I've extracted the computational kernel of CP2K (see PR 29975) for easier
benchmarking. Together with required utility routines to turn it into a
self-contained program and data to test it, I have made it available here:
http://www.pci.unizh.ch/vandevondele/tmp/extracted_collocate.tgz
the summary
--- Comment #1 from jv244 at cam dot ac dot uk 2007-03-02 08:39 ---
Created an attachment (id=13131)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13131&action=view)
gfortran kernel asm
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31021
--- Comment #2 from jv244 at cam dot ac dot uk 2007-03-02 08:39 ---
Created an attachment (id=13132)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13132&action=view)
ifort kernel asm
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31021
--- Comment #73 from jv244 at cam dot ac dot uk 2007-03-02 08:41 ---
I've added PR 31021 to track some performance issue with gfortran on one of
CP2K's kernels.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29975
When I compile a testcase that includes float and inline, it asserts internal
compiler error if it is optimized with -O2. If it is optimized with -O, this
error is not asserted.
% gcc -c test.c -O2 -m4
test.c: In function 'fff':
test.c:19: internal compiler error: in reg_overlap_mentioned_p, at
rt
--- Comment #41 from rguenth at gcc dot gnu dot org 2007-03-02 09:02
---
Thaks for the detective work! I sort of expected the fold patch to be weird or
have no effect - but it was needed only (for the testcase) to get rid of the
target_first == 128 comparison, as that is confusing VRP.
--- Comment #3 from dcb314 at hotmail dot com 2007-03-02 09:07 ---
(In reply to comment #1)
> Created an attachment (id=8382)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8382&action=view) [edit]
> A patch that implements this feature (with some false positives)
For fun, I tried
--- Comment #42 from baldrick at free dot fr 2007-03-02 09:16 ---
Subject: Re: VRP fails to eliminate range checks in Ada code
> Thaks for the detective work! I sort of expected the fold patch to be weird
> or
> have no effect - but it was needed only (for the testcase) to get rid of
--- Comment #3 from rguenth at gcc dot gnu dot org 2007-03-02 09:38 ---
You can attach a compressed version or provide a URL to download from. The
mail
didn't seem to get through.
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #3 from burnus at gcc dot gnu dot org 2007-03-02 09:38 ---
On my "AMD Athlon(tm) 64 X2 Dual Core Processor 4800+", gfortran is in x86_64
mode only 13% slower:
gfortran: Kernel time 5.872366, real 0m33.121s; user 0m32.898s; sys 0m0.088s.
Ifort:Kernel time 5.244328, real 0m
--- Comment #1 from rguenth at gcc dot gnu dot org 2007-03-02 09:42 ---
I count 5 and one such use in i386.c.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31019
Transformations in fold do not properly handle integer types with value ranges
(defined by TYPE_MIN_VALUE and TYPE_MAX_VALUE) that do not correspond to the
types precision.
Citing from PR30911 comment #40
"The problem is in this transformation:
/* Fold (X & C) op (Y & C) as (X ^ Y) & C op
--- Comment #43 from rguenth at gcc dot gnu dot org 2007-03-02 09:53
---
I agree. Let's create another bug for this issue. PR31023.
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
--
--- Comment #4 from jv244 at cam dot ac dot uk 2007-03-02 09:55 ---
(In reply to comment #3)
> On my "AMD Athlon(tm) 64 X2 Dual Core Processor 4800+", gfortran is in x86_64
> mode only 13% slower:
> gfortran: Kernel time 5.872366, real 0m33.121s; user 0m32.898s; sys 0m0.088s.
> Ifort:
--- Comment #4 from dfranke at gcc dot gnu dot org 2007-03-02 09:57 ---
Tobias, do the cases given in PR31016 include the one above?
If yes, this PR could be closed as dupe?!
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31009
--- Comment #2 from ubizjak at gmail dot com 2007-03-02 10:12 ---
Mine, I have a patch in testing.
And this is _definitelly_ not a microoptimization. I'll post the patch shortly
to gcc-patches, please look at some suprising numbers below...
size cc1
textdata bss dec
Using GCC 4.1.2 and compiling with -static -O2, the SPEC CPU2000 benchmark
vortex segfaults on my system (Fedora Core 4, Intel Pentium 4).
