On Mon, 28 Nov 2005, Cameron Hayne wrote:
[...]
> That looks quite similar to the "Nifty Parallel Count" method on the
> page by Gurmeet Singh Manku that I referred to earlier. In his tests,
> this method came out a lot slower than the pre-computed array method
> I mentioned.
>
Beware how you benchmark the speed of alternative routines. Table based
methods are favoured by standalone loops that repeat a short routine over
and over. That's because the table is pretty much guaranteed to be cache
resident (if it fits) during the whole test.
The same routine as used in an actual application might not enjoy such
luxury. (Of course, the standalone loop can be the exact use of the code
in an application; just sayin' one should verify that the benchmark
models the real world accurately.)
Holger
_______________________________________________
Do not post admin requests to the list. They will be ignored.
PerfOptimization-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/perfoptimization-dev/email@hidden
This email sent to email@hidden
