org changed:
What|Removed |Added
Priority|P3 |P2
Target Milestone|--- |4.4.0
http://gcc.
--- Comment #94 from dberlin at gcc dot gnu dot org 2009-02-14 23:06
---
Subject: Re: [4.3/4.4 Regression] Inordinate
compile times on large routines
One of the reasons LCM in RTL is so slow is because it uses a crappy
iteration order.
With the right iteration order, it shoul
--- Comment #9 from bugzilla-gcc at undoso dot eu 2009-02-14 23:06 ---
Created an attachment (id=17302)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17302&action=view)
Proprocessed source of the error
Yes, should have thought to do that before. It was obvious that ccache might
no
--- Comment #9 from nightstrike at gmail dot com 2009-02-14 22:56 ---
Verified to fail on win32 and win64, not just win64. Can someone with
sufficient privileges adjust "Target"?
Verified to work in 4.3, so this is a regression.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39179
--- Comment #8 from rguenth at gcc dot gnu dot org 2009-02-14 22:54 ---
It would be this one: /root/.ccache/scsi_low.tmp..22970.i
Maybe instead of /usr/local/libexec/ccache/world-cc use the system gcc command?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39193
--- Comment #7 from bugzilla-gcc at undoso dot eu 2009-02-14 22:03 ---
Oops, I forgot the next line :
/usr/libexec/cc1 -fpreprocessed /root/.ccache/scsi_low.tmp..22970.i -quiet
-dumpbase scsi_low.tmp..22970.i -march=pentiumpro -mno-align-long-strings
-mpreferred-stack-boundary=2 -mno-mmx
--- Comment #6 from bugzilla-gcc at undoso dot eu 2009-02-14 22:01 ---
I'm probably blind, but I don't see it :
/usr/libexec/cc1 -E -quiet -nostdinc -v -I. -I@ -I@/contrib/altq
-I/usr/obj/usr/src/sys/GENERIC -D_LONGLONG -D_KERNEL -DKLD_MODULE
-DHAVE_KERNEL_OPTION_HEADERS -include /usr/ob
--- Comment #20 from sebor at roguewave dot com 2009-02-14 21:58 ---
Created an attachment (id=17301)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17301&action=view)
Corrected unified demo.
Attached a corrected unified demo with assertions removed and with output on
popular imple
--- Comment #93 from lucier at math dot purdue dot edu 2009-02-14 21:58
---
Subject: Re: [4.3/4.4 Regression] Inordinate compile times on large routines
I instrumented the compiler and looked how many nodes were in each
loop processed by LICM for the Gambit runtime and compiler.
Fo
--- Comment #19 from paolo dot carlini at oracle dot com 2009-02-14 21:54
---
Thanks Martin. Then, as requested by submitter, for now I'm just going to tweak
a bit the behaviour of libstdc++ to bring consistency between unsigned and
signed char platforms, in other terms, value <= 0 and
--- Comment #5 from rguenth at gcc dot gnu dot org 2009-02-14 21:50 ---
Hm, if you add -v as well you should see a separate pre-processing and
compilation sub-commands. The compilation subcommand should refer to the
scsi_low.i file, wherever it is created.
--
http://gcc.gnu.org/bug
--- Comment #18 from sebor at roguewave dot com 2009-02-14 21:41 ---
I was too hasty -- the attached test case is buggy: it's missing a seek to
the beginning of the stream after the first extraction, making the results
for the num_get part meaningless.
(In reply to comment #7)
> "Arbitr
--- Comment #8 from ktietz at gcc dot gnu dot org 2009-02-14 21:25 ---
(In reply to comment #7)
> What happens if you just use
>
> return default_binds_local_p_1 (exp, (TREE_CODE (exp) == VAR_DECL || TREE_CODE
> (exp) == FUNCTION_DECL)
>&& DECL_DLLIMPORT_P (exp));
>
> ?
>
Sam
--- Comment #4 from bugzilla-gcc at undoso dot eu 2009-02-14 21:24 ---
gcc told me it was ignored before. But I tried anyway. The file is still not
created even without -pipe.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39193
--- Comment #17 from sebor at roguewave dot com 2009-02-14 21:21 ---
Created an attachment (id=17300)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17300&action=view)
Unified money_get and num_get test case and results.
Attached is a unified test case with test results on popular
--- Comment #7 from rguenth at gcc dot gnu dot org 2009-02-14 21:12 ---
What happens if you just use
return default_binds_local_p_1 (exp, (TREE_CODE (exp) == VAR_DECL || TREE_CODE
(exp) == FUNCTION_DECL)
&& DECL_DLLIMPORT_P (exp));
?
--
http://gcc.gnu.org/bugzilla/show_bug.
--- Comment #3 from rguenth at gcc dot gnu dot org 2009-02-14 21:10 ---
You need to remove -pipe from the commandline.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39193
--- Comment #2 from bugzilla-gcc at undoso dot eu 2009-02-14 20:12 ---
I have tried to get the file but when I execute
/usr/local/libexec/ccache/world-cc -save-temps -Os -pipe -mmmx -msse -msse2
-msse3 -march=pentiumpro -fno-strict-aliasing -D_KERNEL -DKLD_MODULE -std=c99
-nostdinc -D
--- Comment #6 from ktietz at gcc dot gnu dot org 2009-02-14 20:10 ---
(In reply to comment #2)
> The problem is that targetm.binds_local_p returns true for
>
> type size
> unit size
> align 32 symtab 0 alias set -1 canonical type 0xb785edd0 precision 32
--- Comment #1 from rguenth at gcc dot gnu dot org 2009-02-14 19:24 ---
I suppose _bin_coeff_tab is not properly aligned. Maybe PE support is not
properly honoring DECL_ALIGN. Please try also GCC 4.3.3 which has many
bugfixes compared to 4.3.0.
