create_window is really bad naming, because this function can be called
multiple times, while the name implies that it always creates a new
window. At least the name config_window is not actively misleading.
Allow the backend code to create a GL context on best effort basis,
instead of having to implement separate functions for each variation.
This means there's only a single create_window callback now. Also,
getFunctions() doesn't have the gl3 parameter anymore, which was
confusing and hard to explain.
create_window() tries to create a GL context of any version. The field
MPGLContext.requested_gl_version is taken as a hint whether a GL3 or a
legacy context is preferred. (This should be easy on all platforms.)
The cocoa part always assumes that GL 3 is always available on
OSX 10.7.0 and higher, and miserably fails if it's not. One could try
to put more effort into probing the context, but apparently this
situation never happens, so don't bother. (And even if, mpv should be
able to fall back to vo_corevideo.)
The X11 part doesn't change much, but moving these functions around
makes the diff bigger.
Note about some corner cases:
This doesn't handle CONTEXT_FORWARD_COMPATIBLE_BIT_ARB on OpenGL 3.0
correctly. This was the one thing getFunctions() actually needed the
gl3 parameter, and we just make sure we never use forward compatible
contexts on 3.0. It should work with any version above (e.g. 3.1, 3.2
and 3.3 should be fine). This is because the GL_ARB_compatibility
extension is specified for 3.1 and up only. There doesn't seem to be
any way to detect presence of legacy GL on 3.0 with a forward
compatible context. As a counter measure, remove the FORWARD_COMPATIBLE
flags from the win32 code. Maybe this will go wrong. (Should this
happen, the flag has the be added back, and the win32 will have to
explicitly check for GL 3.0 and add "GL_ARB_compatibility" to the extra
extension string.)
Note about GLX:
Probing GL versions by trying to create a context on an existing window
was (probably) not always possible. Old code used GLX 1.2 to create
legacy contexts, and it required code different from GLX 1.3 even before
creation of the X window (the problem was selections of the X Visual).
That's why there were two functions for window creation (create_window_old
and create_window_gl3). However, the legacy context creation code was
updated to GLX 1.3 in commit b3b20cc, so having different functions for
window creation is not needed anymore.
The backend sub-option for vo_opengl and vo_opengl-old accepted numeric
values (like -1, 0, ...) for compatibility with MPlayer. This was added
in mplayer2 times, and is not important anymore.
You can just use --wid=0 if you really want this.
This only worked/works for X11, and even then it might interact badly
with most desktop environments. All the option did was setting --wid to
0, and the property did nothing.
And support the PIX_FMT_MONOWHITE pixel format. (This is really weird:
unlike PIX_FMT_MONOBLACK, it uses white pixels. I have no idea why
libavcodec doesn't just convert the pixel format on the fly, instead of
bothering everyone with really special pixel formats.)
Recent changes to the OSD code made vo_caca crash when showing OSD.
Since this is a joke VO (== I'd rather not waste my time with it),
remove the OSD support. It wasn't that great anyway.
This fixes the issue that black borders (e.g. on fullscreen) are not
redrawn, even if OSD rendering changes these areas.
In theory, the code could try some clever things to determine whether
clearing the window is really necessary, but that's probably not worth
the trouble and won't bring any significant performance gain, or might
even make things slower (because the GPU can't discard the old
contents).
Also fix redrawing when changing panscan with OSD disabled.
Use the option parser instead of sscanf. Remove the parameter changing
the field dominance (it has been marked deprecated for ages). Add a new
suboption "enabled", which can be used to disable the filter by default,
until it's enabled at runtime:
mpv -vf yadif=enabled=no
For all suboptions, "flat" options were available by separating the
parent option and the sub option with ":", e.g. "--rawvideo:w=123". Drop
this syntax and use "-" as separator. This means even suboptions are
available as normal options now, e.g. "--rawvideo-w=123". The old syntax
doesn't work anymore.
Note that this is completely separate from actual suboptions. For
example, "-rawvideo w=123:h=123" still works. (Not that this syntax is
worth supporting, but it's needed anyway, for for other things like vf
and vo suboptions.)
As a consequence of this change, we also have to add new "no-" prefixed
options for flag suboptions, so that "--no-input-default-bindings"
works. ("--input-no-default-bindings" also works as a consequence of
allowing "-input no-default-bindings" - they are handled by the same
underlying option.)
For --input, always use the full syntax in the manpage. There exist
suboptions other than --input (like --tv, --rawvideo, etc.), but since
they might be handled differently in the future, don't touch these yet.
M_OPT_PREFIXED becomes the default, so remove it. As a minor unrelated
cleanup, get rid of M_OPT_MERGE too and use the OPT_SUBSTRUCT() macro in
some places.
Unrelated: remove the duplicated --tv:buffersize option, fix a typo in
changes.rst.
`--fs-screen` allows to decide what display to go fullscreen into. The
semantics of `--screen` changed and now it is only used to select the windowed
display when starting the application.
This is useful for people using mpv with an external TV. They will start
windowed on their laptop's screen and switch to fullscreen on the TV.
@wm4 worked on the x11 and w32 parts of the code. All is squashed in one
commit for history clarity.
Also some other cosmetic changes. And reformat the two remaining doxygen
comments.
