This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFT] fwprop merge


On Sun, 14 May 2006, Steven Bosscher wrote:

> On Sunday 14 May 2006 02:13, Steven Bosscher wrote:
> > So this is the first part of the changes.  The second bits that clean
> > up cse.c are unchanged.  I have only tested this with a C bootstrap on
> > x86-linux, but I'll give it a full run tomorrow and if it passes, I'll
> > post it as an RFT.
> 
> So this is the fully tested version.
> Bootstrapped and tested on x86_64-suse-linux-gnu, multilib, all languages
> except treelang.
> 
> David, Richard, could you give these patches a re-try on SPEC please?

I have done SPEC2k testing on x86_64 with -O2 and on i686 (x86_64 -m32).
Full results attached, the summaries look like

x86_64:

   164.gzip          1400   165       846    *     1400   159       881    
*
   175.vpr           1400   157       893    *     1400   157       891    
*
   176.gcc                                   X                             
X
   181.mcf           1800   321       561    *     1800   323       557    
*
   186.crafty        1000    64.5    1551    *     1000    66.2    1511    
*
   197.parser        1800   259       696    *     1800   258       698    
*
   252.eon           1300    89.1    1458    *     1300    90.1    1442    
*
   253.perlbmk       1800   160      1122    *     1800   158      1140    
*
   254.gap           1100   117       939    *     1100   115       957    
*
   255.vortex                                X                             
X
   256.bzip2         1500   163       920    *     1500   163       918    
*
   300.twolf         3000   295      1019    *     3000   294      1021    
*

   168.wupwise       1600   150      1064    *     1600       151      
1059*
   171.swim          3100   382       811    *     3100       382       
811*
   172.mgrid         1800   246       732    *     1800       248       
725*
   173.applu         2100   265       793    *     2100       264       
795*
   177.mesa          1400   117      1193    *     1400       119      
1179*
   178.galgel                                     2900       194      
1493*
   179.art           2600   246      1058    *     2600       251      
1036*
   183.equake        1300   123      1056    *     1300       123      
1055*
   187.facerec       1900   227       837    *     1900       227       
835*
   188.ammp          2200   228       965    *     2200       228       
963*
   189.lucas         2000   183      1093    *     2000       183      
1095*
   191.fma3d         2100   258       813    *     2100       257       
816*
   200.sixtrack      1100   258       426    *     1100       258       
426*
   301.apsi          2600   279       932    *     2600       279       
931*


i686:

   164.gzip          1400   179       782    *     1400   181       772    
*
   175.vpr           1400   170       824    *     1400   170       822    
*
   176.gcc                                   X                             
X
   181.mcf           1800   286       630    *     1800   286       630    
*
   186.crafty        1000    97.0    1031    *     1000    97.2    1029    
*
   197.parser        1800   253       713    *     1800   254       707    
*
   252.eon                                   X                             
X
   253.perlbmk       1800   164      1097    *     1800   165      1091    
*
   254.gap           1100   129       853    *     1100   130       848    
*
   255.vortex        1900   162      1175    *     1900   159      1195    
*
   256.bzip2         1500   203       738    *     1500   199       753    
*
   300.twolf         3000   269      1114    *     3000   268      1118    
*

   168.wupwise       1600       179       892*     1600       177       
902*
   171.swim          3100       485       639*     3100       485       
639*
   172.mgrid         1800       369       488*     1800       368       
489*
   173.applu         2100       355       592*     2100       355       
592*
   177.mesa          1400       156       900*     1400       162       
866*
   178.galgel        2900       220      1315*     2900       224      
1296*
   179.art           2600       641       405*     2600       627       
414*
   183.equake        1300       155       836*     1300       154       
846*
   187.facerec       1900       249       762*     1900       250       
761*
   188.ammp          2200       282       779*     2200       282       
779*
   189.lucas         2000       279       717*     2000       279       
717*
   191.fma3d         2100       295       713*     2100       294       
715*
   200.sixtrack      1100       219       502*     1100       218       
504*
   301.apsi          2600       342       760*     2600       341       
762*


I also patched in for the May14 run of 
http://www.suse.de/~rguenther/c++bench and it didn't show any effects
on compile-time, memory-usage or performance.

Richard.

Attachment: fwprop_x86_64-O2.tgz
Description: application/compressed-tar

Attachment: fwprop_i686-O2.tgz
Description: application/compressed-tar


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]