RPi and powerd, was: Re: RPI4 clock speeds and serial port ( temperatures idle and -j4 buildworld buildkernel )

[Mark Millard]

On 2021-Mar-24, at 07:34, tech-lists <tech-lists at zyxst.net> wrote:

> On Tue, Mar 23, 2021 at 12:57:03PM -0700, Mark Millard wrote:
> 
>> 
>> Had a -mcpu=cortext-a72 world and kernel been
>> installed and booted first? 
> 
> yes.
> 
>> This does not show ccache being cleared out
>> before the below. So the times may be examples
>> of "with ccache benefit" times. The contrast
>> with mine and Bob P.'s times suggests a
>> nice time-benefit can occur.
> 
> On this occasion I can't remember for sure if ccache was cleared out. I don't think so.

I've sent time-to-build figures for a from-scratch build
in a configuration somewhat similar to yours other than
devel/ccache not being involved and the specific WITH
and WITHOUT combination used. I included both "my way"
and "your way" figures. I'll not repeat it all here.

At somewhat under 15 hrs 35 min for my experiment with
"your way" (but without devel/ccache), being over 6.5
hrs longer than my normal build configuration, your
"with devel/cache" results taking far less time likely
is mostly attributable to your use of devel/cache and
the particular history that lead to the particular
cache contents that were present when your started
your build.

(Of course, my test is not the same as having devel/ccache
and an empty cache. And I built my configuration that
eliminates about 30 minutes from the llvm/clang part of
the build but has other things that you eliminate.)

Based on the timings and configuration details, your
ccache was almost certainly not cleared. And that has
allowed this comparison/contrast with not having
devel/cache involved.

> The numbered list was to show (for me as well) exactly
> what was done and the order it was done in.

Yep. Thanks.

Caching leads to prior history being involved via
the cache contents that were left behind. It makes
for repeatable testing being messier to set up
(other than the empty-cache case).

But when the contrasts are large enough, things
can still be significantly suggestive.

> I'll rm -rf /usr/obj/* and /var/cache/ccache/* if upgrading between
> major versions, or if it's not building and need to narrow down why. But
> usually leave these dirs intact because (and especially with the ccache
> dir) emptying them negates potential speed benefit.

Yep: and it looks like the time taken can be a
lot less with the devel/ccache use.

If you at some point start from a empty ccache and
empty /usr/obj/ for some reason, you may want to
collect and report the build times and the
configuration it was for. That would be useful for
comparisons with non-empty cache cases.

> make cleanworld &c
> *should* remove all the obj cruft in /usr/obj.
> 


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)