[SYCL][NATIVECPU][CI] Enable Native CPU without OCK on linux build

[PietroGhg]

This PR enables the Native CPU plugin in the build jobs for Linux. It is the first PR in series, with the goal of enabling full e2e testing for the Native CPU plugin on both Linux and Windows.

This initial PR builds the Native CPU plugin without the oneAPI Construction Kit. While the oneAPI Construction Kit is currently a key component for Native CPU, I would prefer to keep things as simple as possible, and enable CI testing for additional features incrementally, in order to gradually assess the impact of the Native CPU plugin in the overall CI workflow.

[PietroGhg]

Hi @intel/dpcpp-devops-reviewers, the main job affected by this PR (SYCL Pre Commit on Linux / build / Build + LIT (pull_request) ) is successful, but the check-sycl target has been skipped, do you know if there's a way for me to force running that target?

[aelovikov-intel]

Hi @intel/dpcpp-devops-reviewers, the main job affected by this PR (SYCL Pre Commit on Linux / build / Build + LIT (pull_request) ) is successful, but the check-sycl target has been skipped, do you know if there's a way for me to force running that target?

This change also triggered post-commit task and check-sycl seemed to be tested there, see https://github.com/intel/llvm/actions/runs/8326107523/job/22781117393?pr=13052.

[aelovikov-intel]

Should we do s/plug in/plugin/g in the PR description?

[PietroGhg]

Hi @intel/dpcpp-devops-reviewers, the main job affected by this PR (SYCL Pre Commit on Linux / build / Build + LIT (pull_request) ) is successful, but the check-sycl target has been skipped, do you know if there's a way for me to force running that target?

This change also triggered post-commit task and check-sycl seemed to be tested there, see https://github.com/intel/llvm/actions/runs/8326107523/job/22781117393?pr=13052.

Yeah you are right, but I see native_cpu=False in the log for the configure step for post commit, do you think I should enable the native cpu plugin for post commit on linux as part of this PR?

[aelovikov-intel]

do you think I should enable the native cpu plugin for post commit on linux as part of this PR

Probably. The main difference in pre- vs post- is that in pre-commit end-to-end tests we use multiple SYCL devices in a single invocation that results in XFAIL-marked tests to be simply skipped. Post-commit uses one SYCL device per llvm-lit invocation and XFAILs are properly tested there.

[PietroGhg]

The error in the post commit linux build pipeline is due to the fact that we run some SYCL executables in the tests in sycl/test/native_cpu, I'll address that in a separate PR and update this one accordingly

[PietroGhg]

@intel/dpcpp-devops-reviewers I've undrafted this PR since the Post Commit Linux Build job executed successfully (in the log you can see that the build was configured with --native_cpu and -DNATIVECPU_USE_OCK=Off), what do you think about these changes?

[PietroGhg]

Hi @aelovikov-intel, coming back to this PR, the pre-commit failure on CUDA is seen in other PRs e.g. #13695. The failure on the Arc job seems to appear also in this run, associated to commit 31f1f1c. I'm not sure about the two perf-test failures, on AMD and CUDA, they fail with cannot find -lstdc++ during cmake config, so the failure seems related to the runner config?