Newsgroups: comp.lang.apl
Path: watmath!watserv2.uwaterloo.ca!torn!cs.utexas.edu!zaphod.mps.ohio-state.edu!swrinde!gatech!gsusgi1.gsu.edu!gsusgi1.gsu.edu!qmdbms
From: qmdbms@gsusgi2.gsu.edu (Brian Schott)
Subject: Re: sorry, but a disappointing performance report
Message-ID: <qmdbms.726353313@gsusgi1.gsu.edu>
Keywords: performance,j
Organization: Georgia State University
References: <1993Jan5.190233.13373@fnbc.com>
Date: Wed, 6 Jan 1993 20:48:33 GMT
Lines: 45

jja@fnbc.com (Joseph Alotta) writes:



>   a=. i. 1e7
>   +/a
>4.999995e11

>for the above, the z interpreter was an order of magnitude (10x) faster  
>than the j interpreter.

What release of j did you use? I understand it has been sped up of late.

>the person who wrote it, a very smart person, says he didn't do anything
>special or anything different than what a good c programmer would
>noramlly do.   my quick inspection of his source code proves this to be  
>true.

>the z interpreter is an order of magnitude slower than compiled c code, so  
>that means that j is 2 orders of magnitude (100x) slower than c.

>this raises the following questions:

>1.  can these results be verified by testing j against apl interpreters?

There is a timing primitive in j. Try:(100) 6!:2 <'+/a' to get the time
average of 100 executions.

>2.  can it be that j is doing additional work that z is not?

>3.  does anyone really care about the speed?

>4.  perhaps maybe someone should think about modifying to j interpreter
>to have it run faster.

>5.  i might be willing to take a look at it, if i could find the
>documentation.  as it is i can't make head or tails out of the code.

Documentation for the code is available for $90 from ISI in Canada.  It is
a terse, but IMHO, clear document. Also, Roger Hui has been awarded a
fellowship type of position by the ACM for the year 1993.  His travel to
sites in North America can be partially funded by the ACM.  He implemented
j and could probably motivate folks to enhance the code.

(B=)
