RE: [boost] Re: A simple monitor for Boost

From: Reid Sweatman (borderland_at_[hidden])
Date: 2000-08-10 21:49:47

Just my unedicated two bits, but personally I'd prefer having the
primitives, even at the expense of some platform-dependent code, and build
the monitor on top of it.  Maybe it would be enough to sort of "deprecate"
the primitives in the documentation.  Something like, "Abandon all hope of
debugging, ye who use these primitives." <g>  I imagine I'd end up using the
monitor most of the time, but I'd still like the option of hitting the
lowest layer if necessary.  At least I'm used to mutexes and semaphores, on
a couple of platforms, anyway.  I'd think another argument for doing it that
way would be that you could implement different styles of monitor, as
necessary.

Reid Sweatman

> -----Original Message-----
> From: Greg Colvin [mailto:gcolvin_at_[hidden]]
> Sent: Thursday, August 10, 2000 2:19 PM
> To: boost_at_[hidden]
> Subject: Re: [boost] Re: A simple monitor for Boost
>
>
> From: "David Abrahams" <abrahams_at_[hidden]>
> > From: "Greg Colvin" <gcolvin_at_[hidden]>
> >
> > > There a many monitor-based languages in which large applications
> > > and operating systems have been written.  There are also proofs
> > > that monitors, mutexes, and semaphores are formally equivalent,
> > > in the sense that given one you can write the others.
> >
> > But without language support, can we expect a class that just
> needs atomic
> > test-and-decrement (for example) to perform well enough when
> implemented in
> > terms of the proposed monitor abstractions?
>
> No.  I think we need an atomic counter and atomic swap as primitives.
>
>
>
> 
>
>
>
>