Preparation for enabling hw filters. mp_image_params can't have an
AVHWFramesContext reference (because it can't hold any allocations, and
isn't meant to hold "active" data in the first place.
So just use a mp_image. It has all real data removed, because that would
essentially leak 1 frame once the decoder or renderer don't need it
anymore.
hw_vaapi.c didn't do much interesting anymore. Other than the function
to create a device for decoding with vaapi-copy, everything can be done
by generic code. Other libavcodec hwaccels are planned to provide the
same API as vaapi. It will be possible to drop the other hw_ files in
the future. They will use this generic code instead.
This was a hack to let libmpv API users pass a d3d device to mpv. It's
not needed anymore for 2 reasons:
1. ANGLE does not have this problem
2. Even native GL via nVidia (where this failed) seems to not require
this anymore
even before the recent refactor the cursor was hidden when moving it to
the top of the screen in fullscreen and placing it on top of the now
visible menu bar.
we need to know when the menu bar is hidden so we don’t create a
‘dead zone’ at the top of the screen where the cursor can’t be hidden.
to determine when the menu bar is visible, and with that the title bar,
we get the height of the menu bar. the height is always 0 when hidden.
furthermore there is no way to get the title bar directly and with that
its height. so we calculate the frame rect of a NSWindowStyleMaskTitled
window from a CGRectZero content frame. the resulting height is the
height of a title bar.
with that we can exclude the top area for the cursor hiding and can be
certain when the menu bar is not hidden.
the cursor couldn’t be hidden when the cursor was at the same position
as the Dock, even if the cursor was next to it. this is especially
annoying in fullscreen since the Dock isn’t actually hidden but is still
reported as being visible. this basically made the part of the screen,
where the Dock resides, a ‘dead zone’. so instead of using the
visibleFrame we will just use the normal frame. there is no problem at
the top area of the screen, since a window can’t be placed above the
menu bar and in fullscreen the menu bar is always reported as not being
on screen.
i suspect this was done so the cursor wasn’t hidden when the it was
placed above the Dock when windowed. with the recent refactor this is
not needed any more.
we can simplify the code because we don't need to change the bool
pointer we were given by the VOCTRL_SET_CURSOR_VISIBILITY event. i
assume this was done to work around some bugs previously to the recent
cursor refactor. i kept that because i thought it was necessary, which
wasn't in the end. after the refactor it only caused some weirdnesses i
tried to work around. without it we can get rid of some special cases
and simplify the code quite a bit.
Implements --hwdec=videotoolbox on iOS. Similar to hwdec_osx.c, but
using CVPixelBuffer APIs available on iOS instead of the equivalent
IOSurface APIs in macOS.
We can drop the custom table.
For some reason, the interop does not accept GL_RGB_RAW_422_APPLE as
internal format for GL_RGB_422_APPLE, so switch the format table to use
GL_RGB (this way both interop and real textures work the same).
Another victim of the apparent requirement of exactly matching texture
formats is kCVPixelFormatType_32BGRA. vo_opengl wants to handle this as
normal RGBA texture, with a swizzle applied in the shader.
CGLTexImageIOSurface2D() rejects this, because it wants the exact
internal format. Just drop the format, because it's useless anyway.
(Maybe this is a bit too fragile...)
All supported pixel formats have a specific "mapping" of CPU data to
textures. This function determines the number and the formats of these
textures. Moving it to a helper will be useful for some hardware decode
interop backends, since they all need similar things.
GL_LUMINANCE_ALPHA is the only reason why per-plane swizzles exist.
Remove per-plane swizzles (again), and regrettably handle them as
special cases (again). Carry along the logical texture format (called
gl_format in some parts of the code, including the new one).
We also don't need a use_integer flag, since the new gl_format member
implies whether it's an integer texture. (Yes, the there are separate
logical GL formats for integer textures. This aspect of the OpenGL API
is hysteric at best.)
This should change nothing about actual rendering logic and GL API
usage.
Originally, there was probably some sort of intention to restrict it to
formats supported by the interop, or something. But in the end it was
overcomplicated nonsense.
In the future, we could use mp_hwdec_ctx.supported_formats or other
mechanisms to handle this in a better way.
mp_hwdec_ctx.ctx is not set to a dummy pointer - hwdec_devices_load() is
only used to detect whether to vo_opengl interop is present, and the
common hwdec code expects that the .ctx field is not NULL.
This also changes videotoolbox-copy to use --videotoolbox-format,
instead of the FFmpeg-set default.
The code for copying a videotoolbox surface to mp_image was duplicated
(with some minor differences - I picked the hw_videotoolbox.c version,
because it was "better"). mp_imgfmt_from_cvpixelformat() is somewhat
duplicated with the vt_formats[] table, but this will be fixed in a
later commit, and moving the function to shared code is preparation.
we reported some unnecessary mouse movements and not all mouse enter
and leave events. that lead to wrongly reported activity on hover areas
like on the OSC or comparable lua scripts. sometimes menu items were
shown that shouldn't be shown or they didn't vanish because of the
missing mouse leave event.
this incorporates @torque's fix for mouse leave events that weren't
triggered during a transition, like going to fullscreen. the
tracking area was updated but the mouse never left that area because
it was never over it.
besides some known cursor visibility bugs the aforementioned changes
also revealed some other bugs that weren't reproducible before because
of the missbehavior.
known issues, in some cases the cursor doesn't show or hide properly.
for example when switching spaces, switching Apps via CMD+Tab or a
system notification. former two could be fixed while keeping our current
blank cursor approach. though the notification case couldn't. there is
no event or similar to detect a notification and the cursor visibility
couldn't be recovered in any way.
new issues, i noticed that our event view isn't initialised yet when the
first VOCTRL_SET_CURSOR_VISIBILITY event gets dispatched, which depends
on the event view to be initialised. so the mouse cursor couldn't be
hidden when mpv was opened and the cursor was within the window bounds.
this wasn't noticeable before because of various bugs and unwanted
behavior that have been fixed with this. now, in case the event view
isn't ready yet, we set the visibility at a later point when the event
view is ready and a helper flag is set.
Fixes#1817#3856#4147
this fixes a small bug with black edges on live resize, due to the
synchronisation with the DisplayLink. we just pause the DisplayLink for
the duration of the live resize. i also added some convenience functions
for reoccurring calls and simplified some DisplayLink related screen
info.
since there are different views on what ontop is, we make the ontop
window level modifiable. at the moment only support for macOS was added.
the default for macOS was changed from 'system' to 'window' since this
fixes an unwanted behaviour in fullscreen and in general causes less
issues with expected behaviour.
Fixes#2376#3974
This replaces the old backend that exclusively used EGL windowing with
one that can also use ANGLE's ability to render to directly to a
texture. The advantage of this is that it allows mpv to create the swap
chain itself and this allows mpv to use a flip-mode swap chain on a HWND
(which avoids problems with DirectComposition) and to use a longer swap
chain that has six backbuffers by default (which reportedly fixes
problems with rendering 24fps video on 24Hz monitors.)
Also, "screenshot window" should now work on DXGI 1.2 and up (Windows 8
and up.)
when the color profile was changed it used the right NSScreen but with
the old colorSpace. this was optimised out by a previous commit because
of a wrong assumption. we need to update the screen so we can get the
new colorSpace. this adds a bit of redundancy since on screen change it
will update screen pointer twice.
the problem here is that dropped files can also be treated as
NSURLPboardType instead of just NSFilenamesPboardType. the 'else if'
could never be reached and was dead code.
this basically reverts ed695ce which tried to fix multiple dropped
URLs, or rather files, and moves the filename check infront of the URL
check. the filename path can handle multiple dropped files, whereas the
URL path can only handle one dropped URL. this assumes that only one URL
can be dropped at a time. it also reverts a603543 because it's not
needed any more.
this also fixes a problem where dropped URLs from Chrome don't conform
to the NSURL class and the readObjectsForClasses method always returned
an empty URL.
