Re: [PATCH 3/3] media: atomisp: Use max() macros

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 29-Sep-24 11:34 PM, David Laight wrote:
> From: Ricardo Ribalda
>> Sent: 27 September 2024 10:42
>>
>> The max() macro produce nicer code and also fixes the following cocci
>> errors:
>>
>> drivers/staging/media/atomisp/pci/sh_css_frac.h:40:17-18: WARNING opportunity for max()
>> drivers/staging/media/atomisp/pci/sh_css_frac.h:50:17-18: WARNING opportunity for max()
>>
>> Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
>> ---
>>  drivers/staging/media/atomisp/pci/sh_css_frac.h | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/staging/media/atomisp/pci/sh_css_frac.h
>> b/drivers/staging/media/atomisp/pci/sh_css_frac.h
>> index 8ba65161f7a9..9642506d2388 100644
>> --- a/drivers/staging/media/atomisp/pci/sh_css_frac.h
>> +++ b/drivers/staging/media/atomisp/pci/sh_css_frac.h
>> @@ -37,7 +37,7 @@ static inline int sDIGIT_FITTING(int v, int a, int b)
>>  	int fit_shift = sFRACTION_BITS_FITTING(a) - b;
>>
>>  	v >>= sSHIFT;
> 
> IIRC right shifts of signed values are undefined.
> (C does not require a cpu to have a right shift that replicates the
> sign bit.)
> 
>> -	v >>= fit_shift > 0 ? fit_shift : 0;
>> +	v >>= max(fit_shift, 0);
> 
> If the shift isn't done the return value is garbage.
> So the code better not let it happen.
> In which case you might as well let the cpu generate a (different)
> random value - so delete the conditional.

Given the history of this code I would no be surprised if some
weird corner case actually relies on the check, so NACK for
dropping the conditional.

> 
>>
>>  	return clamp_t(int, v, sISP_VAL_MIN, sISP_VAL_MAX);
> 
> all three values seem to be 'int' - so no need for the _t variant
> and all the associated casts.

sDIGIT_FITTING() originally was a macro with a bunch of max() + min()
calls nested leading to it expanding to a lot of code after running it
through the pre-processor. When converting this to a static online to
choice was made to with clamp_t() to avoid the overhead of the extra
type checks in regular clamp().

Regards,

Hans




> 
>>  }
>> @@ -47,7 +47,7 @@ static inline unsigned int uDIGIT_FITTING(unsigned int v, int a, int b)
>>  	int fit_shift = uFRACTION_BITS_FITTING(a) - b;
>>
>>  	v >>= uSHIFT;
>> -	v >>= fit_shift > 0 ? fit_shift : 0;
>> +	v >>= max(fit_shift, 0);
>>
>>  	return clamp_t(unsigned int, v, uISP_VAL_MIN, uISP_VAL_MAX);
> 
> as above, but it is just min(v, iISP_VAL_MAX)
> 
> 	David
> 
>>  }
>>
>> --
>> 2.46.1.824.gd892dcdcdd-goog
>>
> 
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)