[Openmp-commits] [openmp] Add openmp support to System z (PR #66081)

[Ilya Leoshkevich via Openmp-commits]

iii-i wrote:

> Can you point me to some test output that shows your issue?

I'm not sure what changed, but in the meantime I cannot reproduce any of the Archer failures. The worst example used to be `projects/openmp/tools/archer/tests/races/taskwait-depend.c`: the race was not detected most of the time. The test simply printed x and y and exited with code 0. I had to manually set the retry count to 1000 to make it pass consistently, but now it succeeds every time. I will file an issue the next time I see something in that area.

> So I'm wondering: why is this then working on Intel, where the frame pointer is also omitted by default. Also, why just this one function and not its callers?

Its caller is the application.

I believe it works on Intel by accident: `-fomit-frame-pointer` does not say the frame pointer must be omitted, it only allows that. In some functions it's still there.

> This should be done as a stand-alone PR

Ok, I will look into this.

https://github.com/llvm/llvm-project/pull/66081