Bump UMF version and switch to dynamic linking

[igchor]

[pbalcer]

Let's target these changes for after 0.9.

[igchor]

Let's target these changes for after 0.9.

Sure, I just wanted to create a PR to do some initial testing

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 12.49%. Comparing base (78ef1ca) to head (beef35a). Report is 153 commits behind head on main.

Files	Patch %	Lines
source/common/umf_helpers.hpp	0.00%	4 Missing :warning:

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

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1430      +/-   ##
==========================================
- Coverage   14.82%   12.49%   -2.33%     
==========================================
  Files         250      239      -11     
  Lines       36220    36000     -220     
  Branches     4094     4086       -8     
==========================================
- Hits         5369     4498     -871     
- Misses      30800    31498     +698     
+ Partials       51        4      -47     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[igchor]

Testing PR on llvm: https://github.com/intel/llvm/pull/13343 Ther is currently an issue with missing hwloc on CI for llvm. I'm trying to address it.

[igchor]

OpenCL e2e test failure are unrelated to the PR (they fail on main as well).

[igchor]

There is one temporary commit on this PR to run e2e tests on my fork of intel/llvm. This is because I've changed how the common library is linked to the adapters/plugins and this requires changes in llvm (as you can see here: https://github.com/intel/llvm/pull/13343)

Also, the OpenCL failures are expected, those tests fail on main as well.

[igchor]

I've removed the tmp commit which mean e2e tests will be failing until we merge changes to intel/llvm. Last successful build is here: https://github.com/oneapi-src/unified-runtime/actions/runs/9842229066/job/27204428422?pr=1430

[igchor]

@kbenzie once https://github.com/intel/llvm/pull/13343 passes and is approved, this should be ready to merge

[kbenzie]

I guess we should wait to merge this until we have confirmation here. Once that's happened please add the ready to merge label @igchor and we'll get this merged.

[igchor]

@kbenzie we have confirmation that intel/llvm is ready for the change: https://github.com/intel/llvm/pull/13343#pullrequestreview-2198112559 @kswiecicki please bump the UR tag once https://github.com/oneapi-src/unified-memory-framework/pull/633 is merged and add ready to merge label

[ldrumm]

This patch does more than two things and has no description.

Why are we switching to dynamic linking? To which version are we bumping? Which parts of the code changes are really required by the bump, and which are in-passing code cleanups?

Right now this patch does everything without rationale. Please split it more logically if you can, remove the fly-by changes, and provide us with a description of what justifies such changes

[pbalcer]

Why are we switching to dynamic linking?

Because that's how UMF is meant to ship with the compiler by default. This has been discussed with the compiler team.

To which version are we bumping?

We are switching to what will become UMF version for the next release.

Which parts of the code changes are really required by the bump, and which are in-passing code cleanups?

All parts of this patch are required to bump up the UMF version. The actual code changes are all very minor fixes required for everything to compile after UMF interface changes. There are no cleanup or feature changes at all here.