[Python-Dev] Par-for/range

[Thomas Wouters]

On Wed, Jul 12, 2000 at 10:03:53PM +0100, Michael Hudson wrote:
> Paul Prescod <paul@prescod.net> writes:
> > Thinking some more...
> > 
> > If we had first class range syntax then it would be relatively easy to
> > recognize the pattern
> > 
> > for i in [x:y:z]:
> > 	....
> > 
> > Then we could generate a special FOR_INTS bytecode. I've heard that "C"
> > has a pretty efficient implementation of that construct. <wink>

> Hmm, yes, but how on earth would you implement that in ceval.c?  You'd
> need another eval loop inside the code for FOR_INTS!  (unless I'm
> being dense, of course).

Nah. We can't use the C for-loop for this, unless the bytecode interpreter
is entirely rewritten. The FOR_INT opcode would be something like:

on entry, stack contains r, i
   where r is range-object specifying (start/)step/end, i is loop counter

 i is incremented by step, checked against end, and if loop hasn't ended
 yet, r, new 'i' and old 'i' are pushed on the stack, so it contains:

r, i+step, i

and the rest of the bytecode can remain the same (assign top value to name,
etc.)

> To invent more statistics on the spot, I'll declare that loop overhead
> is a problem in about 3.4% of all interesting loops; what we need is a
> way to execute the body quicker...

define 'problem' ;) If we can speed things up while keeping the
functionality the same, the syntax clearer and the code only slightly more
complicated, why not ? :-)

Granted, though, if the issue is overall code speed, this isn't going to cut
it.

-- 
Thomas Wouters <thomas@xs4all.net>

Hi! I'm a .signature virus! copy me into your .signature file to help me spread!