This issue was reported before (see
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23449), and seem to be resolved in
GCC 4.1.1. Using -fno-strict-aliasing solves
--- Comment #4 from bernd dot speiser at uni-tuebingen dot de 2007-03-02
10:26 ---
Subject: Re: infinite loop during compilation on x86_64
architecture
rguenth at gcc dot gnu dot org wrote:
> --- Comment #3 from rguenth at gcc dot gnu dot org 2007-03-02 09:38
> ---
> You ca
--- Comment #5 from burnus at gcc dot gnu dot org 2007-03-02 10:43 ---
> Tobias, do the cases given in PR31016 include the one above?
> If yes, this PR could be closed as dupe?!
Actually not. PR 31016 (and related PR 31014) are about cases where one
actually knows that the memory is co
--- Comment #3 from ubizjak at gmail dot com 2007-03-02 11:03 ---
Patch at http://gcc.gnu.org/ml/gcc-patches/2007-03/msg00115.html
--
ubizjak at gmail dot com changed:
What|Removed |Added
--- Comment #4 from uros at gcc dot gnu dot org 2007-03-02 12:27 ---
Subject: Bug 31019
Author: uros
Date: Fri Mar 2 12:26:55 2007
New Revision: 122473
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=122473
Log:
PR target/31019
* config/i386/i386.h (TUNEMASK): Re
reduced testcase:
real*8 function f(x)
t1 = g(0)
if(x .eq. 0) then
f = 0
else if(x .eq. 1) then
f = t1 *log( t1 )
end if
end
dlarnd.f90: In function 'f':
dlarnd.f90:3: internal compiler error: in insert_save, at caller-save.c:731
--
Summary: [dataflow] Crash in caller-save.c due
--- Comment #4 from manu at gcc dot gnu dot org 2007-03-02 14:24 ---
(In reply to comment #3)
> Does the patch need some work to remain up to date ?
The patch is from 2005, so I guess it is outdated. You would need to examine
what are the conflicts and try to fix them. I find easier to
--- Comment #5 from bernd dot speiser at uni-tuebingen dot de 2007-03-02
14:32 ---
I tried to resend the attachment (*.ii file) in compressed form as an e-mail
attachment to gcc-bugs@gcc.gnu.org, after the uncompressed attachment was
supposedly also too big and the earlier mail didn't g
--- Comment #6 from rguenth at gcc dot gnu dot org 2007-03-02 14:38 ---
Looks like e-mails with attachments are dropped. Please try attaching it via
bugzilla to the bug again (compressed). Thanks!
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31015
--- Comment #5 from ubizjak at gmail dot com 2007-03-02 14:51 ---
Patch committed, with somehow smaller code-size saves as per
http://gcc.gnu.org/ml/gcc-patches/2007-03/msg00122.html.
--
ubizjak at gmail dot com changed:
What|Removed |Added
---
--- Comment #5 from ubizjak at gmail dot com 2007-03-02 14:53 ---
Fixed in mainline.
--
ubizjak at gmail dot com changed:
What|Removed |Added
Status|ASSIGNED
--- Comment #5 from ubizjak at gmail dot com 2007-03-02 14:54 ---
Fixed in mainline.
--
ubizjak at gmail dot com changed:
What|Removed |Added
Status|ASSIGNED
--- Comment #7 from bernd dot speiser at uni-tuebingen dot de 2007-03-02
15:02 ---
Created an attachment (id=13133)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13133&action=view)
preprocessed source (compressed)
next try to attach ...
--
http://gcc.gnu.org/bugzilla/show_bu
--- Comment #6 from ubizjak at gmail dot com 2007-03-02 15:02 ---
Fixed in mainline. IMO this is not worth to fix on branches due to comment #5.
--
ubizjak at gmail dot com changed:
What|Removed |Added
--
--- Comment #4 from ubizjak at gmail dot com 2007-03-02 15:09 ---
Still fails on 4.3.0 mainline.
IMO it would be OK if 'make install' exited with a message that 'make all'
should be run before 'make install' instead of uninformative error about
something called fixincludes.
--
http
--- Comment #8 from rguenth at gcc dot gnu dot org 2007-03-02 15:16 ---
That worked. Dup of PR29512. We might consider a backport...
*** This bug has been marked as a duplicate of 29512 ***
--
rguenth at gcc dot gnu dot org changed:
What|Removed
--- Comment #16 from rguenth at gcc dot gnu dot org 2007-03-02 15:16
---
*** Bug 31015 has been marked as a duplicate of this bug. ***
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
-
--- Comment #2 from ubizjak at gmail dot com 2007-03-02 15:16 ---
Closed as WORKSFORME as RH 8.0 is kind of obsolete (I don't have this OS
anymore).
--
ubizjak at gmail dot com changed:
What|Removed |Added
--
--- Comment #13 from ubizjak at gmail dot com 2007-03-02 15:34 ---
Any news about this problem?
