opus - Oct 2017 - Fix DTX is always unavailable when DISABLE_FLOAT_API is not defined

Hi.

I found that DTX is always unavailable when DISABLE_FLAOT_API is not
defined in OPUS_AUTO mode.

As you know analysis_info.valid is alway true except NaN case and
is_silence is alway true except digital zero signal case.

In general, following condition will be alway true except exceptional case.

(analysis_info.valid || is_silence)

But in a code, there is a NOT expression in front of above condition, so
st->silk_mode.useDTX will be always disable

Here is the code.

#ifndef DISABLE_FLOAT_API

       st->silk_mode.useDTX = st->use_dtx && !(analysis_info.valid
||
is_silence);

#else

       st->silk_mode.useDTX = st->use_dtx;

#endif

Is it a bug or are there any reason for this?

In my opinion, the NOT expression should to be removed.

I attached a fixed file.

Please review.


Thanks,

InGyu Kang
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.xiph.org/pipermail/opus/attachments/20171017/e6f4c1c9/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: opus_encoder.c
Type: text/x-csrc
Size: 93618 bytes
Desc: not available
URL:
<http://lists.xiph.org/pipermail/opus/attachments/20171017/e6f4c1c9/attachment-0001.c>

The reason we disable the SILK DTX in that case is because we have a
"global" DTX in the analysis code that works for CELT too.

	Jean-Marc

On 16/10/17 10:23 PM, 강인규 wrote:
> Hi.
> 
> I found that DTX is always unavailable when DISABLE_FLAOT_API is not
> defined in OPUS_AUTO mode.
> 
> As you know analysis_info.valid is alway true except NaN case and
> is_silence is alway true except digital zero signal case.
> 
> In general, following condition will be alway true except exceptional case.
> 
> (analysis_info.valid || is_silence)
> 
> But in a code, there is a NOT expression in front of above condition, so
> st->silk_mode.useDTX will be always disable
> 
> Here is the code.
> 
> #ifndef DISABLE_FLOAT_API
> 
>        st->silk_mode.useDTX = st->use_dtx &&
!(analysis_info.valid ||
> is_silence);
> 
> #else
> 
>        st->silk_mode.useDTX = st->use_dtx;
> 
> #endif
> 
> Is it a bug or are there any reason for this?
> 
> In my opinion, the NOT expression should to be removed.
> 
> I attached a fixed file.
> 
> Please review.
> 
> 
> Thanks,
> 
> InGyu Kang
> 
> 
> 
> _______________________________________________
> opus mailing list
> opus at xiph.org
> http://lists.xiph.org/mailman/listinfo/opus
>