[plt-scheme] nonstandard installation

From: Matthew Flatt (mflatt at cs.utah.edu)
Date: Fri Nov 28 12:09:45 EST 2008

Are these Sparc machines or x86 machines? Sparc would explain slowness
on Robby's test, because the JIT doesn't currently support Sparc ---
and Robby's test is a worst-case example for JIT versus non-JIT
execution.

For start-up, my guess is that the difference is in where the bytecode
files are stored, since start-up time is dominated by loading in
bytecode. Another possibility (which also applies to the difference in
the scaled-down version of Robby's test) is a difference between using
shared libraries versus static builds, if one the other version was
configured with --enable-shared.

Matthew

At Fri, 28 Nov 2008 11:03:17 -0600, "Robby Findler" wrote:
> I think Matthew will have to say more, but the first test you were
> doing was measuring the startup time of mzscheme, which can be
> influenced by lots of things, but different things than the test I
> suggested, which is measuring computation time.
> 
> Robby
> 
> On Fri, Nov 28, 2008 at 11:01 AM, Prabhakar Ragde <plragde at uwaterloo.ca> wrote:
> > Robby Findler wrote:
> >>
> >> Where are you measuring the slowdown? Something like this:
> >>
> >>> (time (let loop ([n (expt 10 8)]) (unless (zero? n) (loop (- n 1)))))
> >>
> >> ? Or some other task?
> >
> > Curiouser.
> >
> > The above takes 372/374/0 on my MacBook Pro (2.33 GHz Intel Core 2 Duo).
> >
> > It takes 101920/111723/0 for the installed Solaris version.
> >
> > When I crank it down to (expt 10 7) and run it, I get 9730/9897/0 for the
> > installed version and 11210/11226/0 for the locally-built version.
> >
> > I think the lesson is to not use Solaris. Pity that that's where all our
> > autotesting is currently done. --PR
> >
> >
> _________________________________________________
>   For list-related administrative tasks:
>   http://list.cs.brown.edu/mailman/listinfo/plt-scheme


Posted on the users mailing list.