Removing MSGLEN in x11_errorhandler() is technically not just a cosmetic
change, but the result is the same anyway.
The string is deallocated by the callee after initialization, so
fallback at runtime passes a deallocated string to libavcodec, which
results in random crashes. Regression introduced by commit 4d016a9.
Fixes#29. When a user used dead input keys (like the accent key), `mpv`
crashed because the code tried to access the 0 element of a characters array
(which was empty).
While I was closing this bug, I refactored some related conditionals to
make the code more readable.
Do this to reduce conflicts with <linux/input.h>, which contains some
conflicting defines.
This changes the meaning of MP_KEY_DOWN:
KEY_DOWN is renamed to MP_KEY_DOWN (cursor down key)
MP_KEY_DOWN is renamed to MP_KEY_STATE_DOWN (modifier for key down state)
Instead of putting codec header data into WAVEFORMATEX and
BITMAPINFOHEADER, pass it directly via AVCodecContext. To do this, we
add mp_copy_lav_codec_headers(), which copies the codec header data
from one AVCodecContext to another (originally, the plan was to use
avcodec_copy_context() for this, but it looks like this would turn
decoder initialization into an even worse mess).
Get rid of the silly CodecID <-> codec_tag mapping. This was originally
needed for codecs.conf: codec tags were used to identify codecs, but
libavformat didn't always return useful codec tags (different file
formats can have different, overlapping tag numbers). Since we don't
go through WAVEFORMATEX etc. and pass all header data directly via
AVCodecContext, we can be absolutely sure that the codec tag mapping is
not needed anymore.
Note that this also destroys the "standard" MPlayer method of exporting
codec header data. WAVEFORMATEX and BITMAPINFOHEADER made sure that
other non-libavcodec decoders could be initialized. However, all these
decoders have been removed, so this is just cruft full of old hacks that
are not needed anymore. There's still ad_spdif and ad_mpg123, bu neither
of these need codec header data. Should we ever add non-libavcodec
decoders, better data structures without the past hacks could be added
to export the headers.
Rearrange some code to make it easier readable. Remove some dead code,
and stop printing AVI headers in demux_lavf. (These are not actual AVI
headers, just for internal use.)
There should be no functional changes, other than reducing output in
verbose mode.
Use codec names instead of FourCCs to identify codecs. Rewrite how
codecs are selected and initialized. Now each decoder exports a list
of decoders (and the codec it supports) via add_decoders(). The order
matters, and the first decoder for a given decoder is preferred over
the other decoders. E.g. all ad_mpg123 decoders are preferred over
ad_lavc, because it comes first in the mpcodecs_ad_drivers array.
Likewise, decoders within ad_lavc that are enumerated first by
libavcodec (using av_codec_next()) are preferred. (This is actually
critical to select h264 software decoding by default instead of vdpau.
libavcodec and ffmpeg/avconv use the same method to select decoders by
default, so we hope this is sane.)
The codec names follow libavcodec's codec names as defined by
AVCodecDescriptor.name (see libavcodec/codec_desc.c). Some decoders
have names different from the canonical codec name. The AVCodecDescriptor
API is relatively new, so we need a compatibility layer for older
libavcodec versions for codec names that are referenced internally,
and which are different from the decoder name. (Add a configure check
for that, because checking versions is getting way too messy.)
demux/codec_tags.c is generated from the former codecs.conf (minus
"special" decoders like vdpau, and excluding the mappings that are the
same as the mappings libavformat's exported RIFF tables). It contains
all the mappings from FourCCs to codec name. This is needed for
demux_mkv, demux_mpg, demux_avi and demux_asf. demux_lavf will set the
codec as determined by libavformat, while the other demuxers have to do
this on their own, using the mp_set_audio/video_codec_from_tag()
functions. Note that the sh_audio/video->format members don't uniquely
identify the codec anymore, and sh->codec takes over this role.
Replace the --ac/--vc/--afm/--vfm with new --vd/--ad options, which
provide cover the functionality of the removed switched.
Note: there's no CODECS_FLAG_FLIP flag anymore. This means some obscure
container/video combinations (e.g. the sample Film_200_zygo_pro.mov)
are played flipped. ffplay/avplay doesn't handle this properly either,
so we don't care and blame ffmeg/libav instead.
OPT_MAKE_FLAGS() used to emit two options (one with "no" prefixed),
but that has been long removed by special casing flag options in the
option parser. OPT_FLAG_ON() used to imply that there's no "no-"
prefixed option, but this hasn't been the case for a while either.
(Conceptually, it has been replaced by OPT_FLAG_STORE().)
Remove OPT_FLAG_OFF, which was unused.
Normally, all flag options can be negated by prepending a "no-", for
example "--no-opt" becomes "--opt=no". Some flag options can't actually
be negated, so add a CONF_TYPE_STORE option type to disallow the "no-"
fallback.
Do the same for choice options. Remove the explicit "no-" prefixed
options, add "no" as choice.
Move the handling of automatic "no-" options from parser-mpcmd.c to
m_config.c, and use it in m_config_set_option/m_config_parse_option.
This makes these options available in the config file. It also
simplifies sub-option parsing, because it doesn't need to handle "no-"
anymore.
mpv -ao help and mpv -vo help shouldn't show the encoding outputs (named
"lavc" on both cases). Also make it impossible to select these manually
when not encoding.
