Why std :: tr1 :: shared_ptr <>. Is reset () so expensive?

Question

Profiling some code that uses shared_ptrs heavily, I found that reset () was unexpectedly expensive.

For instance:

 struct Test { int i; Test() { this->i = 0; } Test(int i) { this->i = i; } } ; ... auto t = make_shared<Test>(1); ... t.reset(somePointerToATestObject); 

Tracking reset () on the last line (in VC ++ 2010), I found that it was creating a new reference counting object.

Is there a cheaper way that reuses an existing link count and doesn't bother a bunch?

Answer 1

In general, you cannot reuse an existing reference counter because there may be other shared_ptr or weak_ptr .

If you can create somePointerToATestObject using make_shared() , then the implementation can use the same heap distribution for both ref and object references. This will save you from one of the heap allocations.