CMPSB compares the byte at [DS:SI] or [DS:ESI] with the
byte at [ES:DI] or [ES:EDI], and sets the flags accordingly.
It then increments or decrements (depending on the direction flag:
increments if the flag is clear, decrements if it is set) SI and
DI (or ESI and EDI).
The registers used are SI and DI if the address size is 16
bits, and ESI and EDI if it is 32 bits. If you need to use
an address size not equal to the current BITS setting, you can
use an explicit a16 or a32 prefix.
The segment register used to load from [SI] or [ESI] can be
overridden by using a segment register name as a prefix (for
example, ES CMPSB). The use of ES for the load from [DI]
or [EDI] cannot be overridden.
CMPSW and CMPSD work in the same way, but they compare a
word or a doubleword instead of a byte, and increment or decrement
the addressing registers by 2 or 4 instead of 1.
The REPE and REPNE prefixes (equivalently, REPZ and
REPNZ) may be used to repeat the instruction up to CX (or
ECX - again, the address size chooses which) times until the
first unequal or equal byte is found.