[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [f-cpu] PS

hi !

nicO wrote:
> What ever fixed number of semaphore you will put they will never have
> enought. So hardware semaphore  must be designed to handel software
> semaphore.
> (I just remind you that a posix semaphore is a counter not a flag.)
> This HW semaphore will be used to protect memory area of the true
> semaphore .So we will need only _one_ semphore to handel that.
> Or we can use an other kind of sync things. Have you ever hurd about
> super step model ?
> It could represent by a wait&sync instruction. This instruction wait
> that every core are stoped with this instruction before releasing them
> in the same time (in NUMA system, caches are updated just before). It's
> much easier to programme multi-cpu application that way (a compiler that
> compile for 2 or more cpu).

i think you'll have to explain that to me by voice and in french...
i seem to miss the deep implication of your idea.
When do we meet again ?

> nicO
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/