[boost] Re: Speed of nonleaky shared pointers

From: Beman Dawes (beman_at_[hidden])
Date: 1999-07-09 11:12:11

At 08:56 AM 7/9/99 -0500, Ed Brey wrote:

>The smart pointer documentation mentions that the efficiency of
attached
>versus detached reference counts in shared pointers came out to be
about
>the same under tests.  It is not clear whether these tests were run
>before or after the fixes were put in to prevent memory leaks if
>allocating the detached reference count throws, but it appears as
though
>there were run before.  Have the tests been rerun since then.  I
>would expect that adding the try blocks could significantly slow the
>detached implementation and may create an imbalance in the force,
>err, the efficiency.

It wasn't "attached" versus "detached" that was tested, but rather
"direct detached" versus "indirect detached".  The various fixes
would apply to both direct and indirect approaches, since both
allocate memory via new and delete.

>Also, have any tests been done to compare code size of attached
versus
>detached, with the exception handling now included.

No, no size tests.

By the way, Kevlin Kenney's "Counted Body Techniques" paper will be
posted on the web site soon.  It will be of interest to people who
care about smart pointers.

--Beman

------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/boost
http://www.egroups.com - Simplifying group communications