[SYCL] Fix test for the case when context has multiple gpu devices (#773)

Author: againull

SYCL/Basic/buffer/buffer_create.cpp

@@ -11,11 +11,15 @@ int main() {
   constexpr int Size = 100;
   queue Queue;
   auto D = Queue.get_device();
-
+  auto NumOfDevices = Queue.get_context().get_devices().size();
   buffer<::cl_int, 1> Buffer(Size);
   Queue.submit([&](handler &cgh) {
     accessor Accessor{Buffer, cgh, read_write};
-    if (D.get_info<info::device::host_unified_memory>())
+    if (NumOfDevices > 1)
+      // Currently the Level Zero plugin uses host allocations for multi-device
+      // contexts because such allocations are accessible by all devices.
+      std::cerr << "Multi GPU should use zeMemAllocHost\n";
+    else if (D.get_info<info::device::host_unified_memory>())
       std::cerr << "Integrated GPU should use zeMemAllocHost\n";
     else
       std::cerr << "Discrete GPU should use zeMemAllocDevice\n";
@@ -26,5 +30,5 @@ int main() {
   return 0;
 }
 
-// CHECK: {{Integrated|Discrete}} GPU should use [[API:zeMemAllocHost|zeMemAllocDevice]]
+// CHECK: {{Integrated|Multi|Discrete}} GPU should use [[API:zeMemAllocHost|zeMemAllocHost|zeMemAllocDevice]]
 // CHECK: ZE ---> [[API]](