Re: Endianness macros capitalization

[Pavel Roskin]

On Sun, 2008-07-06 at 00:54 +0200, Javier Martín wrote:
> El sáb, 05-07-2008 a las 17:30 -0400, Pavel Roskin escribió:
> > They probably should be functions.  We may want to sparse annotate GRUB
> > one day, and then inline functions in the only way to go.
> Hmm... you mean changing this
> 
> #define grub_swap_bytes16(x)  \
> ({ \
>    grub_uint16_t _x = (x); \
>    (grub_uint16_t) ((_x << 8) | (_x >> 8)); \
> })
> 
> ...for this
> 
> inline grub_uint16_t grub_swap_bytes16(uint16_t x)
> {
>   return (x << 8) | (x >> 8);
> }
> 
> and such?

Actually, grub_swap_bytes16 should be avoided in the general code.  We
need to convert macros like grub_cpu_to_le16() to functions.

>  The pro is that we get rid of the ugly hack in the macro
> version that ensures single evaluation, but a con is that we cannot
> _force_ any random compiler to inline anything, so we might end up with
> gross numbers of function calls.

We can force inlining in gcc with always_inline.  But we don't need to
force anything.  Modern versions of gcc can decide for us, but we can
influence the choice by using optimization flags, such as -O2 and -Os.
In come cases, function calls would take less space than the actual code
for byte swapping.

> By the way, what is "sparse annotate"?

There is a utility called sparse:
git://git.kernel.org/pub/scm/linux/kernel/git/josh/sparse.git

It acts like a compiler, but it only produces warnings.  It can
distinguish bitwise types that cannot be treated like normal integers
for arithmetic operations like addition.

If we mark little-endian and big-endian variables as bitwise, sparse
will warn if such variables are used without byte swapping.  Linux uses
it a lot.  Sparse is really a very good checker for that purpose.
Sparse does other checks too.

-- 
Regards,
Pavel Roskin