[wp-hackers] This is a Performance Enhancement? Clarification

[Jacob Santos]

Well, I think I was trolling in my last email. The point is that I grow 
tired trying to argue with people who assume that just, because 
$end-$start = faster than another $end-$start = slower that somehow 
their knowledge of performance is valid and should be accepted. Or worse 
yet, since they are supposed to know PHP well that obviously they know 
something as trivial as performance. That just because someone may not 
know everything about everything related PHP, that his facts and 
evidence are somehow invalid.

I would be more willing to accept if perhaps people didn't try to say 
that run-time compilation equals faster than compile-time compilation. 
Yeah, everyone knows that after the script is already compiled that it 
is faster to compile a function or 20 after the script is running 
(</sarcasm>). Doesn't matter that having to stop the script execution in 
order to compile the new code doesn't make sense that it speeds up 
anything. How does that make sense, that it will speed anything up? [1]

I'm looking for an valid argument with performance measurements and 
proof from either PHP engine programmers, Opcode Cache programmers, etc 
and people that know what they are talking about to explain to me that 
stopping execution to go back to the compile stage is faster than not 
doing that second step. If we can agree that it is not, then the code in 
the last post should never be used for performance and should never be 
referenced that it should be used for performance. Any other reason 
would have been acknowledged. With all of the differing statements out 
there, it makes it really hard to battle with people when you know you 
are correct and the people you are arguing with are misinformed.

Jacob Santos


[1] Of course, as I think about it, a lot of the WordPress code base 
does just this and while it runs slower than other code. As well as 
other code bases that load files dynamically (MVC frameworks). The 
argument is whether or not the advantage of losing a few milliseconds is 
worth the additional feature set that is allowed from the lost.