From: "Rafael J. Wysocki" <rjw@rjwysocki.net>
To: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Cc: Linux PM <linux-pm@vger.kernel.org>, LKML <linux-kernel@vger.kernel.org>
Subject: [PATCH 2/3] cpufreq: intel_pstate: Avoid percentages in limits-related computations
Date: Tue, 14 Mar 2017 16:18:34 +0100 [thread overview]
Message-ID: <3305006.UFgYgyKJxf@aspire.rjw.lan> (raw)
In-Reply-To: <2130954.RIJyMN1BRu@aspire.rjw.lan>
From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
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 into 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>
---
drivers/cpufreq/intel_pstate.c | 60 ++++++++++++++++++++---------------------
1 file changed, 30 insertions(+), 30 deletions(-)
Index: linux-pm/drivers/cpufreq/intel_pstate.c
===================================================================
--- linux-pm.orig/drivers/cpufreq/intel_pstate.c
+++ linux-pm/drivers/cpufreq/intel_pstate.c
@@ -84,6 +84,11 @@ static inline u64 div_ext_fp(u64 x, u64
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
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
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
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
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_ut
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);
- }
-
- /* 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);
+ 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_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);
next prev parent reply other threads:[~2017-03-14 15:26 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-03-14 15:14 [PATCH 0/3] intel_pstate: More fixes related to limits Rafael J. Wysocki
2017-03-14 15:17 ` [PATCH 1/3] cpufreq: intel_pstate: Do not squash policy limits between global ones Rafael J. Wysocki
2017-03-14 15:18 ` Rafael J. Wysocki [this message]
2017-03-14 15:19 ` [PATCH 3/3] cpufreq: intel_pstate: Simplify coordination of P-state limits Rafael J. Wysocki
2017-03-16 12:22 ` [PATCH 0/3] intel_pstate: More fixes related to limits Rafael J. Wysocki
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3305006.UFgYgyKJxf@aspire.rjw.lan \
--to=rjw@rjwysocki.net \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pm@vger.kernel.org \
--cc=srinivas.pandruvada@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.