In the past, src peak was always equal to or higher than dst peak. But
since `--target-peak` got introduced, this could no longer be the case.
This leads to an incorrect result (scaling for peak mismatch in gamma
light) unless some other option (CMS, --linear-scaling, etc.) forces the
linearization.
Fixes#6533
This solves some edge cases when using files with very weird metadata
(e.g. MaxCLL 10k and so forth). Instead of just blindly seeding it with
the tagged metadata, forcibly set the initial state from the detected
values.
Rather than the linear cd/m^2 units, these (relative) logarithmic units
lend themselves much better to actually detecting scene changes,
especially since the scene averaging was changed to also work
logarithmically.
Gamut mapping can take very bright out-of-gamut colors into the
negatives, which completely destroys the color balance (which tone
mapping tries its best to preserve).
This change switches to a logarithmic mean to estimate the average
signal brightness. This handles dark scenes with isolated highlights
much more faithfully than the linear mean did, since the log of the
signal roughly corresponds to the perceptual brightness.
In theory our "eye adaptation" algorithm works in both ways, both
darkening bright scenes and brightening dark scenes. But I've always
just prevented the latter with a hard clamp, since I wanted to avoid
blowing up dark scenes into looking funny (and full of noise).
But allowing a tiny bit of over-exposure might be a good thing. I won't
change the default just yet (better let users test), but a moderate
value of 1.2 might be better than the current 1.0 limit. Needs testing
especially on dark scenes.
The previous approach of using an FIR with tunable hard threshold for
scene changes had several problems:
- the FIR involved annoying hard-coded buffer sizes, high VRAM usage,
and the FIR sum was prone to numerical overflow which limited the
number of frames we could average over. We also totally redesign the
scene change detection.
- the hard scene change detection was prone to both false positives and
false negatives, each with their own (annoying) issues.
Scrap this entirely and switch to a dual approach of using a simple
single-pole IIR low pass filter to smooth out noise, while using a
softer scene change curve (with tunable low and high thresholds), based
on `smoothstep`. The IIR filter is extremely simple in its
implementation and has an arbitrarily user-tunable cutoff frequency,
while the smoothstep-based scene change curve provides a good, tunable
tradeoff between adaptation speed and stability - without exhibiting
either of the traditional issues associated with the hard cutoff.
Another way to think about the new options is that the "low threshold"
provides a margin of error within which we don't care about small
fluctuations in the scene (which will therefore be smoothed out by the
IIR filter).
Instead of desaturating towards luma, we desaturate towards the
per-channel tone mapped version. This essentially proves a smooth
roll-off towards the "hollywood"-style (non-chromatic) tone mapping
algorithm, which works better for bright content, while continuing to
use the "linear" style (chromatic) tone mapping algorithm for primarily
in-gamut content.
We also split up the desaturation algorithm into strength and exponent,
which allows users to use less aggressive desaturation settings without
affecting the overall curve.
Certain low-end Mali GPUs have a rather low precision and overflow
during the PRNG calculations, thereby breaking e.g. deband-grain.
Modify the permute() to avoid this, this does not impact the
quality of PRNG output (noticeably).
This problem was observed on:
GL_VENDOR='ARM', GL_RENDERER='Mali-T720'
GL_VERSION='OpenGL ES 3.1 v1.r15p0-00rel0.bdd9e62cdc8c88e0610a16b5901161e9'
This sacrifices some dynamic range for well-behaved sources, but
prevents catastrophic desaturation on badly mastered / too bright
sources. I think that's the better trade-off. This makes the
desaturation algorithm much "safer" to deploy by default, as well. One
could even argue going up to strength 1.0, which works better for some
sources but worse for others. But I think the current strength is the
best trade-off even after this change.
Mobius isn't well-defined for sig_peak <= 1.0. We can solve this by just
soft-clamping sig_peak to 1.0. Although, in this case, we can just skip
tone mapping altogether since the limit of mobius as sig_peak -> 1.0 is
just a linear function.
