Re: [PATCH 0/7] preempt_count rework -v2

[Linus Torvalds]

On Tue, Sep 10, 2013 at 6:56 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
> +static __always_inline bool __preempt_count_dec_and_test(void)
> +{
> +       unsigned char c;
> +
> +       asm ("decl " __percpu_arg(0) "; sete %1"
> +                       : "+m" (__preempt_count), "=qm" (c));
> +
> +       return c != 0;
> +}
>
> And that's where the sete and test originates from.

We could make this use "asm goto" instead.

An "asm goto" cannot have outputs, but this particular one doesn't
_need_ outputs. You could mark the preempt_count memory as an input,
and then have a memory clobber. I think you need the memory clobber
anyway for that preempt-count thing.

So I _think_ something like

static __always_inline bool __preempt_count_dec_and_test(void)
{
       asm goto("decl " __percpu_arg(0) "\n\t"
                "je %l[became_zero]"
                       : :"m" (__preempt_count):"memory":became_zero);
       return 0;
became_zero:
       return 1;
}

would work.

You need to wrap it in

  #ifdef CC_HAVE_ASM_GOTO

and then have the old "sete" version for older compilers, but for
newer ones you'd get pretty much perfect code. UNTESTED.

             Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/