[FFmpeg-cvslog] r21799 - in trunk/libavcodec: Makefile alsdec.c bgmc.c bgmc.h

[Thilo Borgmann]

Am 13.02.10 20:44, schrieb Justin Ruggles:
> thilo.borgmann wrote:
> 
>> +            for (; i < sb_length; i++) {
>> +                int32_t res = *current_res;
>> +
>> +                if (res == cur_tail_code) {
>> +                    unsigned int max_msb =   (2 + (sx[sb] > 2) + (sx[sb] > 10))
>> +                                          << (5 - delta[sb]);
>> +
>> +                    res = decode_rice(gb, cur_s);
>> +
>> +                    if (res >= 0) {
>> +                        res += (max_msb    ) << cur_k;
>> +                    } else {
>> +                        res -= (max_msb - 1) << cur_k;
>> +                    }
>> +                } else {
>> +                    if (res > cur_tail_code)
>> +                        res--;
>> +
>> +                    if (res & 1)
>> +                        res = -res;
>> +
>> +                    res >>= 1;
>> +
>> +                    if (cur_k) {
>> +                        res <<= cur_k;
>> +                        res  |= get_bits_long(gb, cur_k);
>> +                    }
>> +                }
>> +
>> +            *current_res++ = res;
>> +            }
> 
> wrong indentation on the last line there.

Ups..


> 
> Congrats on getting the arithmetic decoding done.  How is the
> conformance test coverage now?  We should be getting pretty close to
> complete.  Only floating-point support and channel reordering are
> remaining for full spec implementation, right?

Thanks!

Basically yes, although there is also this rls-lms thing. And the spec
says something about encoding all the time ;)

-Thilo