Basically, move vo_opengl above the other VOs (except vo_vdpau). This
changes preferences on Windows and Linux.
Move vo_opengl_old further down and make it the last fallback (before
vo_x11).
vo_caca is crap (no pun intended), and should never be autoprobed.
Add more vf_stereo3d output formats.
Adds high quality green-magenta and yellow-blue dubois
anaglyph 3D output support.
Patch by thomas schorpp, thomas.schorpp gmail
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@35906 b3059339-0415-0410-9bf9-f77b7e298cf2
Conflicts:
video/filter/vf_stereo3d.c
Apply setLevel hacks for fullscreen switching with ontop active only in
fullscreen. This keeps the window correctly on top even when losing focus.
Fixes#21
I also reverted NSScreenSaverWindowLevel to NSNormalWindowLevel + 1, so that
the cmd-tab UI is visible
The main problem this commit addresses is when you switched spaced back and
forth with `--ontop` active the video window didn't recive focus again.
Turns out that setting a normal window level just before conceding focus to
other apps / spaces fixes the problem. I think Cocoa is misbehaving here, and
should probably file a radar.
I found 3 related improvements while fixing this:
* fullscreen_window_level is a dead body from an older implementation.
* Use NSScreenSaverWindowLevel for ontop. This should be a really high window
level. Definitely higher than NSNormalWindowLevel + 1.
* The window level was set correctly only when out of fullscreen.
Make the window resizing menu items calculate the new window size based on the
video size and not the current window size.
This only makes a difference when using `--autofit`.
This functionality looked smart but created problems with some kinds of
multi touch events. Moreover some events coming from the windows server – like
hovering a corner for window resize – didn't cause the player to wake up
immediately.
The "correct" non hacky way to implement async event polling with cocoa would
be having the vanilla cocoa event loop driving the player and setting up mpv's
terminal FDs as event sources for the cocoa event loop.
Fixes#20
Remove screenshot_force and associated logic. Always try to use the
screenshot video filter before trying taking screenshots with the VO,
which means that --vf=screenshot now takes the role of --vf=screenshot_force.
(To make this clear, not adding a video filter is still the recommended
way to take screenshots; we just change how VF screenshots are forced.)
Preferring VO over VF and having --vf=screenshot_force used to make
sense when not all VOs supported screenshots, and some VOs had somewhat
broken screenshots (like vo_xv taking screenshots with OSD in it). But
all these issues are fixed now, so just get rid of the cruft.
This commit makes `is_osx_version_at_least` cache the result of reading
`/System/Library/CoreServices/SystemVersion.plist`. Since that is a file
read operation it was bad to use this function frequently (i.e.: when
processing user events).
Remove `is_lion_or_above` (introduced in c9396c0a) as that was a more
specialized wrapper which had the only advantage of adding it's own cache.
Don't force VOs to pick an arbitrary default Visual and Colormap. They
still can override them if needed. This simplifies the X11 VO interface.
Always create a Colormap for simplicity. Using CopyFromParent fails if
the selected visual is not the same of that of the parent window, which
happens for me with vo_opengl.
vo_vdpau and vo_xv explicitly set CWBorderPixel, do that in x11_common
instead (it was already done for native windows, but not for slave mode
windows).
What gl_common did was incorrect in theory: freeing a colormap while a
window uses it will change the colormap of the window to "None", and
the color mapping for such windows is "undefined".
The original video mode wasn't reliably restored, so just store the
mode separately.
For some reason, window decorations got into the picture, possibly due
to an incorrect initial window position or something like this.
Normally, the window is positioned and sized such that it covers the
screen entirely, even though the window still has decorations and is
not in fullscreen mode (fullscreen wouldn't be correct, because the
virtual desktop size is not screen size). Hack-fix by forcing window
decorations off when VM-switching.
All in all, VM switching is still buggy and useless.
Some parts for initiating mode switches were duplicated in every VO
supporting X11 (except vo_opengl/gl_common, which didn't support mode
switching). Move this to x11_common.c.
Note that this might be slightly risky: is it really guaranteed that no
VO needed to do "special" setup that depends on X parameters changing
after a mode switch, such as bit depth, visuals etc.? From what I can
see, this shouldn't be the case (X probably can't even change depth on
the fly). Even if this should be a one-way road, VM switching is in
general very useless, and its implementation buggy, so it can just be
removed should unfixable problems arise.
Do this because we want to remove the global variables required with the
old code. In particular, there doesn't seem any way to set a secure user
data pointer with xlib.
XSelectInput() causes a BadAccess error when some of the requested event
flags are reserved (for exclusive flags like ButtonPress). The custom
error handler caught this and set a global variable, so that the code
could retry the XSelectInput() call without the conflicting flags. Use a
different approach that doesn't need a custom error handler. (Although
we still assume that the error handler doesn't terminate the program.)
Move things that are used by vo_xv only into vo_xv, same for vo_x11.
Rename some functions exported by x11_common, like vo_init to
vo_x11_common. Make functions not used outsode of x11_common.c private
to that file. Eliminate all global variables defined by x11_common
(except error handler and colormap stuff).
