In a (C++) class I took recently the teacher mentioned that using the -> operator was a tiny bit slower than using dot notation and dereferencing your pointer manually (e.g. (*ptr)).
- Is this true, and, if so, why?
- Does this apply to C as well?
Asked
Active
Viewed 93 times
4
Marcus Hansson
- 816
- 2
- 8
- 17
-
5Since `a->b` is just syntactic sugar for `(*a).b`, I highly doubt that. – unwind Sep 24 '15 at 08:53
-
1@unwind Well since you can overload `->` and `*`, you could in principle be mischievous and insert a `sleep(n)` into the `->` implementation. But yeah … – Konrad Rudolph Sep 24 '15 at 08:54
-
3@unwind yeah, that's what I thought as well @KonradRudolph `Some men just want to watch the world burn` :) – Marcus Hansson Sep 24 '15 at 09:01
1 Answers
6
The -> operator is neither slower or faster than . operator. The fact is that dereferencing something is slower than just access to a memory location, because there is one more indirection. And this is a fact of life, either in C and C++ and any other language.
In C++, you have also references, so you can dereference something using the . too! So the problem here is not arrow-vs-dot, the problem is if the compiler can go straight to a value or if it must search for its address before.
Paolo M
- 12,403
- 6
- 52
- 73
-
`you have also references` yeah, but he definitely meant dereferencing a *pointer*, not "dereference a reference" (or whatever it's called). – Marcus Hansson Sep 24 '15 at 09:06
-
The OP (as I understood) was not talking about pointer vs. direct object access but different notations of accessing an object through a pointer, namely `p->member` vs. `(*p).member`, which are absolutely and categorically identical (in their default implementation). To Marcus: Get your money back. – Peter - Reinstate Monica Sep 24 '15 at 09:33
-
1`Get your money back` I'm Swede. University's free. But if I'd payed for it, I would definitely demand my money back. – Marcus Hansson Sep 24 '15 at 09:43
-
@PeterSchneider Oh... I guess you're right! Well, then the answer would turn into a "**NO**" :) – Paolo M Sep 24 '15 at 09:48
-
That dereferencing something is slower " this is a fact of life, either in C and C++ and any other language." While that is correct, optimizing compilers -- most notably JIT compilers which have more information at hand -- can possibly optimize away the dereferencing if they can prove that the reference keeps pointing to the same object (which in C++ for example is always true for "references"). – Peter - Reinstate Monica Sep 24 '15 at 10:20
-
As an example, using a variable instead of a literal also involves an indirection, as I layed out elsewhere (cf. http://stackoverflow.com/a/32736116/3150802). When playing around with the code I saw that the compiler would translate `int i=11; int f(){ return i*i; }` into a simple "return 121" unless I excplicitly switched optimizing off. Comparable optimizations should happen with pointers. – Peter - Reinstate Monica Sep 24 '15 at 10:25