|
Forcing Functions To Be Called Non-inlined
Submitted by |
Occasionaly, one needs to call a function (usually small one) and make
sure that it'll not be inlined. Often such little functions are auto-inlined
by the optimizer, when compiling in release/optimized mode. Although there is
a way to prevent function from inlining by prototyping it with the _cdecl
keyword, sometimes all you need just specific calls to be non-inlined.
Here is the macro:
#define NON_INLINED_CALL(a) (((int)(a)+1)?(a):(0)) |
Now, by forcing some fake calculation (int(a)+1) we stop the compiler
from optimzing given line, also by using v ? (a) : (0), we are forcing the
compiler to use the _cdecl way of calling the function.
Example:
static int test(int a, int b, int c)
{
return a*b*c;
}
#define NON_INLINED_CALL(a) (((int)(a)+1)?(a):(0))
extern int results[2];
extern int a, b, c;
extern int d, e, f;
void test2( void )
{
results[0] = test(a,b,c); // Eventually with good optimizations this will be inlined
results[1] = NON_INLINED_CALL(test)(d,e,f); // No inlining here
} |
Under MSVC we can construct simple and faster macro
#define NON_INLINED_CALL_MSVC(a) ((a) ? (a) : 0) |
but the latter will not stop GCC optimizer, and we'll not get the results we wanted to.
P.S. I didn't test any other compilers - only GCC 2.99(ps2-ee one), and MSVC6.00
|
The zip file viewer built into the Developer Toolbox made use
of the zlib library, as well as the zlibdll source additions.
|