Subj : Re: To anybody at Microsoft who cares...  :)
To   : comp.os.ms-windows.programmer.win32,comp.programming.threads
From : Ziv Caspi
Date : Sat Jan 08 2005 01:03 am

Does this help?

Multiprocessor Considerations for Kernel-Mode Drivers
http://www.microsoft.com/whdc/driver/kernel/MP_issues.mspx

Ziv Caspi

"SenderX" <xxx@xxx.com> wrote in message 
news:f0EEc.3478$MB3.1944@attbi_s04...
> Do you guys have any idea when/if you are going to give a "detailed
> description" of the InterlockedXXX API's memory visibility semantics?
>
> I think you are going for the straight forward method of acquire/release,
> but your docs need to be tightened up. For one, does
> InterlockedCompareExchangeAcquire/Release execute the membar upon cas
> failure, on non-tso CPU's? You are forcing programmers to disassemble your
> atomic-ops to try to ease their minds. I know that acquire/release has
> basically the same memory visibility rules that pthread_mutex_lock/unlock
> has, but you still need to document "exactly what your acquire/release
> semantics are", and give detailed description on how they "apply to your
> interlocked API"...
>
>
>
> MSDN would be a great place for this... Put it under something like:
>
> "The Microsoft Memory Visibility Standard: Acquire/Release/Full"
>
> :O
>
>
>
> P.S.
>
> I also wouldn't hurt to document how acquire/release "applies to all of 
> your
> sync primitives"...
>
> ;)
>
>

.