avfilter: Deduplicate default video inputs/outputs

Lots of video filters use a very simple input or output:
An array with a single AVFilterPad whose name is "default"
and whose type is AVMEDIA_TYPE_VIDEO; everything else is unset.

Given that we never use pointer equality for inputs or outputs*,
we can simply use a single AVFilterPad instead of dozens; this
even saves .data.rel.ro (8312B here) as well as relocations.

*: In fact, several filters (like the filters in vf_lut.c)
already use the same outputs; furthermore, ff_filter_alloc()
duplicates the input and output pads so that we do not even
work with the pads directly.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>

1 files changed, 1 insertions, 8 deletions

diff --git a/libavfilter/vf_tonemap.c b/libavfilter/vf_tonemap.c
index d1087e6bd9..f34448d0e9 100644
--- a/libavfilter/vf_tonemap.c
+++ b/libavfilter/vf_tonemap.c
@@ -307,13 +307,6 @@ static const AVFilterPad tonemap_inputs[] = {
     },
 };
 
-static const AVFilterPad tonemap_outputs[] = {
-    {
-        .name         = "default",
-        .type         = AVMEDIA_TYPE_VIDEO,
-    },
-};
-
 const AVFilter ff_vf_tonemap = {
     .name            = "tonemap",
     .description     = NULL_IF_CONFIG_SMALL("Conversion to/from different dynamic ranges."),
@@ -321,7 +314,7 @@ const AVFilter ff_vf_tonemap = {
     .priv_size       = sizeof(TonemapContext),
     .priv_class      = &tonemap_class,
     FILTER_INPUTS(tonemap_inputs),
-    FILTER_OUTPUTS(tonemap_outputs),
+    FILTER_OUTPUTS(ff_video_default_filterpad),
     FILTER_PIXFMTS(AV_PIX_FMT_GBRPF32, AV_PIX_FMT_GBRAPF32),
     .flags           = AVFILTER_FLAG_SLICE_THREADS,
 };