OSDev.org

Hi

This is my first real attempt at any AT&T inline asm and this is driving me arround the bend:

asm ("movl %esp, %0": "=g" (temp) : : "memory");

I want to move a value from esp to a local variable called temp but I get this error:

operand number missing after %-letter

What am I doing wrong?

thanks
srg

srg wrote: asm ("movl %esp, %0": "=g" (temp) : : "memory");
...
operand number missing after %-letter

asm ("movl %%esp, %0": "=g" (temp) : : "memory");

AT&T sucks.

Candy wrote:

srg wrote: asm ("movl %esp, %0": "=g" (temp) : : "memory");
...
operand number missing after %-letter

asm ("movl %%esp, %0": "=g" (temp) : : "memory");

AT&T sucks.

Well I'm still concidering coding any assembly as seperate routines in NASM, I'm nice and familliar with that.

One, GNU as (the assembler used for inline ASM in gcc) can also to Intel syntax.

Two, AT&T doesn't suck any more than any other ASM dialect, just because a user used the wrong syntax. If I forget a '$' in my Perl script, that's just as wrong and doesn't say Perl sucks.

Three, personally I consider AT&T syntax much superior because I have previous experience on 6502 and 68k CPUs - I consider Intel syntax to be seriously broken.

Solar wrote: One, GNU as (the assembler used for inline ASM in gcc) can also to Intel syntax.

In a broken way, with still most of the same problems Intel-syntax people had with at&t syntax.

Two, AT&T doesn't suck any more than any other ASM dialect, just because a user used the wrong syntax. If I forget a '$' in my Perl script, that's just as wrong and doesn't say Perl sucks.

I wasn't referring to the user error but more to the unintuitive (yes, I know it's only past experience, don't flame me now) concept of using one percent-sign with registers, but if you use arguments you have to modify them all to use 2 percent signs.

Three, personally I consider AT&T syntax much superior because I have previous experience on 6502 and 68k CPUs - I consider Intel syntax to be seriously broken.

I can understand that, have done 68HC11 programming myself. Still, for x86es the intel syntax is lots better (imo). I still can't figure out how exactly to encode a sib operation in AT&T syntax, but in intel syntax its just [xxx] with on the place of XXX some addition of some operands you'd like. Lots more intuitive than a 8(ebx, eax, 4) something to me...