CUDA：未解决的外部函数'cudaMemcpyAsync'

Question

我不断收到编译错误：

ptxas fatal   : Unresolved extern function 'cudaMemcpyAsync'

在文件buffer.cuh中：

__device__ void markBuffer(volatile bool* is_ready_for_write_list, void* volatile * data_list, void* data, size_t num_samples_per_read, size_t sample_offset, void** tensor_list) {
    size_t index = getIndex(num_samples_per_read, sample_offset);

    data_list[index] = data;
    cudaMemcpyAsync((tensor_list)[index], ((int*)(data_list[index])) - (num_samples_per_read - 1), num_samples_per_read * 4, cudaMemcpyDeviceToDevice, NULL);
    is_ready_for_write_list[index] = true;
}

buffer.cuh被包含在nv_wavenet.cuh中，而nv_wavenet.cuh被包含在nv_wavenet_test.cu中。

我正在使用：

nvcc -arch=sm_61 -std=c++11  -g --use_fast_math -G -g -maxrregcount 128 nv_wavenet_test.cu matrix.cpp nv_wavenet_reference.cpp -o nv_wavenet_test