trjhtr

This code is intended to be included with X86-16 writing ASCIIZ strings directly to video and is dependent upon some of the declarations in that code. When combined with that code a string declared @ 1000:0 in the following manner would;

AscTxt: db 01, 84H ; View page 0 output to 4
 db 'This is an example', 01, 40H ; View page 4 output to 0
 db 'This is replacing whatever is on top line of page 0'
 db 0, 01, 0FH, 0, 0 ; Waits for response then view page 0

As an example, let's start with this.

This is the next thing you would see after entering your .COM filename @ command prompt.

Notice how the cursor has not moved to EOS. That is by design. Then after pressing any key except ESC, we'll revert back to page 0 with top line being replaced.

; =============================================================================================
; Change display and/or destination page

; ENTER: 00 - 7 = Set do not change current page being displayed
; 6-4 = Display page. Ignored if equal to active or but 7 set
; 3 = Set, do not change display page
; 2-0 = Destination page. Ignored if already pointing there or bit 3 set.

; LEAVE: DI = New pointer when applicable
; DX = volatile, all others unchanged.

; FLAGS: Undefined
; ---------------------------------------------------------------------------------------------

 F01: cmp al, 1
 jnz F02
 lodsb ; Read only parameter for this function

 ; Real mode has a few limitations in addressing far data, so this just seems to be the
 ; most practical way of addressing data in BDA

 push bx

 ; Test bit 7 to determine if display page should be changed. If so, then bits 6-4 are
 ; ignored.

 test al, 10000000B ; Is bit 7 on
 jnz .cDest ; if so we are not changing display page

 push ax ; Still going to need low nibble
 shr ax, 4 ; Shift page number into low nibble
 cmp al, [fs:DispPg] ; Check BDA if anything actually needs done
 jz .cDest - 1 ; Already on that page

 ; The intent of procedurd e as a whole is to avoid BIOS as much as possible, but I
 ; did not want to implement code to manipulate controller.

 mov ah, SAP
 int VIDEO
 pop ax ; Restore value in low nibble

 .cDest: test al, 1000B ; Is bit 3 on
 jnz .exit ; if so we are not changing destination page ; Return to instruction just before label .next

 ; Check if there is any need to do anything by determining if destination video segment
 ; is already being pointed too.

 and al, 7
 mov dx, es ; Get current video segment
 mov dl, al
 add dl, 0B8H
 cmp dh, dl ; ZF will be set if same segments
 jz .exit

 ; Set new segment and then determine offset based on that pages cursor position from
 ; x/y coordinates specified in BDA.

 shl dx, 8
 mov es, dx ; Set new segment

 ; Caret position of new page or even one that has been written to before is assumed
 ; to be the starting point of next write.

 mov bx, Cursors ; Point to beginning of array of vectors
 shl ax, 1
 add bl, al
 mov dx, [bx] ; Points into arrary of vectors in BDA

 mov di, dx ; Just in case we are already at 0,0.
 or dx, dx
 jz .exit

 ; If position is top/left, not much point multiplying by zero

 mov al, dh
 imul ax, 80 ; 80 x 25 x 16 color assumed
 and di, 0FFH
 add di, ax
 shl di, 1 ; ES:DI set to new page & offset

 ; Restore non-volatile

 .exit: pop bx
 ret

 F02: ret

The initial comment: not clear what is entering (deducting it's about byte [ds:si]) and I would rather use "arguments" or "input" word. Not clear the description is about bits. Not all arguments are described (the code does use also ds:si address for lodsb). Typos. Modified registers are incomplete too (again si). Etc..

F01: cmp al, 1
 jnz F02

Why name like F01? Was it just for this review, or is it actual label? Use rather something more descriptive.

Why does the function end when al is not equal to 1? That should be mentioned in description, that al must be 1 to make it work.

I would move push bx lot more closer to the area where it's modified, so it can be easily seen with eye, where the whole push/pop block is, and span over fewer branching points (which is always error prone, to keep correct stack across some branching). You can actually surround just few instructions with push/pop bx, not having any branching at all.

 push ax ; Still going to need low nibble
 shr ax, 4 ; Shift page number into low nibble
 cmp al, [fs:DispPg] ; Check BDA if anything actually needs done
 jz .cDest - 1 ; Already on that page

Bug: if ah contains some low bits, you will get wrong display page value. Should be shr al,4 instead.

Why .cDest-1? First why cDest, that's not English, and why -1, you are on x86 powerful machine, there absolutely no reason to save bytes on symbol names or amount of symbols. 1988 called, they want your habits back.

Some more typos... don't you have spellcheck in your editor? :-o (I'm using simple kate for NASM sources, and Shift+Ctrl+O will switch spellchecking on/off, so it doesn't bother me on instructions, but I can use it to review comments from time to time).

 mov ah, SAP
 int VIDEO

This will not compile, you should rather provide working source for review. By building it yourself before posting you will prevent some kind of accident, like cutting out more than you did want. Also I have no idea what is SAP or VIDEO, so no way to review it and put any advice on that. Except stop using cryptic short symbol names which can't be comprehended by reading like plain English.

 mov bx, Cursors ; Point to beginning of array of vectors
 shl ax, 1
 add bl, al

Do just shl al, 1 if you are using only al later. This at first did look as bug to me, because you have anything in ah, then I realized you will avoid that by using al only.

Actually this will fail if Cursors array spans over 256B boundary, then the add bl,al will produce wrong address.

But you didn't show how you define Cursors, maybe you are sure it is well aligned and will always fit into 256B "page".

Still maybe consider more robust version doing whole and ax,7 add ax,ax add bx,ax to calculate with full 16b values.

And the whole set cursor code is a bit fishy, I mean the calculation of di based on [x, y], don't you have such code already somewhere?

Are you optimizing for speed that you can't afford to call that? (probably not, as you are still in real mode)

Also imul ax, 80 is again risky as you don't specify ah content (would work with my modification above to and ax,7).

I would probably write separate subroutine to calculate di based on [x, y] coordinates, like for example:

; input: dh:dl = y:x of cursor
; output: di = text mode offset for cursor position
 getDiForCursorPosition:
 push dx
 movzx di, dh ; di = y
 imul di, 80 ; di = 80 * y
 movzx dx, dl ; dx = x
 add di, dx ; di = 80 * y + x
 add di, di ; di = (80 * y + x) * 2
 pop dx
 ret

And use that from everywhere (after you will check it works as expected, I didn't debug it :) ).