storeforward.asm

section .data

align 64
buf1: dq 0

section .text

global storeload_L
proc_frame storeload_L
    alloc_stack 8
[endprolog]
    align 64
    ; 16 is good, 15 is bad
    ; 0 is inconsistent
    times 16 nop
    mov r8, rcx
    rdtscp
    mov rcx, r8
    mov r8d, eax
    jmp .entry
align 4096
.entry:
    ; these nops have to be here for some reason
    ; a single 3-byte nop does *not* work
    times 3 nop
    ; this lea can only be right here,
    ; not before the align, not even before the 3 nops
    lea rdx, [rel buf1]
.loop:
    mov [rdx], rax
    mov rax, [rdx]
    sub ecx, 1
    jnz .loop

    rdtscp
    sub eax, r8d
    add rsp, 8
    ret
endproc_frame