memcpy based on unified-addressing

termoshttrequested to merge
gpu-memcpy-unified into master

Simplify memcpy trait before try to implement async memcpy #4

  • Rewrite Memcpy trait implementations using auto-detection by cuMemcpy and CU_MEMORYTYPE_UNIFIED for cuMemcpy3D
    • Each memories (e.g. DeviceMemory) will be casted into a Rust slice to share the implementation
  • Remove Managed trait
Edited by termoshtt

Merge request reports