Bitshift is returning incorect results

[piginablanket]

Can someone explain why this returns -13002 instead of 52534, like it should be?

BitShift(3442917988, 16)

Edited March 29, 2014 by piginablanket

[DatMCEyeBall]

32-bit integer wrap around?

Edited March 29, 2014 by DatMCEyeBall

[Malkey]

BitShift(3442917988, 16) is

3442917988 = 11001101 00110110 11000010 01100100  =  0xCD36C264
SHIFT right by 16
Equals
-13002     = 11111111 11111111 11001101 00110110  =  0xFFFFCD36


BitXOR(BitShift(-13002, 16), 0xFFFF0000) is

-13002     = 11111111 11111111 11001101 00110110  =  0xFFFFCD36
XOR
-65536     = 11111111 11111111 00000000 00000000  =  0xFFFF0000
Equals
52534      = 00000000 00000000 11001101 00110110  =  0x0000CD36
 
A work around when the 32 binary bit is a "1" :-

;ConsoleWrite(BitXOR(BitShift(3442917988, 16), 0xFFFF0000) & @LF) ; Returns 52534

Local $iH = 0xCD36C264 ; 0x4D36C264

ConsoleWrite((BitAND($iH, 0x80000000) ? BitXOR(BitShift($iH, 16), 0xFFFF0000) : BitShift($iH, 16)) & @LF)

 
 
Note the behavior of the 32nd bit when it is a "1" or a "0".

         
Instead of
 
v---- 32nd bit is a "1"                                           v-- C(Hex) is 1100(Binary)
11001101 00110110 11000010 01100100  =  0xCD36C264
as above,

we have:-
BitShift(0x4D36C264, 16) is

v---- 32nd bit is a "0"                                         v-- 4 is 0100
01001101 00110110 11000010 01100100 = 0x4D36C264
SHIFT right by 16
Equals
00000000 00000000 01001101 00110110 = 0x00004D36
|<---------------------->| all zeros here

When the 32bit is a "1", all those above zeros become ones.

That is strange behavior.

Edit: Improved work-around example.

Edited March 30, 2014 by Malkey

[jchd]

There is nothing strange here. Docs clearly state that "Bit operations are performed as 32-bit integers". Read "signed 32-bit integers".

So 3442917988 is in fact -852049308.

[czardas]

It is termed an arithmetic shift which I imagine has some neat applications for mathematical calculations. The alternative, logical shift, is also very useful; but there is no such native function in AutoIt so you have to code your own.

Edited March 30, 2014 by czardas

[jchd]

That.

Also see >this post for a 64-bit version of bitwise operations.

Edited March 30, 2014 by jchd