You will (almost) never realize the execution speed of C in perl....
That depends on what you're doing. If you want to compare the languages themselves (okay, you can use C's standard library), I can suggest several algorithms and data structures and tasks which will be as fast in Perl as in C.
... which is why people do not write (eg) 3D API bindings for perl.
A far larger problem is data marshalling instead of opcode dispatch.
I can suggest several algorithms and data structures and tasks which will be as fast in Perl as in C.
I keep seeing people say this, but I've never managed to see it happen. Usually Perl is fast enough, but I've never managed to outperform the equivalent C program. Could you please provide some examples of what is faster in Perl than C, because I really am interested.
Many of these algorithms are for arbitrary string searching; it's difficult to outperform Perl's regex engine. (Please note that the portion of my post which you did not quite implied that you can't pull in PCRE.) I suspect, but can't immediately prove, that code heavy with malloc/free temporary data may reach a point where it's faster in Perl than in C, but that depends on overcoming the overhead of op dispatch in Perl.
I wrote bindings for OGRE. I found them to be fairly fast, actually, though of course most of the execution is done in C++ land.
You probably don't mean Perl specifically, but interpreted languages generally. But from what I see, a lot of times there are no bindings for Perl when there are for Python and Ruby. I think the reason for not writing bindings is not the speed (Perl is probably faster than Ruby, I don't know about Python) but rather the ease of writing the bindings. For example, for Python there is Py++; we don't have the equivalent of that for Perl.