Re: master 86c4e5a2fb3: Fix implicit declaration of bswap_{32,64}

Re: master 86c4e5a2fb3: Fix implicit declaration of bswap_{32,64}

[Eli Zaretskii]

> branch: master
> commit 86c4e5a2fb3fd6b7acb8a3fc10e1e7c2eb8012a9
> Author: Andreas Schwab <schwab@linux-m68k.org>
> Commit: Andreas Schwab <schwab@linux-m68k.org>
> 
>     Fix implicit declaration of bswap_{32,64}
>     
>     * src/data.c: Move include of <byteswap.h> ...
>     * src/lisp.h: ... here.

Wouldn't it be better to use __builtin_bswap32 when it's available?

Paul, why doesn't Gnulib do that in its byteswap.h header?

Re: master 86c4e5a2fb3: Fix implicit declaration of bswap_{32,64}

[Andreas Schwab]

On Mär 30 2024, Eli Zaretskii wrote:

> Wouldn't it be better to use __builtin_bswap32 when it's available?

The compiler is smart enough to do that by itself.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: master 86c4e5a2fb3: Fix implicit declaration of bswap_{32,64}

[Collin Funk]

On 3/30/24 1:34 AM, Andreas Schwab wrote:
>> Wouldn't it be better to use __builtin_bswap32 when it's available?
> The compiler is smart enough to do that by itself.

The macro would evaluate it's arguments multiple times though right?

This line:

    uint32_t result = bswap_32 (value++);

should increment 'value' once, but would do it four times with a
macro.

Collin

Re: master 86c4e5a2fb3: Fix implicit declaration of bswap_{32,64}

[Paul Eggert]

On 3/30/24 02:16, Eli Zaretskii wrote:
> Wouldn't it be better to use __builtin_bswap32 when it's available?
> 
> Paul, why doesn't Gnulib do that in its byteswap.h header?

I think Gnulib byteswap.h predates __builtin_bswap32. Not sure it's 
worth the hassle of using __builtin_bswap32 if available, as (as Andreas 
mentioned) that won't make Emacs go faster in the usual gcc -O2 case.