🚀 go-pugleaf

In article <3480B9C0.5861@email.adr> of Sat, 29 Nov 1997 16:56:32 in
comp.lang.pascal.borland, Mike <bad@email.adr> wrote:
>Dr John Stockton wrote:
>> 
>> In article <347FC4E5.7D00@email.adr> of Fri, 28 Nov 1997 23:31:49 in
>> borland.public.turbopascal, Mike <bad@email.adr> wrote:
>> >
>> >> With the upload of this article, I should be uploading the new
>> >>         http://www.merlyn.demon.co.uk/programs/del_timr.pas
>> >> which also includes the Delete routines of myself, Mike, & Pedt - uses
>> >> conditional compilation, see the code.
>> 
>> >Is a printout available of the results on your system?
>> 
>> Yes, if you have a printer!  The results should be approximately system-
>> independent; timing self-calibrates.  Results depend, no doubt, on what
>> sizes you use; I use random ones.
>
>Actually, I was interested in comparing the results on *your* system.

<< Who else has a 486/33 these days? even I am nearly ready to upgrade.
I'm running an exhaustive correctness test of Inject on my older
machine, a PPC640 - all string lengths, all positions in string and some
not in - been running since about 11/29-20:00, should finish 12/04-??
SFSG  >>

I've reported the results here before; here's a time table :

     Remove/Delete    Inject/Insert   Run-Time Checks (RSIc)
          2/3             2/3               ON  
          2/5             1/2               OFF

These have been re-measured, as there have been minor changes to make
Inject infallible (uploaded yesterday evening (GMT)).

There is no point in being very exact, since the average speedup depends
on the mix of string lengths and other parameters, and I guess on
processor type; my results are for what the number converges to using
uniformly-distributed random data, including some off-range cases.  Some
would prefer to concentrate on shorter strings and remain on-range; some
would average differently; some would redo the method in asm (as the
library one is), but I do not have that skill.  All I really claim is
that the newer method (which, I believe, is what Delphi uses) seems
usefully faster even when implemented in Pascal.
 
-- 
John Stockton, Surrey, UK.    jrs@merlyn.demon.co.uk    Turnpike v1.12    MIME.
  Web URL: http://www.merlyn.demon.co.uk/ - FAQqish topics, acronyms and links.
  Correct 4-line sig separator is as above, a line comprising "-- " (SoRFC1036)
  Before a reply, quote with ">" / "> ", known to good news readers (SoRFC1036)