5 - 10 = 5 + (-10) = 0x05 + (0xF5 + 1) = 0x05 + 0xF6 = 0xFB
00000101
11110101
+ 00000001
==========
11111011
And this goes on for 16 or 32 or 64 bits 0 + 1 + 0 = 1, carry 0
You are right in the sense that this is not being done. Subtraction is an addition with the second operand inverted and the translation in the translation. Some processor families invert some of them. It looks like you are looking at what inverts execution to subtract.
, 5 - 10 (), 10 - 5 , ( ).
- , , , , .