This is the mail archive of the gcc@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]

Re: [Patch] C++ inlining heuristics changed (3.0.1)


On Wed, 22 Aug 2001, Kurt Garloff wrote:
> Gerald, could you test as well?

Done. And, I'm extremely positively surprised and impressed. First the
build benchmarks, then run-time benchmarks:

  -O3		GCC 2.95.3  GCC 3.0   GCC 3.0.1	3.0.1+patch

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

Summary: With your patch, my binaries are smaller than ever before, and
compile-time is below 3.0.1 and even approaching the one with 2.95.3

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

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).


I *strongly* opt for adding Kurt's first patch which is incredibly simple
to both branches as soon as possible (and then wait for his copyright
assignment etc. before we can consider his more complicated patches).

(IMNSHO we also need this patch so that we can assess the true quality of
further patches more realistically and set a benchmark which cannot be
beaten as easily as current GCC 3.0.x and mainline.)

Okay to install his patch on both branches?

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


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