[OPENMP][NVPTX]Perform memory flush if number of threads to sync is 1 or less.

[ABataev]

Summary

According to the OpenMP standard, barrier operation must perform
implicit flush operation. Currently, if there is only one thread in the
team, barrier does not flush the memory. Patch fixes this problem.

[ABataev]

Addressed comments.

[ABataev]

Updated test.

[ABataev]

Rebase

[grokos]

This looks consistent with the changes in D62397, I think it's good to go.

libomptarget/deviceRTLs/nvptx/src/sync.cu

if (numberOfActiveOMPThreads > 1) {
             WARPSIZE * ((numberOfActiveOMPThreads + WARPSIZE - 1) / WARPSIZE);
 
         PRINT(LD_SYNC,
               "call kmpc_barrier with %d omp threads, sync parameter %d\n",
               (int)numberOfActiveOMPThreads, (int)threads);
         // Barrier #1 is for synchronization among active threads.
         named_sync(L1_BARRIER, threads);
       }
+    } else {
+      // Still need to flush the memory per the standard.
+      __kmpc_flush(loc_ref);
     } // numberOfActiveOMPThreads > 1
     PRINT0(LD_SYNC, "completed kmpc_barrier\n");
   }
 }
 
 // Emit a simple barrier call in SPMD mode.  Assumes the caller is in an L0
 // parallel region and that all worker threads participate.
 EXTERN void __kmpc_barrier_simple_spmd(kmp_Ident *loc_ref, int32_t tid) {

libomptarget/deviceRTLs/nvptx/test/parallel/barrier.c

+// RUN: %compile-run-and-check
+
+#include <omp.h>
+#include <stdio.h>
+
+int main(int argc, char *argv[]) {
+  int data, out, flag = 0;
+#pragma omp target teams num_teams(2) map(tofrom                               \
+                                          : out) map(to                        \
+                                                     : data, flag)             \
+    thread_limit(1)
+#pragma omp parallel num_threads(1)
+  {
+    if (omp_get_team_num() == 0) {
+      /* Write to the data buffer that will be read by thread in team 1 */
+      data = 42;
+/* Flush data to thread in team 1 */
+#pragma omp barrier
+      /* Set flag to release thread in team 1 */
+#pragma omp atomic write
+      flag = 1;
+    } else if (omp_get_team_num() == 1) {
+      /* Loop until we see the update to the flag */
+      int val;
+      do {
+#pragma omp atomic read
+        val = flag;
+      } while (val < 1);
+      out = data;
+#pragma omp barrier
+    }
+  }
+  // CHECK: out=42.
+  /* Value of out will be 42 */
+  printf("out=%d.\n", out);
+  return !(out == 42);
+}