* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 [not found] ` <573B4C2A.4010205@intel.com> @ 2016-05-17 23:16 ` Sedat Dilek 2016-05-17 23:18 ` Rafael J. Wysocki 0 siblings, 1 reply; 15+ messages in thread From: Sedat Dilek @ 2016-05-17 23:16 UTC (permalink / raw) To: Rafael J. Wysocki; +Cc: Pandruvada, Srinivas, linux-pm On 5/17/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: > On 5/17/2016 5:30 PM, Sedat Dilek wrote: >> Hi, > > Hi, > > You have a broken address of linux-pm (it leads to nowhere). > Grrr, yeah you are right. >> I wonder if the Kconfig logic is correct when using >> CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL=y in combination with >> intel_pstate driver for X86 architecture >> >> The available scaling-governors for intel_pstate are: performance >> powersave. >> >> Only acpi-cpufreq can use the new schedutil governor. > > The right way to say that would be that intel_pstate doesn't work with > cpufreq governors. > > scalling_governor is just used by intel_pstate as an interface for > choosing the mode it works in and it always uses its own built-in > "governor". And BTW, intel_pstate "powersave" is not the same as the > cpufreq's "powersave" even. > >> Shouldn't it be prevented from selecting schedutil governor as default >> when CONFIG_X86_INTEL_PSTATE=y is set? > > That should apply to "ondemand" too and it doesn't, so no. > > "schedutil" is just a regular cpufreq governor and it doesn't care about > intel_pstate. > >> It makes no real sense to me. >> >> Ubuntu has in it's kernels CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y. >> >> What are the alternatives? >> Set CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y. > > You can choose every cpufreq governor as the default. There are 6 of > then now in the tree. > >> BTW, I cannot change my scaling-driver... >> ...intel_pstate -> acpi-cpufreq... >> >> root# for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n >> acpi-cpufreq > /sys/devices/system/cpu/cpufreq/${p}/scaling_driver ; >> done >> >> Attached is also a modified /etc/init.d/ondemand script from my >> Ubuntu/precise AMD64 to use and fallback to schedutil governor. > > No, you can't change the scaling driver this way. > > It actually is not possible to change from intel_pstate to anything else > now. To run without intel_pstate, you need to add > "intel_pstate=disable" to the kernel command line. > OK, I will try intel_pstate=disable in my k-c-l. Is there no way to switch between acpi-cpufreq and intel_pstate on the fly? Did not look in the Kconfig sections... Can I built both as modules? Then unload and load the other? - Sedat - ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-17 23:16 ` [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 Sedat Dilek @ 2016-05-17 23:18 ` Rafael J. Wysocki 2016-05-19 11:04 ` Sedat Dilek 0 siblings, 1 reply; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-17 23:18 UTC (permalink / raw) To: sedat.dilek; +Cc: Pandruvada, Srinivas, linux-pm On 5/18/2016 1:16 AM, Sedat Dilek wrote: > On 5/17/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >> On 5/17/2016 5:30 PM, Sedat Dilek wrote: >>> Hi, >> Hi, >> >> You have a broken address of linux-pm (it leads to nowhere). >> > Grrr, yeah you are right. > >>> I wonder if the Kconfig logic is correct when using >>> CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL=y in combination with >>> intel_pstate driver for X86 architecture >>> >>> The available scaling-governors for intel_pstate are: performance >>> powersave. >>> >>> Only acpi-cpufreq can use the new schedutil governor. >> The right way to say that would be that intel_pstate doesn't work with >> cpufreq governors. >> >> scalling_governor is just used by intel_pstate as an interface for >> choosing the mode it works in and it always uses its own built-in >> "governor". And BTW, intel_pstate "powersave" is not the same as the >> cpufreq's "powersave" even. >> >>> Shouldn't it be prevented from selecting schedutil governor as default >>> when CONFIG_X86_INTEL_PSTATE=y is set? >> That should apply to "ondemand" too and it doesn't, so no. >> >> "schedutil" is just a regular cpufreq governor and it doesn't care about >> intel_pstate. >> >>> It makes no real sense to me. >>> >>> Ubuntu has in it's kernels CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y. >>> >>> What are the alternatives? >>> Set CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y. >> You can choose every cpufreq governor as the default. There are 6 of >> then now in the tree. >> >>> BTW, I cannot change my scaling-driver... >>> ...intel_pstate -> acpi-cpufreq... >>> >>> root# for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n >>> acpi-cpufreq > /sys/devices/system/cpu/cpufreq/${p}/scaling_driver ; >>> done >>> >>> Attached is also a modified /etc/init.d/ondemand script from my >>> Ubuntu/precise AMD64 to use and fallback to schedutil governor. >> No, you can't change the scaling driver this way. >> >> It actually is not possible to change from intel_pstate to anything else >> now. To run without intel_pstate, you need to add >> "intel_pstate=disable" to the kernel command line. >> > OK, I will try intel_pstate=disable in my k-c-l. > > Is there no way to switch between acpi-cpufreq and intel_pstate on the fly? No, there's no way to do that ATM. > Did not look in the Kconfig sections... > Can I built both as modules? > Then unload and load the other? No, intel_pstate can't be built as a module. Thanks, Rafael ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-17 23:18 ` Rafael J. Wysocki @ 2016-05-19 11:04 ` Sedat Dilek 2016-05-19 11:54 ` Rafael J. Wysocki 0 siblings, 1 reply; 15+ messages in thread From: Sedat Dilek @ 2016-05-19 11:04 UTC (permalink / raw) To: Rafael J. Wysocki; +Cc: Pandruvada, Srinivas, linux-pm On 5/18/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: > On 5/18/2016 1:16 AM, Sedat Dilek wrote: >> On 5/17/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >>> On 5/17/2016 5:30 PM, Sedat Dilek wrote: >>>> Hi, >>> Hi, >>> >>> You have a broken address of linux-pm (it leads to nowhere). >>> >> Grrr, yeah you are right. >> >>>> I