> usecs := usecs_per_jiffie * (4G-delta_clocks) / clocks_per_jiffie
wheee, it's "delta_clocks-4G", which is "delta_clock" by definition, so
this isnt a bug :((
the other two remarks (2 unnecessary divisions and time drifting
instead of maximising) are still valid .... i guess :)
-- mingo