Re: proposal to add -Wheader-guard option
On Tue, Feb 4, 2014 at 12:45 PM, Markus Trippelsdorf wrote: > On 2014.02.04 at 12:36 +0530, Prathamesh Kulkarni wrote: >> Ping ? > > Patches should be posted to the gcc-patc...@gcc.gnu.org list. > Please follow up there. Redirected to gcc-patc...@gcc.gnu.org list. I apologize for posting here. > > -- > Markus
FW: Git repo lagging behing
Since the owner of the repo http://gcc.gnu.org/git/?p=gcc.git;a=summary is marked as being: gcc@gcc.gnu.org moving this thread of the gcc mailing list. Paulo Matos -Original Message- From: gcc-help-ow...@gcc.gnu.org [mailto:gcc-help-ow...@gcc.gnu.org] On Behalf Of Paulo Matos Sent: 04 February 2014 14:08 To: gcc-h...@gcc.gnu.org Subject: Git repo lagging behing Hi, Don't know if anybody noticed but something's wrong with the git repo. It's lagging behind. Last commit there was: 18 hours agotejohnson 2014-02-03 Teresa Johnson which was certainly not the last commit on trunk. Paulo Matos
Merge epilog loop & loop version due to alias/alignment in vectorization?
Hi,
One of biggest issues we have with GCC vectorization is bloated code size.
For example, vectorized version is 2.5 times of non-vectorized one for the
following simple code. One reason is that GCC often creates one loop copy
because of aliasing/alignment and one epilog loop because of loop iteration
constraint.
void foo (int *a, int *b, int N)
{
int i;
for (i = 0; i < N; i++)
{
a[i] = b[i];
}
}
Looking closely, the epilog loop and alignement/aliasing loop are almost
identical, just different in initial values for some variables entering
the loop. Can they be merged into one in such situations? If yes, any
suggestion on how to implement it?
...
:
# i_39 = PHI
_41 = (long unsigned int) i_39;
_42 = _41 * 4;
_43 = a_7(D) + _42;
_44 = b_9(D) + _42;
_45 = *_44;
*_43 = _45;
i_47 = i_39 + 1;
if (N_4(D) > i_47)
goto ;
else
goto ;
:
goto ;
:
# i_51 = PHI
tmp.6_56 = (int) ratio_mult_vf.5_38;
if (niters.3_34 == ratio_mult_vf.5_38)
goto ;
else
goto ;
:
# i_50 = PHI
goto ;
:
goto ;
:
:
# i_24 = PHI <0(12), i_32(14)>
_26 = (long unsigned int) i_24;
_27 = _26 * 4;
_28 = a_7(D) + _27;
_29 = b_9(D) + _27;
_30 = *_29;
*_28 = _30;
i_32 = i_24 + 1;
if (N_4(D) > i_32)
goto ;
else
goto ;
...
Thanks,
Bingfeng
Re: Merge epilog loop & loop version due to alias/alignment in vectorization?
See also http://gcc.gnu.org/ml/gcc/2013-08/msg00259.html
There are some concerns, but it would be interesting to do some
benchmarking of this.
David
On Tue, Feb 4, 2014 at 8:27 AM, Bingfeng Mei wrote:
> Hi,
> One of biggest issues we have with GCC vectorization is bloated code size.
> For example, vectorized version is 2.5 times of non-vectorized one for the
> following simple code. One reason is that GCC often creates one loop copy
> because of aliasing/alignment and one epilog loop because of loop iteration
> constraint.
>
> void foo (int *a, int *b, int N)
> {
> int i;
> for (i = 0; i < N; i++)
> {
> a[i] = b[i];
> }
> }
>
> Looking closely, the epilog loop and alignement/aliasing loop are almost
> identical, just different in initial values for some variables entering
> the loop. Can they be merged into one in such situations? If yes, any
> suggestion on how to implement it?
>
> ...
