Re: Contest - hardware requirements
Christian Cuvier wrote:
>
> Some optimisations may require some assumptions on the benchmark hardware.
Yes, but if you are trying that as your 'edge' then this contest isn't
really for you. I strongly suggest that algorithm optimisations will give
you much more leverage than the way you use the hardware.
> Since there's hardly any access to OS functions that would give some needed
> indications, could you, Derek, state:
> - any minimal amount of available free RAM the submitted program can count on
> at run time;
90MB +- 20MB
But I doubt you can get away without using the page file at some stage.
> - are you using a 25-line or 43/50-line display?
25 line, full screen (non-windowed) DOS display. However all output is
being redirected to a file and the programs thus do not display anything
to the screen.
Currently, when I do multiple runs of a program, the variance between runs
is < 1% (0.05 seconds for the longest ones) with most runs being the same
duration. Usually the first run is a tad longer than others. I said I
was going to do five runs, throw away the two longest and average the
remaining three, but it has worked out in practice that the shortest
duration of the five runs is the one I use. I figure that if the program
ran that fast at least once, then it can run that fast.
And by the way, there's no sheep station riding on the outcome this.
--
Derek Parnell
Melbourne, Australia
|
Not Categorized, Please Help
|
|
