[FFmpeg-devel] [RFC][PATCH 1/3] lavu: Add av_crc2() and av_crc_init2()

[James Almer]

On 27/01/13 1:55 PM, Michael Niedermayer wrote:
> On Sun, Jan 27, 2013 at 03:52:13AM -0300, James Almer wrote:
>> Functionally the same as av_crc() and av_crc_init(), but
>> supporting CRCs of up to 64 bits.
>>
>> Signed-off-by: James Almer <jamrial at gmail.com>
> [...]
>> @@ -90,6 +98,12 @@ const AVCRC *av_crc_get_table(AVCRCId crc_id)
>>  uint32_t av_crc(const AVCRC *ctx, uint32_t crc,
>>                  const uint8_t *buffer, size_t length)
>>  {
>> +    return (uint32_t)av_crc2(ctx, crc, buffer, length);
>> +}
>> +
>> +uint64_t av_crc2(const AVCRC *ctx, uint64_t crc,
>> +                 const uint8_t *buffer, size_t length)
>> +{
>>      const uint8_t *end = buffer + length;
>>  
>>  #if !CONFIG_SMALL
>> @@ -101,8 +115,9 @@ uint32_t av_crc(const AVCRC *ctx, uint32_t crc,
>>              crc ^= av_le2ne32(*(const uint32_t *) buffer); buffer += 4;
>>              crc = ctx[3 * 256 + ( crc        & 0xFF)] ^
>>                    ctx[2 * 256 + ((crc >> 8 ) & 0xFF)] ^
>> +                                 (crc >> 32)          ^
>>                    ctx[1 * 256 + ((crc >> 16) & 0xFF)] ^
>> -                  ctx[0 * 256 + ((crc >> 24)       )];
>> +                  ctx[0 * 256 + ((crc >> 24) & 0xFF)];
>>          }
>>      }
>>  #endif
> 
> use of 64bit variables to calculate 32bit crcs would likely cause a
> significant performance loss on 32bit architectures

Should I instead create completely separate functions exclusively for CRC64 
then?

Doing so would also deal with Paul's concerns about changing the variable
types for a few structs.

Regards.