[plt-scheme] From hash-table to generator

From: Jens Axel Søgaard (jensaxel at soegaard.net)
Date: Sat May 12 05:23:11 EDT 2007

Eli Barzilay skrev:
 > On May 11, Jens Axel Søgaard wrote:
 >> In my sample application I get:
 >>    Above version:    cpu time: 23594 real time: 25469 gc time: 3389
 >>    Call/cc version:  cpu time: 18125 real time: 18344 gc time: 11515
 >>    List version:     cpu time: 1843 real time: 1843 gc time: 578
 >> The conclusion must therefore be, that you need to
 >> implement hash-table->generator as a primitive, if you want
 >> to avoid Unnecessary allocation.
 > These numbers show that context switches are expensive -- and the ones
 > involved with threads are a little more expensive than those with
 > continuation.  Allocating a list removes such switching, but it leads
 > to allocation (which is not a big deal, but my guess is that you're
 > worried about really big tables).


 > So how about a hybrid approach: use either the continuation or the
 > thread+channel code, but pass along blocks of N values from the table.
 > This way you're protected from really big tables, and with something
 > like N=1000 the context switching would be reduced to a minimum, and
 > the run time would be very close to what you'd get from a primitive
 > iterator.

Excellent suggestion.

Jens Axel Søgaard

Posted on the users mailing list.