~~~
Here are my last Maxima and Fortran saves. You are correct.
I recalculated and got a time ratio of 23 instead of 196
between the Maxima and Fortran program compiles.
Thank you very much for taking the interest and time check this
and keep me on the straight and narrow. I stand corrected.
I will send you copy of my .wxm and .f95 in a separate email.
I find this an interesting thread.. to categorize programs by
their symbolic vs numeric capabilities. Surely a bit of
an oversimplification since Maxima, Fortran, Matlab, Octave
each is both symbolic and numeric to some extent.
"Many people know and love Matlab" and the reason is of course coding and
calculation speed, say finding the pseudo inverse of a 1000 row matrix
is not all that uncommon among science students, and one could argue there
is less symbolics in Matlab than there the was in APL and that APL is
some subset of Maxima symbolics. I have never liked the term 'computer algebra'
applied to Maxima, because Maxima is more at Knowledge base Mathematics,
orders of magnitude faster than any mathematics encyclopedia, or web sorted
lexical search than it is to just algebra. In addition Maxima number routines
have excellent precision and range in floats, yet "who really cares about
Lapack, parallel Lapack, etc "
The reasons for the time differences seems to come down to the story that
the lisp and maxima symbolic implementation, at the opposite extreme of the
annoyingly explicit fortran declarations, can not be easily or practically be
undone to define an inheritable numeric integer index variable class
and must remain symbolic for some hair ball non denumerable state Touring machine
algorithm coding capability. But let me not be hasty and take a closer look at
the " assume(x>0); for i:0 thru 10*x step x do print(i) " and the
" for i:10^1000 thru 10^1000+4 do print("hello"); " examples which
look as interesting as they are surprising, maybe I can still
learn something more here.
Thanks
Ed
~~~
At 02:54 PM 5/27/2008, "S. Newhouse" <sen1 at math.msu.edu> wrote:
...I tried your original code with maxima, intel fortran, and gfortran on
>my linux box running Fedora core 6. The fortrans are faster but at most
>an order of 30, not 196.
>If you want me to try your newer code, just send it to me in private email.
>
>Here are the maxima results....
>fortran result....