[Shootout-list] Re: sumcol - when the correct answer is wrong :-)

Brent Fulgham bfulg@pacbell.net
Tue, 21 Sep 2004 10:34:27 -0700 (PDT)


--- Brian Hurt <bhurt@spnz.org> wrote:
> > Maybe we should have a new test
> > - where success was either providing the right
> > answer, or failing at runtime (see what 
> > performance is like with runtime checks enabled); 
> > and failure was silently ignoring the error ;-)
> 
> Was that an error?  There are good reasons for a
> language to define that behavior.  Unless the test 
> is "how well does the language handle big 
> ints?"

I disagree.  I don't think it is reasonable behavior
to allow integer overlow, unless the developer
explicitely codes to allow this behavior.  Silently
handling a catastrophic condition like this is why
cancer patients receive overdoses of radiation,
satellites crash into planets, and corporate loan
balances turn into "credit" values.

> Unless you're testing the performance of big ints,
> I'd recommend no test require an integer larger 
> than +/- 2^30 to be represented.

Most tests are confined to 32-bit values.  However,
because modern systems are so fast we can't capture
data at a fine enough resolution on certain tests
without resorting to very large values.

-Brent