boost: Initializing shared pointer reset vs make_shared

Question

Apart of the size of generated code, what's the difference between using reset() for initializing a shared pointer over the function make_shared()?

Case 1 by using reset()

boost::shared_ptr<A> pA;
pA.reset(new A());

Case 2 by using make_shared()

boost::shared_ptr<A> pA;
pA = boost::make_shared<A>();

In general, is it a good practice to use reset over make_shared to reduce the size of executable?

Answer 1

reset(new T(...)) allocates a heap block, constructs the object, allocates a new heap block for the reference counter and initializes the reference counter.

make_shared<T>(...) allocates a heap block slightly larger than required for the object and constructs the object and the reference counter in the same heap block.

The chance is high that make_shared() runs faster and requires less memory.

But there is a small drawback if you are using an IDE like Microsoft Visual Studio: Intellisense is not able to show you the names of the parameters used in the constructor. The code is working correctly but editing the make_shared() call is uncomfortable.

Answer 2

make_shared<T> creates the reference counter in the same chunk of memory which is allocated for T. It's an optimization. reset does not do this.