Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Since we can have a display in studio range with (SPU) in full range. And since
the YUV->RGB add a studio to full range conversion we need to undo it when
displaying on a studio range swapchain.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The transfer manipulation is only done if the source doesn't match the display.

On systems older than Windows 10, RGB BT.709 Gamma 2.2 is always picked.

Use the ST2084 target for HLG sources as well.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

It will only give a list on Windows 10.

The DXGI_COLOR_SPACE_TYPE field is generated based on the values in the
structure to make sure the values match the DXGI value.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The saturate should be done after the pixel is ready for EOTF.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The standard mentions the 12.0 coefficient. The rendering problem is elsewhere.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

This was untested so far but now we have proper PQ output handled by Windows.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

In other words, if a YUV->RGB matrix conversion will be used

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Steve Lhomme authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Fix #17666

This reverts commit 8447e3d8.

Source of undersized buffers on format change.

Since audio format is derived from fmt.i_codec, the update
path is inconsistent, not updating the aout format in case
fmt.i_codec changes (only comparing the derived audio.i_format)

otherwise computed dts/pts will always be incorrect if no vui

Since pictures are held while being rendered by the GPU.

Ref #18116

Ref #18116

No non iframe output with single NAL blocks regression.

did not work with audio less track or misconfigured
track like PMF, and then no pcr was issued

avoid confusion and errors.
tracks id are different.

If the decoder is draining, more pictures or blocks can be outputted, so wait
the drain to finish before querying outputs.

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

As a result of dbae0e30, a comment was incorrectly split up so that it
implies that each split applies to one particular value (which is not
correct).

These changes fix that by putting both entities into a single comment,
hopefully signalling that either "0x01" or "0x02" means "closed
captions/XDS" (and not each for one).

--

Thanks for Francois Cartegnie for pointing out the mistake:

 - https://mailman.videolan.org/pipermail/vlc-devel/2017-March/112291.html



Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The type yield by va_arg( args, T ) is already guaranteed to be T;
there is no need for the explicit cast.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

A switch should be easier to maintain, while also improving
readability.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The type yield by va_arg( args, T ) is already guaranteed to be T;
there is no need for the explicit cast.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The type yield by "va_arg( args, T )" is already guaranteed to be T,
there's no need for the explicit extra cast.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

And do not tag blocks with switching slices

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

This function has been unused since 2011.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

The last usage of this function was removed by fe5df3d5 in March 2012.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

Filip Roséen authored 8 years ago and Hugo Beauzée-Luyssen committed 8 years ago

Last usage was removed in June 2008 as part of the changed instroduced
by 95addd57.

Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>