Re: [Haskell-cafe] Slow IO

Am Montag, 11. September 2006 18:22 schrieben Sie:

On 9/11/06, Daniel Fischer wrote:

...

The problem spec states that the input file contains about 500 test cases, each given by between 1 and 100,000 lines, each line containing a single word of between 2 and 1000 letters. So the file should be about 12.5G on average.

I don't think that that necessarily follows. Although I've never seen Not necessarily of course. But as the problem contains a warning about large input/output data, I assumed the worst reasonable case (uniform distribution).

the input file, of course, I imagine that many cases are fairly small, but designed to test the accuracy of your algorithm. A few are large (in one way or another) to test the extremes of your algorithm. But the overall size of the input file is probably much, much smaller than that estimate. (Maybe 1MB? Maybe 10MB?)

That'd be peanuts with ByteString, 1MB even without.

...

A time limit of 7s is given.

That's CPU time, and thus not including I/O, right? I couldn't find

Doesn't IO use the CPU? But seriously, I would have thought that's what 'time' lists under user and that includes IO-time as far as I can tell.

the answer on the SPOJ site. Their FAQ is a forum, but I don't see it there:

http://www.spoj.pl/forum/viewforum.php?f=6&sid=6c8fb9c3216c3abd1e720f8b4b56 82b3

In any case, the problem can't be too large; the top twenty programs all finished in under 0.35 (CPU seconds?). Even if yours is a tenth as fast as the C and C++ programs, that would be 3.5s -- twice as fast as it needs to be.

http://www.spoj.pl/ranks/WORDS1/

Of course, you don't have access to these other programs for comparison; but I hope that this gives you a better idea of the size (and manageability) of the task.

Good luck with it!

--Tom Phoenix

-- "In My Egotistical Opinion, most people's C programs should be indented six feet downward and covered with dirt." -- Blair P. Houghton