- 論壇徽章:
- 11
|
此言差矣，sgi stl對(duì)小于128字節(jié)的小內(nèi)存是有很高效的內(nèi)存池算法的。
malloc 的效率低是公認(rèn)了的 ...
ydfgic 發(fā)表于 2011-08-26 13:09 ![]()
malloc 的效率是誰(shuí)公認(rèn)低了？
sgi stl 內(nèi)存池自然有其優(yōu)點(diǎn)， 但毛病也不小
1。 無(wú)法處理大塊內(nèi)存申請(qǐng)吧， 如果處理大塊內(nèi)存申請(qǐng)， 要么內(nèi)存塊的跨度必須大， 就是浪費(fèi)內(nèi)存； 要么要求鏈表數(shù)量足夠多， 如果鏈表數(shù)量足夠多， 循環(huán)又要多了（分配2， 結(jié)果該鏈表沒(méi)有， 去4的里面找， 也沒(méi)有， 去6的鏈表找， 再?zèng)]有。。。。）
2。 無(wú)法內(nèi)存回收吧， 除非它再加上buddy 系統(tǒng)， 可是至少sgi stl 沒(méi)有； 它就是一個(gè)動(dòng)態(tài)平衡算法， 適應(yīng)的峰值不高的應(yīng)用， 如果一瞬間要求2G的內(nèi)存， 然后剩下的1個(gè)月， 只要求1M的內(nèi)存， 它的表現(xiàn)不咋地
* Why use this malloc?
0042
0043
This is not the fastest, most space-conserving, most portable, or
0044
most tunable malloc ever written. However it is among the fastest
0045
while also being among the most space-conserving, portable and tunable.
0046
Consistent balance across these factors results in a good general-purpose
0047
allocator for malloc-intensive programs.
0048
0049
The main properties of the algorithms are:
0050
* For large (>= 512 bytes) requests, it is a pure best-fit allocator,
0051
with ties normally decided via FIFO (i.e. least recently used).
0052
* For small (<= 64 bytes by default) requests, it is a caching
0053
allocator, that maintains pools of quickly recycled chunks. 這個(gè)是不是sgi stl 的基本手法？？？
0054
* In between, and for combinations of large and small requests, it does
0055
the best it can trying to meet both goals at once.
0056
* For very large requests (>= 128KB by default), it relies on system
0057
memory mapping facilities, if supported.
0058 |
|
免費(fèi)注冊(cè)
發(fā)表于 2011-08-25 21:06
