Move asynchronous zero (#226)

CHANGELOG.md

@@ -8,6 +8,7 @@ This project adheres to [Semantic Versioning](http://semver.org/).
 ### Added
 ### Changed
 - Made the library header only
+- Moved asynchronous `::zero` from `Device` to `Stream`
 - Replaced `include_cuda_code` helper with `target_embed_source`
 ### Removed
 

include/cudawrappers/cu.hpp

@@ -415,8 +415,6 @@ class DeviceMemory : public Wrapper<CUdeviceptr> {
 
   void zero(size_t size) { checkCudaCall(cuMemsetD8(_obj, 0, size)); }
 
-  void zero(size_t size, Stream &stream);
-
   const void *parameter()
       const  // used to construct parameter list for launchKernel();
   {
@@ -484,6 +482,10 @@ class Stream : public Wrapper<CUstream> {
     checkCudaCall(cuMemPrefetchAsync(devPtr, size, dstDevice, _obj));
   }
 
+  void zero(CUdeviceptr devPtr, size_t size) {
+    checkCudaCall(cuMemsetD8Async(devPtr, 0, size, _obj));
+  }
+
   void launchKernel(Function &function, unsigned gridX, unsigned gridY,
                     unsigned gridZ, unsigned blockX, unsigned blockY,
                     unsigned blockZ, unsigned sharedMemBytes,
@@ -534,10 +536,6 @@ class Stream : public Wrapper<CUstream> {
   }
 };
 
-inline void DeviceMemory::zero(size_t size, Stream &stream) {
-  checkCudaCall(cuMemsetD8Async(_obj, 0, size, stream));
-}
-
 inline void Event::record(Stream &stream) {
   checkCudaCall(cuEventRecord(_obj, stream._obj));
 }

tests/test_cu.cpp

@@ -87,7 +87,7 @@ TEST_CASE("Test zeroing cu::DeviceMemory", "[zero]") {
 
     cu::Stream stream;
     stream.memcpyHtoDAsync(mem, src, size);
-    mem.zero(size, stream);
+    stream.zero(mem, size);
     stream.memcpyDtoHAsync(tgt, mem, size);
     stream.synchronize();