Most authors have agreed. Like with vo.c, we don't take module
declarations into consideration (see 0e09533c73).
Notable issues:
115bfb9762: the author has (probably) not agreed. The vf.c changes were
removed with the removal of filter DR, though (see c54fc507da).
7b25afd742: the same author adds VOCTRLs for deinterlacing switching at
runtime. Put them under a HAVE_GPL ifdef just to be sure. (It looks like
we could remove the VOCTRLs immediately, as they're needed only for some
compatibility things, but no need to do that yet.)
02b199e5e9: the author had a conditional agreement to LGPL, which
doesn't allow us to change it just yet, but the code added here was
completely removed anyway. (These days, the pts is passed as mp_image
field, and put_image is gone.)
3532cd532e: same author, but code removed with DR removal.
f0626e2f8d: same author, but code was moved to mp_image.c.
e5b4b495c3: agreed to LGPLv3+ only, but the code was removed in
cfa1f9e082 anyway.
086c324692: author was not asked - minor warning fix, but no mpv
includes malloc.h anymore.
e9d0a1d609: author was not asked - removed again in 33b62af947.
c260a1139d: author could not be reached - but this code was removed when
mpv changed the image allocation code to essentially use FFmpeg's
pixdesc.
Until now, the texture pointer was stored in plane 1, and the texture
array index was in plane 2. Move this down to plane 0 and plane 1. This
is to align it to the new WIP D3D11 decoding API in Libav, where we
decided that there is no reason to avoid setting plane 0, and that it
would be less weird to start at plane 0.
If the VO doesn't support a format output by vf_lavfi, no conversion
filter was inserted, and filter chain creation failed.
This is because vf_lavfi doesn't properly follow the format negotiation
model, which means the format negotiation pass does not catch all cases
where conversion is needed. Specifically, vf_lavfi supports that all
output formats are supported for any given input format, but then does
not actually call vf_next_query_format() in reconfig() to check which
format it uses, but outputs whatever it gets from libavfilter.
I think this is ok to avoid excessive complexity in vf_lavfi.c, but it
also means adding more kludges to vf.c. I justify this (and the code
duplication) with the idea that the current filter chain code will die
anyway at some point.
The .log field additions for c->first/c->last are strictly speaking not
needed, but useful for debugging.
Now you can for example do "--vf=hue=h=60" - there is no "hue" filter in
mpv, so libavfilter's will be used.
This has certain caveats (see manpage).
The point of this is providing a relatively smooth transition path to
removing our own filter stuff.
The plan is to nuke the custom filter chain completely. It's not clear
what will happen to the still needed builtin filters (mostly hardware
deinterlacing and vf_vapoursynth). Most likely we'll replace them with
different filter chain concept (whose main purpose will be providing
builtin things and bridging to libavfilter).
The undocumented "warn" options are there to disable deprecation
warnings when the player inserts filter automatically.
The same will be done to audio filters, at a later point.
Basically, see the example in input.rst.
This is better than the "old" vf-toggle method, because it doesn't
require the user to duplicate the filter string in mpv.conf and
input.conf.
Some aspects of this changes are untested, so enjoy your alpha testing.
I'm not sure what's going on here, but it appears kodi switches forward
and backwards references for advanced VPP deinterlacing modes. This in
turn makes deinterlacing with these modes apparently work. If you don't
switch the directions, you get a stuttering mess.
As far as the libva trace dump is concerned, this makes mpv's libva
deinterlacing API use behave like kodi's, and appears to reproduce
smooth video with advanced libva deinterlacing enabled.
I'm hearing that Mesa actually does it correctly, and I'm not sure what
will happen there. For now, passing "reversal-bug=no" as sub-option to
the vavpp filter will undo this behavior.
Don't give the driver more forward/backward refernces than it requested
in num_forward_references/num_backward_references. This shouldn't
matter, I'm just trying to play it safe.
When playing with VOs which do not provide mp_hwdec_ctx, vf->hwdec_devs
will remain NULL. This would make it crash on hwdec_devices_get_first(),
even if no hardware decoding or filters using hardware decoding were
involved.
Fixes#4064.
The FFmpeg versions we support all have the APIs we were checking for.
Only Libav missed them. Simplify this by explicitly checking for FFmpeg
in the code, instead of trying to detect the presence of the API.
Requires a bunch of hacks:
- we access AVFilterLink.hw_frames_ctx. This is not a public API in
FFmpeg and Libav. Newer FFmpeg provides an accessor
(av_buffersink_get_hw_frames_ctx), but it's not available in Libav or
the current FFmpeg release or Libav. We need this value after filter
graph creation, so We have no choice but to access this.
One alternative is making filter creation and format negotiation
fully lazy (i.e. delay it and do it as filters are output), but this
would be a huge change.
So for now, we knowingly violate FFmpeg's and Libav's ABI and API
constraints because they don't provide anything better.
On newer FFmpeg, we use the (quite ugly) accessor, though.
