Alex Tweedly wrote:
On 31/10/2014 04:04, Richard Gaskin wrote:
What I find most remarkable is the way your benchmark makes the value
of this ID caching seem rather trivial.
...
It is a trivial example - but the speed-up of 50X seems significant.
In relative terms it certainly seems so, but in absolute time just how
much is being saved? That is, what is the time required to resolve an
object ID without caching?
This reminds me of a test I ran yesterday on a different issue, similar
only in that whether or not it's a valuable thing to do may differ
depending on whether you're looking at total-time or time-per-iteration:
I was measuring how much overhead is involved in making a call to a
function relative to performing the same code in-line.
The total time saved was 15.897949 ms, which seems a worthy number until
we consider that it was running for 100,000 iterations.
In a single iteration it was only saving 0.000159 ms, which for the
purposes of the test made it clear that the performance "penalty" of
factoring large handlers is probably not worth the time we spend trying
to wrap our head around really long complex handlers.
With ID caching we have a nuance that further qualifies its value: it's
very specific, caching only ID references but not the more commonly-used
forms like names or, as often used in loops, ordinal numbers.
That said, I tend to be somewhat obsessive about performance because I
see every nanosecond saved as a penny in a bank account; over time I can
save up enough to have some milliseconds free to spend on new features
like real-time display updates or background processing.
So I'm always grateful for anything that boosts performance in the
engine, even those that may seem small, since over the long run
everything adds up.
Probably worth adding it to the compare-engine-versions benchmark; I'll
do that once a framework / sample of that happens.
I hope we can get that project going. It'll be very helpful for many
purposes.
--
Richard Gaskin
Fourth World Systems
Software Design and Development for the Desktop, Mobile, and the Web
____________________________________________________________________
ambassa...@fourthworld.com http://www.FourthWorld.com
_______________________________________________
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode
