FASM: further optimized prime number counter using array that expands with each prime number found, to not waste extra division ops (i.e. don't % 9, 15, 21, 25, 27, etc)

prime64_M.asm

format PE64 console 3.1

entry @f

include 'win64a.inc'

MAX_ITERATIONS = 100000000

; 1kb EXE 1337
section '' import code data \
			readable writeable executable
	library msvcrt,'MSVCRT.DLL'
	import msvcrt,\
		printf,'printf'

	align 10h
	@@:
		xor		r13	,	r13
		inc		r13
		mov		r14	,	r13
		add		r14b,	2
	@:
		xor		r12	,	r12
	@@:
		xor		rdx	,	rdx
		mov		rax	,	r14
		mov		r15	,	[p + r12 * 8]
		div		r15
		test	rdx	,	rdx
		jz		@f

		inc		r12
		cmp		r12	,	r13
		jb		@b

		invoke	printf,	fmt,	r14
		mov		[p + r13 * 8],	r14
		inc		r13
	@@:
		add		r14	,	2
		cmp		r14	,	-3
		jb		@

		ret
	fmt	db	'%u',13,0
	
	align	10h
	p	dq	3 ; array, start with 1 item
		rq	MAX_ITERATIONS

prime64_M_L.asm

; linux / WSL
format ELF64 executable 3

MAX_ITERATIONS = 100000000

segment executable writeable readable
	_start:
		xor		rcx	,	rcx
		inc		rcx
		mov		rbp	,	3
	@:
		xor		rbx	,	rbx
	@@:
		xor		rdx	,	rdx
		mov		rax	,	rbp
		mov		rdi	,	[p + rbx * 8]
		div		rdi
		test	rdx	,	rdx
		jz		@f

		inc		rbx
		cmp		rbx	,	rcx
		jb		@b

		mov		rsi	,	rbp
		mov		[p + rcx * 8],	rbp
		inc		rcx
	@@:
		add		rbp	,	2
		cmp		rbp	,	-3
		jb		@

		ret
	
	align	10h
	p	dq	3
		rq	MAX_ITERATIONS

using gdb on linux version to see registers and control pausing and running of the program