There shouldn't be any functional changes, and only code is moved
around. There are some minor simplifications in the X11 init code, as
we completely remove the ability to initialize X11 and X11+VO
separately (see commit b4d9647 "mplayer: do not create X11 state in player frontend"),
and the respective functions are conflated into vo_x11_init() and
vo_x11_uninit().
Dithering was disabled if the input bit depth was not larger than the
output bit depth of the screen framebuffer. But since scaling, RGB
conversion, and other filters change the number of significant bits
anyway, dithering could still benefit image quality even in these
cases. Always do dithering, unless dithering is completely disabled.
The original intention of this mechanism was not to change the image
needlessly when playing video that matches the native bit depth of the
screen.
The decoder returns with AVFrame.format not correctly set for some h264
files (strangely only some). We have to access AVCodecContext.pix_fmt
instead. On newer libavcodec versions, it's the other way around: the
AVCodecContext.pix_fmt may be incorrectly set on pixel format changes,
and you are supposed to use AVFrame.format.
The same problem probably exists on older ffmpeg versions too.
Exactly the same issue as with the previous commit. Just like the vdpau
code, this was apparently copy-pasted from the vo_x11 code, even though
it doesn't make much sense.
Using vdpau on an X server configured to a bit depth of 30 (10 bit per
component) failed finding a visual. The cause was a hack that tried to
normalize the bit depth to 24 if it was not a known depth. It's unknown
why/if this is needed, but the following things speak against it:
- it prevented unusual bit depths like 30 bit from working
- it wasn't needed with normal bit depth like 24 bit
- it's probably copy-pasted from vo_x11 (where this code possibly makes
sense, unlike in vo_vdpau)
Just remove this code and look for a visual with native depth.
-x/-y were rather useless and obscure. The only use I can see is
forcing a specific aspect ratio without having to calculate the aspect
ratio float value (although --aspect takes values of the form w:h).
This can be also done with --geometry and --no-keepaspect. There was
also a comment that -x/-y is useful for -vm, although I don't see how
this is useful as it still messes up aspect ratio.
-xy is mostly obsolete. It does two things: a) set the window width to
a pixel value, b) scale the window size by a factor. a) is already done
by --autofit (--autofit=num does exactly the same thing as --xy=num, if
num >= 8). b) is not all that useful, so we just drop that
functionality.
--autofit=WxH sets the window size to a maximum width and/or height,
without changing the window's aspect ratio.
--autofit-larger=WxH does the same, but only if the video size is
actually larger than the window size that would result when using
the --autofit=WxH option with the same arguments.
This also means the option is verified on program start, not when the VO
is created. The actual code becomes a bit more complex, because the
screen width/height is not available at program start.
The actual parsing code is still the same, with its unusual sscanf()
usage.
Now the calculations of the final display size are done after the filter
chain. This makes the difference between display aspect ratio and window
size a bit more clear, especially in the -xy case.
With an empty filter chain, the behavior of the options should be the
same, except that they don't affect vo_image and vo_lavc anymore.
Drop queued frames on seek. Reset the internal state of some filters
that seem to need it as well: at least vf_divtc still produced some
frames using the previous PTS.
This fixes weird behavior with some filters on seeking. In particular,
this could lead to A/V desync or apparent lockups due to the PTS of
filtered frames being too far away from audio PTS.
This commit does only the minimally required work to fix these PTS
related issues. Some filters have state dependent on previously filtered
frames, and these are not automatically reset with this commit (even
vf_divtc and vf_softpulldown reset the PTS info only). Filters that
actually require a full reset can implement VFCTRL_SEEK_RESET.
Options that take pixel format names now also accept ffmpeg names.
mpv internal names are preferred. We leave this undocumented
intentionally, and may be removed once libswscale stops printing
ffmpeg pixel format names to the terminal (or if we stop passing the
SWS_PRINT_INFO flag to it, which makes it print these).
(We insist on keeping the mpv specific names instead of dropping them
in favor of ffmpeg's name due to NIH, and also because ffmpeg always
appends the endian suffixes "le" and "be".)
<ctype.h> is needed at least for isalnum(). Most time this worked,
because some ffmpeg or Libav versions recursively include this header
from libavutil/common.h. Fix it so it always works.
This printed per-frame statistics into a file, like bitrate or frame
type. Not very useful and accesses obscure AVCodecContext fields
(danger of deprecation/breakage), so get rid of it.
This was a "broken misfeature" according to Libav developers. It wasn't
implemented for modern codecs (like h264), and has been removed from
Libav a while ago (the AVCodecContext field has been marked as
deprecated and its value is ignored). FFmpeg still supports it, but
isn't much useful due to aforementioned reasons.
Remove the code to enable it.
This was an awkward hack that attempted to avoid the use of 16 bit
textures, while still allowing rendering 10-16 bit YUV formats. The
idea was that even if the hardware doesn't support 16 bit textures,
an A8L8 textures could be used to convert 10 bit (etc.) to 8 bit in
the shader, instead of doing this on the CPU.
This was an experiment, disabled by default, and was (probably) rarely
used. I've never heard of this being used successfully. Remove it.
This is more correct. Not all frame specific fields are in AVFrame,
such as colorspace and color_range, and these are still queried
through the decoder context.
This field contained the "average" bit depth per pixel. It serves no
purpose anymore. Remove it.
