>>>>> "todd" == Todd Lyons <[EMAIL PROTECTED]> writes:

Hi

>> For the APIC routing all interrupts to only one CPU, it is not normal.

todd> I think that's what people are bugging about in the Cooker list at the
todd> moment.  I never noticed it until someone in the list mentioned it.  The
todd> complaint is that it's impacting performance, I personally have not seen
todd> anything as a result of it.

If you have lots of interrupts (think for instance of a network
server), you will notice the drop in performance :(

>> Notice that for P4, it is normal, as some designer decided that this
>> was the right thing to do :(  

todd> Hmmm, so it's done for the P4 like that at the hardware level?
todd> Interesting.  Thanks for the 

In the P4 only way to fix it is to spread the interrupts by hand
between the processors.  Dual Pentium (<P4) use an algorithm that  is
a some special case of round robin.  I don't remember the details, but
basically was that each interrupt goes to each cpu in round robin order.

Later, Juan.

-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Reply via email to