I want some parts are pinned on host, the others are pageable. For example,
cudaMallocManaged(&data, size, …)
cudaMemAdvise(data+offset, pageSize, cudaMemAdviseSetPreferredLocation, cpuid)
cudaMemAdvise(data+offset, pageSize, cudaMemAdviseSetAccessedBy, gpuid)
But the executed results is not the same as the one using cudaHostAlloc(&data, size, …)
Is any way to achieve this?