Only vo_opengl_old still uses this in order to allocate a buffer that is
shared between all planes.
This used to mean that there is more than one plane. This is not very
useful: IMGFMT_Y8 was not considered planar, but it's just a Y plane,
and could be treated the same as actual planar formats. On the other
hand, IMGFMT_NV12 is partially packed, and usually requires special
handling, but was considered planar.
Change its meaning. Now the flag is set if the format has a separate
plane for each component. IMGFMT_Y8 is now planar, IMGFMT_NV12 is not.
As an odd special case, IMGFMT_MONO (1 bit per pixel) is like a planar
RGB format with a single plane.
In theory, vf_sub could take any format supported by swscale. But to be
sure that it's reasonably fast, only 420P was allowed. However, other
similar 8 bit planar formats will be just as fast and there's no reason
to exclude them. Even for completely different formats there doesn't
seem to be any significant advantage to force vf_sub to convert to a
simpler/more common format.
This did random things with some image formats, including 10 bit
formats. Fixes the mp_image_clear() function too.
This still has some caveats:
- doesn't clear alpha to opaque (too hard with packed RGB, and is rarely
needed)
- sets luma to 0 for mpeg-range YUV, instead of the lowest possible
value (e.g. 16 for 8 bit)
Needed because mplayer* basically tracks DAR, which makes it harder for
filters which want to keep the SAR.
The contents of this function are duplicated in a few filters, and will
be used instead of custom code as the filters are updated later.
This makes it behave like vo_lavc.
Unfortunately, the code for setting up the OSD dimensions (mp_osd_res)
is copied from vo_lavc, but it doesn't look like something that should
be factored out.
redraw_frame() copied the image into the currently visible buffer. This
resulted in flicker when doing heavy OSD redrawing (like changing the
subtitle size to something absurdly large).
Use the same logic as draw_image instead.
In order to support OSD redrawing for vo_xv and vo_x11, draw_bmp.c
included an awkward "backup" mechanism to copy and restore image
regions that have been changed by OSD/subtitles.
Replace this by a much simpler mechanism: keep a reference to the
original image, and use that to restore the Xv/X framebuffers.
In the worst case, this may increase cache pressure and memory usage,
even if no OSD or subtitles are rendered. In practice, it seems to be
always faster.
It's not easy to tell whether the OSD/subs are empty, or if something is
drawn. In general you have to use osd_draw() with a custom callback. If
nothing is visible, the callback is never invoked. (The actual reason
why this is so "hard" is the implementation of osd_libass.c, which
doesn't allow separating rendering and drawing of OSD elements, because
all OSD elements share the same ASS_Renderer.)
To simplify avoiding copies, make osd_draw_on_image() instead of the
caller use mp_image_make_writeable(). Introduce osd_draw_on_image_p(),
which works like osd_draw_on_image(), but gets the new image allocation
from an image pool. This is supposed to be an optimization, because it
reduces the frequency of large allocations/deallocations for image data.
The result of this is that the frequency of copies needed in conjunction
with vf_sub, screenshots, and vo_lavc (encoding) should be reduced.
vf_sub now always does true pass-through if no subs are shown.
Drop the pts check from vf_sub. This didn't make much sense.
Remove the strange things the old mp_image_setfmt() code did to the
image format parameters. This includes setting chroma shift to 31 for
gray (Y8) formats and more.
Y8 + vo_opengl_old didn't actually work for unknown reasons (regression
in this branch). Fix this. The difference is that Y8 is now interpreted
as gray RGB (LUMINANCE texture) instead of involving YUV (and levels)
conversion.
Get rid of distinguishing RGB and BGR. There wasn't really any good
reason for this.
Remove mp_get_chroma_shift() and IMGFMT_IS_YUVP16*(). mp_imgfmt_desc
gives the same information and more.
Even though #ifdef ACCURATE is removed, the result should be about the
same. The fallback is only used by packed YUV formats (YUYV, NV12), and
doing 16 bit for them instead of 8 bit is not useful.
A side effect is that Y8 (gray) is not converted drawing subs, and for
alpha formats, the alpha plane is not removed. This means the number of
planes after upsampling can be 1-4 (1: gray, 2: gray+alpha, 3: planar,
4: planar+alpha). The code has to be adjusted accordingly to work on the
color planes only. Also remove the workaround for the chroma shift 31
hack.
mplayer's video chain traditionally used FourCCs for pixel formats. For
example, it used IMGFMT_YV12 for 4:2:0 YUV, which was defined to the
string 'YV12' interpreted as unsigned int. Additionally, it used to
encode information into the numeric values of some formats. The RGB
formats had their bit depth and endian encoded into the least
significant byte. Extended planar formats (420P10 etc.) had chroma
shift, endian, and component bit depth encoded. (This has been removed
in recent commits.)
Replace the FourCC mess with a simple enum. Remove all the redundant
formats like YV12/I420/IYUV. Replace some image format names by
something more intuitive, most importantly IMGFMT_YV12 -> IMGFMT_420P.
