[OpenMP][test]Flip bit-fields in 'struct flags' for big-endian in test cases

[xingxue-ibm]

This patch flips bit-fields in struct flags for big-endian in test cases to be consistent with the definition of the structure in libomp kmp.h.

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[iii-i]

The change looks good to me.

One thing though: grepping for 'struct kmp_depend_info' shows two other places where we have these bit fields: openmp/libomptarget/include/OpenMP/InternalTypes.h and openmp/runtime/test/tasking/hidden_helper_task/common.h. Would it make sense to make this change there as well?

[xingxue-ibm]

The change looks good to me.

One thing though: grepping for 'struct kmp_depend_info' shows two other places where we have these bit fields: openmp/libomptarget/include/OpenMP/InternalTypes.h and openmp/runtime/test/tasking/hidden_helper_task/common.h. Would it make sense to make this change there as well?

Thanks for checking, @iii-i! Changed openmp/runtime/test/tasking/hidden_helper_task/common.h for big-endian as well. openmp/libomptarget/include/OpenMP/InternalTypes.h seems to be a header for the target runtime. Since I won't be able to test changes, I'd refrain from touching it and will leave it for future if the target runtime needs to support big-endian.

[xingxue-ibm]

Gentle ping...Are there any further comments or concerns?

[kkwli]

LGTM

[jprotze]

I suggest to add a comment in kmp.h, that provides the hint of explicit copy of the struct in tests and that any changes to the struct should be applied to the tests.
Can you add this comment as part of your patch?
Overall looks good to me.

[jprotze]

I was wondering whether this test is compiler(clang) specific, or is this portable?
Since kmp.h has the same code, it make sense to replicate the pattern here.

[xingxue-ibm]

I was wondering whether this test something that is compiler(clang) specific, or is this portable?

Checked compilers in Compiler Explorer, the test works with gcc, clang, icc, icx, and zig cc. Does not seem to work with msvc.

[jprotze]

This is something that would also need to be addressed in kmp.h.

Does this versioning have a correctness implication, or is this a performance optimization?
In the latter case, from my perspective, a viable approach would be hardcode the endianess, if the two macros are undefined.

It's your choice to fix it as part of this patch or whether you prefer to fix it in a separate pr:)

[xingxue-ibm]

I would do it in a separate PR.

[xingxue-ibm]

I suggest to add a comment in kmp.h, that provides the hint of explicit copy of the struct in tests and that any changes to the struct should be applied to the tests. Can you add this comment as part of your patch? Overall looks good to me.

Thanks very much, @jprotze! Yeah, it will be helpful to mention that test cases have an explicit copy of the struct. I will add the comment to kmp.h.

[jprotze]

I'm not sure whether you have the permissions to merge, if not just leave a comment and Kelvin or I will hit the merge button

[xingxue-ibm]

I'm not sure whether you have the permissions to merge, if not just leave a comment and Kelvin or I will hit the merge button

Thanks, I have the permission.