Skip to contentSkip to conversation

hwmon: (lm75) Tune resolution and sample time per chip

PATCH In review Jean Delvare 2 participants 12 messages Mar 12, 2013 – Mar 14, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 4/5 hwmon: (lm75) Tune resolution and sample time per chip

Most LM75-compatible chips can either sample much faster or with a much better resolution than the original LM75 chip. So far the lm75 driver did not let the user take benefit of these improvements. Do it now.

I decided to almost always configure the chip to use the best resolution possible, which also means the longest sample time. The only chips for which I didn't are the DS75, DS1775 and STDS75, because they are really too slow in 12-bit mode (1.2 to 1.5 second worst case) so I went for 11-bit mode as a more reasonable tradeoff. This choice is dictated by the fact that the hwmon subsystem is meant for system monitoring, it has never been supposed to be ultra-fast, and as a matter of fact we do cache the sampled values in almost all drivers.

If anyone isn't pleased with these default settings, they can always introduce a platform data structure or DT support for the lm75. That being said, it seems nobody ever complained that the driver wouldn't refresh the value faster than every 1.5 second, and the change made it faster for all chips even in 12-bit mode, so I don't expect any complaint.

---
 Documentation/hwmon/lm75 |    7 ++++---
 drivers/hwmon/lm75.c     |   42 ++++++++++++++++++++++++++++++++++++------
 2 files changed, 40 insertions(+), 9 deletions(-)
drivers/hwmon/lm75.c +36 -6
@@ -169,6 +169,7 @@ lm75_probe(struct i2c_client *client, co
 	int status;
 	u8 set_mask, clr_mask;
 	int new;
+	enum lm75_type kind = id->driver_data;
 
 	if (!i2c_check_functionality(client->adapter,
 			I2C_FUNC_SMBUS_BYTE_DATA | I2C_FUNC_SMBUS_WORD_DATA))
@@ -187,30 +188,59 @@ lm75_probe(struct i2c_client *client, co
 	set_mask = 0;
 	clr_mask = LM75_SHUTDOWN;		/* continuous conversions */
 
-	switch (id->driver_data) {
+	switch (kind) {
 	case adt75:
 		clr_mask |= 1 << 5;		/* not one-shot mode */
+		data->resolution = 12;
+		data->sample_time = HZ / 8;
 		break;
 	case ds1775:
 	case ds75:
 	case stds75:
-		clr_mask |= 3 << 5;		/* 9-bit mode */
+		clr_mask |= 3 << 5;
+		set_mask |= 2 << 5;		/* 11-bit mode */
+		data->resolution = 11;
+		data->sample_time = HZ;
+		break;
+	case lm75:
+	case lm75a:
+		data->resolution = 9;
+		data->sample_time = HZ / 2;
+		break;
+	case max6625:
+		data->resolution = 9;
+		data->sample_time = HZ / 4;
+		break;
+	case max6626:
+		data->resolution = 12;
+		data->resolution_limits = 9;
+		data->sample_time = HZ / 4;
+		break;
+	case tcn75:
+		data->resolution = 9;
+		data->sample_time = HZ / 8;
 		break;
 	case mcp980x:
+		data->resolution_limits = 9;
+		/* fall through */
 	case tmp100:
 	case tmp101:
+		set_mask |= 3 << 5;		/* 12-bit mode */
+		data->resolution = 12;
+		data->sample_time = HZ;
+		clr_mask |= 1 << 7;		/* not one-shot mode */
+		break;
 	case tmp105:
 	case tmp175:
 	case tmp275:
 	case tmp75:
-		clr_mask |= 3 << 5;		/* 9-bit mode */
+		set_mask |= 3 << 5;		/* 12-bit mode */
 		clr_mask |= 1 << 7;		/* not one-shot mode */
+		data->resolution = 12;
+		data->sample_time = HZ / 2;
 		break;
 	}
 
-	data->resolution = 9;
-	data->sample_time = HZ + HZ / 2;
-
 	/* configure as specified */
 	status = lm75_read_value(client, LM75_REG_CONF);
 	if (status < 0) {
Documentation/hwmon/lm75 +4 -3
@@ -67,7 +67,8 @@ the temperature falls below the Hysteres
 All temperatures are in degrees Celsius, and are guaranteed within a
 range of -55 to +125 degrees.
 
