Re: isapnp probe is slow

[christos%astron.com@localhost (Christos Zoulas)]

In article <20080330205421.GY10654%snowdrop.l8s.co.uk@localhost>,
David Laight  <david%l8s.co.uk@localhost> wrote:
>On Sat, Mar 29, 2008 at 11:46:26PM +0000, Andrew Doran wrote:
>> > > 
>> > > Would disabling isapnp based on the age of the system make sense?
>> > > Alternatively, what other ways are there to speed it up?
>> > 
>> > I thought the isapnp grope was relatively fast?
>> 
>> I think it takes about 4 seconds or so.
>
>WTF is it doing - I've written the isapnp enumeration algorithm (a long
>time ago) and don't remember it requiring any long delays.
>
>I've found a definition of the algorithm...
>32 iows, 1ms delay, followed by 72 iors alternating 30us and 250us delays.
>
>That ought to be a small fraction of a second - even if it has to be
>repeated because the selection 'read' port is otherwise assigned.

Buggy cards that don't spin up quickly enough. It used to take less
time, but delays were increased to accommodate for buggy hardware.

christos