Current mainline still has severe problems:
-msse3 -O2 -mfpmath=sse -ffast-math
GCC 4.3 -ffast-math double performance:
ALGORITHM NB REPSTIME MFLOPS
= = =
--- Comment #7 from ubizjak at gmail dot com 2007-03-02 15:37 ---
(In reply to comment #6)
> This looks like a straightforward fix to build_common_tree_nodes2.
Is it possible to provide a patch for this? This warning is generated during
povray compilation, and povray is part of SPEC2006
The invalid code (missing '{' in second line)
template class A{
template void B() (static_cast (0)); }
};
creates an ICE with gcc-4.1.2:
crash.cc:2: internal compiler error: in grokfield, at cp/decl2.c:896
--
Summary: ICE on invalid code
Product: gcc
Versio
--- Comment #1 from axel-freyn at gmx dot de 2007-03-02 15:47 ---
Created an attachment (id=13134)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13134&action=view)
testcase
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31026
--- Comment #1 from burnus at gcc dot gnu dot org 2007-03-02 16:14 ---
Another thing which currently fails in gfortran (and g95) is:
module x
implicit none
integer, parameter :: d=8
interface
real(d) function y()
import
end functio
--- Comment #2 from burnus at gcc dot gnu dot org 2007-03-02 16:16 ---
(Example for the latter is: http://users.erols.com/dnagle/pub/pthread.f03,
which also needs ISO_C_BINDING)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30922
--- Comment #18 from paolo at gcc dot gnu dot org 2007-03-02 16:21 ---
Subject: Bug 28080
Author: paolo
Date: Fri Mar 2 16:21:30 2007
New Revision: 122478
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=122478
Log:
2007-03-02 Paolo Carlini <[EMAIL PROTECTED]>
* testsu
--- Comment #4 from rguenth at gcc dot gnu dot org 2007-03-02 16:23 ---
VRP can do this on the mainline. Fixed. (I'll add a testcase)
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
-
--- Comment #34 from michael dot klein at fazi dot de 2007-03-02 16:25
---
Could the be related to
http://sources.redhat.com/ml/binutils/2004-08/msg00187.html?
I'm getting those linker errors with both gcc 4.1.1 and gcc 4.1.2 on
x86-Solaris with GNU ld from binutils-070227, but only wh
Compiler segfaults in simple virtual inheritance situation. If inheritance is
changed to non-virtual then code compiles correctly. Also code compiles
correctly if seemingly irrelevant bits of code are removed.
No headers are required for the code to trigger the bug and none are provided.
--
There are a lot of feature test macros in the i386/x86_64 compiler of the form:
(x86_some_var & (1 << ix86_arch))
These tests could be micro-optimized, either by storing 1 << ix86_arch into a
global variable, or by having a global variable that is the result of the and
and the shift, so that a sim
--- Comment #1 from v dot lesk at ic dot ac dot uk 2007-03-02 17:02 ---
//Couldnt include attachment so attaching (short) code here. See next comment
for compiler output.
class A
{
private:
protected:
public:
};
template
class C:
virtual public A
{
private:
protected:
public:
C()
--- Comment #2 from v dot lesk at ic dot ac dot uk 2007-03-02 17:03 ---
Compiler command and output for this bug :
g++ -v -save-temps -c -O0 ./source.C
Using built-in specs.
Target: i586-mandriva-linux-gnu
Configured with: ../configure --prefix=/usr --libexecdir=/usr/lib
--with-slibdir=/
--- Comment #14 from pinskia at gcc dot gnu dot org 2007-03-02 17:42
---
> Please consider the fact that all benchmarks are using -ffast-math nowadays.
> ;)
Please also consider the fact that the register allocator has been broken since
20 years ago :) :) :) :).
