Sucks, but better than freezing forever given the (to me) unpredictable
RPI behavior. This will be good enough to drop out of vsync timing mode,
or to abort playback.
Recreate all dispmanx objects after mode changes signalled by the TV
callback. This is needed since dispmanx objects are marked as invalid
and cease working.
One important point is that the vsync callbacks will stop coming when
this happens, so restoring the callback is important.
Note that the MMAL renderer itself does not get trashed by the firmware
on such events, but we completely reconfigure it anyway when it happens.
A minor simplification. Most callers don't need this, and there's no
good reason why the caller should provide an "initializer" like this.
(This function calls mp_image_new_dummy_ref(), which has no reason
for an initializer either.)
The active texture and some pixelstore parameters are now always reset
to defaults when entering and leaving the renderer. Could be important
for libmpv.
As of ffmpeg git master, only the libavdevice decklink wrapper supports
this. Everything else has dropped support.
You're now supposed to use AV_CODEC_ID_WRAPPED_AVFRAME, which works a
bit differently. Normal AVFrames should still work for these encoders.
Since what we're doing is a linear blend of the four colors, we can just
do it for free by using GPU sampling.
This requires significantly fewer texture fetches and calculations to
compute the final color, making it much more efficient. The code is also
much shorter and simpler.
Until now, we have made the assumption that a driver will use only 1
hardware surface format. the format is dictated by the driver (you
don't create surfaces with a specific format - you just pass a
rt_format and get a surface that will be in a specific driver-chosen
format).
In particular, the renderer created a dummy surface to probe the format,
and hoped the decoder would produce the same format. Due to a driver
bug this required a workaround to actually get the same format as the
driver did.
Change this so that the format is determined in the decoder. The format
is then passed down as hw_subfmt, which allows the renderer to configure
itself with the correct format. If the hardware surface changes its
format midstream, the renderer can be reconfigured using the normal
mechanisms.
This calls va_surface_init_subformat() each time after the decoder
returns a surface. Since libavcodec/AVFrame has no concept of sub-
formats, this is unavoidable. It creates and destroys a derived
VAImage, but this shouldn't have any bad performance effects (at
least I didn't notice any measurable effects).
Note that vaDeriveImage() failures are silently ignored as some
drivers (the vdpau wrapper) support neither vaDeriveImage, nor EGL
interop. In addition, we still probe whether we can map an image
in the EGL interop code. This is important as it's the only way
to determine whether EGL interop is supported at all. With respect
to the driver bug mentioned above, it doesn't matter which format
the test surface has.
In vf_vavpp, also remove the rt_format guessing business. I think the
existing logic was a bit meaningless anyway. It's not even a given
that vavpp produces the same rt_format for output.
In the past, --video-unscaled also disabled zooming and aspect ratio
corrections. But this didn't make much sense in terms of being a useful
option. The new behavior just sets the initial video size to be
unscaled, but it's still affected by zoom commands and aspect ratio
corrections.
To get the old behavior back, --video-aspect=0 --video-zoom=0 need to be
added as well (in the general case). Most of the time it should not make
a difference though.
Also, there seems to have been some additional dst_rect clamping code
inside src_dst_split_scaling that didn't seem to either be necessary nor
ever get triggered. (The code immediately above it already makes sure to
crop the video if it's larger than the dst_rect)
No idea why it was there, but I just removed it.
Apply basic transformations like rotation by 90° and mirroring when
sampling from the source textures. The original idea was making this
part of img_tex.transform, but this didn't work: lots of code plays
tricks on the transform, so manipulating it is not necessarily
transparent, especially when width/height are switched. So add a new
pre_transform field, which is strictly applied before the normal
transform.
This fixes most glitches involved with rotating the image.
Cropping and rotation are now weirdly separated, even though they could
be done in the same step. I think this is not much of a problem, and
has the advantage that changing panscan does not trigger FBO
reallocations (I think...).
Typically happens with some implementations if no context is currrent,
or is otherwise broken. This is particularly relevant to the opengl_cb
API, because the API user will have no other indication what went wrong.
The underlying intention of this code is to make changing
--videotoolbox-format at runtime work. For this reason, the format can't
just be statically setup, but must be read from the option at runtime.
This means the format is not fixed anymore, and we have to make sure the
renderer is property reinitialized if the format changes. There is
currently no way to trigger reinit on this level, which is why the
mp_image_params.hw_subfmt field was introduced.
One sketchy thing remains: normally, the renderer is supposed to be
involved with VO format negotiation, which would ensure that the VO
can take the format at all. Since the hw_subfmt is not part of this
format negotiation, it's implied the get_vt_fmt() callback only
returns formats supported by the renderer. This is not necessarily
clear because vo_opengl checks this with converted_imgfmt separately.
None of this matters in practice though, because we know all formats
are always supported.
(This still requires somehow triggering decoder reinit to make the
change effective.)
The sync-by-display mode relies on using the vsync statistics for
timing. As a consequence discontinuities must be handled somehow. Until
now we have done this by completely resetting these statistics.
This can be somewhat annoying, especially if the GL driver's vsync
timing is not ideal. So after e.g. a seek it could take a second until
it locked back to the proper values.
Change it not to reset all statistics. Some state obviously has to be
reset, because it's a discontinuity. To make it worse, the driver's
vsync behavior will also change on such discontinuities. To compensate,
we discard the timings for the first 2 vsyncs after each discontinuity
(via num_successive_vsyncs). This is probably not fully ideal, and
num_total_vsync_samples handling in particular becomes a bit
questionable.
The past behavior was a bit weird, especially when zooming out. There
was no simple way to zoom in or out in consistent increments using
keybindings alone.
The new behavior preserves most of the old behavior's semantics but
scales out to infinity better. It coincidentally also makes it
really easy to get clean power of 2 ratios (e.g. 2x, 4x, 8x and their
inverses).
Fixes#3004.
This makes the black point closer (chromatically) to the white point, by
ensuring channels keep their consistent brightness ratios as they go
down to zero.
I also raised the 3DLUT version as this changes semantics and is a
separate commit from the previous one.
This commit refactors the 3DLUT loading mechanism to build the 3DLUT
against the original source characteristics of the file. This allows us,
among other things, to use a real BT.1886 profile for the source. This
also allows us to actually use perceptual mappings. Finally, this
reduces errors on standard gamut displays (where the previous 3DLUT
target of BT.2020 was unreasonably wide).
This also improves the overall accuracy of the 3DLUT due to eliminating
rounding errors where possible, and allows for more accurate use of
LUT-based ICC profiles.
The current code is somewhat more ugly than necessary, because the idea
was to implement this commit in a working state first, and then maybe
refactor the profile loading mechanism in a later commit.
Fixes#2815.
This also draws it after color management etc. In a nutshell, this
change makes the transparency checkerboard independent of upscaling,
panning, cropping etc. It will always be the same apparent size and
position (relative to the window).
It will also be independent of the video colorspace and such things.
(Note: This might cause white imbalance issues if playing a file with a
white point that does not match the display, in absolute colorimetric
mode. But that's uncommon, especially in conjunction with transparent
image files, so it's not a primary concern here)
Until now, we've let the windowing backend decide. But since they
usually require premultiplied alpha, and premultiplied alpha is easier
to handle, hardcode it.
The recent changes fixed rotation handling, but reversed the rotation
direction. The direction is expected to be counter-clockwise, because
demuxers export video rotation metadata as such.
This has been completely broken since commit 93546f0c. But even before,
rotation handling did not make too much sense. In particular, it rotated
the contents of the cropped image, instead of adjusting the crop
rectangle as well. The result was that things like panscan or zooming
did not behave as expected with rotation applied.
The same is true for vertical flipping. Flipping is triggered by
negative image stride. OpenGL does not support flipping the image on
upload, so it's done as part of the rendering. It can be triggered with
--vf=flip, but other filters and even decoders could setup negative
stride to flip the image.
Fix these issues by applying transforms to texture coordinates properly,
and by making rotation and flipping part of these transforms.
This still doesn't work properly for separated scaling. The issue is
that we'd have to adjust how the passes are done. For now, pick a very
stupid solution by rotating the image to a FBO, and then scaling from
that. This has the avantage that the scale logic doesn't have to be
complicated for such a rare case. It could be improved later.
Prescaling is apparently still broken. I don't know if chroma
positioning works properly either. None of this should affect the case
with no rotation.
gl_transform_vec() assumed column-major, while everything else seemed to
assumed row-major memory organization for gl_transform.m. Also,
gl_transform_trans() seems to contain additional confusion.
This didn't matter until now, as everything has been orthogonal, this
the swapped matrix entries were always 0.
If the texture count is lower than 4, entries in va.textcoord[] will
remain uninitialized. While this is unlikely to be a problem (since
these values are unused on the shader side too), it's not nice and might
explain some things which have shown up in valgrind.
Fix by always initializing the whole thing.
Instead of reallocating almost all of the shader string several times
per pass, build it into a fixed buffer that will be reallocated as
needed.
While this still uses a linear search and full comparison of the shader
text, this will compare the shader's string length first before doing a
full comparison as a nice side effect. (That's also why the fragment
shader is compared first - it's more likely to be different for
different cache entries than the vertex shader stub.)
Glitches when resizing are still possible, but are reduced. Other VOs
could support this too, but don't need to do so.
(Totally avoiding glitches would be much more effort, and probably not
worth the trouble. How about you just watch the video the player is
playing, instead of spending your time resizing the window.)
Until now, we have tried to create a GL 3.0 context. The main reason for
this is that many Mesa-based drivers did not support anything better.
But some drivers (Mesa AMD) will not report a higher OpenGL version,
because their compatibility mode is restricted. While later GL features
are reported as extensions just fine, there doesn't seem to be a way to
determine or enable higher GLSL versions.
Add some more shitty hacks to try to deal with this messed up situation,
and try to probe each interesting GL version separately (starting with
3.3, then 3.2 etc.). Other backends might suffer from similar problems,
but these will have to deal with it on their own.
Probably fixes#2938, or maybe not.
Prevents an infinite loop of configure event and set_fullscreen
request on Weston and other compositors respecting the protocol.
Fixes#2817
This reverts commit eb6b2b6e50.
converted_imgfmt will be used by the renderer logic to build an
appropriate shader chain. It doesn't influence the format of any
textures. Thus it doesn't matter whether the hw video surface is mapped
as RGB or RGBA. What matters is if the video actually contains alpha or
not. Since virtually all hardware decoder do not support alpha in any
way, this can be hardcoded as "no alpha".
This avoids unnecessary GPU work.
This also gets rid of the kind of hard to read texture swizzle setup and
turns it into something dumber.
Assumes that we don't create any FBOs with 2 channel formats. (Only the
video source textures are handled by this commit.)
Previously, gl->DXOpenDeviceNV was called twice using dxva2 with dxinterop. AMD
drivers refused to allow this. With this commit, context_dxinterop sets its own
implementation of MPGetNativeDisplay, which can return either a
IDirect3DDevice9Ex or a dxinterop_device_HANDLE depending on the "name" request
string. hwdec_dxva2gldx then requests both of these avoiding the need to call
gl->DXOpenDeviceNV a second time.
Drag&drop mechanisms typically support multiple types for the drop data.
Move most of the logic which types are accepted and preferred to
event.c, where the data is also interpreted.
(Maybe sorting the types by assigning scores is over-engineered, since
they're already sorted by preference, but it's actually not much more
code.)
Not very interesting/meaningful yet, but preparation for the next
commit.
Reduces VO access and makes the code more self-contained. (One day the
windowing backend code should not access the VO anymore. We're just not
quite there yet.)
Like dxinterop, this uses StretchRect or RGB conversion. This is unavoidable as
long as we use the dxva2 API, as there is no way to access the raw hardware
decoded Direct3D9 surfaces.
The default of 1.0 was basically making half the algorithm do nothing,
since it turned off all diagonal contributions. The upstream default is
0.6, and this produces a more reasonable image.
The values were changed to reflect an upstream change in the source for
the super-xBR implementation.