Based on testing with real-world non-HDR BT.2020 clips, clipping the
color space looks better than attempting to gamut map using a tone
mapping shader that's (by now) optimized for HDR content.
If anything, we'd have to develop a separate gamut mapping shader that
works in LCh space.
This solves a number of problems simultaneously:
1. When outputting HLG, this allows tuning the OOTF based on the display
characteristics.
2. When outputting PQ or other HDR curves, this allows soft-limiting the
output brightness using the tone mapping algorithm.
3. When outputting SDR, this allows HDR-in-SDR style output, by
controlling the output brightness directly.
Closes#5521
The HLG OOTF is defined as a one-parameter family of OOTFs depending on
the display's peak luminance. With the preceding change to OOTF scale
and handling, we no longer have any issues with outputting values in
whatever signal range we need.
So as a result, it's easy for us to support a tunable OOTF which may
(drastically) alter the display brightness. In fact, this is also the
only correct way to do it, because the HLG appearance depends strongly
on the OOTF configuration. For the OOTF, we consult the mastering
display's tagging (via src.sig_peak). For the inverse OOTF, we consult
the output display's target peak.
The primary need for this change is the fact that the OOTF was
incorrectly scaled, due to the fact that the application of the OOTF can
itself change the required normalization peak. (Plus, an oversight in
pass_inverse_ootf meant we forgot to normalize at the end of it)
The linearize/delinearize functions still normalize the scale since it's
used in a number of places throughout gpu/video.c, but the color
management function now converts to absolute scale right away, instead
of in an awkward way inside the tone mapping branch. The OOTF functions
now work in absolute scale only.
In addition, minor changes have been made to the way normalization is
handled for tone mapping - we now divide out the dst_peak *after* peak
detection, in order to make the scale of the peak detection buffer
consistent even if the dst_peak were to (hypothetically) change
mid-stream. In theory, we could also do this for desaturation, but doing
the desaturation before tone mapping has the advantage of preserving
much more brightness than the other way around - and even mid-stream
changes are not that drastic here.
Finally, some preparation work has been done for allowing the user to
customize the `dst.sig_peak` in the future.
The major changes are as follows:
1. Use `uint32_t` instead of `unsigned int` for the SSBO size
calculation. This doesn't really matter, since a too-big buffer will
still work just fine, but since `uint` is a 32-bit integer by
definition this is the correct way to do it.
2. Pre-divide the frame_sum by the num_wg immediately at the end of a
frame. This change was made to prevent overflow. At 4K screen size,
this code is currently already very at risk of overflow, especially
once I started playing with longer averaging sizes. Pre-dividing this
out makes it just about fit into 32-bit even for worst-case PQ
content. (It's technically also faster and easier this way, so I
should have done it to begin with). Rename `frame_sum` to `frame_avg`
to clearly signal the change in semantics.
3. Implement a scene transition detection algorithm. This basically
compares the current frame's average brightness against the
(averaged) value of the past frames. If it exceeds a threshold, which
I experimentally configured, we reset the peak detection SSBO's state
immediately - so that it just contains the current frame. This
prevents annoying "eye adaptation"-like effects on scene transitions.
4. As a result of the previous change, we can now use a much larger
buffer size by default, which results in a more stable and less
flickery result. I experimented with values between 20 and 256 and
settled on the new value of 64. (I also switched to a power-of-2
array size, because I like powers of two)
The current peak detection algorithm was very bugged (which contributed
to the excessive cross-frame flicker without long normalization) and
also didn't take into account the frame average brightness level.
The new algorithm both takes into account frame average brightness (in
addition to peak brightness), and also computes the values in a more
stable/correct way. (The old path was basically undefined behavior)
In addition to improving the algorithm, we also switch to hable tone
mapping by default, and try to enable peak computation automatically
whever possible (compute shaders + SSBOs supported). We also make the
desaturation milder, after extensive testing during libplacebo
development.
I also had to compensate a bit for the representational differences
between mpv and libplacebo (libplacebo treats 1.0 as the reference peak,
but mpv treats it as the nominal peak), but it shouldn't have caused any
problems.
