unified-runtime Wrap linker flags on Windows for IntelLLLVM

The Intel C++ compiler requires linker flags to be wrapped, because CMake passes them through the compiler driver. Prefix linker options with LINKER:. CMake will transorm it to the appropriate flag for the compiler driver: (Nothing for MSVC or clang-cl, and /Qoption,link for icx), so this will work for other compilers and for earlier CMake versions too.

[!NOTE]
Requires updating the fetched versions of UMF and Level Zero to versions that include the fix for the same problem. I tested locally with -DUMF_TAG=e6ff45e1636bd172117bab6d9f4d00638113f592 -DUR_LEVEL_ZERO_LOADER_TAG=v1.18.1.

Fixes: #2178

Oct 08 '24 07:10 Maetveis

@lukaszstolarczuk @ldorau @lisanna-dettwyler

Oct 08 '24 07:10 Maetveis

I can try and add an icx build in the CI, if you want...? (I'm doing such workflow for UMF at the moment)

Oct 08 '24 08:10 lukaszstolarczuk

Ping, can we get this merged? I don't have merge access myself.

Oct 18 '24 10:10 Maetveis

Ping, can we get this merged? I don't have merge access myself.

@oneapi-src/unified-runtime-level-zero-write still need to approve this.

Oct 18 '24 10:10 kbenzie

@kbenzie can you approve the workflows? I rebased to see if the CI failures go away or if I will need to investigate.

Oct 23 '24 12:10 Maetveis

Ping @oneapi-src/unified-runtime-level-zero-write

Oct 23 '24 12:10 jsji

@kbenzie can you approve the workflows? I rebased to see if the CI failures go away or if I will need to investigate.

The pull request labeler job is expected to fail for PRs from people who aren't a memory of the UR traige team. If there are e2e failures that doesn't necessarily block merging as long as we can determine they are not related to these changes.

Oct 23 '24 13:10 kbenzie

@nrspruit @kbenzie Can this get merged? I'm fairly certain the CI failures are not caused by this change.

Oct 30 '24 15:10 Maetveis

just going to rebase to see if we can get a clean run of the e2e l0 job

Oct 30 '24 16:10 aarongreig

@aarongreig bump, should I look into the failures? I am not familiar with L0 and I don't expect runtime failures from this. This blocking some other future work for me.

Nov 05 '24 12:11 Maetveis

No, those failures are expected.

Nov 05 '24 12:11 pbalcer

No, those failures are expected.

Okay, Thank you. Lets get this merged then pretty please :).

Nov 05 '24 13:11 Maetveis

ping @callumfare @nrspruit @pbalcer

Nov 08 '24 11:11 Maetveis

@Maetveis Please resolve the conflicts and I'll make sure this gets merged soon

Nov 08 '24 12:11 callumfare

@Maetveis Please resolve the conflicts and I'll make sure this gets merged soon

Thanks! Done.

Nov 08 '24 13:11 Maetveis

Since #2100 was reverted and it had a CMake fix that was needed here (the source of the conflict previously) I restored this PR to how it was before conflict resolution.

Ping @callumfare, please merge this ASAP.

Nov 12 '24 08:11 Maetveis