Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1674004 > unrolled thread

[PATCH 0/4 v2] x86,cpufreq: unify APERF/MPERF computation

Back to article view | Back to linux.kernel

Contents

  [PATCH 0/4 v2] x86,cpufreq: unify APERF/MPERF computation Len Brown <lenb@kernel.org> - 2017-06-24 07:20 +0200
    [PATCH 2/4 v2] x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF Len Brown <lenb@kernel.org> - 2017-06-24 07:20 +0200
      Re: [PATCH 2/4 v2] x86: use common aperfmperf_khz_on_cpu() to  calculate KHz using APERF/MPERF Thomas Gleixner <tglx@linutronix.de> - 2017-06-24 11:00 +0200
        Re: [PATCH 2/4 v2] x86: use common aperfmperf_khz_on_cpu() to  calculate KHz using APERF/MPERF "Rafael J. Wysocki" <rafael@kernel.org> - 2017-06-24 14:10 +0200
    [PATCH 3/4] intel_pstate: delete scheduler hook in HWP mode Len Brown <lenb@kernel.org> - 2017-06-24 07:20 +0200

#1674004 — [PATCH 0/4 v2] x86,cpufreq: unify APERF/MPERF computation

Hi Rafael, Thomas,

Please see the updated 2nd patch in this series -- in response to tglx review.
Patch 1,3,4 are unchanged.

thanks,
-Len

This patch series has 3 goals:

1. Make "cpu MHz" in /proc/cpuinfo supportable.

2. Make /sys/.../cpufreq/scaling_cur_freq meaningful
   and consistent on modern x86 systems.

3. Use 1. and 2. to remove scheduler and cpufreq overhead

There are 3 main changes since this series was proposed
about a year ago:

This update responds to distro feedback to make /proc/cpuinfo
"cpu MHz" constant.  Originally, we had proposed making it return
the same dynamic value as cpufreq sysfs.

Some community members suggested that sysfs MHz values should
be meaninful, even down to 10ms intervals.  So this has been
changed, versus the original proposal to not re-compute
at intervals shorter than 100ms.

(For those who really care about observing frequency, the
 recommendation remains to use turbostat(8) or equivalent utility,
 which can reliably measure concurrent intervals of arbitrary length)

The intel_pstate sampling mechanism has changed.
Originally this series removed an intel_pstate timer in HWP mode.
Now it removes the analogous scheduler call-back.

Most recently, in response to posting this patch on the list
about 10-days ago, the patch to remove frequency calculation
from inside intel_pstate was dropped, in order to maintain compatibility
with tracing scripts.  Also, the order of the last two patches
has been exchanged.

Please let me know if you see any issues with this series.

thanks!
Len Brown, Intel Open Source Technology Center

The following changes since commit 3c2993b8c6143d8a5793746a54eba8f86f95240f:

  Linux 4.12-rc4 (2017-06-04 16:47:43 -0700)

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux.git x86

for you to fetch changes up to 68516d288d3968fe22d6c8984a7bcbdcdbed351d:

  intel_pstate: skip scheduler hook when in "performance" mode. (2017-06-23 22:01:46 -0700)

----------------------------------------------------------------
Len Brown (4):
      x86: do not use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz"
      x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF
      intel_pstate: delete scheduler hook in HWP mode
      intel_pstate: skip scheduler hook when in "performance" mode.

 arch/x86/kernel/cpu/Makefile     |  1 +
 arch/x86/kernel/cpu/aperfmperf.c | 79 ++++++++++++++++++++++++++++++++++++++++
 arch/x86/kernel/cpu/proc.c       | 10 +----
 drivers/cpufreq/cpufreq.c        | 12 +++++-
 drivers/cpufreq/intel_pstate.c   | 18 +++------
 include/linux/cpufreq.h          |  2 +
 6 files changed, 100 insertions(+), 22 deletions(-)
 create mode 100644 arch/x86/kernel/cpu/aperfmperf.c

[toc] | [next] | [standalone]

#1674005 — [PATCH 2/4 v2] x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF

From: Len Brown <len.brown@intel.com>

The goal of this change is to give users a uniform and meaningful
result when they read /sys/...cpufreq/scaling_cur_freq
on modern x86 hardware, as compared to what they get today.

