Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Thu, 15 Jul 2021 08:50:21 -0700
From:   Yury Norov <yury.norov@gmail.com>
To:     Steven Rostedt <rostedt@goodmis.org>
Cc:     linux-kernel@vger.kernel.org,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Rasmus Villemoes <linux@rasmusvillemoes.dk>,
        Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <senozhatsky@chromium.org>,
        Tejun Heo <tj@kernel.org>
Subject: Re: [PATCH 2/2] bitmap: introduce for_each_set_bitrange
Message-ID: <YPBZPbCgJPjV2qPW@yury-ThinkPad>
References: <20210709034519.2859777-1-yury.norov@gmail.com>
 <20210709034519.2859777-3-yury.norov@gmail.com>
 <20210709095950.6a451ccb@oasis.local.home>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210709095950.6a451ccb@oasis.local.home>
Precedence: bulk

On Fri, Jul 09, 2021 at 09:59:50AM -0400, Steven Rostedt wrote:
> On Thu,  8 Jul 2021 20:45:19 -0700
> Yury Norov <yury.norov@gmail.com> wrote:
> 
> > bitmap_list_string() is very ineffective when printing bitmaps with long
> > ranges of set bits because it calls find_next_bit for each bit. We can do
> > better by detecting ranges of set bits.
> > 
> > This patch introduces a macro for_each_set_bitrange and uses it in
> > bitmap_list_string(). In my environment, before/after is 943008/31008 ns.
> > 
> > Signed-off-by: Yury Norov <yury.norov@gmail.com>
> > ---
> >  include/linux/find.h |  7 +++++++
> >  lib/vsprintf.c       | 40 ++++++++++++++++------------------------
> >  2 files changed, 23 insertions(+), 24 deletions(-)
> > 
> > diff --git a/include/linux/find.h b/include/linux/find.h
> > index ae9ed52b52b8..1a5ed45dc81b 100644
> > --- a/include/linux/find.h
> > +++ b/include/linux/find.h
> > @@ -301,6 +301,13 @@ unsigned long find_next_bit_le(const void *addr, unsigned
> >  	     (bit) < (size);					\
> >  	     (bit) = find_next_zero_bit((addr), (size), (bit) + 1))
> >  
> > +#define for_each_set_bitrange(b, e, addr, size)			\
> 
> The above needs a kerneldoc header.

OK. 

> 
> > +	for ((b) = find_next_bit((addr), (size), 0),		\
> > +	     (e) = find_next_zero_bit((addr), (size), (b) + 1);	\
> > +	     (b) < (size);					\
> > +	     (b) = find_next_bit((addr), (size), (e) + 1),	\
> > +	     (e) = find_next_zero_bit((addr), (size), (b) + 1))
> > +
> >  /**
> >   * for_each_set_clump8 - iterate over bitmap for each 8-bit clump with set bits
> >   * @start: bit offset to start search and to store the current iteration offset
> > diff --git a/lib/vsprintf.c b/lib/vsprintf.c
> > index 87acf66f0e4c..1ee54dace71e 100644
> > --- a/lib/vsprintf.c
> > +++ b/lib/vsprintf.c
> > @@ -1240,38 +1240,30 @@ char *bitmap_list_string(char *buf, char *end, unsigned long *bitmap,
> >  			 struct printf_spec spec, const char *fmt)
> >  {
> >  	int nr_bits = max_t(int, spec.field_width, 0);
> > -	/* current bit is 'cur', most recently seen range is [rbot, rtop] */
> > -	int cur, rbot, rtop;
> > -	bool first = true;
> > +	char *start = buf;
> > +	int b, e;
> >  
> >  	if (check_pointer(&buf, end, bitmap, spec))
> >  		return buf;
> >  
> > -	rbot = cur = find_first_bit(bitmap, nr_bits);
> > -	while (cur < nr_bits) {
> > -		rtop = cur;
> > -		cur = find_next_bit(bitmap, nr_bits, cur + 1);
> > -		if (cur < nr_bits && cur <= rtop + 1)
> > -			continue;
> > +	for_each_set_bitrange(b, e, bitmap, nr_bits) {
> > +		buf = number(buf, end, b, default_dec_spec);
> > +		if (e == b + 1)
> > +			goto put_comma;
> 
> Using a goto to skip a few lines instead of just having the reverse
> conditional is rather sloppy IMO.
> 
> 		if (e != b + 1) {
> 			if (buf < end)
> 				*buf = '-';
> 			buf++;
> 			buf = number(buf, end, e - 1, default_dec_spec);
> 		}
> 
> Is much clearer.
 	
I don't think it's clearer, but as you wish.

> >  
> > -		if (!first) {
> > -			if (buf < end)
> > -				*buf = ',';
> > -			buf++;
> > -		}
> > -		first = false;
> > +		if (buf < end)
> > +			*buf = '-';
> >  
> > -		buf = number(buf, end, rbot, default_dec_spec);
> > -		if (rbot < rtop) {
> > -			if (buf < end)
> > -				*buf = '-';
> > -			buf++;
> > +		buf = number(++buf, end, e - 1, default_dec_spec);
> > +put_comma:
> > +		if (buf < end)
> > +			*buf = ',';
> > +		buf++;
> > +	}
> >  
> > -			buf = number(buf, end, rtop, default_dec_spec);
> > -		}
> > +	if (buf > start)
> > +		buf--;
> 
> If the above is to undo the last comma, please put back the first logic.
> 
> -- Steve

You're asking me to move part of the logic inside the loop which generally
should be avoided. Is there any particular reason to do this?

> 
> >  
> > -		rbot = cur;
> > -	}
> >  	return buf;
> >  }
> >