The anti-ringing code was basically not working at all, the new
algorithm _significantly_ improves the result (reduces ringing).
This is a fresh implementation from scratch that carries with it
significantly less baggage and verbosity from the previous (ported)
version.
The actual values for the masks and such were copied from the
current code. Behavior and performance should be unaffected.
An important difference between the old code and the new code is that
the new code always explicitly samples from the first component, rather
than being able to process multiple planes at once.
Since prescale-luma only affects luma, I deemed this unnecessary. May
change in the future, if prescale-chroma ever gets implemented. But
prescaling multiple planes would be slow to do this way. (Better would
be to generalize it to differently-sized vectors)
Instead of hard-coding the logic and planes to skip, factor this out
to a reusible function, and instead add the number of relevant
coordinates to the texture state.
Since prescale now literally only affects the luma plane (and the
filters are all designed for luma-only operation either way), the option
has been renamed and the documentation updated to clarify this.
This is a pretty major rewrite of the internal texture binding
mechanic, which makes it more flexible.
In general, the difference between the old and current approaches is
that now, all texture description is held in a struct img_tex and only
explicitly bound with pass_bind. (Once bound, a texture unit is assumed
to be set in stone and no longer tied to the img_tex)
This approach makes the code inside pass_read_video significantly more
flexible and cuts down on the number of weird special cases and
spaghetti logic.
It also has some improvements, e.g. cutting down greatly on the number
of unnecessary conversion passes inside pass_read_video (which was
previously mostly done to cope with the fact that the alternative would
have resulted in a combinatorial explosion of code complexity).
Some other notable changes (and potential improvements):
- texture expansion is now *always* handled in pass_read_video, and the
colormatrix never does this anymore. (Which means the code could
probably be removed from the colormatrix generation logic, modulo some
other VOs)
- struct fbo_tex now stores both its "physical" and "logical"
(configured) size, which cuts down on the amount of width/height
baggage on some function calls
- vo_opengl can now technically support textures with different bit
depths (e.g. 10 bit luma, 8 bit chroma) - but the APIs it queries
inside img_format.c doesn't export this (nor does ffmpeg support it,
really) so the status quo of using the same tex_mul for all planes is
kept.
- dumb_mode is now only needed because of the indirect_fbo being in the
main rendering pipeline. If we reintroduce p->use_indirect and thread
a transform through the entire program this could be skipped where
unnecessary, allowing for the removal of dumb_mode. But I'm not sure
how to do this in a clean way. (Which is part of why it got introduced
to begin with)
- It would be trivial to resurrect source-shader now (it would just be
one extra 'if' inside pass_read_video).
Why was this done so stupidly, with so many complicated special cases,
before? Declare it once so the shader bits don't have to figure out where
and when to do so themselves.
The WGL_NV_DX_interop spec says that a shared IDirect3DSurface9 must not
be lockable, but off-screen plain surfaces are always lockable and using
them causes Nvidia drivers to crash. Use a rendertarget for the shared
surface instead.
This also changes the name of the DX_interop handle for the rendertarget
to match the name of the DirectX object (rather than the GL one) to
match the convention used in context_dxinterop.c.
Apple crap (namely hardware decoding interop) forces us to use rectangle
textures for input. But after that we continue with normal textures.
This was not considered for debanding, and the sampler type used for it
can be different depending on the exact render chain. Simply use the
target type of the input texture.
* use mp_HRESULT_to_str/mp_LastError_to_str
* make some messages non-identical
* replace "GL" -> "OpenGL"
* change some MP_FATAL to MP_ERR that don't actually kill the vo
A client API user is allowed to call mpv_opengl_cb_uninit_gl() followed
by mpv_opengl_cb_init_gl(). This crashed; fix it by fixing the lifetime
of ctx->gl.
It thinks that integer_conv_fbo[index] is implied to be accessed with up
to index=5. Although that is theoretical only, it has a point that this
makes no sense. Use the same constant for the array allocation, to make
it more uniform and robust.
Fixes CID 1350060.
The ctx->redrawing field signals whether flip_page() should block. Do
not block if a black frame (i.e. nothing) is to be rendered.
Also, frame==NULL can never happen.
This makes it possible to set video size and position using the
--geometry and/or --autofit options. It's also possible to switch
between fullscreen/non-fullscreen playback during runtime.
This is the "unicode" version of it. It appears Firefox uses it now?
I'm not sure if we still need to support the old variant, but hopefully
not.
Fixes#2782.
Since there can be multiple backends for a single API (vaapi can use GLX
or EGL), not logging the exact backend name is annoying. So add it. At
the same time, there is no need to duplicate the name as used by the
--hwdec options, so replace it with using the numeric hwdec API ID.
GLES requires this. Some more common sampler types have default
precisions, but not usampler2D. Newer ANGLE builds verify this more
strictly than older builds, so this wasn't caught before.
Fixes#2761.
GLES does not support high bit depth fixed point textures for unknown
reasons, so direct 10 bit input is not possible. But we can still use
integer textures, which are supported by GLES 3.0. These store integer
data just like the standard fixed point textures, except they are not
normalized on sampling. They also don't support bilinear filtering, and
require a special sampler ("usampler2D").
While these texture formats enable us to shuffle the data to the GPU,
they're rather impractical with the requirements mentioned above and our
current architecture. One problem is that most code assumes it can
always use bilinear scaling (even if bilinear is never used when using
appropriate scale/cscale options). Another is that we don't have any
concept of running a function on a texture in an uniform way.
So for now, run a simple conversion step through a FBO. The FBO will use
the rgba16f format normally, which gives enough bits for 10 bit, and
will at least gracefully degrade with higher depth input.
This is bound to be much slower than a more "direct" method, but at
least it works and is simple to implement.
The odd change of function call order in init_video() is to properly
disable "dumb mode" (no FBO use) if these texture formats are in use.
This was never reset - absolutely can't be right. If the renderer
somehow switches back to another codepath, it certainly has to be reset.
Maybe this was hard to hit, as the normalization is going to be
idempotent in simpler cases (like rendering RGBA input).
Also get rid of the "merged" variable.