Modern x86 processors include the hardware needed
to accurately calculate frequency over an interval --
APERF, MPERF, and the TSC.

Here we provide an x86 routine to make this calculation
on supported hardware, and use it in preference to any
driver driver-specific cpufreq_driver.get() routine.

MHz is computed like so:

MHz = base_MHz * delta_APERF / delta_MPERF

MHz is the average frequency of the busy processor
over a measurement interval.  The interval is
defined to be the time between successive invocations
of aperfmperf_khz_on_cpu(), which are expected to to
happen on-demand when users read sysfs attribute
cpufreq/scaling_cur_freq.

As with previous methods of calculating MHz,
idle time is excluded.

base_MHz above is from TSC calibration global "cpu_khz".

This x86 native method to calculate MHz returns a meaningful result
no matter if P-states are controlled by hardware or firmware
and/or if the Linux cpufreq sub-system is or is-not installed.

When this routine is invoked more frequently, the measurement
interval becomes shorter.  However, the code limits re-computation
to 10ms intervals so that average frequency remains meaningful.

Discerning users are encouraged to take advantage of
the turbostat(8) utility, which can gracefully handle
concurrent measurement intervals of arbitrary length.

Signed-off-by: Len Brown <len.brown@intel.com>
---
 arch/x86/kernel/cpu/Makefile     |  1 +
 arch/x86/kernel/cpu/aperfmperf.c | 79 ++++++++++++++++++++++++++++++++++++++++
 drivers/cpufreq/cpufreq.c        | 12 +++++-
 include/linux/cpufreq.h          |  2 +
 4 files changed, 93 insertions(+), 1 deletion(-)
 create mode 100644 arch/x86/kernel/cpu/aperfmperf.c

diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile
index 5200001..cdf8249 100644
--- a/arch/x86/kernel/cpu/Makefile
+++ b/arch/x86/kernel/cpu/Makefile
@@ -21,6 +21,7 @@ obj-y			+= common.o
 obj-y			+= rdrand.o
 obj-y			+= match.o
 obj-y			+= bugs.o
+obj-$(CONFIG_CPU_FREQ)	+= aperfmperf.o
 
