* [PATCH 001/007] cpufreq_conservative: aligning of codebase with ondemand
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
@ 2006-03-26 9:54 ` Dominik Brodowski
2006-03-26 9:55 ` [PATCH 002/007] cpufreq_conservative: alter default responsiveness Dominik Brodowski
` (6 subsequent siblings)
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:54 UTC (permalink / raw)
To: davej, cpufreq
From: Alexander Clouter <alex@digriz.org.uk>
Since the conservative govenor was released its codebase has drifted from the
the direction and updates that have been applied to the ondemand govornor.
This patch addresses the lack of updates in that period and brings
conservative back up to date. The resulting diff file between
cpufreq_ondemand.c and cpufreq_conservative.c is now much smaller and shows
more clearly the differences between the two.
Another reason to do this is ages ago, knowingly, I did a piss poor attempt
at making conservative less responsive by knocking up
DEF_SAMPLING_RATE_LATENCY_MULTIPLIER by two orders of magnitude. I did fix
this ages ago but in my dis-organisation I must have toasted the diff and
left it the way it was. About two weeks ago a user contacted me saying he
was having problems with the conservative governor with his AMD Athlon XP-M
2800+ as /sys/devices/system/cpu/cpu0/cpufreq/conservative showed
sampling_rate_min 9950000
sampling_rate_max 1360065408
Nine seconds to decide about changing the frequency....not too responsive :)
Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_conservative.c | 53 ++++++++++++++++----------------
1 files changed, 27 insertions(+), 26 deletions(-)
2c906b317b2d9c7e32b0d513e102bd68a2c49112
diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
index ac38766..adecd31 100644
--- a/drivers/cpufreq/cpufreq_conservative.c
+++ b/drivers/cpufreq/cpufreq_conservative.c
@@ -35,12 +35,7 @@
*/
#define DEF_FREQUENCY_UP_THRESHOLD (80)
-#define MIN_FREQUENCY_UP_THRESHOLD (0)
-#define MAX_FREQUENCY_UP_THRESHOLD (100)
-
#define DEF_FREQUENCY_DOWN_THRESHOLD (20)
-#define MIN_FREQUENCY_DOWN_THRESHOLD (0)
-#define MAX_FREQUENCY_DOWN_THRESHOLD (100)
/*
* The polling frequency of this governor depends on the capability of
@@ -53,10 +48,14 @@
* All times here are in uS.
*/
static unsigned int def_sampling_rate;
-#define MIN_SAMPLING_RATE (def_sampling_rate / 2)
+#define MIN_SAMPLING_RATE_RATIO (2)
+/* for correct statistics, we need at least 10 ticks between each measure */
+#define MIN_STAT_SAMPLING_RATE (MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10))
+#define MIN_SAMPLING_RATE (def_sampling_rate / MIN_SAMPLING_RATE_RATIO)
#define MAX_SAMPLING_RATE (500 * def_sampling_rate)
-#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (100000)
-#define DEF_SAMPLING_DOWN_FACTOR (5)
+#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000)
+#define DEF_SAMPLING_DOWN_FACTOR (1)
+#define MAX_SAMPLING_DOWN_FACTOR (10)
#define TRANSITION_LATENCY_LIMIT (10 * 1000)
static void do_dbs_timer(void *data);
@@ -136,7 +135,7 @@ static ssize_t store_sampling_down_facto
unsigned int input;
int ret;
ret = sscanf (buf, "%u", &input);
- if (ret != 1 )
+ if (ret != 1 || input > MAX_SAMPLING_DOWN_FACTOR || input < 1)
return -EINVAL;
mutex_lock(&dbs_mutex);
@@ -173,8 +172,7 @@ static ssize_t store_up_threshold(struct
ret = sscanf (buf, "%u", &input);
mutex_lock(&dbs_mutex);
- if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD ||
- input < MIN_FREQUENCY_UP_THRESHOLD ||
+ if (ret != 1 || input > 100 || input < 0 ||
input <= dbs_tuners_ins.down_threshold) {
mutex_unlock(&dbs_mutex);
return -EINVAL;
@@ -194,8 +192,7 @@ static ssize_t store_down_threshold(stru
