Re: [PATCH] use tail pointer for LOOP

[Thien-Thi Nguyen <ttn@gnuvola.org>]

() tomas@tuxteam.de
() Thu, 17 Jun 2010 16:05:01 +0200

   Here you go:

     copy1: (1.115929 5 0.7795429999999997)
     copy2: (1.210733 5 0.8172469999999996)
     copy3: (1.2816079999999999 5 0.8502519999999998)
     copy4: (1.295846 5 0.9325959999999993)

   (I did implement the garbage-collect-before-each-run as you suggested in
   your other mail: this can be nicely seen in the constant number "5"
   above).

   I won't spam this list with the modified source (unless someone clamours
   for it, that is).

   So without the GC the gap narrows significantly. But still: reverse's
   the king of the hill.

   OTOH, the differences in GC times start to be a considerable fraction of
   the whole difference, so I might be measuring noise anyway.

You can test this hypothesis (or rather the sensitivity of the system to this
effect) by permuting the order.  E.g., what if you do ‘copy1’ last?

   Seems we can't beat an old Lisp idiom (my guess is that Lisp is alien
   technology, but hey).

Don't give up, yet.

Another idea is to 10x the input length.

thi