Re: array of strings in a function

[Craig Carey <research@ijs.co.nz>]

On Sat, 18 Oct 2003 10:05:30 +0000, Marius Amado Alves
<amado.alves@netcabo.pt> wrote:

>On Sat, 2003-10-18 at 08:48, Craig Carey wrote:
>> I timed some Ada 95 Strings packages. Ada 95 vendors are guided by the
>> design of Ada and then their customers get faster code from the
>> Internet. 
>> 
>> Here are the results of the test. 
>> SP Speed/Hz = 171560  1/12.5 : my no-strings array of pointers scheme
>> SV Speed/Hz =  13760  1      : MAA with my Striunli strings
>...
>
>Nice work! Could you explain the speed units tho?

That 13,760 is the number of loop iterations divided by the number of
seconds. Timing was done in a 900 MHz Windows 200x PC running -O2 Gnat
3.15p, and the loop contained just this line:

     AV_Show (+"one" + "two" + "three" + "four" + "five" + "six");

The numbers record that your solution is 29 times slower than that
solution I e-mailed that had an array aggregate of records containing
a length and a pointer to the 1st character of the string. (29 approx
equals "12.5" times "2.3".) So far only one Ada free strings package
mentioned here is designed to be fastest. Then when tested it is found
to be fastest. I can't comment on nearly invisble compilers such as
the 1.1MSLOC IBM Ada compiler.

Here are the URLs of the strings packages that were being promoted at
comp.lang.ada:

[1] My Striunli (was "Seafood") package: http://www.ijs.co.nz/code/
[2] Adalog strings package of Mr Rosen: http://www.adalog.fr/compo2.htm
[3] The Charles Containers" routines of Mr Heaney
    http://home.earthlink.net/~matthewjheaney/charles/index.html


Here is the code with the timed loops:
http://www.ijs.co.nz/code/tmp-comp.lang,ada-18-oct-2003-string-speed
                                    [continued.]   -test-program.ada


-----

Can someone tell us if IBM is promoting its compiler ?.

Hopefully the ARG will stone cold deny IBM a right to exercise a
voting power at ARG decision making meetings with the reason being
that the company public claims to an interest in Ada is that it is
the first to make a public show of losing a 1.1E6 SLOC compiler.

The SLOC size is mentioned by Mr P. Leroy of IBM (who votes at
ARG meetings), here at an Ada-Comment message:

   http://www.ada-auth.org/cgi-bin/cvsweb.cgi/AIs/AI-00286.TXT

Pehaps ACT would rush to disagree but I suppose they should have
a greater influence in Ada's future.

For the whole year I was expecting some newsletter from IBM
conceding that they own an Ada compiler but I got nothing. One of
the Rational staff seemed to try to get that to happen. Incredibly
nothing on Ada at the website.

------------------------------------------------------------------------
At 2003-10-16 09:46 -0400 Thursday, Edge Subscriber wrote:
>
>Dear Rational Edge Subscriber,
>
>The October 2003 issue of our ezine, from IBM Rational software,
>is ready for you at http://www.therationaledge.com
>
>This month, we focus on modeling and use cases -- especially for large
> systems, integration projects, and systems engineering efforts.
>
>And please take our READER SURVEY! We want to know more about you and
> your software development issues, in order to keep valuable content
> coming your way.
>
>Sincerely,
>
>Mike Perrow
>Editor-in-Chief
>The Rational Edge
------------------------------------------------------------------------

I have not figured out the name of the city that IBM is making those
decisions in.



Craig Carey
   Mailing lists: http://www.ijs.co.nz/ada_95.htm