On Sun, Aug 23, 2009 at 10:38 PM, Robert Dodier<robert.dod...@gmail.com> wrote:
> On 8/23/09, William Stein <wst...@gmail.com> wrote:
>
>> In fact, I am mildly surprised that the above two timings between ecl
>> and maxima are so different. I wonder why it is an order of magnitude
>> faster to compute "(+ 1 2)" than to compute "(MEVAL #$ 1 + 2$)".
>> Perhaps in the second case the Maxima parser has to parse "1+2"?
>
> Well, the Maxima parser is slow, but also (MEVAL '((MPLUS) 1 2))
> goes through the simplifier, which eventually calls (+ 1 2)
> as a special case. The simplifier attempts to apply various
> identities, so it's really much more complex than just adding two
> integers. I suppose that adding numbers is an important enough
> special case that some effort could be spent trying to speed it up.
> But most of the fun is in symbolic stuff, so that's probably why
> it hasn't been done yet.
>
> Not sure in this case whether the parser or simplifier dominates.
>
>> Robert Dodier -- do you have any crazy ideas why Maxima might get
>> slower to do the same thing over time? Is it caching all output
>> somewhere? What other caching might be going on?
>
> This kind of stuff drives me nuts ... It appears the problem is
> that Maxima is keeping a list of all input streams that have
> ever been opened, and traversing the list every time an input
> line is parsed. I'm guessing that Sage creates a little input file
> each time it calls Maxima or something like that which creates
> a new input stream every time. (In that case a work-around is
> to avoid creating so many new input streams. Maybe open a
> socket and send stuff through that?)
Wow, many thanks for looking into this!!
Regarding your hypothesis, in theory I do not think Sage does any such
thing. When Sage uses Maxima via pexpect, the *only* thing it does is
type expressions in via the tty. This is *exactly* the same as when
you use Maxima. Maxima absolutely cannot tell the difference between
you using it and Sage using it. Here's a picture:
Sage <----> Pseudotty <---> Maxima
You <-----> Bash/ssh/Pseudotty <---> Maxima
There is one exception -- if Sage is going to send some huge input to
Maxima, it will write it to a file, then tell Maxima (via the tty) to
read in that file. That is orthogonal to our problem.
> That is in the code which annotates input with
> debugging info. Try this: either edit maxima/src/nparse.lisp
> and clobber the existing definition of ADD-LINEINFO with:
>
> (defun add-lineinfo (x) x)
OK, I'm testing this now. I made this the last line of nparse.lisp.
The result -- no go. Any other ideas? Anything I can add in debug
wise? I wonder if there is a way for us to recreate this outside of
Sage... that's the other thing I can try I guess.
./sage -f maxima-5.16.3.p3.spkg
...
real 3m18.849s
user 2m34.570s
sys 0m37.220s
Successfully installed maxima-5.16.3.p3
You can safely delete the temporary build directory
/scratch/wstein/build/sage-4.1/spkg/build/maxima-5.16.3.p3
Making Sage/Python scripts relocatable...
Making script relocatable
Finished installing maxima-5.16.3.p3.spkg
wst...@sage:~/build/sage-4.1$ ./sage
----------------------------------------------------------------------
| Sage Version 4.1, Release Date: 2009-07-09 |
| Type notebook() for the GUI, and license() for information. |
----------------------------------------------------------------------
Loading Sage library. Current Mercurial branch is: ref3
sage: timeit('maxima.eval("2+2")')
5 loops, best of 3: 565 µs per loop
sage: timeit('maxima.eval("2+2")')
625 loops, best of 3: 1.18 ms per loop
sage: timeit('maxima.eval("2+2")')
125 loops, best of 3: 2.86 ms per loop
sage: timeit('maxima.eval("2+2")')
125 loops, best of 3: 3.87 ms per loop
Argh!
William
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