ret = sscanf (buf, "%u", &input);
mutex_lock(&dbs_mutex);
- if (ret != 1 || input > MAX_FREQUENCY_DOWN_THRESHOLD ||
- input < MIN_FREQUENCY_DOWN_THRESHOLD ||
+ if (ret != 1 || input > 100 || input < 0 ||
input >= dbs_tuners_ins.up_threshold) {
mutex_unlock(&dbs_mutex);
return -EINVAL;
@@ -337,7 +334,6 @@ static void dbs_check_cpu(int cpu)
*/
/* Check for frequency increase */
-
idle_ticks = UINT_MAX;
for_each_cpu_mask(j, policy->cpus) {
unsigned int tmp_idle_ticks, total_idle_ticks;
@@ -357,7 +353,7 @@ static void dbs_check_cpu(int cpu)
/* Scale idle ticks by 100 and compare with up and down ticks */
idle_ticks *= 100;
up_idle_ticks = (100 - dbs_tuners_ins.up_threshold) *
- usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
+ usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
if (idle_ticks < up_idle_ticks) {
down_skip[cpu] = 0;
@@ -398,6 +394,7 @@ static void dbs_check_cpu(int cpu)
struct cpu_dbs_info_s *j_dbs_info;
j_dbs_info = &per_cpu(cpu_dbs_info, j);
+ /* Check for frequency decrease */
total_idle_ticks = j_dbs_info->prev_cpu_idle_up;
tmp_idle_ticks = total_idle_ticks -
j_dbs_info->prev_cpu_idle_down;
@@ -414,12 +411,14 @@ static void dbs_check_cpu(int cpu)
freq_down_sampling_rate = dbs_tuners_ins.sampling_rate *
dbs_tuners_ins.sampling_down_factor;
down_idle_ticks = (100 - dbs_tuners_ins.down_threshold) *
- usecs_to_jiffies(freq_down_sampling_rate);
+ usecs_to_jiffies(freq_down_sampling_rate);
if (idle_ticks > down_idle_ticks) {
- /* if we are already at the lowest speed then break out early
+ /*
+ * if we are already at the lowest speed then break out early
* or if we 'cannot' reduce the speed as the user might want
- * freq_step to be zero */
+ * freq_step to be zero
+ */
if (requested_freq[cpu] == policy->min
|| dbs_tuners_ins.freq_step == 0)
return;
@@ -434,9 +433,8 @@ static void dbs_check_cpu(int cpu)
if (requested_freq[cpu] < policy->min)
requested_freq[cpu] = policy->min;
- __cpufreq_driver_target(policy,
- requested_freq[cpu],
- CPUFREQ_RELATION_H);
+ __cpufreq_driver_target(policy, requested_freq[cpu],
+ CPUFREQ_RELATION_H);
return;
}
}
@@ -507,13 +505,16 @@ static int cpufreq_governor_dbs(struct c
if (dbs_enable == 1) {
unsigned int latency;
/* policy latency is in nS. Convert it to uS first */
+ latency = policy->cpuinfo.transition_latency / 1000;
+ if (latency == 0)
+ latency = 1;
- latency = policy->cpuinfo.transition_latency;
- if (latency < 1000)
- latency = 1000;
-
- def_sampling_rate = (latency / 1000) *
+ def_sampling_rate = latency *
DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
+
+ if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
+ def_sampling_rate = MIN_STAT_SAMPLING_RATE;
+
dbs_tuners_ins.sampling_rate = def_sampling_rate;
dbs_tuners_ins.ignore_nice = 0;
dbs_tuners_ins.freq_step = 5;
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 002/007] cpufreq_conservative: alter default responsiveness
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
2006-03-26 9:54 ` [PATCH 001/007] cpufreq_conservative: aligning of codebase with ondemand Dominik Brodowski
@ 2006-03-26 9:55 ` Dominik Brodowski
2006-03-26 9:55 ` [PATCH 003/007] cpufreq_conservative: make for_each_cpu() safe Dominik Brodowski
` (5 subsequent siblings)
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:55 UTC (permalink / raw)
To: davej, cpufreq
From: Alexander Clouter <alex@digriz.org.uk>
The sensible approach to making conservative less responsive than ondemand :)
As mentioned in patch [1/4]. We do not want conservative to shoot through
all the frequencies, its point (by default) is to slowly move through them.
