Re: 2048 CPUs [was: Re: New filesystem for Linux]

[Albert Cahalan]

> >> No intel document guarantees you that if more CPUs
> >> simultaneously execute locked cmpxchg in a loop that a
> >
> > If we are talking 2048 cpus, we are talking ia64.
>
> IA64 spinlock is locked cmpxchg, if failed than pause (i386 equivalent of
> rep nop) read the value, and if unlocked, try cmpxchg again.
>
> There is no fairness in it.

I suppose we could use something better.

There is the MCS lock, the related CLH lock, and IBM's
improvement on the MCS lock. As with RCU, we'd need
to get IBM's permission to use their lock. (so, how did we
get permission for RCU?) The basic MCS lock is also
patented I think.

http://www.cs.rochester.edu/~scott/professional/Dijkstra/presentation.html
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/