[PATCH 28/31] iio: gyro: use parity32 in adxrs450.c

[PATCH 28/31] iio: gyro: use parity32 in adxrs450.c

[zhaoxiu.zeng]

From: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>

Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
---
 drivers/iio/gyro/adxrs450.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/drivers/iio/gyro/adxrs450.c b/drivers/iio/gyro/adxrs450.c
index a330d42..f1f19fc20 100644
--- a/drivers/iio/gyro/adxrs450.c
+++ b/drivers/iio/gyro/adxrs450.c
@@ -108,9 +108,7 @@ static int adxrs450_spi_read_reg_16(struct iio_dev *indio_dev,
 
 	mutex_lock(&st->buf_lock);
 	tx = ADXRS450_READ_DATA | (reg_address << 17);
-
-	if (!(hweight32(tx) & 1))
-		tx |= ADXRS450_P;
+	tx |= !parity32(tx) * ADXRS450_P;
 
 	st->tx = cpu_to_be32(tx);
 	ret = spi_sync_transfer(st->us, xfers, ARRAY_SIZE(xfers));
@@ -144,9 +142,7 @@ static int adxrs450_spi_write_reg_16(struct iio_dev *indio_dev,
 
 	mutex_lock(&st->buf_lock);
 	tx = ADXRS450_WRITE_DATA | (reg_address << 17) | (val << 1);
-
-	if (!(hweight32(tx) & 1))
-		tx |= ADXRS450_P;
+	tx |= !parity32(tx) * ADXRS450_P;
 
 	st->tx = cpu_to_be32(tx);
 	ret = spi_write(st->us, &st->tx, sizeof(st->tx));
-- 
2.5.5

Re: [PATCH 28/31] iio: gyro: use parity32 in adxrs450.c

[Jonathan Cameron]

On 27/03/16 08:42, zhaoxiu.zeng wrote:
> From: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
> 
> Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
Interesting.  Whilst obviously correct I wonder if this obscures the
intent of the code a little. Lars, what do you think?

Jonathan
> ---
>  drivers/iio/gyro/adxrs450.c | 8 ++------
>  1 file changed, 2 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/iio/gyro/adxrs450.c b/drivers/iio/gyro/adxrs450.c
> index a330d42..f1f19fc20 100644
> --- a/drivers/iio/gyro/adxrs450.c
> +++ b/drivers/iio/gyro/adxrs450.c
> @@ -108,9 +108,7 @@ static int adxrs450_spi_read_reg_16(struct iio_dev *indio_dev,
>  
>  	mutex_lock(&st->buf_lock);
>  	tx = ADXRS450_READ_DATA | (reg_address << 17);
> -
> -	if (!(hweight32(tx) & 1))
> -		tx |= ADXRS450_P;
> +	tx |= !parity32(tx) * ADXRS450_P;
>  
>  	st->tx = cpu_to_be32(tx);
>  	ret = spi_sync_transfer(st->us, xfers, ARRAY_SIZE(xfers));
> @@ -144,9 +142,7 @@ static int adxrs450_spi_write_reg_16(struct iio_dev *indio_dev,
>  
>  	mutex_lock(&st->buf_lock);
>  	tx = ADXRS450_WRITE_DATA | (reg_address << 17) | (val << 1);
> -
> -	if (!(hweight32(tx) & 1))
> -		tx |= ADXRS450_P;
> +	tx |= !parity32(tx) * ADXRS450_P;
>  
>  	st->tx = cpu_to_be32(tx);
>  	ret = spi_write(st->us, &st->tx, sizeof(st->tx));
> 

Re: [PATCH 28/31] iio: gyro: use parity32 in adxrs450.c

[Lars-Peter Clausen]

On 03/28/2016 10:35 AM, Jonathan Cameron wrote:
> On 27/03/16 08:42, zhaoxiu.zeng wrote:
>> From: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
>>
>> Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
> Interesting.  Whilst obviously correct I wonder if this obscures the
> intent of the code a little. Lars, what do you think?

The parity function is newly introduced in this series and can be more
efficient that  just hw_weight() & 1 on certain architectures. Since the
result is the same using it is certainly an improvement. But ...

[...]
>> -	if (!(hweight32(tx) & 1))
>> -		tx |= ADXRS450_P;
>> +	tx |= !parity32(tx) * ADXRS450_P;

... this should still be

if (!parity32(tx))
	tx |= ADXRS450_P;

Otherwise it's a bit too much obfuscated for my taste. Just leave it to the
compiler to optimize it as it sees it fit.