|
Boost : |
From: David Abrahams (dave_at_[hidden])
Date: 2008-04-28 11:10:00
on Sun Apr 27 2008, "Michael Fawcett" <michael.fawcett-AT-gmail.com> wrote:
> On Sun, Apr 27, 2008 at 9:38 PM, David Abrahams
> <dave_at_[hidden]> wrote:
>>
>> on Wed Mar 05 2008, Sebastian Redl <sebastian.redl-AT-getdesigned.at> wrote:
>>
>> > Michael Fawcett wrote:
>> >> I forgot to mention before that this is using VS 2005 as the
>> >> compiler.
>>
>> 64- or 32- bit? It matters.
>
> 32-bit.
That's a poor test of any other compiler, or of what's possible, then.
MS used the setjmp/longjmp approach to EH with that compiler, for
compatibility with "structured exceptions." The 64-bit compiler has a
"real" EH implementation ;-)
One more thing to note about the case of terminating graph algorithms
with exceptions: these algorithms have *lots* of hooks for each
iteration of the process, each of which might need to terminate the
algorithm
(e.g. http://www.boost.org/doc/libs/1_35_0/libs/graph/doc/DFSVisitor.html),
so that can add up to quite a few tests for each iteration of the inner
loop, which can be avoided if EH is used.
-- Dave Abrahams Boost Consulting http://boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk