>This means it does not hamper performance during these cases
But it's the high load case that it worth to improve. It's when the load
of my machine is very high that I want to see it to scale well.
I agree it may cut down latency when the machine is not under load of
course but I simply don't find it interesting enough.
>Oh yes, there are other options available btw. There are machines
>where DMA controllers exist which can be put into a zero'ing or
>loopback mode to do the zero'ing work. And thus have no "cpu cost"
>except to write the address to the register on the controller.
And run the I/O completation irqs and controller logic and probably
protecting the cache with irq-spinlocks (to insert the page in the
available list in that complatation irq).
Doing it with an ad-hoc controller is going to improve performance of
course as you won't pay the cost anymore. That's a different scenario. The
only problem is that I am not aware of any zeroing hardare DMA controller
on my machines.
Andrea
-
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/