--
rguenth at gcc dot gnu dot org cha
--- Comment #1 from rguenth at gcc dot gnu dot org 2009-02-14 19:21 ---
We need preprocessed source of the testcase which you can obtain by appending
-save-temps to the command line. The file will be called scsi_low.i.
--
rguenth at gcc dot gnu dot org changed:
What|
Hi,
I try the following code:
#define MAX_N 40
#define MAX_R 25
#ifndef FALSE
#define FALSE 0
#define TRUE 1
#endif
static unsigned int bin_coeff_tab[ MAX_N ][ MAX_R ];
static int is_calculated = FALSE;
static void
bin_coeff_init( void )
{
int i, j;
for( i = 0; i < MAX_N; i++ )
--- Comment #5 from rguenth at gcc dot gnu dot org 2009-02-14 18:49 ---
Jason, can we fix this in the C++ frontend? Having both TREE_STATIC and
DECL_EXTERNAL set seems bogus.
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
-
--- Comment #4 from rguenth at gcc dot gnu dot org 2009-02-14 18:48 ---
Btw, this looks like similar to PR36207 which was fixed in the Ada frontend.
Still I think that i386_pe_binds_local_p is completely bogus.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39179
Here is the gcc log :
===> scsi_low (all)
/usr/local/libexec/ccache/world-cc -Os -pipe -mmmx -msse -msse2 -msse3
-march=pentiumpro -fno-strict-aliasing -D_KERNEL -DKLD_MODULE -std=c99
-nostdinc -DHAVE_KERNEL_OPTION_HEADERS -include
/usr/obj/usr/src/sys/GENERIC/opt_global.h -I. -I@ -I@/contrib/al
--enable-multilib --enable-decimal-float
--with-long-double-128 --with-included-gettext --enable-stage1-checking
--enable-checking=release --with-tune=k8 --with-cpu=k8 --with-arch=k8
--with-gnu-as --with-as=/usr/local/bin/as --with-gnu-ld
--with-ld=/usr/local/bin/ld --with-gmp=/usr/local --with-mpfr=/usr
--- Comment #3 from ebotcazou at gcc dot gnu dot org 2009-02-14 17:16
---
> The problem is that targetm.binds_local_p returns true for
>
> type size
> unit size
> align 32 symtab 0 alias set -1 canonical type 0xb785edd0 precision 32
> min max >
> r
--- Comment #2 from rguenth at gcc dot gnu dot org 2009-02-14 17:03 ---
The problem is that targetm.binds_local_p returns true for
unit size
align 32 symtab 0 alias set -1 canonical type 0xb785edd0 precision 32
min max >
readonly used public static unsigned exter
--- Comment #1 from rguenth at gcc dot gnu dot org 2009-02-14 16:55 ---
Single-file testcase, reproduces with a x86_64-pc-mingw32 cross at any
optimization level. get_symbol_constant_value returns zero for K::k.
struct K {
static const unsigned k;
};
extern "C" void abort (void);
i
--- Comment #1 from kargl at gcc dot gnu dot org 2009-02-14 16:00 ---
I'm inclined to think that this should be closed with WONTFIX.
Deallocation of an allocated array when a subprogram returns
is a fundamental concept of Fortran. A user should know what
to expect and code accordingly.
--- Comment #92 from stevenb dot gcc at gmail dot com 2009-02-14 14:42
---
Subject: Re: [4.3/4.4 Regression] Inordinate
compile times on large routines
Re: Comment #88
I think the patch is perfectly acceptable as a stop-gap solution. I
don't think we have anything better fo
--- Comment #6 from hjl dot tools at gmail dot com 2009-02-14 14:37 ---
Fixed.
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
Status|UNCONFIRM
nu-as --with-as=/usr/local/bin/as --with-gnu-ld
--with-ld=/usr/local/bin/ld --with-gmp=/usr/local --with-mpfr=/usr/local
--without-ppl
Thread model: posix
gcc version 4.4.0 20090214 (experimental) [trunk revision 144177] (GCC)
Tada,
Rob
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39186
--- Comment #1 from pinskia at gcc dot gnu dot org 2009-02-14 10:05 ---
This has since been fixed for 4.4.0 (on the trunk).
*** This bug has been marked as a duplicate of 36254 ***
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
---
--- Comment #16 from pinskia at gcc dot gnu dot org 2009-02-14 10:05
---
*** Bug 39191 has been marked as a duplicate of this bug. ***
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
-
--- Comment #1 from domob at gcc dot gnu dot org 2009-02-14 09:50 ---
Confirmed.
--
domob at gcc dot gnu dot org changed:
What|Removed |Added
CC|
> gfortran test.f90
test.f90:1.10:
IF (i==2) b(3)
1
Error: Cannot assign to a named constant at (1)
> cat test.f90
IF (i==2) b(3)
END
(The correct code would have been, the error message is confusing, syntax error
would have been better)
IF (i==2) CALL b(3)
END
--
Summary:
In this code:
//Load a style icon as a pixmap
QPixmap loadImagePixmap(const QString& fileName)
{
for (int i = 0; i < cacheOfPixmaps.size(); ++i) {
if(ScacheOfPixmaps.at(i)==fileName)
return cacheOfPixmaps.at(i);
}
QString fullPath=load
--- Comment #7 from doko at ubuntu dot com 2009-02-14 09:08 ---
seen as well on sparc-linux, x86_64-linux and i486-linux (when built with
-mcpu=i686 -mtune=i586), although triggered in different files.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38292
40 matches
Mail list logo
