Minki/linux
1
0
Fork 1
mirror of https://github.com/torvalds/linux.git synced 2024-11-22 12:11:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

cpufreq: intel_pstate: Do not use PID-based P-state selection

Browse Source 
All systems with a defined ACPI preferred profile that are not
"servers" have been using the load-based P-state selection algorithm
in intel_pstate since 4.12-rc1 (mobile systems and laptops have been
using it since 4.10-rc1) and no problems with it have been reported
to date.  In particular, no regressions with respect to the PID-based
P-state selection have been reported.  Also testing indicates that
the P-state selection algorithm based on CPU load is generally on par
with the PID-based algorithm performance-wise, and for some workloads
it turns out to be better than the other one, while being more
straightforward and easier to understand at the same time.

Moreover, the PID-based P-state selection algorithm in intel_pstate
is known to be unstable in some situation and generally problematic,
the issues with it are hard to address and it has become a
significant maintenance burden.

For these reasons, make intel_pstate use the "powersave" P-state
selection algorithm based on CPU load in the active mode on all
systems and drop the PID-based P-state selection code along with
all things related to it from the driver.  Also update the
documentation accordingly.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This commit is contained in:
Rafael J. Wysocki 2017-07-25 00:10:59 +02:00
parent 520eccdfe1
commit 9d0ef7af1f
2 changed files with 9 additions and 326 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
Documentation/admin-guide/pm/intel_pstate.rst
View File

@ -167,35 +167,17 @@ is set.
``powersave``
.............
Without HWP, this P-state selection algorithm generally depends on the
processor model and/or the system profile setting in the ACPI tables and there
are two variants of it.
One of them is used with processors from the Atom line and (regardless of the
processor model) on platforms with the system profile in the ACPI tables set to
"mobile" (laptops mostly), "tablet", "appliance PC", "desktop", or
"workstation". It is also used with processors supporting the HWP feature if
that feature has not been enabled (that is, with the ``intel_pstate=no_hwp``
argument in the kernel command line). It is similar to the algorithm
Without HWP, this P-state selection algorithm is similar to the algorithm
implemented by the generic ``schedutil`` scaling governor except that the
utilization metric used by it is based on numbers coming from feedback
registers of the CPU. It generally selects P-states proportional to the
current CPU utilization, so it is referred to as the "proportional" algorithm.
current CPU utilization.
The second variant of the ``powersave`` P-state selection algorithm, used in all
of the other cases (generally, on processors from the Core line, so it is
referred to as the "Core" algorithm), is based on the values read from the APERF
and MPERF feedback registers and the previously requested target P-state.
It does not really take CPU utilization into account explicitly, but as a rule
it causes the CPU P-state to ramp up very quickly in response to increased
utilization which is generally desirable in server environments.
Regardless of the variant, this algorithm is run by the driver's utilization
update callback for the given CPU when it is invoked by the CPU scheduler, but
not more often than every 10 ms (that can be tweaked via ``debugfs`` in `this
particular case <Tuning Interface in debugfs_>`_). Like in the ``performance``
case, the hardware configuration is not touched if the new P-state turns out to
be the same as the current one.
This algorithm is run by the driver's utilization update callback for the
given CPU when it is invoked by the CPU scheduler, but not more often than
every 10 ms. Like in the ``performance`` case, the hardware configuration
is not touched if the new P-state turns out to be the same as the current
one.
This is the default P-state selection algorithm if the
:c:macro:`CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE` kernel configuration option
@ -720,34 +702,7 @@ P-state is called, the ``ftrace`` filter can be set to to
gnome-shell-3409 [001] ..s. 2537.650850: intel_pstate_set_pstate <-intel_pstate_timer_func
<idle>-0 [000] ..s. 2537.654843: intel_pstate_set_pstate <-intel_pstate_timer_func
Tuning Interface in ``debugfs``
-------------------------------
The ``powersave`` algorithm provided by ``intel_pstate`` for `the Core line of
processors in the active mode <powersave_>`_ is based on a `PID controller`_
whose parameters were chosen to address a number of different use cases at the
same time. However, it still is possible to fine-tune it to a specific workload
and the ``debugfs`` interface under ``/sys/kernel/debug/pstate_snb/`` is
provided for this purpose. [Note that the ``pstate_snb`` directory will be
present only if the specific P-state selection algorithm matching the interface
in it actually is in use.]
The following files present in that directory can be used to modify the PID
controller parameters at run time:
| ``deadband``
| ``d_gain_pct``
| ``i_gain_pct``
| ``p_gain_pct``
| ``sample_rate_ms``
| ``setpoint``
Note, however, that achieving desirable results this way generally requires
expert-level understanding of the power vs performance tradeoff, so extra care
is recommended when attempting to do that.
.. _LCEU2015: http://events.linuxfoundation.org/sites/events/files/slides/LinuxConEurope_2015.pdf
.. _SDM: http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-software-developer-system-programming-manual-325384.html
.. _ACPI specification: http://www.uefi.org/sites/default/files/resources/ACPI_6_1.pdf
.. _PID controller: https://en.wikipedia.org/wiki/PID_controller

