Temporarily disabled unnormalized-clamp-linear-float.cpp

[DenisBakhvalov]

Related issue: intel/llvm#8814

[cperkinsintel]

@DenisBakhvalov - this test is already marked XFAIL for LevelZero and Windows. Why is it necessary to further disable it? If we don't want the test run at all, we can switch XFAIL: to UNSUPPORTED:

[DenisBakhvalov]

@cperkinsintel, based on what I saw, it sporadically passes on Windows + LevelZero, even though it is supposed to fail.

[cperkinsintel]

@DenisBakhvalov , well let's just switch XFAIL: to UNSUPPORTED: if "sometimes passing" is a problem. Then the test won't be run at all.
Adding an outside requirement just makes it more complicated, especially since we want to maintain this test for other devices and platforms to catch any regressions. There are outstanding tickets for this problem. Once the L0 team get their fixes in, it'd be nice to be able to reactive this test easily.

[DenisBakhvalov]

@DenisBakhvalov , well let's just switch XFAIL: to UNSUPPORTED: if "sometimes passing" is a problem. Then the test won't be run at all.
Adding an outside requirement just makes it more complicated, especially since we want to maintain this test for other devices and platforms to catch any regressions. There are outstanding tickets for this problem. Once the L0 team get their fixes in, it'd be nice to be able to reactive this test easily.

OK.

[cperkinsintel]

LGTM. Thanks!

[dm-vodopyanov]

Just ran checks in #202 one more time, and now this test fails with OpenCL on Linux:
http://134.191.240.166:8010/blue/organizations/jenkins/LLVM-Test-Suite-CI%2FLLVM-Test-Suite-Lin/detail/LLVM-Test-Suite-Lin/455/pipeline#step-177-log-475

[DenisBakhvalov]

Just ran checks in #202 one more time, and now this test fails with OpenCL on Linux:
http://134.191.240.166:8010/blue/organizations/jenkins/LLVM-Test-Suite-CI%2FLLVM-Test-Suite-Lin/detail/LLVM-Test-Suite-Lin/455/pipeline#step-177-log-475

I think it's a different one. The failing test in your CI is Sampler/normalized-none-linear-float.cpp. Nevertheless, the issue seems a little annoying.

[cperkinsintel]

@dm-vodopyanov - I'll take care of it. We had to adjust the nonnormalized test last week. I'll update this one too. The OCL:CPU has an error where it gets flaky results on certain edge cases. There's an outstanding ticket on it, but I'll adjust the test in the interim.

[dm-vodopyanov]

Thanks a lot Chris! Denis, yes, sorry, this is completely another test, I didn't match the whole names.