And I repeat again, t
--- Comment #1 from ubizjak at gmail dot com 2007-03-02 17:45 ---
*** This bug has been marked as a duplicate of 31019 ***
--
ubizjak at gmail dot com changed:
What|Removed |Added
--
--- Comment #6 from ubizjak at gmail dot com 2007-03-02 17:45 ---
*** Bug 31028 has been marked as a duplicate of this bug. ***
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31019
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Severity|blocker |normal
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31027
--- Comment #5 from jv244 at cam dot ac dot uk 2007-03-02 18:15 ---
> >
> > grid_fast.F:483: note: not vectorized: can't determine dependence between
> > (*coef_447)[D.1967_2320] and (*coef_447)[D.1967_2320]
> > DO icoef=1,coef_max
> > coef(icoef,1)=coef(i
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Target Milestone|--- |4.3.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30970
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Target Milestone|--- |4.3.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30413
--- Comment #1 from geoffk at gcc dot gnu dot org 2007-03-02 18:59 ---
This is because gccspec.c is adding -shared-libgcc when it sees an Objective-C
file, but inconsistently; it doesn't handle -x options, it doesn't handle the
-objC flag. On Darwin, it's wrong to add -shared-libgcc lik
the following is tiny missed optimization, as it fails to link at -O3
read(5,*) igmin
DO ig=igmin,0
ig2=1-ig
if (ig.EQ.ig2) CALL link_error()
ENDDO
END
since ig can only have values in the interval [-huge,0], ig2 can only have
values in [1,huge] and so ig can never be equal to ig2. Code sim
--- Comment #35 from hjl at lucon dot org 2007-03-02 19:43 ---
(In reply to comment #34)
> Could the be related to
> http://sources.redhat.com/ml/binutils/2004-08/msg00187.html?
>
> I'm getting those linker errors with both gcc 4.1.1 and gcc 4.1.2 on
> x86-Solaris with GNU ld from binut
--- Comment #1 from pinskia at gcc dot gnu dot org 2007-03-02 20:04 ---
*** This bug has been marked as a duplicate of 30965 ***
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
--
--- Comment #1 from pinskia at gcc dot gnu dot org 2007-03-02 20:04 ---
*** Bug 31029 has been marked as a duplicate of this bug. ***
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
---
--- Comment #6 from levengcc at pleetveern dot com 2007-03-02 21:58 ---
(In reply to comment #4)
> Patch posted here:
> http://gcc.gnu.org/ml/gcc-patches/2006-03/msg01197.html
>
This patch would meet my needs if the -I- option is removed from gcc. Without
it
or the -I- option, I would
--- Comment #4 from simartin at gcc dot gnu dot org 2007-03-02 22:09
---
Subject: Bug 28253
Author: simartin
Date: Fri Mar 2 22:09:20 2007
New Revision: 122490
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=122490
Log:
2007-03-02 Simon Martin <[EMAIL PROTECTED]>
PR
--- Comment #5 from simartin at gcc dot gnu dot org 2007-03-02 22:14
---
Fixed on the mainline
--
simartin at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #8 from pbrook at gcc dot gnu dot org 2007-03-02 22:51 ---
Fixed.
trunk r122437 http://gcc.gnu.org/ml/gcc-cvs/2007-03/msg00020.html
4.2 r122489 http://gcc.gnu.org/ml/gcc-cvs/2007-03/msg00072.html
--
pbrook at gcc dot gnu dot org changed:
What|Removed
--
burnus at gcc dot gnu dot org changed:
What|Removed |Added
AssignedTo|unassigned at gcc dot gnu |burnus at gcc dot gnu dot
|dot org
--- Comment #4 from burnus at gcc dot gnu dot org 2007-03-02 23:03 ---
Subject: Bug 30873
Author: burnus
Date: Fri Mar 2 23:03:26 2007
New Revision: 122495
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=122495
Log:
2007-03-02 Paul Thomas <[EMAIL PROTECTED]>
Tobias
--- Comment #12 from skunk at iskunk dot org 2007-03-02 23:36 ---
Here's my minimal test case. Compile with "-O3 -Wall -c":
#include
void frob(int *pi);
int main(void)
{
int i;
printf("i = %d\n", i);
frob(&i);
return 0;
}
No warning from 4.0.3 nor 4.1.2
--
--- Comment #10 from rakdver at gcc dot gnu dot org 2007-03-03 00:39
---
Subject: Bug 29902
Author: rakdver
Date: Sat Mar 3 00:38:56 2007
New Revision: 122501
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=122501
Log:
PR tree-optimization/29902
* tree-ssa-loop-
--- Comment #11 from rakdver at gcc dot gnu dot org 2007-03-03 00:41
---
> Can this patch be backported to 4.2?
Done.
--
rakdver at gcc dot gnu dot org changed:
What|Removed |Added
-
--- Comment #5 from patchapp at dberlin dot org 2007-03-03 03:01 ---
Subject: Bug number PR30873
A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2007-03/msg00110.html
--
http://gcc.gnu.org/bugzilla/sh
--- Comment #3 from patchapp at dberlin dot org 2007-03-03 03:01 ---
Subject: Bug number PR30882
A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2007-03/msg00111.html
--
http://gcc.gnu.org/bugzilla/sh
70 matches
Mail list logo
