[libcxx-commits] [libcxxabi] Adding Separate OpenMP Offloading Backend to `libcxx/include/__algorithm/pstl_backends` (PR #66968)

[Anton Rydahl via libcxx-commits]

AntonRydahl wrote:

I have also added a number of tests that verify that vectors of primitive data types are offloaded to the GPU. They use `omp_is_initial_device()` to test whether the loop bodies run on the host or the device. 

If there are no available devices, the new tests simply don't run:
```c++
  if (omp_get_num_devices() < 1)
    return 0;
``` 
Also, I have introduced a new feature such that they only run when the OpenMP PSTL backend is enabled:
```c++
// REQUIRES: openmp_pstl_backend
```

https://github.com/llvm/llvm-project/pull/66968