Re: 2us Interrupt latency's for Linux 2.0.xx

Linus Torvalds (torvalds@transmeta.com)
4 Sep 1998 17:07:14 GMT


In article <Pine.LNX.3.96.980904012157.6390A-100000@z.ml.org>,
Gregory Maxwell <linker@z.ml.org> wrote:
>
>If NT uses the IO-APIC then why is it limited to the 16 normal INTs?
>Furthermore, if they use IO-APIC how could they screw up their int
>handling so badily their their int latency is so high dispite the ultra
>fast IO-APIC.

Note that the hardware overhead of interrupt handling is usually still
much much less than the software overhead. Bad interrupt latency comes
from bad code, not from bad hardware.

The old-style XT-PIC is slower enough that you can see the effects under
Linux, but that's with a fairly optimized interrupt handling path. I
suspect the thing is completely in the noise when running NT.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/faq.html