Re: [AD] More AMD64 stuff

[Evert Glebbeek <eglebbk@xxxxxxxxxx>]

> > Currently, the only thing this is used for is to port over the CPU 
> > identification code. This seems to work like a charm, but I'm not too 
> > comfortable with the way I had to grab the arguments passed to the 
> > function: %rdi gets the first argument, %rsi gets the second one. I did 
> > some googling and checking around and this does seem to be normal (and 
it 
> > certainly simplifies the code significantly), but I'd like to hear more 
> > opinions on it.
> 
> 
> clue: 0.0%  sorry i have absolutely zero understanding of this %rdi 
thingy.

On the i386, the normal convention is to pass arguments on the stack and 
set up a stackframe to read them in (but I think you can also specify that 
you want to use registers). On the AMD64, at least when interfacing with a 
higher level language, the convention seems to be that the arguments 
themselves are passed in registers: %rdi (the 64 bit version of the %edi 
register) holds the first argument, %rsi holds the second argument, etc.

> i tend to think its best left upto the compiler if possible.

Yeah, you're probably right. If someone feels adventurous, the output from 
gcc would probably be a good place to start. It'll be a mess to get the 
alignment of structure members right though.

> in MSVC7  intrinsics can be used to make nice C-like code that uses 
> mmx/sse instructions.  does GCC support them?

I'm not sure. I've used inline code like

inline int bsr(unsigned int n)
{
   asm ("bsr %%eax, %%eax": "=a" (n) : "a" (n) );
   return n;
}

but I'm not sure if that's the sort of thing you're refering to?

> whats with the  sprintf  and  strcat ?
> my spelling usually contains 'N's

The test programme doesn't use snprintf or strncat. You're right that it 
should probably be changed for that, so feel free to submit a patch for 
this ;)

Evert