regulator: tps62360: Simplify tps62360_set_voltage_time_sel implementation

For linear mappings, we can use below equation to get the voltage difference between new_selector and old_selector:

abs(new_selector - old_selector) * rdev->desc->uV_step

---
 drivers/regulator/tps62360-regulator.c |   13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)

index e534269..d044a58 100644
@@ -179,17 +179,10 @@ static int tps62360_set_voltage_time_sel(struct regulator_dev *rdev,
 		unsigned int old_selector, unsigned int new_selector)
 {
 	struct tps62360_chip *tps = rdev_get_drvdata(rdev);
-	int old_uV, new_uV;
 
-	old_uV = regulator_list_voltage_linear(rdev, old_selector);
-	if (old_uV < 0)
-		return old_uV;
-
-	new_uV = regulator_list_voltage_linear(rdev, new_selector);
-	if (new_uV < 0)
-		return new_uV;
-
-	return DIV_ROUND_UP(abs(old_uV - new_uV), tps->change_uv_per_us);
+	return DIV_ROUND_UP(abs(new_selector - old_selector) *
+			    rdev->desc->uV_step,
+			    tps->change_uv_per_us);
 }
 
 static int tps62360_set_mode(struct regulator_dev *rdev, unsigned int mode)

1.7.9.5

Applied, thanks. There's a new framework feature that Yadwinder Singh has contributed which should allow this to be moved to the regulator desc.

2012/6/18 Mark Brown <[email protected]>:

In this case, tps->change_uv_per_us is is calculated by :
tps->change_uv_per_us = DIV_ROUND_UP(32000, BIT(ramp_ctrl));

Which means tps->change_uv_per_us may be not alignment to mV.

This make me think if we should change the unit of ramp_delay to uV/uS rather than mV/uS. So we can avoid truncate issue.

Regards,
Axel

Yes, we should have the ramp delay in uV/uS. many times, the ramp delay come as 0.025mV/uS which will not be possible if we do mv/us.