25. Sign Extending Signed Value

• Consider:

          .DATA
          mem8    SBYTE   -5
          mem16   SWORD   +5
          mem32   SDWORD  -5
  
          .CODE
          .
          .
          .
          mov     al, mem8    ; Load 8-bit -5 (FBh)
          cbw                 ; Convert to 16-bit -5 (FFFBh) in AX
          mov     ax, mem16   ; Load 16-bit +5
          cwd                 ; Convert to 32-bit +5 (0000:0005h) in DX:AX
          mov     ax, mem16   ; Load 16-bit +5
          cwde                ; Convert to 32-bit +5 (00000005h) in EAX
          mov     eax, mem32  ; Load 32-bit -5 (FFFFFFFBh)
          cdq                 ; Convert to 64-bit -5
                              ;   (FFFFFFFF:FFFFFFFBh) in EDX:EAX
  

• Sign extending instructions efficiently convert unsigned values as well, provided the sign bit is zero.

• This example, for instance, correctly widens mem16 whether you treat the variable as signed or unsigned.

• The processor does not differentiate between signed and unsigned values.

• For instance, the value of mem8 in the previous example is literally 251 (0FBh) to the processor.

• It ignores the human convention of treating the highest bit as an indicator of sign.

• The processor can ignore the distinction between signed and unsigned numbers because binary arithmetic works the same in either case.

• The programmer, not the processor, must keep track of which values are signed or unsigned, and treat them accordingly.