[boost] Re: Empty Base Optimization

From: scleary_at_[hidden]
Date: 1999-11-29 08:29:49

> Darin -
>
>> I suggest we remove the empty inline destructor definition. Does the it
>> make things more efficient? When I tested it with the compilers I use, it
>> made no difference in the code generated. If it doesn't make
>> things any better, I'd prefer to omit it.
>
> In at least one compiler (Borland C++) the presence of a destructor
results
> in exception handling prolog-epilog code even when the destructor is empty
> - given that its presence serves no purpose, it should indeed be left out
>

Yes, it should be left out, since implicit destructors and especially
trivial destructors, both generated by the compiler (not user-defined), are
much more easily optimized.

For the generic Boost version, it should be removed.  I have a habit of
putting explicit destructors in every class I write because Borland C++
Builder (which I use) has an obscure bug dealing with stack unwinding that
is worked around by putting an explicit destructor in every unwound class.

        -Steve