Re: C++ inlining heuristics changed (3.0.1)

[Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>]

As promised, updated with data for your second patch.

First the build benchmarks, then run-time benchmarks:

  -O3		GCC 2.95.3  GCC 3.0   GCC 3.0.1  patch-1   patch-2

  Compile-time	6:00        32:15     7:46       6:41      7:21
  Binary size	4377664     6213988   4160356    3948196   4029092

Summary: With both patches, my binaries are smaller than ever before, and
compile-time is below 3.0.1 and even approaching the one with 2.95.3; the
second patch is more expensive in terms of compile-time and binary size.

      STRATCOMP1-ALL |   5.68  |  99.41  |  98.18  | 101.35  |  98.23
   STRATCOMP-770.2-Q |   1.28  |   1.46  |   1.99  |   1.45  |   1.44
               2QBF1 |  24.34  |  40.88  |  37.64  |  30.51  |  30.47
          PRIMEIMPL2 |  19.26  |  28.22  |  38.08  |  21.71  |  20.18
            ANCESTOR |  12.65  |  11.66  |  17.32  |  12.85  |  13.09
       3COL-SIMPLEX1 |  12.07  |  11.21  |  16.79  |  12.93  |  12.54
        3COL-LADDER1 |  53.49  |  76.91  |  84.97  |  65.03  |  63.42
      3COL-N-LADDER1 |  23.73  |  27.67  |  41.06  |  24.48  |  23.07
        3COL-RANDOM1 |  18.61  |  26.30  |  36.56  |  21.52  |  20.13
          HP-RANDOM1 |  14.19  |  15.05  |  20.58  |  15.95  |  15.66
       HAMCYCLE-FREE |   2.06  |   2.47  |   3.85  |   2.89  |   2.79
             DECOMP2 |  31.70  |  32.45  |  51.61  |  30.85  |  31.03
        BW-P4-Esra-a | 106.77  | 125.14  | 167.70  | 116.59  | 112.88
        BW-P5-nopush |   8.41  |   9.68  |  13.06  |   8.93  |   8.86
       BW-P5-pushbin |   7.72  |   9.03  |  12.91  |   8.30  |   8.01
     BW-P5-nopushbin |   2.70  |   3.08  |   4.38  |   2.83  |   2.82
              3SAT-1 |  59.35  |  85.68  | 115.94  |  66.39  |  59.89
   3SAT-1-CONSTRAINT |  31.12  |  48.77  |  68.35  |  36.69  |  34.94
        HANOI-Towers |   6.10  |   6.56  |   9.46  |   5.89  |   5.94
              RAMSEY |  13.63  |  15.00  |  21.84  |  14.41  |  14.66
             CRISTAL |  14.88  |  15.26  |  22.10  |  17.11  |  16.81
             HANOI-K |  59.49  |  80.59  | 121.81  |  66.69  |  63.78
           21-QUEENS |  18.48  |  25.02  |  38.21  |  19.85  |  18.72
   MSTDir[V=13,A=40] |  36.38  |  36.07  |  47.96  |  34.20  |  34.25
   MSTDir[V=15,A=40] |  37.24  |  35.80  |  47.81  |  34.23  |  34.52
 MSTUndir[V=13,A=40] |  19.07  |  19.30  |  26.65  |  18.63  |  18.50
 MSTUndir[V=15,A=40] | 311.58  | 310.65  | 423.87  | 300.44  | 303.05
         TIMETABLING |  23.98  |  25.47  |  37.90  |  25.28  |  24.41

Summary: Still worse than 2.95.3 in most cases, but significantly better
than 3.0.1 (which in turn is quite a bit worse when compared to 3.0). The
second patch is an improvement over the first one.

I'm still waiting for approval to commit the first patch (which doesn't
require the paperwork) on both branches.

Gerald
-- 
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/