How to let a loop run for a while before checking for breakcondition?

[matt at vazor.com]

IMHO you're better off avoiding all this...  It makes the code
unnecessarily complicated when you're not sure if this is a performance
bottleneck or not...

Common advice is to write the code as simple as possible first, then go
back and optimize as needed using the profiler.  This is where
extending in C becomes useful, but don't go to that level if it isn't
really needed.

KISS -- Keep It Simple Stupid

thx

m

Hendrik van Rooyen wrote:
> "Claudio Grondi" <claudio.grondi at freenet.de> wrote:
>
>
> | Diez B. Roggisch wrote:
> | > Claudio Grondi schrieb:
> | >
> | >>
> | >> Sometimes it is known in advance, that the time spent in a loop will
> | >> be in order of minutes or even hours, so it makes sense to optimize
> | >> each element in the loop to make it run faster.
> | >> One of instructions which can sure be optimized away is the check for
> | >> the break condition, at least within the time where it is known that
> | >> the loop will not reach it.
> | >>
> | >> Any idea how to write such a loop?
> | >>
> | >> e.g.
> | >>
> | >> counter = 2*64
> | >>
> | >> while counter(BUT DON'T CHECK IT THE FIRST ONE HOUR LONG):
> | >
> | >
> | > now = time.time()
> | > while time.time() - now < 3600.0 or some_other_condition:
> | >    ...
> | >
> | >
> | > The short circuiting of the or will prevent the execution of
> | > some_other_condition.
> | >
> | >>   ... do something ... # and decrease the counter
> | >>
> | >> Thanks for any hint, but in particular if related to timers on the
> | >> Windows 2000/XP system I am mainly working with.
> | >>
> | >> What do you think about this idea? Does it make sense?
> | >
> | > What idea?
> | This one you haven't probably got from what I have written.
> | I thought, that the introductory text gives enough context to be able to
> | see what I mean, but I was apparently wrong.
> |
> | The idea is to speed up a loop by using a timer interrupt interfering
> | with the loop, so that only after the timer interrupt would occur, the
> | loop will start to check its break condition in each iteration.
> | No checking of any kind in the loop should happen up to that time to
> | minimize the number of operations in each iteration within the loop
> | itself (i.e. the loop more or less won't know, that there is a timer on
> | its way to change the loops behavior at a later time).
> |
> | I hope this above helps to understand what I would like to achieve.
> |
> | Claudio Grondi
>
> I don't think this is usefully possible in python - the problem is that you will
> simply replace one check - The expiry of the counter - with another - to see if
> the interrupt has occurred already -
>
> That said - the way I would do it would be to do something like this (in
> horrible pseudo code):
>
> loop_start:
>     do_something()
>     jump loop_start
>     if counter > end_value:
>         break
>     jump loop_start
> loop_end:
>
>
> Interrupt_routine:
>     replace the first jump to loop_start with a bunch of no - ops
>     return
>
> I don't think you can do this in python - it involves altering the running
> loop - but hey maybe I can learn something here...
>
> This example sort of exposes the break for what it is - a jump statement in
> disguise - "look you cant recognise me - I am wearing dark glasses" - and
> "continue" is exactly like that too - the only difference is that the one jumps
> to the end, and the other to the beginning of the loop...
> 
> - Hendrik