- mp_image_params doesn't (and can't) have a field for the frames
context AVBufferRef. So we pass it via vf_set_proto_frame(), and even
more hacks.
- if a filter needs a hw context, but we haven't created one yet
(because normally we create them lazily), it will fail at init.
- we allow any hw format now, although this could go horrible wrong.
Why all this effort? We could move hw deinterlacing filters etc. to
FFmpeg, which is a very worthy goal.
Instead of using the awful older "API" that passed the parameters
formatted as string. AVBufferSrcParameters is also a prerequisite for
hardware frame filtering support.
Looks quite like a bug. If you have a filter chain with only the
dynaudnorm filter, and send call av_buffersrc_add_frame(s, NULL), then
subsequent av_buffersink_get_frame() calls will return EAGAIN instead of
EOF.
This was apparently caused by a recent change in FFmpeg.
Some other circumstances (which I didn't fully analyze and which is due
to the playloop's absurd temporary-EOF behavior on seeks) then led the
decoder loop to send data again, but since libavfilter was stuck in the
EOF state now, it could never recover. It kept sending new input (due to
missing output), until the demuxer refused to return more audio packets.
Each time a filter error was printed.
Fortunately, it's pretty easy to workaround. We just mark the p->eof
flag as we send an EOF frame to libavfilter. The p->eof flag is used
only to recover from temporary EOF: it resets the filter if new data is
available again. We don't care much about av_buffersink_get_frame()
returning a broken EAGAIN state in this situation and essentially ignore
it, meaning if we get EAGAIN after sending EOF, we assume effectively
that EOF was fully reached.
We always want to use __declspec(selectany) to declare GUIDs, but
manually including <initguid.h> in every file that used GUIDs was
error-prone. Since all <initguid.h> does is define INITGUID and include
<guiddef.h>, we can remove all references to <initguid.h> and just
compile with -DINITGUID to get the same effect.
Also, this partially reverts 622bcb0 by re-adding libuuid.a to the
build, since apparently some GUIDs (such as GUID_NULL) are not declared
in the source file, even when INITGUID is set.
Normally I'd prefer a bunch of smaller functions with fewer parameters
over a single function with a lot of parameters. But future changes will
require messing with the parameters in a slightly more complex way, so a
combined function will be needed anyway. The now-unused "global"
parameter is required for later as well.
Just a minor refactor along the planned option change. This commit will
make it easier to update (i.e. copy) the VO options without copying
_all_ options. For now, behavior should be equivalent, though.
(The VO options were put into a separate struct quite early - when all
global variables were removed from the source code. It wasn't clear
whether the separate struct would have any actual purpose, but it seems
it will now. Awesome, huh.)
Instead of letting it keep decoding by trying to find a new frame,
"plug" the frame queue by not removing it. (Or actually, by putting
it back instead of discarding it.)
Matters for seamless looping (following commits), and possibly some
other corner cases.
The added function vf_unread_output_frame() is a bit of a sin, but still
reasonable, since its implementation is trivial.
This makes the difference between passing VA_FRAME_PICTURE or
VA_BOTTOM_FIELD for progressive frames (that should be force-
deinterlaced) to VAProcPipelineParameterBuffer.flags. VA-VPP doesn't
really seem to care, and we can get rid of mp_refqueue_is_interlaced()
entirely. It could be argued it's better to pass field flags instead of
the progressive flag.
"Real" frame flag vs. what we pretend it to be. It always used the real
flag, and thus never deinterlaced unflagged frames, even if the
suboption was set to "no".
I made this call up because I sort of thought this makes senssssse. I'm
now convinced that it does not, and even is actively harmful. I'm still
quite in the dark how the DirectD 11 video API is supposed to work with
synchronization, but at least for normal graphics this call would not
make much sense.
For example it should be set to IMGFMT_D3D11NV12 if it isn't already.
Otherwise, an assertion in vf.c could trigger.
This probably couldn't be provoked yet.
This involves multiple changes:
1. Brightness metadata is split into nominal peak and signal peak.
For a quick and dirty explanation: nominal peak is the brightest value
that your color space can represent (i.e. the brightness of an encoded
1.0), and signal peak is the brightest value that actually occurs in
the video (i.e. the brightest thing that's displayed).
2. vo_opengl uses a new decision logic to figure out the right nom_peak
and sig_peak for all situations. It also does a better job of picking
the right target gamut/colorspace to use for the OSD. (Which still is
and still should be treated as sRGB). This change in logic also
fixes#3293 en passant.
3. Since it was growing rapidly, the logic for auto-guessing / inferring
the right colorimetry configuration (in pass_colormanage) was split from
the logic for actually performing the adaptation (now pass_color_map).
Right now, the new logic doesn't do a whole lot since HDR metadata is
still ignored (but not for long).
This has two reasons:
1. I tend to add new fields to this metadata, and every time I've done
so I've consistently forgotten to update all of the dozens of places in
which this colorimetry metadata might end up getting used. While most
usages don't really care about most of the metadata, sometimes the
intend was simply to “copy” the colorimetry metadata from one struct to
another. With this being inside a substruct, those lines of code can now
simply read a.color = b.color without having to care about added or
removed fields.
