[boost] Re: Smart Pointer Timing Test Results

From: Miki Jovanovic (miki_at_[hidden])
Date: 2000-02-01 10:30:52

boris fomitchev <fb-_at_[hidden]> wrote: 
> What I was speaking of was actually "faster shared ptr"
> implementation by Dave Abrahams - it has same size as the original one
> and no other penalties compared to it.

That would explain it :-). Nothing like a juicy misunderstanding.

> To fight new(), shared_ptr may use allocator<T> (if they happy to
have SGI
> allocator<>),
> or use Dave's implementation.  I think the latter is better.

Hmm, I will examine the implementation of these allocators. I was
earlier convinced that they are definitely beneficial for optimising
speed critical segments. I am curious, what are the disadvantages of
using allocators extensively, for as many of classes as possible.

Cheers,

Miki Jovanovic.