By default its ten times less responsive.
Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_conservative.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
e8a02572252f9115c2b8296c40fd8b985f06f872
diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
index adecd31..3ca3cf0 100644
--- a/drivers/cpufreq/cpufreq_conservative.c
+++ b/drivers/cpufreq/cpufreq_conservative.c
@@ -509,7 +509,7 @@ static int cpufreq_governor_dbs(struct c
if (latency == 0)
latency = 1;
- def_sampling_rate = latency *
+ def_sampling_rate = 10 * latency *
DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 003/007] cpufreq_conservative: make for_each_cpu() safe
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
2006-03-26 9:54 ` [PATCH 001/007] cpufreq_conservative: aligning of codebase with ondemand Dominik Brodowski
2006-03-26 9:55 ` [PATCH 002/007] cpufreq_conservative: alter default responsiveness Dominik Brodowski
@ 2006-03-26 9:55 ` Dominik Brodowski
2006-03-26 9:56 ` [PATCH 004/007] cpufreq_conservative: alternative initialise approach Dominik Brodowski
` (4 subsequent siblings)
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:55 UTC (permalink / raw)
To: davej, cpufreq
From: Alexander Clouter <alex@digriz.org.uk>
All these changes should make cpufreq_conservative safe in regards to the x86
for_each_cpu cpumask.h changes and whatnot.
Whilst making it safe a number of pointless for loops related to the cpu
mask's were removed. I was never comfortable with all those for loops,
especially as the iteration is over the same data again and again for each
CPU you had in a single poll, an O(n^2) outcome to frequency scaling.
The approach I use is to assume by default no CPU's exist and it sets the
requested_freq to zero as a kind of flag, the reasoning is in the source ;)
If the CPU is queried and requested_freq is zero then it initialises the
variable to current_freq and then continues as if nothing happened which
should be the same net effect as before?
Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_conservative.c | 91 +++++++++++++++-----------------
1 files changed, 42 insertions(+), 49 deletions(-)
08a28e2e98aa821cf6f15f8a267beb2f33377bb9
diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
index 3ca3cf0..7498f25 100644
--- a/drivers/cpufreq/cpufreq_conservative.c
+++ b/drivers/cpufreq/cpufreq_conservative.c
@@ -294,31 +294,40 @@ static struct attribute_group dbs_attr_g
static void dbs_check_cpu(int cpu)
{
unsigned int idle_ticks, up_idle_ticks, down_idle_ticks;
+ unsigned int tmp_idle_ticks, total_idle_ticks;
unsigned int freq_step;
unsigned int freq_down_sampling_rate;
- static int down_skip[NR_CPUS];
- static int requested_freq[NR_CPUS];
- static unsigned short init_flag = 0;
- struct cpu_dbs_info_s *this_dbs_info;
- struct cpu_dbs_info_s *dbs_info;
-
+ static unsigned short down_skip[NR_CPUS];
+ static unsigned int requested_freq[NR_CPUS];
+ static unsigned int init_flag = NR_CPUS;
+ struct cpu_dbs_info_s *this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
struct cpufreq_policy *policy;
- unsigned int j;
- this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
if (!this_dbs_info->enable)
return;
- policy = this_dbs_info->cur_policy;
-
- if ( init_flag == 0 ) {
- for_each_online_cpu(j) {
- dbs_info = &per_cpu(cpu_dbs_info, j);
- requested_freq[j] = dbs_info->cur_policy->cur;
+ if ( init_flag != 0 ) {
+ for_each_cpu(init_flag) {
+ down_skip[init_flag] = 0;
+ /* I doubt a CPU exists with a freq of 0hz :) */
+ requested_freq[init_flag] = 0;
}
- init_flag = 1;
+ init_flag = 0;
}
+ /*
+ * If its a freshly initialised cpu we setup requested_freq. This
+ * check could be avoided if we did not care about a first time
+ * stunted increase in CPU speed when there is a load. I feel we
+ * should be initialising this to something. The removal of a CPU
+ * is not a problem, after a short time the CPU should settle down
+ * to a 'natural' frequency.
+ */
+ if (requested_freq[cpu] == 0)
+ requested_freq[cpu] = this_dbs_info->cur_policy->cur;
+
+ policy = this_dbs_info->cur_policy;
+
/*
* The default safe range is 20% to 80%
* Every sampling_rate, we check
@@ -335,20 +344,15 @@ static void dbs_check_cpu(int cpu)
/* Check for frequency increase */
idle_ticks = UINT_MAX;
- for_each_cpu_mask(j, policy->cpus) {
- unsigned int tmp_idle_ticks, total_idle_ticks;
- struct cpu_dbs_info_s *j_dbs_info;
-
- j_dbs_info = &per_cpu(cpu_dbs_info, j);
- /* Check for frequency increase */
- total_idle_ticks = get_cpu_idle_time(j);
- tmp_idle_ticks = total_idle_ticks -
- j_dbs_info->prev_cpu_idle_up;
- j_dbs_info->prev_cpu_idle_up = total_idle_ticks;
- if (tmp_idle_ticks < idle_ticks)
- idle_ticks = tmp_idle_ticks;
- }
+ /* Check for frequency increase */
+ total_idle_ticks = get_cpu_idle_time(cpu);
+ tmp_idle_ticks = total_idle_ticks -
+ this_dbs_info->prev_cpu_idle_up;
+ this_dbs_info->prev_cpu_idle_up = total_idle_ticks;
+
+ if (tmp_idle_ticks < idle_ticks)
+ idle_ticks = tmp_idle_ticks;
/* Scale idle ticks by 100 and compare with up and down ticks */
idle_ticks *= 100;
@@ -357,13 +361,9 @@ static void dbs_check_cpu(int cpu)
if (idle_ticks < up_idle_ticks) {
down_skip[cpu] = 0;
- for_each_cpu_mask(j, policy->cpus) {
- struct cpu_dbs_info_s *j_dbs_info;
+ this_dbs_info->prev_cpu_idle_down =
+ this_dbs_info->prev_cpu_idle_up;
- j_dbs_info = &per_cpu(cpu_dbs_info, j);
- j_dbs_info->prev_cpu_idle_down =
- j_dbs_info->prev_cpu_idle_up;
- }
/* if we are already at full speed then break out early */
if (requested_freq[cpu] == policy->max)
return;
@@ -388,21 +388,14 @@ static void dbs_check_cpu(int cpu)
if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor)
return;
- idle_ticks = UINT_MAX;
- for_each_cpu_mask(j, policy->cpus) {
- unsigned int tmp_idle_ticks, total_idle_ticks;
- struct cpu_dbs_info_s *j_dbs_info;
-
- j_dbs_info = &per_cpu(cpu_dbs_info, j);
- /* Check for frequency decrease */
- total_idle_ticks = j_dbs_info->prev_cpu_idle_up;
- tmp_idle_ticks = total_idle_ticks -
- j_dbs_info->prev_cpu_idle_down;
- j_dbs_info->prev_cpu_idle_down = total_idle_ticks;
+ /* Check for frequency decrease */
+ total_idle_ticks = this_dbs_info->prev_cpu_idle_up;
+ tmp_idle_ticks = total_idle_ticks -
+ this_dbs_info->prev_cpu_idle_down;
+ this_dbs_info->prev_cpu_idle_down = total_idle_ticks;
- if (tmp_idle_ticks < idle_ticks)
- idle_ticks = tmp_idle_ticks;
- }
+ if (tmp_idle_ticks < idle_ticks)
+ idle_ticks = tmp_idle_ticks;
/* Scale idle ticks by 100 and compare with up and down ticks */
idle_ticks *= 100;
@@ -491,7 +484,7 @@ static int cpufreq_governor_dbs(struct c
j_dbs_info = &per_cpu(cpu_dbs_info, j);
j_dbs_info->cur_policy = policy;
- j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(j);
+ j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(cpu);
j_dbs_info->prev_cpu_idle_down
= j_dbs_info->prev_cpu_idle_up;
}
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 004/007] cpufreq_conservative: alternative initialise approach
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
` (2 preceding siblings ...)
