[SYCL][Fusion][Tests] Test group functions, group algoriths and reduction

[victor-eds]

Add tests for fusing kernels with the same ND-range using group functions, group algorithms and reductions.

[victor-eds]

Failure unrelated to this change

[victor-eds]

@intel/llvm-gatekeepers can I get this merged, please?

[aelovikov-intel]

Looks like this new test fails in Nightly every day since its introduction: https://github.com/intel/llvm/actions/runs/7620444887/job/20755732583

@victor-eds , @sommerlukas , please fix or disable the test while investigation is in progress.

[sommerlukas]

Looks like this new test fails in Nightly every day since its introduction: https://github.com/intel/llvm/actions/runs/7620444887/job/20755732583

@victor-eds , @sommerlukas , please fix or disable the test while investigation is in progress.

Thanks for reporting this @aelovikov-intel, we'll investigate.

Are there any notable differences in the Nightly pipeline, as the test doesn't seem to fail in pre-commit here or post-commit since then?

[aelovikov-intel]

Are there any notable differences in the Nightly pipeline, as the test doesn't seem to fail in pre-commit here or post-commit since then?

Nothing applicable to this test comes to my mind. We have different XFAIL processing (see test-e2e/README), but that is irrelevant here.

[aelovikov-intel]

we'll investigate.

If the timeline is unclear, please disable the test for now.

[sommerlukas]

we'll investigate.

If the timeline is unclear, please disable the test for now.

We have a PR for disabling: #12540

[sommerlukas]

Are there any notable differences in the Nightly pipeline, as the test doesn't seem to fail in pre-commit here or post-commit since then?

Nothing applicable to this test comes to my mind. We have different XFAIL processing (see test-e2e/README), but that is irrelevant here.

As the problem seems to only occur for OpenCL CPU, the relevant difference might be that OpenCL CPU is only tested in nightly.

[aelovikov-intel]

As the problem seems to only occur for OpenCL CPU, the relevant difference might be that OpenCL CPU is only tested in nightly.

We do run opencl:cpu in pre-commit, it's part of the Intel Gen12 runner (uses three devices at once - opencl cpu/gpu + level_zero gpu).

I suspect the difference is that in Nightly it's the AMD CPU and not Intel for opencl:cpu target.