Rogers and BitTorrent: another datapoint

[Christopher Browne]

On 11/1/06, Andrej Marjan <amarjan-e+AXbWqSrlAAvxtiuMwx3w at public.gmane.org> wrote:
> In a nutshell: running natively compiled code through a smart optimizing
> VM can make it faster than native.

There are other lessons also available...

CMUCL contains not one, but TWO compilers:
 - One, called Python, which generates optimized binaries for several
architectures.
 - Another which generates a bytecode-compiled form which is very
compact; about 1/6 the size of machine code.  (After all, machine code
has to do lots of register allocation work as well as code alignment)

The bytecode form isn't as quick, but it is SMALL.  You can thus
partition your application into portions that are compiled in
different fashions.  Generally, you'd use bytecode for things that run
infrequently, and compile to machine code for tight loops that you
might even want to expand by loop unrolling and such.

Horses for courses...  The infrequently used code doesn't chew much
memory, leaving more for more useful stuff.
-- 
http://www3.sympatico.ca/cbbrowne/bytecode.html
Oddly enough, this is completely standard behaviour for shells. This
is a roundabout way of saying `don't use combined chains of `&&'s and
`||'s unless you think Gödel's theorem is for sissies'.
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists