This allows clients to initialize 3D image objects with data in a BC1,
BC2, or BC3 (aka DXTn, aka S3TC) format. Since Metal doesn't support
this natively, DXTn-compressed image data are manually decompressed
prior to upload or copying. This particular algorithm was derived from
Wine, from code originally authored by Connor McAdams. It's somewhat
primitive, and could probably stand some vectorization, but it's a good
first start.
The meat of the algorithm is pulled into a header, "MVKDXTnCodec.def",
intended to be used by both CPU and shader code. CPU uploads, which in
MoltenVK happen when binding a `VkDeviceMemory` object to an image, run
the code normally. GPU uploads, i.e. copying from a buffer to a texture,
run the algorithm as a compute shader, possibly storing the results to
an intermediate buffer to then copy directly to the texture.
Intermediate buffers are used when uploading to mip levels higher than
0, since desktop Metal doesn't support directly writing to higher mip
levels from a shader.
Only uploads are supported. Downloads are not implemented yet.
Theoretically, this could be extended to other compressed formats as
well, allowing MoltenVK to support all texture compression formats on
all platforms.
Decrease minimum swapchain images to 2.
Add MVKPhysicalDeviceMetalFeatures::minSwapchainImageCount & maxSwapchainImageCount.
Minor refactoring of mvkClamp() & MVKPhysicalDevice::getSurfacePresentModes().
Update VK_MVK_MOLTENVK_SPEC_VERSION to 15.
Update MoltenVK version to 1.0.30.
Instead of assuming that `kMVKFormatNone` means the format is
compressed. This makes the code doing these checks more obvious and
self-documenting.
Since I changed the public MoltenVK-specific API, I also bumped the
`VK_MVK_moltenvk` spec revision.
Since a pipeline might be shared between multiple draws, but those draws
could use different textures, it isn't tenable to have a single aux
buffer shared between those draws. Instead, we'll set up constant
buffers using the `setXxxBytes()` method on the command encoder. That
way, separate draws using the same pipeline won't interfere with each
other.
This change also uses separate buffers for the vertex and fragment
shaders, since they may have different sets of textures.
Null the `sampler` field of the stored `VkDescriptorImageInfo` if the
descriptor uses an immutable sampler. That way, we won't try to
double-free a sampler.
Retain immutable samplers when the set is created. Release them when the
set is destroyed.
In which I am proved wrong again. While Metal indeed won't assert if the
pixel formats don't match, the results of attempting such a copy are...
mixed, to say the least. Sometimes, nothing is copied at all. Other
times, pixels are copied in weird ways that don't seem to conform at all
to expectation. (This is most apparent in the output of the CTS's image
copying tests.) It's best, then, to ensure that the source and
destination have the same format.
Unfortunately, in the process of writing this, my fears regarding not
being able to make uncompressed views of compressed images were
confirmed. So we can't do this for compressed images.
All that is required is that we can either read the format
(`TRANSFER_SRC`) or sample from it (`SAMPLED_IMAGE`). Every image format
that Metal supports supports those two features.
This marks the depth/stencil formats as able to be the source of a
blit. The spec requires in particular that `D16_UNORM` and `D32_SFLOAT`
support that.
Compressed formats only support reading and filtering. They also can't
be used to create linear textures, so they can't be used for buffer
views, either.
Use `MTLVertexFormatChar`, `UChar`, `Short`, `UShort`, and `Half` when
available. Add `MTLVertexFormatUChar4Normalized_BGRA` for
`VK_FORMAT_B8G8R8A8_UNORM`.
These are only available on Metal 2.0 (iOS 11, macOS 10.13) and up, so
this adds what MoltenVK was using before as a fallback, similar to how
some `MTLPixelFormat`s have substitutes.