Fixes#4036
this optimises two things and fix a minor bug.
1. we always updated the display refresh rate on any mode change whether
it was the current screen or not. now we only update the refresh rate
when the mode changed happened on the current screen.
2. the windowDidChangeScreen event doesn't exclusively trigger on screen
changes so we updated the display refresh rate in cases where it wasn't
needed at all. since we manually keep track of the current screen, we
can easily test if the screen really changed.
3. weirdly on initWithContentRect accessing the screen of the window
always returned the main screen instead of the screen the window is
created on. since we already use the window init method with the screen
as argument, overwrite that method instead and use the screen argument.
we are dealing with several problems here, which weren't apparent
because we always initialised a new displaylink for the display refresh
rate retrieval, previously to commit 449eb20 and bug 9490b62.
just changing the display with CVDisplayLinkSetCurrentCGDisplay
can cause inconsistent refresh rates and discontinuity in timestamps.
this can either lead to bogus values for the Actual display refresh rate
or retrieving the refresh rate of the previous display if we immediately
try to get a new value. since the Actual refresh rate is computed i
assume that it at least needs one refresh period to actual return
something useful.
furthermore when changing the screen and updating the displaylink, it
seems that the retrieved refresh rates for the screen mpv wasn't opened
on are being estimated in a sub-optimal way. as an example, when moving
my window to my second screen the Actual refresh rate was always a
constant 60Hz, even though it is supposed to fluctuate a little bit.
though if mpv was started on the secondary screen the Actual refresh
rate fluctuated around 59.94Hz like expected. in that case my primary
screen always reported a constant 60Hz instead.
for the first problem we moved the actual retrieval of the refresh rate
to the very last moment when mpv actual requests a new value and not
when the refresh rate changed. we only update the displaylink itself
when a possible refresh rate change is detected. this gives the
displaylink some time to calculate the new refresh rate. for the second
problem, instead of setting the new display we completely uninitialise
the old dislaylink and create a new one for the new screen. this gives
us properly estimated refresh rates.
additionally we also optimised the display refresh rate fallback
heuristic. it will never be 0 anymore and we prevent it from returning
bogus values with a simple threshold for the difference of the Actual
and Nominal refresh rate.
This should allow us to create the device in situations when
Direct3DCreate9 normally fails, for example if no user is logged in.
While the later use-case is not very interesting, I hope it to work in
some other situations as well, for example while certain drivers are in
exclusive full screen mode.
This is available since Windows 7, so I'm removing the old call
completely.
setting a window level other than NSNormalWindowLevel always sets
NSWindowCollectionBehaviorTransient, which prevents certain things to
work properly. examples are automatic switching to the active Space when
mpv is made active and (de-)miniaturizing. latter always lead to a
vanishing window.
Fixes#1757#1884
The lock was disabled recently. This commit gets rid of the dummied out
calls. The main reason for removing it is that there is no apparent need
for it anymore, and the new FFmpeg vaapi code does not use or provide
such a lock (there are some places which we cannot control and which do
vaapi API calls, like frame destructors).
Apparently this is the maximum that can be preserved. There is also
something about the decoder being able only to use 3 frames at a time,
and I'm assuming these are part of the 8 frames.
It was basically inverted. Not sure how this even happened. Hopefully
it's more an "I don't know what I was doing" instead of an "I don't know
what I am doing" case.
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.
This can be useful in other contexts.
Note that we end up setting AVCodecContext.width/height instead of
coded_width/coded_height now. AVCodecParameters can't set coded_width,
but this is probably more correct anyway.
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.
vo_opengl used to have it as sub-option, which made it very hard to pass
down option values to backends in a generic way (even if these options
were completely backend-specific). For --opengl-dcomposition we used a
VOFLAG to deal with this. Fortunately, sub-options are gone, and we can
just add it as global option.
Move the option to context_angle.c and add it as global option. I
thought about adding a mechanism to let backends declare options, which
would get magically picked up my m_config instead of having to add them
to the global option list manually (similar to VO vo_driver.options),
but decided against this complexity just for 1 or 2 backends. Likewise,
it could have been added as a single option to avoid the boilerplate of
an option struct, but then again there are probably going to be more
angle suboptions, and it's cleaner.
Scale the window by the assumed DPI scaling factor, using 96 DPI as
base. For example, a screen that reports 192 DPI is assumed to have a
DPI scale factor 2. The window will then be created with twice the size.
For robustness reasons, we accept only integer DPI scales between 1 and
9. We also error out if the X and Y scales are very different, as this
most likely indicates a multiscreen system with botched size reporting.
I'm not sure if reading the X server's DPI is such a good idea - maybe
the Xrdb "Xft.dpi" value should be used instead. The current method
follows what xdpyinfo does.
This can be disabled with --hidpi-window-scale=no.
our constrainFrameRect prevents our window from positioning itself ontop
of the menubar, which is unwanted for a fullscreen window. this always
positioned our window vertically at -22/-23pt when going into fullscreen
because of the menubar. this bug doesn't show on newer versions of OS X
since the various flags we set force the window position. on OS X 10.9
though the fullscreen window was shifted 22pt downwards. even though
this bug doesn't show on newer OS X versions, it should still be fixed
for a possible behaviour changes in future version.
Fixes#4044
everytime we switched to a new video file a new displaylink was
initialised and started, but the old one was not stopped and released
beforehand. this lead to several displaylink callback calls per swap,
depending on how many files were switched beforehand. moving the
displaylink init call to the cocoa init functions will ever only init
one displaylink.
Fixes#4031
we are calling the method on a NSWindow object that may not respond to
that call, since its a method of MpvVideoWindow. add the method to our
protocol and rename that protocol to reflect the change.
in some circumstances cocoa isn't able to enter or exit fullscreen but
we still set window sizes and flags accordingly. this leaves us in a
hanging state between fullscreen and window. it also prevents the
toggleFullscreen method and its events to work properly afterwards. in
that state it's impossible to enter or exit this 'semi-fullscreen'.
add a proper fallback to recover from this state.
Fixes#4035
Introduce the --opengl-hwdec-interop option, which replaces
--hwdec-preload. The new option allows explicit selection of the interop
backend.
This is relatively complex, and I would have preferred not to add this,
but it's probably useful to debug certain problems. In exchange, the
"new" option documents that pretty much any but the simplest use of it
will not be forward compatible.
Replace the old code, that played games to evade thread-safety issues,
with newer thread-safe option access functions.
This also means mp_opengl_create() doesn't need to cache the hwdec
settings anymore. (They're applied in mpv_opengl_cb_init_gl() instead.)
Tried to decode a High 4:2:2 file, since libavcodec code seemed to
indicate that it's supported. Well, it decodes to garbage.
I couldn't find out why ffmpeg.c actually appears to reject this
correctly. The API seems to be fine with, just that the output is
garbage.
Add a hack for now.
Successful decoding of a frame resets ctx->hwdec_fail_count to 0 - which
us ok, but prevents fallback if it fails if --vd-lavc-software-fallback
is set to something higher than 1.
Just fail it immediately, since failing here always indicates some real
error (or OOM), not e.g. a video parsing error or such, which we try to
tolerate via the error counter.
Use the libavutil vdpau frame allocation code instead of our own "old"
code. This also uses its code for copying a video surface to normal
memory (used by vdpau-copy).
Since vdpau doesn't really have an internal pixel format, 4:2:0 can be
accessed as both nv12 and yuv420p - and libavutil prefers to report
yuv420p. The OpenGL interop has to be adjusted accordingly.