 obj-$(CONFIG_PROC_FS)	+= proc.o
 obj-$(CONFIG_X86_FEATURE_NAMES) += capflags.o powerflags.o
diff --git a/arch/x86/kernel/cpu/aperfmperf.c b/arch/x86/kernel/cpu/aperfmperf.c
new file mode 100644
index 0000000..d869c86
--- /dev/null
+++ b/arch/x86/kernel/cpu/aperfmperf.c
@@ -0,0 +1,79 @@
+/*
+ * x86 APERF/MPERF KHz calculation for
+ * /sys/.../cpufreq/scaling_cur_freq
+ *
+ * Copyright (C) 2017 Intel Corp.
+ * Author: Len Brown <len.brown@intel.com>
+ *
+ * This file is licensed under GPLv2.
+ */
+
+#include <linux/jiffies.h>
+#include <linux/math64.h>
+#include <linux/percpu.h>
+#include <linux/smp.h>
+
+struct aperfmperf_sample {
+	unsigned int	khz;
+	unsigned long	jiffies;
+	u64	aperf;
+	u64	mperf;
+};
+
+static DEFINE_PER_CPU(struct aperfmperf_sample, samples);
+
+/*
+ * aperfmperf_snapshot_khz()
+ * On the current CPU, snapshot APERF, MPERF, and jiffies
+ * unless we already did it within 10ms
+ * calculate kHz, save snapshot
+ */
+static void aperfmperf_snapshot_khz(void *dummy)
+{
+	u64 aperf, aperf_delta;
+	u64 mperf, mperf_delta;
+	struct aperfmperf_sample *s = this_cpu_ptr(&samples);
+
+	/* Don't bother re-computing within 10 ms */
+	if (time_before(jiffies, s->jiffies + HZ/100))
+		return;
+
+	rdmsrl(MSR_IA32_APERF, aperf);
+	rdmsrl(MSR_IA32_MPERF, mperf);
+
+	aperf_delta = aperf - s->aperf;
+	mperf_delta = mperf - s->mperf;
+
+	/*
+	 * There is no architectural guarantee that MPERF
+	 * increments faster than we can read it.
+	 */
+	if (mperf_delta == 0)
+		return;
+
+	/*
+	 * if (cpu_khz * aperf_delta) fits into ULLONG_MAX, then
+	 *	khz = (cpu_khz * aperf_delta) / mperf_delta
+	 */
+	if (div64_u64(ULLONG_MAX, cpu_khz) > aperf_delta)
+		s->khz = div64_u64((cpu_khz * aperf_delta), mperf_delta);
+	else	/* khz = aperf_delta / (mperf_delta / cpu_khz) */
+		s->khz = div64_u64(aperf_delta,
+			div64_u64(mperf_delta, cpu_khz));
+	s->jiffies = jiffies;
+	s->aperf = aperf;
+	s->mperf = mperf;
+}
+
+unsigned int arch_freq_get_on_cpu(int cpu)
+{
+	if (!cpu_khz)
+		return 0;
+
+	if (!static_cpu_has(X86_FEATURE_APERFMPERF))
+		return 0;
+
+	smp_call_function_single(cpu, aperfmperf_snapshot_khz, NULL, 1);
+
+	return per_cpu(samples.khz, cpu);
+}
diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index 26b643d..6e7424d 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -632,11 +632,21 @@ show_one(cpuinfo_transition_latency, cpuinfo.transition_latency);
 show_one(scaling_min_freq, min);
 show_one(scaling_max_freq, max);
 
+__weak unsigned int arch_freq_get_on_cpu(int cpu)
+{
+	return 0;
+}
+
 static ssize_t show_scaling_cur_freq(struct cpufreq_policy *policy, char *buf)
 {
 	ssize_t ret;
+	unsigned int freq;
 
-	if (cpufreq_driver && cpufreq_driver->setpolicy && cpufreq_driver->get)
+	freq = arch_freq_get_on_cpu(policy->cpu);
+	if (freq)
+		ret = sprintf(buf, "%u\n", freq);
+	else if (cpufreq_driver && cpufreq_driver->setpolicy &&
+			cpufreq_driver->get)
 		ret = sprintf(buf, "%u\n", cpufreq_driver->get(policy->cpu));
 	else
 		ret = sprintf(buf, "%u\n", policy->cur);
diff --git a/include/linux/cpufreq.h b/include/linux/cpufreq.h
index a5ce0bbe..905117b 100644
--- a/include/linux/cpufreq.h
+++ b/include/linux/cpufreq.h
@@ -883,6 +883,8 @@ static inline bool policy_has_boost_freq(struct cpufreq_policy *policy)
 }
 #endif
 
+extern unsigned int arch_freq_get_on_cpu(int cpu);
+
 /* the following are really really optional */
 extern struct freq_attr cpufreq_freq_attr_scaling_available_freqs;
 extern struct freq_attr cpufreq_freq_attr_scaling_boost_freqs;
-- 
2.7.4

[toc] | [prev] | [next] | [standalone]

#1674053 — Re: [PATCH 2/4 v2] x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF

