rocr: Allocate AQL queue on device memory

- Use HSA_ALLOCATE_QUEUE_DEV_MEM=1 to create AQL queue in device
memory.
- Before writing AQL packet header to the queue use an SFENCE to ensure
that there is no reodering of the writes over PCIE

Change-Id: I5eacdc35108c4a1e245c75ae349b7495451aa60d

runtime/hsa-runtime/core/driver/kfd/amd_kfd_driver.cpp

@@ -117,6 +117,11 @@ KfdDriver::AllocateMemory(const core::MemoryRegion &mem_region,
            ? 1
            : kmt_alloc_flags.ui32.GTTAccess);
 
+  kmt_alloc_flags.ui32.Uncached =
+      (alloc_flags & core::MemoryRegion::AllocateUncached
+            ? 1
+            : kmt_alloc_flags.ui32.Uncached);
+
   if (m_region.IsLocalMemory()) {
     // Allocate physically contiguous memory. AllocateKfdMemory function call
     // will fail if this flag is not supported in KFD.