[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Read times improved in AKCL 1.179

AKCL 1.179 contains, among other things, a nice performance
improvement to the Lisp function READ, thanks to Bill Schelter.  Here
are some stats obtained concerning the cost of READing a 1 megabyte
file.  In each case the results reflect the cpu, not real/elapsed,
time reported by the form

  (time (with-open-file (fl "foo" :direction :input) (read fl) nil))

However, in each case, the cpu time was within 3% of the reported
real/elapsed time.  Times are in seconds.  In each Lisp the above form
was evaluated twice to (somewhat) mitigate the randomness of the time
spent in GC.  (One might hope, for example, that the space allocated
during and discarded after the first READ would be speedily
reallocable during the second READ.)

                   First READ                    Second READ

AKCL 1.179            38                             24
Lucid 3.0.2           78                             67
KCL (June 87)         79                             72
Allegro 3.1.beta.22  140                            121

These are times running on a Sun-3/280.

(The file "foo" is actually the file "goedel.events" available by
anonymous ftp as /pub/nqthm/goedel.events from rascal.ics.utexas.edu.
This is N. Shankar's input to Moore's and my prover, checking Rosser's
version of Goedel's incompleteness theorem for Shoenfield's version of
first order logic using Cohen's Z2 axioms for finite set theory.)