Inlining for optimization

[areza123]

For optimized coding I use the following guideline:
Inline methods that - have no loops, are non virtual, are non-recursive, are not (esp.derived) class constructors / destructors and do not make calls to virtual functions.

The important point missing here is a specification on the number of lines of code the inlined function should contain. Initially I thought I would inline only those methods that contained at most 7 lines of code. This includes the length of functions that would be called from my inlined function. Is this spec ok ? I am looking for a second opinion - may be based on some mathematics.

Cheers

 

[PerFnurt]

I follow this small guidline (not based on math, sorry):

Based on the paradigm of encapsulation:
Use implicit (ie implementing in the class header) inline if and only if it is a minimal function, ie basically just a return statement. I'm definately against cluttering the class' interface with implementation details, but I think simple returns are reasonable.

Based on focus on what is important:
Use "normal" inline if it is proven or very likely that the gain of inline is measurable. Typically on a few time critical every-so-often called functions.
If they are really-really time critical - consider if inline assembler will gain performance.







/Per

_{if (typos) cout << &quot;My fingers are faster than my brain. Sorry for the typos.&quot;;}