> :
> # i_39 = PHI
> _41 = (long unsigned int) i_39;
> _42 = _41 * 4;
> _43 = a_7(D) + _42;
> _44 = b_9(D) + _42;
> _45 = *_44;
> *_43 = _45;
> i_47 = i_39 + 1;
> if (N_4(D) > i_47)
> goto ;
> else
> goto ;
>
> :
> goto ;
>
> :
> # i_51 = PHI
> tmp.6_56 = (int) ratio_mult_vf.5_38;
> if (niters.3_34 == ratio_mult_vf.5_38)
> goto ;
> else
> goto ;
>
> :
> # i_50 = PHI
> goto ;
>
> :
> goto ;
>
> :
>
> :
> # i_24 = PHI <0(12), i_32(14)>
> _26 = (long unsigned int) i_24;
> _27 = _26 * 4;
> _28 = a_7(D) + _27;
> _29 = b_9(D) + _27;
> _30 = *_29;
> *_28 = _30;
> i_32 = i_24 + 1;
> if (N_4(D) > i_32)
> goto ;
> else
> goto ;
> ...
>
> Thanks,
> Bingfeng
Re: FW: Git repo lagging behing
On Tue, Feb 04, 2014 at 03:59:45PM +, Paulo Matos wrote: > Don't know if anybody noticed but something's wrong with the git repo. It's > lagging behind. Last commit there was: > 18 hours ago tejohnson 2014-02-03 Teresa Johnson > > which was certainly not the last commit on trunk. I believe one of the overseers I don't want to name (hi Frank) will have a look at this. Marek
Re: Draft C bindings for IEEE 754-2008 part 4 now available
Revised versions of parts 3 and 4 are now available: http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1789.pdf http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1790.pdf (note that these are not yet expected to address all comments on the previous versions; see also N1791 and N1792, with minutes of the last floating-point group meeting and agenda for the next). Again, any comments are best sent direct to WG14. -- Joseph S. Myers jos...@codesourcery.com
question about gtype-desc.c
Hello Jakub, I think I have found a fix to _Cilk_for to be structured the way you requested. I am currently trying to clean up my code so that I can send you a patch. Now, I am trying to remove the 2 fields I put in gimple_omp_for_iter: loop_count and grain. But, it is giving the following errors when I remove it. Do you know what is causing it and how I can fix it? gtype-desc.c: In function âvoid gt_ggc_mx_gimple_statement_base(void*)â: gtype-desc.c:1721: error: âstruct gimple_omp_for_iterâ has no member named âloop_countâ gtype-desc.c:1721: error: âstruct gimple_omp_for_iterâ has no member named âloop_countâ gtype-desc.c:1722: error: âstruct gimple_omp_for_iterâ has no member named âgrainâ gtype-desc.c:1722: error: âstruct gimple_omp_for_iterâ has no member named âgrainâ gtype-desc.c: In function âvoid gt_pch_nx_gimple_statement_base(void*)â: gtype-desc.c:4854: error: âstruct gimple_omp_for_iterâ has no member named âloop_countâ gtype-desc.c:4854: error: âstruct gimple_omp_for_iterâ has no member named âloop_countâ gtype-desc.c:4855: error: âstruct gimple_omp_for_iterâ has no member named âgrainâ gtype-desc.c:4855: error: âstruct gimple_omp_for_iterâ has no member named âgrainâ gtype-desc.c: In function âvoid gt_pch_p_21gimple_statement_base(void*, void*, void (*)(void*, void*), void*)â: gtype-desc.c:8176: error: âstruct gimple_omp_for_iterâ has no member named âloop_countâ gtype-desc.c:8178: error: âstruct gimple_omp_for_iterâ has no member named âgrainâ These files are in the build directory. I deleted everything and reconfigured but the error reappears. Thanks, Balaji V. Iyer.
