Ken Moffat wrote:
On Sun, Jul 24, 2016 at 01:06:40PM -0500, Bruce Dubbs wrote:
Checking, CONFIG_HZ=1000 seems to be the default, at least on my systems. I
have never set it directly that I recall, but I generally start on a new
system with 'make defconfig' and that sets it for me.
If someone starts with an old config file and runs 'make oldconfig', that
may account for propagating a poor setting for this parameter.
From the kernel help:
Allows the configuration of the timer frequency. It is customary to have
the timer interrupt run at 1000 Hz but 100 Hz may be more beneficial for
servers and NUMA systems that do not need to have a fast response for user
interaction and that may experience bus contention and cacheline bounces
as a result of timer interrupts. Note that the timer interrupt occurs on
each processor in an SMP environment leading to NR_CPUS * HZ number of
timer interrupts per second.
-- Bruce
ISTR that over the years I have had various values there, including
1000 at one time (but I probably didn't try playing a DVD on vlc
while I was running like that).
But I suspect I dropped back to 250 because my machines were
underpowered and I had not noticed any benefit from 1000.
What I normally do is build several kernels during the -rc stages.
On my intels I have not yet hit any video problems (I've been lucky,
a lot of people seem to still get problems on intel because there is
such a wide range of hardware). But on AMD my A4 and my A10 had
problems - in each case my reports led to them getting fixed before
the release. So I consider that testing kernels is worthwhile ;-)
And for a new kernel I look at the questions, and the help, in
menuconfig and make a decision about what I think will be best -
then I carry that forward unless it turns out to give obvious
problems. So, even if I had started with defconfig (and that never
used to match my cheap AMD machines!) I diverge from it over time.
Interestingly, my i7 haswell has 8 cores (with hyperthreading), my
i3 has 2/4, my phenom has 4 (arguably), so even at 250 there are a
lot of interrupts.
When a machine is running at 2GHz, 1K interrupts does not appear to be a
lot. Most interrupts do not require any action so it is generally
if ( ! do_something ) return
2 million+ cycles between interrupts seems like a lot to me. Remember
that for video a lot of the work is offloaded to the video card or the
video section of the processor and is thus run in parallel to the main
cpu(s)/cores.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