Add img_fourcc.h, which contains the old IDs for code that actually uses
FourCCs. Change the way demuxers, that output raw video, identify the
video format: they set either MP_FOURCC_RAWVIDEO or MP_FOURCC_IMGFMT to
request the rawvideo decoder, and sh_video->imgfmt specifies the pixel
format. Like the previous hack, this is supposed to avoid the need for
a complete codecs.cfg entry per format, or other lookup tables. (Note
that the RGB raw video FourCCs mostly rely on ffmpeg's mappings for NUT
raw video, but this is still considered better than adding a raw video
decoder - even if trivial, it would be full of annoying lookup tables.)
The TV code has not been tested.
Some corrective changes regarding endian and other image format flags
creep in.
These functions weren't specific to video filters and were misplaced
in vf.c. Move them to mp_image.c.
Fix the big endian test in vf_mpi_clear/mp_image_clear, which has been
messed up in 74df1d.
According to DOCS/OUTDATED-tech/colorspaces.txt, the following formats
are supposed to be palettized:
IMGFMT_BGR8
IMGFMT_RGB8,
IMGFMT_BGR4_CHAR
IMGFMT_RGB4_CHAR
IMGFMT_BGR4
IMGFMT_RGB4
Of these, only BGR8 and RGB8 are actually treated as palettized in some
way. ffmpeg has only one palettized format (AV_PIX_FMT_PAL8), and
IMGFMT_BGR8 was inconsistently mapped to packed non-palettized RGB
formats too (AV_PIX_FMT_BGR8). Moreover, vf_scale.c contained messy
hacks to generate a palette when AV_PIX_FMT_BGR8 is output. (libswscale
does not support AV_PIX_FMT_PAL8 output in the first place.)
Get rid of all of this, and introduce IMGFMT_PAL8, which directly maps
to AV_PIX_FMT_PAL8. Remove the palette creation code from vf_scale.c.
IMGFMT_BGR8 maps to AV_PIX_FMT_RGB8 (don't ask me why it's swapped),
without any palette use. Enabling it in vo_x11 or using it as vf_scale
input seems to give correct results.
mp_image_alloc_planes() allocated images with minimal stride, even if
the resulting stride was unaligned. It was the responsibility of
vf_get_image() to set an image's width to something larger than
required to get an aligned stride, and then crop it. Always allocate
with aligned strides instead.
Get rid of IMGFMT_IF09 special handling. This format is not used
anymore. (IF09 has 4x4 chroma sub-sampling, and that is what it was
mainly used for - this is still supported.) Get rid of swapped chroma
plane allocation. This is not used anywhere, and VOs like vo_xv,
vo_direct3d and vo_sdl do their own swapping.
Always round chroma width/height up instead of down. Consider 4:2:0 and
an uneven image size. For luma, the size was left uneven, and the chroma
size was rounded down. This doesn't make sense, because chroma would be
missing for the bottom/right border.
Remove mp_image_new_empty() and mp_image_alloc_planes(), they were not
used anymore, except in draw_bmp.c. (It's still allowed to setup
mp_images manually, you just can't allocate image data with them
anymore - this is also done in draw_bmp.c.)
Replace the internal pixel format stuff with code that queries the
libavutil list of pixel format descriptors.
Trying to map IMGFMT_IS_RGB() etc. turned out extremely hacky.
Instead of using a callback to "capture" the image next time the filter
function is called, do it the other way around: on every filter
invocation, create a reference to the image, and return it if a
screenshot is requested. This also fixes the 1-frame delay when taking
screenshots with the filter.
This also allows simplifying screenshot.c.
VFCAP_TIMER disables any additional waiting done by mpv in the
playloop. Remove VFCAP_TIMER, but re-use the idea for vo_image and
vo_lavc.
This means --untimed doesn't have to be passed when using --vo=image.
This allows avoiding a copy.
The logic about using the AVFrame.reference field if buffer hints are
not available is similar to mplayer-svn's direct rendering code. This is
needed because h264 decoding doesn't set buffer hints.
<wm4> michaelni: couldn't it set buffer hints from AVFrame.reference then?
<michaelni> i see no reason ATM why that wouldnt be possible
OK...
Remove mp_image.width/height. The w/h members are the ones to use.
width/height were used internally by vf_get_image(), and sometimes for
other purposes.
Remove some image flags, most of which are now useless or completely
unused. This includes VFCAP_ACCEPT_STRIDE: the vf_expand insertion in
vf.c does nothing.
Remove some other unused mp_image fields.
Some rather messy changes in vo_opengl[_old] to get rid of legacy
mp_image flags and fields. This is left from when vo_gl supported DR.
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
The code was entirely correct, as the VO doesn't report VFCAP_ACCEPT_STRIDE
in query_format. Add stride capability in preparation for changing the
video chain: soon all VOs will have to support arbitrary strides.
The code assumed mp_image_alloc() would allocate an image large enough
for corevideo's stride, which doesn't have to be the case. If
corevideo's stride was larger than the stride of mp_image, the memcpy()
would write beyond the mp_image allocation.
This probably didn't actually happen, but fix the code to be more
correct anyway.
Setting the size of a mp_image must be done with mp_image_set_size()
now. Do this to guarantee that the redundant fields (like chroma_width)
are updated consistently. Replacing the redundant fields by function
calls would probably be better, but there are too many uses of them,
and is a bit less convenient.
Most code actually called mp_image_setfmt(), which did this as well.
This commit just makes things a bit more explicit.
