[OpenMP] Add missing test for pinned memory API

[carlo.bertolli]

Summary

I accidentally left out the test for the pinned API introduced by D138933. Adding it back.

[jdoerfert]

Should this have a target? Do you want to check a and b after?

[carlo.bertolli]

Updated test with checks of a and b pointers.

[carlo.bertolli]

In D140077#3996828, @jdoerfert wrote:

Should this have a target? Do you want to check a and b after?

Good points. This is just the host API for now, so no way to test a target. I do have a test for this in a dependent patch (https://reviews.llvm.org/D139208). I will update that tomorrow to make the missing link between this api and the libomptarget implementation.
I added checks for a and b after allocation and initialization.

[jdoerfert]

LG, one nit, is pinned_alloc2 really pinned? If not, maybe rename it.

openmp/runtime/test/api/omp_pinned.c

+// RUN: %libomp-compile-and-run
+// RUN: env OMP_ALLOCATOR=omp_default_mem_space:pinned=true %libomp-run
+
+#include <omp.h>
+
+int main() {
+  omp_alloctrait_t pinned_trait[1] = {{omp_atk_pinned, omp_atv_true}};
+  omp_allocator_handle_t pinned_alloc =
+      omp_init_allocator(omp_default_mem_space, 1, pinned_trait);
+  omp_allocator_handle_t pinned_alloc2 = omp_get_default_allocator();
+  double *a = (double *)omp_alloc(10 * sizeof(double), pinned_alloc);
+  double *b = (double *)omp_alloc(10 * sizeof(double), pinned_alloc2);
+
+#pragma omp parallel for
+  for (int i = 0; i < 10; i++) {
+    a[i] = 0;
+    b[i] = 1;
+  }
+
+  omp_free(a, pinned_alloc);
+  omp_free(b, pinned_alloc2);
+
+  return 0;
+}