2. It makes the type definitions nicer for upcoming refactors.
In going through all of the usages, I also expanded a few where I felt
that omitting the “young” fields was a bug.
See previous commit. (The mixing case was never supported, so this has
equivalent functionality.)
This also implicitly fixes a bug: the old code allocated image data for
the cropped surface size only, which means the
video_surface_get_bits_y_cb_cr call could write beyond the allocated
image memory.
Until now, we've always converted vdpau video surfaces to RGB, and then
mapped the resulting RGB texture. Change this so that the surface is
mapped as NV12 plane textures.
The reason this wasn't done until now is because vdpau surfaces are
mapped in an "interlaced" way as separate fields, even for progressive
video. This requires messy reinterleraving. It turns out that even
though it's an extra processing step, the result can be faster than
going through the video mixer for RGB conversion.
Other than some potential speed-gain, doing this has multiple other
advantages. We can apply our own color conversion, which is important in
more complex cases. We can correctly apply debanding and potentially
other processing that requires chroma-specific or in-YUV handling.
If deinterlacing is enabled, this switches back to the old RGB
conversion method. Until we have at least a primitive deinterlacer in
vo_opengl, this will stay this way. The d3d11 and vaapi code paths are
similar. (Of course these don't require any crazy field reinterleaving.)
Otherwise stale references will survive forever. Could leak hardware
video surfaces.
In particular, the mpv vdpau code crashed with an assertion when exiting
after toggling deinterlacing, because not all references were released.
Instead of doing HDR tone mapping on an ad-hoc basis inside
pass_colormanage, the reference peak of an image is now part of the
image params (alongside colorspace, gamma, etc.) and tone mapping is
done whenever peak_src != peak_dst.
To get sensible behavior when mixing HDR and SDR content and displays,
target-brightness is a generic filler for "the assumed brightness of SDR
content".
This gets rid of the weird display_scaled hack, sets the framework
for multiple HDR functions with difference reference peaks, and allows
us to (in a future commit) autodetect the right source peak from
the HDR metadata.
(Apart from metadata, the source peak can also be controlled via
vf_format. For HDR content this adjusts the overall image brightness,
for SDR content it's like simulating a different exposure)
Commit 0348cd08 was too naive/simple, and always inserted the d3d11vpp
filter if any d3d11 output image formats were supported, even if it
makes no sense. For example --vf=format=rgb8 already breaks it.
It needs to take the set of supported input formats into account. the
weird format negotiation makes this hard. As a simple and cheap
solution, make some assumptions about the supported formats of a filter.
I hope to simplify this one day by using another format negotiation
algorithm, but this can probably wait.
We now have a video filter that uses the d3d11 video processor, so it
makes no sense to have one in the VO interop code. The VO uses it for
formats not directly supported by ANGLE (so the video data is converted
to a RGB texture, which ANGLE can take in).
Change this so that the video filter is automatically inserted if
needed. Move the code that maps RGB surfaces to its own inteorp backend.
Add a bunch of new image formats, which are used to enforce the new
constraints, and to automatically insert the filter only when needed.
The added vf mechanism to auto-insert the d3d11vpp filter is very dumb
and primitive, and will work only for this specific purpose. The format
negotiation mechanism in the filter chain is generally not very pretty,
and mostly broken as well. (libavfilter has a different mechanism, and
these mechanisms don't match well, so vf_lavfi uses some sort of hack.
It only works because hwaccel and non-hwaccel formats are strictly
separated.)
The RGB interop is now only used with older ANGLE versions. The only
reason I'm keeping it is because it's relatively isolated (uses only
existing mechanisms and adds no new concepts), and because I want to be
able to compare the behavior of the old code with the new one for
testing. It will be removed eventually.
If ANGLE has NV12 interop, P010 is now handled by converting to NV12
with the video processor, instead of converting it to RGB and using the
old mechanism to import that as a texture.
Main use: deinterlacing.
I'm not sure how to select the deinterlacing mode at all. You can
enumate the available video processors, but at least on Intel, all of
them either signal support for all deinterlacers, or none (the latter is
apparently used for IVTC). I haven't found anything that actually tells
the processor _which_ algorithm to use.
Another strange detail is how to select top/bottom fields and field
dominance. At least I'm getting quite similar results to vavpp on Linux,
so I'm content with it for now.
Future plans include removing the D3D11 video processor use from the
ANGLE interop code.
This makes vf_vdpaupp use the deinterlacer helper code already used by
vf_vavpp. I nice side-effect is that this also removes some traces of
code originating from vo_vdpau.c, so we can switch it to LGPL.
Extend the refqueue helper with a deint setting. If not set,
mp_refqueue_should_deint() always returns false, which slightly
simplifies vf_vdpaupp. It's of no consequence to vf_vavpp (other than it
has to set it to get expected behavior).