What do people think about adaptive polling?

Simon 'corecode' Schubert corecode at fs.ei.tum.de
Tue Dec 12 05:24:27 PST 2006

Hasso Tepper wrote:
Simon 'corecode' Schubert wrote:
I think that's the equivalent of fast and slow interrupts in bsd.  we
basically (with some exceptions, like serial io) do all the work in the
slow interrupt handler (= interrupt thread).
As far as I understand it, no. Correct me if I'm wrong, but fast and slow
interrupts is about splitting interrupts into two kinds - serial io is
example of fast interrupt and network io is example of slow interrupt.
yes.  i was under the impression that we just read the data from the port in the fast interrupt and then wake up the so-called bottom half.  i could be wrong, of course.

Sofirq in Linux is about splitting work done to serve one interrupt into two
parts - minimal and very fast interrupt handler schedules task done in
sofirq thread.
maybe i understand dragonfly's interrupt processing wrong, but i would look at our interrupt system as almost entirely "soft".  the hard, fast part is generic, which simply records the event of an interrupt.  because of that the interrupt thread gets woken up, which is essentially a "soft" one.  thinking about this, i could also be entirely wrong (but i hope i'm not).

Serve - BSD     +++  RENT this banner advert  +++    ASCII Ribbon   /"\
Work - Mac      +++  space for low €€€ NOW!1  +++      Campaign     \ /
Party Enjoy Relax   |   http://dragonflybsd.org      Against  HTML   \
Dude 2c 2 the max   !   http://golden-apple.biz       Mail + News   / \
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pgp00019.pgp
Type: application/octet-stream
Size: 252 bytes
Desc: "Description: OpenPGP digital signature"
URL: <http://lists.dragonflybsd.org/pipermail/kernel/attachments/20061212/447a6949/attachment-0020.obj>

More information about the Kernel mailing list