Re: [Gcl-devel] Boyer benchmark results

[Camm Maguire]

Thanks for this, Vadim!  As you can tell I decided not to post this
additional example to the release notes page for simplicity reasons.
It is clear that there is a fair degree of noise in these numbers.
The broad conclusion of your run below is basically the same -- the
median and geometric mean of the gcl runtimes are slightly lower than
those for cmucl.  All we are and should be claiming given the state of
this imperfect science is that 'gcl is *about as fast* as cmucl', a
statement which of course will vary given the task at hand, but which
appears to be broadly supported by these results in aggregate.

Take care, 

"Vadim V. Zhytnikov" <address@hidden> writes:

> Camm Maguire writes:
> > Greetings!
> > "Vadim V. Zhytnikov" <address@hidden> writes:
> >
> >>Camm Maguire writes:
> >>
> >>>Greetings!
> >>>Thanks Vadim!  Have added this result to the page.  Please give it a
> >>>look over for possible errors.
> >>>Take care,
> >>
> >>Minor correction.  Table heading reads: CMUCL 18e-9,
> >>CLISP 2.33-2 - these are Debian package release numbers.
> >>But I use my own ALT Linux builds so maybe just
> >>CMUCL 18e and CLISP 2.33.1 would be more precise.
> >>
> > OK, done.
> >
> >>I think it is good idea to put your gabliel.tgz
> >>somewhere on GCL Savannah site and provide reference.
> >>Let's everyone who cares get it and check our
> >>results.
> >>
> > I wanted to commit them to the cvs tree, but there does not appear to
> > be a permissible license/copyright anywhere.  Richard Gabriel offers a
> > pdf of the book from which these were taken on his website under the
> > 'creative commons' license, which, to my understanding, is not
> > compatible with the (L)GPL as it forbids commercial use.
> > I will try to contact him for clarification.
> > Take care,
> >
> >>I'd like to repeat gabriel test suite on yet one
> >>more machine with Pentium III CPU - I have some
> >>733Mhz one near at hand.
> >>
> >>
> 
> Yet another banchmark now with Pentium III
> 
> Pentium III 733MHz, Linux 2.4.22, 384Mb
> 
>           GCL 2.6.2 ANSI   CMUCL 18e   CMU/GCL
> 
> BOYER        21.420        23.280      1.087
> BROWSE       48.450      53.020      1.094
> CTAK         4.260       2.040       0.479
> DDERIV       17.470      18.040      1.033
> DERIV        15.880      12.000      0.756
> DESTRU-MOD   16.720      15.220      0.910
> DESTRU       23.500      15.230      0.648
> DIV2         16.430      12.840      0.781
> FFT-MOD      1.020       1.920       1.882
> FFT          0.980       1.930       1.969
> FPRINT       9.480       27.500      2.901
> FREAD        5.770       12.160      2.107
> FRPOLY       57.030      60.200      1.056
> PUZZLE-MOD   6.390       66.030     10.333
> PUZZLE       6.770       63.580      9.391
> STAK         10.030      15.560      1.551
> TAK-MOD      7.230       8.720       1.206
> TAK          7.300       8.890       1.218
> TAKL         2.710       4.550       1.679
> TAKR         4.120       5.240       1.271
> TPRINT       17.950      29.150      1.624
> TRAVERSE     169.100     106.000     0.627
> TRIANG-MOD   133.470     885.950     6.638
> TRIANG       149.620     157.240     1.051
> 
> 
> --------------   SESSION   ------------------
> 
>         For gcl Common Lisp
> ??? ??? 23 22:14:58 MSD 2004
> 
> BOYER        21.420
> BROWSE       48.450
> CTAK         4.260
> DDERIV       17.470
> DERIV        15.880
> DESTRU-MOD   16.720
> DESTRU       23.500
> DIV2         16.430
> FFT-MOD      1.020
> FFT          0.980
> FPRINT       9.480
> FREAD        5.770
> FRPOLY       57.030
> PUZZLE-MOD   6.390
> PUZZLE       6.770
> STAK         10.030
> TAK-MOD      7.230
> TAK          7.300
> TAKL         2.710
> TAKR         4.120
> TPRINT       17.950
> TRAVERSE     169.100
> TRIANG-MOD   133.470
> TRIANG       149.620
> 
> --------------   SESSION   ------------------
> 
>         For lisp Common Lisp
> ??? ??? 23 22:29:31 MSD 2004
> 
> BOYER        23.280
> BROWSE       53.020
> CTAK         2.040
> DDERIV       18.040
> DERIV        12.000
> DESTRU-MOD   15.220
> DESTRU       15.230
> DIV2         12.840
> FFT-MOD      1.920
> FFT          1.930
> FPRINT       27.500
> FREAD        12.160
> FRPOLY       60.200
> PUZZLE-MOD   66.030
> PUZZLE       63.580
> STAK         15.560
> TAK-MOD      8.720
> TAK          8.890
> TAKL         4.550
> TAKR         5.240
> TPRINT       29.150
> TRAVERSE     106.000
> TRIANG-MOD   885.950
> TRIANG       157.240
> 
> ---------------------------------------------
> 
> It looks like Pentium 4 (I mean 2.4GHz Xeon) desn't
> like some CMUCL tests.
> 
> BTW, on Athlon I've tried -march=i586 vs -march=athlon
> but found no difference.
> 
> -- 
>       Vadim V. Zhytnikov
> 
>       <address@hidden>
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah