Skip to contentSkip to conversation

hwmon: (adm1031) Increase update rate

PATCH In review Ira W. Snyder 2 participants 4 messages Mar 30, 2010 – Apr 6, 2010 Hardware monitoring

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hwmon: (adm1031) Increase update rate

The adm1031 chip is capable of an 8 Hz update rate, configurable at runtime with the fan filter register. Increase the update rate to get faster temperature readings.

---

This may be a bit controversial. The ADM1031 chip may use more power at a higher update rate. In my (embedded) application, this doesn't matter. The faster update rate is more important.

Does anyone have thoughts on this? It is easy enough for me to carry myself, but I'd much rather have it upstream.

Ira

 drivers/hwmon/adm1031.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)
drivers/hwmon/adm1031.c +5 -1
index 1644b92..c7195cc 100644
@@ -36,6 +36,7 @@
 #define ADM1031_REG_FAN_DIV(nr)		(0x20 + (nr))
 #define ADM1031_REG_PWM			(0x22)
 #define ADM1031_REG_FAN_MIN(nr)		(0x10 + (nr))
+#define ADM1031_REG_FAN_FILTER		(0x23)
 
 #define ADM1031_REG_TEMP_OFFSET(nr)	(0x0d + (nr))
 #define ADM1031_REG_TEMP_MAX(nr)	(0x14 + 4 * (nr))
@@ -919,6 +920,9 @@ static void adm1031_init_client(struct i2c_client *client)
 				ADM1031_CONF1_MONITOR_ENABLE);
 	}
 
+	/* Increase the ADC sampling rate -- temperatures update at 8 Hz */
+	read_val = adm1031_read_value(client, ADM1031_REG_FAN_FILTER);
+	adm1031_write_value(client, ADM1031_REG_FAN_FILTER, read_val | 0x1C);
 }
 
 static struct adm1031_data *adm1031_update_device(struct device *dev)
@@ -929,7 +933,7 @@ static struct adm1031_data *adm1031_update_device(struct device *dev)
 
 	mutex_lock(&data->update_lock);
 
-	if (time_after(jiffies, data->last_updated + HZ + HZ / 2)
+	if (time_after(jiffies, data->last_updated + HZ / 8)
 	    || !data->valid) {
 
 		dev_dbg(&client->dev, "Starting adm1031 update\n");
Signature · 7 lines
1.5.4.3


_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
#1 lore raw

Hi Ira,

Quoted 10 lines · Ira W. Snyder

The adm1031 chip is capable of an 8 Hz update rate, configurable at runtime with the fan filter register. Increase the update rate to get faster temperature readings.

---

This may be a bit controversial. The ADM1031 chip may use more power at a higher update rate. In my (embedded) application, this doesn't matter. The faster update rate is more important.

If you use the automatic fan speed control capabilities, then indeed a faster update rate can be benefical. But as you suspected, it is somewhat controversial, due to the higher power consumption possibly for no good reason. Also, the BIOS/firmware might have already configured the chip and traditionally the hwmon drivers preserve these settings as much as possible.

Quoted 2 lines

Does anyone have thoughts on this? It is easy enough for me to carry myself, but I'd much rather have it upstream.

I wouldn't want it upstream as is. However I can think of two alternative implementations which I would be happy with:

1* Let the adm1031 driver accept platform data, and the update rate be part of that platform data. Then you can instantiate your device explicitly and pass the desired update rate.

2* Expose the update frequency through sysfs, and make it writable. After all, so many monitoring devices have a configurable update rate that it would make sense to have a standard attribute for it. If you go that route, we have to standardize on it first, and add it to Documentation/hwmon/sysfs-interface. Then implement it in the adm1031 driver and possibly other drivers.

Note that both options aren't mutually exclusive, although one of them should be enough to satisfy your need.

Quoted 34 lines
 drivers/hwmon/adm1031.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)
drivers/hwmon/adm1031.c +5 -1
index 1644b92..c7195cc 100644
@@ -36,6 +36,7 @@
 #define ADM1031_REG_FAN_DIV(nr)		(0x20 + (nr))
 #define ADM1031_REG_PWM			(0x22)
 #define ADM1031_REG_FAN_MIN(nr)		(0x10 + (nr))
+#define ADM1031_REG_FAN_FILTER		(0x23)
 
 #define ADM1031_REG_TEMP_OFFSET(nr)	(0x0d + (nr))
 #define ADM1031_REG_TEMP_MAX(nr)	(0x14 + 4 * (nr))
@@ -919,6 +920,9 @@ static void adm1031_init_client(struct i2c_client *client)
 				ADM1031_CONF1_MONITOR_ENABLE);
 	}
 
+	/* Increase the ADC sampling rate -- temperatures update at 8 Hz */
+	read_val = adm1031_read_value(client, ADM1031_REG_FAN_FILTER);
+	adm1031_write_value(client, ADM1031_REG_FAN_FILTER, read_val | 0x1C);
 }
 
 static struct adm1031_data *adm1031_update_device(struct device *dev)
@@ -929,7 +933,7 @@ static struct adm1031_data *adm1031_update_device(struct device *dev)
 
 	mutex_lock(&data->update_lock);
 
-	if (time_after(jiffies, data->last_updated + HZ + HZ / 2)
+	if (time_after(jiffies, data->last_updated + HZ / 8)
 	    || !data->valid) {
 
 		dev_dbg(&client->dev, "Starting adm1031 update\n");
Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
#2 lore raw
In reply to Ira W. Snyder
Quoted 14 lines · Jean Delvare

Hi Ira,

Quoted 10 lines · Ira W. Snyder

The adm1031 chip is capable of an 8 Hz update rate, configurable at runtime with the fan filter register. Increase the update rate to get faster temperature readings.

---

This may be a bit controversial. The ADM1031 chip may use more power at a higher update rate. In my (embedded) application, this doesn't matter. The faster update rate is more important.

Quoted 7 lines

If you use the automatic fan speed control capabilities, then indeed a faster update rate can be benefical. But as you suspected, it is somewhat controversial, due to the higher power consumption possibly for no good reason. Also, the BIOS/firmware might have already configured the chip and traditionally the hwmon drivers preserve these settings as much as possible.

Yep, that makes perfect sense.

Quoted 17 lines
Quoted 2 lines

Does anyone have thoughts on this? It is easy enough for me to carry myself, but I'd much rather have it upstream.

I wouldn't want it upstream as is. However I can think of two alternative implementations which I would be happy with:

1* Let the adm1031 driver accept platform data, and the update rate be part of that platform data. Then you can instantiate your device explicitly and pass the desired update rate.

2* Expose the update frequency through sysfs, and make it writable. After all, so many monitoring devices have a configurable update rate that it would make sense to have a standard attribute for it. If you go that route, we have to standardize on it first, and add it to Documentation/hwmon/sysfs-interface. Then implement it in the adm1031 driver and possibly other drivers.

Quoted 3 lines

Note that both options aren't mutually exclusive, although one of them should be enough to satisfy your need.

I like the idea of a sysfs node. Does "update_rate" (in milliseconds) sound ok, or do you have another name you'd prefer?

Thanks,
Ira

Quoted 27 lines
Quoted 27 lines
 drivers/hwmon/adm1031.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)
drivers/hwmon/adm1031.c +4 -0
index 1644b92..c7195cc 100644
@@ -36,6 +36,7 @@
 #define ADM1031_REG_FAN_DIV(nr)		(0x20 + (nr))
 #define ADM1031_REG_PWM			(0x22)
 #define ADM1031_REG_FAN_MIN(nr)		(0x10 + (nr))
+#define ADM1031_REG_FAN_FILTER		(0x23)
 
 #define ADM1031_REG_TEMP_OFFSET(nr)	(0x0d + (nr))
 #define ADM1031_REG_TEMP_MAX(nr)	(0x14 + 4 * (nr))
@@ -919,6 +920,9 @@ static void adm1031_init_client(struct i2c_client *client)
 				ADM1031_CONF1_MONITOR_ENABLE);
 	}
 
+	/* Increase the ADC sampling rate -- temperatures update at 8 Hz */
+	read_val = adm1031_read_value(client, ADM1031_REG_FAN_FILTER);
+	adm1031_write_value(client, ADM1031_REG_FAN_FILTER, read_val | 0x1C);
 }
 
 static struct adm1031_data *adm1031_update_device(struct device *dev)
@@ -929,7 +933,7 @@ static struct adm1031_data *adm1031_update_device(struct device *dev)
 
Quoted 12 lines
Quoted 7 lines
drivers/hwmon/adm1031.c +1 -1
 	mutex_lock(&data->update_lock);
 
-	if (time_after(jiffies, data->last_updated + HZ + HZ / 2)
+	if (time_after(jiffies, data->last_updated + HZ / 8)
 	    || !data->valid) {
 
 		dev_dbg(&client->dev, "Starting adm1031 update\n");
Signature · 1 line
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#3 lore raw
In reply to Ira W. Snyder
Quoted 42 lines · Ira W. Snyder
Quoted 14 lines · Jean Delvare

Hi Ira,

Quoted 10 lines · Ira W. Snyder

The adm1031 chip is capable of an 8 Hz update rate, configurable at runtime with the fan filter register. Increase the update rate to get faster temperature readings.

---

This may be a bit controversial. The ADM1031 chip may use more power at a higher update rate. In my (embedded) application, this doesn't matter. The faster update rate is more important.

Quoted 7 lines

If you use the automatic fan speed control capabilities, then indeed a faster update rate can be benefical. But as you suspected, it is somewhat controversial, due to the higher power consumption possibly for no good reason. Also, the BIOS/firmware might have already configured the chip and traditionally the hwmon drivers preserve these settings as much as possible.

Yep, that makes perfect sense.

Quoted 17 lines
Quoted 2 lines

Does anyone have thoughts on this? It is easy enough for me to carry myself, but I'd much rather have it upstream.

I wouldn't want it upstream as is. However I can think of two alternative implementations which I would be happy with:

1* Let the adm1031 driver accept platform data, and the update rate be part of that platform data. Then you can instantiate your device explicitly and pass the desired update rate.

2* Expose the update frequency through sysfs, and make it writable. After all, so many monitoring devices have a configurable update rate that it would make sense to have a standard attribute for it. If you go that route, we have to standardize on it first, and add it to Documentation/hwmon/sysfs-interface. Then implement it in the adm1031 driver and possibly other drivers.

Quoted 6 lines
Quoted 3 lines

Note that both options aren't mutually exclusive, although one of them should be enough to satisfy your need.

I like the idea of a sysfs node. Does "update_rate" (in milliseconds) sound ok, or do you have another name you'd prefer?

Sounds good, no objection.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors