As we know we share cudaMalloc buffers between different processes through IPC. Can we share cudaMallocManged buffers?
Hi,
Unified memory use same data point for CPU and GPU memory.
But cudaDeviceSynchronize() is required to switching the process.
You can check our document for examples:
[url]https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#um-unified-memory-programming-hd[/url]
Thanks.