Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Sat, 23 Mar 2019 20:31:53 -0700
From:   Nathan Chancellor <natechancellor@gmail.com>
To:     Sultan Alsawaf <sultan@kerneltoast.com>
Cc:     akpm@linux-foundation.org, linux-kernel@vger.kernel.org,
        arnd@arndb.de, keescook@chromium.org, linux@rasmusvillemoes.dk,
        rostedt@goodmis.org, torvalds@linux-foundation.org,
        viro@zeniv.linux.org.uk
Subject: Re: [RFCv2] string: Use faster alternatives when constant arguments
 are used
Message-ID: <20190324033153.GA15815@archlinux-ryzen>
References: <20190324014445.28688-1-sultan@kerneltoast.com>
 <20190324022406.GA18988@sultan-box.localdomain>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20190324022406.GA18988@sultan-box.localdomain>
User-Agent: Mutt/1.11.4 (2019-03-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sat, Mar 23, 2019 at 07:24:06PM -0700, Sultan Alsawaf wrote:
> I messed up the return value for strcat in the first patch. Here's a fixed
> version, ready for some scathing reviews.
> 
> From: Sultan Alsawaf <sultan@kerneltoast.com>
> 
> When strcpy, strcat, and strcmp are used with a literal string, they can
> be optimized to memcpy or memcmp calls. These alternatives are faster
> since knowing the length of a string argument beforehand allows
> traversal through the string word at a time without being concerned
> about looking for the terminating zero character. In some cases, the
> replaced calls to memcpy or memcmp can even be optimized out completely
> for a significant speed up.
> 
> Signed-off-by: Sultan Alsawaf <sultan@kerneltoast.com>
> ---
>  include/linux/string.h | 30 ++++++++++++++++++++++++++++++
>  1 file changed, 30 insertions(+)
> 
> diff --git a/include/linux/string.h b/include/linux/string.h
> index 7927b875f..59c301c0e 100644
> --- a/include/linux/string.h
> +++ b/include/linux/string.h
> @@ -476,4 +476,34 @@ static __always_inline size_t str_has_prefix(const char *str, const char *prefix
>  	return strncmp(str, prefix, len) == 0 ? len : 0;
>  }
>  
> +/*
> + * Replace some common string helpers with faster alternatives when one of the
> + * arguments is a constant (i.e., literal string). This uses strlen instead of
> + * sizeof for calculating the string length in order to silence compiler
> + * warnings that may arise due to what the compiler thinks is incorrect sizeof
> + * usage. The strlen calls on constants are folded into scalar values at compile
> + * time, so performance is not reduced by using strlen.
> + */
> +#define strcpy(dest, src)							\
> +	__builtin_choose_expr(__builtin_constant_p(src),			\
> +		memcpy((dest), (src), strlen(src) + 1),				\
> +		(strcpy)((dest), (src)))
> +
> +#define strcat(dest, src)							\
> +	__builtin_choose_expr(__builtin_constant_p(src),			\
> +		({								\
> +			memcpy(strchr((dest), '\0'), (src), strlen(src) + 1);	\
> +			(dest);							\
> +		}),								\
> +		(strcat)((dest), (src)))
> +
> +#define strcmp(dest, src)							\
> +	__builtin_choose_expr(__builtin_constant_p(dest),			\
> +		__builtin_choose_expr(__builtin_constant_p(src),		\
> +			(strcmp)((dest), (src)),				\
> +			memcmp((dest), (src), strlen(dest) + 1)),		\
> +		__builtin_choose_expr(__builtin_constant_p(src),		\
> +			memcmp((dest), (src), strlen(src) + 1),			\
> +			(strcmp)((dest), (src))))
> +
>  #endif /* _LINUX_STRING_H_ */
> -- 
> 2.21.0
> 

Explicitly cc'ing some folks who have touched include/linux/string.h in
the past and might want to take a look at this.

Nathan