Difference between revisions of "Talk:Script"
m (Fixed my signature.) |
m (Corrected text.) |
||
| Line 4: | Line 4: | ||
--[[User:ThePiachu|ThePiachu]] 11:37, 20 December 2011 (GMT) | --[[User:ThePiachu|ThePiachu]] 11:37, 20 December 2011 (GMT) | ||
| − | The byte vectors in the stack are specified as being signed integers | + | The byte vectors in the stack are specified as being signed integers of variable-length. Then there's an explanation that these integers are considered false if they are either zero or negative zero, which is 0x80. For this to be the case, the elements should be represented in an old binary format called sign-magnitude, which is important to state explicitly, since today virtually all computers use two's complement as representation, and there's no such thing as a negative zero. There's even another representation, one's complement, where negative zero looks like 0xff. |
Reading the source code of the application, I see that arithmetic operations expect unsigned integers, for example, operations OP_2MUL and OP_2DIV are implemented as byte-shifting, which wouldn't work with signed representations. | Reading the source code of the application, I see that arithmetic operations expect unsigned integers, for example, operations OP_2MUL and OP_2DIV are implemented as byte-shifting, which wouldn't work with signed representations. | ||
| + | |||
| + | It seems to me that, at best, variable-length sign-magnitued integer format is only used for testing for truth, although I haven't read all the code yet. | ||
--[[User:Jpierre|Jean-Pierre Rupp]] 10:43, 4 March 2012 (GMT) | --[[User:Jpierre|Jean-Pierre Rupp]] 10:43, 4 March 2012 (GMT) | ||
Revision as of 10:49, 4 March 2012
xOP_IFDUP 115 0x73 x x / x x If the input is true or false, duplicate it.
Shouldn't it be: "If the input is true, duplicate it."? --ThePiachu 11:37, 20 December 2011 (GMT)
The byte vectors in the stack are specified as being signed integers of variable-length. Then there's an explanation that these integers are considered false if they are either zero or negative zero, which is 0x80. For this to be the case, the elements should be represented in an old binary format called sign-magnitude, which is important to state explicitly, since today virtually all computers use two's complement as representation, and there's no such thing as a negative zero. There's even another representation, one's complement, where negative zero looks like 0xff.
Reading the source code of the application, I see that arithmetic operations expect unsigned integers, for example, operations OP_2MUL and OP_2DIV are implemented as byte-shifting, which wouldn't work with signed representations.
It seems to me that, at best, variable-length sign-magnitued integer format is only used for testing for truth, although I haven't read all the code yet.
--Jean-Pierre Rupp 10:43, 4 March 2012 (GMT)
