This is substantially more accurate than the method used in CAT97, and
improves the perceptual quality of color adaptation, including color
temperature adaptation and color blindness simulation.

Also fixes an out-of-date comment.

Kacper Michajłow authored 10 months ago and Niklas Haas committed 10 months ago

This makes cache entries compatible between 32-bit and 64-bit builds and
avoids issues with loading cache wrong.

This was mistakenly added to the list of device-level function pointers,
which doesn't even make any sense as it's not a device-level function.

Caught by a warning in new versions of the vulkan validation layers,
since the result is always NULL.

Avoids triggering undefined behavior of GPU upload/download paths.

While avoiding Infinity and NaN.

For style reasons.

In the worst case, merely improves style. In the best case, allows
per-file caching to work.

For copying terminal output.

Misaki Kasumi authored 11 months ago and Niklas Haas committed 10 months ago

It is a GLES 2.0 extension, and we do not enable 16f formats with GLES 2.0 at present.

This is already called and stored.

Capabilities like vertex, texel buffer support or readback shouldn't
influence the sorting, because this requirement will typically be
explicitly filtered for when searching for such formats.

This capability sorting only exists to allow prioritizing formats that
have the optional ability to leverage more performant code paths, e.g.
by allowing blits, linear sampling or texture storage. And in either
case, we wouldn't want to pick a worse rendering format just for some
minor side benefit from having it be compute-dispatchable when blending
or some other obscure edge case like that.

(And lastly, there is also the user friendliness consideration of
wanting "primitive" texture formats near the top of the list)

introduced in Linux v5.14

llvm-rc doesn't allow multi-byte codes in narrow utf-8.

Save `ref->img.rect` before it gets indirectly rounded and use that
for calculations.

Fixes: #324

And remove the now redundant sig variable. All cache entries should be
invalidated by the previous commit anyway, but change the seed just to
be on the safe side.

This also makes them work on GL 3.0, since we get that for free.

The existing OpenGL code can transparently make use of these features by
just correcting the checks.

Fixes: 8e6062a2

This reverts commit 4fdb9c0d.

Caused build issues, will update when they are fixed upstream.

Oops?

Not sure this matters for anything. To be on the safe side, only force
it if there was an exact match.

Conform to a style more consistent with other infodumps in libplacebo,
and move it out of random logic and into a common place.

Again, the explicit parameters here are not making things clearer.

Just make `icc` itself mutable. It's already complicated enough anyway,
the explicit parameter list is not simplifying things.

For consistency

Supersedes: videolan/libplacebo!643

Annoyingly we have to memcpy this array if we want to pretty-print the
Auto (%s) case like we used to. Oh well. This could probably be done
more elegantly, but it's definitely not worth the time investment.

Shocking, I know.

Motivated by !643, but also generally useful.

Explicit OOTF support was removed and baked into the EOTF step a long
time ago.