[FFmpeg-devel] [PATCH] RoQ video encoder, take 2

[matthieu castet]

Hi,

Reimar D?ffinger wrote:
> Hello,
>> +typedef struct
>> +{
>> +    uint32_t eval_se[3];
>> +
>> +    roq_subcel_evaluation_t subCels[4];
>> +
>> +    int8_t motionX, motionY;
>> +    uint8_t cbEntry;
>> +
>> +    uint32_t sourceX, sourceY;
>> +} cel_evaluation_t;
> 
> Unless there is a special reason not to, try using just "int" instead of
> "int8_t".
> 
Yes, and for the question why, there are some architecture (like arm) 
where everything is 32 bits. Using int8_t force the compiler to add 
extra instructions to emulate 8 bits operations (overflow, ...).