As of macOS Big Sur and iOS/tvOS 14, the `discard_fragment()` function
in MSL is defined to have demote semantics; that is, fragment shader
output is discarded, but the fragment shader thread continues to run as
a helper invocation. This is very useful for Direct3D emulation, since
this is the semantic that HLSL `discard` has.

Signed-off-by: Chip Davis <chip@holochip.com>

…pple Silicon.

Use an explicit gradient to make it sample the correct level.

Update SPIRV-Cross to pull in the change needed for this.

Work around problems with explicit LoD with arrayed depth images on Apple Silicon.

Approved-by: Steven Winston

As of macOS Big Sur and iOS/tvOS 14, the `discard_fragment()` function
in MSL is defined to have demote semantics; that is, fragment shader
output is discarded, but the fragment shader thread continues to run as
a helper invocation. This is very useful for Direct3D emulation, since
this is the semantic that HLSL `discard` has.

Signed-off-by: Chip Davis <chip@holochip.com>

Advertise the VK_EXT_shader_demote_to_helper_invocation extension.

# Conflicts:
#	Docs/MoltenVK_Runtime_UserGuide.md
#	MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h
#	MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
#	MoltenVK/MoltenVK/GPUObjects/MVKRenderPass.mm
#	MoltenVK/MoltenVK/Layers/MVKExtensions.def

# Conflicts:
#	Docs/MoltenVK_Runtime_UserGuide.md
#	MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
#	MoltenVK/MoltenVK/Layers/MVKExtensions.def

# Conflicts:
#	MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h
#	MoltenVK/MoltenVK/GPUObjects/MVKRenderPass.mm

…dating from upstream/main.

…r time.

2.) remove validation from the driver layer.

Leave some room for XFB buffers and the argument buffers by limiting
maximum vertex input bindings to 16.

# Conflicts:
#	Docs/MoltenVK_Runtime_UserGuide.md
#	ExternalRevisions/SPIRV-Cross_repo_revision
#	MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
#	MoltenVK/MoltenVK/Layers/MVKExtensions.def

reverting the revision change.

…o transform_feedback_public

# Conflicts:
#	ExternalRevisions/SPIRV-Cross_repo_revision

…o transform_feedback_public

Fix the formatting of some added parts to be more consistent with the
rest of MoltenVK.

Since the vertex shader will be writing directly to the transform
feedback buffers, it occurs to me that no copying at all is necessary.

Sets up the pipeline to have two stages and compiles the vertex shader
as a compute shader, but doesn't yet generate the passthrough vertex
shader.

…ransform_feedback

What the heck is wrong with me...?

… pass-through shader.

… shader.

Still broken. In particular, it doesn't handle structs or arrays
properly. Arrays need an offset added to the XFB offset for each
element. Structs aren't handled at all.

Sort captured outputs by offset. Sort others by output location.

Use packed types if the offset isn't aligned.

We still need to emit the array of clip distances for the rasterizer;
this is just for the fragment shader interface.

Fix clip/cull distance attribute for the fragment shader interface.

Make sure SPIRV-Cross knows about them.

…ers.