Preemption is a potential problem, but it doesn't break it more than it
already is.
This requires a bug fix to FFmpeg's vdpau code, or vdpau-copy (as well
as taking screenshots) will fail. Libav has fixed this bug ages ago.
In a way it can be reused. For now, sw_format and initial_pool_size
determination are still vaapi-specific. I'm hoping this can be eventally
moved to libavcodec in some way. Checking the supported_formats array is
not really vaapi-specific, and could be moved to the generic code path
too, but for now it would make things more complex.
hw_cuda.c can't use this, but hw_vdpau.c will in the following commit.
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.
Because it allows easier testing of filters + hwdec.
Make the texture setup code a bit more generic so it doesn't get too
much of a mess. We also use the GL renderer utility function
gl_find_unorm_format(), which saves us additional work with OpenGL's
semi-redundant format specifiers.
If hardware decoding is enabled (via --hwdec anything), the player was
printing an informational message that software decdoing is used.
Basically, this confuses users, because they think there is a problem or
such. Just disable the message, it's semi-useless anyway.
This was suggested on IRC, after yet another user was asking why this
message was shown (with a follow up discussion which CPUs can decode
what kind of video codecs).
EGL rendering + new decode API didn't work due to a certain libva bug
with sort-of legacy API use hitting again. It will report the wrong
vaapi pixel format. It's old code and always nv12 anyway, so stop
worrying about it.
There are going to be users who have a Mesa installation which do not
support 10 bit, but a GPU which can decode to 10 bit. So it's probably
better not to hardcode whether it is supported.
Introduce a more general way to signal supported formats from renderer
to decoder. Obviously this is imperfect, because it still isn't part of
proper format negotation (for example, what if there's a vavpp filter,
which accepts anything). Still slightly better than before.
I don't know any way to probe for vaapi dmabuf/EGL dmabuf support
properly (in particular testing specific formats, not just general
availability). So we stay with the current approach and try to create
and map dummy surfaces on init to probe for support. Overdo it and check
all formats that AVHWFramesConstraints reports, instead of only NV12 and
P010 surfaces.
Since we can support unknown formats now, add explicitly checks to the
EGL/dmabuf mapper code to reject unsupported formats. I also noticed
that libavutil signals support for RGB0/BGR0, but couldn't get it to
work. Remove the DRM formats that are unused/didn't work the way I tried
to use them.
With this, 10 bit decoding + rendering should work, provided you have
a capable CPU and a patched Mesa. The required Mesa patch adds support
for the R16 and GR32 formats. It was sent by a Kodi developer to the
Mesa developer mailing list and was not accepted yet.
For convenience. Since we still have code that works even if creating a
AVHWDeviceContext fails, failure is ignored. (Although currently, it
succeeds creation even with the stale/abandoned vdpau wrapper driver.)
For surfaces allocated by libavutil, we assume that the sw_format (i.e.
in hw_subfmt in mp_image_params) is always correct. The API guarantees
that it explicitly sets the equivalent vaapi format on surface
allocation.
For surfaces allocated by mpv's old vaapi code, we explicitly retrieve
the format right after decoding. Unless the driver magically changes the
format asynchronously, it will still be correct once the surface reaches
the renderer.
In both cases, checking the format again is obviously redundant. In
addition, it doesn't require us to maintain a libva fourcc <-> mpfmt
table and the va_fourcc_to_imgfmt() function. This also unbreaks 10 bit
rendering support (still disabled by default).
This does not work, because Mesa has no support for the proposed
DRM_FORMAT_R16 and DRM_FORMAT_GR16 formats. It's also untested of
course.
As long as video/decode/vaapi.c doesn't hand down P010 surfaces, this is
fine anyway.
