INT 15 to return memory size

I am writing a program for DOS to return the extended memory size but it is not working properly. The code I have is below

Code:

	_asm 
      {
        xor ax,ax       /* Clear accumulator*/
        mov ah,0x88      /* move hex 88 into AH*/
        nop
	nop
	nop
	int 15h
	mov memory, ax
	}
		
	printf("\nThe extended memory size is %d (0x%X).\n", memory, memory);

This is returning a value of 0, but my ram is 512MB.

What am I doing wrong? Does it have to do with the registers being 16 bit?

Also I will need to get the base memory size as well, but that is next on the list after I can figure this part out.

Probably register size limitation. If your memory is 512MB, then the number of 1K blocks, which is what the interrupt returns, would be 524288. That would require an integer of at least 19 bits to represent.

According to one place I found, "works only on 80286 and 80386 machines". My guess is the amount of RAM during those machines was never anticipated to go that high.

http://www.htl-steyr.ac.at/~morg/pci...s/inte7dps.htm

You'll probably need to look at function 0xe820.

> What am I doing wrong? Does it have to do with the registers being 16 bit?
And your compiler being over 20 years old, and that the OS for which it was written never knew that more than 1MB of memory ever existed.

Backwards compatibility only stretches so far. You've reached the end of the road - time to upgrade.

Take a look at Detecting Memory (x86) - OSDev Wiki. It lists ah=88 as one of the backups with many issues. Like kennedy said, use eax=e820. The link I gave has an example of how to use the output of eax=e820.

ok i am using INT 15 ax=0xe820 now but i am having issues with the upper 16 bits of the ax register. i looked at the DOS.h file but it doesn't have anyway to write to the eax register, and if i try something like:

Code:

_asm
{
move eax, 0x0000e820
}

my compiler doesn't like "eax", i can only use ax

Like I said - powerpoint presentation using stone chisels.

When are you going to realise you have the wrong tool for the job?

I don't want to do this, it is an assignment

Well I hope you mention some of the difficulties you overcame in your report.

I think it's mov.

Originally Posted by nonoob

I think it's mov.

Wow. I've read that line numerous times. Read yours too. Never saw the difference in the move and the mov. I had to read yours like 5 times to figure out the difference.

Isn't it amazing that we get so that we read the real thing even though what is printed is wrong.

Good catch nonoob!

Hehe, maybe you still had some Motorola 68000 cobwebs in your brain.

WTF is up with those nops and the xor ax,ax before you move 88h into ah?(I just now took the time to actually look at the code, before I just noticed the fact OP is using ah=88.)

Code:

xor ax,ax
mov ah,*

Just use "mov ax,8800h"

Code:

nop
nop
nop
...

WHY!?

Just kidding, but really, none of that is needed. I realize there's no need for so-called optimization, but there's also no reason to literally waste clock cycles . nops take up space in the cache and cause extra fetching for no reason in this case.

One important thing to note about eax=e820 is that even though e820 can fit in ax, it's expected that the top word of eax be cleared. I doubt it would cause your computer to explode, but it's better to be safe.

Originally Posted by Salem

> What am I doing wrong? Does it have to do with the registers being 16 bit?
And your compiler being over 20 years old, and that the OS for which it was written never knew that more than 1MB of memory ever existed.

Backwards compatibility only stretches so far. You've reached the end of the road - time to upgrade.

Every time you boot an Intel computer, it starts up in 16-bit mode and the OS boot loader has to use a BIOS request exactly like this to figure out how much RAM is available. Day-to-day programmers don't have to do things like this any more but it's by no means outdated. We all still have BIOSes and they are all still 16-bit.

Intel doesn't make computers .

> the OS for which it was written never knew that more than 1MB of memory ever existed.
MS-DOS 5.0 introduced HIMEM.SYS, which is the driver used to enable the A20 line extending the addressable memory to 16MB . Talking about backwards compatibility, it's stupid how much we have retained from the time of DOS. The old way of enabling the A20 was through a pin in the keyboard controller! This way is still maintained and used today(although there is a BIOS int to do it). The A20 originated from an attempt in the 80286 to keep compatibility with the 8086. Such a waste of time, huh?