Re: question about gtype-desc.c
On Tue, Feb 04, 2014 at 10:12:13PM +, Iyer, Balaji V wrote: > Hello Jakub, > I think I have found a fix to _Cilk_for to be structured the way you > requested. I am currently trying to clean up my code so that I can send you a > patch. > > Now, I am trying to remove the 2 fields I put in gimple_omp_for_iter: > loop_count and grain. But, it is giving the following errors when I remove it. > > Do you know what is causing it and how I can fix it? > > gtype-desc.c: In function âvoid gt_ggc_mx_gimple_statement_base(void*)â: > gtype-desc.c:1721: error: âstruct gimple_omp_for_iterâ has no member named > âloop_countâ > gtype-desc.c:1721: error: âstruct gimple_omp_for_iterâ has no member named > âloop_countâ > gtype-desc.c:1722: error: âstruct gimple_omp_for_iterâ has no member named > âgrainâ > gtype-desc.c:1722: error: âstruct gimple_omp_for_iterâ has no member named > âgrainâ > gtype-desc.c: In function âvoid gt_pch_nx_gimple_statement_base(void*)â: > gtype-desc.c:4854: error: âstruct gimple_omp_for_iterâ has no member named > âloop_countâ > gtype-desc.c:4854: error: âstruct gimple_omp_for_iterâ has no member named > âloop_countâ > gtype-desc.c:4855: error: âstruct gimple_omp_for_iterâ has no member named > âgrainâ > gtype-desc.c:4855: error: âstruct gimple_omp_for_iterâ has no member named > âgrainâ > gtype-desc.c: In function âvoid gt_pch_p_21gimple_statement_base(void*, > void*, void (*)(void*, void*), void*)â: > gtype-desc.c:8176: error: âstruct gimple_omp_for_iterâ has no member named > âloop_countâ > gtype-desc.c:8178: error: âstruct gimple_omp_for_iterâ has no member named > âgrainâ > > These files are in the build directory. I deleted everything and reconfigured > but the error reappears. You mean you've started a new build directory, or just removed *.o or something similar? I'd suggest to rm s-gtype gtype-desc.c and make s-gtype. Jakub
RE: question about gtype-desc.c
> -Original Message- > From: gcc-ow...@gcc.gnu.org [mailto:gcc-ow...@gcc.gnu.org] On Behalf Of > Jakub Jelinek > Sent: Tuesday, February 4, 2014 5:23 PM > To: Iyer, Balaji V > Cc: gcc@gcc.gnu.org > Subject: Re: question about gtype-desc.c > > On Tue, Feb 04, 2014 at 10:12:13PM +, Iyer, Balaji V wrote: > > Hello Jakub, > > I think I have found a fix to _Cilk_for to be structured the way you > requested. I am currently trying to clean up my code so that I can send you a > patch. > > > > Now, I am trying to remove the 2 fields I put in gimple_omp_for_iter: > loop_count and grain. But, it is giving the following errors when I remove it. > > > > Do you know what is causing it and how I can fix it? > > > > gtype-desc.c: In function âvoid > gt_ggc_mx_gimple_statement_base(void*)â: > > gtype-desc.c:1721: error: âstruct gimple_omp_for_iterâ has no member > > named âloop_countâ > > gtype-desc.c:1721: error: âstruct gimple_omp_for_iterâ has no member > > named âloop_countâ > > gtype-desc.c:1722: error: âstruct gimple_omp_for_iterâ has no member > > named âgrainâ > > gtype-desc.c:1722: error: âstruct gimple_omp_for_iterâ has no member > > named âgrainâ > > gtype-desc.c: In function âvoid > gt_pch_nx_gimple_statement_base(void*)â: > > gtype-desc.c:4854: error: âstruct gimple_omp_for_iterâ has no member > > named âloop_countâ > > gtype-desc.c:4854: error: âstruct gimple_omp_for_iterâ has no member > > named âloop_countâ > > gtype-desc.c:4855: error: âstruct gimple_omp_for_iterâ has no member > > named âgrainâ > > gtype-desc.c:4855: error: âstruct gimple_omp_for_iterâ has no member > > named âgrainâ > > gtype-desc.c: In function âvoid > gt_pch_p_21gimple_statement_base(void*, void*, void (*)(void*, void*), > void*)â: > > gtype-desc.c:8176: error: âstruct gimple_omp_for_iterâ has no member > > named âloop_countâ > > gtype-desc.c:8178: error: âstruct gimple_omp_for_iterâ has no member > > named âgrainâ > > > > These files are in the build directory. I deleted everything and > > reconfigured > but the error reappears. > > You mean you've started a new build directory, or just removed *.o or > something similar? I'd suggest to rm s-gtype gtype-desc.c and make s-gtype. I did the following cd build_dir rm -rf * ../gcc/configure --prefix=$PWD/../install --enable-languages="c,c++" make Should I do make s-gtype before I do "make" above? > > Jakub
Re: question about gtype-desc.c