This is still not quite the same as libplacebo, since libplacebo also
allows tagging the desired scene average brightness on the output, and
it also supports reading the scene average brightness from static
metadata (MaxFALL) where available. But those changes are a bit more
involved. It's possible we could also read this from metadata in the
future, but we have problems communicating with AVFrames as it is and I
don't want to touch the mpv colorimetry structs for the time being.
ra_d3d11 uses the SPIR-V compiler to translate GLSL to SPIR-V, which is
then translated to HLSL. This means it always exposes the same GLSL
version that the SPIR-V compiler supports (4.50 for shaderc/glslang.)
Despite claiming to support GLSL 4.50, some features that are tied to
the GLSL version in OpenGL are not supported by ra_d3d11 when targeting
legacy Direct3D feature levels.
This includes two features that mpv relies on:
- Reading from gl_FragCoord in the fragment shader (requires FL 10_0)
- textureGather from any texture component (requires FL 11_0)
These features have been exposed as new RA caps.
Normally such code is didsabled by have_mglsl==false in
check_gl_features(), but apparently not this one.
Just fix it. Seems also more readable.
Fixes#5069.
Comparing mpv's implementation against the ACES ODR reference samples
and algorithms, it seems like they're happy desaturating highlights
_way_ more aggressively than mpv currently does. And indeed, looking at
some example clips like The Redwoods (which is actually well-mastered),
the current desaturation produces unnatural-looking brightness fringes
where the sky meets the treeline.
Adjust the algorithm to make it apply to a much larger, more gradual
brightness region; and change the interpretation of the parameter. As a
bonus, the new parameter is actually sanely scaled (higher values = more
desaturation). Also, make it scale based on the signal level instead of
the luminance, to avoid under-desaturating bright blues.
This makes the radeon driver shut up about frequently updating
STATIC_DRAW UBOs (--opengl-debug), and also reduces the amount of
synchronization necessary for vulkan uniform buffers.
Also add some extra debugging/tracing code paths. I went with a
flags-based approach in case we ever want to extend this.
Almost as fast as the old code, but more general. Notably, glslang
doesn't support nested arrays.
(cf. https://github.com/KhronosGroup/glslang/issues/1057)
Also much cleaner code-wise, so I think I'll keep it even if glslang
implements array_of_arrays.
This is done in several steps:
1. refactor MPGLContext -> struct ra_ctx
2. move GL-specific stuff in vo_opengl into opengl/context.c
3. generalize context creation to support other APIs, and add --gpu-api
4. rename all of the --opengl- options that are no longer opengl-specific
5. move all of the stuff from opengl/* that isn't GL-specific into gpu/
(note: opengl/gl_utils.h became opengl/utils.h)
6. rename vo_opengl to vo_gpu
7. to handle window screenshots, the short-term approach was to just add
it to ra_swchain_fns. Long term (and for vulkan) this has to be moved to
ra itself (and vo_gpu altered to compensate), but this was a stop-gap
measure to prevent this commit from getting too big
8. move ra->fns->flush to ra_gl_ctx instead
9. some other minor changes that I've probably already forgotten
Note: This is one half of a major refactor, the other half of which is
provided by rossy's following commit. This commit enables support for
all linux platforms, while his version enables support for all non-linux
platforms.
Note 2: vo_opengl_cb.c also re-uses ra_gl_ctx so it benefits from the
--opengl- options like --opengl-early-flush, --opengl-finish etc. Should
be a strict superset of the old functionality.
Disclaimer: Since I have no way of compiling mpv on all platforms, some
of these ports were done blindly. Specifically, the blind ports included
context_mali_fbdev.c and context_rpi.c. Since they're both based on
egl_helpers, the port should have gone smoothly without any major
changes required. But if somebody complains about a compile error on
those platforms (assuming anybody actually uses them), you know where to
complain.
Wessel Dankers
Niklas Haas
sfan5
James Ross-Gowan
wm4