Often requested. The main argument, that prominent scalers like sharpen
change the image even if no scaling happens, disappeared anyway.
("sharpen", unsharp masking, is neither prominent nor a scaler anymore.
This is an artifact from MPlayer, which fuses unsharp masking with
bilinear scaling in order to make it single-pass, or so.)
Some VOs had support for these - remove them.
Typically, these formats will have only some use in cases where using
RGB software conversion with libswscale is faster than letting the
VO/GPU do it (i.e. almost never). For the sake of testing this case,
keep IMGFMT_RGB565. This is the least messy format, because it has no
padding/alpha bits with unknown semantics.
Note that decoding to these formats still works. We'll let libswscale
repack the data to whatever the VO in use can take.
Only request the current screen configuration instead of polling for new
screens, too. We're not interested in detecting any new screens as we're
merely enumerating what is currently connected and configured.
On some hardware (like mine) calling XRRGetScreenResources will stall
X11 for about 10 to 20 seconds. This has annoyed me for a few months
now and almost made me switch to VLC ;)
Signed-off-by: wm4 <wm4@nowhere>
Do this to make the license situation less confusing.
This change should be of no consequence, since LGPL is compatible with
GPL anyway, and making it LGPL-only does not restrict the use with GPL
code.
Additionally, the wording implies that this is allowed, and that we can
just remove the GPL part.
This covers source files which were added in mplayer2 and mpv times
only, and where all code is covered by LGPL relicensing agreements.
There are probably more files to which this applies, but I'm being
conservative here.
A file named ao_sdl.c exists in MPlayer too, but the mpv one is a
complete rewrite, and was added some time after the original ao_sdl.c
was removed. The same applies to vo_sdl.c, for which the SDL2 API is
radically different in addition (MPlayer supports SDL 1.2 only).
common.c contains only code written by me. But common.h is a strange
case: although it originally was named mp_common.h and exists in MPlayer
too, by now it contains only definitions written by uau and me. The
exceptions are the CONTROL_ defines - thus not changing the license of
common.h yet.
codec_tags.c contained once large tables generated from MPlayer's
codecs.conf, but all of these tables were removed.
From demux_playlist.c I'm removing a code fragment from someone who was
not asked; this probably could be done later (see commit 15dccc37).
misc.c is a bit complicated to reason about (it was split off mplayer.c
and thus contains random functions out of this file), but actually all
functions have been added post-MPlayer. Except get_relative_time(),
which was written by uau, but looks similar to 3 different versions of
something similar in each of the Unix/win32/OSX timer source files. I'm
not sure what that means in regards to copyright, so I've just moved it
into another still-GPL source file for now.
screenshot.c once had some minor parts of MPlayer's vf_screenshot.c, but
they're all gone.
Should take care of the planned FFmpeg AV_PIX_FMT_P010 addition. (This
will eventually be needed when doing HEVC Main 10 decoding with DXVA2
copyback.)
Commit 3909e4cd ended up losing the ability to tune the gaussian window,
which this commit trivially reintroduces.
The constant scaling factor (present in the code copied from glumpy)
also goes against filter_kernels.c conventions, which is that f(0.0) = 1
(and the invoking code takes care of normalization), and has been
removed.
The values of this new gaussian function corresponds to different
functions when compared against the old version. To translate the old
values p1 to the new values p2 requires solving 2^(e/p1) = e^(2/p2) or
p2 = p1 * 2/(e * ln(2)) ≈ p1 * 1.0615
In other words, to get the old default in the new function requires
setting scale-param1 to 1.0615. (The new function is *slightly* sharper
by default)
(Though most users should probably not notice the change)
To get a uniform license for this file, relicense the mpv parts to BSD
as well.
But leave the door open for a later change to LGPL. (All non-Glumpy code
was written within mpv, and all mpv authors have agreed to LGPL
relicensing.)
Closes#2688.
This file claims to be based on the "MPlayer VA-API patch", but this is
untrue. Only some glue code was copied from hwdec_vaglx.c, and this glue
code was never in MPlayer or the MPlayer VA-API patch in any form, and
instead part of the mpv-original way we do hardware decoding OpenGL
interop. The EGL interop method didn't exist at the time the MPlayer
VA-API patch was created either.
This commit replaces code based on AGG, taken from this source file:
http://vector-agg.cvs.sourceforge.net/viewvc/vector-agg/agg-2.5/include/agg_image_filters.h
The intention is that filter_kernels.c can be relicensed to LGPL or BSD.
Because the AGG author died, full replacement is the only way to achieve
it.
This affects only some filter functions. These are exclusively
mathematical functions for computing filter coefficients. (Other parts
in filter_kernel.c were originally written by me, with heavy additions
and refactoring done by other mpv contributors.) While the code is
mostly just well-known mathematical formulas written down in C form,
AGG copyright could perhaps be claimed anyway.
To remove the AGG code, I replaced it with the filter functions from:
https://github.com/glumpy/glumpy/blob/master/glumpy/library/build-spatial-filters.py
These functions conveniently compute exactly the same thing in mpv,
Glumpy, AGG (and about anything that will filter images using the same
mathematical principles).
First I ported the Python code in the file to C. Then I replaced all
functions in filter_kernels.c with this code that could be replaced.
Then I investigated whether the remaining functions were based on AGG
code and took appropriate action:
hanning(), hamming(), quadric(), bicubic(), kaiser(), blackman(),
spline16(), spline36(), gaussian(), sinc() were taken straight from
Glumpy.
For sinc(), re-add the "fabs(x) < 1e-8" check, which was added in commit
586dc557 for unknown reasons.
gaussian() loses its filter parameter for some reason. (Well, who cares,
not my problem.)
The really awkward thing is that the text for hanning() and hamming()
does not change. In theory these functions are now based on Glumpy code,
but it seems like this can be neither proven nor denied. (The same
happened in some other cases with at least a few lines of code.)
sphinx() was added in commit 586dc557, and looks suspiciously like
sinc() as well. Replace the first 3 lines of the body with the ported
function (of which 2 lines do not change; the first uses code only in
mpv, and the second is just "return 1.0;"). The 4th line is only similar
on an abstract level (and that because of the mathematical relation
between these functions). Although the original sinc() was probably used
as template for it, with the other lines replaced, I don't think you
could make the claim that it falls under AGG copyright.
jinc() was added in commit 26baf5b9, but the code for it might be based
on sinc(). Rewrite it based on the "new" sinc(). Some of the same
remarks as with sphinx() apply.
cubic_bc() was ported from Glumpy's Mitchell(). (As far as I'm aware,
with the default parameters it's called "the" Mitchell-Netravali filter,
but in mpv this function is used to generate a whole group of filters.)
spline64() was added in commit a8b67c66, and was probably derived from
spline36(). Re-derive it from the "new" spline36().
triangle() could be considered derived from the original bilinear().
This is this in the original commit:
static double bilinear(kernel *k, double x)
{
return 1.0 - x;
}
This _might_ be based on AGG's image_filter_bilinear:
struct image_filter_bilinear
{
static double radius() { return 1.0; }
static double calc_weight(double x)
{
return 1.0 - x;
}
};
Considering that the "framework" was written by me, and the only part
from AGG taken is "return 1.0 - x;", and this part is trivial and was
later thoroughly replaced, this is probably not under the AGG copyright.
I'm hoping this doesn't introduce regressions. But the main focus is not
being productive anyway, and I didn't rigorously check unintended
changes in functionality.
Apparently, the firmware will ignore pixel_x/pixel_y if the numeric
value of them gets too high (even if they indicate square pixel aspect
ratio). Even worse, the destination rectangle is ignored completely,
and the video frame is simply stretched to the screen. I suspect this
is an overflow or weird sanity check within the firmware.
Work it around by limiting the fields to 16000, which is an arbitrary
but apparently working limit.
GLSL in GLES 2.0 did not have line continuation in its preprocessor.
This broke shader compilation. It also broke subtitle rendering in
vo_rpi, which reuses some of the OpenGL code.
Line continuation was finally added in GLES 3.0, which is perhaps the
reason why ANGLE accepted it.
Untested, but should be fine. Broken by commit 0a0bb905.
Also fix the include statement in context_rpi.c, which caused another
compilation failure. Also untested. (Because I'm lazy.)
Fixes#2638.
gcc 4.8 does not support C11 thread local storage. This is a bit
annoying, so add a hack to use the gcc specific __thread extension if
C11 TLS is not available.
(This is used for the extremely silly mpv-internal way hwdec modules
access some platform specific handles. Disabling it simply made
hwdec_vaegl.c always fail initialization.)
Fixes#2631.
Add a "blend-tiles" choice to the "alpha" sub-option. This is pretty
simplistic and uses the GL raster position to derive the tiles. A weird
consequence is that using --vo=opengl and --vo=opengl-hq gives different
scaling behavior (screenspace pixel size vs. source video pixel size
16x16 tiles), but it seems we don't have easy access to the original
texture coordinates. Using the rasterpos is probably simpler.
Make this option the default.
long is 64 bits on x86_64 on Linux, which means the check for the corner
case of computing the depth mask is wrong.
Also, X11 compositors seem to expect premultiplied alpha.
MPlayer traditionally always used the display aspect ratio, e.g. 16:9,
while FFmpeg uses the sample (aka pixel) aspect ratio.
Both have a bunch of advantages and disadvantages. Actually, it seems
using sample aspect ratio is generally nicer. The main reason for the
change is making mpv closer to how FFmpeg works in order to make life
easier. It's also nice that everything uses integer fractions instead
of floats now (except --video-aspect option/property).
Note that there is at least 1 user-visible change: vf_dsize now does
not set the display size, only the display aspect ratio. This is
because the image_params d_w/d_h fields did not just set the display
aspect, but also the size (except in encoding mode).
Since alpha isn't pulled through the colormatrix (maybe it should?), we
reject alpha formats with odd sizes, such as yuva444p10.
But the awful tex_mul path in vo_opengl does this anyway (at some points
even explicitly), which means there will be a subtle difference in
handling of 16 bit yuv alpha formats. Make it consistent and always
apply the range adjustment to the alpha component. This also means odd
sizes like 10 bit are supported now.
This assumes alpha uses the same "shifted" range as the yuv color
channels for depths larger than 8 bit. I'm not sure whether this is
actually the case.
Now common.c only contains the code for the function loader, while
context.c contains the backend loader/dispatcher.
Not calling it "backend.c", because the central struct is called
MPGLContext.
This is used for dithering, although I'm not aware of anyone who got
higher than 8 bit depth support to work on Linux.
Also put this into egl_helpers.c. Since EGL is pseudo-portable at best I
have no hope that the EGL context creation code in all the backends can
be fully shared. But some self-contained functionality can definitely be
shared.
Store the determined framebuffer depth in struct GL instead of
MPGLContext. This means gl_video_set_output_depth() can be removed, and
also justifies adding new fields describing framebuffer/backend
properties to struct GL instead of having to add more functions just to
shovel the information around.
Keep in mind that mpgl_load_functions() will wipe struct GL, so the
new fields must be set before calling it.
Although the source file is named w32.c, the backend name was "win"
until recently. It was accidentally changed to "w32"; fix it.
Fixes#2608 (the manual is correct).
When a Direct3D 9Ex device fails to reset, it gets put into the lost
state, so set the lost_device flag and don't attempt to present until
the device moves out of that state. Failure to recreate the size-
dependent objects should set lost_device as well, since we shouldn't try
to present in that state.
Also, it looks like I was too eager to remove code that sets priv
members to NULL and I accidentally removed some that was needed.
Direct3D doesn't like 0-sized swapchain dimensions, even when those
dimensions are automatically set. Manually set them to a size that isn't
zero instead.
Why not.
Also, instead of disabling hue/saturation for RGB, just don't apply
them. (They don't make sense for conversion matrixes other than YUV, but
I can't be bothered to keep the fine-grained disabling of UI controls
either.)
