Move copy_to_host source to cpp file

Should fix some issues with duplicate symbols seen during builds.

Author: jakirkham

CHANGELOG.md

@@ -16,6 +16,7 @@
 - PR #215 Catch polymorphic exceptions by reference instead of by value
 - PR #221 Fix segfault calling rmmGetInfo when uninitialized
 - PR #225 Avoid invoking Python operations in c_free
+- PR #230 Fix duplicate symbol issues with `copy_to_host`
 
 
 

CMakeLists.txt

@@ -109,6 +109,7 @@ link_directories("${CMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES}" # CMAKE_CUDA_IMPLICIT
 
 add_library(rmm SHARED
             src/rmm.cpp
+            src/device_buffer.cpp
             src/memory_manager.cpp
             thirdparty/cnmem/src/cnmem.cpp
             src/mr/default_memory_resource.cpp)

include/rmm/device_buffer.hpp

@@ -393,28 +393,5 @@ class device_buffer {
                                     ///< allocate/deallocate device memory
 };
 
-/**--------------------------------------------------------------------------*
- * @brief Copies rmm::device_buffer to a preallocated host buffer.
- *
- * Copies device memory asynchronously on the specified stream
- *
- * @throws std::runtime_error if `hb` is `nullptr` or copy fails
- *
- * @param db `rmm::device_buffer` to copy to host
- * @param hb host allocated buffer to copy data to
- * @param stream CUDA stream on which the device to host copy will be performed
- *-------------------------------------------------------------------------**/
-void copy_to_host(const device_buffer& db, void* hb, cudaStream_t stream = 0) {
-  if (hb == nullptr) {
-    throw std::runtime_error{"Cannot copy to `nullptr`."};
-  }
-  cudaError_t err = cudaMemcpyAsync(hb,
-                                    db.data(),
-                                    db.size(),
-                                    cudaMemcpyDeviceToHost,
-                                    stream);
-  if (err != cudaSuccess) {
-    throw std::runtime_error{"Failed to copy to host."};
-  }
-}
+void copy_to_host(const device_buffer& db, void* hb, cudaStream_t stream = 0);
 }  // namespace rmm

src/device_buffer.cpp

@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2018, NVIDIA CORPORATION.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <rmm/device_buffer.hpp>
+
+namespace rmm {
+/**--------------------------------------------------------------------------*
+ * @brief Copies rmm::device_buffer to a preallocated host buffer.
+ *
+ * Copies device memory asynchronously on the specified stream
+ *
+ * @throws std::runtime_error if `hb` is `nullptr` or copy fails
+ *
+ * @param db `rmm::device_buffer` to copy to host
+ * @param hb host allocated buffer to copy data to
+ * @param stream CUDA stream on which the device to host copy will be performed
+ *-------------------------------------------------------------------------**/
+void copy_to_host(const device_buffer& db, void* hb, cudaStream_t stream) {
+  if (hb == nullptr) {
+    throw std::runtime_error{"Cannot copy to `nullptr`."};
+  }
+  cudaError_t err = cudaMemcpyAsync(hb,
+                                    db.data(),
+                                    db.size(),
+                                    cudaMemcpyDeviceToHost,
+                                    stream);
+  if (err != cudaSuccess) {
+    throw std::runtime_error{"Failed to copy to host."};
+  }
+}
+}  // namespace rmm