Warning: the video filter chain still sets up mp_images manually,
and vf_get_image() is not updated.
Refcounting will conceptually allocate and free images all the time
when using the filter chain. Add a pool that makes these reallocations
cheap.
This only affects the image data, not mp_image structs and similar small
allocations. Small allocations are always fast with reasonable memory
managers, while large image data will trigger mmap/munmap calls each
time.
Note that if the codec doesn't support DR1, the image has to be copied.
There is no other way to guarantee that the image will be valid after
decoding the next image.
The only important codec that doesn't support DR1 yet is rawvideo. It's
likely that ffmpeg/Libav will fix this at some time. For now, this
decoder uses an evil hack and puts pointers to the packet data into the
returned frame. This means the image will actually get invalid as soon
as the corresponding video packet is free'd, so copying the image is the
only reasonable thing anyway.
Replace libavcodec's native buffer allocation with code taken from
ffplay/ffmpeg's libavfilter support. The code in lavc_dr1.c is directly
copied from cmdutils.c. Note that this is quite arcane code, which
contains some workarounds for decoder bugs and the like. This is not
really a maintainance burden, since fixes from ffmpeg can be directly
applied to the code in lavc_dr1.c.
It's unknown why libavcodec doesn't provide such a function directly.
avcodec_default_get_buffer() can't be reused for various reasons.
There's some hope that the work known as The Evil Plan [1] will make
custom get_buffer implementations unneeded.
The DR1 support as of this commit does nothing. A future commit will
use it to implement ref-counting for mp_image (similar to how AVFrame
will be ref-counted with The Evil Plan.)
[1] http://lists.libav.org/pipermail/libav-devel/2012-December/039781.html
Deprecate the hardware specific video codec entries (like ffh264vdpau).
Replace them with the --hwdec switch, which requests that a specific
hardware decoding API should be used. The codecs.conf entries will be
removed at a later time, but for now they are useful for testing and
compatibility.
Instead of --vc=ffh264vdpau, --hwdec=vdpau should be used.
Add a fallback if hardware decoding fails. Most hardware decoders
(including vdpau) support only a subset of h264, and having such a
fallback is supposed to enable a better user experience.
This was buggy and didn't even work in the simplest cases. It was
disabled when multithreading was used, and always disabled for h264.
A better alternative (reference counting) will be added later.
Hardware decoding still uses the ffmpeg DR mechanism, but has been
decoupled from mpv's DR in the previous commit.
vdpau hardware decoding used the DR (direct rendering) path to let the
decoder query a surface from the VO. Special-case the HW decoding path
instead, to make it separate from DR.
This mutated the variable for the thread count option
(lavc_param->threads) on decoder initialization. This didn't have any
practical relevance, unless formats supporting hardware video decoding
and other formats were played in the same mpv instance. In this case,
hardware decoding would set threads to 1, and all files played after
that would use only one thread as well even with software decoding.
Remove XvMC leftover (CODEC_CAP_HWACCEL).
Simplify the decoder pixel format handling by making it handle only
the case vd_lavc needs: a video stream always decodes to a single
pixel format.
Remove the handling for multiple pixel formats, and remove the
codecs.conf pixel format declarations that are left.
Remove the handling of "ambiguous" pixel formats like YV12 vs. I420 (via
VDCTRL_QUERY_FORMAT etc.). This is only a problem if the video chain
supports I420, but not YV12, which doesn't seem to be the case anywhere,
and in fact would not have any advantage.
Make the "flip" flag a global per-codec flag, rather than a pixel format
specific flag. (Some ffmpeg decoders still return a flipped image, so
this has to be done manually.) Also fix handling of the flip operation:
do not overwrite the global flip option, and make the --flip option
invert the codec flip option rather than overriding it.
Slices allowed filtering or drawing video in horizontal bands or
blocks. This allowed working on the video in smaller units. In theory,
this could bring a performance win by lowering cache pressure, as you
didn't have to keep the whole video frame in cache while filtering,
only the slice.
In practice, the slice code path was barely used for the following
reasons:
- Multithreaded decoding with ffmpeg didn't use slices. The ffmpeg
slice callback was disabled, because it can be called from another
thread, and the mplayer video chain is not thread-safe.
- There was nothing that would turn "full" images into appropriate
slices, so slices were rarely used.
- Most filters didn't actually support slices.
On the other hand, supporting slices lead to code duplication and more
complex code in general. I made some experiments and didn't find any
actual measurable performance improvements when using slices. Even
ffmpeg removed slices based filtering from libavfilter in favor of
simpler code.
The most broken thing about the slices code path is that slices can't
be queued, like it is done for images in vo.c.
For some reason, libavcodec abuses the slices rendering code path for
hardware decoding: in that case, the only purpose of the draw callback
is to pass a vdpau video surface object to video output. (It is unclear
to me why this had to use the slices code, instead of just returning an
AVFrame with the required vdpau state.)
Make this code separate within mpv, so that the internal slices code
path is not used for hardware decoding. Pass the vdpau state with
VOCTRL_HWDEC_DECODER_RENDER instead.
Remove the mencoder specific VOCTRLs.
Remove VOCTRL_DRAW_IMAGE and always set vo_driver.draw_image in VOs.