WGL_NV_DX_interop is widely supported by Nvidia and AMD drivers. It
allows a texture to be shared between Direct3D and WGL, so that
rendering can be done with WGL and presentation can be done with
Direct3D. This should allow us to work around some persistent WGL
issues, such as dropped frames with some driver/OS combos, drivers that
buffer frames to increase performance at the cost of latency, and the
inability to disable exclusive fullscreen mode when using WGL to render
to a fullscreen window.
The addition of a DX_interop backend might also enable some cool
Direct3D-specific enhancements in the future, such as using the
GetPresentStatistics API to get accurate frame presentation timestamps.
Note that due to a driver bug, this backend is currently broken on
Intel. It will appear to work as long as the window is not resized too
often, but after a few changes of size it will be unable to share the
newly created renderbuffer with GL. See:
https://software.intel.com/en-us/forums/graphics-driver-bug-reporting/topic/562051
With default setting, the matrix for fruit dithering requires 12 bits
precision (values from 0/4096 to 4095/4096). But 16-bit float
provides only 10 bits. In addition, when `dither-size-fruit=8` is
set, 16 bits are required from the texture format.
Fix this by attempting to use 16 bit integer texture first. This is
still not precise, but should be better than using a half float.
The recent LUT adjustment changes broke interpolation.
The concatenation of the shader stages is a bit messy, and it seems like
sampler_prelude is not a good place to add this macro. Always add the
macro to every shader instead. (While this doesn't seem too elegant,
this isn't too inelegant either, and goes these problems out of the
way.)
The computation of the tex_mul variable was broken in multiple ways.
This variable is used e.g. by debanding for moving expansion of 10 bit
fixed-point input to normalized range to another stage of processing.
One obvious bug was that the rgb555 pixel format was broken. This format
has component_bits=5, but obviously it's already sampled in normalized
range, and does not need expansion. The tex_mul-free code path avoids
this by not using the colormatrix. (The code was originally designed to
work around dealing with the generally complicated pixel formats by only
using the colormatrix in the YUV case.)
Another possible bug was with 10 bit input. It expanded the input by
bringing the [0,2^10) range to [0,1], and then treating the expanded
input as 16 bit input. I didn't bother to check what this actually
computed, but it's somewhat likely it was wrong anyway. Now it uses
mp_get_csp_mul(), and disables expansion when computing the YUV matrix.
It turns out that with accurate lookup we can decrease the
default size of texture now. Do it to compensate the performance
loss introduced by the LUT_POS macro.
Define a macro to correct the coordinate for lookup texture. Cache
the corrected coordinate for 1D filter and use mix() to minimize the
performance impact.
It always was a weird artifact - VOCTRLs are meant _not_ to require
special handling in the code that passes them through (like in vo.c).
Removing it is also interesting to further reduce the dependency of
backends on struct vo. Just get rid of it.
Removing it is somewhat inconvenient, because in many situations the UI
window is created after the first VOCTRL_UPDATE_WINDOW_TITLE. This means
these backends have to store it in a new field in their own context.
If the sampling point is placed diagonally, the radius difference
could be as large as sqrt(2.0). And a loosened check with (radius - 1)
would potentially include pixels out of the range.
Fix the check to handle those corner case properly to avoid
unnecessary texture lookup and improve the performance a bit.
There are claims that nnedi3.c doesn't constitute its own new
implementation, but is derived from existing HLSL or OpenCL shaders
distributed under the LGPLv3 license.
Until these are resolved, do the "correct" thing and require
--enable-gpl3 to build nnedi.
I guess gl_video->global was originally meant to be optional, but now it
crashes in some newer code with vo_opengl_cb, which tries to init it
with this field set to NULL (because normally it's not needed).
Probably fixes#2542.
Don't use the average FPS if there are likely skipped vsyncs. Note that
we don't use the normal skip detection, as it is unreliable if the real
and assumed display FPS differ too much. The normal skip detection is
still in place as it's more reliable in the case when vsync jitters
much, but the display FPS is relatively exact.
Further improvement over commit 41f2c653.
At least I hope so.
Deriving the duration from the pts was not really correct. It doesn't
include speed adjustments, and becomes completely wrong of the user e.g.
changes the playback speed by a huge amount. Pass through the accurate
duration value by adding a new vo_frame field.
The value for vsync_offset was not correct either. We don't need the
error for the next frame, but the error for the current one. This wasn't
noticed because it makes no difference in symmetric cases, like 24 fps
on 60 Hz.
I'm still not entirely confident in the correctness of this, but it sure
is an improvement.
Also, remove the MP_STATS() calls - they're not really useful to debug
anything anymore.
This was just converting back and forth between int64_t/microseconds and
double/seconds. Remove this stupidity. The pts/duration fields are still
in microseconds, but they have no meaning in the display-sync case (also
drop printing the pts field from opengl/video.c - it's always 0).
If we switched away from the system FPS, we were remaining in this mode
ssentially forever. There's no reason to do so; switch back if the
estimated FPS gets worse again. Improvement over commit 41f2c653.
Commit 12eb8b2d accidentally disabled framedropping in the audio timing
case. It tried to replace the last_flip field with the prev_vsync one,
which didn't work because prev_sync is reset to 0 if the timing code is
used. Fix it by always setting it properly. This field must (or should)
be reinitialized to something sensible when switching to display sync
timing mode; since prev_vsync is not reset anymore, the check when to
reinitialize this field has to be adjusted as well.
It's a bit weird that update_vsync_timing_after_swap() now does some
minor work for timing mode too, but I guess it's ok, if only to avoid
additional fields and timer calls.
If the system-reported display FPS (returned by the VO backends, or
forced with --display-fps) is too imprecise (deviating frame duration by
more than 1%). This works if the display FPS is off by almost 1 (typical
for old/bad/broken OS APIs). Actually it even works if the FPs is
completely wrong.
Is it a good idea? Probably not. It might be better to only output a
warning message. But unless there are reports about it going terribly
wrong, I'll go with this for now.
Actually I'm not content with the detection added in commit 44376d2d. It
triggers too often if vsync is very jittery. It's easy to avoid this: we
add more samples to the detection by reusing the drift computation loop.
If there's a significant step in the drift, we consider it a drop.
This adds basic support for ICC profiles. Per-monitor profiles are
supported. WCS profiles are not supported, but there is an API for
converting WCS profiles to ICC, so they might be supported in future.