This can be tested by removing the code that disables P010 output:
diff --git a/video/decode/vaapi.c b/video/decode/vaapi.c
--- a/video/decode/vaapi.c
+++ b/video/decode/vaapi.c
@@ -55,13 +55,6 @@ static int init_decoder(struct lavc_ctx *ctx, int w, int h)
assert(!ctx->avctx->hw_frames_ctx);
- // If we use direct rendering, disallow 10 bit - it's probably not
- // implemented yet, and our downstream components can't deal with it.
- if (!p->own_ctx && required_sw_format != AV_PIX_FMT_NV12) {
- MP_WARN(ctx, "10 bit surfaces are currently supported.\n");
- return -1;
- }
-
Rendering support in Mesa probably doesn't exist yet. In theory it might
be possible to use VPP to convert the surfaces to 8 bit (like we do it
with dxva2/d3d11va as ANGLE doesn't support rendering 10 bit surface
either), but that too would require explicit mechanisms. This can't be
implemented either until I have a GPU with actual support.
Other hwdecs will also be able to use this (as soon as they are switched
to use AVHWFramesContext).
As an additional feature, failing to copy back the frame counts as
hardware decoding failure and can trigger fallback. This can be done
easily now, because it needs no way to communicate this from the hwaccel
glue code to the common code.
The old code is still required for the old decode API, until we either
drop or rewrite it. vo_vaapi.c's OSD code (fuck...) also uses these
surface functions to a higher degree.
mp_image_hw_download() is a libavutil wrapper added in the previous
commit. We drop our own code completely, as everything is provided by
libavutil and our helper wrapper.
This breaks the screenshot code, so that has to be adjusted as well.
Makes va_surface_download() call mp_image_hw_download() for
libavutil-allocated surfaces, which in turn calls
av_hwframe_transfer_data().
mp_image_hw_download() is actually not specific to vaapi, and can be
used for any hw surface allocated by libavutil.
Mostly affects conversion of the colorimetric parameters.
Not changing AV_FRAME_DATA_MASTERING_DISPLAY_METADATA handling - that's
too messy, as decoders typically output it for keyframes only, and would
require weird caching that can't even be done on the level of the frame
rewrapping functions.
This fixes direct rendering with hwdec_vaegl.c.
The code duplication between update_image_params() and
mp_image_copy_fields_from_av_frame() is quite annoying,
bit will have to be resolved in another commit.
AVHWDeviceContext.user_opaque is reserved to libavutil under certain
circumstances, while AVHWFramesContext.user_opaque is truly free for use
by us. It's slightly simpler too.
Dummy out the locking around all libva calls, which in theory shouldn't
be needed anyway. Two years ago, these were added to prevent crashes
with vaapi decoding and direct rendering (vo_opengl/vo_vaapi) active.
It's not clear whether these are really needed - theory strongly points
towards no. Some developers familiar with vaapi expressed similar
thoughts. But past experience says differently. So let's try
without the locking for a while and see what happens.
A recent commit added code that checks some HAVE_ symbols in this file.
No config.h include was added, so they could be unavailable and break
compilation (in practice, just --hwdec=vaapi-copy would break).
Not sure how I missed this, maybe waf defined these symbols on the
compiler command line for some reason.
The old API is deprecated, and libavcodec prints a warning at runtime.
The new API is a bit nicer and does many things for you, such as
managing the underlying hwaccel decoder. libavutil also provides code
for managing surfaces (we use their surface pool).
The new code does not contain any code from the original MPlayer VAAPI
patch (that was used as base for some of the vaapi code in mpv). Thus
the new code is LGPL.
The new API actually does not add any visible symbols, so the only way
to detect it is a version check. Of course, the versions overlap
between FFmpeg and Libav, which requires additional care. The new
API did not get merged into FFmpeg yet, so there's no check for
FFmpeg.
We usually attach some significant metadata and context to "our"
surfaces. Surfaces created by libavutil (such as we plan to do it when
using the new vaapi decode API in the following commit) don't have this
context, so e.g. copy decoding mode won't work.
Add tons of hacks to make this somehow work.
Eventually we will use libavutil's mechanisms and drop the hacks.