wonder if the Kconfig logic is correct when using >>>> CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL=y in combination with >>>> intel_pstate driver for X86 architecture >>>> >>>> The available scaling-governors for intel_pstate are: performance >>>> powersave. >>>> >>>> Only acpi-cpufreq can use the new schedutil governor. >>> The right way to say that would be that intel_pstate doesn't work with >>> cpufreq governors. >>> >>> scalling_governor is just used by intel_pstate as an interface for >>> choosing the mode it works in and it always uses its own built-in >>> "governor". And BTW, intel_pstate "powersave" is not the same as the >>> cpufreq's "powersave" even. >>> >>>> Shouldn't it be prevented from selecting schedutil governor as default >>>> when CONFIG_X86_INTEL_PSTATE=y is set? >>> That should apply to "ondemand" too and it doesn't, so no. >>> >>> "schedutil" is just a regular cpufreq governor and it doesn't care about >>> intel_pstate. >>> >>>> It makes no real sense to me. >>>> >>>> Ubuntu has in it's kernels CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y. >>>> >>>> What are the alternatives? >>>> Set CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y. >>> You can choose every cpufreq governor as the default. There are 6 of >>> then now in the tree. >>> >>>> BTW, I cannot change my scaling-driver... >>>> ...intel_pstate -> acpi-cpufreq... >>>> >>>> root# for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n >>>> acpi-cpufreq > /sys/devices/system/cpu/cpufreq/${p}/scaling_driver ; >>>> done >>>> >>>> Attached is also a modified /etc/init.d/ondemand script from my >>>> Ubuntu/precise AMD64 to use and fallback to schedutil governor. >>> No, you can't change the scaling driver this way. >>> >>> It actually is not possible to change from intel_pstate to anything else >>> now. To run without intel_pstate, you need to add >>> "intel_pstate=disable" to the kernel command line. >>> >> OK, I will try intel_pstate=disable in my k-c-l. >> >> Is there no way to switch between acpi-cpufreq and intel_pstate on the >> fly? > > No, there's no way to do that ATM. > >> Did not look in the Kconfig sections... >> Can I built both as modules? >> Then unload and load the other? > > No, intel_pstate can't be built as a module. > It's really a pity that I cannot switch on the fly between cpufreq/schedutil vs. intel_pstate/x86 Hmm, I have to read your replies carefully and this topic is still confusing for me. I tried in the meantime "intel_pstate=disable" in my kernel-command-line. $ dmesg | egrep -i 'intel_pstate|cpufreq|cpuidle|cpu[0-3]' [ 0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-4.6.0-4-iniza-small root=UUID=001AADA61AAD9964 loop=/ubuntu/disks/root.disk ro intel_pstate=disable [ 0.000000] ACPI: SSDT 0x00000000DAFEB000 000804 (v01 PmRef Cpu0Ist 00003000 INTL 20061109) [ 0.000000] Kernel command line: BOOT_IMAGE=/boot/vmlinuz-4.6.0-4-iniza-small root=UUID=001AADA61AAD9964 loop=/ubuntu/disks/root.disk ro intel_pstate=disable [ 0.003982] CPU0: Thermal monitoring enabled (TM1) [ 0.075695] smpboot: CPU0: Intel(R) Core(TM) i5-2467M CPU @ 1.60GHz (family: 0x6, model: 0x2a, stepping: 0x7) [ 0.342317] cpuidle: using governor ladder [ 0.358253] cpuidle: using governor menu [ 0.422578] ACPI: SSDT 0xFFFF8800D58F9000 000688 (v01 PmRef Cpu0Cst 00003001 INTL 20061109) [ 0.855129] microcode: CPU0 sig=0x206a7, pf=0x10, revision=0x28 [ 0.855157] microcode: CPU1 sig=0x206a7, pf=0x10, revision=0x28 [ 0.855190] microcode: CPU2 sig=0x206a7, pf=0x10, revision=0x28 [ 0.855227] microcode: CPU3 sig=0x206a7, pf=0x10, revision=0x28 [ 85.914640] cpufreq: CPU0: Fast frequency switching not enabled [ 85.914645] cpufreq: Registered transition notifiers: [ 85.914702] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [ 85.914948] cpufreq: CPU1: Fast frequency switching not enabled [ 85.914950] cpufreq: Registered transition notifiers: [ 85.914954] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [ 85.915093] cpufreq: CPU2: Fast frequency switching not enabled [ 85.915095] cpufreq: Registered transition notifiers: [ 85.915099] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [ 85.915188] cpufreq: CPU3: Fast frequency switching not enabled [ 85.915190] cpufreq: Registered transition notifiers: [ 85.915193] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 Some sysfs-kungfu with acpi-cpufreq/schedutil ... # LC_ALL=C ls -lR /sys/devices/system/cpu/cpufreq/policy0/ /sys/devices/system/cpu/cpufreq/policy0/: total 0 -r--r--r-- 1 root root 4096 May 19 12:54 affected_cpus -r--r--r-- 1 root root 4096 May 19 12:54 bios_limit -r-------- 1 root root 4096 May 19 12:54 cpuinfo_cur_freq -r--r--r-- 1 root root 4096 May 19 12:54 cpuinfo_max_freq -r--r--r-- 1 root root 4096 May 19 12:54 cpuinfo_min_freq -r--r--r-- 1 root root 4096 May 19 12:54 cpuinfo_transition_latency -r--r--r-- 1 root root 4096 May 19 12:54 freqdomain_cpus -r--r--r-- 1 root root 4096 May 19 12:54 related_cpus -r--r--r-- 1 root root 4096 May 19 12:54 scaling_available_frequencies -r--r--r-- 1 root root 4096 May 19 12:54 scaling_available_governors -r--r--r-- 1 root root 4096 May 19 12:54 scaling_cur_freq -r--r--r-- 1 root root 4096 May 19 12:54 scaling_driver -rw-r--r-- 1 root root 4096 May 19 11:46 scaling_governor -rw-r--r-- 1 root root 4096 May 19 12:54 scaling_max_freq -rw-r--r-- 1 root root 4096 May 19 12:54 scaling_min_freq -rw-r--r-- 1 root root 4096 May 19 12:54 scaling_setspeed drwxr-xr-x 2 root root 0 May 19 12:54 stats /sys/devices/system/cpu/cpufreq/policy0/stats: total 0 -r--r--r-- 1 root root 4096 May 19 13:00 time_in_state -r--r--r-- 1 root root 4096 May 19 13:00 total_trans -r--r--r-- 1 root root 4096 May 19 13:00 trans_table # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat /sys/devices/system/cpu/cpufreq/${p}/scaling_driver ; done acpi-cpufreq acpi-cpufreq acpi-cpufreq acpi-cpufreq # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done schedutil schedutil schedutil schedutil # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat /sys/devices/system/cpu/cpufreq/${p}/scaling_available_governors ; done conservative ondemand userspace powersave schedutil performance