I'm just not sure if anyone actually uses them.
Reloading the ICC profile when it's changed in the control panel is also
not supported. This might be possible by using the WCS APIs and watching
the registry for changes, but there is no official API for it, and as
far as I can tell, no other Windows programs can do it.
Return the estimated/ideal flip time to the timing logic (meaning
vo_get_delay() returns a smoothed out time). In addition to this add
some lame but working drift compensation. (Useful especially if the
display FPS is wrong by a factor such as 1.001.)
Also remove some older leftovers. The vsync_interval_approx and
last_flip fields are redundant or unneeded.
For the vo-delayed-frame-count property.
Slightly less dumb than the previous one (which was removed earlier),
but still pretty dumb. But this also seems to be relatively robust, even
with strong vsync jittering.
This logic was kind of questionable anyway, and --display-sync should
give much better results. (I would even go as far as saying that the
FPS-dependent framedrop code made things worse in some situations. Not
all, though.)
This is simply the average refresh rate. Including "bad" samples is
actually an advantage, because the property exists only for
informational purposes, and will reflect problems such as the driver
skipping a vsync.
Also export the standard deviation of the vsync frame duration
(normalized to the range 0-1) as vsync-jitter property.
The OSD takes up an entire fullscreen dispmanx layer. Although the GPU
should be able to handle it (possibly even without any disadvantages),
it'll still be useful for debugging performance issues.
This is a hack, but unfortunately the DwmGetCompositionTimingInfo
heuristic does not work in all cases (with multiple-monitors on Windows
8.1 and even with a single monitor in Windows 10.) See the comment in
mp_w32_is_in_exclusive_mode() for more details.
It should go without saying that if any better method of doing this
reveals itself, this hack should be dropped.
It doesn't have any real purpose anymore. Up until now, it was still
implemented by vo_wayland, but since we changed how the frame callbacks
work, even that appears to be pointless.
Originally, the plan was to somehow extend this mechanism to all
backends and to magically fix frame scheduling, but since we can't hope
for proper mechanisms even on wayland, this idea looks way less
interesting.
The D3D9 backend does not support GLES 3, which makes it pretty useless.
But it still might be a legitimate replacement of vo_direct3d.c on
Windows 7 machines.
Note that we could just use:
eglGetDisplay(EGL_D3D11_ELSE_D3D9_DISPLAY_ANGLE)
But for now I'll leave the old code. Maybe this can exclude use of
software rendering backends (EGL_PLATFORM_ANGLE_DEVICE_TYPE_WARP_ANGLE).
Since I'm not sure, I won't touch it.
Running mpv with default config will now pick up ANGLE by default. Since
some think ANGLE is still not good enough for hq features, extend the
"es" option to reject GLES backends, and add to to the opengl-hq preset.
One consequence is that mpv will by default use libswscale to convert
10 bit video to 8 bit, before it reaches the VO.
I decided that I actually can't stand how vo_opengl unnecessarily puts
the video through 3 shader stages (instead of 1). Thus, what was meant
to be a fallback for weak OpenGL implementations, the dumb-mode, now
becomes default if the user settings allow it.
The code required to check for the settings isn't so wild, so I guess
it's manageable. I still hope that one day, our rendering logic can
generate ideal shader stages for this case too.
Note that in theory, dumb-mode could be reenabled at runtime due to a
color management 3D LUT being set, so a separate dumb_mode field is
required. The dumb-mode option can't just be overwritten.
Unfortunately, color management can still not work, because no GLES
version specified so far support fixed-point 16 bit textures. Maybe
we could use integer textures, but these don't support filtering.
Using float textures would be another possibility.
Polar scalers use 1D textures, because they're slightly faster on some
GPUs than 2D textures. But 2D textures work too, so add support for
them.
Allows using these scalers with ANGLE.
Just like commit f9a2fc59. There are probably some more such cases.
The vec2 constructor calls are probably fine, but don't bother with
confusing inconsistencies.
While desktop GL's glTexImage2D() essentially accepts anything, GLES is
much stricter. The combination of allowed formats/types/internal formats
is exactly specified. The GLES 3.0.4 specification lists them in
table 3.2. (The ANGLE API validation code references this table.)
The table could probably be extended into a general declarative table
about GL formats covering other uses, but this would be a big
non-trivial project, so don't bother and accept a minor degree
of duplication with other tables.
Note that the format and type do (or should) not matter here, because
no image data is transferred to the GPU.
We don't only need float textures for advanced scaling - we also need
them to be filterable with GL_LINEAR. On GLES, this is not supported
until GLES 3.1, but some implementation expose them with extensions.
This makes advanced scaling sort-of work for GLES 3.0 (on ANGLE). It's
still not very advisable, as 8 bits might not be enough to avoid
debanding. (Ironically, the debanding filter can be enabled, and does
not raise any GL errors - but probably doesn't do anything useful.)
Turns out glGetTexLevelParameter, which is missing in ANGLE, is a
GLES3.1 function. Removing it from the list of core GLES3 functions
makes ANGLE work in GLES3 mode.
ANGLE is a GLES2 implementation for Windows that uses Direct3D 11 for
rendering, enabling vo_opengl to work on systems with poor OpenGL
drivers and bypassing some of the problems with native GL, such as VSync
in fullscreen mode.
Unfortunately, using GLES2 means that most of vo_opengl's advanced
features will not work, however ANGLE is under rapid development and
GLES3 support is supposed to be coming soon.
Something goes wrong somewhere. Don't bother, it's only needed for
compatibility with our absolute baseline (GL 2.1/GLES 2).
On the other hand, we can process nv12 formats just fine.
For the sake of vaapi interop, we want to use EGL, but on the other
hand, but because driver developers are full of shit, vdpau interop will
not work on EGL (even if the driver supports EGL). The latter happens
with both nvidia and AMD Mesa drivers.
Additionally, EGL vaapi interop support can apparently only detected at
runtime by actually using it. While hwdec_vaegl.c already does this, it
would require initializing libva on _every_ system, which will cause
libav to print an unpreventable bullshit message to the terminal.