2006-03-26 9:55 ` [PATCH 003/007] cpufreq_conservative: make for_each_cpu() safe Dominik Brodowski
@ 2006-03-26 9:56 ` Dominik Brodowski
2006-03-26 9:57 ` [PATCH 005/007] cpufreq_ondemand: Warn if it cannot run due to too long transition latency Dominik Brodowski
` (3 subsequent siblings)
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:56 UTC (permalink / raw)
To: davej, cpufreq
From: Alexander Clouter <alex@digriz.org.uk>
Venki, author of cpufreq_ondemand, came up with a neater way to remove the
initialiser code from the main loop of my code and out to the point when the
governor is actually initialised.
Not only does it look but it also feels cleaner, plus its simpler to
understand. It also saves a bunch of pointless conditional statements in the
main loop.
Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_conservative.c | 55 ++++++++++----------------------
1 files changed, 18 insertions(+), 37 deletions(-)
a159b82770ab84e1b5e0306fa65e158188492b16
diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
index 7498f25..a152d2c 100644
--- a/drivers/cpufreq/cpufreq_conservative.c
+++ b/drivers/cpufreq/cpufreq_conservative.c
@@ -65,6 +65,8 @@ struct cpu_dbs_info_s {
unsigned int prev_cpu_idle_up;
unsigned int prev_cpu_idle_down;
unsigned int enable;
+ unsigned int down_skip;
+ unsigned int requested_freq;
};
static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info);
@@ -297,35 +299,12 @@ static void dbs_check_cpu(int cpu)
unsigned int tmp_idle_ticks, total_idle_ticks;
unsigned int freq_step;
unsigned int freq_down_sampling_rate;
- static unsigned short down_skip[NR_CPUS];
- static unsigned int requested_freq[NR_CPUS];
- static unsigned int init_flag = NR_CPUS;
struct cpu_dbs_info_s *this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
struct cpufreq_policy *policy;
if (!this_dbs_info->enable)
return;
- if ( init_flag != 0 ) {
- for_each_cpu(init_flag) {
- down_skip[init_flag] = 0;
- /* I doubt a CPU exists with a freq of 0hz :) */
- requested_freq[init_flag] = 0;
- }
- init_flag = 0;
- }
-
- /*
- * If its a freshly initialised cpu we setup requested_freq. This
- * check could be avoided if we did not care about a first time
- * stunted increase in CPU speed when there is a load. I feel we
- * should be initialising this to something. The removal of a CPU
- * is not a problem, after a short time the CPU should settle down
- * to a 'natural' frequency.
