🚀 go-pugleaf

In article <4e54vc$7sk@orac.mon.rnb.com>, hdavies@kzin.mon.rnb.com (Hugh J.E. Davies) writes:
|> In article <4e3a2u$eoa@wcap.centerline.com>, chase@centerline.com (David Chase) writes:
|> >> Randal L. Schwartz (merlyn@stonehenge.com) wrote:
|> >> : Tom Christiansen has had a long-standing bet that someone can't give
|> >> : him a program in C that he can't make run no more than "e" times
|> >> : slower (about 2.8 for you non-math-geeks) in Perl.  So far, no one's
|> >> : done it.
|> >
|> >Is this really true?  What's the bet?  Is a particular piece of hardware
|> >specified?  On modern RISC machines, carefully crafted C can multiply
|> >large matrices at >50% of peak Mflops.  To compete in Perl, I think
|> >you'd have to run the "contest" on a particularly crufty piece of
|> >hardware.
|> 
|> Forgive my ignorance, and please correct me if I'm wrong, but if
|> the 'C' is running faster because of faster hardware, wouldn't perl
|> also? What would make a 'C' program run faster but not the 'perl'
|> interpreter?

Super-mini's (Convex amoung others) often have specialized vector and matrix 
hardware that allow it to perform the same operation on large portions of
the vector with a single instruction cycle.  This requires specialized 
optimizing compilers.  PERL wouldn't take advantage of the specialized 
hardware/instruction set and thus would have a worse benchmark.

My guess is that the generic term "modern RISC machines" is probably wrong.
Tom's original claim was probably for a UNIX desktop box and would
still hold true.

Murf

-- 
John A. Murphy (better known as Erin's dad)  jam@philabs.research.philips.com
345 Scarborough Road
Briarcliff Manor, NY 10510    One one-trillionith of a surprise: picoboo
(914)945-6216 		      millihellen: The beauty needed to launch 1 ship