It doesn't understand that TARRAY_GROW never returns NULL, so teach it.

Turns out clang was right in warnung us about the possible division by
zero: we only asserted > ...

We previously just ignored this, even when the transfer queue required
it. I have no idea how I ...

This must be a multiple of 4, however our API does not require this
(rightfully so). So just cop...

Needed for compatibility with older vulkan versions

This limitation was causing issues with a very real use case: uploading
cropped texture data fro...

We have confirmed that these structs need to be used when creating an
image or buffer that will ...

Upstream has decided against renaming shaderc_shared.so to shaderc.so,
so right now this only fi...

This skips glslang by default if it's set to 'auto' when shaderc is also
found. (However, we sti...

I had previously assumed that _GENERAL was the only possible layout for
external API interop, bu...

This should hopefully make it slightly clearer how libplacebo expects
users to interact with its...

Make it clear that the contents of the memory are not useful while the
objects are not exported....

The exported memory/sync handles don't really make sense if we don't
expose the device UUID they...

These should make the intent more clear and reduce the risk of
confusion.

This reverts commit 52c21bbc3e3c109457cfe26dfc4c016a0e542523.

This reverts commit adab9b76e7fff029712d71326eb6224f2d77f7b5.

Forgot to actually mark them as dedicated, which led to them never
getting cleaned up.

We already have logic to pre-cache swapchain frames, so it's trivial for
us to fetch the first s...

Just use some dirty hacks to signal the VkSemaphore, for lack of any
actual external API usage h...

These must eventually be ref-counted once we start using them to
synchronize access to actual GP...

This change introduces the generic definition of an exported semaphore
pair. Such a pair can be ...

This is similar to the old vulkan hold/release_external but more
generalized and more powerful. ...

Edge case that can arise in the test suite.

When we introduce exportable semaphores, we'll have handles which are
not memory-backed. To avoi...

I missed this assertion when I added support for exportable textures.

Forgot this one in the previous commit.

I'm not 100% sure that this is guaranteed to be a power of two - and
even if it is currently, it...

This mirrors a similar change in mpv, which helped reduce vsync jitter
measurements by including...

The vertex buffer might be larger than the amount of vertex information
we're actually rendering...

This is analogous to the existing functionality for buffers. A
texture may be created, requestin...

ra -> gpu

This makes no real sense in gpu.h, since the pl_gpu doesn't actually
need this information in an...

No reason to have the space inside the string.

Ever since SPIRV-Cross has started emulating std140 layout when
cross-compiling SPIR-V to D3D11,...

Rather than editing this every time there's a new extension, it's
probably saner in the long run...

Since there's no better way to check for the existence of this value
other than by literally pro...

Rename subX/Y to sub_x/y for consistency

The code as written always used the subsampled weights for chroma, even
if the actual chroma pla...

While the access mask is ignored for a queue family ownership transfer,
the pipeline stage is no...

See the justification in the comments. Works around build issues on some
platforms. Not the clea...

Oops.

That brings back down the CPU usage of the API2 demo to basically zero.
Doesn't change th...

The renderer no longer has any abort() code paths, and the API version
indicator is now stable o...

The underlying device memory may get reused by libplacebo, thus
invalidating the contents of any...

Now it actually works correctly for the Cp channel

For debugging purposes

But add an extra assert to make sure we catch bugs due to forgetting to
add other non-linear col...

This may be useful in cases where users need access to a shared VkImage
but don't want to bother...

With the new instance_params field, we no longer need to separate the
`pl_vk_inst`.

Especially in the offline processing case (where no separate VkSurface
needs to be created), hav...

Replaced by a better one upstream.

Closes #48.

This comes in two variants, PQ and HLG. Normally it should be easy
enough to infer this from the...

This claims to implement the EOTF, but for HLG it omits the OOTF - so we
should clearly mark it ...

A stray "break" (left over from a previous iteration of the code) caused
the code for finding a ...

Explicitly defaulting the buffer type was also not the right thing,
since it e.g. conflicted wit...

The previous behavior assumed that the ordering in the vulkan driver's
memory types would be ord...

This is statically limited to 64 kB by the spec, so enforce these limits
by requiring host-mappe...

Since pl_tex_recreate already explicitly invalidates the contents of the
texture (due to the pos...

There's really no need to require this

Since users may want to allocate texture transfer buffers in GPU memory
for the purposes of uplo...

We create the fd when allocating the memory, since we only need one to
represent the entire allo...

This is useful to share memory with e.g. CUDA or other external APIs.

Closes #43

Handles are uint64_t on LP32, cf. VK_DEFINE_NON_DISPATCHABLE_HANDLE.

On Android arm32, VKAPI_PTR is defined to force the "hardfloat" calling
convention.

This is harmless since we cast it back to the correct type before
actually using it, but it's st...

These are false positives, silenced by some application of assert().

When using push descriptors these fields were never correctly NULL'd.

This is entirely not an error, but the way the code was written it
resulted in an undefined memo...

Introduced by the target format checking. Sometimes `target` is NULL.

It doesn't make sense to assume this will execute unconditionally,
especially when so many thing...

We forgot to add the offset of the vk_memslice into the vk_buf_write
when synchronizing via buf_...

Also fix a typo (was still saying pl_tex_recreate).

Right now, for some reason, when something involves visible writes, we
submit it as part of the ...

Due to the way the comment was formulated, it sounded like multiple
accesses by libplacebo to th...

This is can be used to apply AV1 film grain onto planes directly on the
GPU. Eventually, the pl_...

The current code treated SH_LUT_TEXTURE like SH_LUT_LINEAR for no good
reason (which was also wr...

No need to waste GPU time on the SH_LUT macro and texturing here,
texelFetch turns out to be fas...

Most of the time this layout is not actually needed.

These are used in several contexts, so give them better names.

A shader may fail due to any number of error conditions (that previously
resulted in just an err...

Reword an existing error and also add checking for a new error
condition, previously forgotten.

Since fa9df7d the shader cache actually completely broke for compute
shaders, since the tfmt che...

Fix 1050 -> 1050 Ti, also fix typo and some other improvements