- */
- if (requested_freq[cpu] == 0)
- requested_freq[cpu] = this_dbs_info->cur_policy->cur;
-
policy = this_dbs_info->cur_policy;
/*
@@ -360,12 +339,12 @@ static void dbs_check_cpu(int cpu)
usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
if (idle_ticks < up_idle_ticks) {
- down_skip[cpu] = 0;
+ this_dbs_info->down_skip = 0;
this_dbs_info->prev_cpu_idle_down =
this_dbs_info->prev_cpu_idle_up;
/* if we are already at full speed then break out early */
- if (requested_freq[cpu] == policy->max)
+ if (this_dbs_info->requested_freq == policy->max)
return;
freq_step = (dbs_tuners_ins.freq_step * policy->max) / 100;
@@ -374,18 +353,18 @@ static void dbs_check_cpu(int cpu)
if (unlikely(freq_step == 0))
freq_step = 5;
- requested_freq[cpu] += freq_step;
- if (requested_freq[cpu] > policy->max)
- requested_freq[cpu] = policy->max;
+ this_dbs_info->requested_freq += freq_step;
+ if (this_dbs_info->requested_freq > policy->max)
+ this_dbs_info->requested_freq = policy->max;
- __cpufreq_driver_target(policy, requested_freq[cpu],
+ __cpufreq_driver_target(policy, this_dbs_info->requested_freq,
CPUFREQ_RELATION_H);
return;
}
/* Check for frequency decrease */
- down_skip[cpu]++;
- if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor)
+ this_dbs_info->down_skip++;
+ if (this_dbs_info->down_skip < dbs_tuners_ins.sampling_down_factor)
return;
/* Check for frequency decrease */
@@ -399,7 +378,7 @@ static void dbs_check_cpu(int cpu)
/* Scale idle ticks by 100 and compare with up and down ticks */
idle_ticks *= 100;
- down_skip[cpu] = 0;
+ this_dbs_info->down_skip = 0;
freq_down_sampling_rate = dbs_tuners_ins.sampling_rate *
dbs_tuners_ins.sampling_down_factor;
@@ -412,7 +391,7 @@ static void dbs_check_cpu(int cpu)
* or if we 'cannot' reduce the speed as the user might want
* freq_step to be zero
*/
- if (requested_freq[cpu] == policy->min
+ if (this_dbs_info->requested_freq == policy->min
|| dbs_tuners_ins.freq_step == 0)
return;
@@ -422,11 +401,11 @@ static void dbs_check_cpu(int cpu)
if (unlikely(freq_step == 0))
freq_step = 5;
- requested_freq[cpu] -= freq_step;
- if (requested_freq[cpu] < policy->min)
- requested_freq[cpu] = policy->min;
+ this_dbs_info->requested_freq -= freq_step;
+ if (this_dbs_info->requested_freq < policy->min)
+ this_dbs_info->requested_freq = policy->min;
- __cpufreq_driver_target(policy, requested_freq[cpu],
+ __cpufreq_driver_target(policy, this_dbs_info->requested_freq,
CPUFREQ_RELATION_H);
return;
}
@@ -489,6 +468,8 @@ static int cpufreq_governor_dbs(struct c
= j_dbs_info->prev_cpu_idle_up;
}
this_dbs_info->enable = 1;
+ this_dbs_info->down_skip = 0;
+ this_dbs_info->requested_freq = policy->cur;
sysfs_create_group(&policy->kobj, &dbs_attr_group);
dbs_enable++;
/*
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 005/007] cpufreq_ondemand: Warn if it cannot run due to too long transition latency
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
` (3 preceding siblings ...)
2006-03-26 9:56 ` [PATCH 004/007] cpufreq_conservative: alternative initialise approach Dominik Brodowski
@ 2006-03-26 9:57 ` Dominik Brodowski
2006-03-26 9:57 ` [PATCH 006/007] cpufreq_ondemand: keep ignore_nice_load value when it is reselected Dominik Brodowski
` (2 subsequent siblings)
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:57 UTC (permalink / raw)
To: davej, cpufreq
From: Eric Piel <Eric.Piel@tremplin-utc.net>
Display a warning if the ondemand governor can not be selected due to a
transition latency of the cpufreq driver which is too long.