conservative ondemand userspace powersave schedutil performance conservative ondemand userspace powersave schedutil performance conservative ondemand userspace powersave schedutil performance # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat /sys/devices/system/cpu/cpufreq/${p}/scaling_available_frequencies ; done 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 # cat /sys/devices/system/cpu/cpufreq/schedutil/rate_limit_us 10000 Just did not found it... How do I enable "Fast frequency switching" on my cpu-cores? - Sedat - ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 11:04 ` Sedat Dilek @ 2016-05-19 11:54 ` Rafael J. Wysocki [not found] ` <CA+icZUXV3mz9Yr05o5zHLOcXwYRWydtXZOuLnE10Zy3RwzOHUg@mail.gmail.com> 0 siblings, 1 reply; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 11:54 UTC (permalink / raw) To: sedat.dilek Cc: Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: > On 5/18/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >> On 5/18/2016 1:16 AM, Sedat Dilek wrote: >>> On 5/17/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >>>> On 5/17/2016 5:30 PM, Sedat Dilek wrote: >>>>> Hi, >>>> Hi, >>>> >>>> You have a broken address of linux-pm (it leads to nowhere). >>>> >>> Grrr, yeah you are right. >>> >>>>> I wonder if the Kconfig logic is correct when using >>>>> CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL=y in combination with >>>>> intel_pstate driver for X86 architecture >>>>> >>>>> The available scaling-governors for intel_pstate are: performance >>>>> powersave. >>>>> >>>>> Only acpi-cpufreq can use the new schedutil governor. >>>> The right way to say that would be that intel_pstate doesn't work with >>>> cpufreq governors. >>>> >>>> scalling_governor is just used by intel_pstate as an interface for >>>> choosing the mode it works in and it always uses its own built-in >>>> "governor". And BTW, intel_pstate "powersave" is not the same as the >>>> cpufreq's "powersave" even. >>>> >>>>> Shouldn't it be prevented from selecting schedutil governor as default >>>>> when CONFIG_X86_INTEL_PSTATE=y is set? >>>> That should apply to "ondemand" too and it doesn't, so no. >>>> >>>> "schedutil" is just a regular cpufreq governor and it doesn't care about >>>> intel_pstate. >>>> >>>>> It makes no real sense to me. >>>>> >>>>> Ubuntu has in it's kernels CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y. >>>>> >>>>> What are the alternatives? >>>>> Set CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y. >>>> You can choose every cpufreq governor as the default. There are 6 of >>>> then now in the tree. >>>> >>>>> BTW, I cannot change my scaling-driver... >>>>> ...intel_pstate -> acpi-cpufreq... >>>>> >>>>> root# for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n >>>>> acpi-cpufreq > /sys/devices/system/cpu/cpufreq/${p}/scaling_driver ; >>>>> done >>>>> >>>>> Attached is also a modified /etc/init.d/ondemand script from my >>>>> Ubuntu/precise AMD64 to use and fallback to schedutil governor. >>>> No, you can't change the scaling driver this way. >>>> >>>> It actually is not possible to change from intel_pstate to anything else >>>> now. To run without intel_pstate, you need to add >>>> "intel_pstate=disable" to the kernel command line. >>>> >>> OK, I will try intel_pstate=disable in my k-c-l. >>> >>> Is there no way to switch between acpi-cpufreq and intel_pstate on the >>> fly? >> >> No, there's no way to do that ATM. >> >>> Did not look in the Kconfig sections... >>> Can I built both as modules? >>> Then unload and load the other? >> >> No, intel_pstate can't be built as a module. >> > > It's really a pity that I cannot switch on the fly between > cpufreq/schedutil vs. intel_pstate/x86 > Hmm, I have to read your replies carefully and this topic is still > confusing for me. > > I tried in the meantime "intel_pstate=disable" in my kernel-command-line. > > $ dmesg | egrep -i 'intel_pstate|cpufreq|cpuidle|cpu[0-3]' > [ 0.000000] Command line: > BOOT_IMAGE=/boot/vmlinuz-4.6.0-4-iniza-small > root=UUID=001AADA61AAD9964 loop=/ubuntu/disks/root.disk ro > intel_pstate=disable > [ 0.000000] ACPI: SSDT 0x00000000DAFEB000 000804 (v01 PmRef > Cpu0Ist 00003000 INTL 20061109) > [ 0.000000] Kernel command line: > BOOT_IMAGE=/boot/vmlinuz-4.6.0-4-iniza-small > root=UUID=001AADA61AAD9964 loop=/ubuntu/disks/root.disk ro > intel_pstate=disable > [ 0.003982] CPU0: Thermal monitoring enabled (TM1) > [ 0.075695] smpboot: CPU0: Intel(R) Core(TM) i5-2467M CPU @ 1.60GHz > (family: 0x6, model: 0x2a, stepping: 0x7) > [ 0.342317] cpuidle: using governor ladder > [ 0.358253] cpuidle: using governor menu > [ 0.422578] ACPI: SSDT 0xFFFF8800D58F9000 000688 (v01 PmRef > Cpu0Cst 00003001 INTL 20061109) > [ 0.855129] microcode: CPU0 sig=0x206a7, pf=0x10, revision=0x28 > [ 0.855157] microcode: CPU1 sig=0x206a7, pf=0x10, revision=0x28 > [ 0.855190] microcode: CPU2 sig=0x206a7, pf=0x10, revision=0x28 > [ 0.855227] microcode: CPU3 sig=0x206a7, pf=0x10, revision=0x28 > [ 85.914640] cpufreq: CPU0: Fast frequency switching not enabled > [ 85.914645] cpufreq: Registered transition notifiers: > [ 85.914702] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 > [ 85.914948] cpufreq: CPU1: Fast frequency switching not enabled > [ 85.914950] cpufreq: Registered transition notifiers: > [ 85.914954] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 > [ 85.915093] cpufreq: CPU2: Fast frequency switching not enabled > [ 85.915095] cpufreq: Registered transition notifiers: > [ 85.915099] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 > [ 85.915188] cpufreq: CPU3: Fast frequency switching not enabled > [ 85.915190] cpufreq: Registered transition notifiers: > [ 85.915193] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 > > Some sysfs-kungfu with acpi-cpufreq/schedutil ... > > # LC_ALL=C ls -lR /sys/devices/system/cpu/cpufreq/policy0/ > /sys/devices/system/cpu/cpufreq/policy0/: > total 0 > -r--r--r-- 1 root root 4096 May 19 12:54 affected_cpus > -r--r--r-- 1 root root 4096 May 19 12:54 bios_limit > -r-------- 1 root root 4096 May 19 12:54 cpuinfo_cur_freq > -r--r--r-- 1 root root 4096 May 19 12:54 cpuinfo_max_freq > -r--r--r-- 