determine number of cyccles for cpu operations

cpu_cycles.asm

section .data
  fmt         db "Value: %ld", 0x0a, 0  ; Format string for printf
  iterations  dq 0x3b9aca00  ; 1000000000

section .bss
  number      resq 1         ; Reserve 1 quadword (64 bits or 8 bytes) for number used for printf
  sum         resq 1         ; Reserve 1 quadword (64 bits or 8 bytes) for sum
  cyclesa   resq 1         ; Reserve 1 quadword (64 bits or 8 bytes) for rdtsc output
  cyclesb   resq 1         ; Reserve 1 quadword (64 bits or 8 bytes) for rdtsc output

section .text
global main
extern printf

get_cyclesA:
  rdtsc
  shl    rdx, 32
  mov    rcx,rax
  or     rcx,rdx
  mov    [cyclesa], rcx    ; Move the value in rax to cpucycles
  ret                   ; Return to the caller (address is already on the stack)

get_cyclesB:
  rdtsc
  shl    rdx, 32
  mov    rcx,rax
  or     rcx,rdx
  mov    [cyclesb], rcx    ; Move the value in rax to cpucycles
  ret                   ; Return to the caller (address is already on the stack)

print_number:
  lea   rdi, [fmt]   ; parameter  1  for printf
  mov   rsi, [number]    ; Load the value at rsp (the reserved stack space) into rsi
  xor   eax,  eax     ; Zero out eax to indicate no floating point parameters for printf
  call  printf
  ret                   ; Return to the caller (address is already on the stack)



main:
  push  rbp         ; pushes rbp immediately below the return address.
  mov   rbp, rsp    ; makes rbp point to that object (stack address)
  sub rsp, 8                         ; Align the stack to 16 bytes, i.e, you pushed rbp so you get 8 bytes misaligned


  call get_cyclesA
    mov    edx, [iterations]   ; set loop variable 1000000000
add_loop:
    mov    rax, QWORD [sum]  ; read memory to rax
    add    rax,0x1    ; add 1 to rax
    mov    QWORD [sum],rax   ; write rax to memory
    sub    edx,0x1    ; subtract 1 from loop variable
    jne    add_loop


  call get_cyclesB

  mov r9, [cyclesa]
  mov [number], r9
  call print_number ; print cycles start

  mov r8, [cyclesb]
  mov [number], r8
  call print_number ; print cycles end

  mov rax, [cyclesb]
  mov rbx, [cyclesa]
  sub rax, rbx
  mov [number], rax
  call print_number  ; print difference

  mov rax, [cyclesb]
  mov rbx, [cyclesa]
  sub rax, rbx
  mov r12, [iterations]
  xor rdx, rdx
  div r12  ; rax / r10   rax has answer remainder in rdx
  mov [number], rax
  call print_number ; print average

add rsp, 8                         ; Restore the stack pointer
  xor   eax, eax    ; return 0
  pop   rbp
  ret

Makefile

cpu_cycles.elf: cpu_cycles.asm
	nasm -f elf64 -g -F dwarf $< -o cpu_cycles.o
	# ld -o $@ lfm.o
	gcc -no-pie -o $@ cpu_cycles.o
	./$@