Signed-off-by: Eric Piel <eric.piel@tremplin-utc.net>
Acked-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_ondemand.c | 5 ++++-
1 files changed, 4 insertions(+), 1 deletions(-)
ff8c288d7d1a368b663058cdee1ea0adcdef2fa2
diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
index 69aa1db..6430489 100644
--- a/drivers/cpufreq/cpufreq_ondemand.c
+++ b/drivers/cpufreq/cpufreq_ondemand.c
@@ -395,8 +395,11 @@ static int cpufreq_governor_dbs(struct c
return -EINVAL;
if (policy->cpuinfo.transition_latency >
- (TRANSITION_LATENCY_LIMIT * 1000))
+ (TRANSITION_LATENCY_LIMIT * 1000)) {
+ printk(KERN_WARNING "ondemand governor failed to load "
+ "due to too long transition latency\n");
return -EINVAL;
+ }
if (this_dbs_info->enable) /* Already enabled */
break;
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 006/007] cpufreq_ondemand: keep ignore_nice_load value when it is reselected
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
` (4 preceding siblings ...)
2006-03-26 9:57 ` [PATCH 005/007] cpufreq_ondemand: Warn if it cannot run due to too long transition latency Dominik Brodowski
@ 2006-03-26 9:57 ` Dominik Brodowski
2006-03-26 9:57 ` [PATCH 007/007] cpufreq_ondemand: add range check Dominik Brodowski
2006-03-27 19:53 ` [git pull] cpufreq governor updates for 2.6.17 Dave Jones
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:57 UTC (permalink / raw)
To: davej, cpufreq
From: Eric Piel <Eric.Piel@tremplin-utc.net>
Keep the value of ignore_nice_load of the ondemand governor even after
the governor has been deselected and selected back. This is the behavior
of the other exported values of the ondemand governor and it's much more
user-friendly.
Signed-off-by: Eric Piel <eric.piel@tremplin-utc.net>
Acked-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_ondemand.c | 3 +--
1 files changed, 1 insertions(+), 2 deletions(-)
9cbad61b41f0b6f0a4c600fe96d8292ffd592b50
diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
index 6430489..cd846f5 100644
--- a/drivers/cpufreq/cpufreq_ondemand.c
+++ b/drivers/cpufreq/cpufreq_ondemand.c
@@ -84,6 +84,7 @@ struct dbs_tuners {
static struct dbs_tuners dbs_tuners_ins = {
.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
.sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR,
+ .ignore_nice = 0,
};
static inline unsigned int get_cpu_idle_time(unsigned int cpu)
@@ -434,8 +435,6 @@ static int cpufreq_governor_dbs(struct c
def_sampling_rate = MIN_STAT_SAMPLING_RATE;
dbs_tuners_ins.sampling_rate = def_sampling_rate;
- dbs_tuners_ins.ignore_nice = 0;
-
dbs_timer_init();
}
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 007/007] cpufreq_ondemand: add range check
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
` (5 preceding siblings ...)
2006-03-26 9:57 ` [PATCH 006/007] cpufreq_ondemand: keep ignore_nice_load value when it is reselected Dominik Brodowski
@ 2006-03-26 9:57 ` Dominik Brodowski
2006-03-27 19:53 ` [git pull] cpufreq governor updates for 2.6.17 Dave Jones
7 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-26 9:57 UTC (permalink / raw)
To: davej, cpufreq
From: Dominik Brodowski <linux@dominikbrodowski.net>
Assert that cpufreq_target is, at least, called with the minimum frequency
allowed by this policy, not something lower. It triggered problems on ARM.
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
drivers/cpufreq/cpufreq_ondemand.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
7c9d8c0e84d395a01289ebd1597758939a875a86
diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
index cd846f5..956d121 100644
--- a/drivers/cpufreq/cpufreq_ondemand.c
+++ b/drivers/cpufreq/cpufreq_ondemand.c
@@ -351,6 +351,9 @@ static void dbs_check_cpu(int cpu)
freq_next = (freq_next * policy->cur) /
(dbs_tuners_ins.up_threshold - 10);
+ if (freq_next < policy->min)
+ freq_next = policy->min;
+
if (freq_next <= ((policy->cur * 95) / 100))
__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
}
--
1.2.4
^ permalink raw reply related [flat|nested] 11+ messages in thread* Re: [git pull] cpufreq governor updates for 2.6.17
2006-03-26 9:53 [git pull] cpufreq governor updates for 2.6.17 Dominik Brodowski
` (6 preceding siblings ...)
