Hi,
I've gone ahead and added vector and bytevector ops to the VM. That
should provide a considerable speedup to programs that use those data
structures.
As far as brainfuck goes:
On Wed 24 Jun 2009 14:03, Andy Wingo <wi...@pobox.com> writes:
> brainfuck@(guile-user)> ,c ++]
brainfuck@(guile-user)> ,c ++]
Disassembly of #<objcode b7073dc0>:
0 (make-int8:0) ;; 0
1 (load-symbol "make-vector") ;; make-vector
16 (link-now)
17 (variable-ref)
18 (make-int16 117 48) ;; 30000
21 (make-int8:0) ;; 0
22 (call 2)
24 (local-set 1)
26 (local-set 0)
The prolog, as before
28 (local-ref 1)
30 (local-ref 0)
32 (local-ref 1)
34 (local-ref 0)
36 (vector-ref)
37 (make-int8:1) ;; 1
38 (add)
39 (vector-set)
One +
40 (local-ref 1)
42 (local-ref 0)
44 (local-ref 1)
46 (local-ref 0)
48 (vector-ref)
49 (make-int8:1) ;; 1
50 (add)
51 (vector-set)
The other +.
52 (void)
53 (return)
Here we push "the unspecified value" and return.
As far as performance goes, I took the fibonacci program from
http://www.hevanet.com/cristofd/brainfuck/fib.b:
>++++++++++>+>+[
[+++++[>++++++++<-]>.<++++++[>--------<-]+<<<]>.>>[
[-]<[>+<-]>>[<<+>+>-]<[>+<-[>+<-[>+<-[>+<-[>+<-[>+<-
[>+<-[>+<-[>+<-[>[-]>+>+<<<-[>+<-]]]]]]]]]]]+>>>
]<<<
]
This program doesn't terminate; you will have to kill it.
Daniel B Cristofani (cristofdathevanetdotcom)
http://www.hevanet.com/cristofd/brainfuck/
and let it run for about 3 seconds, wall time. The last number that it
printed was:
271651113667355860622427102019076226843364138595316927370091609703323530354645264253440817026771386351610030043817601548972412867287338615983131
I then switched back to Scheme, did a (define x that-number), then:
(define (find-fibo n a b)
(let ((c (+ a b)))
(write c) (newline)
(if (= c x)
n
(find-fibo (1+ n) b c))))
,t (find-fibo 3 1 1)
=> 688
clock utime stime cutime cstime gctime
0.15 0.01 0.00 0.00 0.00 0.00
That is to say, brainfuck computed and printed the 688 first elements of
the fibonacci sequence in about 3 seconds.
The "write" is find-fibo in there to test output speed. Without it
find-fibo executes in 0.01 s. With it, and my stupid gnome-terminal,
0.15 or 0.20 s.
I think that means the brainfuck implementation is doing pretty well, to
do bignum math at that speed, only having 1+ and 1- as its operators...
Andy
--
http://wingolog.org/
