[game_edu] Feedback request on this programming lecture snippit

[Richard Fleming]

Looking for general feedback on this, good/bad/what are you smoking/ect, before I inflict it on my students


Macro and Micro Optimization

Many times on the net you will find articles and blog posts proclaiming the evils of optimization.  “Don't do it!  Not until you absolutely must!” They all cry.  “Profile and measure before you try!”  is another rally cry.  And this is good advice, but only for certain kinds of optimizations.  The problem is that very few of them make a distinction between kinds of optimization, which is what I will do here.

Macro Vs. Micro
I sort optimizations into two groups: Micro and Macro.  When you read a blog post on the evils of early optimization, they are talking about Micro optimization.  This is where you take small and isolated sections of code, and change them so they run as fast as possible.  This kind of optimization tends to create code that is very hard to read, complex, and bug prone unless you are very careful.  It can also take a lot of effort and time.  So the net is correct on this issue, avoid it until performance degrades bad enough that you feel you must.  Also, measuring or profiling is critical.  Humans are bad at guessing where performance bottlenecks are.  You can actually make things worse if you 'optimize' a spot that doesn't need it.

Macro optimization however is different.  You want to do this from the very start.  Macro optimization is optimization at the architectural level.  Do I use a list or a queue for this data structure?  How about trees, do I need a red-black tree, or perhaps a b-tree?  The point here is to examine what you need to do and come up with an architecture that makes sense will work well.  This involves knowing the language and target platform really well.  A macro optimization that works well for a desktop app might be horrible for a mobile target, and vice versa.

Now you might look at this and wonder: Doesn't this sound a lot like micro optimizations?  I mean, it's a lot of effort and time, and didn't you just finish saying humans are bad at guessing performance bottlenecks?  Well, yes and yes, but some things are different here.  First and foremost, you should be doing a technical design doc for your game anyways.  If you just grab a regular design document and start coding, you are begging for trouble.  You really, really need to sit down and do some thinking on how you want to structure the code.  As for performance, we know algorithmic performance.  Called 'Big O Notation', you can tell exactly how an algorithm will perform in different cases.  Lists are constant time for insertions.  The cost of bubble sort increase exponentially with added items.  These costs are known and proven.  Keeping these things in mind will help prevent the need for micro optimization.  Then there are the platform issues that I mentioned.  In C#, you have value types like integers, and then everything else is an Object.  If you pass an integer into a parameter that expects an Object, it 'boxes' the integer by creating a new Object to contain it.  Not that big of a deal (in most cases) on a desktop, but this will cripple you in the Compact Framework if it happens even moderately.

So the take-away:  How should you optimize?  Always start with macro optimization.  If your code is an unholy mess of spaghetti logic, no amount of micro optimization can save you.  As you implement the code, continually measure the performance.  Once you notice the performance dropping below acceptable levels, break out the profiling tools and find that one line that is draining all of your performance.



Dragons are awesome and dragons are fun.
But if you get one angry...
then you'll be well done.
---------------------------------------
Richard Fleming
RC 332G, x4838
Interested in creating computer games?
Then check out JCCC's Game Development AA Degree!
http://www.jccc.edu/home/depts.php/1287

The information contained in this e-mail and any attachments thereto ("e-mail") is sent by the Johnson County Community College ("JCCC") and is intended to be confidential and for the use of only the individual or entity named above. The information may be protected by federal and state privacy and disclosures acts or other legal rules. If the reader of this message is not the intended recipient, you are notified that retention, dissemination, distribution or copying of this e-mail is strictly prohibited. If you have received this e-mail in error please immediately notify JCCC by email reply and immediately and permanently delete this e-mail message and any attachments thereto. Thank you.