Re: Ada Shootout program for K-Nucleotide (patches)

[Jacob Sparre Andersen <sparre@nbi.dk>]

Georg Bauhaus wrote:

> I don't know, though, whether or not exchanging one "&" for one less
> Stream_IO.Write is also good when strings are larger, because of the
> copying in "&", CPU cache sizes, ...

I fear that it might be slower in that case. - But it may not be relevant here.

> Since the test has been run in a VM on top of Vista, I don't trust
> the timings that much anyway.  Maybe the order of tests has an
> effect, too?  Does somebody have a stable Linux (or Unix) test
> environment?  Could you try this?

I tried on my Debian/stable workstation (made the program repeat the
test twice in case of initialisation effects):

Print_Old, SHORT_STRINGS:  12.368589000
Print_Old, MEDIUM_STRINGS:  1.143411000
Print_Old, LONG_STRINGS:  0.553761000
Print, SHORT_STRINGS:  13.874468000
Print, MEDIUM_STRINGS:  3.012395000
Print, LONG_STRINGS:  2.274855000
Print_Old, SHORT_STRINGS:  12.254596000
Print_Old, MEDIUM_STRINGS:  1.140442000
Print_Old, LONG_STRINGS:  0.573983000
Print, SHORT_STRINGS:  13.666659000
Print, MEDIUM_STRINGS:  3.014188000
Print, LONG_STRINGS:  2.286064000

The program was apparently using a nearly constant 94% of the CPU
resources during the whole run (except for the first few seconds).

Greetings,

Jacob
-- 
"Politicians are a lot like diapers. They should be changed
 frequently, and for the same reasons."