[UR][L0 v2] Set pointer kernel arguments only for queue's associated device #20179
Conversation
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
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
force-pushed
the
set_arg_pending_args_v2
branch
from
f961f83 to
3148745
Compare
againull
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
againull
force-pushed
the
set_arg_pending_args_v2
branch
from
3148745 to
51f881d
Compare
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
force-pushed
the
set_arg_pending_args_v2
branch
from
51f881d to
fc827a1
Compare
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
pbalcer
reviewed
Sep 26, 2025
Comment on lines
429
to
434
| // In multi-device context store the raw pointer value and defer setting the | ||
| // argument until we know the device where kernel is being submitted. | ||
| if (hKernel->getContext()->getDevices().size() > 1) { | ||
| hKernel->addPendingPointerArgument(argIndex, pArgValue); | ||
| return UR_RESULT_SUCCESS; | ||
| } |
There was a problem hiding this comment.
tbh I'd rather we have the same path for both scenarios. But this is going to be rewritten/removed with the transition to enqueue kernel with args, so up to you if you want to change it.
| return UR_RESULT_ERROR_INVALID_KERNEL_ARGUMENT_INDEX; | ||
| } | ||
|
|
||
| pending_pointer_args.push_back({argIndex, pArgValue}); |
There was a problem hiding this comment.
This is a potential allocation in the hot path. Might be worth testing performance.
There was a problem hiding this comment.
Sure, here is the successful perf testing for this PR: https://github.com/intel/llvm/actions/runs/18106004233
againull
added a commit
that referenced
this pull request
Sep 26, 2025
…ice (#20177) In multi-device context, ensure that pointer kernel arguments are set only for the device associated with the queue being used for kernel launch. Previously, arguments were set for all devices in the kernel's device map, which was unnecessary and potentially incorrect when launching on a specific device. Same will be done for v2 adapter here: #20179
github-actions bot
pushed a commit
to oneapi-src/unified-runtime
that referenced
this pull request
Sep 27, 2025
In multi-device context, ensure that pointer kernel arguments are set only for the device associated with the queue being used for kernel launch. Previously, arguments were set for all devices in the kernel's device map, which was unnecessary and potentially incorrect when launching on a specific device. Same will be done for v2 adapter here: intel/llvm#20179
againull
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
kswiecicki
approved these changes
Sep 30, 2025
| uint32_t groupSizeY, uint32_t groupSizeZ, | ||
| ze_command_list_handle_t commandList, wait_list_view &waitListView) { | ||
| auto hZeKernel = getZeHandle(hDevice); | ||
| auto &deviceKernelOpt = deviceKernels[deviceIndex(hDevice)]; |
There was a problem hiding this comment.
nit: I'm wondering whether getZeHandle should also perform this .has_value() check.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
No description provided.