Is there a faster way of dividing by 2, with sign, in assembly than the one in the exam开发者_JS百科ple below?
...
mov ecx, 2
idiv ecx
push eax #push the result
...
Sure:
sar eax, 1
The sar
opcode differs from shr
in that the most significant (sign) bit is preserved in sar
, and it is set to 0 in shr
. The Arithmetic shift page on Wikipedia shows much more detail about this operation in a general context.
Note that on a 2's complement machine (which the x86 is) this actually calculates floor(eax / 2)
. In particular, that means that for an integer x:
- for x = 0, the result is 0
- for x > 0, the result is floor(x / 2)
- for x < 0, the result is also floor(x / 2), or -ceil(-x / 2)
The latter result gives results that may be unexpected. For example, -3 sar 1 results in -2, not -1. On the other hand, 3 sar 1 results in 1.
精彩评论