LOOPE - x86-64 instruction reference

loopeLoop while equal 
Decrements rcx by one and jumps to the short target if the counter is non-zero and ZF is 1.
8086 and later
Intel's reference: LOOP/LOOPcc 
CFPFAFZFSFDFOF –––––––
Changes no flag
Reads ZF
Forms  loope   
 loope , ecx   386 and later 
 loope , rcx  
Operands imm8 A branch target, written as a label or address. NASM calculates the displacement from the end of the instruction. The displacement must fit a signed byte, -128 to 127. 
loope / loopz additionally require ZF = 1; loopne / loopnz require ZF = 0. The instructions do not change flags.
In 64-bit mode, the default counter is rcx; a32 selects ecx. A 16-bit cx counter is unavailable in this mode. A zero initial counter wraps to its maximum value when decremented.
Write a target label or address. NASM calculates a signed 8-bit displacement from the end of the instruction (−128 through 127 bytes). There is no longer near encoding.
 global _start

section .text
_start:
        mov rcx, 5          ; the counter this family reads
        xor rax, rax
again:
        add rax, rcx
        cmp rax, 0          ; the conditional forms read the flags too
        loope again         ; jumps while rcx is not zero and the last comparison was equal

        mov rax, 60         ; syscall 60: exit
        xor rdi, rdi
        syscall