Maxima: compiling with cmucl, sbcl and gcl.

Hi Robert,

Unfortunately, it is not possible to send the program as this calls a 
number of functions which I have written, which in turn call other functions, 
etc.

However, what I wanted to tell you is that in the second half of the program 
the function mysimprec() (see below) is called about 5 times. 

>From the running times given below, it appears that Maxima 5.10 is much 
faster than Maxima 5.12 in running this function.

(1) First half of program (second half commented out):
amd64: suse linux 10.1, 64 bit, Maxima 5.10, cmucl: 5.8 sec
amd64: suse linux 10.1, 64 bit, Maxima 5.12, cmucl: 6.8 sec

(2) First half of program plus one call to mysimprec():
amd64: suse linux 10.1, 64 bit, Maxima 5.10, cmucl: 6.6 sec
amd64: suse linux 10.1, 64 bit, Maxima 5.12, cmucl: 86.7 sec

(3) Whole program:
amd64: suse linux 10.1, 64 bit, Maxima 5.10, cmucl: 14.7 sec
amd64: suse linux 10.1, 64 bit, Maxima 5.12, cmucl: 170.5 sec

Best regards,

C. Frangos.



On Friday 06 July 2007 06:44, you wrote:
> On 7/5/07, C. Frangos <cfrangos at telkomsa.net> wrote:
> > amd64: suse linux 10.1, 64 bit, Maxima 5.10, cmucl: 14.7 sec
> > amd64: suse linux 10.1, 64 bit, Maxima 5.10, sbcl: 17.7 sec
> > amd64: suse linux 10.1, 64 bit, Maxima 5.10, clisp: 64.3 sec
> >
> > amd64: suse linux 10.1, 64 bit, Maxima 5.12, cmucl: 170.5 sec
> > amd64: suse linux 10.1, 64 bit, Maxima 5.12, sbcl: 200.5 sec
> > amd64: suse linux 10.1, 64 bit, Maxima 5.12, clisp: 633.4 sec
>
> This is very intriguing. I would certainly like to know why
> 5.12 is apparently much slower. I don't believe we have had
> other reports of that, so I'm guessing the reason depends on
> the details of the program you are running. In the absence of
> the program, I don't think we can tell. So I'll take this opportunity
> to encourage you to post the program if you can.
>
> best
> Robert Dodier


mysimprec(s) := block(

[indexlist,sx,nf,listsimpf,nsx,nmin,simpf,sxnew],

listsimpf : [fullratsimp, trigsimp, factor, trigreduce, dividethru],

nf : length(listsimpf),

sx : s,

nmin : length(string(sx)),

indexlist : [],

for i1 : 1 thru nf do (

simpf : listsimpf[i1],

sxnew : apply(simpf,[sx]),

nsx : length(string(sxnew)),

if (nsx < nmin) then (
indexlist : endcons(simpf,indexlist),
sx : sxnew,
nmin : nsx
)

),

display(indexlist),

return(sx)

);





dividethru(a) := block(

[num, den, quo, result],

if not atom(a) and op(a)="//" then
( 
  num : part(a,1), 
  
  den : part(a,2),
  
  quo : divide(num,den), 

  result : quo[1]+quo[2]/den
) 

else 

(result : a),

return(result)

);