| 作者 |
给楼下的绽铃子大师提几个建议 |
 |
| 所跟贴 |
"发明的关键" should be: -- GOOG08 - (319 Byte) 2010-3-31 周三, 22:36 (839 reads) |
网客JT [博客]

头衔: 海归上校 声望: 博导
加入时间: 2008/01/11 文章: 1021
海归分: 97019
|
|
作者:网客JT 在 海归商务 发贴, 来自【海归网】 http://www.haiguinet.com
Yeah, I may not understand, so please help me out:
Could you please explain to me,
"theoretical upper limit of cache replacement algorithms" -- is it regarding ALL possible software applications, or only specific applications be tested/optimized for?
I am aware of The theoretically optimal page replacement algorithm
(also known ad Belady's optimal page replacement policy), which is a theoretical concept, that can not be implemented for GENERAL PURPOSE OPERATING SYSTEM,unless the software application to run on it iis either known before hand or has specific memory access patterns.
So how is this "theoretical upper limit of cache replacement algorithms" defined here? -- in regard of software applications being executed on the CPU/OS.
Mind you, optimal solution regarding special type of software applications is still meaningful & maybe useful, but a far cry from the earth-shattering claims being made here.
| GOOG08 写道: | Sorry, you might not understand too.
The key is:
"Our studies also indicate that WLRU is very close to the theoretical upper limit of cache replacement algorithms. This means any further improvement in CPU cache performance will have to come from changes to the software. "
Do you know their studies? |
作者:网客JT 在 海归商务 发贴, 来自【海归网】 http://www.haiguinet.com
|
|
|
| 返回顶端 |
|
 |
|
| |
|
|
|
您不能在本论坛发表新主题, 不能回复主题, 不能编辑自己的文章, 不能删除自己的文章, 不能发表投票, 您 不可以 发表活动帖子在本论坛, 不能添加附件不能下载文件, |
|
|