Add 'regression' tests to SYCL/InvokeSimd

[fveselov]

'Regression' category tests

[sarnex]

LGTM, we can merge once CI passes :)

[v-klochkov]

I have looked through this huge PR. It looks good to me. Thanks to @sarnex for being main reviewer of this PR.

Maybe just one question that can be fixed easily.
The folder names used in InvokeSimd folder use 3 different styles:

• SomeFolder // GOOD // e.g. 'InvokeSimd'
• something // Ok // e.g. 'regression'. IMO 'Regression would be more consistent, but Ok.
• SOME_FOLDER // not Ok, IMO // e.g. 'IMPL_SUBGROUP'. Can it be renamed to 'ImplicitSubgroup', please?

[fveselov]

I have looked through this huge PR. It looks good to me. Thanks to @sarnex for being main reviewer of this PR.

Maybe just one question that can be fixed easily. The folder names used in InvokeSimd folder use 3 different styles:

• SomeFolder // GOOD // e.g. 'InvokeSimd'
• something // Ok // e.g. 'regression'. IMO 'Regression would be more consistent, but Ok.
• SOME_FOLDER // not Ok, IMO // e.g. 'IMPL_SUBGROUP'. Can it be renamed to 'ImplicitSubgroup', please?

Agree, renamed both 'regression' and 'IMPL_SUBGROUP' directories.

I also created one more PR (#1562) doing a similar renames for the "Feature" and "Spec" categories of InvokeSimd.

[fveselov]

Windows and Linux CIs failures are unrelated to changes:

[2023-02-01T10:35:43.252Z] Failed Tests (3):
[2023-02-01T10:35:43.252Z] SYCL :: ESIMD/api/svm_gather_scatter_scalar_off.cpp
[2023-02-01T10:35:43.252Z] SYCL :: ESIMD/dword_atomic_cmpxchg_scalar_off.cpp
[2023-02-01T10:35:43.252Z] SYCL :: ESIMD/dword_atomic_smoke_scalar_off.cpp

[v-klochkov]

Windows and Linux CIs failures are unrelated to changes:

[2023-02-01T10:35:43.252Z] Failed Tests (3):
[2023-02-01T10:35:43.252Z] SYCL :: ESIMD/api/svm_gather_scatter_scalar_off.cpp
[2023-02-01T10:35:43.252Z] SYCL :: ESIMD/dword_atomic_cmpxchg_scalar_off.cpp
[2023-02-01T10:35:43.252Z] SYCL :: ESIMD/dword_atomic_smoke_scalar_off.cpp

@fineg74 - please take a look at these fails, they must be caused by one of your recent changes in these tests.
Well, it also possible that the fails showed up because the older/yesterday version of compiler was used to run this llvm-test-suite changes.

[fineg74]

@v-klochkov It seems the compiler change for scalar offset didn't go into latest nightly build for some reason that is used to run tests. The latest commit that went into nightly build was 8:23 AM commit while scalar offset change was merged at 15:35
Will follow up tomorrow if the problem persists