On Fri, 23 Jun 2017, Len Brown wrote:
> This x86 native method to calculate MHz returns a meaningful result
> no matter if P-states are controlled by hardware or firmware
> and/or if the Linux cpufreq sub-system is or is-not installed.
> 
> When this routine is invoked more frequently, the measurement
> interval becomes shorter.  However, the code limits re-computation
> to 10ms intervals so that average frequency remains meaningful.
> 
> Discerning users are encouraged to take advantage of
> the turbostat(8) utility, which can gracefully handle
> concurrent measurement intervals of arbitrary length.
> 
> Signed-off-by: Len Brown <len.brown@intel.com>

Reviewed-by: Thomas Gleixner <tglx@linutronix.de>

Raphael, please take the whole lot through the cpufreq tree.

Thanks,

	tglx

[toc] | [prev] | [next] | [standalone]

#1674073 — Re: [PATCH 2/4 v2] x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF

On Sat, Jun 24, 2017 at 10:56 AM, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Fri, 23 Jun 2017, Len Brown wrote:
>> This x86 native method to calculate MHz returns a meaningful result
>> no matter if P-states are controlled by hardware or firmware
>> and/or if the Linux cpufreq sub-system is or is-not installed.
>>
>> When this routine is invoked more frequently, the measurement
>> interval becomes shorter.  However, the code limits re-computation
>> to 10ms intervals so that average frequency remains meaningful.
>>
>> Discerning users are encouraged to take advantage of
>> the turbostat(8) utility, which can gracefully handle
>> concurrent measurement intervals of arbitrary length.
>>
>> Signed-off-by: Len Brown <len.brown@intel.com>
>
> Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
>
> Raphael, please take the whole lot through the cpufreq tree.

I will, thanks!

Rafael

[toc] | [prev] | [next] | [standalone]

#1674006 — [PATCH 3/4] intel_pstate: delete scheduler hook in HWP mode

From: Len Brown <len.brown@intel.com>

The cpufreq/scaling_cur_freq sysfs attribute is now provided by
shared x86 cpufreq code on modern x86 systems, including
all systems supported by the intel_pstate driver.

In HWP mode, maintaining that value was the sole purpose of
the scheduler hook, intel_pstate_update_util_hwp(),
so it can now be removed.

Signed-off-by: Len Brown <len.brown@intel.com>
---
 drivers/cpufreq/intel_pstate.c | 14 +++-----------
 1 file changed, 3 insertions(+), 11 deletions(-)

diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
index b7de5bd..4ec5668 100644
--- a/drivers/cpufreq/intel_pstate.c
+++ b/drivers/cpufreq/intel_pstate.c
@@ -1732,16 +1732,6 @@ static void intel_pstate_adjust_pstate(struct cpudata *cpu, int target_pstate)
 		fp_toint(cpu->iowait_boost * 100));
 }
 
-static void intel_pstate_update_util_hwp(struct update_util_data *data,
-					 u64 time, unsigned int flags)
-{
-	struct cpudata *cpu = container_of(data, struct cpudata, update_util);
-	u64 delta_ns = time - cpu->sample.time;
-
-	if ((s64)delta_ns >= INTEL_PSTATE_HWP_SAMPLING_INTERVAL)
-		intel_pstate_sample(cpu, time);
-}
-
 static void intel_pstate_update_util_pid(struct update_util_data *data,
 					 u64 time, unsigned int flags)
 {
@@ -1933,6 +1923,9 @@ static void intel_pstate_set_update_util_hook(unsigned int cpu_num)
 {
 	struct cpudata *cpu = all_cpu_data[cpu_num];
 
+	if (hwp_active)
+		return;
+
 	if (cpu->update_util_set)
 		return;
 
@@ -2557,7 +2550,6 @@ static int __init intel_pstate_init(void)
 		} else {
 			hwp_active++;
 			intel_pstate.attr = hwp_cpufreq_attrs;
-			pstate_funcs.update_util = intel_pstate_update_util_hwp;
 			goto hwp_cpu_matched;
 		}
 	} else {
-- 
2.7.4

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web