Enable Sysadmin
Inject external data in policy evaluations with Conftest
Separating policy from data enables more robust and reusable policy definitions that allow you to factor external data sources in compliance evaluation.
source
(author: Anthony Critelli)
Inject external data in policy evaluations with Conftest
Separating policy from data enables more robust and reusable policy definitions that allow you to factor external data sources in compliance evaluation.
source
(author: Anthony Critelli)
#学习
《An Alternative TAGE-like Conditional Branch Predictor》
提出一套自适应的 Allocation Throttling 方法,可以消除 Cold entry 问题和 Allocation Rate 不合适的问题
https://dl.acm.org/doi/10.1145/3226098
《An Alternative TAGE-like Conditional Branch Predictor》
提出一套自适应的 Allocation Throttling 方法,可以消除 Cold entry 问题和 Allocation Rate 不合适的问题
https://dl.acm.org/doi/10.1145/3226098
Enable Sysadmin
How to run containers on Mac with Podman
Go beyond the basics, learn what happens under the hood when running Podman on your Mac, and create a flexible container environment that meets your needs.
source
(author: Marcelo Paternostro)
How to run containers on Mac with Podman
Go beyond the basics, learn what happens under the hood when running Podman on your Mac, and create a flexible container environment that meets your needs.
source
(author: Marcelo Paternostro)
Enable Sysadmin
Big data basics: What sysadmins need to know
Learn what big data is, how data is processed and visualized, and key big data terms to know.
source
(author: Gabrielle Stenzel)
Big data basics: What sysadmins need to know
Learn what big data is, how data is processed and visualized, and key big data terms to know.
source
(author: Gabrielle Stenzel)
Enable Sysadmin
How to encrypt Bash shell variables with Ansible Vault
Use Ansible Vault to share encrypted Bash environment variables across projects.
source
(author: Randy Romero)
How to encrypt Bash shell variables with Ansible Vault
Use Ansible Vault to share encrypted Bash environment variables across projects.
source
(author: Randy Romero)
讲道理我感觉没有load value driven的那一篇有价值
毕竟各种nn本来延迟就高,各种减精度以后也没有比原来好多少
现在前端都在卷每周期两条taken,这个延迟很难解决
而且最关键的是你profile出来的训练数据如何保证fair,编译器那边的pgo都还是经常负优化,你在benchmark上能有收益万一放到正常程序就寄了呢
毕竟各种nn本来延迟就高,各种减精度以后也没有比原来好多少
现在前端都在卷每周期两条taken,这个延迟很难解决
而且最关键的是你profile出来的训练数据如何保证fair,编译器那边的pgo都还是经常负优化,你在benchmark上能有收益万一放到正常程序就寄了呢
近几年顶会的bpu论文以Intel的那一篇sw/hw co-design的 Branch Prediction Is Not a Solved Problem 为起始,都开始搞cnn/rnn/rl,然后搞个ISA扩展让编译器做profiling以后插hint进去
另外最近几天猛翻bpu的文献,发现有的文章是真的水啊…2023年了拿个gshare改改hash都能发出去,感觉有点ai水文那味了
#XiangShan
首次真实感受到技术壁垒是什么
现在香山的主预测器就是一个典型的tage-sc的mpki,但是比Intel和苹果的高50%
这个事情其实学界骂了Intel很久了,自从我们伟大的Andre Seznec不再活跃以后,后面的新bpu公开的很少,都成商业机密了
整个学术界要做baseline只能拿tage-sc-l
首次真实感受到技术壁垒是什么
现在香山的主预测器就是一个典型的tage-sc的mpki,但是比Intel和苹果的高50%
这个事情其实学界骂了Intel很久了,自从我们伟大的Andre Seznec不再活跃以后,后面的新bpu公开的很少,都成商业机密了
整个学术界要做baseline只能拿tage-sc-l