Prev: Creating new codecvt facets
Next: Is the Mixin pattern accepted in all camps?
From: Larry Evans on 5 Jul 2010 23:45
Is there any reason, other than compiler optimization limitations,
why, in the following code, dispatch_funvec should be any slower
than dispatch_switch. Are current implementations of dispatch_funvec
slower just because it's harder to optimize than the dispatch_switch.

TIA.

-regards,
Larry

-[---- cut here ----
unsigned value=0;

#define MAX_INDEX 3

template
< unsigned Index
>
unsigned fun(void)
{
value+=Index+1;
return value;
}

typedef unsigned(*fun_type)(void);
#ifdef CONSTEXPR_ENABLED
constexpr
#else
const
#endif
fun_type
fun_vec[MAX_INDEX]=
{ fun<0>
, fun<1>
, fun<2>
};

int dispatch_funvec(unsigned index)
{
return fun_vec[index]();
}

int dispatch_switch(unsigned index)
{
switch (index)
{
case 0:
return fun<0>();
case 1:
return fun<1>();
case 2:
return fun<2>();
default:
return -1;
}
}

-]---- cut here ----

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

From: Bart van Ingen Schenau on 6 Jul 2010 04:12
On Jul 6, 4:45 pm, Larry Evans <cppljev...(a)gmail.com> wrote:
> Is there any reason, other than compiler optimization limitations,
> why, in the following code, dispatch_funvec should be any slower
> than dispatch_switch. Are current implementations of dispatch_funvec
> slower just because it's harder to optimize than the dispatch_switch.

Basically, yes. dispatch_funvec is very much harder to optimise than
dispatch_switch, because indirect function calls (through a function
pointer) are very hard to expand inline. With functions as small (and
similar) as in your example, that could make a noticeable difference.

Bart v Ingen schenau


--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

 | 
Pages: 1
Prev: Creating new codecvt facets
Next: Is the Mixin pattern accepted in all camps?