[9fans] /dev/cputemp

[9fans] /dev/cputemp

[erik quanstrom]

i dug out the code i wrote some time ago to read cpu
temperature (or the closest stand in for it i could find)
on most modern amd and intel processors.  it's wrapped
up with a change to extend the model macro so that
conroe l processors don't appear to be xeons.

/n/sources/contrib/quanstro/src/9/pc/^(cpuid0.s dat.h devarch.c io.h)

you'll need to add cpuid0.s to l.s.

here's how it works.  minooka obviously has two processors:

	minooka; cat /dev/cputemp
	32±1
	31±1

- erik

[9fans] /dev/cputemp

[Jessica Yu]

Just curious, is this not a thing in the nix kernel? grep'd the nix
sources and it didn't seem to be in devarch.c, it's in 9/pc/ though;
is there another way to grab cpu temp?

I ask because there seems to be a significant temperature change on my
test machine between the old nix kernel and some of the new scheduler
changes, although I wanted some numbers to back that observation (or
maybe I'm imagining things!). Not a super big deal but just something
I was just curious about. :-)

Jessica

Re: [9fans] /dev/cputemp

[erik quanstrom]

On Tue May 27 17:59:41 EDT 2014, jyu@cowsay.org wrote:
> Just curious, is this not a thing in the nix kernel? grep'd the nix
> sources and it didn't seem to be in devarch.c, it's in 9/pc/ though;
> is there another way to grab cpu temp?
>
> I ask because there seems to be a significant temperature change on my
> test machine between the old nix kernel and some of the new scheduler
> changes, although I wanted some numbers to back that observation (or
> maybe I'm imagining things!). Not a super big deal but just something
> I was just curious about. :-)

yes, i did not bring cputemp into the nix kernel because i thought the code
was ugly, and the temperatures were not necessarly accurate.  but if you want
to pull it in, that would be fine.

- erik