On 13-Oct-06, at 2:34 AM, Ronald Vogelaar wrote:
----- Original Message ----- From: "Frank Condello"
<[EMAIL PROTECTED]>
For some apps you can see improvements by reducing function calls
in the order of thousands or even just hundreds, never mind
millions. I've seen real-world measurable speed ups by manually
inlining methods and otherwise reducing function calls by using
pointers. I mean "seen" literally, saving just a few milliseconds
per frame in a game really adds up for example - even a seemingly
modest 10FPS jump can turn an unplayable game into an enjoyable
experience.
I deliberately used an example of which the code (incrementing a
counter) is among the simplest to achieve for a compiler.
Therefore, I figured, measured time would most effectively and
truthfully render the overhead of a function call over longhand code.
Now, I'm not familiar with gaming (I don't even play computer
games, let alone create them), and I'm not sure I follow you when
you speak of reducing function calls by using pointers to create a
significant speedup. I just don't see where the speedup would have
to come from. Could you give an example?
I'm referring specifically to memoryblocks:
// function call
block.SingleValue(n) = 0.0
// no function call
blockPtr = block
blockPtr.Single(n) = 0.0
For 3D animations you're often pushing around arrays of hundreds or
thousands of vectors every frame, and in my experience the direct Ptr
manipulation is easily 5-10 times faster. Admittedly Memoryblock
accessors like SingleValue are more expensive than a simple
"increment" method, but I'd imagine inlining all my Vector and
Quaternion class functions would yield immediate benefits. It's not
uncommon to write math libraries as a set of macros in C code for the
same reasons.
Frank.
<http://developer.chaoticbox.com/>
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives of this list here:
<http://support.realsoftware.com/listarchives/lists.html>
