Instead of hard-coding the bindings and position in pl_shader_res, place
them dynamically once w...

Debanding in particular requires linear sampling, and sh_bind in general
requires 2D, sampleable...

By just clearing the output_w/h again. This basically makes it an
invisible operation as far as ...

This makes no more sense, since it's part of the color_repr.

This adds the secondary shader as a subpass of the first shader;
exposing it as a function (much...

This copies all elements from one array into another.

This is a helper built on top of talloc, which maintains a refcounted
container. The function `t...

This was missing for whatever reason.

Left-over from an earlier API design.

v0.1.2

Fixes:
- Add missing header file
- Cosmetic fixes in the generated shaders

Improvements...

In particular, the device limits

- enable IRC notifications
- enable only for whitelisted branches
- enable test framework + bett...

Forgot to commit this, it seems..

No need to waste a VA on this (or force the user to pass one in)

Compute shaders don't have this (for obvious reasons)

Missing space after layout()

Omitting this doesn't actually break anything due to the way sh_fresh
works, but it's still the ...

I've also found some independent resources describing this exact
algorithm. In particular, one o...

Turns out both Vulkan and OpenGL require this, and D3D11 apparently as
well.

And out of the params struct, where it doesn't really fit. (Due to the
fact that the params stru...

Instead of having the user pass in seed values, instead pass in a global
`index` on pl_shader ge...

Returns floats (in pixels) instead of integers in half-pixels. No real
reason not to. Floats are...

- get rid of the useless padding field, since padded textures aren't a
thing in practice, and ...

In particular, the device limits

- enable IRC notifications
- enable only for whitelisted branches
- enable test framework + bett...

Forgot to commit this, it seems..

Currently only supports ordered dither and white noise. I plan on adding
blue noise as well; per...

Omitting this doesn't actually break anything due to the way sh_fresh
works, but it's still the ...

No need to waste a VA on this (or force the user to pass one in)

Compute shaders don't have this (for obvious reasons)

I've also found some independent resources describing this exact
algorithm. In particular, one o...

Missing space after layout()

The default settings are changed (since the recent fixes have made async
compute stable to enabl...

This corresponds to a typical standard gamut SDR RGB monitor.

Not only are these unnecessary, but so is the env var check in this
case. Probably for the better.

Was missing a stub function

Since assert() can be compiled out, and users will vehemently argue for
their right to do so, ju...

I didn't like the nested blocks. I also think `usewinloc` sounds sort of
weird, so change the fo...

To cross compile using mingw32:

$ meson --cross-file cross_mingw_x86_64.txt build-w64

or

$ me...

The _create_local function need the msvcr100.dll that is available starting
Windows 7.

Remove one duplicate definition, and move the dispatch table to the
bottom.

Instead of recreating the object completely. One fewer allocation, not
that it matters..

Needed for the definition of pl_matrix3x3 etc.

Even for coherent buffers, we technically need to ensure writes are made
visible to the host by ...

After lots of testing and careful cross-examination between HDR UHD
clips and their SDR counterp...

This accomplishes several things:

1. vec4 colors are now explicitly always normalized such that...

Basically a wrapper around PL_ALIGN2 so users don't have to NIH this.
Might add more such wrappe...

As a rule of thumb, if there's a pl_foo_default_params, then specifying
a const struct pl_foo_pa...

Apparently atexit() in libraries is bad news, and pthread_once_t (in
printf.c) leaks memory anyw...

This is to avoid everybody having to reimplement the same logic
internally. It also defaults the...

The QF index and the qinfo index are not necessarily the same. In fact,
the former can be arbitr...

New extensions will be guarded by their presence in vulkan.h, so I can
add new enumeration membe...

Let's make sure we don't end up in confusing scenarios where code only
randomly works on some pl...

Since nobody can agree on adding `printf_l` to the standard, we have to
do this using OS-depende...

This is basically equivalent to hardware bilinear/bicubic sampling.

Re-used some of the logic from the polar sampling setup code

The mpv code hard-coded this to linear, but we need to make it dynamic.

Since e.g. `uselocale` can override `setlocale` for that thread, and we
have absolutely no way o...

The color attachment pass should be using the access mask of the color
attachment write, *not* w...

It's common to have e.g. build.foo or build.clang

See mpv commit c2d4fd0ef43e44c5c8caee13ae486ae05784c3a5 for more
information.

Using install_headers with subdirectories like that breaks completely.
Solve it by just using su...

This is component order invariant, so make it so. Fixes a bug where
a2bgr10 etc. are not correct...

For sorting, we try a few heuristics but ultimately maintain the
guarantee that formats with "mo...

1. Some capabilities now also imply other capabilities
2. Texture creation is now *always* suppo...

Fragment shader is still TODO.

Also, the test desperately needs to be improved. Right now it ju...

Thank god for LIBPLACEBO_LEAK_REPORT

size_t can never be < 0 anyway.

First of all, make sure we warn the user if NDEBUG was set; with a
suitably evil sounding env va...

This fixes a bug where pl_dispatch would use the format type, possibly
leading to undefined beha...

Needed for symmetry

No reason to make this dynamically allocated. It makes zero sense to
have more namespaces than t...

This makes the build directory a bit saner

In order to release the resources associated with a shader without
submitting anything.

This was brittle and ugly and slow and annoying anyway. Instead, turn
this into a user problem. ...

This avoids all potential headaches due to multiple evaluation of the
compute shader position macro

This required making some (semi-significant) changes to shaders/ra as
well. This provides the in...

Still missing the fake vertex simulation; but this is a start.

For shaders which require access to "configured" objects, this tracks
the object's lifetime and ...

Taken from mpv, with some modifications to fit the new system.

Not entirely sure I'm happy with...

This is more convenient code-wise. Also adds some much-needed helpers.