From: Joost Diepenmaat <joost@...>
Date: 2005-07-16T13:05:48+09:00
Subject: Re: [QUIZ] Sampling (#39)

On Sat, Jul 16, 2005 at 12:41:23PM +0900, Joost Diepenmaat wrote:
> On Sat, Jul 16, 2005 at 11:51:27AM +0900, James Edward Gray II wrote:
> > On Jul 15, 2005, at 11:38 AM, jason r tibbetts wrote:
> > 
> > >James,
> > >
> > >What are your machine's specs, just for comparison's sake? I don't  
> > >want to get too discouraged by my 5+ year-old Sun beater  
> > >workstation. :)
> > 
> > Dual 2.0 Ghx G5.  2 Gigs of RAM.
> > 
> > >I've also found that the printing of the results (in the form of an  
> > >array) is taking several minutes in and of itself. Since optimizing  
> > >the printout probably isn't the point of the quiz, can anyone  
> > >recommend something faster than arr.each{ | elem | p elem }
> > 
> > IO is certainty a part of this quiz.  The benchmark library is your  
> > friend...  ;)
> 
> One of the interesting findings for me was that
> 
>     puts array_of_strings
> 
> is *very* much faster than
> 
>     puts array_of_integers
> 
> Joost.

And by the way:

 > time ./sample2   5_000_000 1_000_000_000 >big.txt 
 real    2m5.395s
 user    1m39.892s
 sys     0m11.576s

 > head big.txt 
 141
 441
 854
 876
 968
 1123
 1276
 1317
 1557
 1587

 > tail big.txt 
 999998078
 999998164
 999998558
 999998626
 999998650
 999998814
 999998912
 999999127
 999999160
 999999385

Slightly more than 2 minutes wallclock time. That means it's just fast
enough to keep me from getting bored and killing the process :-) I
managed to shave off about 8 minutes run time by trying a couple of
different strategies with the IO and int -> string conversions. The 
"sample algorithm" itself is the second-dumbest I could think of - very
easy to prove correct, though :-)

system info:

Single Intel(R) Celeron(R) CPU 2.40GHz w / 1GB of memory running Linux
2.6.11

Joost.