Date:   Thu, 9 Mar 2023 22:41:01 -0800
From:   Eric Biggers <ebiggers@kernel.org>
To:     Yangtao Li <frank.li@vivo.com>
Cc:     tytso@mit.edu, adilger.kernel@dilger.ca,
        hsiangkao@linux.alibaba.com, linux-ext4@vger.kernel.org,
        linux-erofs@lists.ozlabs.org, linux-kernel@vger.kernel.org,
        viro@zeniv.linux.org.uk
Subject: Re: [PATCH] ext4: convert to DIV_ROUND_UP() in
 mpage_process_page_bufs()
Message-ID: <ZArQ/WaBrw0LR56v@sol.localdomain>
References: <ZArLbO1ckmcXwQf0@sol.localdomain>
 <20230310062738.69663-1-frank.li@vivo.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230310062738.69663-1-frank.li@vivo.com>
Precedence: bulk

On Fri, Mar 10, 2023 at 02:27:38PM +0800, Yangtao Li wrote:
> > Please don't do this.  This makes the code compile down to a division, which is
> > far less efficient.  I've verified this by checking the assembly generated.
> 
> How much is the performance impact? So should the following be modified as shift operations as well?
> 
> fs/erofs/namei.c:92:    int head = 0, back = DIV_ROUND_UP(dir->i_size, EROFS_BLKSIZ) - 1;
> fs/erofs/zmap.c:252:    const unsigned int totalidx = DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
> fs/erofs/decompressor.c:14:#define LZ4_MAX_DISTANCE_PAGES       (DIV_ROUND_UP(LZ4_DISTANCE_MAX, PAGE_SIZE) + 1)
> fs/erofs/decompressor.c:56:                                     DIV_ROUND_UP(distance, PAGE_SIZE) + 1 :
> fs/erofs/decompressor.c:70:     unsigned long bounced[DIV_ROUND_UP(LZ4_MAX_DISTANCE_PAGES,
> fs/erofs/data.c:84:     nblocks = DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
> 

No, compilers can handle division by a power-of-2 constant just fine.  It's just
division by a non-constant that is inefficient.

- Eric