Comments on: Speed Problems: C vs. Gambit Scheme

By: effiltUnina

effiltUnina — Fri, 28 Oct 2011 13:23:03 +0000

Hi!
send you post: to my @aqiuuyhw twitter

By: Andrew

Andrew — Tue, 04 May 2010 23:44:48 +0000

Or you could take your original code, convert all the integers to floats and compile it with a different scheme compiler.

Using bigloo 3.3a I get the following timings…
bash-3.00$ bigloo -O3 -suffix scheme -native pi.scheme -o pi3 bash-3.00$ time ./pi3 3.1415949166667


real    0m1.913s

user    0m1.910s

sys     0m0.000s

bash-3.00$ time pi-c

Estimate of pi is: 3.14159492

real 0m1.570s user 0m1.560s sys 0m0.010s

Using the following code…

(module pi (main main2))


(define (main2 args)

  (display (string-append (number->string (exact->inexact (main '()))) "\n")))
(define (main args)

  (let* ((partitions 12000.)

         (increment (/ 1. partitions))

         (count 0.))
    (do ((y 0. (+ 1. y)))

        ((= partitions y) (* 4. (/ count (* partitions partitions))))

      (let ((b (* increment (+ y .5))))

        (do ((x 0. (+ 1. x)))

((= partitions x)) (let ((a (* increment (+ x .5)))) (if (< (+ (* a a) (* b b)) 1.) (set! count (+ 1. count)))))))))

By: Axio

Axio — Mon, 22 Mar 2010 08:40:17 +0000

Yeah. It’s the same code with more explicit typing and better precalculation.

By: Axio

Axio — Mon, 22 Mar 2010 08:25:42 +0000

/tmp % time gsi-gambit loop2
3.1415949166666666
gsi-gambit loop2 101.80s user 1.65s system 97% cpu 1:45.65 total
[adrien]/tmp % time ./loop-c
Estimate of pi is: 3.14159492
./loop-c 1.11s user 0.00s system 99% cpu 1.124 total

(declare (block) (standard-bindings) (extended-bindings) (not safe))

(define (main args)
(let ((num-partitions 12000.)
(num-partitions-ex 12000))
(let ((interval (fl/ 1. num-partitions)))
(let loop ((i 0) (circle-count 0))
(if (< i num-partitions-ex)
(let ((new-circle-count
(let loop2 ((j 0) (circle-count circle-count))
(if (< j num-partitions-ex)
(let ((a2 (flexpt (fl* (+ i .5) interval) 2.))
(b2 (flexpt (fl* (+ j .5) interval) 2.)))
(loop2 (+ j 1)
(if (string (exact->inexact (main ‘()))) “\n”))

I guess that I’m still far from Gambiteer’s performance, but it’s still better than your original code ^^

By: Gambiteer

Gambiteer — Wed, 17 Mar 2010 02:24:53 +0000

See

http://paste.lisp.org/display/96491#4

By: Gambiteer

Gambiteer — Wed, 17 Mar 2010 02:17:56 +0000

(declare (standard-bindings) (extended-bindings) (block) (not safe))

(define (main args) (let* ((partitions 12000) (increment (fl/ (fixnum->flonum partitions)))) (let y-loop ((y 0) (count 0)) (if (fx= partitions y) (* 4 (/ count (* partitions partitions))) (let ((b (fl* increment (fl+ (fixnum->flonum y) 0.5)))) (let x-loop ((x 0) (count count)) (if (fx= partitions x) (y-loop (fx+ y 1) count) (let ((a (fl* increment (fl+ (fixnum->flonum x) 0.5)))) (if (flstring (exact->inexact (main '()))) "\n"))

By: Gambiteer

Gambiteer — Wed, 17 Mar 2010 02:16:20 +0000

(declare (standard-bindings)
(extended-bindings)
(block)
(not safe))

(define (main args)
(let* ((partitions 12000)
(increment (fl/ (fixnum->flonum partitions))))
(let y-loop ((y 0)
(count 0))
(if (fx= partitions y)
(* 4 (/ count (* partitions partitions)))
(let ((b (fl* increment (fl+ (fixnum->flonum y) 0.5))))
(let x-loop ((x 0)
(count count))
(if (fx= partitions x)
(y-loop (fx+ y 1)
count)
(let ((a (fl* increment (fl+ (fixnum->flonum x) 0.5))))
(if (flstring (exact->inexact (main ‘()))) “\n”))

:~> gcc -O3 -Wall -o pic pic.c
:~> time ./pic
Estimate of pi is: 3.14159492
1.230u 0.000s 0:01.23 100.0% 0+0k 0+0io 0pf+0w
:~> gsc pi
:~> time gsi pi
3.1415949166666666
1.250u 0.000s 0:01.25 100.0% 0+0k 0+0io 0pf+0w