I am trying to write a c++ program with inline assembly
that enables svga graphics. Below is source code for
a test-program that I have written. It works (at least with
borland turbo c++ 3.0 and my machine :) )
but it is very slow; filling screen takes some seconds. It would be great if someone could help and tell how to speed
things up.

#include // For the ldiv();
#include // For the getch();

// Initialize svga in resolution of 800x600 and 256 colors.
// This is the first procedure to be called.
void init_800_600()
asm mov ax,0x04f02; // svga.
asm mov bx,0x0103; // 800x600 and 256 colors.
asm int 0x10; // Do it.

// This procedure is called int the end. It sets the video-mode
// to text.
void quit()
asm mov ah,0x00; // Set video-mode.
asm mov al,0x03; // Back to normal.
asm int 0x10; // Do it.

// We can't use all the memory at the same time because
// of the 64 k limit. This procedure flips the page.
void flip_page(short page)
asm mov dx,page; // Move the page-number to dx.
asm mov bx,0; // bx is now 0.
asm mov ax,0x04f05; // svga.
asm int 0x10; // Do it.

// Draws a pixel.
// Calculation of the page and pixel's location in memory
// takes probably too much time and should be done
// some other manner?
void pixel(long x, long y, unsigned char color)
long location = ( y * 800L + x ); // Calculate the place in memory.

ldiv_t page = ldiv(location,65536L); // We cannot use all the memory at the
flip_page(page.quot); // same time because of the 64 k limit.
// Calculate the page.

// Calculate the new location in the page.
unsigned int new_location = location-(page.quot*65536);

// Plot a pixel.
asm mov ax,0x0A000;
asm mov es,ax;
asm mov al,color;
asm mov di, new_location;
asm mov es:[di], al;

int main(void)
init_800_600(); // Initialize the svga in 800x600
// and 256 colors.

// Fill the screen.
for ( unsigned int y = 0; y < 600; y++ )
for ( unsigned int x = 0; x < 800; x++ )

getch(); // Wait for a key.
quit(); // Set video-mode to normal.
return 0; // All OK :)


  • Here is good improvement:

    When you fill, say, 65536 pixels from the page #0 you call flip_page() 65536 times, but page is still #0, so it is useless code... What you should do is every time you call 'flip_page()' you store in global variable the page number and use it to decide if you should flip it:

    int glb_RamPage = -1;

    void flip (int nPage)
    if (nPage != glb_RamPage) {
    _asm // flip it here
    _asm // ...
    glb_RamPage = nPage;

    Here is another improvement:

    When you draw horizontal lines you should use 'memset()' or
    _asm rep stosb (or stosw (or stosd))
    Of course, you have to see if the line does not cross the banking line of 64Kb, if it is the you divide it on 2 line segments. For vertical lines also, do not calculate the offset every time - if scan line is 800 bytes then just add these bytes to the address and set a pixel...


Sign In or Register to comment.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


In this Discussion