Transfer cost
The time spent moving data between machines, or between host and GPU memory. Often the reason an operation is not worth accelerating.
The reason a GPU can measure slower than a CPU on small work: the data has to get there and back. Worth checking whenever an operation that should be faster is not.
GPU & accelerationPerformance & optimization
Related terms
- GPU offloading: Moving computationally intensive operations onto a GPU. Eugo applies this automatically when a GPU is available and the data transfer is worth its cost.
- Serialization: Converting Python objects into bytes so they can be sent to another machine. Large task arguments are serialized and copied to every worker that needs them.