https://hgpu.org/?p=9195
SemCache: Semantics-aware Caching for Efficient GPU Offloading