Skip to content

Conversation

hdelan
Copy link
Contributor

@hdelan hdelan commented Jun 28, 2024

Batch merge of PRs:

#1792
#1787
#1782
#1783
#1781
#1773
#1765
#1763

@hdelan hdelan requested review from a team as code owners June 28, 2024 12:30
@hdelan hdelan requested a review from steffenlarsen June 28, 2024 12:30
@github-actions github-actions bot added loader Loader related feature/bug conformance Conformance test suite issues. specification Changes or additions to the specification experimental Experimental feature additions/changes/specification level-zero L0 adapter specific issues cuda CUDA adapter specific issues hip HIP adapter specific issues command-buffer Command Buffer feature addition/changes/specification labels Jun 28, 2024
@JackAKirk
Copy link
Contributor

LGTM for #1792 and #1765

@hdelan
Copy link
Contributor Author

hdelan commented Jun 28, 2024

These are all already approved and Ready to merge individually so I don't think we need any additional reviews?

@hdelan
Copy link
Contributor Author

hdelan commented Jun 28, 2024

At least in this previous batch PR #1774 we just got an approval from Benie and then it was merged.

@callumfare
Copy link
Contributor

@hdelan I don't think we do, I'm happy to merge once tests are all passing. But obviously if anyone wants to check things before then that's ok

JackAKirk and others added 7 commits June 28, 2024 13:56
this aspect is true only for sm_90 cuda gpus.

Signed-off-by: JackAKirk <[email protected]>
Signed-off-by: JackAKirk <[email protected]>
An event created with interop doesn't have an associated queue. So get
an active device through the context.
get_program is the same as getProgram.
We can only call hipMemAdvise on managed pointers allocated via
hipMallocManaged, so can only support the UR entry point on pointers
allocated via urUSMSharedAlloc.

When faced with other pointers, we're supposed to ignore them and result
"success", but instead this PR continues the tradition of returning
UR_RESULT_ERROR_ADAPTER_SPECIFIC with a warning message providing the
user more information. This should all be fixed up later when there's a
better warning mechanism.
AD2605 and others added 2 commits June 28, 2024 15:35
set property cudaFuncAttributeNonPortableClusterSizeAllowed only if cluster launch is used

set has_property_cluster_launch only if cluster property is used

fix cluster dimensions being set in accordance to grid dimensions

fix ordering of cluster dims for workDim 2

fix compilation errors

review comments 1

review comments 1

increase cluster size upon launch to check CU_FUNC_ATTRIBUTE_NON_PORTABLE_CLUSTER_SIZE_ALLOWED flag being added
- Fix incorrect conditions for copy engine usage that were reported on coverity.
@hdelan
Copy link
Contributor Author

hdelan commented Jun 28, 2024

I have squashed #1765

@callumfare
Copy link
Contributor

The CI jobs all passed before the commits from #1765 were squashed, and the intel/llvm pipeline passed, so I'm merging this now rather than waiting for them to rerun

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
command-buffer Command Buffer feature addition/changes/specification conformance Conformance test suite issues. cuda CUDA adapter specific issues experimental Experimental feature additions/changes/specification hip HIP adapter specific issues level-zero L0 adapter specific issues loader Loader related feature/bug specification Changes or additions to the specification
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants