Re: [boost] Fw: [atomic] review results

Subject: Re: [boost] Fw: [atomic] review results
From: Peter Dimov (pdimov_at_[hidden])
Date: 2011-11-08 12:25:43

Tim Blechmann wrote:
> isn't atomic_is_lock_free only defined for integral types?

I'm not aware of any such requirement.

> and why isn't atomic::is_lock_free a static member function?

I don't know.

> dispatching per-object would actually make sense, because there may be
> platforms which require objects to be aligned to certain memory boundaries 
> for
> double-width CAS.

It can be read that way, but it doesn't make much sense from the user's 
perspective for objects to randomly be lock-free or non-lock-free. 
std::atomic<> should ensure the necessary alignment.