1 root root 4096 May 19 12:54 cpuinfo_min_freq > -r--r--r-- 1 root root 4096 May 19 12:54 cpuinfo_transition_latency > -r--r--r-- 1 root root 4096 May 19 12:54 freqdomain_cpus > -r--r--r-- 1 root root 4096 May 19 12:54 related_cpus > -r--r--r-- 1 root root 4096 May 19 12:54 scaling_available_frequencies > -r--r--r-- 1 root root 4096 May 19 12:54 scaling_available_governors > -r--r--r-- 1 root root 4096 May 19 12:54 scaling_cur_freq > -r--r--r-- 1 root root 4096 May 19 12:54 scaling_driver > -rw-r--r-- 1 root root 4096 May 19 11:46 scaling_governor > -rw-r--r-- 1 root root 4096 May 19 12:54 scaling_max_freq > -rw-r--r-- 1 root root 4096 May 19 12:54 scaling_min_freq > -rw-r--r-- 1 root root 4096 May 19 12:54 scaling_setspeed > drwxr-xr-x 2 root root 0 May 19 12:54 stats > > /sys/devices/system/cpu/cpufreq/policy0/stats: > total 0 > -r--r--r-- 1 root root 4096 May 19 13:00 time_in_state > -r--r--r-- 1 root root 4096 May 19 13:00 total_trans > -r--r--r-- 1 root root 4096 May 19 13:00 trans_table > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat > /sys/devices/system/cpu/cpufreq/${p}/scaling_driver ; done > acpi-cpufreq > acpi-cpufreq > acpi-cpufreq > acpi-cpufreq > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat > /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done > schedutil > schedutil > schedutil > schedutil > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat > /sys/devices/system/cpu/cpufreq/${p}/scaling_available_governors ; > done > conservative ondemand userspace powersave schedutil performance > conservative ondemand userspace powersave schedutil performance > conservative ondemand userspace powersave schedutil performance > conservative ondemand userspace powersave schedutil performance > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat > /sys/devices/system/cpu/cpufreq/${p}/scaling_available_frequencies ; > done > 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 > 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 > 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 > 1601000 1600000 1500000 1400000 1300000 1200000 1100000 1000000 900000 800000 > > # cat /sys/devices/system/cpu/cpufreq/schedutil/rate_limit_us > 10000 > > Just did not found it... > How do I enable "Fast frequency switching" on my cpu-cores? It should be enabled automatically by schedutil if acpi-cpufreq is in use, but the cpufreq_stat module had registered a transition notifier before and that prevented the fast switching from being enabled. Is schedutil the default governor in your Kconfig? If you set that, it should start before cpufreq_stat is loaded. The other way to work around it is to unload acpi-cpufreq and cpufreq_stat and then load the driver again. I have a plan to rework cpufreq_stat so it doesn't use the notifier, but that'll take some time to complete. ^ permalink raw reply [flat|nested] 15+ messages in thread
[parent not found: <CA+icZUXV3mz9Yr05o5zHLOcXwYRWydtXZOuLnE10Zy3RwzOHUg@mail.gmail.com>]
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 [not found] ` <CA+icZUXV3mz9Yr05o5zHLOcXwYRWydtXZOuLnE10Zy3RwzOHUg@mail.gmail.com> @ 2016-05-19 12:53 ` Rafael J. Wysocki 2016-05-19 13:41 ` Sedat Dilek 0 siblings, 1 reply; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 12:53 UTC (permalink / raw) To: sedat.dilek Cc: Rafael J. Wysocki, Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: > On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >>> On 5/18/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >>>> On 5/18/2016 1:16 AM, Sedat Dilek wrote: >>>>> On 5/17/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >>>>>> On 5/17/2016 5:30 PM, Sedat Dilek wrote: [cut] > [ Sorry, forgot to attach my linux-config and dmesg-output ] > > No, schedutil is not my default-governor. > >> The other way to work around it is to unload acpi-cpufreq and >> cpufreq_stat and then load the driver again. >> > > I have in my linux-config... > > # > # CPU Frequency scaling > # > CONFIG_CPU_FREQ=y > CONFIG_CPU_FREQ_GOV_ATTR_SET=y > CONFIG_CPU_FREQ_GOV_COMMON=y > CONFIG_CPU_FREQ_STAT=y > CONFIG_CPU_FREQ_STAT_DETAILS=y > 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_ONDEMAND is not set > # CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set > # CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL is not set > CONFIG_CPU_FREQ_GOV_PERFORMANCE=y > CONFIG_CPU_FREQ_GOV_POWERSAVE=y > CONFIG_CPU_FREQ_GOV_USERSPACE=y > CONFIG_CPU_FREQ_GOV_ONDEMAND=y > CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y > CONFIG_CPU_FREQ_GOV_SCHEDUTIL=y > > # > # CPU frequency scaling drivers > # > CONFIG_X86_INTEL_PSTATE=y > CONFIG_X86_PCC_CPUFREQ=y > CONFIG_X86_ACPI_CPUFREQ=y > CONFIG_X86_ACPI_CPUFREQ_CPB=y > CONFIG_X86_POWERNOW_K8=y > # CONFIG_X86_AMD_FREQ_SENSITIVITY is not set > CONFIG_X86_SPEEDSTEP_CENTRINO=y > # CONFIG_X86_P4_CLOCKMOD is not set > > Which of the Kconfigs should be compiled as a module (if available as tristate)? > > CONFIG_CPU_FREQ_STAT That needs to be a module. I guess that should be sufficient, because then you should be able to simply unload it, but do that before you switch the governor to schedutil. > CONFIG_X86_ACPI_CPUFREQ That can be a module too. > CONFIG_X86_ACPI_CPUFREQ_CPB You don't need this one on Intel hardware. ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 12:53 ` Rafael J. Wysocki @ 2016-05-19 13:41 ` Sedat Dilek 2016-05-19 13:46 ` Rafael J. Wysocki 0 siblings, 1 reply; 15+ messages in thread From: Sedat Dilek @ 2016-05-19 13:41 UTC (permalink / raw) To: Rafael J. Wysocki Cc: Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: > On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>> wrote: >>>> On 5/18/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >>>>> On 5/18/2016 1:16 AM, Sedat Dilek wrote: >>>>>> On 5/17/16, Rafael J. Wysocki <rafael.j.wysocki@intel.com> wrote: >>>>>>> On 5/17/2016 5:30 PM, Sedat Dilek wrote: > > [cut] > >> [ Sorry, forgot to attach my linux-config and dmesg-output ] >> >> No, schedutil is not my default-governor. >> >>> The other way to work around it is to unload acpi-cpufreq and >>> cpufreq_stat and then load the driver again. >>> >> >> I have in my linux-config... >> >> # >> # CPU Frequency scaling >> # >> CONFIG_CPU_FREQ=y >> CONFIG_CPU_FREQ_GOV_ATTR_SET=y >> CONFIG_CPU_FREQ_GOV_COMMON=y >> CONFIG_CPU_FREQ_STAT=y >> CONFIG_CPU_FREQ_STAT_DETAILS=y >> 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_ONDEMAND is not set >> # CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set >> # CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL is not set >> CONFIG_CPU_FREQ_GOV_PERFORMANCE=y >> CONFIG_CPU_FREQ_GOV_POWERSAVE=y >> CONFIG_CPU_FREQ_GOV_USERSPACE=y >> CONFIG_CPU_FREQ_GOV_ONDEMAND=y >> CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y >> CONFIG_CPU_FREQ_GOV_SCHEDUTIL=y >> >> # >> # CPU frequency scaling drivers >> # >> CONFIG_X86_INTEL_PSTATE=y >> CONFIG_X86_PCC_CPUFREQ=y >> CONFIG_X86_ACPI_CPUFREQ=y >> CONFIG_X86_ACPI_CPUFREQ_CPB=y >> CONFIG_X86_POWERNOW_K8=y >> # CONFIG_X86_AMD_FREQ_SENSITIVITY is not set >> CONFIG_X86_SPEEDSTEP_CENTRINO=y >> # CONFIG_X86_P4_CLOCKMOD is not set >> >> Which of the Kconfigs should be compiled as a module (if available as >> tristate)? >> >> CONFIG_CPU_FREQ_STAT > > That needs to be a module. > > I guess that should be sufficient, because then you should be able to > simply unload it, but do that before you switch the governor to > schedutil. > >> CONFIG_X86_ACPI_CPUFREQ > > That can be a module too. > >> CONFIG_X86_ACPI_CPUFREQ_CPB > > You don't need this one on Intel hardware. > With these changes... $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small /boot/config-4.6.0-5-iniza-small CPU_FREQ_STAT y -> m X86_ACPI_CPUFREQ y -> m X86_POWERNOW_K8 y -> m ...I still see "Fast frequency switching not enabled"... $ sudo modprobe -r -v acpi_cpufreq [sudo] password for wearefam: rmmod acpi_cpufreq $ sudo modprobe -r -v cpufreq_stats rmmod cpufreq_stats $ sudo modprobe -v cpufreq_stats insmod /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko $ sudo modprobe -v acpi_cpufreq insmod /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/acpi-cpufreq.ko # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done performance performance performance performance # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n schedutil > /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done schedutil schedutil schedutil schedutil $ dmesg | tail [ 756.114412] cpufreq: CPU0: Fast frequency switching not enabled [ 756.114416] cpufreq: Registered transition notifiers: [ 756.114424] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] [ 756.114725] cpufreq: CPU1: Fast frequency switching not enabled [ 756.114727] cpufreq: Registered transition notifiers: [ 756.114731] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] [ 756.114996] cpufreq: CPU2: Fast frequency switching not enabled [ 756.114998] cpufreq: Registered transition notifiers: [ 756.115002] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] [ 756.115294] cpufreq: CPU3: Fast frequency switching not enabled [ 756.115296] cpufreq: Registered transition notifiers: [ 756.115299] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] Do I have to enable some CPU setting in the BIOS? - Sedat - ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 13:41 ` Sedat Dilek @ 2016-05-19 13:46 ` Rafael J. Wysocki 2016-05-19 13:54 ` Sedat Dilek 0 siblings, 1 reply; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 13:46 UTC (permalink / raw) To: sedat.dilek Cc: Rafael J. Wysocki, Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: > On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>> wrote: [cut] > > With these changes... > > $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small > /boot/config-4.6.0-5-iniza-small > CPU_FREQ_STAT y -> m > X86_ACPI_CPUFREQ y -> m > X86_POWERNOW_K8 y -> m > > ...I still see "Fast frequency switching not enabled"... > > $ sudo modprobe -r -v acpi_cpufreq > [sudo] password for wearefam: > rmmod acpi_cpufreq > $ sudo modprobe -r -v cpufreq_stats > rmmod cpufreq_stats > > $ sudo modprobe -v cpufreq_stats > insmod /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko Skip the above modprobe. :-) > $ sudo modprobe -v acpi_cpufreq > insmod /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/acpi-cpufreq.ko > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat > /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done > performance > performance > performance > performance > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n > schedutil > /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; > done > > # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat > /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done > schedutil > schedutil > schedutil > schedutil > > $ dmesg | tail > > [ 756.114412] cpufreq: CPU0: Fast frequency switching not enabled > [ 756.114416] cpufreq: Registered transition notifiers: > [ 756.114424] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] > [ 756.114725] cpufreq: CPU1: Fast frequency switching not enabled > [ 756.114727] cpufreq: Registered transition notifiers: > [ 756.114731] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] > [ 756.114996] cpufreq: CPU2: Fast frequency switching not enabled > [ 756.114998] cpufreq: Registered transition notifiers: > [ 756.115002] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] > [ 756.115294] cpufreq: CPU3: Fast frequency switching not enabled > [ 756.115296] cpufreq: Registered transition notifiers: > [ 756.115299] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 [cpufreq_stats] > > Do I have to enable some CPU setting in the BIOS? No, that's a software issue purely. ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 13:46 ` Rafael J. Wysocki @ 2016-05-19 13:54 ` Sedat Dilek 2016-05-19 14:04 ` Rafael J. Wysocki 0 siblings, 1 reply; 15+ messages in thread From: Sedat Dilek @ 2016-05-19 13:54 UTC (permalink / raw) To: Rafael J. Wysocki Cc: Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: > On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>> wrote: >>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>> wrote: > > [cut] > >> >> With these changes... >> >> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >> /boot/config-4.6.0-5-iniza-small >> CPU_FREQ_STAT y -> m >> X86_ACPI_CPUFREQ y -> m >> X86_POWERNOW_K8 y -> m >> >> ...I still see "Fast frequency switching not enabled"... >> >> $ sudo modprobe -r -v acpi_cpufreq >> [sudo] password for wearefam: >> rmmod acpi_cpufreq >> $ sudo modprobe -r -v cpufreq_stats >> rmmod cpufreq_stats >> >> $ sudo modprobe -v cpufreq_stats >> insmod >> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko > > Skip the above modprobe. :-) > If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do not see anything with "cpu fast switching" in my logs. ( I double-checked to have schedutil as running cpufreq-governor. ) - Sedat - >> $ sudo modprobe -v acpi_cpufreq >> insmod >> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/acpi-cpufreq.ko >> >> # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat >> /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done >> performance >> performance >> performance >> performance >> >> # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C echo -n >> schedutil > /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; >> done >> >> # for p in policy0 policy1 policy2 policy3 ; do LC_ALL=C cat >> /sys/devices/system/cpu/cpufreq/${p}/scaling_governor ; done >> schedutil >> schedutil >> schedutil >> schedutil >> >> $ dmesg | tail >> >> [ 756.114412] cpufreq: CPU0: Fast frequency switching not enabled >> [ 756.114416] cpufreq: Registered transition notifiers: >> [ 756.114424] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 >> [cpufreq_stats] >> [ 756.114725] cpufreq: CPU1: Fast frequency switching not enabled >> [ 756.114727] cpufreq: Registered transition notifiers: >> [ 756.114731] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 >> [cpufreq_stats] >> [ 756.114996] cpufreq: CPU2: Fast frequency switching not enabled >> [ 756.114998] cpufreq: Registered transition notifiers: >> [ 756.115002] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 >> [cpufreq_stats] >> [ 756.115294] cpufreq: CPU3: Fast frequency switching not enabled >> [ 756.115296] cpufreq: Registered transition notifiers: >> [ 756.115299] cpufreq: cpufreq_stat_notifier_trans+0x0/0xc0 >> [cpufreq_stats] >> >> Do I have to enable some CPU setting in the BIOS? > > No, that's a software issue purely. > ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 13:54 ` Sedat Dilek @ 2016-05-19 14:04 ` Rafael J. Wysocki 2016-05-19 14:24 ` Sedat Dilek 2016-05-19 14:24 ` Rafael J. Wysocki 0 siblings, 2 replies; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 14:04 UTC (permalink / raw) To: sedat.dilek Cc: Rafael J. Wysocki, Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: > On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>> wrote: >>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>> wrote: >> >> [cut] >> >>> >>> With these changes... >>> >>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>> /boot/config-4.6.0-5-iniza-small >>> CPU_FREQ_STAT y -> m >>> X86_ACPI_CPUFREQ y -> m >>> X86_POWERNOW_K8 y -> m >>> >>> ...I still see "Fast frequency switching not enabled"... >>> >>> $ sudo modprobe -r -v acpi_cpufreq >>> [sudo] password for wearefam: >>> rmmod acpi_cpufreq >>> $ sudo modprobe -r -v cpufreq_stats >>> rmmod cpufreq_stats >>> >>> $ sudo modprobe -v cpufreq_stats >>> insmod >>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >> >> Skip the above modprobe. :-) >> > > If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do > not see anything with "cpu fast switching" in my logs. > ( I double-checked to have schedutil as running cpufreq-governor. ) That means it has been enabled. ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 14:04 ` Rafael J. Wysocki @ 2016-05-19 14:24 ` Sedat Dilek 2016-05-19 14:25 ` Rafael J. Wysocki 2016-05-19 14:24 ` Rafael J. Wysocki 1 sibling, 1 reply; 15+ messages in thread From: Sedat Dilek @ 2016-05-19 14:24 UTC (permalink / raw) To: Rafael J. Wysocki Cc: Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: > On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> >>> wrote: >>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>> wrote: >>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>>> wrote: >>> >>> [cut] >>> >>>> >>>> With these changes... >>>> >>>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>>> /boot/config-4.6.0-5-iniza-small >>>> CPU_FREQ_STAT y -> m >>>> X86_ACPI_CPUFREQ y -> m >>>> X86_POWERNOW_K8 y -> m >>>> >>>> ...I still see "Fast frequency switching not enabled"... >>>> >>>> $ sudo modprobe -r -v acpi_cpufreq >>>> [sudo] password for wearefam: >>>> rmmod acpi_cpufreq >>>> $ sudo modprobe -r -v cpufreq_stats >>>> rmmod cpufreq_stats >>>> >>>> $ sudo modprobe -v cpufreq_stats >>>> insmod >>>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >>> >>> Skip the above modprobe. :-) >>> >> >> If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do >> not see anything with "cpu fast switching" in my logs. >> ( I double-checked to have schedutil as running cpufreq-governor. ) > > That means it has been enabled. > How can I check if cpu fast switching is really enabled? I checked with the dmesg before. - Sedat - ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 14:24 ` Sedat Dilek @ 2016-05-19 14:25 ` Rafael J. Wysocki 0 siblings, 0 replies; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 14:25 UTC (permalink / raw) To: sedat.dilek Cc: Rafael J. Wysocki, Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 4:24 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: > On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >> On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>> wrote: >>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>> wrote: >>>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>>>> wrote: >>>> >>>> [cut] >>>> >>>>> >>>>> With these changes... >>>>> >>>>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>>>> /boot/config-4.6.0-5-iniza-small >>>>> CPU_FREQ_STAT y -> m >>>>> X86_ACPI_CPUFREQ y -> m >>>>> X86_POWERNOW_K8 y -> m >>>>> >>>>> ...I still see "Fast frequency switching not enabled"... >>>>> >>>>> $ sudo modprobe -r -v acpi_cpufreq >>>>> [sudo] password for wearefam: >>>>> rmmod acpi_cpufreq >>>>> $ sudo modprobe -r -v cpufreq_stats >>>>> rmmod cpufreq_stats >>>>> >>>>> $ sudo modprobe -v cpufreq_stats >>>>> insmod >>>>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >>>> >>>> Skip the above modprobe. :-) >>>> >>> >>> If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do >>> not see anything with "cpu fast switching" in my logs. >>> ( I double-checked to have schedutil as running cpufreq-governor. ) >> >> That means it has been enabled. >> > > How can I check if cpu fast switching is really enabled? > I checked with the dmesg before. Please see the message I've just sent: http://marc.info/?l=linux-pm&m=146366787406347&w=2 ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 14:04 ` Rafael J. Wysocki 2016-05-19 14:24 ` Sedat Dilek @ 2016-05-19 14:24 ` Rafael J. Wysocki 2016-05-19 14:50 ` Sedat Dilek 1 sibling, 1 reply; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 14:24 UTC (permalink / raw) To: Rafael J. Wysocki Cc: sedat.dilek, Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 4:04 PM, Rafael J. Wysocki <rafael@kernel.org> wrote: > On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>> wrote: >>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>>> wrote: >>> >>> [cut] >>> >>>> >>>> With these changes... >>>> >>>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>>> /boot/config-4.6.0-5-iniza-small >>>> CPU_FREQ_STAT y -> m >>>> X86_ACPI_CPUFREQ y -> m >>>> X86_POWERNOW_K8 y -> m >>>> >>>> ...I still see "Fast frequency switching not enabled"... >>>> >>>> $ sudo modprobe -r -v acpi_cpufreq >>>> [sudo] password for wearefam: >>>> rmmod acpi_cpufreq >>>> $ sudo modprobe -r -v cpufreq_stats >>>> rmmod cpufreq_stats >>>> >>>> $ sudo modprobe -v cpufreq_stats >>>> insmod >>>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >>> >>> Skip the above modprobe. :-) >>> >> >> If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do >> not see anything with "cpu fast switching" in my logs. >> ( I double-checked to have schedutil as running cpufreq-governor. ) > > That means it has been enabled. Actually, you can check if it has been enabled by looking and the wakeups statistics from powertop. Without fast switching you'll see kWork with sugov_work under "Description" somewhere close to the top of the list. ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 14:24 ` Rafael J. Wysocki @ 2016-05-19 14:50 ` Sedat Dilek 2016-05-19 16:05 ` Rafael J. Wysocki 0 siblings, 1 reply; 15+ messages in thread From: Sedat Dilek @ 2016-05-19 14:50 UTC (permalink / raw) To: Rafael J. Wysocki Cc: Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: > On Thu, May 19, 2016 at 4:04 PM, Rafael J. Wysocki <rafael@kernel.org> > wrote: >> On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> >> wrote: >>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>> wrote: >>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>> wrote: >>>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek >>>>>>>> <sedat.dilek@gmail.com> >>>>>>>> wrote: >>>> >>>> [cut] >>>> >>>>> >>>>> With these changes... >>>>> >>>>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>>>> /boot/config-4.6.0-5-iniza-small >>>>> CPU_FREQ_STAT y -> m >>>>> X86_ACPI_CPUFREQ y -> m >>>>> X86_POWERNOW_K8 y -> m >>>>> >>>>> ...I still see "Fast frequency switching not enabled"... >>>>> >>>>> $ sudo modprobe -r -v acpi_cpufreq >>>>> [sudo] password for wearefam: >>>>> rmmod acpi_cpufreq >>>>> $ sudo modprobe -r -v cpufreq_stats >>>>> rmmod cpufreq_stats >>>>> >>>>> $ sudo modprobe -v cpufreq_stats >>>>> insmod >>>>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >>>> >>>> Skip the above modprobe. :-) >>>> >>> >>> If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do >>> not see anything with "cpu fast switching" in my logs. >>> ( I double-checked to have schedutil as running cpufreq-governor. ) >> >> That means it has been enabled. > > Actually, you can check if it has been enabled by looking and the > wakeups statistics from powertop. > > Without fast switching you'll see kWork with sugov_work under > "Description" somewhere close to the top of the list. > The software-repos for Ubuntu/precise lists me a PowerTOP version 1.97. Not sure if this version has all bits to show me cpu fast switching (short: c-f-s). IOW, it is a bit uncool to see it "not enabled" in my logs. I would expect also to see the successful initialization/availibility in my logs. Does turbostats (shipped with the kernel-sources) support c-f-s? - Sedat - [1] list PowerTOP v2.8 [1] https://01.org/powertop ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 14:50 ` Sedat Dilek @ 2016-05-19 16:05 ` Rafael J. Wysocki 2016-05-19 16:16 ` Sedat Dilek 0 siblings, 1 reply; 15+ messages in thread From: Rafael J. Wysocki @ 2016-05-19 16:05 UTC (permalink / raw) To: sedat.dilek Cc: Rafael J. Wysocki, Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On Thu, May 19, 2016 at 4:50 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: > On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >> On Thu, May 19, 2016 at 4:04 PM, Rafael J. Wysocki <rafael@kernel.org> >> wrote: >>> On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> >>> wrote: >>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>> wrote: >>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>>> wrote: >>>>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek >>>>>>>>> <sedat.dilek@gmail.com> >>>>>>>>> wrote: >>>>> >>>>> [cut] >>>>> >>>>>> >>>>>> With these changes... >>>>>> >>>>>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>>>>> /boot/config-4.6.0-5-iniza-small >>>>>> CPU_FREQ_STAT y -> m >>>>>> X86_ACPI_CPUFREQ y -> m >>>>>> X86_POWERNOW_K8 y -> m >>>>>> >>>>>> ...I still see "Fast frequency switching not enabled"... >>>>>> >>>>>> $ sudo modprobe -r -v acpi_cpufreq >>>>>> [sudo] password for wearefam: >>>>>> rmmod acpi_cpufreq >>>>>> $ sudo modprobe -r -v cpufreq_stats >>>>>> rmmod cpufreq_stats >>>>>> >>>>>> $ sudo modprobe -v cpufreq_stats >>>>>> insmod >>>>>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >>>>> >>>>> Skip the above modprobe. :-) >>>>> >>>> >>>> If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do >>>> not see anything with "cpu fast switching" in my logs. >>>> ( I double-checked to have schedutil as running cpufreq-governor. ) >>> >>> That means it has been enabled. >> >> Actually, you can check if it has been enabled by looking and the >> wakeups statistics from powertop. >> >> Without fast switching you'll see kWork with sugov_work under >> "Description" somewhere close to the top of the list. >> > > The software-repos for Ubuntu/precise lists me a PowerTOP version 1.97. > Not sure if this version has all bits to show me cpu fast switching > (short: c-f-s). Any version of powertop should be sufficient. You just need to go into the "wakups" screen and look for "sugov_work" in the output. If it's not there, you have fast switching enabled. > IOW, it is a bit uncool to see it "not enabled" in my logs. > I would expect also to see the successful initialization/availibility > in my logs. The reason why "not enabled" is visible is because that indicates some conflict with cpufreq transition notifiers and we want to know about those. > > Does turbostats (shipped with the kernel-sources) support c-f-s? That's not a hardware feature, so why would it? Why do you care so much about the fast switching in the first place? ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 2016-05-19 16:05 ` Rafael J. Wysocki @ 2016-05-19 16:16 ` Sedat Dilek 0 siblings, 0 replies; 15+ messages in thread From: Sedat Dilek @ 2016-05-19 16:16 UTC (permalink / raw) To: Rafael J. Wysocki Cc: Rafael J. Wysocki, Pandruvada, Srinivas, linux-pm@vger.kernel.org On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: > On Thu, May 19, 2016 at 4:50 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote: >> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>> On Thu, May 19, 2016 at 4:04 PM, Rafael J. Wysocki <rafael@kernel.org> >>> wrote: >>>> On Thu, May 19, 2016 at 3:54 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>> wrote: >>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>> On Thu, May 19, 2016 at 3:41 PM, Sedat Dilek <sedat.dilek@gmail.com> >>>>>> wrote: >>>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>>> On Thu, May 19, 2016 at 2:10 PM, Sedat Dilek >>>>>>>> <sedat.dilek@gmail.com> >>>>>>>> wrote: >>>>>>>>> On 5/19/16, Rafael J. Wysocki <rafael@kernel.org> wrote: >>>>>>>>>> On Thu, May 19, 2016 at 1:04 PM, Sedat Dilek >>>>>>>>>> <sedat.dilek@gmail.com> >>>>>>>>>> wrote: >>>>>> >>>>>> [cut] >>>>>> >>>>>>> >>>>>>> With these changes... >>>>>>> >>>>>>> $ ./scripts/diffconfig /boot/config-4.6.0-4-iniza-small >>>>>>> /boot/config-4.6.0-5-iniza-small >>>>>>> CPU_FREQ_STAT y -> m >>>>>>> X86_ACPI_CPUFREQ y -> m >>>>>>> X86_POWERNOW_K8 y -> m >>>>>>> >>>>>>> ...I still see "Fast frequency switching not enabled"... >>>>>>> >>>>>>> $ sudo modprobe -r -v acpi_cpufreq >>>>>>> [sudo] password for wearefam: >>>>>>> rmmod acpi_cpufreq >>>>>>> $ sudo modprobe -r -v cpufreq_stats >>>>>>> rmmod cpufreq_stats >>>>>>> >>>>>>> $ sudo modprobe -v cpufreq_stats >>>>>>> insmod >>>>>>> /lib/modules/4.6.0-5-iniza-small/kernel/drivers/cpufreq/cpufreq_stats.ko >>>>>> >>>>>> Skip the above modprobe. :-) >>>>>> >>>>> >>>>> If I do not insmod cpufreq_stats (only insmod acpi_cpufreq) then I do >>>>> not see anything with "cpu fast switching" in my logs. >>>>> ( I double-checked to have schedutil as running cpufreq-governor. ) >>>> >>>> That means it has been enabled. >>> >>> Actually, you can check if it has been enabled by looking and the >>> wakeups statistics from powertop. >>> >>> Without fast switching you'll see kWork with sugov_work under >>> "Description" somewhere close to the top of the list. >>> >> >> The software-repos for Ubuntu/precise lists me a PowerTOP version 1.97. >> Not sure if this version has all bits to show me cpu fast switching >> (short: c-f-s). > > Any version of powertop should be sufficient. You just need to go > into the "wakups" screen and look for "sugov_work" in the output. If > it's not there, you have fast switching enabled. > I needed to hack Makefile.am and configure.ac of powertop v2.8 to be buildable with my old autoconf/automake installed here. With *no cpu-fast-switching* I can see *sugov_work* in the Overview section. PowerTOP 2.8 Overview Idle stats Frequency stats Device stats Tunables Summary: 237.6 wakeups/second, 0.0 GPU ops/seconds, 0.0 VFS ops/sec and 7.2% CPU use Usage Events/s Category Description 100.0% Device Audio codec hwC0D0: Realtek 100.0% Device Audio codec hwC0D3: Intel 1.1 ms/s 103.4 kWork sugov_work ... >> IOW, it is a bit uncool to see it "not enabled" in my logs. >> I would expect also to see the successful initialization/availibility >> in my logs. > > The reason why "not enabled" is visible is because that indicates some > conflict with cpufreq transition notifiers and we want to know about > those. > Anyway, I expect both status to see for cpu-fast-switching. >> >> Does turbostats (shipped with the kernel-sources) support c-f-s? > > That's not a hardware feature, so why would it? > > Why do you care so much about the fast switching in the first place? > I know zero about it - in this minute and before this minute :-). - Sedat - ^ permalink raw reply [flat|nested] 15+ messages in thread
end of thread, other threads:[~2016-05-19 16:16 UTC | newest]
Thread overview: 15+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <CA+icZUV0Mxu+fvrmpJmnDjLmxpSd83hDM5od0_1kw4gEm5aaEw@mail.gmail.com>
[not found] ` <573B4C2A.4010205@intel.com>
2016-05-17 23:16 ` [cpufreq-4.7] New schedutil governor in combination with intel_pstate/x86 Sedat Dilek
2016-05-17 23:18 ` Rafael J. Wysocki
2016-05-19 11:04 ` Sedat Dilek
2016-05-19 11:54 ` Rafael J. Wysocki
[not found] ` <CA+icZUXV3mz9Yr05o5zHLOcXwYRWydtXZOuLnE10Zy3RwzOHUg@mail.gmail.com>
2016-05-19 12:53 ` Rafael J. Wysocki
2016-05-19 13:41 ` Sedat Dilek
2016-05-19 13:46 ` Rafael J. Wysocki
2016-05-19 13:54 ` Sedat Dilek
2016-05-19 14:04 ` Rafael J. Wysocki
2016-05-19 14:24 ` Sedat Dilek
2016-05-19 14:25 ` Rafael J. Wysocki
2016-05-19 14:24 ` Rafael J. Wysocki
2016-05-19 14:50 ` Sedat Dilek
2016-05-19 16:05 ` Rafael J. Wysocki
2016-05-19 16:16 ` Sedat Dilek
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).