Path: news.daimi.aau.dk!jacobse
From: jacobse@daimi.aau.dk (Jacob Seligmann)
Newsgroups: comp.lang.beta,comp.lang.lisp
Subject: Re: Comparison: Beta - Lisp
Date: 9 Sep 1994 11:03:38 GMT
Organization: DAIMI, Computer Science Dept. at Aarhus University
Lines: 73
Message-ID: <34pfea$6ee@belfort.daimi.aau.dk>
References: <34n2qe$d74@nz12.rz.uni-karlsruhe.de> <lenngrayCvunsr.448@netcom.com>
NNTP-Posting-Host: clematis.daimi.aau.dk
Xref: news.daimi.aau.dk comp.lang.beta:15 comp.lang.lisp:13162

Lenny Gray (lenngray@netcom.com) wrote:

> Bruno Haible (haible@ma2s2.mathematik.uni-karlsruhe.de) wrote:
>
> : ...
> :    some integer array hacking
> :          C            4.2 sec
> :          Mjolner    111 sec
> :          GCL        288 sec
> :          CLISP      415 sec
> :    (All timings on a 486/33.)
> : ...

Ouch!

> Are these numbers right?  I've seriously used GCL and CLISP myself and
> had some arguments with a "true believer Lisper" who thought "Lisp _does_
> compete reasonably with C for numeric stuff", but I never bothered to do
> the timing tests, and always assumed it wasn't this bad.  Is it, really?

I don't know what Bruno's "some integer array hacking" program does, 
but the performance ratios shown do not reflect my personal experience. 
For example, here's a very simple "some integer array hacking" program 
I wrote for the occasion:

   /* siah.c - some integer array hacking in C */
   void main(void)
   {
    int a[10000], i, j;
    for (j=0; j<1000; j++)
      for (i=0; i<10000; i++)
        a[i] = i*4;
   }

   (* siah.bet - some integer array hacking in BETA *)
   ORIGIN '~beta/basiclib/v1.4/betaenv'
   --program:descriptor--
   (#
      a: [10000]@integer;
   do 
      (for j:1000 repeat
           (for i:a.range repeat i*4 -> a[i] for);
      for);
   #)

And here's the run-times (486/66; BETA compiler v5.0(1); gcc v2.5.8):

   BETA      4.41
   gcc       3.10
   gcc -O6   1.24

In this case, the ratio between BETA and unoptimized C was 1.4, while 
the ratio between BETA and optimized C was 3.6.

Bear in mind that the BETA compiler does not currently perform nearly 
as much optimization as gcc does. Also, the code generated by the BETA 
compiler contained an index check (which could have been removed, had 
the optimizer been smarter) at each assignment, while the C code 
obviously did not.

> Also, I was interested in Beta until one minute ago, because of this.

Hopefully, your interest has been reawakened.

/Jacob Seligmann
------------------------------------------------------------------------
Mjolner Informatics ApS             Phone:   (+45) 86 20 20 00 ext. 2754
Science Park Aarhus                 Direct:  (+45) 86 20 20 11 - 2754
Gustav Wieds Vej 10                 Fax:     (+45) 86 20 12 22
DK-8000 Aarhus C, Denmark           Email:   jacobse@mjolner.dk       
------------------------------------------------------------------------
                             BETA is better                             
------------------------------------------------------------------------