On Tue, Feb 04, 2014 at 10:26:20PM +, Iyer, Balaji V wrote: > > > gtype-desc.c:8176: error: âstruct gimple_omp_for_iterâ has no member > > > named âloop_countâ > > > gtype-desc.c:8178: error: âstruct gimple_omp_for_iterâ has no member > > > named âgrainâ > > > > > > These files are in the build directory. I deleted everything and > > > reconfigured > > but the error reappears. > > > > You mean you've started a new build directory, or just removed *.o or > > something similar? I'd suggest to rm s-gtype gtype-desc.c and make s-gtype. > > I did the following > > cd build_dir > rm -rf * > ../gcc/configure --prefix=$PWD/../install --enable-languages="c,c++" > make > > Should I do make s-gtype before I do "make" above? What you did should be more than enough. But then, I didn't see loop_count nor grain fields of gimple_omp_for_iter in your previously posted patches (only your ChangeLog mentioned it). So, if you start with wiping up the build directory, I think you have to grep where the fields appear. Jakub
RE: question about gtype-desc.c
> -Original Message- > From: Jakub Jelinek [mailto:ja...@redhat.com] > Sent: Tuesday, February 4, 2014 5:36 PM > To: Iyer, Balaji V > Cc: gcc@gcc.gnu.org > Subject: Re: question about gtype-desc.c > > On Tue, Feb 04, 2014 at 10:26:20PM +, Iyer, Balaji V wrote: > > > > gtype-desc.c:8176: error: âstruct gimple_omp_for_iterâ has no > > > > member named âloop_countâ > > > > gtype-desc.c:8178: error: âstruct gimple_omp_for_iterâ has no > > > > member named âgrainâ > > > > > > > > These files are in the build directory. I deleted everything and > > > > reconfigured > > > but the error reappears. > > > > > > You mean you've started a new build directory, or just removed *.o > > > or something similar? I'd suggest to rm s-gtype gtype-desc.c and make s- > gtype. > > > > I did the following > > > > cd build_dir > > rm -rf * > > ../gcc/configure --prefix=$PWD/../install --enable-languages="c,c++" > > make > > > > Should I do make s-gtype before I do "make" above? > > What you did should be more than enough. But then, I didn't see > loop_count nor grain fields of gimple_omp_for_iter in your previously > posted patches (only your ChangeLog mentioned it). So, if you start with > wiping up the build directory, I think you have to grep where the fields > appear. > OK. Found the issue. I had the 2 fields enclosed inside a #if 0 and #endif I guess whatever that is reading the fields to create gtype-desc.c does not understand C/C++ pre-processing, and thus it didn't ignore the fields in between #if 0. Thanks, Balaji V. Iyer.
Re: git mirror hasn't been updated for 8 hours
Hi - There appears to have been a problem with bernie's cron-based git-svn-fetch job, due to a corrupted (emptied) .git/svn/.cache/XXX.db file. With that file removed, the cron job is running again. A little bit extra logging was added to help pinpoint the cause next time. - FChE
Re: question about gtype-desc.c
On Tue, 2014-02-04 at 23:07 +, Iyer, Balaji V wrote:
>
> > -Original Message-
> > From: Jakub Jelinek [mailto:ja...@redhat.com]
> > Sent: Tuesday, February 4, 2014 5:36 PM
> > To: Iyer, Balaji V
> > Cc: gcc@gcc.gnu.org
> > Subject: Re: question about gtype-desc.c
> >
> > On Tue, Feb 04, 2014 at 10:26:20PM +, Iyer, Balaji V wrote:
> > > > > gtype-desc.c:8176: error: âstruct gimple_omp_for_iterâ has no
> > > > > member named âloop_countâ
> > > > > gtype-desc.c:8178: error: âstruct gimple_omp_for_iterâ has no
> > > > > member named âgrainâ
> > > > >
> > > > > These files are in the build directory. I deleted everything and
> > > > > reconfigured
> > > > but the error reappears.
[...]
> > What you did should be more than enough. But then, I didn't see
> > loop_count nor grain fields of gimple_omp_for_iter in your previously
> > posted patches (only your ChangeLog mentioned it). So, if you start with
> > wiping up the build directory, I think you have to grep where the fields
> > appear.
> >
>
> OK. Found the issue. I had the 2 fields enclosed inside a #if 0 and #endif
>
> I guess whatever that is reading the fields to create gtype-desc.c does not
> understand C/C++ pre-processing, and thus it didn't ignore the fields in
> between #if 0.
[...]
Yes. gengtype, the utility creating gtype-desc.c, does not care about
the C++ preprocessing. Ideally, I believe it should, but that is hard
work.
--
Basile STARYNKEVITCH http://starynkevitch.net/Basile/
email: basilestarynkevitchnet mobile: +33 6 8501 2359
8, rue de la Faiencerie, 92340 Bourg La Reine, France
*** opinions {are only mine, sont seulement les miennes} ***
