On Wed, Jul 06, 2005 at 02:59:37PM -0400, Alan Cox wrote:
On Wed, Jul 06, 2005 at 02:18:40PM -0400, Brian Gerst wrote:
> What about the bloat and slowdown because of the unnecessary spinlocks
> and more complex scheduler code?
The locks shouldn't matter. AMD got locked operations on exclusive cache
lines right. Its generally only Intel boxes that had high lock costs.
And the only UP EM64T's I've seen have HT. Though, that could be
disabled, but I expect that the folks disabling it are in a minority.
Not sure on the scheduler.
It'll degenerate to a single runqueue on UP, so there shouldn't
be any of the complicated balancing going on. Bloat-wise, it's
a tiny amount of change.
Dave