前端方向预测器实现方式讨论
#4200
Replies: 1 comment
-
|
方向预测器的latency显然和性能密切相关,能少一拍就少一拍,目前评估使用的工艺能够支持1 cycle 的bubble,因此使用这样的设计。另外据我了解,很多高性能的处理器也会使用包括但不限于ahead pipeline等方式尽可能压缩方向预测器的latency,1 cycle bubble并不奇怪。 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
前段时间简单对比了一下香山和boom的前端,我发现香山的前端第二级btb(ftb)和tage都是一个cycle bubble
https://docs.xiangshan.cc/zh-cn/latest/frontend/bp/
然后boom的前端btb是一个cycle bubble,但方向预测器latency远大于香山。
https://docs.boom-core.org/en/latest/sections/branch-prediction/index.html
然后看到另一篇三星的论文,看起来他们的方向预测器latency也高于btb,然后btb对于always taken有特殊优化才实现这种场景下1cycle的taken branch bubble。
当前在香山上是出于什么考虑tage的latency压缩的这么低,还是经过算法调优后选择牺牲精度保证latency?
Beta Was this translation helpful? Give feedback.
All reactions