riscv-boom's loop
loopThe Loop Predictor specializes in predicting loop exit branches.It learns the iteration count of loops and predicts the loop exit when the learned count is reached. Structure: Small table (16 set
riscv-boom's tage
TAGE is the most advanced branch predictor in BOOM. It uses multiple prediction tables indexed by different history lengths to achieve high accuracy. Key Concepts: Multiple tables with geometrically
riscv-boom's btb
btbThe Branch Target Buffer (BTB) caches branch target addresses to provide fast prediction of where branches will go. Structure: Set-associative cache (default: 128 sets, 2 ways) Each entry stores:
riscv-boom's bim
bimThis module implements a Bimodal Branch Predictor using 2-bit saturating counters. The prediction is based on the MSB of the counter: Counter = 0 (00): Strongly Not Taken Counter = 1 (01
riscv-boom's ubtb
ubtbThis is the set-associative variant of the Micro BTB, as opposed to the fully-associative FAMicroBTB. It uses indexed addressing like a cache. Structure: 256 sets by default (configurable) Each e
Branch Prediction for riscv-boom
Branch PredictionBOOM uses two levels of branch prediction: a fast Next-Line Predictor (NLP) a slower but more complex Backing Predictor (BPD) In this case, the NLP is a Branch Target Buffer and the
预测器关键参数
分支预测器的性能指标上一篇文章研究了一个供学习研究的分支预测器 biriscv_npc，顺便量化了这个分支预测器的性能（唯一指标就是分支准确率，已经包含各个分支指令执行次数的权重，想想为什么）。 这篇文章来研究各个参数对于分支预测器的影响，以及为什么影响。 BTB 容量处理器的 BTB 大小是有限的（比如 512 个条目）。当需要存入第 513 个分支指令的预测信息时，必须覆盖掉某一个旧条目。选择
biriscv_npc 性能分析
分支预测器性能统计基本信息当 branch_request_i 拉高： RAS 学习如果 BTB 命中或者不命中： btb_pc_q 更新：btb_pc_q[btb_wr_entry_r] <= branch_source_i btb_is_call_q 更新：btb_is_call_q[btb_wr_entry_r]<= branch_is_call_i btb_is_re