24 May
2011
24 May
'11
11:18 p.m.
Your claim is " It is a fast as the popular smart pointer * boost::shared_ptr<T>*". Yet, in single-threaded code and shared_ptr using new instead of make_shared, block_ptr still takes 3.3x as long as shared_ptr.
That is *a lot* of overhead...
It would be interesting to see a comparison of random access pattern times via block_ptr vs shared_ptr (both with and without make_shared). If I've grokked the block_ptr implementation details correctly (low probability but possible), I'd expect to see the comparison be more favorable for block_ptr than the case Phil reported. - Rhys