; array2_64.asm   a 2D array of characters, indexing and printing
; Assemble:	nasm -f elf64  array2_64.asm
; Link:		gcc -o array2_64  array2_64.o
; Run:		./array2_64

        extern	printf		; the C function, to be called
        SECTION .data		; Data section, initialized variables
nrow:	dq 	5		; 5 rows
ncol:	dq 	7		; 7 columns
fmtc:	db 	"%c", 0		; print one character at a time
fmtend:	db 	10, 0		; end a line
fmtdbg:	db	"i=%ld, j=%ld, k=%ld", 10, 0
star:	db 	'*'		; one character '*'
spc:	db 	' '		; space
	
	section .bss
a2:	resb 	5*7		; two dimensional array of bytes
i:	resq 	1		; row subscript
j:	resq 	1 		; col subscript
	
        SECTION .text           ; Code section.
        global main		; the standard gcc entry point
main:				; the program label for the entry point
        push    rbp		; set up stack frame

	
; clear a2 to space
	mov 	rax,0		; i=0
	mov	[i],rax
loopi:
	mov	rax,[i]
	mov 	rbx,0		; j=0
	mov	[j],rbx
loopj:
	mov	rax,[i]
	mov	rbx,[j]
	imul 	rax,[ncol]	; i*ncol
	add  	rax, rbx	; i*ncol + j
	mov 	dl, [spc]	; need just character, byte
	mov 	[a2+rax],dl	; store space

;	mov	rdi, fmtdbg
;	mov	rsi, [i]
;	mov	rdx, [j]
;	mov	rcx, rax
;	mov	rax, 0
;	call	printf
	
	mov	rbx,[j]
	inc 	rbx		; j++
	mov	[j],rbx
	cmp 	rbx,[ncol]      ; j<ncol
	jne 	loopj

	mov	rax,[i]
	inc 	rax		; i++
	mov	[i],rax
	cmp	rax,[nrow]	; i<ncol
	jne 	loopi
; end clear a2 to space

; put in a few star
	mov	rax,0		; i = 0
	mov	rbx,0		; j = 0
	imul 	rax,[ncol]	; i*ncol
	add  	rax, rbx	; i*ncol + j
	mov	dl, [star]
	mov 	[a2+rax],dl	; store star a2[0][0]

	mov	rax,1		; i = 1
	mov	rbx,1		; j = 1
	imul 	rax,[ncol]
	add  	rax, rbx
	mov	dl, [star]
	mov 	[a2+rax],dl	; store star a2[1][1]

	mov	rax,2		; i = 2
	mov	rbx,2		; j = 2
	imul 	rax,[ncol]
	add  	rax, rbx
	mov	dl, [star]
	mov 	[a2+rax],dl	; store star a2[2][2]
	
	mov	rax,3		; i = 3
	mov	rbx,3		; j = 3
	imul 	rax,[ncol]
	add  	rax, rbx
	mov	dl, [star]
	mov 	[a2+rax],dl	; store star a2[3][3]


; print
	mov 	rax,0		; i=0
	mov	[i],rax
ploopi:
	mov	rax,[i]
	mov 	rbx,0		; j=0
	mov	[j],rbx
ploopj:
	mov	rax,[i]
	mov	rbx,[j]
	mov 	dl, [spc]
;	mov 	[a2+rax*[ncol]+rbx],dl ; won't accept
				; compute double subscript
	imul 	rax,[ncol]
	add  	rax, rbx

	mov 	rdi, fmtc
	mov 	rsi, [a2+rax]
	mov	rax, 0
	call 	printf
	
	mov	rbx,[j]
	inc 	rbx		; j++
	mov	[j],rbx
	cmp 	rbx,[ncol]      ; j<ncol
	jne 	ploopj

	mov 	rdi, fmtend
	mov	rax, 0
	call 	printf		; end line
	
	mov	rax,[i]
	inc 	rax		; i++
	mov	[i],rax
	cmp	rax,[nrow]	; i<ncol
	jne 	ploopi
; print a2

	
	
	pop	rbp		; restore stack
	mov	rax,0		; normal, no error, return value
	ret			; return