diff --git a/lib/gc/ExecutionEngine/OpenCLRuntime/OpenCLRuntimeWrappers.cpp b/lib/gc/ExecutionEngine/OpenCLRuntime/OpenCLRuntimeWrappers.cpp
index f63d94316..6adaac2bd 100644
--- a/lib/gc/ExecutionEngine/OpenCLRuntime/OpenCLRuntimeWrappers.cpp
+++ b/lib/gc/ExecutionEngine/OpenCLRuntime/OpenCLRuntimeWrappers.cpp
@@ -21,7 +21,7 @@
 #ifdef _WIN32
 #define OCL_RUNTIME_EXPORT __declspec(dllexport)
 #else
-#define OCL_RUNTIME_EXPORT
+#define OCL_RUNTIME_EXPORT __attribute__((visibility("default")))
 #endif // _WIN32
 
 namespace {
@@ -73,7 +73,7 @@ struct CLExtTable {
   clSharedMemAllocINTEL_fn allocShared;
   clMemBlockingFreeINTEL_fn blockingFree;
   clSetKernelArgMemPointerINTEL_fn setKernelArgMemPtr;
-  clEnqueueMemcpyINTEL_fn enqueneMemcpy;
+  clEnqueueMemcpyINTEL_fn enqueueMemcpy;
   CLExtTable() = default;
   CLExtTable(cl_platform_id plat) {
     allocDev =
@@ -84,7 +84,7 @@ struct CLExtTable {
         (clMemBlockingFreeINTEL_fn)queryCLExtFunc(plat, MemBlockingFreeName);
     setKernelArgMemPtr = (clSetKernelArgMemPointerINTEL_fn)queryCLExtFunc(
         plat, SetKernelArgMemPointerName);
-    enqueneMemcpy =
+    enqueueMemcpy =
         (clEnqueueMemcpyINTEL_fn)queryCLExtFunc(plat, EnqueueMemcpyName);
   }
 };
@@ -390,6 +390,14 @@ extern "C" OCL_RUNTIME_EXPORT void gpuMemFree(GPUCLQUEUE *queue, void *ptr) {
   }
 }
 
+extern "C" OCL_RUNTIME_EXPORT void gpuMemCopy(GPUCLQUEUE *queue, void *dst,
+                                              void *src, uint64_t size) {
+  auto func = queue->ext_table_ ? queue->ext_table_->enqueueMemcpy
+                                : (clEnqueueMemcpyINTEL_fn)queryCLExtFunc(
+                                      queue->device_, EnqueueMemcpyName);
+  CL_SAFE_CALL(func(queue->queue_, true, dst, src, size, 0, nullptr, nullptr));
+}
+
 extern "C" OCL_RUNTIME_EXPORT cl_program
 gpuModuleLoad(GPUCLQUEUE *queue, const unsigned char *data, size_t dataSize) {
   if (queue) {