Re: drawing performance
Re: drawing performance
- Subject: Re: drawing performance
- From: Georg Tuparev <email@hidden>
- Date: Sun, 7 Dec 2003 01:16:54 +0100
On Saturday, December 6, 2003, at 04:52 PM, Shawn Erickson wrote:
From the get go of a project performance should be a major concern and
design factor however you should never make most optimizations without
measuring the affects and necessity of those changes. In others words
don't make them assuming they are needed or help. You should set
performance metrics, threshold, requirements, etc. and continuously
measure against them while developing. It will be an iterative > process.
That is what they did with Safari (they tried not to allow performance
regressions as the project progress).
In general one should favor simpler code as long as the performance of
it is sufficient.
Let us turn this last statement upside-down. After years of
programming, I slowly start realizing that thinking about optimization
is a way of actually simplifying the design and making the code more
expressive (I use the word "thinking" instead of "doing" on purpose). I
was surprised by my own realization - after all this is contrary to
everything I was ever told about programming. And then I make some
experiments with my team. Some of the stories (XP) were done with the
performance in mind, and for the others, we wrote non-optimized code
first, then we wrote a story that sets a performance target, and we
fiddled, refactored, and redesigned until we met the target. In over
70% of the cases the subjective view of the team was that the final
design was simpler and clearer if we considered performance from the
beginning....
-- georg --
"War is God's way of teaching Americans about geography."
Ambrose Bierce, writer (1842-1914)
_______________________________________________
cocoa-dev mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/cocoa-dev
Do not post admin requests to the list. They will be ignored.