Try to counter these huge loads of bullshit by adding more fucking
bullshit.
We want the following behavior:
- VO probed, backend probed: only accept non-sw, fail completely
otherwise
- VO forced, backend probed: use the first non-sw, or if none is found,
fall back to the first working sw backend
- VO probed, backend forced: (I don't care about this case)
- VO forced, backend forced: just use that backend
Also, on backend probe failure the vo->probed field was left in its old
state.
This adds support for the progress indicator taskbar extension
that was introduced with Windows 7 and Windows Server 2008 R2.
I don’t like this solution because it keeps its own state and
introduces another VOCTRL, but I couldn’t come up with anything
less messy.
closes#2399
In the display-sync, non-interpolation case, and if the display refresh
rate is higher than the video framerate, we duplicate display frames by
rendering exactly the same screen again. The redrawing is cached with a
FBO to speed up the repeat.
Use glBlitFramebuffer() instead of another shader pass. It should be
faster.
For some reason, post-process was run again on each display refresh.
Stop doing this, which should also be slightly faster. The only
disadvantage is that temporal dithering will be run only once per video
frame, but I can live with this.
One aspect is messy: clearing the background is done at the start on the
target framebuffer, so to avoid clearing twice and duplicating the code,
only copy the part of the framebuffer that contains the rendered video.
(Which also gets slightly messy - needs to compensate for coordinate
system flipping.)
Currently, vo.c will always continue to render the currently queued
frame, which sets last_flip, which in turn confuses vo_get_delay(),
which in turn will show a bogus A/V desync message on unpause. So just
reset it again on unpause.
I guess the removed code is an old leftover, and makes no sense anymore.
Should fix weird A/V diff dropouts when frames are being dropped with
display-sync.
If the player sends a frame with duration==0 to the VO, it can trivially
underrun. Don't panic, but keep the correct time.
Also, returning the absolute time from vo_get_next_frame_start_time()
just to turn it into a float with relative time was silly. Rename it and
make it return what the caller needs.
"Missed" implies the frame was dropped, but what really happens is that
the following frame will be shown later than intended (due to the
current frame skipping a vsync).
(As of this commit, this property is still inactive and always
returns 0. See git blame for details.)
Apparently Windows treats windows that use OpenGL, cover an entire
screen and have the WS_POPUP style set or are topmost windows as
exclusive fullscreen windows that bypass DWM and cannot be covered
by other windows.
This means we can’t use dwmflush in fullscreen mode, and it also
means that no other window can cover mpv, and it makes the screen
flicker when switching to fullscreen mode.
This can be avoided by not setting the WS_POPUP flag.
Users can still access the old behavior by enabling stay-on-top
(which IMO at least makes sense—now we just need to get dwmflush
autodetection right to avoid nasty surprises).
fixes#2177
This applies to unexpected freezes or deadlocks, not e.g. normal
framedrops. The verbose messages also might remind an API user if the
API usage is incorrect, such as not calling mpv_opengl_cb_draw() when a
redraw request was issued.
The nnedi3 prescaler requires a normalized range to work properly,
but the original implementation did the range normalization after
the first step of the first pass. This could lead to severe quality
degradation when debanding is not enabled for NNEDI3.
Fix this issue by passing `tex_mul` into the shader code.
Fixes#2464
vo_opengl_cb is a special case, because we somehow have to render video
asynchronously, all while "trusting" the API user to do it correctly.
This didn't quite work, and a while ago a compromise using a timeout to
prevent theoretically possible deadlocks was added.
Make it even more synchronous. Basically, go all the way, and
synchronize rendering between VO and user renderer thread to the
full extent possible.
This means the silly frame queue is dropped, and we event attempt to
synchronize the GL SwapBuffer call (via mpv_opengl_cb_report_flip()).
The changes introduced with commit dc33eb56 are effectively dropped. I
don't even remember if they mattered.
In the future, we might make all VOs fetch asynchronously from a frame
queue, which would mostly remove the differences between vo_opengl and
vo_opengl_cb, but this will take a while (if it will even be done).
Pick the correct GLSL version from the GL_SHADING_LANGUAGE_VERSION
string. Might be somewhat questionable, as we expect the minor version
number not to have leading 0s.
Should help with cases when the reported GLSL version is much higher
than the equivalent of the reported GL version. This problem was
observed in combination with GL_ARB_uniform_buffer_object, which
can't be used if the declared GLSL version is too low.
Simplifies some auto detection matters.
I _still_ don't want to remove the lazy loading mechanism, because it's
still slightly useful for filters using the hwdec APIs. My main
motivation for not always preloading them is actually that libva prints
random useless crap to the terminal with no way to prevent this.
Notes:
- Unfortunately the only way to talk to EGL from within DRM I could find
involves linking with GBM (generic buffer management for Mesa.)
Because of this, I'm pretty sure it won't work with proprietary NVidia
drivers, but then again, last time I checked NVidia didn't offer
proper screen resolution for VT.
- VT switching doesn't seem to work at all. It's worth mentioning that
using vo_drm before introduction of VT switcher had an anomaly where
user could switch to another VT and input text to it, while video
played on top of that VT. However, that isn't the case with drm_egl:
I can't switch to other VT during playback like this. This makes me
think that it's either a limitation coming from my firmware or from
EGL/KMS itself rather than a bug with my code. Nonetheless, I still
left (untestable) VT switching code in place, in case it's useful to
someone else.
- The mode_id, connector_id and device_path should be configurable for
power users and people who wish to watch videos on nonprimary screen.
Unfortunately I didn't see anything that would allow OpenGL backends
to register their own set of options. At the same time, adding them to
global namespace is pointless.
- A few dozens of lines could be shared with vo_drm (setting up VT
switching, most of code behind page flipping). I don't have any strong
opinion on this.
- Sometimes I get minor visual glitches. I'm not sure if there's a race
condition of some sort, unitialized variable (doubtful), or if it's
buggy driver. (I'm using integrated Intel HD Graphics 4400 with Mesa)
- .config and .control are very minimal.
Signed-off-by: wm4 <wm4@nowhere>