Make draw_image mandatory: change some VOs (like vo_x11) to support it,
and remove the image-to-slices fallback in vf_vo.
Remove vo_driver.is_new. This member indicated whether draw_image is
supported unconditionally, which is now always the case.
draw_image_pts is a hack until the video filter chain is changed to
include the PTS as field in mp_image. Then vo_vdpau and vo_lavc will
be changed to use draw_image.
The check determined whether the argument for .align is in bytes, or
log2(bytes). Apparently it's always in bytes for ELF i386 systems, and
this check is used for x86 inline assembler only. Even if this
assumption should be wrong, it likely won't cause much damage: the
existing code uses it only in the form ".align 4", which means in the
worst case it will try to align to 16 bytes, which doesn't cause any
problems (unless the object file format does not support such a high
alignment).
Update the filters that used this.
Quoting the GNU as manual:
For other systems, including ppc, i386 using a.out format, arm and
strongarm, it is the number of low-order zero bits the location counter
must have after advancement. For example `.align 3' advances the
location counter until it a multiple of 8. If the location counter is
already a multiple of 8, no change is needed.
The premultiplied-alpha hack is changed:
- The first stage now uses a colormod of black with an unmodified
texture. This saves on applying the AND mask of 0xFF000000 to keep
alpha only.
- The second stage no longer uses an AND mask, but only an OR mask of
0xFF000000 to cancel out alpha.
- The texture uploads are no longer done using SDL_LockTexture,
SDL_ConvertPixels, SDL_UnlockTexture when the mpv pixel format matches
the OSD's pixel format. Instead, SDL_UploadTexture is used, which
saves a copy when using the "opengl" renderer.
Strictly speaking, 444P is higher quality than YV12, but doing this is
not very useful when playing 10 bit video with -vo opengl-old on a GPU
that doesn't support 16 bit textures.
For 9-15 bit material, cutting off the lower bits leads to significant
quality reduction, because these formats leave the most significant bits
unused (e.g. 10 bit padded to 16 bit, transferred as 8 bit -> only
2 bits left). 16 bit formats still can be played like this, as cutting
the lower bits merely reduces quality in this case.
This problem was encountered with the following GPU/driver combination:
OpenGL vendor string: Intel Open Source Technology Center
OpenGL renderer string: Mesa DRI Intel(R) 915GM x86/MMX/SSE2
OpenGL version string: 1.4 Mesa 9.0.1
It appears 16 bit support is rather common on GPUs, so testing the
actual texture depth wasn't needed until now. (There are some other Mesa
GPU/driver combinations which support 16 bit only when using RG textures
instead of LUMINANCE_ALPHA. This is due to OpenGL driver bugs.)
The extension checking logic was broken, which reported OpenGL 3 if the
OpenGL .so exported OpenGL 3-only symbols, even if the reported OpenGL
version is below 3.0. Fix it and simplify the code a bit. Also never
fail hard if required functions are not found. The caller should check
the capability flags instead. Give up on the idea that we should print
a warning if essential functions are not found (makes loading of ancient
legacy-only extensions easier).
This was experienced with the following version strings:
OpenGL vendor string: Intel Open Source Technology Center
OpenGL renderer string: Mesa DRI Intel(R) 915GM x86/MMX/SSE2
OpenGL version string: 1.4 Mesa 9.0.1
(Possibly reports a very old version because it has no GLSL support,
and thus isn't even GL 2.0 compliant.)
Change from gamma 2.2 to the slightly more precise 1/0.45 as per BT.709.
https://www.itu.int/rec/R-REC-BT.709-5-200204-I/en mentions a value of
γ=0.45 for the conceptual non-linear precorrection of video signals.
This is approximately the inverse of 2.22, and not 2.20 as the code had
been using until now.
This mainly serves as a fallback for platforms where nothing better is
available; also as a debugging help. Both the audio and video driver are
not first class - the audio driver lacks delay detection, and the video
driver only supports a single YUV color space.
Configure options: --disable-sdl2 to disable SDL 2.0+ detection,
--disable-sdl to disable SDL 1.2+ detection. Both options need to be
specified to turn off SDL support entirely.
This is better than having just the operating system type decide the
wakeup period, as e.g. when compiling for Win32/cygwin, a wakeup period
of 0.5 would work perfectly fine.
Instead, the default wakeup period is now only decided by availability
of a working select() system call (which is the case on cygwin but not
mingw and MSVC) AND a vo that can provide an event file descriptor or a
similar hack (vo_corevideo). vos that cannot do either need polling for
event handling and now can set the wakeup period to 0.02 in the vo code.
Using vf_screenshot on Libav printed useless/misleading error messages
when playing 10 bit h264 with a VO that supports 8 bit yuv420p only:
Unsupported format 444p14le
Unsupported format 444p14be
...
The cause of this is that vf_scale is inserted to handle the format
conversion, and tries to find a pixel format with best quality. This
includes the 14 bit and 12 bit formats, which don't exist on Libav.
vf_screenshot tries to query whether Libav's libswscale supports it,
resulting in these error messages. (In theory, vf_scale is missing this
check, but it doesn't matter in practice.)
Since this warning is rather useless anyway, because all input video
comes from libavcodec, and only the conversion into the other could
possibly fail. Silence the warning by raising it to verbose message
level.
Closes#7.
