Messages in this thread | | | Subject | Re: [patch] Re: spin_unlock optimization(i386) | Date | Sun, 21 Nov 1999 15:13:40 +0000 (GMT) | From | Alan Cox <> |
| |
> yep, this should be possible. I remember having hacked in something like > that a year ago but i saw crashes - although that might be an unrelated > thing.
The PPro documentation says "Writes by a single processor are observed in the same order by all processors", as well as (on a per cpu basis) "Writes to memory are always carried out in program order".
Reads are allowed to pass. But the lock bset can handle this (we might spin in error for an instant).
"Writes from all processors on the system bus can be observed in different orders"
This is coming back from a spinlock and only cpu gets to clear it so who cares
Unfortunately it then goes on to say "Intel does not guarantee that future processors will support this model"
If the memory area is marked as write-back, which seems to be the case for RAM then writes can be buffered and combined and reads can be speculative.
Alan
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |