[FFmpeg-devel] [GSoC] avcodec/als: Add ALS encoder

[Thilo Borgmann]

Hi,

> Patch attached.
> 
> It fixes the fate tests.
> However, there's a slight bug in the encoder in handling the last frame.
> I'll definitely fix it later.
> I hope the patch can be merged in this state.

no. The last frame has to be handled properly before merging happens.
I'm at LinuxCon & ELCE for the next two weeks, so don't expect too much input from me before Oct. 14th. Please fix whatever is possible until then.

-Thilo
 
> - Umair
> 
> On Mon, Sep 12, 2016 at 12:53 PM, Umair Khan <omerjerk at gmail.com> wrote:
> 
>> Hi,
>>
>> Sorry for late reply. I was travelling a bit.
>> Attached is the patch which includes the following changes than the
>> previous one -
>>
>> 1. Removed changes of the file libavformat/movenc.c, as I had added
>> this file in the initial commit by mistake.
>> 2. Removed the assembly code.
>> 3. Make changes as suggested by Michael (av_assertX and header_size).
>>
>> As far as fate tests are concerned, I haven't checked that yet. I'll
>> see what's the issue there.
>>
>> On Mon, Aug 29, 2016 at 11:34 PM, Michael Niedermayer
>> <michael at niedermayer.cc> wrote:
>>> On Mon, Aug 29, 2016 at 10:47:59PM +0530, Umair Khan wrote:
>>>> Hi,
>>>>
>>>> On Sun, Aug 28, 2016 at 4:26 PM, Michael Niedermayer
>>>> <michael at niedermayer.cc> wrote:
>>>>> On Sun, Aug 28, 2016 at 01:34:46PM +0530, Umair Khan wrote:
>>>>>> Hi,
>>>>>>
>>>>>> Patches attached. :)
>>>>>>
>>>>>> - Umair
>>>>>
>>>>>>  Changelog |    1 +
>>>>>>  1 file changed, 1 insertion(+)
>>>>>> d3f30e62d803d967bd5c27dc5dfad278ce5c02e9
>> 0001-Changelog-Add-entry-for-ALS-encoder.patch
>>>>>> From 020370545a82c8c1304ec9c177a75e27e59b84e8 Mon Sep 17 00:00:00
>> 2001
>>>>>> From: Umair Khan <omerjerk at gmail.com>
>>>>>> Date: Sat, 27 Aug 2016 22:22:02 +0530
>>>>>> Subject: [PATCH 1/2] Changelog: Add entry for ALS encoder
>>>>>>
>>>>>> Signed-off-by: Umair Khan <omerjerk at gmail.com>
>>>>>> ---
>>>>>>  Changelog | 1 +
>>>>>>  1 file changed, 1 insertion(+)
>>>>>>
>>>>>> diff --git a/Changelog b/Changelog
>>>>>> index b903e31..90c15ad 100644
>>>>>> --- a/Changelog
>>>>>> +++ b/Changelog
>>>>>> @@ -15,6 +15,7 @@ version <next>:
>>>>>>  - True Audio (TTA) muxer
>>>>>>  - crystalizer audio filter
>>>>>>  - acrusher audio filter
>>>>>> +- ALS encoder
>>>>>>
>>>>>>
>>>>>>  version 3.1:
>>>>>> --
>>>>>> 2.7.4 (Apple Git-66)
>>>>>>
>>>>>
>>>>> [...]
>>>>>> +static int calc_short_term_prediction(ALSEncContext *ctx, ALSBlock
>> *block,
>>>>>> +                                       int order)
>>>>>> +{
>>>>>> +    ALSSpecificConfig *sconf = &ctx->sconf;
>>>>>> +    int i, j;
>>>>>> +
>>>>>> +    int32_t *res_ptr = block->res_ptr;
>>>>>> +    int32_t *smp_ptr = block->cur_ptr;
>>>>>> +
>>>>>> +    assert(order > 0);
>>>>>
>>>>> should be av_assertX (X=0/1/2)
>>>>>
>>>>>
>>>>> [...]
>>>>>> +int ff_window_init(WindowContext *wctx, enum WindowType type, int
>> length,
>>>>>> +                   double param)
>>>>>> +{
>>>>>> +    if (!length || length < -1)
>>>>>> +        return AVERROR(EINVAL);
>>>>>> +
>>>>>> +    wctx->type   = type;
>>>>>> +    wctx->length = length;
>>>>>> +    wctx->param  = param;
>>>>>> +
>>>>>> +    switch (type) {
>>>>>> +    case WINDOW_TYPE_RECTANGLE:
>>>>>> +        rectangle_init(wctx);
>>>>>> +        break;
>>>>>> +    case WINDOW_TYPE_WELCH:
>>>>>> +        WINDOW_INIT(welch)
>>>>>> +        break;
>>>>>> +    case WINDOW_TYPE_SINERECT:
>>>>>> +        WINDOW_INIT(sinerect)
>>>>>> +        break;
>>>>>> +    case WINDOW_TYPE_HANNRECT:
>>>>>> +        WINDOW_INIT(hannrect)
>>>>>> +        break;
>>>>>> +    default:
>>>>>> +        return AVERROR(EINVAL);
>>>>>> +    }
>>>>>> +
>>>>>
>>>>>> +    if (HAVE_MMX)
>>>>>> +        ff_window_init_mmx(wctx);
>>>>>
>>>>> breaks build on non x86 as the function declaration / prototype is
>>>>> not there in that case
>>>>
>>>> What should I do with this then? I'm not too aware of how the whole
>>>> code works because I didn't originally write it.
>>>> So, I'll need some help here. :)
>>>
>>> IIRC the declaration / prototype is under #if
>>> but the call is not under #if
>>> thus if the condition on the #if is untrue this fails to build
>>>
>>> (this is not the same as the functions implementation being missing
>>>  for if(0) code, that one works with all supported platforms)
>>>
>>> its probably best to split the whole *_mmx code out into a seperate
>>> patch and also either the call must be under #if or the declaration
>>> must be available independant of an #if
>>>
>>>
>>> [...]
>>> --
>>> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>>>
>>> Democracy is the form of government in which you can choose your dictator
>>>
>>> _______________________________________________
>>> ffmpeg-devel mailing list
>>> ffmpeg-devel at ffmpeg.org
>>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>>
>>
>>
>>
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel at ffmpeg.org
>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel