[SYCL][Bindless] Allow 3-channel image formats #15210
Merged
Conversation
This file contains 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
This patch doesn't implement support for 3-channel formats,
it just allows it in the spec and implementation
so that a backend could implement it later (e.g. Level Zero).
* Removed check from `image_descriptor::verify`
* Updated spec to allow 3-channel formats
* Very simple test, just verifies that CUDA backend still fails
* Missing some PTX intrinsics, use `float4` instead of `float4`
when compiling for CUDA. The more important part is checking that
the CUDA backend throws an exception when trying to create image.
isaacault
approved these changes
Aug 28, 2024
DBDuncan
approved these changes
Aug 28, 2024
cppchedy
approved these changes
Aug 28, 2024
nrspruit
reviewed
Aug 28, 2024
| @@ -0,0 +1,109 @@ | |||
| // REQUIRES: cuda | |||
There was a problem hiding this comment.
Why are you limiting this to CUDA adapter?
There was a problem hiding this comment.
Because it requires extra L0 work, I see you added it here: https://github.com/oneapi-src/unified-runtime/pull/2022/files
gmlueck
reviewed
Aug 28, 2024
sycl/doc/extensions/experimental/sycl_ext_oneapi_bindless_images.asciidoc
Show resolved
Hide resolved
wenju-he
reviewed
Aug 29, 2024
Co-authored-by: Wenju He <[email protected]>
Currently no way to query whether a backends supports 3 channel formats.
# Conflicts: # sycl/doc/extensions/experimental/sycl_ext_oneapi_bindless_images.asciidoc
# Conflicts: # sycl/doc/extensions/experimental/sycl_ext_oneapi_bindless_images.asciidoc
ProGTX
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
|
@jbrodman are you OK with these changes? We can build on top of them later. |
ProGTX
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
|
@intel/llvm-gatekeepers please merge this, there will be follow up work to add queries and to properly test on L0 |
YixingZhang007
pushed a commit
to YixingZhang007/llvm
that referenced
this pull request
Oct 30, 2024
This patch doesn't implement support for 3-channel formats, it just allows it in the spec and implementation so that a backend could implement it later (e.g. Level Zero). * Removed check from `image_descriptor::verify` * Updated spec to allow 3-channel formats * Very simple test, just verifies that CUDA backend still fails * Missing some PTX intrinsics, use `float4` instead of `float4` when compiling for CUDA. The more important part is checking that the CUDA backend throws an exception when trying to create image. --------- Co-authored-by: Wenju He <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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 patch doesn't implement support for 3-channel formats, it just allows it in the spec and implementation
so that a backend could implement it later (e.g. Level Zero).
image_descriptor::verifyfloat4instead offloat4when compiling for CUDA. The more important part is checking that the CUDA backend throws an exception when trying to create image.