avatar
Articles
110
Tags
24
Categories
52
Home
Archives
Tags
Categories
About
Logoxixi's blog
Home
Archives
Tags
Categories
About

xixi's blog

RIP-Tagged Region Prefetcher(AMD US12204459B2)[2022]: Data cache region prefetcher
Created2026-04-27|cpucacheprefetchdata-pfspatial
AMD 专利(US12204459B2) [2022] 解读:RIP-Tagged Region Prefetcher - 用 RIP 和 cache-line offset 标记 region footprint,训练非顺序区域访问并在后续同 RIP miss 时触发预取。
Region Pattern Matching(AMD US12360907B2)[2025]: Region pattern-matching hardware prefetcher
Created2026-04-27|cpucacheprefetchdata-pfspatial
AMD 专利(US12360907B2) [2025] 解读:Region Pattern Matching - 将 region 内访问 footprint 记录成 subdivision bitmap,并通过 region type / direct pattern table 在相似 region 上重放预取。
Cache Architechture-Aware Cache Replacement
Created2026-04-23|cpucachereplacement
Cache 架构感知的替换策略简介
Prefetch-Aware Cache Replacement
Created2026-04-23|cpucachereplacement
预取数据通常构成 cache 的很大一部分,所以与预取器的交互是设计 cache 替换策略时需要考虑的重要因素 设计预取感知 cache 替换策略有两个主要目标: 替换策略应避免 cache pollution,这种污染是由不准确的预取引起的 替换策略应优先淘汰那些可以被预取的 cacheline ,而不是那些难以预取的 cacheline Belady 的 MIN 算法在无预取的情况下被证明是最优的,但在存在预取器的情况下是不完备的,因为它忽略了第二个设计目标:降低可预取 cacheline 的优先级 降低 Cache Pollution此类解决方案可分为两大类: 第一类从预取器获取反馈,以识别可能不准确的预取请求 此类策略通常需要与预取器进行明确的协同设计,或对现有预取器进行修改 例 1,Ishii 等人利用 AMPM 预取器的内部状态来告知预取 cacheline 的插入优先级 例 2,KPC 通过协同设计预取器来提供关于置信度和预计重用时间的反馈。这些信息随后被用于决定是将预取插入 L2 Cache 还是 L3 Cache,同时也用于确定插入 cacheline...
Multi-Core-Aware Cache Replacement
Created2026-04-23|cpucachereplacement
多核感知的 Cache 管理与替换策略
Criticality-Driven Cache Optimizations
Created2026-04-23|cpucachereplacement
Criticality 感知的 Cache 替换策略简介
Cost-Aware Cache Replacement Policies
Created2026-04-22|cpucachereplacement
Cost 感知的 Cache 替换策略简介
Bingo (2019 HPCA): Bingo Spatial Data Prefetcher
Created2026-04-17|cpucacheprefetchdata-pf
2019 HPCA Bingo 论文解读:用一张统一的 history table 通过长事件/短事件双重查找来兼顾 spatial prefetch 的 accuracy 和 coverage
Kairos (2025 MICRO): Elevating Temporal Prefetching Through Instruction Correlation
Created2026-04-17|cpucacheprefetchdata-pf
2025 MICRO Kairos 论文解读:通过指令关联识别 key IP,提升片上 temporal prefetcher 的元数据利用率
Streamline (2026 HPCA): Streamlined On-Chip Temporal Prefetching
Created2026-04-17|cpucacheprefetchdata-pf
2026 HPCA Streamline 论文解读:用 stream-based 元数据表示替代 pairwise 表示,提升片上时序预取器的存储效率、预取覆盖率和带宽效率
123…11
avatar
xixi
Articles
110
Tags
24
Categories
52
Follow Me
Announcement
This is my blog, for learning and communication purposes only. If there is any copyright infringement, please leave a message via email on GitHub.
Recent Posts
Offset Based Pointer Prefetcher(Arm US10445241B2)[2019]: Prefetching using offset data to access a pointer within a current data element for use in prefetching a subsequent data element2026-05-06
Pointer Array Prefetcher(AMD US12050916B2)[2024]: Array of pointers prefetching2026-05-06
Apple Data Prefetcher Overview2026-05-03
Load-to-AGU Early Bypass(Apple US9710268B2)[2017]: Reducing latency for pointer chasing loads2026-05-03
Pointer Chasing Prediction(Apple US9116817B2)[2015]: Pointer chasing prediction2026-05-03
Categories
  • cpu92
    • bpu7
      • btb4
      • restore1
    • cache45
      • LLC1
      • coherence2
      • prefetch28
Tags
HPCA ARM 2022 MICRO alibaba 2024 2020 2019 sifive 2025 2023 ASPLOS ventana intel 2018 ISCA Arm Tenstorrent AMD 2014 2026 Apple patent SC
Archives
  • May 2026 8
  • April 2026 29
  • March 2026 17
  • February 2026 9
  • November 2025 5
  • October 2025 16
  • September 2025 26
Website Info
Article Count :
110
Unique Visitors :
Page Views :
Last Update :
© 2025 - 2026 By xixiFramework Hexo|Theme Butterfly
Life is so Beautify!