Re: [boost] Exception specifications in smart_ptr.hpp

From: Greg Colvin (greg_at_[hidden])
Date: 2000-07-21 17:33:00

From: David Abrahams <abrahams_at_[hidden]>
> From: "Valentin Bonnard" <Bonnard.V_at_[hidden]>
> To: <boost_at_[hidden]>
> Sent: Friday, July 21, 2000 2:38 PM
> Subject: Re: [boost] Exception specifications in smart_ptr.hpp
> 
> 
> > Greg Colvin wrote:
> >
> > > For non-inline functions "throws nothing" specs may allow
> > > optimizations that help more than their overhead hurts.
> >
> > In many compilers exceptions cost nothing in term of runtime.
> > In some compilers exception specs cost a lot.

I didn't say "help runtime" I said "help".

> > I don't know of any compilers for which exceptions specs
> > cost nothing but empty specs help in term of runtime. Could
> > you name one ?
> 
> I think IBM's compiler has zero-time-overhead EH where empty exception specs
> can help save space if used very carefully. Saving space can ultimately help
> with runtime because of image size, swapping, etc... but I think it's a
> marginal case if you're interested in speed. With a smart compiler, it's
> executable image size which can be significantly reduced by using empty
> exception specifications.
> 
> > I can't think of many real-world cases where non empty
> > specs could help. For inline functions is doesn't help
> > because the compiler would have as much info by performing
> > inline substitution.
> 
> Agreed, 100%.

Yep.