Clean up TensorDeviceThreadPool.h
- Gets rid of unused or mostly unused methods in TensorDeviceThreadPool.h
- Reduces the amount of type erasure in
parallelForand enables perfect forwarding of parameter packs inenqueuewhen c++20 is used. - Adds missing
std::movein TensorExecutor.h. - Simplifies TensorReduction.h.
Edited by Rasmus Munk Larsen