| 32 位的寄存器容量是 4 字節(jié), 如果內(nèi)存中的數(shù)據(jù)都按 4*n 字節(jié)對(duì)齊, 肯定會(huì)加快吞吐速度;
但事實(shí)并非如此, 不同大小的數(shù)據(jù)可能會(huì)讓寄存器別別扭扭地去處理, 從而降低了運(yùn)行速度!
如果使用對(duì)齊, 就會(huì)浪費(fèi)掉一些內(nèi)存空間; 其實(shí)這是一個(gè)需要權(quán)衡 "速度" 與 "內(nèi)存" 得失的問題.
準(zhǔn)備使用的測(cè)試文件: ------------------------------------------------------------------------------------------------------------------------------------------
.586
.modelflat, stdcall
include?? windows.inc
include?? kernel32.inc
include?? masm32.inc
include?? debug.inc
includelib kernel32.lib
includelib masm32.lib
includelib debug.lib
.data
?? ;聲明三個(gè)字節(jié)變量
??? v1 db 0
??? v2 db 0
??? v3 db 0
.code
main proc
?? ;三個(gè)字節(jié)變量的默認(rèn)偏移地址如下(在沒有對(duì)齊約束的情況下, 它們各占一個(gè)字節(jié)):
??? PrintDec offset v1 ;4206592
??? PrintDec offset v2 ;4206593
??? PrintDec offset v3 ;4206594
??? ret
main endp
end main
------------------------------------------------------------------------------------------------------------------------------------------
ALIGN: 指定對(duì)齊邊界 ------------------------------------------------------------------------------------------------------------------------------------------ ; Test11_2.asm
.586
.modelflat, stdcall
include?? windows.inc
include?? kernel32.inc
include?? masm32.inc
include?? debug.inc
includelib kernel32.lib
includelib masm32.lib
includelib debug.lib
.data
??? v1 db0
??? align 4 ;讓下一個(gè)變量的起始地址保證是 4 的倍數(shù)
??? v2 db0
??? v3 db0
.code
main proc
??? PrintDec offset v1 ;4206592
??? PrintDec offset v2 ;4206596 (!)
??? PrintDec offset v3 ;4206597
??? ret
main endp
end main
------------------------------------------------------------------------------------------------------------------------------------------
;align 后面的參數(shù)是 2n, 還可測(cè)試下: 1、2、8、16
EVEN 是偶對(duì)齊, 相當(dāng)于 ALIGN 2 ------------------------------------------------------------------------------------------------------------------------------------------ ; Test11_3.asm
.586
.modelflat, stdcall
include?? windows.inc
include?? kernel32.inc
include?? masm32.inc
include?? debug.inc
includelib kernel32.lib
includelib masm32.lib
includelib debug.lib
.data
??? v1 db0
??? even?? ;(!)
??? v2 db0
??? v3 db0
.code
main proc
??? PrintDec offset v1 ;4206592
??? PrintDec offset v2 ;4206594 (!)
??? PrintDec offset v3 ;4206595
??? ret
main endp
end main
------------------------------------------------------------------------------------------------------------------------------------------
ORG 可以指定從當(dāng)前位置跨越指定書目的字節(jié)再安排下一個(gè)數(shù)據(jù): ------------------------------------------------------------------------------------------------------------------------------------------ ; Test11_4.asm
.586
.modelflat, stdcall
include?? windows.inc
include?? kernel32.inc
include?? masm32.inc
include?? debug.inc
includelib kernel32.lib
includelib masm32.lib
includelib debug.lib
.data
??? v1 db0
??? org100 ;(!)
??? v2 db0
??? v3 db0
.code
main proc
??? PrintDec offset v1 ;4206592
??? PrintDec offset v2 ;4206692 (!)
??? PrintDec offset v3 ;4206693
??? ret
main endp
end main ------------------------------------------------------------------------------------------------------------------------------------------ 這些偽指令不對(duì)局部變量產(chǎn)生影響, 因?yàn)閭沃噶钭饔迷诰幾g之前, 局部變量使用內(nèi)存是程序運(yùn)行之后的事.
關(guān)于局部變量的對(duì)齊和速度, 除了編譯的工作外, 我們?cè)诎才抛兞款愋秃痛涡驎r(shí)也應(yīng)該考慮到 "對(duì)齊" 的問題. ? 轉(zhuǎn)載于：----博客圓--------- | 
