summaryrefslogtreecommitdiffstats
path: root/drivers/cpufreq/intel_pstate.c
diff options
context:
space:
mode:
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>2017-03-14 16:18:34 +0100
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>2017-03-15 16:52:29 +0100
commite4c204ced0ac25e02e58679f07096c5bac0b0d96 (patch)
tree55363eb8c97522ad447b9c7e3126f71509582780 /drivers/cpufreq/intel_pstate.c
parent3f8ed54aee491bbb83656592c2d0ad7b78d045ca (diff)
downloadblackbird-op-linux-e4c204ced0ac25e02e58679f07096c5bac0b0d96.tar.gz
blackbird-op-linux-e4c204ced0ac25e02e58679f07096c5bac0b0d96.zip
cpufreq: intel_pstate: Avoid percentages in limits-related computations
Currently, intel_pstate_update_perf_limits() first converts the policy minimum and maximum limits into percentages of the maximum turbo frequency (rounding up to an integer) and then converts these percentages to fractions (by using fixed-point arithmetic to divide them by 100). That introduces a rounding error unnecessarily, because the fractions can be obtained by carrying out fixed-point divisions directly on the input numbers. Rework the computations in intel_pstate_hwp_set() to use fractions instead of percentages (and drop redundant local variables from there) and modify intel_pstate_update_perf_limits() to compute the fractions directly and percentages out of them. While at it, introduce percent_ext_fp() for converting percentages to fractions (with extended number of fraction bits) and use it in the computations. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Diffstat (limited to 'drivers/cpufreq/intel_pstate.c')
-rw-r--r--drivers/cpufreq/intel_pstate.c56
1 files changed, 28 insertions, 28 deletions
diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
index ee12641ee010..08e134ffba68 100644
--- a/drivers/cpufreq/intel_pstate.c
+++ b/drivers/cpufreq/intel_pstate.c
@@ -84,6 +84,11 @@ static inline u64 div_ext_fp(u64 x, u64 y)
return div64_u64(x << EXT_FRAC_BITS, y);
}
+static inline int32_t percent_ext_fp(int percent)
+{
+ return div_ext_fp(percent, 100);
+}
+
/**
* struct sample - Store performance sample
* @core_avg_perf: Ratio of APERF/MPERF which is the actual average
@@ -850,7 +855,6 @@ static void intel_pstate_hwp_set(struct cpufreq_policy *policy)
u64 value, cap;
for_each_cpu(cpu, policy->cpus) {
- int max_perf_pct, min_perf_pct;
struct cpudata *cpu_data = all_cpu_data[cpu];
s16 epp;
@@ -864,16 +868,14 @@ static void intel_pstate_hwp_set(struct cpufreq_policy *policy)
else
hw_max = HWP_HIGHEST_PERF(cap);
- max_perf_pct = perf_limits->max_perf_pct;
- min_perf_pct = perf_limits->min_perf_pct;
- min = hw_max * min_perf_pct / 100;
+ min = fp_ext_toint(hw_max * perf_limits->min_perf);
rdmsrl_on_cpu(cpu, MSR_HWP_REQUEST, &value);
value &= ~HWP_MIN_PERF(~0L);
value |= HWP_MIN_PERF(min);
- max = hw_max * max_perf_pct / 100;
+ max = fp_ext_toint(hw_max * perf_limits->max_perf);
value &= ~HWP_MAX_PERF(~0L);
value |= HWP_MAX_PERF(max);
@@ -1223,7 +1225,7 @@ static ssize_t store_max_perf_pct(struct kobject *a, struct attribute *b,
limits->max_perf_pct);
limits->max_perf_pct = max(limits->min_perf_pct,
limits->max_perf_pct);
- limits->max_perf = div_ext_fp(limits->max_perf_pct, 100);
+ limits->max_perf = percent_ext_fp(limits->max_perf_pct);
intel_pstate_update_policies();
@@ -1260,7 +1262,7 @@ static ssize_t store_min_perf_pct(struct kobject *a, struct attribute *b,
limits->min_perf_pct);
limits->min_perf_pct = min(limits->max_perf_pct,
limits->min_perf_pct);
- limits->min_perf = div_ext_fp(limits->min_perf_pct, 100);
+ limits->min_perf = percent_ext_fp(limits->min_perf_pct);
intel_pstate_update_policies();
@@ -2078,36 +2080,34 @@ static void intel_pstate_clear_update_util_hook(unsigned int cpu)
static void intel_pstate_update_perf_limits(struct cpufreq_policy *policy,
struct perf_limits *limits)
{
+ int32_t max_policy_perf, min_policy_perf;
- limits->max_policy_pct = DIV_ROUND_UP(policy->max * 100,
- policy->cpuinfo.max_freq);
- limits->max_policy_pct = clamp_t(int, limits->max_policy_pct, 0, 100);
+ max_policy_perf = div_ext_fp(policy->max, policy->cpuinfo.max_freq);
+ max_policy_perf = clamp_t(int32_t, max_policy_perf, 0, int_ext_tofp(1));
if (policy->max == policy->min) {
- limits->min_policy_pct = limits->max_policy_pct;
+ min_policy_perf = max_policy_perf;
} else {
- limits->min_policy_pct = DIV_ROUND_UP(policy->min * 100,
- policy->cpuinfo.max_freq);
- limits->min_policy_pct = clamp_t(int, limits->min_policy_pct,
- 0, 100);
+ min_policy_perf = div_ext_fp(policy->min,
+ policy->cpuinfo.max_freq);
+ min_policy_perf = clamp_t(int32_t, min_policy_perf,
+ 0, max_policy_perf);
}
- /* Normalize user input to [min_policy_pct, max_policy_pct] */
- limits->min_perf_pct = max(limits->min_policy_pct,
- limits->min_sysfs_pct);
- limits->min_perf_pct = min(limits->max_policy_pct,
- limits->min_perf_pct);
- limits->max_perf_pct = min(limits->max_policy_pct,
- limits->max_sysfs_pct);
- limits->max_perf_pct = max(limits->min_policy_pct,
- limits->max_perf_pct);
+ /* Normalize user input to [min_perf, max_perf] */
+ limits->min_perf = max(min_policy_perf,
+ percent_ext_fp(limits->min_sysfs_pct));
+ limits->min_perf = min(limits->min_perf, max_policy_perf);
+ limits->max_perf = min(max_policy_perf,
+ percent_ext_fp(limits->max_sysfs_pct));
+ limits->max_perf = max(min_policy_perf, limits->max_perf);
- /* Make sure min_perf_pct <= max_perf_pct */
- limits->min_perf_pct = min(limits->max_perf_pct, limits->min_perf_pct);
+ /* Make sure min_perf <= max_perf */
+ limits->min_perf = min(limits->min_perf, limits->max_perf);
- limits->min_perf = div_ext_fp(limits->min_perf_pct, 100);
- limits->max_perf = div_ext_fp(limits->max_perf_pct, 100);
limits->max_perf = round_up(limits->max_perf, EXT_FRAC_BITS);
limits->min_perf = round_up(limits->min_perf, EXT_FRAC_BITS);
+ limits->max_perf_pct = fp_ext_toint(limits->max_perf * 100);
+ limits->min_perf_pct = fp_ext_toint(limits->min_perf * 100);
pr_debug("cpu:%d max_perf_pct:%d min_perf_pct:%d\n", policy->cpu,
limits->max_perf_pct, limits->min_perf_pct);
OpenPOWER on IntegriCloud