José Romero <j...@2.71828.com.ar> writes: Hi Jose, very interesting results. I always had the feeling that even compiled Elisp is not much faster than PicoLisp, but I thought that was just a 'feeling'. I would like to add your results to the wiki page - let me know if I shouldn't.
> On Mon, 23 Apr 2012 02:22:21 +0200 > Thorsten <quintf...@googlemail.com> wrote: > >> >> Hi List, >> out of curiosity I did a little speedtest with PicoLisp and Emacs Lisp >> on a 64bit Arch Linux (see >> http://picolisp.com/46850/62704437697738713~!wiki?PILvsEL). >> >> The results are quite favorable for PicoLisp - it is about 4 times >> faster than interpreted Emacs Lisp and 2 times faster than compiled >> Emacs Lisp. >> > > Nice, I did my own tests on a 32 bit machine, not such a dramatic > speedup for picolisp by my numbers (about the same speed as bytecode) > but this thing is only really measuring function call/arithmetic cost. > > $ cat > fibo.el << . >> (defun fibo (N) >> (if (> 2 N) >> 1 >> (+ (fibo (1- N)) (fibo (- N 2))) ) ) >> >> (fibo 35) >> . > $ time emacs --no-site-file --script fibo.el > > real 0m13.854s > user 0m13.829s > sys 0m0.016s > $ emacs --no-site-file -batch -f batch-byte-compile fibo.el > Wrote /home/cyborgar/tmp/sptest/fibo.elc > $ time emacs --no-site-file --script fibo.elc > > real 0m5.882s > user 0m5.828s > sys 0m0.008s > $ cat > fibo.l << . >> (de fibo (N) >> (if (> 2 N) >> 1 >> (+ (fibo (dec N)) (fibo (- N 2))) ) ) >> (fibo 35) >> . > $ time pil fibo.l -bye > > real 0m5.662s > user 0m5.624s > sys 0m0.004s > > I tested with the "extensive list manipulations" code Alex tested CLisp > and SBCL a while back, with pretty nice results also: > > $ cat > tst.el << . >> (defun tst () >> (mapcar >> (lambda (X) >> (cons >> (car X) >> (reverse (delete (car X) (cdr X))) ) ) >> '((a b c a b c) (b c d b c d) (c d e c d e) (d e f d e f)) ) ) >> (dotimes (i 1000000) (tst)) >> . > $ time emacs --no-site-file --script tst.el > > real 0m8.311s > user 0m8.273s > sys 0m0.032s > $ emacs --no-site-file -batch -f > batch-byte-compile tst.el Wrote /home/cyborgar/tmp/sptest/tst.elc > $ time emacs --no-site-file --script tst.elc > > real 0m5.622s > user 0m5.604s > sys 0m0.012s > $ cat > tst.l << . >> (de tst () >> (mapcar >> '((X) >> (cons >> (car X) >> (reverse (delete (car X) (cdr X))) ) ) >> '((a b c a b c) (b c d b c d) (c d e c d e) (d e f d e f)) ) ) >> (do 1000000 (tst)) >> . > $ time pil tst.l -bye > > real 0m1.208s > user 0m1.192s > sys 0m0.012s > > Looks like the emacs compiler can't improve much in that function and > it's still 4.6-6.9x slower than picolisp, whoops :) > > $ uname -a > Linux icz 2.6.32-5-686 #1 SMP Mon Jan 16 16:04:25 UTC 2012 i686 > GNU/Linux > $ cat /proc/cpuinfo |grep "model name" | cut -d: -f2 > Pentium(R) Dual-Core CPU T4200 @ 2.00GHz > Pentium(R) Dual-Core CPU T4200 @ 2.00GHz > > Cheers, > José -- cheers, Thorsten -- UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
