mem-cache: Aligned the BOP algorithm with RTL and added a flag.

[zephyrols]

1. Aligned the BOP

• aligned hash algorithm
• aligned tag position

2. Add flags for BOP

• enable_dynamic_external_offset: for opening add external offset
• enable_dynamic_depth: for opening dynamic prefetch depth
• enable_early_stop: for opening learning phase early stop

Change-Id: Ie5adde018f00d5440b63631a0dfc1c797f80373a

[zephyrols]

old-bop-hash is the baseline(commit: https://github.com/OpenXiangShan/GEM5/commit/c9bd2b3a38a1e776df6094bb9f5f6c7879b5e71c) v2 run with script kmh.py with no other param v3 run with script xiangshan.py with only one param --ideal-kmhv3

This can significantly enhance GemsFDTD. Other benchmarks have been increased or decreased to varying degrees. Overall, the performance has improved. The performance display of 50% checkpoints is as follows:

[XiangShanRobot]

[Generated by GEM5 Performance Robot] commit: a99f547ac750dca56a5ba73fed70dda0331a7d19 workflow: gem5 Ideal Performance Test

Ideal Performance

Overall Score

	PR	Master	Diff(%)
Score	19.56	19.59	-0.18 🔴

[XiangShanRobot]

[Generated by GEM5 Performance Robot] commit: a99f547ac750dca56a5ba73fed70dda0331a7d19 workflow: gem5 Performance Test

Standard Performance

Overall Score

	PR	Master	Diff(%)
Score	15.81	15.81	-0.04 🔴