vo/vulkan: sync requested extensions and features from ffmpeg

[philipl]

Over time, the list of requested extensions and features has diverged from
ffmpeg (libhwcontext_vulkan), and while this isn't guaranteed to cause
problems, as it depends on driver behaviour, it definitely makes it hard to
compare validation errors between the two implementations.

There is a large list of feature activations that have to be handled by an
update to libplacebo, which I'm making separately.

[github-actions]

Download the artifacts for this pull request:

Windows

• mpv-aarch64-pc-windows-msvc
• mpv-i686-w64-mingw32
• mpv-x86_64-pc-windows-msvc
• mpv-x86_64-w64-mingw32

macOS

• mpv-macos-13-intel
• mpv-macos-14-arm
• mpv-macos-15-arm

[kasper93]

it definitely makes it hard to compare validation errors between the two implementations.

Remember that since those are optional extensions, they might not exists, and validation errors or the implementation should work fine both with them not loaded.

[philipl]

it definitely makes it hard to compare validation errors between the two implementations.

Remember that since those are optional extensions, they might not exists, and validation errors or the implementation should work fine both with them not loaded.

For sure. But I mean when doing a side-by-side comparison between ffmpeg and mpv, in which case the Vulkan environment will be the same.

[kasper93]

For the record about Vulkan headers version supported. Current Android NDK (r28c) have 275 headers. Debian stable (trixie) (released two days ago) has 309, while oldstable (bookworm) has 239. Ubuntu 24.04 (LTS) has 275.

So, if we ignore oldstable, we could bump requirement to 275, but this covers only 2/7 of those extensions. Probably not worth it.

Of course those are only headers, but we aimed to not require side loading them, at least for latest Debian and NDK, which are generally what's lagging behind. Unless there is a need to do so.

[philipl]

For the record about Vulkan headers version supported. Current Android NDK (r28c) have 275 headers. Debian stable (trixie) (released two days ago) has 309, while oldstable (bookworm) has 239. Ubuntu 24.04 (LTS) has 275.

So, if we ignore oldstable, we could bump requirement to 275, but this covers only 2/7 of those extensions. Probably not worth it.

Of course those are only headers, but we aimed to not require side loading them, at least for latest Debian and NDK, which are generally what's lagging behind. Unless there is a need to do so.

Still seems worth bumping as a matter of principle. If we let debian oldstable hold us back, we'll be here forever. There's a bunch of other clean up we can do now that trixie is out, right?

[kasper93]

LGTM. I didn't check exact versions, but I trust you on that.

[sfan5]

Can't test but it compiles fine for Android.

[philipl]

I'm going to take another pass at this later this week to add all the extensions/features I initial proposed to add directly into libplacebo. Based on the discussion around that, they should all be here too.

[philipl]

I have updated to add the additional extensions/features that I had previously added directly in libplacebo. As discussed, I'm putting all of them here so that no libplacebo changes are made.