[racket] deceptive perfomance for typed racket in integer division

From: Pierpaolo Bernardi (olopierpa at gmail.com)
Date: Wed Dec 12 08:12:54 EST 2012

On Tue, Dec 11, 2012 at 7:09 PM, Neil Toronto <neil.toronto at gmail.com> wrote:

> I've had good success running DrRacket and selecting "No debugging or
> profiling" and deselecting "Preserve stacktrace" in the Language dialog.
> Most of the time, I get performance similar to command-line Racket. I think
> the only significant difference between running a module in that environment
> and running it using command-line Racket is that in DrRacket, GC pauses tend
> to be longer because there's a larger heap.

hmmmm... every time I tried, in the past, I got significantly slower
times in DrRacket, with "No debugging" and "no stacktrace".

In this particular occasion I obtained around 20s when I tried in
DrRacket the first time. Then, I got always around 10s.

Probably I hit a big gc the first time.  hmmmm...

Posted on the users mailing list.