I am working on an application that uses both the CUDA Driver API and Runtime API. I want to confirm if using cuCtxSynchronize (Driver API) to synchronize a buffer, followed by cudaMemcpy (Runtime API) to copy the data, ensures proper synchronization between the two APIs.
Does the cuCtxSynchronize guarantee that all operations on the buffer from the Driver API are completed before the cudaMemcpy is executed?
Any clarification on the interoperability and synchronization behavior between these APIs would be greatly appreciated.
Thank you!