https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91257

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
                 CC|                            |rguenth at gcc dot gnu.org
           Assignee|rguenth at gcc dot gnu.org         |unassigned at gcc dot 
gnu.org

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
Otherwise I'm done here.  Updated numbers for trunk (release-checking,
non-bootstrapped build with GCC 4.8 host compiler) on the original testcase

                               time   peak rss
-O3 -fwrapv -m32               112s   4.7GB
-O2 -fwrapv -m32               111s   4.7GB
-O2 -fwrapv -m32 -fno-ipa-icf  301s   4.7GB
-O1 -fwrapv -m32               204s   4.5GB
-O1                            202s   4.5GB

and on the testcase with just gen_opt (the original has gen_opt{,1,2,3,4,5})

-O3 -fwrapv -m32                50s   1.7GB
-O2 -fwrapv -m32                50s   1.7GB
-O1                             34s   1.4GB

so it's clear that ICF simply shrinks the testcase by identifying all
functions as 1:1 identical for the original testcase with -O2+.

Numbers show there's really no IPA component to the testcase then
and looking at a reduced one is fully sufficicent.

Time report for plain -O2:

Time variable                                   usr           sys          wall
              GGC
 tree VRP                           :   5.88 ( 12%)   0.24 ( 25%)   6.10 ( 12%)
    598 kB (  0%)
 tree Early VRP                     :   0.81 (  2%)   0.00 (  0%)   0.81 (  2%)
    152 kB (  0%)
 tree copy propagation              :   1.42 (  3%)   0.00 (  0%)   1.42 (  3%)
      0 kB (  0%)
 tree PTA                           :   7.77 ( 16%)   0.25 ( 26%)   8.02 ( 16%)
   7755 kB (  1%)
 tree SSA rewrite                   :   1.74 (  3%)   0.00 (  0%)   1.75 (  3%)
    629 kB (  0%)
 dominator optimization             :   2.29 (  5%)   0.00 (  0%)   2.32 (  5%)
    707 kB (  0%)
 tree CCP                           :   3.57 (  7%)   0.04 (  4%)   3.61 (  7%)
     63 kB (  0%)
 tree PRE                           :   1.69 (  3%)   0.02 (  2%)   1.71 (  3%)
    357 kB (  0%)
 tree FRE                           :   4.15 (  8%)   0.00 (  0%)   4.19 (  8%)
    152 kB (  0%)
 tree backward propagate            :   1.04 (  2%)   0.02 (  2%)   1.06 (  2%)
      0 kB (  0%)
 tree forward propagate             :   1.29 (  3%)   0.00 (  0%)   1.31 (  3%)
      0 kB (  0%)
 out of ssa                         :   9.27 ( 19%)   0.00 (  0%)   9.28 ( 18%)
     47 kB (  0%)
 TOTAL                              :  49.80          0.97         50.79       
 747309 kB


Time report for plain -O1:

Time variable                                   usr           sys          wall
              GGC
 tree copy propagation              :   1.40 (  4%)   0.00 (  0%)   1.41 (  4%)
      0 kB (  0%)
 tree PTA                           :   5.78 ( 17%)   0.19 ( 32%)   5.97 ( 18%)
   7679 kB (  1%)
 tree SSA rewrite                   :   1.63 (  5%)   0.00 (  0%)   1.63 (  5%)
    629 kB (  0%)
 dominator optimization             :   2.29 (  7%)   0.01 (  2%)   2.30 (  7%)
    694 kB (  0%)
 tree CCP                           :   3.53 ( 11%)   0.07 ( 12%)   3.60 ( 11%)
     63 kB (  0%)
 tree FRE                           :   2.74 (  8%)   0.02 (  3%)   2.75 (  8%)
    101 kB (  0%)
 tree backward propagate            :   1.04 (  3%)   0.01 (  2%)   1.05 (  3%)
      0 kB (  0%)
 tree forward propagate             :   1.29 (  4%)   0.00 (  0%)   1.29 (  4%)
      0 kB (  0%)
 out of ssa                         :   6.22 ( 19%)   0.00 (  0%)   6.22 ( 18%)
     55 kB (  0%)
 TOTAL                              :  33.38          0.59         33.98       
 744713 kB

Reply via email to