2006-03-26 9:57 ` [PATCH 007/007] cpufreq_ondemand: add range check Dominik Brodowski
@ 2006-03-27 19:53 ` Dave Jones
2006-03-27 19:58 ` Dave Jones
7 siblings, 1 reply; 11+ messages in thread
From: Dave Jones @ 2006-03-27 19:53 UTC (permalink / raw)
To: cpufreq
On Sun, Mar 26, 2006 at 11:53:39AM +0200, Dominik Brodowski wrote:
> Hi Dave,
>
> Please pull from
> git://git.kernel.org/pub/scm/linux/kernel/git/brodo/cpufreq-2.6.git/
> for a few cpufreq updates for 2.6.17. The diffstat and list of changes
> follows, the patches will be sent out as individual mails as replies to this
> message.
I'm not entirely sure what happened here.
I git pull'd from this, and ended up with a 59,886 line delta against Linus' tree.
As you sent the diffs too, I'll just apply-mbox them.
Dave
--
http://www.codemonkey.org.uk
^ permalink raw reply [flat|nested] 11+ messages in thread* Re: [git pull] cpufreq governor updates for 2.6.17
2006-03-27 19:53 ` [git pull] cpufreq governor updates for 2.6.17 Dave Jones
@ 2006-03-27 19:58 ` Dave Jones
2006-03-27 21:06 ` Dominik Brodowski
0 siblings, 1 reply; 11+ messages in thread
From: Dave Jones @ 2006-03-27 19:58 UTC (permalink / raw)
To: cpufreq
On Mon, Mar 27, 2006 at 02:53:10PM -0500, Dave Jones wrote:
> On Sun, Mar 26, 2006 at 11:53:39AM +0200, Dominik Brodowski wrote:
> > Hi Dave,
> >
> > Please pull from
> > git://git.kernel.org/pub/scm/linux/kernel/git/brodo/cpufreq-2.6.git/
> > for a few cpufreq updates for 2.6.17. The diffstat and list of changes
> > follows, the patches will be sent out as individual mails as replies to this
> > message.
>
> I'm not entirely sure what happened here.
> I git pull'd from this, and ended up with a 59,886 line delta against Linus' tree.
> As you sent the diffs too, I'll just apply-mbox them.
pilot-error on my part, the pull succeeded.
Dave
--
http://www.codemonkey.org.uk
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [git pull] cpufreq governor updates for 2.6.17
2006-03-27 19:58 ` Dave Jones
@ 2006-03-27 21:06 ` Dominik Brodowski
0 siblings, 0 replies; 11+ messages in thread
From: Dominik Brodowski @ 2006-03-27 21:06 UTC (permalink / raw)
To: Dave Jones; +Cc: cpufreq
On Mon, Mar 27, 2006 at 02:58:10PM -0500, Dave Jones wrote:
> On Mon, Mar 27, 2006 at 02:53:10PM -0500, Dave Jones wrote:
> > On Sun, Mar 26, 2006 at 11:53:39AM +0200, Dominik Brodowski wrote:
> > > Hi Dave,
> > >
> > > Please pull from
> > > git://git.kernel.org/pub/scm/linux/kernel/git/brodo/cpufreq-2.6.git/
> > > for a few cpufreq updates for 2.6.17. The diffstat and list of changes
> > > follows, the patches will be sent out as individual mails as replies to this
> > > message.
> >
> > I'm not entirely sure what happened here.
> > I git pull'd from this, and ended up with a 59,886 line delta against Linus' tree.
> > As you sent the diffs too, I'll just apply-mbox them.
>
> pilot-error on my part, the pull succeeded.
In fact, it was my mistake, as I cloned a copy of Linus' tree, checked that
you didn't have anything new in your cpufreq tree since Linus' last pull, but
forgot to reset my cpufreq tree to the HEAD of your cpufreq tree... Sorry
about that,
Dominik
^ permalink raw reply [flat|nested] 11+ messages in thread