https://hgpu.org/?p=18604
Hashing, Caching, and Synchronization: Memory Techniques for Latency Masking Multithreaded Applications