[Vulkan][DirectX] Add support for array textures, starting with Texture2DArray and RWTexture2DArray - #1468
Open
Icohedron wants to merge 7 commits into
Open
Conversation
…rray
Introduces the Texture2DArray and RWTexture2DArray resource kinds and the
plumbing they need to be described, validated and laid out, without wiring
them into any back-end yet.
- OutputProperties gains ArraySlices (default 1). Buffer size validation
multiplies the mip chain by the slice count.
- Array texture data is laid out slice-major: the full mip chain of slice 0,
then slice 1, and so on. This matches D3D12's `Mip + Slice * MipLevels`
subresource indexing, so the same YAML feeds every back-end.
computeTightTextureUploadLayout() emits one footprint per (mip, slice).
- TextureCreateDesc gains ArraySlices and IsArray. IsArray is not derivable
from the slice count: a Texture2DArray with a single slice still needs an
array-typed view.
- Resources are validated at parse time: ArraySlices may only exceed 1 for
an array kind, and must be at least 1.
Adding a ResourceKind means every exhaustive switch over it has to handle the
new values, so the descriptor and image-view mappings are filled in here and
Metal asserts on the kinds it cannot create. Creating and copying layered
images is left to the per-back-end commits that follow.
Assisted by: Claude Opus 5
Describes the slice count on the resource and selects array-typed views:
- getDXResourceDesc() sets DepthOrArraySize from ArraySlices, which is what
makes GetCopyableFootprints report the per-(mip, slice) subresources that
the upload and readback paths already walk.
- SRV and UAV creation pick TEXTURE2DARRAY over TEXTURE2D based on IsArray,
filling in FirstArraySlice/ArraySize.
Assisted by: Claude Opus 5
Maps the new kinds onto layered 2D images:
- Images are created with arrayLayers taken from the slice count, and image
views expose every layer.
- Copy regions and barriers iterate (slice, mip) rather than mip alone, so
uploads and readbacks address each subresource individually. Vulkan
staging buffers stay tightly packed and slice-major, matching the layout
the harness produces.
Assisted by: Claude Opus 5
Icohedron
force-pushed
the
users/icohedron/texture2darray
branch
from
August 21, 2026 02:33
b73f57c to
841135a
Compare
This was
linked to
issues
Aug 21, 2026
Assisted by: Claude Opus 5
2 tasks
bob80905
approved these changes
Aug 25, 2026
bob80905
left a comment
bob80905
left a comment
Contributor
There was a problem hiding this comment.
Overall excellent contribution imo, just a couple things. None are blocking.
| I.setError(Twine("Referenced buffer ") + R.Name + " not found!"); | ||
| } | ||
| if (R.BufferPtr) | ||
| if (auto Err = validateTextureResource(R)) |
Contributor
There was a problem hiding this comment.
[SUGGESTION]: The convention here seems to be to check the resource type before validating for that type. You've baked in the texture type check into the validation function, but why not check whether the resource is a texture type, and if so, then call the validation function?
So something like if (R.BufferPtr && R.isTexture()) ?
| float4 mainPS(float4 Pos : SV_POSITION) : SV_Target { | ||
| uint2 GTid = (uint2)Pos.xy; | ||
|
|
||
| // Texture2DArray::CalculateLevelOfDetail takes a **float2**, not the float3 |
Contributor
There was a problem hiding this comment.
nit: It may be possible to make this comment less verbose.
| Out[6] = Tex.SampleLevel(BorderSamp, float3(1.25, 0.5, 1), 0); // -> border | ||
| Out[7] = Tex.SampleLevel(MirrorSamp, float3(1.75, 0.5, 1), 0); // -> Blue | ||
|
|
||
| // The slice coordinate is *not* subject to the address mode: it is always |
Contributor
There was a problem hiding this comment.
Likewise with this comment.
|
|
||
| [numthreads(1, 1, 1)] | ||
| void main() { | ||
| // Every slice carries its own mip chain. Slice 0 is mip 0 [Red, Green] over |
Contributor
There was a problem hiding this comment.
Likewise, generally might be worth doing a light comment skim and making sure they're trimmed and concise.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR adds support for array textures in the VK and DX backends.
To start with, only Texture2DArray and RWTexture2DArray are supported.
The remaining array texture types will be added in subsequent PRs.
This PR also fixes some of incorrectly-linked XFAILs in texture tests, and adds some missing docs about UAVs for mipmapped textures that also apply to array textures.
I also snuck in tests for RWTexture2D since we are implementing RWTexture2DArray with this PR anyways.
Assisted by: Claude Opus 5
Stack created with GitHub Stacks CLI • Give Feedback 💬