While grebbing through the latest 2.2.0-pre5 I wonder why the "=A"
constraint isn't used for (atleast) the rdtsc asm's.
IMHO this optimizes quite better in functions returning 'long long'
(aka edx:eax).
The following test program
static inline unsigned long long read_rdtsc(void)
{
unsigned long long d;
__asm__ __volatile__ ("rdtsc" : "=A" (d) );
return d;
}
unsigned long long d;
unsigned long long xxx(void)
{
d = read_rdtsc();
return d;
}
compiles (-O2) as
xxx:
pushl %ebp
movl %esp,%ebp
#APP
rdtsc
#NO_APP
movl %eax,d
movl %edx,d+4
movl %ebp,%esp
popl %ebp
ret
which is IMHO realy optimal, much better as what you get using constructs
such as
__asm__("rdtsc":"=a" (eax), "=d" (edx));
Are there any problems with "=A' and the compilers, that I so far haven't
seen?
Hans
<lermen@fgan.de>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/