🚀 go-pugleaf

Gert Doering wrote:

> Tom Christiansen <tchrist@mox.perl.com> writes:
> 
> >Oh my! But perl really is very slow to start up: on my abysmally slow
> >and overloaded sparc classic, it takes 4.58 user seconds and 14.22
> >systems seconds to run an empty perl program -- ONE HUNDRED TIMES.
> >That's a whopping 0.0458 user seconds and 0.1422 system seconds per
> >run.  Obviously this is too much overhead for a small program.
> 
> Ummm... on my SCO Unix machine (486dx2-66, 40 Mb RAM), it takes about 5-10
> seconds to start up *per invocation*... now *that* is what I'd call slow
> startup :(

Wow, something is seriously wrong with your machine!  On my housemate's
SCO system (3.2v4.2, 486/66DX2, 32MB RAM), starting perl 5.000 100 times
takes 1.68s user time, 2.99s sys time.  That's over 4x as fast as Tom's
SPARC Classic.  Over 100x faster than your system.  Uh, is that realtime
or CPU time?  Run `timex perl </dev/null`.  Maybe it's a DNS problem
causing perl to pause during startup?

Maybe we'd better define our benchmarks; an "empty perl program" could
have several meanings.  I ran:

  timex ksh -c 'i=0; while [ $i -lt 100 ]; do let i=i+1; perl; done' </dev/null

The quoted times are the average of four runs (there was little variance
between the runs).

>Bela<