276
drivers/cpufreq/intel_pstate.c
View File

@ -172,28 +172,6 @@ struct vid_data {
int32_t ratio;
};
/**
* struct _pid - Stores PID data
* @setpoint: Target set point for busyness or performance
* @integral: Storage for accumulated error values
* @p_gain: PID proportional gain
* @i_gain: PID integral gain
* @d_gain: PID derivative gain
* @deadband: PID deadband
* @last_err: Last error storage for integral part of PID calculation
*
* Stores PID coefficients and last error for PID controller.
*/
struct _pid {
int setpoint;
int32_t integral;
int32_t p_gain;
int32_t i_gain;
int32_t d_gain;
int deadband;
int32_t last_err;
};
/**
* struct global_params - Global parameters, mostly tunable via sysfs.
* @no_turbo: Whether or not to use turbo P-states.
@ -223,7 +201,6 @@ struct global_params {
* @last_update: Time of the last update.
* @pstate: Stores P state limits for this CPU
* @vid: Stores VID limits for this CPU
* @pid: Stores PID parameters for this CPU
* @last_sample_time: Last Sample time
* @aperf_mperf_shift: Number of clock cycles after aperf, merf is incremented
* This shift is a multiplier to mperf delta to
@ -258,7 +235,6 @@ struct cpudata {
struct pstate_data pstate;
struct vid_data vid;
struct _pid pid;
u64 last_update;
u64 last_sample_time;
@ -283,28 +259,6 @@ struct cpudata {
static struct cpudata **all_cpu_data;
/**
* struct pstate_adjust_policy - Stores static PID configuration data
* @sample_rate_ms: PID calculation sample rate in ms
* @sample_rate_ns: Sample rate calculation in ns
* @deadband: PID deadband
* @setpoint: PID Setpoint
* @p_gain_pct: PID proportional gain
* @i_gain_pct: PID integral gain
* @d_gain_pct: PID derivative gain
*
* Stores per CPU model static PID configuration data.
*/
struct pstate_adjust_policy {
int sample_rate_ms;
s64 sample_rate_ns;
int deadband;
int setpoint;
int p_gain_pct;
int d_gain_pct;
int i_gain_pct;
};
/**
* struct pstate_funcs - Per CPU model specific callbacks
* @get_max: Callback to get maximum non turbo effective P state
@ -333,15 +287,6 @@ struct pstate_funcs {
};
static struct pstate_funcs pstate_funcs __read_mostly;
static struct pstate_adjust_policy pid_params __read_mostly = {
.sample_rate_ms = 10,
.sample_rate_ns = 10 * NSEC_PER_MSEC,
.deadband = 0,
.setpoint = 97,
.p_gain_pct = 20,
.d_gain_pct = 0,
.i_gain_pct = 0,
};
static int hwp_active __read_mostly;
static bool per_cpu_limits __read_mostly;
@ -509,56 +454,6 @@ static inline void intel_pstate_exit_perf_limits(struct cpufreq_policy *policy)
}
#endif
static signed int pid_calc(struct _pid *pid, int32_t busy)
{
signed int result;
int32_t pterm, dterm, fp_error;
int32_t integral_limit;
fp_error = pid->setpoint - busy;
if (abs(fp_error) <= pid->deadband)
return 0;
pterm = mul_fp(pid->p_gain, fp_error);
pid->integral += fp_error;
/*
* We limit the integral here so that it will never
* get higher than 30. This prevents it from becoming
* too large an input over long periods of time and allows
* it to get factored out sooner.
*
* The value of 30 was chosen through experimentation.
*/
integral_limit = int_tofp(30);
if (pid->integral > integral_limit)
pid->integral = integral_limit;
if (pid->integral < -integral_limit)
pid->integral = -integral_limit;
dterm = mul_fp(pid->d_gain, fp_error - pid->last_err);
pid->last_err = fp_error;
result = pterm + mul_fp(pid->integral, pid->i_gain) + dterm;
result = result + (1 << (FRAC_BITS-1));
return (signed int)fp_toint(result);
}
static inline void intel_pstate_pid_reset(struct cpudata *cpu)
