added mutlti device context test

Author: t4c1

SYCL/Basic/multi_device_context.cpp

@@ -0,0 +1,67 @@
+// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %HOST_RUN_PLACEHOLDER %t.out
+// RUN: %CPU_RUN_PLACEHOLDER %t.out
+// RUN: %GPU_RUN_PLACEHOLDER %t.out
+
+// REQUIRES: cuda
+
+#include <stdlib.h>
+#include <sycl.hpp>
+
+sycl::event add(sycl::queue q, sycl::buffer<int> buff, int *usm,
+                sycl::event e) {
+  return q.submit([&](sycl::handler &cgh) {
+    auto acc = buff.get_access<sycl::access::mode::read_write>(cgh);
+    cgh.depends_on(e);
+    cgh.single_task([=]() { acc[0] += *usm; });
+  });
+}
+
+int main() {
+  sycl::platform plat = sycl::platform::get_platforms()[0];
+  auto devices = plat.get_devices();
+  if (devices.size() < 2) {
+    std::cout << "Need two devices for the test!" << std::endl;
+    return 0;
+  }
+
+  sycl::device dev1 = devices[0];
+  sycl::device dev2 = devices[1];
+
+  sycl::context ctx{{dev1, dev2}};
+
+  sycl::queue q1{ctx, dev1};
+  sycl::queue q2{ctx, dev2};
+
+  int a = 1;
+  int b = 2;
+  int c = 4;
+  int d = 5;
+  {
+    sycl::buffer<int> buff1(&a, 1);
+    sycl::buffer<int> buff2(&b, 1);
+
+    // test copying usm
+    int *usm1 = sycl::malloc_device<int>(1, q1);
+    int *usm2 = sycl::malloc_device<int>(1, q2);
+    sycl::event e1 = q1.memcpy(usm1, &c, 1);
+    sycl::event e2 = q2.memcpy(usm2, &d, 1);
+
+    // test combination of usm and buffers in a kernel
+    sycl::event e3 = add(q1, buff1, usm1, e1);
+    sycl::event e4 = add(q2, buff2, usm2, e2);
+
+    // change values in usm to ensure results are distinct
+    sycl::event e5 = q1.memcpy(usm1, &d, 1, e3);
+    sycl::event e6 = q2.memcpy(usm2, &c, 1, e4);
+
+    // use each buffer on the other device than before - tests that copying
+    // between devices works
+    add(q1, buff2, usm1, e5);
+    add(q2, buff1, usm2, e6);
+  }
+  assert(a == 1 + 2 * 4);
+  assert(b == 2 + 2 * 5);
+
+  return 0;
+}