[Vorbis-dev] Building a new decoder - some questions on the spec (residue decode)

[Peter Steele]

I didn't hear anything from anybody but I have resolved the problem by
looking through the libvorbis code.

I did find that the specification document for residue decode is not
100% correct and that libvorbis does not implement it as it is written
(from a logic perspective). The key differences being :

1) limit_residue_begin = residue_begin (does != max(residue_begin,actualSize))
2) building classifications information has incorrect logic in the spec



On Sat, Mar 28, 2015 at 7:50 AM, Peter Steele <steelehere at gmail.com> wrote:
> Hi
>
> I am trying to build a new decoder in Java, I know there are a couple
> of java decoders already but they don't implement the bit of the spec
> I am having questions on
>
> Question 1)
>
> In section 8.6.2 it mentions
>
> "However, it is also possible to set [residue_begin] and [residue_end]
> to specify
> a range partially or wholly beyond the maximum vector size. Before
> beginning residue
> decode, limit [residue_begin] and [residue_end] to the maximum
> possible vector size
> as follows."
>
> The code segment then has
>
> 1) [actual\_size] = current blocksize/2;
> 2) if residue encoding is format 2
> 3) [actual\_size] = [actual\_size] * [ch];
> 4) [limit\_residue\_begin] = maximum of ([residue\_begin],[actual\_size]);
> 5) [limit\_residue\_end] = maximum of ([residue\_end],[actual\_size]);
>
> In a real world vorbis file I see
>
> residue_begin = 0;
> residue_end = 256;
> actual_size = 8;
>
> following that code I would get
>
> limit_residue_begin = 8
> limit_residue_end = 256
>
> Is this right? It feels wrong having an actual size of 8 yet the
> difference of end-begin being 248.
>
> Question 2)
>
> Further down in 8.6.2 it has a line of code
>
> 18) if ([vqbook] is not ’unused’) {
>
> Can someone tell me the full definition of 'unused' here? I have an
> instance currently where the vqbook returned is not of type 1 or 2 so
> doesn't have any vqvectors (The index returned from the
> classifications was 0 which is the default up populated value from the
> code in 8.6.1 which was retrieved in the header packet decode)
>
> Many Thanks
>
> Peter