🦈 IMP: Prefetching Transposition Table Entries

[TheBlackPlague]

🎯 Summary

This PR aims to improve memory access performance by a notable amount. This is done by prefetching the Transposition Table Entry couple instructions before it is needed. The CPU fetches it into the L1 cache (as specified by the Prefetch hint) in parallel with the instructions that would run just before the entry is needed so that when it is required, it may be taken instantly from the L1 cache (high-speed access).

👏 Acknowledgements

• Jay Honnold: Jay initially suggested this for StockNemo; however, due to the nature of the .NET Runtime and its tendency to move memory around, the prefetching wasn't practical (and never gained ELO). This changes in StockDory's C++ World. Awesome suggestion, Jay! 🙌

📈 ELO

STC:

ELO   | 12.71 +- 7.53 (95%)
SPRT  | 10.0+0.10s Threads=1 Hash=16MB
LLR   | 2.97 (-2.94, 2.94) [0.00, 5.00]
GAMES | N: 4760 W: 1474 L: 1300 D: 1986

LTC:

ELO   | 14.66 +- 7.98 (95%)
SPRT  | 60.0+0.60s Threads=1 Hash=256MB
LLR   | 3.00 (-2.94, 2.94) [0.00, 5.00]
GAMES | N: 3888 W: 1121 L: 957 D: 1810