I've been writing a bootloader in assembly (a game). The bootloader uses the bios WAIT function (int 0x15, ah 0x86) for delays between frames. I'm debugging with BOCHS and everything works just splendid (timing is perfect). I also made a bootable iso with the isogenimage tool and tested my bootloader in virtualbox and everything works as expected there too. So I think it's safe to say that delays work in virtual environments.
Now here's the peculiar part: when I write the bootloader to a usb drive and boot from it, the WAIT function waits FAR longer than on the virtual machines. I'd estimate it's 2-3 times longer. At least. Such long delays between frames are absolutely unacceptably for a game.
What in the world could cause this behaviour? Is it possible that my buggy bios clock runs at a slower speed (>18 microseconds)? But surely, that would violate some IBM standard? It's inexplicable to me.
By the way, if I remove the WAIT interrupt, the bootloader runs as expected (just too fast for my purposes), so it's not my actual code that's running slow. Plus, I'm running on a recent ~2013 laptop, so it can't be because of low performance.
Any insight, appreciated!
EDIT: I had the idea to test it on another physical computer and IT ALSO RAN EXTREMELY SLOW! So maybe it's not a "shitty bios" :) The odds of it running slow on two computers are pretty low (I think), so that might mean something...
CODE:
;MACRO CONSTANTS
%define p_width 10 ; player width
%define p_left_offset 20 ; space between left margin and player's left side
%define p_right_offset 290 ; space between right margin and player's right side; for linear player drawing purposes (xres - p_width - p_left_offset)
%define beak_width 5
%define xres 320 ; VGA resolution width
%define yres 200 ; VGA resolution height
%define buffer_addr 0x1000 ; start of offscrean buffer
[bits 16]
[org 0x7c00]
section .text
boot:
jmp start
times 3-($-$$) DB 0x90 ; Support 2 or 3 byte encoded JMPs before BPB.
; Dos 4.0 EBPB 1.44MB floppy
OEMname: db "mkfs.fat" ; mkfs.fat is what OEMname mkdosfs uses
bytesPerSector: dw 512
sectPerCluster: db 1
reservedSectors: dw 1
numFAT: db 2
numRootDirEntries: dw 224
numSectors: dw 2880
mediaType: db 0xf0
numFATsectors: dw 9
sectorsPerTrack: dw 18
numHeads: dw 2
numHiddenSectors: dd 0
numSectorsHuge: dd 0
driveNum: db 0
reserved: db 0
signature: db 0x29
volumeID: dd 0x2d7e5a1a
volumeLabel: db "NO NAME "
fileSysType: db "FAT12 "
start:
mov ax, 0x13 ; VGA 16bit colors 320x200 mode
int 0x10 ; call video update bios
mov ax, 0xA000 ; Video memory startaddr
mov es, ax ; Real buffer video segment into segment register (for segment:offset format)
mov ax, buffer_addr ; temp storage of virtual buffer addr
mov gs, ax ; Virtual buffer segment stored in GS
xor ax, ax
mov ds, ax ; zero offset to access vars defined in ;Data secition
gameloop:
call clear_buffer ; prepare to draw by clearing memory buffer
call draw_player ; draw player to memory buffer
call switch_buffers ; copy memory buffer to vram
add word [y_offset], 0x03 ; downward velocity, moves player around
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; HERE IS THE WAIT INTERRUPT THAT CAUSES THE PROBLEM ON REAL HARDWARE
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; WAIT (sleep) for a little bit
mov ah, 0x86 ; specify for int 0x15 WAIT interupt
mov cx, 0x0006 ; high word of wait time
mov dx, 0xffff ; low word of wait time
int 0x15 ; waits for cx:dx 1,000,000ths of a second
jmp gameloop
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
hlt
; Draws the player square on screen, with y offset of y_offset
draw_player:
xor si, si ; make sure si is 0; this will be pixel offset accumulatore
xor bx, bx ; this will be row counter; paintnig will stop if bl == p_width
mov ax, xres ; used to get space above player
mul word [y_offset] ; result will be in dx:ax (high, low bytes)
add si, ax ; add top offset to pixel accumulator
accum_p_offset:
inc bl ; keep track of rows drawn.
add si, p_left_offset ; now offset is at starting point to paint
mov cx, si ; cx looks into the future
add cx, p_width ; cx will be the goal pixel
draw_p_pixel: ; draws the player pixels
inc si ; go to next pixel (compensate for 0-start)
mov [gs:si], byte 0x0E ; move white pixel to virtual memory at offset ax
cmp si, cx ; check if ax caught up with cx
jne draw_p_pixel ; if not, draw next player pixel
ret
clear_buffer:
mov cx, 32000 ; number of times loop is performaned. 320 * 200 / 2 (because moving words)
xor si, si ; clear si to make sure acumulator starts at first pixel
zero_buf:
mov [gs:si], word 0x0101 ; set background to blue
add si, 2 ; increment accumulator
loop zero_buf ; jumps to zero_buf if cx is not equal to 0, then decrements cx
ret
; copies contents from memory buffer to vram
switch_buffers:
push ds ; save old data-segment (needed for getting the address of databytes later on)
mov si, gs ; use si as temp for gs (virtual memory segment)
mov ds, si ; move into datasegment the addr of virtual memory segment for movsw instruction
xor si, si ; clear accumulators for string operation
xor di, di
mov cx, 32000 ; how many words to copy 320 * 200 / 2
cld ; copy direction
rep movsw ; repeat copy from ds:si to es:di until cx is 0 (copies buffer from memory to vram)
pop ds ; restores data segment to reference databytes properly after switch_buffers
ret
;DATA
y_offset:
dw 0x16
; FILLER
times 510 - ($-$$) db 0
dw 0xaa55
