lseward@randvax.UUCP (Larry Seward) (10/31/84)
Forwarded news... Received: from mit-mc.arpa by rand-unix.ARPA; Mon, 29 Oct 84 11:17:21 pst Message-Id: <8410291917.AA16013@rand-unix.ARPA> Date: 29 October 1984 14:16-EDT From: Jeffrey P. Golden <JPG@MIT-MC> Subject: MACSYMA vs. SMP In answer to Jim Purtillo's recent mail: I have used SMP very little, so I am not a good candidate to answer your mail. It would be nice if people who have ready access to both SMP and MACSYMA could come up with some test suites of examples that come up in regular use (i.e. not canned examples produced by Inference or Symbolics) to compare for capability, ease of use, and speed. Some classes of examples where I would think MACSYMA would shine over SMP include: multivariate factorization, indefinite integration especially where the Risch code is used, and problems where exact answers are required. SMP tends to introduce floating point numbers pretty early in the game. It would be great if some comparisons could be run by those who have SMP and MACSYMA on the same machine.
purtilo@uiucdcsb.UUCP (11/05/84)
....... Regarding the MACSYMA vs. SMP question recently posed .... Thanks to all who responded in mail: to briefly summarize what came in, uniformly all who bothered to comment at all commented in favor of macsyma. Folks tended to base their opinions NOT on relative speed or any hard-to-judge criteria such as "ease of use", but on correctness. The examples cited by Jeff Golden in an earlier news item (e.g. multi- variate factorization or too-early introduction of floating point) were noted by many. To follow up on a suggestion Jeff made (in the aforementioned note), I think this forum might be a pretty good place to contrast use of the various systems available as applied in different ways. Perhaps a reasonable ``test suite'' of problems could ultimately be derived; but since we all know the dangers of comparing systems solely on performance on a limited set of problems, I submit that the value will not so much be just the set of problems assembled as it will be the value of deciding exactly what/how we want to measure. Anyway, there is precedent: every once in a while such an article appears in SIGSAM, e.g., so perhaps this forum could be used to assist in (1) making it easier for the general public to try the tests for themselves, and (2) allowing the public to have more direct feedback on what is tested. Cheers - Jim