{
struct _pid *pid = &cpu->pid;
pid->p_gain = percent_fp(pid_params.p_gain_pct);
pid->d_gain = percent_fp(pid_params.d_gain_pct);
pid->i_gain = percent_fp(pid_params.i_gain_pct);
pid->setpoint = int_tofp(pid_params.setpoint);
pid->last_err = pid->setpoint - int_tofp(100);
pid->deadband = int_tofp(pid_params.deadband);
pid->integral = 0;
}
static inline void update_turbo_state(void)
{
u64 misc_en;
@ -911,82 +806,6 @@ static void intel_pstate_update_policies(void)
cpufreq_update_policy(cpu);
}
/************************** debugfs begin ************************/
static int pid_param_set(void *data, u64 val)
{
unsigned int cpu;
*(u32 *)data = val;
pid_params.sample_rate_ns = pid_params.sample_rate_ms * NSEC_PER_MSEC;
for_each_possible_cpu(cpu)
if (all_cpu_data[cpu])
intel_pstate_pid_reset(all_cpu_data[cpu]);
return 0;
}
static int pid_param_get(void *data, u64 *val)
{
*val = *(u32 *)data;
return 0;
}
DEFINE_SIMPLE_ATTRIBUTE(fops_pid_param, pid_param_get, pid_param_set, "%llu\n");
static struct dentry *debugfs_parent;
struct pid_param {
char *name;
void *value;
struct dentry *dentry;
};
static struct pid_param pid_files[] = {
{"sample_rate_ms", &pid_params.sample_rate_ms, },
{"d_gain_pct", &pid_params.d_gain_pct, },
{"i_gain_pct", &pid_params.i_gain_pct, },
{"deadband", &pid_params.deadband, },
{"setpoint", &pid_params.setpoint, },
{"p_gain_pct", &pid_params.p_gain_pct, },
{NULL, NULL, }
};
static void intel_pstate_debug_expose_params(void)
{
int i;
debugfs_parent = debugfs_create_dir("pstate_snb", NULL);
if (IS_ERR_OR_NULL(debugfs_parent))
return;
for (i = 0; pid_files[i].name; i++) {
struct dentry *dentry;
dentry = debugfs_create_file(pid_files[i].name, 0660,
debugfs_parent, pid_files[i].value,
&fops_pid_param);
if (!IS_ERR(dentry))
pid_files[i].dentry = dentry;
}
}
static void intel_pstate_debug_hide_params(void)
{
int i;
if (IS_ERR_OR_NULL(debugfs_parent))
return;
for (i = 0; pid_files[i].name; i++) {
debugfs_remove(pid_files[i].dentry);
pid_files[i].dentry = NULL;
}
debugfs_remove(debugfs_parent);
debugfs_parent = NULL;
}
/************************** debugfs end ************************/
/************************** sysfs begin ************************/
#define show_one(file_name, object) \
static ssize_t show_##file_name \
@ -1661,44 +1480,6 @@ static inline int32_t get_target_pstate_use_cpu_load(struct cpudata *cpu)
return target;
}
static inline int32_t get_target_pstate_use_performance(struct cpudata *cpu)
{
int32_t perf_scaled, max_pstate, current_pstate, sample_ratio;
u64 duration_ns;
/*
* perf_scaled is the ratio of the average P-state during the last
* sampling period to the P-state requested last time (in percent).
*
* That measures the system's response to the previous P-state
* selection.
*/
max_pstate = cpu->pstate.max_pstate_physical;
current_pstate = cpu->pstate.current_pstate;
perf_scaled = mul_ext_fp(cpu->sample.core_avg_perf,
div_fp(100 * max_pstate, current_pstate));
/*
* Since our utilization update callback will not run unless we are
* in C0, check if the actual elapsed time is significantly greater (3x)
* than our sample interval. If it is, then we were idle for a long
* enough period of time to adjust our performance metric.
*/
duration_ns = cpu->sample.time - cpu->last_sample_time;
if ((s64)duration_ns > pid_params.sample_rate_ns * 3) {
sample_ratio = div_fp(pid_params.sample_rate_ns, duration_ns);
perf_scaled = mul_fp(perf_scaled, sample_ratio);
} else {
sample_ratio = div_fp(100 * (cpu->sample.mperf << cpu->aperf_mperf_shift),
