Re: [boost] Performance: shared_ptr

From: Peter Dimov (pdimov_at_[hidden])
Date: 2003-11-15 10:54:46

Matthew Wilson wrote:
> Hi Boosters
>
> I'm just doing some research on reference-counting, and contrasting
> self-counting types with externally counted types, the latter
> effected with the rather excellent shared pointer technique.
>
> [As a side question, is it true that Nicolai Josuttis was the
> inventor of the technique?]

Which specific technique do you have in mind?

> Naturally, one of the shared pointer's I've used is
> boost::shared_ptr. The measurements have demonstrated some surprising
> performance characteristics for shared_ptr in particular, and I want
> to solicit the opinion of the expert's here before I make any bold
> claims, as some results are just too surprising for me to trust them.
>
> There are five types tested.
>
> 1. A self-counting type, using an ancient Synesis (my company)
> template ref-counting bolt-in class
> 2. A self-counting type, using a custom-written (for this test)
> template ref-counting bolt-in class
> 3. Externally counting, using boost::shared_ptr
> 4. Externally counting, using a custom-written (for this test) shared
> pointer type
> 5. Externally counting, using a custom-written (for this test) shared
> pointer type, which has a custom allocator for the shared count
> members

Based on the results below, I assume that 4 & 5 do not support weak pointers
and probably use atomic integer instructions for the count updates?

You may be interested in adding:

6. boost::intrusive_ptr

7. shared_ptr, but with #define BOOST_SP_USE_QUICK_ALLOCATOR

8. shared_ptr, but with detail/shared_count.hpp replaced with

    http://www.pdimov.com/cpp/shared_count_x86_exp2.hpp

to the list.

Would it be possible for you to share the source of the benchmarks?