这两天有个Java的面试,job description里有句Understanding of different OS Memory management strategies and heap structure,只是什么意思?我应该看些什 么做准备呢?
h*o
3 楼
Journal of Materials Science July 1990, Volume 25, Issue 7, pp 3101-3104 Production of ultra fine SiC powder from SiC bulk by arc-plasma irradiation under different atmospheres and its application to photocatalysts Yasufumi Nariki, Yasunobu Inoue, Kohichi Tanaka http://link.springer.com/article/10.1007/BF00587657 Thank you.
【在 s****y 的大作中提到】 : 这两天有个Java的面试,job description里有句Understanding of different OS : Memory management strategies and heap structure,只是什么意思?我应该看些什 : 么做准备呢?
【在 s****y 的大作中提到】 : 这两天有个Java的面试,job description里有句Understanding of different OS : Memory management strategies and heap structure,只是什么意思?我应该看些什 : 么做准备呢?
h*o
8 楼
Thank you very much!
g*g
9 楼
能大致说清楚常见的GC算法就差不多了。
S*A
10 楼
OS Memory management strategies 是要在不同 OS 层面上说的,例如 page, mmap 之类的。 Java GC 说到底是个 application 层面上的,应该和 OS 不是一个 level。
【在 s****y 的大作中提到】 : 这两天有个Java的面试,job description里有句Understanding of different OS : Memory management strategies and heap structure,只是什么意思?我应该看些什 : 么做准备呢?
w*x
12 楼
right
【在 S*A 的大作中提到】 : OS Memory management strategies : 是要在不同 OS 层面上说的,例如 page, mmap 之类的。 Java GC : 说到底是个 application 层面上的,应该和 OS 不是一个 level。
some tips from SM's ppt 1) because we have optimized TLB. sometimes virtual is not as slow as what we thought 2) if n is not that big. Linear search can beat O(logn) BST 3) false sharing is the worst thing when we have multiple cores 4) forward traversal and backward traversal are almost the same 5) hyper threading "cores" are sharing L1D and L1I 6) inline has its downsides 7) PGO and WPO 8) Cache-oblivious algorithm design. 9) Other cache technology issues: Memory banks. Associativity. Inclusive vs. exclusive content.
S*A
43 楼
1-5 我不用他说都很明白。 6 是说 inlined code size blow out code cache 吧,还有 其他的吗? 7 需要解释一下 PGO 和 WPO 是什么东西。 8 就是assume worse case always cache missing 吧。 如何设计 data structure 分布在尽量少的 cache line 里面。 9 这些user space编程几乎不太可能考虑这个方面的东西。 例如 memory bank 根本管不着。
【在 b*******s 的大作中提到】 : some tips from SM's ppt : 1) because we have optimized TLB. sometimes virtual is not as slow as what : we thought : 2) if n is not that big. Linear search can beat O(logn) BST : 3) false sharing is the worst thing when we have multiple cores : 4) forward traversal and backward traversal are almost the same : 5) hyper threading "cores" are sharing L1D and L1I : 6) inline has its downsides : 7) PGO and WPO : 8) Cache-oblivious algorithm design.