http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48702

--- Comment #25 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-05-26 
10:01:33 UTC ---
I have benchmarked SPEC2k6 on the 4.6 branch with the patch (and required
backports) with -O3 -ffast-math -funroll-loops, Base is unpatched, Peak
is patched.  Average of three runs on a Intel(R) Core(TM) i7 CPU X 980 @
3.33GHz, the noise in the tests is +-1s.  So from a performance point
the patch doesn't seem to have problems.

                                  Estimated                       Estimated
                Base     Base       Base        Peak     Peak       Peak
Benchmarks      Ref.   Run Time     Ratio       Ref.   Run Time     Ratio
-------------- ------  ---------  ---------    ------  ---------  ---------
400.perlbench    9770        334       29.2 *    9770        337       29.0 *  
401.bzip2        9650        467       20.7 *    9650        459       21.0 *  
403.gcc          8050        289       27.9 *    8050        289       27.8 *  
429.mcf          9120        302       30.2 *    9120        304       30.0 *  
445.gobmk       10490        413       25.4 *   10490        413       25.4 *  
456.hmmer        9330        375       24.9 *    9330        374       24.9 *  
458.sjeng       12100        446       27.1 *   12100        446       27.1 *  
462.libquantum  20720        332       62.4 *   20720        332       62.4 *  
464.h264ref     22130        540       41.0 *   22130        540       41.0 *  
471.omnetpp      6250        300       20.9 *    6250        305       20.5 *  
473.astar        7020        373       18.8 *    7020        373       18.8 *  
483.xalancbmk    6900        225       30.7 *    6900        226       30.5 *  
 Est. SPECint_base2006                 28.3
 Est. SPECint2006                                                      28.3

                                  Estimated                       Estimated
                Base     Base       Base        Peak     Peak       Peak
Benchmarks      Ref.   Run Time     Ratio       Ref.   Run Time     Ratio
-------------- ------  ---------  ---------    ------  ---------  ---------
410.bwaves      13590        347       39.2 *   13590        347       39.2 *  
416.gamess      19580        740       26.5 *   19580        738       26.5 *  
433.milc         9180        382       24.0 *    9180        381       24.1 *  
434.zeusmp       9100        336       27.1 *    9100        336       27.1 *  
435.gromacs      7140        318       22.4 *    7140        318       22.4 *  
436.cactusADM   11950        422       28.3 *   11950        426       28.1 *  
437.leslie3d     9400        316       29.8 *    9400        315       29.9 *  
444.namd         8020        397       20.2 *    8020        398       20.2 *  
447.dealII      11440        315       36.3 *   11440        314       36.5 *  
450.soplex       8340        220       37.9 *    8340        220       37.9 *  
453.povray       5320        186       28.6 *    5320        187       28.4 *  
454.calculix     8250        365       22.6 *    8250        365       22.6 *  
459.GemsFDTD    10610        344       30.9 *   10610        340       31.2 *  
465.tonto        9840        380       25.9 *    9840        378       26.0 *  
470.lbm         13740        254       54.0 *   13740        255       53.9 *  
481.wrf         11170        351       31.8 *   11170        352       31.8 *  
482.sphinx3     19490        483       40.4 *   19490        485       40.2 *  
 Est. SPECfp_base2006                  30.0
 Est. SPECfp2006                                                       30.0

Reply via email to