Fastcode Move B&V 8.0

From: Dennis (marianndkc_at_home3.gvdnet.dk)
Date: 02/26/05 

Date: Sat, 26 Feb 2005 16:47:41 +0100

Hi All

I am working on a new release of the Move B&V.

John O'Harrow has suggested some changes to the benchmark. I would like to
discuss this one first:

"I would like to propose the then maximum array size used within the
TinyMovesWithoutCacheFlushes procedure (SubBenchmark 4) is reduced from its
current value of 51 down to a more realistic value of 32. Where did we get
51 from anyway?

Experiments have shown that the vast majority of moves within the RTL (and
user code) are of between 1 and 32 bytes). I believe changing this limit
will therefore give a more realistic benchmark."

Do you agree that the vast majority of moves are between 1 and 32 bytes?

I think I remember something about some measurements, but I need you to
refresh my memories ;-)

Best regards
Dennis

Relevant Pages

  • Re: combining array refs
    ... > John W. Krahn wrote: ... >>> what is the most effecient way to combine multiple array refs into one ... > You might want to Benchmark this. ... > I know from the perl cookbook that joining two hashes can be done this ...
    (perl.beginners)
  • Re: Running 100 Balls
    ... makes 100 has achieved a level of ability beyond the vast majority of ... players. ... BUT, when I think of hitting a benchmark of 50 or 100, ...
    (rec.sport.billiard)
  • Re: Text munging problem
    ... John W. Krahn am Samstag, ... use Benchmark q; ... Rate twomaps onemap subst ... Hmm... ...
    (perl.beginners)
  • Re: wanted fast routines
    ... >>>Ergo, the whole benchmark stinks. ... If David Frank can come up with a more ... >>>realistic benchmark, then I'd be happy to repeat my comparison. ...
    (comp.lang.fortran)
  • Re: New FastCode Uppercase and Lowercase functions
    ... > Hi John and all ... >> These functions are considerably faster than all of my previous IA32, ... > Which benchmark did you use? ...
    (borland.public.delphi.language.basm)