[racket] Everything type-checks; on to benchmarking and optimization

From: Matthias Felleisen (matthias at ccs.neu.edu)
Date: Mon Feb 17 08:27:39 EST 2014

That is odd. Do share the code so others can inspect it. 

(The trick with this assignment is to make the two matrixes large enough so that they fit, barely fit, don't fit in certain levels of the memory hierarchy. This is perhaps 10-year old thinking but I am sure it hasn't gotten much better.) 

On Feb 16, 2014, at 11:43 PM, Bloch Stephen wrote:

> I'm teaching a junior-level algorithms course, and assigned my students to implement three different matrix-multiplication algorithms and benchmark them.  So I had to do it myself too :-)
> In untyped Racket, most of the runs took seconds or minutes, but others (only twice as large, so I would expect roughly 8 times the run-time) have been running for days without result.  So I decided to try porting everything to TR, to see whether that saved any significant (presumably constant-factor) time.  Once I'd gotten everything to type-check in TR, I revised the untyped version to resemble the typed version as closely as possible.
> The results so far: the typed version is taking 20-30 times longer than the untyped version.  Huh?
> I've installed the optimization coach, and am looking through its recommendations.  For the untyped version, it suggested a couple of inlinings.  For the typed version, nothing.
> Stephen Bloch
> sbloch at adelphi.edu
> GPG  key at http://adelphi.edu/sbloch/sbloch.pubkey.asc
> ____________________
>  Racket Users list:
>  http://lists.racket-lang.org/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.racket-lang.org/users/archive/attachments/20140217/9890e8dc/attachment.html>

Posted on the users mailing list.