[FFmpeg-devel] [PATCH 2/4] avfilter/avfilter: Add numbers of (in|out)pads directly to AVFilter

[Andreas Rheinhardt]

Nicolas George:
> Andreas Rheinhardt (12021-08-12):
>> Up until now, an AVFilter's lists of input and output AVFilterPads
>> were terminated by a sentinel and the only way for the user to get
>> the length of these lists was by using avfilter_pad_count(). This
>> has two drawbacks: first, sizeof(AVFilterPad) is not negligible
>> (i.e. 64B on 64bit systems); second, getting the size involves
>> a function call instead of just reading the data.
>>
>> This commit therefore changes this. The sentinels are removed and new
>> fields nb_inputs and nb_outputs are added to AVFilter that contain the
>> number of elements of the respective AVFilterPad array.
>> Given that AVFilter.(in|out)puts are the only arrays of zero-terminated
>> AVFilterPads an API user has access to (AVFilterContext.(in|out)put_pads
>> are not zero-terminated and they already have a size field) the argument
>> to avfilter_pad_count() is always one of these lists, so it just has to
>> find the filter the list belongs to and read said number. This is slower
>> than before, but users are expected to switch to AVFilter.nb_(in|out)puts;
>> and furthermore, avfilter_pad_count() is probably never called in hot
>> loops anyway.
>>
>> This saves about 49KiB from the binary; notice that these sentinels are
>> not in .bss despite being zeroed; in fact, they are in .data.rel.ro
>> due to the non-sentinels.
>>
>> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at outlook.com>
> 
> I think it is a good change.
> 
> We could shave ~3.5k extra by making the fields uint16_t; lavfi is
> absolutely not ready to handle filters with thousands of inputs, there
> are loops everywhere.
> 
> OTOH, I do not like the fact that it requires more boilerplate code.

When I made this I found it acceptable because it also removed the
boilerplate { NULL } at the end of the structures. And because it could
be easily automated.

> 
>> @@ -330,6 +329,7 @@ const AVFilter ff_asrc_aevalsrc = {
>>      .priv_size     = sizeof(EvalContext),
>>      .inputs        = NULL,
>>      .outputs       = aevalsrc_outputs,
>> +    .nb_outputs    = FF_ARRAY_ELEMS(aevalsrc_outputs),
>>      .priv_class    = &aevalsrc_class,
>>  };
> 
> A macro would solve that:
> 
> #define FILTER_INPUTS(array) \
>     .inputs = (array),
>     .nb_inputs = FF_ARRAY_ELEMS(array)

That would indeed be even better.

> 
> Also, I am rather not for using these fields directly. Keeping them
> private and leaving avfilter_pad_count() gives us more freedom,
> including the freedom to change the type from uint16_t to unsigned later
> without break.
> 
But then I'd rather replace avfilter_pad_count() with something that
does not require to loop over all filters. Something like unsigned
avfilter_pads_count(const AVFilter *filter, int is_input).

- Andreas