-The LM75 only updates its values each 1.5 seconds; reading it more often
+The driver caches the values for a period varying between 1 second for the
+slowest chips and 125 ms for the fastest chips; reading it more often
 will do no harm, but will return 'old' values.
 
 The original LM75 was typically used in combination with LM78-like chips
@@ -78,8 +79,8 @@ The LM75 is essentially an industry stan
 LM75 clones not listed here, with or without various enhancements,
 that are supported. The clones are not detected by the driver, unless
 they reproduce the exact register tricks of the original LM75, and must
-therefore be instantiated explicitly. The specific enhancements (such as
-higher resolution) are not currently supported by the driver.
+therefore be instantiated explicitly. Higher resolution up to 12-bit
+is supported by this driver, other specific enhancements are not.
 
 The LM77 is not supported, contrary to what we pretended for a long time.
 Both chips are simply not compatible, value encoding differs.
Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
maintainer Acked-by on v1 4/5
#1 lore raw
Quoted 14 lines · Jean Delvare

Most LM75-compatible chips can either sample much faster or with a much better resolution than the original LM75 chip. So far the lm75 driver did not let the user take benefit of these improvements. Do it now.

I decided to almost always configure the chip to use the best resolution possible, which also means the longest sample time. The only chips for which I didn't are the DS75, DS1775 and STDS75, because they are really too slow in 12-bit mode (1.2 to 1.5 second worst case) so I went for 11-bit mode as a more reasonable tradeoff. This choice is dictated by the fact that the hwmon subsystem is meant for system monitoring, it has never been supposed to be ultra-fast, and as a matter of fact we do cache the sampled values in almost all drivers.

Quoted 7 lines

If anyone isn't pleased with these default settings, they can always introduce a platform data structure or DT support for the lm75. That being said, it seems nobody ever complained that the driver wouldn't refresh the value faster than every 1.5 second, and the change made it faster for all chips even in 12-bit mode, so I don't expect any complaint.

I have been using update-interval to make it configurable. Good example is the LM90 driver.

Quoted 1 line
Quoted 113 lines
---
 Documentation/hwmon/lm75 |    7 ++++---
 drivers/hwmon/lm75.c     |   42 ++++++++++++++++++++++++++++++++++++------
 2 files changed, 40 insertions(+), 9 deletions(-)
drivers/hwmon/lm75.c +36 -6
@@ -169,6 +169,7 @@ lm75_probe(struct i2c_client *client, co
 	int status;
 	u8 set_mask, clr_mask;
 	int new;
+	enum lm75_type kind = id->driver_data;
 
 	if (!i2c_check_functionality(client->adapter,
 			I2C_FUNC_SMBUS_BYTE_DATA | I2C_FUNC_SMBUS_WORD_DATA))
@@ -187,30 +188,59 @@ lm75_probe(struct i2c_client *client, co
 	set_mask = 0;
 	clr_mask = LM75_SHUTDOWN;		/* continuous conversions */
 
-	switch (id->driver_data) {
+	switch (kind) {
 	case adt75:
 		clr_mask |= 1 << 5;		/* not one-shot mode */
+		data->resolution = 12;
+		data->sample_time = HZ / 8;
 		break;
 	case ds1775:
 	case ds75:
 	case stds75:
-		clr_mask |= 3 << 5;		/* 9-bit mode */
+		clr_mask |= 3 << 5;
+		set_mask |= 2 << 5;		/* 11-bit mode */
+		data->resolution = 11;
+		data->sample_time = HZ;
+		break;
+	case lm75:
+	case lm75a:
+		data->resolution = 9;
+		data->sample_time = HZ / 2;
+		break;
+	case max6625:
+		data->resolution = 9;
+		data->sample_time = HZ / 4;
+		break;
+	case max6626:
+		data->resolution = 12;
+		data->resolution_limits = 9;
+		data->sample_time = HZ / 4;
+		break;
+	case tcn75:
+		data->resolution = 9;
+		data->sample_time = HZ / 8;
 		break;
 	case mcp980x:
+		data->resolution_limits = 9;
+		/* fall through */
 	case tmp100:
 	case tmp101:
+		set_mask |= 3 << 5;		/* 12-bit mode */
+		data->resolution = 12;
+		data->sample_time = HZ;
+		clr_mask |= 1 << 7;		/* not one-shot mode */
+		break;
 	case tmp105:
 	case tmp175:
 	case tmp275:
 	case tmp75:
-		clr_mask |= 3 << 5;		/* 9-bit mode */
+		set_mask |= 3 << 5;		/* 12-bit mode */
 		clr_mask |= 1 << 7;		/* not one-shot mode */
+		data->resolution = 12;
+		data->sample_time = HZ / 2;
 		break;
 	}
 
-	data->resolution = 9;
-	data->sample_time = HZ + HZ / 2;
-
 	/* configure as specified */
 	status = lm75_read_value(client, LM75_REG_CONF);
 	if (status < 0) {
Documentation/hwmon/lm75 +4 -3
@@ -67,7 +67,8 @@ the temperature falls below the Hysteres
 All temperatures are in degrees Celsius, and are guaranteed within a
 range of -55 to +125 degrees.
 
-The LM75 only updates its values each 1.5 seconds; reading it more often
+The driver caches the values for a period varying between 1 second for the
+slowest chips and 125 ms for the fastest chips; reading it more often
 will do no harm, but will return 'old' values.
 
 The original LM75 was typically used in combination with LM78-like chips
@@ -78,8 +79,8 @@ The LM75 is essentially an industry stan
 LM75 clones not listed here, with or without various enhancements,
 that are supported. The clones are not detected by the driver, unless
 they reproduce the exact register tricks of the original LM75, and must
-therefore be instantiated explicitly. The specific enhancements (such as
-higher resolution) are not currently supported by the driver.
+therefore be instantiated explicitly. Higher resolution up to 12-bit
+is supported by this driver, other specific enhancements are not.
 
 The LM77 is not supported, contrary to what we pretended for a long time.
 Both chips are simply not compatible, value encoding differs.
Signature · 7 lines
-- 
Jean Delvare

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

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

#2 lore raw
In reply to Jean Delvare
Quoted 22 lines · Guenter Roeck
Quoted 14 lines · Jean Delvare

Most LM75-compatible chips can either sample much faster or with a much better resolution than the original LM75 chip. So far the lm75 driver did not let the user take benefit of these improvements. Do it now.

I decided to almost always configure the chip to use the best resolution possible, which also means the longest sample time. The only chips for which I didn't are the DS75, DS1775 and STDS75, because they are really too slow in 12-bit mode (1.2 to 1.5 second worst case) so I went for 11-bit mode as a more reasonable tradeoff. This choice is dictated by the fact that the hwmon subsystem is meant for system monitoring, it has never been supposed to be ultra-fast, and as a matter of fact we do cache the sampled values in almost all drivers.

Quoted 7 lines

If anyone isn't pleased with these default settings, they can always introduce a platform data structure or DT support for the lm75. That being said, it seems nobody ever complained that the driver wouldn't refresh the value faster than every 1.5 second, and the change made it faster for all chips even in 12-bit mode, so I don't expect any complaint.

Quoted 2 lines

I have been using update-interval to make it configurable. Good example is the LM90 driver.

Indeed. The LM90 case is a little different though, as the update rate has an effect primarily on power consumption, and for extreme values on accuracy if I remember correctly (due to less/no averaging.) While in the LM75 case it has no effect on power consumption but on resolution.

Are you suggesting that I should rename lm75_data.sample_time to update_interval, and change it from jiffies to ms, for consistency?

That I should add a read-only update_interval to report the update interval to user-space?

That I should make the update_interval attribute writable and change the resolution on the fly accordingly, on a per-chip basis? This would be non-trivial, as each chip have their own conversion times for each supported resolution.

Quoted 3 lines
Quoted 1 line

Thanks,

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
maintainer
#3 lore raw
In reply to Jean Delvare
Quoted 23 lines · Jean Delvare
Quoted 22 lines · Guenter Roeck
Quoted 21 lines · Jean Delvare

Most LM75-compatible chips can either sample much faster or with a much better resolution than the original LM75 chip. So far the lm75 driver did not let the user take benefit of these improvements. Do it now.

I decided to almost always configure the chip to use the best resolution possible, which also means the longest sample time. The only chips for which I didn't are the DS75, DS1775 and STDS75, because they are really too slow in 12-bit mode (1.2 to 1.5 second worst case) so I went for 11-bit mode as a more reasonable tradeoff. This choice is dictated by the fact that the hwmon subsystem is meant for system monitoring, it has never been supposed to be ultra-fast, and as a matter of fact we do cache the sampled values in almost all drivers.

If anyone isn't pleased with these default settings, they can always introduce a platform data structure or DT support for the lm75. That being said, it seems nobody ever complained that the driver wouldn't refresh the value faster than every 1.5 second, and the change made it faster for all chips even in 12-bit mode, so I don't expect any complaint.

Quoted 8 lines
Quoted 2 lines

I have been using update-interval to make it configurable. Good example is the LM90 driver.

Indeed. The LM90 case is a little different though, as the update rate has an effect primarily on power consumption, and for extreme values on accuracy if I remember correctly (due to less/no averaging.) While in the LM75 case it has no effect on power consumption but on resolution.

Hi Jean,

There are always multiple dimensions. Sample time, resolution, accuracy, and power consumption all interact with each other. I have been using update_interval so far to report/configure it all. "resolution" (or possibly tempX_resolution) was suggested as possible additional sysfs attribute at some point.

Quoted 3 lines

Are you suggesting that I should rename lm75_data.sample_time to update_interval, and change it from jiffies to ms, for consistency?

Would be a start.

Quoted 8 lines

That I should add a read-only update_interval to report the update interval to user-space?

That I should make the update_interval attribute writable and change the resolution on the fly accordingly, on a per-chip basis? This would be non-trivial, as each chip have their own conversion times for each supported resolution.

Actually that was the idea, though I did not think about complexity. If it is too complex it might not be worth the effort.

Thanks,
Guenter

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

#4 lore raw
In reply to Jean Delvare

Hi Guenter,

Quoted 12 lines · Guenter Roeck
Quoted 4 lines · Jean Delvare

Indeed. The LM90 case is a little different though, as the update rate has an effect primarily on power consumption, and for extreme values on accuracy if I remember correctly (due to less/no averaging.) While in the LM75 case it has no effect on power consumption but on resolution.

There are always multiple dimensions. Sample time, resolution, accuracy, and power consumption all interact with each other. I have been using update_interval so far to report/configure it all. "resolution" (or possibly tempX_resolution) was suggested as possible additional sysfs attribute at some point.

Quoted 4 lines
Quoted 2 lines

Are you suggesting that I should rename lm75_data.sample_time to update_interval, and change it from jiffies to ms, for consistency?

Would be a start.

I slept over it and I don't think I'll do that change. lm90's update_interval refers to how frequently the chip starts a new measurement. The chips supported by this driver don't do continuous sampling, they sleep between measurements. The conversion time is constant.

lm75's sample_time is for how long the driver can cache the values. I ensured that it was always larger than the maximum possible conversion time, but for some of the supported chips this maximum is a lot more than the typical conversion time. LM75-like chips do continuous sampling, and we don't know the exact conversion time, so we simply don't know the update_interval value. All we can do is estimate its maximum.

So it's not as simple as renaming a variable, changing its unit and exporting it to sysfs. Some more thinking and care is required, which I have no time for ATM.

Quoted 10 lines
Quoted 7 lines

That I should add a read-only update_interval to report the update interval to user-space?

That I should make the update_interval attribute writable and change the resolution on the fly accordingly, on a per-chip basis? This would be non-trivial, as each chip have their own conversion times for each supported resolution.

Actually that was the idea, though I did not think about complexity. If it is too complex it might not be worth the effort.

For me it's not worth it. I think my patches already represent a significant improvement to the lm75 driver, it will make readings faster and more accurate out of the box. I don't want to spend more time on this, I'd rather spend time reviewing pending patches ;)

Of course if someone else steps up later to work on this, I have no objection.

Signature · 6 lines
Jean Delvare

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

Hi Guenter,

Hi Jean,

Quoted 5 lines
Quoted 1 line

If it is too complex it might not be worth the effort.

For me it's not worth it. I think my patches already represent a significant improvement to the lm75 driver, it will make readings faster and more accurate out of the box. I don't want to spend more

Makes sense.

Quoted 2 lines

time on this, I'd rather spend time reviewing pending patches ;)

Does that mean I can now submit the 20-something patches I didn't bother to submit due to lack of reviewers ? Just asking :-)

Guenter

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

#6 lore raw
In reply to Jean Delvare
Quoted 2 lines · Guenter Roeck

Does that mean I can now submit the 20-something patches I didn't bother to submit due to lack of reviewers ? Just asking :-)

Nah, I'd rather review the patches you've already sent... starting with ltc2978 and coretemp.

Which doesn't prevent you from posting patches anyway. And even pushing them upstream if nobody has time to review them and you're confident they'll do more good than harm.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
maintainer
#7 lore raw
In reply to Jean Delvare
Quoted 11 lines · Jean Delvare
Quoted 2 lines · Guenter Roeck

Does that mean I can now submit the 20-something patches I didn't bother to submit due to lack of reviewers ? Just asking :-)

Nah, I'd rather review the patches you've already sent... starting with ltc2978 and coretemp.

Which doesn't prevent you from posting patches anyway. And even pushing them upstream if nobody has time to review them and you're confident they'll do more good than harm.

I might actually do that at some point.

Key question right now though is what to do with the nct6775 driver ... I would really like to get it off my back. It just seems impossible to get anyone to review it. Not really surprising given its size. On the other side there is real demand for it, to the point that at least archlinux already provides it. I wonder if I should just push it into 3.10 and wait for any fallout. Do you have an opinion ?

Thanks,
Guenter

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

#8 lore raw
In reply to Jean Delvare
Quoted 6 lines · Guenter Roeck

Key question right now though is what to do with the nct6775 driver ... I would really like to get it off my back. It just seems impossible to get anyone to review it. Not really surprising given its size. On the other side there is real demand for it, to the point that at least archlinux already provides it. I wonder if I should just push it into 3.10 and wait for any fallout. Do you have an opinion ?

Yes, please push it into 3.10. It's a new driver, so you're not going to cause any regression. Worst case, driver doesn't work properly for some users, or some feature is broken, this can't really be worse than having no driver at all.

And if this ultimately lets us remove support for some chips from the w83627ehf driver and make it look sane again, I'm all for it.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
maintainer
#9 lore raw
In reply to Jean Delvare
Quoted 8 lines · Jean Delvare
Quoted 6 lines · Guenter Roeck

Key question right now though is what to do with the nct6775 driver ... I would really like to get it off my back. It just seems impossible to get anyone to review it. Not really surprising given its size. On the other side there is real demand for it, to the point that at least archlinux already provides it. I wonder if I should just push it into 3.10 and wait for any fallout. Do you have an opinion ?

Quoted 5 lines

Yes, please push it into 3.10. It's a new driver, so you're not going to cause any regression. Worst case, driver doesn't work properly for some users, or some feature is broken, this can't really be worse than having no driver at all.

Ok, I'll do that.

Quoted 3 lines

And if this ultimately lets us remove support for some chips from the w83627ehf driver and make it look sane again, I'm all for it.

Makes sense. This also makes one of the patches I submitted for w83627ehf unnecessary, as it affects nct6775/nct6776 support.

Thanks,
Guenter

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

#10 lore raw
In reply to Jean Delvare
Quoted 6 lines · Guenter Roeck
Quoted 2 lines · Jean Delvare

And if this ultimately lets us remove support for some chips from the w83627ehf driver and make it look sane again, I'm all for it.

Makes sense. This also makes one of the patches I submitted for w83627ehf unnecessary, as it affects nct6775/nct6776 support.

Not necessarily an issue, having both drivers overlap for a period of time is desirable for smooth transition.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
maintainer
#11 lore raw
In reply to Jean Delvare
Quoted 11 lines · Jean Delvare
Quoted 6 lines · Guenter Roeck
Quoted 2 lines · Jean Delvare

And if this ultimately lets us remove support for some chips from the w83627ehf driver and make it look sane again, I'm all for it.

Makes sense. This also makes one of the patches I submitted for w83627ehf unnecessary, as it affects nct6775/nct6776 support.

Not necessarily an issue, having both drivers overlap for a period of time is desirable for smooth transition.

Agreed. More a matter of time needed for the code review. I'll let you decide.

Dropping nct6775/nct6776 from the w83627ehf driver will reduce its size by more than 500 LOC, and its size by more than 7k (49k -> 41.5k) on x86_64.

Guenter

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