[UR] Enable UMF tracking by default

[igchor]

It is needed by adapters's urUSMFree implementations where we rely on umfPoolByPtr (which always returns null if tracking is disabled).

@kbenzie should I also make a PR to the main branch? We plan to move the UMF to a separate repo anyway and I expect this option (UMF_ENABLE_POOL_TRACKING) to not be publicly exposed to UR users anymore.

[igchor]

I think this is ready to be merged. The failing tests turned out to just be tests that unexpectedly succeeded (most of the CUDA and L0 tests with UsePoolEnabled work after this change).

[lukaszstolarczuk]

At this moment, if someone sets UMF_ENABLE_POOL_TRACKING manually to OFF, these tests will fail. Do we expect anyone to do that? Perhaps we should remove this option from CMake, after all?

[igchor]

@lukaszstolarczuk you are right, but I think I would actually prefer to do this after we merge / cherry-pick this change. Removing the option would also mean that we can remove some of the existing code so the change would be bigger.

[jandres742]

Fixes to test-usm in L0 in #1105 depending on this patch.

[jandres742]

@igchor :
is there anything extra to do on this PR, or could we ask @kbenzie to move it to "ready to merge" ?

[kbenzie]

@kbenzie should I also make a PR to the main branch? We plan to move the UMF to a separate repo anyway and I expect this option (UMF_ENABLE_POOL_TRACKING) to not be publicly exposed to UR users anymore.

I think we can just target adapters for now. I'm hoping we can merged adapters into main branch soon and then stop using the adapters branch so this problem will go away.

[kbenzie]

@igchor :
is there anything extra to do on this PR, or could we ask @kbenzie to move it to "ready to merge" ?

Might be prudent to rebase on top of adapters to make sure its working with the latest set of changes.

[igchor]

@igchor :
is there anything extra to do on this PR, or could we ask @kbenzie to move it to "ready to merge" ?

Might be prudent to rebase on top of adapters to make sure its working with the latest set of changes.

Done. Do we want to wait until changes from #1084 are cherry-picked? If we merge this first, there might be some extra Windows failures.

[kbenzie]

Done. Do we want to wait until changes from #1084 are cherry-picked? If we merge this first, there might be some extra Windows failures.

If that's a dependency, then yes.

[kbenzie]

Done. Do we want to wait until changes from #1084 are cherry-picked? If we merge this first, there might be some extra Windows failures.

If that's a dependency, then yes.

This will be included in #1072 which we'll aiming to merged to adapters next.

[codecov-commenter]

Codecov Report

Attention: 22 lines in your changes are missing coverage. Please review.

Comparison is base (fe2735a) 15.79% compared to head (3276f70) 15.79%.
Report is 4 commits behind head on adapters.

Files	Patch %	Lines
test/conformance/testing/include/uur/fixtures.h	0.00%	15 Missing ⚠️
test/conformance/kernel/urKernelSetArgSampler.cpp	0.00%	4 Missing ⚠️
...ance/kernel/urKernelSetSpecializationConstants.cpp	0.00%	2 Missing ⚠️
test/conformance/kernel/urKernelSetArgPointer.cpp	0.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@             Coverage Diff              @@
##           adapters    #1071      +/-   ##
============================================
- Coverage     15.79%   15.79%   -0.01%     
============================================
  Files           223      223              
  Lines         31351    31365      +14     
  Branches       3511     3511              
============================================
+ Hits           4951     4953       +2     
- Misses        26349    26361      +12     
  Partials         51       51              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kbenzie]

@igchor any idea why the checks failed?

[igchor]

@kbenzie there was a conflict in test/conformance/kernel/kernel_adapter_level_zero.match. It should be fixed now and there's one less test failing for L0 kernel test.

[kbenzie]

Thanks @igchor

[fabiomestre]

I have updated the target branch of this PR from the adapters branch to the main branch.
Development in UR is moving back to main. The adapters branch will soon be deleted.