Re: inline functions

junky_fellow@xxxxxxxxxxx wrote:
> What is an inline function ?

A function declared with the function specifier "inline". It's
new in the C99 standard (but a lot of compilers allowed it as
an extension already prior to that).

> When should we declare a particular function as inline ?

When you have a short function where the overhead of calling
the function is too large, i.e. if there's more time spend in
calling the function and returning from it than in what the
function actually does. Potential candidates could be e.g.

inline int_abs( int x ) {
return x >= 0 ? x : -x;
}

> If a function is declared to be inline, does that mean
> that the entire code for that function is subtituted there

Not necessarily. The only thing the standard says is that:
"Making a function an inline function suggests that calls to
the function be as fast as possible.". Please note the word
"suggests", i.e. the compiler is free to do nothing. So what
the compiler exactly does with inlined functions is left to
the discretion of the implementor. One way to do it is, of
course, to put the code for the function everywhere where
it's called.

> just like macros? Then why not use macros ?

Functions and macros are completely different beasts. A
macro is nothing else than a text replacement that happens
at a very early stage of the compilation. Therefore macros
can lead to very strange results unless used very carefully.
Consider e.g.

#define ABS( x ) ( ( x ) >= 0 ? ( x ) : ( -x ) )

This looks quite nice and simple, doesn't it? But try to call
it like this

int a = -1;
int b = ABS( a++ );

and think about what's going to be the result. (Hint: you will end
up with 'b' set to 0 and 'a' being 1. You would expect it to be the
other way round, wouldn't you? To figure out why do the text replace-
ment for the macro by hand and see what's the resulting code.) That's
something that won't happen with a function (inlined or not), since
function arguments are evaluated before passed to the function.

Regards, Jens
--
\ Jens Thoms Toerring ___ Jens.Toerring@xxxxxxxxxxxxxxxxxxx
\__________________________ http://www.toerring.de
.

Relevant Pages

  • Re: Alternatives to #define?
    ... >> But macros can be difficult to follow too. ... >> large parts of the code by switching a compiler flag. ... > inline void doStuff() ...
    (comp.lang.cpp)
  • Re: inline methods
    ... "inlined" by the compiler if it is declared final (necessary, ... If I declare a class to be final, this implies that all of the methods ... Will javac inline the methods? ...
    (comp.lang.java.programmer)
  • Re: Alternatives to #define?
    ... > but for example logging macros and testing macros can be difficult to read ... But I do like how the compiler has an opportunity to remove ... inline void doStuff() ...
    (comp.lang.cpp)
  • [PATCH] [9/11] x86: Replace NSC/Cyrix specific chipset access macros by inlined functions.
    ... With the macros a line like this fails: ... * NSC/Cyrix CPU indexed register access macros ... static inline void sync_core ... +static inline void setCx86 ...
    (Linux-Kernel)
  • Re: [PATCH 1/1] i386: Geodes TSC is not neccessary to mark tu unstable
    ... Replace NSC/Cyrix specific chipset access macros by inlined functions. ... * NSC/Cyrix CPU indexed register access macros ... static inline void sync_core ... +static inline void setCx86 ...
    (Linux-Kernel)