Re: [PATCH v4 2/5] m68k/bitops: use __builtin_{clz,ctzl,ffs} to evaluate constant expressions

[Finn Thain]

On Sun, 28 Jan 2024, Vincent Mailhol wrote:

> The compiler is not able to do constant folding on "asm volatile" code.
> 
> Evaluate whether or not the function argument is a constant expression
> and if this is the case, return an equivalent builtin expression.
> 
> On linux 6.7 with an allyesconfig and GCC 13.2.1, it saves roughly 11 KB.
> 
>   $ size --format=GNU vmlinux.before vmlinux.after
>     text       data        bss      total filename
>     60457964   70953697    2288644  133700305 vmlinux.before
>     60441196   70957057    2290724  133688977 vmlinux.after
> 
> Reference: commit fdb6649ab7c1 ("x86/asm/bitops: Use __builtin_ctzl() to evaluate constant expressions")
> Link: https://git.kernel.org/torvalds/c/fdb6649ab7c1
> 
> Reviewed-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> Signed-off-by: Vincent Mailhol <mailhol.vincent@xxxxxxxxxx>
> ---
>  arch/m68k/include/asm/bitops.h | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/arch/m68k/include/asm/bitops.h b/arch/m68k/include/asm/bitops.h
> index a8b23f897f24..02ec8a193b96 100644
> --- a/arch/m68k/include/asm/bitops.h
> +++ b/arch/m68k/include/asm/bitops.h
> @@ -469,6 +469,9 @@ static __always_inline unsigned long ffz(unsigned long word)
>  {
>  	int res;
>  
> +	if (__builtin_constant_p(word))
> +		return __builtin_ctzl(~word);
> +
>  	__asm__ __volatile__ ("bfffo %1{#0,#0},%0"
>  			      : "=d" (res) : "d" (~word & -~word));
>  	return res ^ 31;

If the builtin has the desired behaviour, why do we reimplement it in asm? 
Shouldn't we abandon one or the other to avoid having to prove (and 
maintain) their equivalence?

> @@ -490,6 +493,9 @@ static __always_inline unsigned long ffz(unsigned long word)
>  	!defined(CONFIG_M68000)
>  static __always_inline unsigned long __ffs(unsigned long x)
>  {
> +	if (__builtin_constant_p(x))
> +		return __builtin_ctzl(x);
> +
>  	__asm__ __volatile__ ("bitrev %0; ff1 %0"
>  		: "=d" (x)
>  		: "0" (x));
> @@ -522,6 +528,9 @@ static __always_inline int ffs(int x)
>  {
>  	int cnt;
>  
> +	if (__builtin_constant_p(x))
> +		return __builtin_ffs(x);
> +
>  	__asm__ ("bfffo %1{#0:#0},%0"
>  		: "=d" (cnt)
>  		: "dm" (x & -x));
> @@ -540,6 +549,9 @@ static __always_inline int fls(unsigned int x)
>  {
>  	int cnt;
>  
> +	if (__builtin_constant_p(x))
> +		return x ? BITS_PER_TYPE(x) - __builtin_clz(x) : 0;
> +
>  	__asm__ ("bfffo %1{#0,#0},%0"
>  		: "=d" (cnt)
>  		: "dm" (x));
>