[boost] Re: Re: Platform Neutrality - without reinterpret_cast<>andifdef

From: Eric Woodruff (Eric.Woodruff_at_[hidden])
Date: 2002-08-05 15:07:03

I can understand the hit taken in the readability of the mutex
implementation for "efficiency," but it is unacceptable for thread. I've
read boost's biases and the thread implementation is a certain violation of
the heart of boost's principles.

----- Original Message -----
From: William E. Kempf
Newsgroups: gmane.comp.lib.boost.devel
Sent: Monday, 2002:August:05 3:35 PM
Subject: Re: Re: Platform Neutrality - without reinterpret_cast<>andifdef

----- Original Message -----
From: "Eric Woodruff" <Eric.Woodruff_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Monday, August 05, 2002 2:17 PM
Subject: [boost] Re: Platform Neutrality - without reinterpret_cast<>
andifdef

> PlatformSpecificThread is a concrete class--not a template. There is no
> "selection", it is only defined in the localized thread implementation.
The
> key is the pre-declaration. I can see how it might be considered the PIMPL
> idiom, however, doesn't that idiom require a base class and polymorphism
> (virtual methods)? This has none of those.
>
> Isn't the problem with mutexs only to be blamed on the allocator used?

Possibly, and a small memory allocator could improve things here a little,
though the need for thread safety will still cause some (possibly
significant) overhead with the PIMPL idiom.

> Otherwise the entire object-oriented paradigm is meaningless.

Not at all.  Just because OO can cause unnecessary overhead in one case does
not lead to the conclusion that it's meaningless.  The idea of using PIMPL
idioms, with or with out polymorphic implementations, solely for the sake of
removing conditional compilation may work, but it's not the most efficient
way to accomplish the goal.

Bill Kempf
_______________________________________________
Unsubscribe & other changes:
http://lists.boost.org/mailman/listinfo.cgi/boost