[SYCL] Enable checking the result of max_num_work_groups query with exceeded launch limits on more backends (hip and opencl) #15369
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
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
GeorgeWeb
force-pushed
the
georgi/unsupported-max-coop-wgs-ur-api
branch
from
5e88e70
to
15a5ccf
Compare
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
|
See related oneapi-src/unified-runtime#2038 change. |
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
…xceeded launch limits on more backends (hip and opencl)
GeorgeWeb
force-pushed
the
georgi/unsupported-max-coop-wgs-ur-api
branch
from
15a5ccf
to
b5d35b0
Compare
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
| @@ -116,14 +116,14 @@ if(SYCL_UR_USE_FETCH_CONTENT) | |||
| CACHE PATH "Path to external '${name}' adapter source dir" FORCE) | |||
| endfunction() | |||
|
|
|||
| set(UNIFIED_RUNTIME_REPO "https://github.com/oneapi-src/unified-runtime.git") | |||
| set(UNIFIED_RUNTIME_REPO "https://github.com/GeorgeWeb/unified-runtime.git") | |||
There was a problem hiding this comment.
is this right? That the sycl product depends on your personal fork of the UR?
There was a problem hiding this comment.
It is a temp change for testing with oneapi-src/unified-runtime#2038. Of course it isn't intended to land like this. It will be modified back to https://github.com/oneapi-src/unified-runtime.git with the correct merged commit in oneapi-src/unified-runtime:main after the UR portion gets merged.
| if (dev.get_backend() == sycl::backend::ext_oneapi_cuda) { | ||
| // Note: Level-Zero currently always returns a non-zero value. | ||
| // TODO: Remove the backend condition once the Level-Zero API issue is fixed. | ||
| if (dev.get_backend() != sycl::backend::ext_oneapi_level_zero) { |
There was a problem hiding this comment.
This is a small thing, but can we improve the assert message? It seems that the message we want some other developer in the future to understand is that the query SHOULD have failed (but didn't).
cperkinsintel
approved these changes
Sep 26, 2024
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aarongreig
approved these changes
Oct 7, 2024
aarongreig
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aarongreig
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
aarongreig
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aarongreig
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
|
@intel/llvm-gatekeepers please merge |
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.
The HIP and OpenCL backend implementations of the query had a default return
1group implementation, which is an incorrect assumptions. They will now be marked as Unsupported with the accompanying UR chagnes (see oneapi-src/unified-runtime#2038), so for these cases thekernel_queue_specific::max_num_work_groupslaunch query will rely on the fallback that returns either1or0groups based on hardware resource limitation checks for the kernel.