GNU Assembler not making sense

[scanish]

For reasons unknown I have to rewrite my startup assembly code for my kernel compatible for the GNU toolchain. I used the Netwide before, with Intel syntax, and my chums told me they want to stick strictly to the GNU chain. So I thought of it as a very nice opportunity to get familiar with AT&T syntax. But now it drives me mad.
I worked through a small reference on Memory Reference in Intel and in AT&T, but the hints there just give me errors I cannot work out.
This is the code I need to translate:

Code: Select all

	mov eax, _isr%1
	mov word [idt + 8 * %1], ax
	mov word [idt + 8 * %1 + 2], OS_CODE_DESC
	mov word [idt + 8 * %1 + 4], 0x8e00
	shr eax, 16
	mov word [idt + 8 * %1 + 6], ax

And my tries look like this:

Code: Select all

	movl isr\num, %eax
	movl idt, %ebx
	movw %ax, 0(%ebx, 8, \num)
	movw $OS_CODE_DESC, 2(%ebx, 8, \num)
	movw $0x8e00, 4(%ebx, 8, \num)
	shrl $16, %eax
	movw %ax, 6(%ebx, 8, \num)

So. The problem is basically: Any position I try for the references either makes AS tell me it's a wrong syntax or a wrong scale.

Please help...

[jrepan]

You can use .intel_syntax directive to use Intel syntax with GNU assembler.

[scanish]

You can? But the dotted directives are the same as with AT&T? I didn't know as supports Intel, sorry.

[jrepan]

But the dotted directives are the same as with AT&T?

I am quite sure they are (but I may be wrong, because I use NASM too). The simplest way to know for sure, is to just try and see if it works.

[scanish]

Thanks for the help. The .intel_syntax is not quite the same as with nasm, it still expects % on registers and $ on who knows what, but there was no quick documentary on it

It turned out the entire problem was that the gas does not like to have direct values in memory references. It wants them in a register, unlike nasm. But no documentary or translation site had any hint on that.

Sorry to have made that thread.

[Owen]

It would have helped if you told us specifically where the error was

[Solar]

Thanks for the help. The .intel_syntax is not quite the same as with nasm, it still expects % on registers and $ on who knows what, but there was no quick documentary on it

There's always ".intel_syntax no_prefix" to try.

Yes, GAS is somewhat under-documented.

[qw]

This is incorrect:

Code: Select all

0(%ebx, 8, \num)

GAS syntax is

Code: Select all

displacement(base, index, scale)

So it should be

Code: Select all

8 * \num(%ebx)

I haven't tested this, additional errors may occur.

[scanish]

Yes, thanks, that works, because \num is a constant.

Code: Select all

8*\num+2(%ebx)

Works perfectly... thanks a lot!

[qw]

You're welcome!