home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.symbolic
- Path: sparky!uunet!utcsri!helios.physics.utoronto.ca!alchemy.chem.utoronto.ca!mroussel
- From: mroussel@alchemy.chem.utoronto.ca (Marc Roussel)
- Subject: Space efficiency (Was: The Real Meaning of Efficiency?)
- Message-ID: <1992Nov20.051351.18180@alchemy.chem.utoronto.ca>
- Organization: Department of Chemistry, University of Toronto
- References: <18NOV199217533011@reg.triumf.ca> <1eghvuINNi6i@crcnis1.unl.edu>
- Date: Fri, 20 Nov 1992 05:13:51 GMT
- Lines: 30
-
- Others have rightly pointed out that a concept of time efficiency that
- does not include the programmer's time is inadequate in a modern
- research environment. Unfortunately, almost everyone seems to have
- ignored space efficiency in designing their symbolic algebra libraries. I
- spent several frustrating hours yesterday trying to do a tricky
- calculation with Maple on a busy machine, all because the size of the
- computation kept growing beyond available resources. (We have 128MB of
- memory, of which about 30MB was available during this session. I did
- eventually manage to run my job by changing my approach to the problem,
- but the relatively trivial computation I was engaged in should have fit
- in 30MB. The library routines involved were obviously written by someone from
- the "RAM is cheap" school of thought.) What I and, I'm sure, many
- other scientists could really use is the ability to tell a symbolic
- algebra program whether to trade time for space or vice-versa. Quite
- often, making some sort of compromise is the only way to get the
- calculation through the machine. I'm perfectly willing to compromise on
- turn-around time if it's the difference between successfully completing
- a calculation and staring at yet another "Object too large" message.
- I do know about FORM. One of these days, I'll have to take a good
- hard look at it. My preliminary impression (based on inspecting the
- manual) is that the type of computations which I most commonly carry
- out are not naturally expressed in FORM's language.
- I would be interested to hear whether the major players (the Symbolic
- Computation Group, Wolfram Research, the vendors of various Macsyma variants,
- etc.) worry much about space efficiency when they are creating new
- library code. I get the impression that the answer is no, at least for
- Maple and Mathematica.
-
- Marc R. Roussel
- mroussel@alchemy.chem.utoronto.ca
-