Stuart Henderson wrote:
On 2007/11/26 18:21, Daniel Ouellet wrote:
One more question for you if I may.
As far as you know, does this difference interrupt processing is also use
for UBS devices as well by any chance?
yes, and everything else that uses interrupts.
(/sys/arch/i386/config/GENERIC.MP uses ioapic, .../GENERIC does not).
save dmesg from the various different kernels and use diff to see what
changes between them.
I didn't see in your posts to this thread whether you've compared
with/without acpi? (recent snapshots should enable acpi automatically
if you have >1 cpu/core and you'll see several acpi* devices show up).
Both had acpi enable, but also just a few days ago Theo disable ACPI on
i386 for a bug that Miod needs to fix, if I remember the right name.
I sure can test it again for the i386, but I did that many times over
and no problem.
Only problem with the amd64.mp.
I will however for the sake of completeness load the latest on i386 as I
run the one for November 15, enable acpi, compare the dmesg and send it
in. I did that for myself however and I do see a bunch of numbers in the
amd64 one that I have no clue as to what they mean however, but that's
new, however the bug is not, so I kind of look over that as being new
stuff that will come in one day and left it to that.
Give me about 5 or may be 10 minutes and I will send the diff between
the i386.mp and amd64.mp
Most likely they will not be the exact same date however as the amd is
always behind, but still I will do it.