cpu->sample.tsc);
if (sample_ratio < int_tofp(1))
perf_scaled = 0;
}
cpu->sample.busy_scaled = perf_scaled;
return cpu->pstate.current_pstate - pid_calc(&cpu->pid, perf_scaled);
}
static int intel_pstate_prepare_request(struct cpudata *cpu, int pstate)
{
int max_pstate = intel_pstate_get_base_pstate(cpu);
@ -1741,23 +1522,6 @@ static void intel_pstate_adjust_pstate(struct cpudata *cpu, int target_pstate)
fp_toint(cpu->iowait_boost * 100));
}
static void intel_pstate_update_util_pid(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 < pid_params.sample_rate_ns)
return;
if (intel_pstate_sample(cpu, time)) {
int target_pstate;
target_pstate = get_target_pstate_use_performance(cpu);
intel_pstate_adjust_pstate(cpu, target_pstate);
}
}
static void intel_pstate_update_util(struct update_util_data *data, u64 time,
unsigned int flags)
{
@ -1792,7 +1556,7 @@ static struct pstate_funcs core_funcs = {
.get_turbo = core_get_turbo_pstate,
.get_scaling = core_get_scaling,
.get_val = core_get_val,
.update_util = intel_pstate_update_util_pid,
.update_util = intel_pstate_update_util,
};
static const struct pstate_funcs silvermont_funcs = {
@ -1825,7 +1589,7 @@ static const struct pstate_funcs knl_funcs = {
.get_aperf_mperf_shift = knl_get_aperf_mperf_shift,
.get_scaling = core_get_scaling,
.get_val = core_get_val,
.update_util = intel_pstate_update_util_pid,
.update_util = intel_pstate_update_util,
};
static const struct pstate_funcs bxt_funcs = {
@ -1879,8 +1643,6 @@ static const struct x86_cpu_id intel_pstate_cpu_ee_disable_ids[] = {
{}
};
static bool pid_in_use(void);
static int intel_pstate_init_cpu(unsigned int cpunum)
{
struct cpudata *cpu;
@ -1911,8 +1673,6 @@ static int intel_pstate_init_cpu(unsigned int cpunum)
intel_pstate_disable_ee(cpunum);
intel_pstate_hwp_enable(cpu);
} else if (pid_in_use()) {
intel_pstate_pid_reset(cpu);
}
intel_pstate_get_cpu_pstates(cpu);
@ -2270,12 +2030,6 @@ static struct cpufreq_driver intel_cpufreq = {
static struct cpufreq_driver *default_driver = &intel_pstate;
static bool pid_in_use(void)
{
return intel_pstate_driver == &intel_pstate &&
pstate_funcs.update_util == intel_pstate_update_util_pid;
}
static void intel_pstate_driver_cleanup(void)
{
unsigned int cpu;
@ -2310,9 +2064,6 @@ static int intel_pstate_register_driver(struct cpufreq_driver *driver)
global.min_perf_pct = min_perf_pct_min();
if (pid_in_use())
intel_pstate_debug_expose_params();
return 0;
}
@ -2321,9 +2072,6 @@ static int intel_pstate_unregister_driver(void)
if (hwp_active)
return -EBUSY;
if (pid_in_use())
intel_pstate_debug_hide_params();
cpufreq_unregister_driver(intel_pstate_driver);
intel_pstate_driver_cleanup();
@ -2391,24 +2139,6 @@ static int __init intel_pstate_msrs_not_valid(void)
return 0;
}
#ifdef CONFIG_ACPI
static void intel_pstate_use_acpi_profile(void)
{
switch (acpi_gbl_FADT.preferred_profile) {
case PM_MOBILE:
case PM_TABLET:
case PM_APPLIANCE_PC:
case PM_DESKTOP:
case PM_WORKSTATION:
pstate_funcs.update_util = intel_pstate_update_util;
}
}
#else
static void intel_pstate_use_acpi_profile(void)
{
}
#endif
static void __init copy_cpu_funcs(struct pstate_funcs *funcs)
{
pstate_funcs.get_max = funcs->get_max;
@ -2420,8 +2150,6 @@ static void __init copy_cpu_funcs(struct pstate_funcs *funcs)
pstate_funcs.get_vid = funcs->get_vid;
pstate_funcs.update_util = funcs->update_util;
pstate_funcs.get_aperf_mperf_shift = funcs->get_aperf_mperf_shift;
intel_pstate_use_acpi_profile();
}
#ifdef CONFIG_ACPI