| Newsgroups |
gmane.linux.lfs.automated |
| Message-ID |
<Y9mzaQTripcUK/[email protected]> |
On Wed, Feb 01, 2023 at 12:09:17AM +0100, Pierre Labastie wrote:
> On Tue, 2023-01-31 at 16:00 -0600, Bruce Dubbs wrote:
> > On 1/31/23 15:18, Ken Moffat ([email protected] via alfs-discuss Mailing
> > List)
> > wrote:
> > > On Tue, Jan 31, 2023 at 07:01:52PM +0100, Pierre Labastie wrote:
> > > > On Tue, 2023-01-31 at 17:26 +0100, Dragan wrote:
> > > > > I've successfully build r11.2-297.
> > > > >
> > > > > Log attached.
> > > > >
> > > > > Dragan
> > > >
> > > > Hmmm, weird that you need 130 sec to build binutils pass 1 at -j16. It's
> > > > around
> > > > 30 sec for me... Have you had a look at the cpuidle governor for your
> > > > kernel?
> > > >
> > > > Pierre
> > > > PS: By tweaking the cpuidle governor, I could decrease the SBU at -j1 from
> > > > 130
> > > > to 80 seconds...
> > >
> > > For information, on my Ryzen 4750G a -j1 build of binutils from
> > > r11.2-296 on a host system running r11.2-150 took 103.999 seconds.
> > > Building binutils pass 1 on the completed system took 127.307
> > > seconds.
> > >
> > > In each case the host system, but not binutils pass 1, had some
> > > optimzations and hardening, and both were running a 6.1.8 kernel.
> > >
> > > You guys with your your souped-up hardware ;-)
> >
> > I just did a fresh build this morning. For me, my binutils-pass1 was also 104
> > seconds at -j1. However at -j22, -pass2 was 34 seconds.
> >
> > I do not normally change the cpufreq governor or scaling_governor. The
> > scaling_governor is currently set to 'userspace'.
> >
> > Ryzen 9 3900X
> >
>
> Intel processors have this 'intel_pstate' scaling governor, which uses directly
> hardware monitoring to set the state. It is quite different from other
> governors, so I'd recommend reading /usr/share/doc/linux-<version>/admin-
> guide/pm/intel_pstate.rst, which has a reference to the other governors if you
> want to get the whole picture. I think (not totally sure) I had to change the
> CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE kernel config option to get acceptable
> performance on my intel 12900.
>
> Pierre
>
On my i3 skylake (currently inoperable) I used
(details last saved for 5.19)
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL is not set
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
# CONFIG_CPU_FREQ_GOV_POWERSAVE is not set
# CONFIG_CPU_FREQ_GOV_USERSPACE is not set
# CONFIG_CPU_FREQ_GOV_ONDEMAND is not set
# CONFIG_CPU_FREQ_GOV_CONSERVATIVE is not set
CONFIG_CPU_FREQ_GOV_SCHEDUTIL=y
#
# CPU frequency scaling drivers
#
CONFIG_X86_INTEL_PSTATE=y
CONFIG_X86_PCC_CPUFREQ=y
# CONFIG_X86_AMD_PSTATE is not set
CONFIG_X86_ACPI_CPUFREQ=y
On my i7 haswell, confi last saved for 6.1.3, I'm using the same.
On my AMDs I use PERFORMANCE, although on the 4750 I see that I have
CONFIG_X86_INTEL_PSTATE=y
# CONFIG_X86_PCC_CPUFREQ is not set
CONFIG_X86_AMD_PSTATE=y
# CONFIG_X86_AMD_PSTATE_UT is not set
CONFIG_X86_ACPI_CPUFREQ=y
CONFIG_X86_ACPI_CPUFREQ_CPB=y
#
On the 4750 I tried the recent AMD changes back in (I think) 6.0-rc,
but my hardware is not supported.
ĸen
--
When it was cold, we used to sit by a candle for warmth.
If it got really cold we lit it.
--
http://lists.linuxfromscratch.org/sympa/info/alfs-discuss
Unsubscribe: See the above information page