diff mbox

[media] mceusb: Optimize DIV_ROUND_CLOSEST call

Message ID 20120901205357.1a75d8a1@endymion.delvare (mailing list archive)
State New, archived
Headers show

Commit Message

Jean Delvare Sept. 1, 2012, 6:53 p.m. UTC
DIV_ROUND_CLOSEST is faster if the compiler knows it will only be
dealing with unsigned dividends.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
---
 drivers/media/rc/mceusb.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Guenter Roeck Sept. 2, 2012, 4:54 a.m. UTC | #1
On Sat, Sep 01, 2012 at 08:53:57PM +0200, Jean Delvare wrote:
> DIV_ROUND_CLOSEST is faster if the compiler knows it will only be
> dealing with unsigned dividends.
> 
> Signed-off-by: Jean Delvare <khali@linux-fr.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Guenter Roeck <linux@roeck-us.net>
> Cc: Mauro Carvalho Chehab <mchehab@infradead.org>

Tested-by: Guenter Roeck <linux@roeck-us.net>

> ---
>  drivers/media/rc/mceusb.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> --- linux-3.6-rc3.orig/drivers/media/rc/mceusb.c	2012-08-04 21:49:27.000000000 +0200
> +++ linux-3.6-rc3/drivers/media/rc/mceusb.c	2012-09-01 18:53:32.053042123 +0200
> @@ -627,7 +627,7 @@ static void mceusb_dev_printdata(struct
>  			break;
>  		case MCE_RSP_EQIRCFS:
>  			period = DIV_ROUND_CLOSEST(
> -					(1 << data1 * 2) * (data2 + 1), 10);
> +					(1U << data1 * 2) * (data2 + 1), 10);
>  			if (!period)
>  				break;
>  			carrier = (1000 * 1000) / period;
> 
> 
> -- 
> Jean Delvare
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Mauro Carvalho Chehab Sept. 18, 2012, 3:49 p.m. UTC | #2
Em 01-09-2012 15:53, Jean Delvare escreveu:
> DIV_ROUND_CLOSEST is faster if the compiler knows it will only be
> dealing with unsigned dividends.
> 
> Signed-off-by: Jean Delvare <khali@linux-fr.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Guenter Roeck <linux@roeck-us.net>
> Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
> ---
>  drivers/media/rc/mceusb.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> --- linux-3.6-rc3.orig/drivers/media/rc/mceusb.c	2012-08-04 21:49:27.000000000 +0200
> +++ linux-3.6-rc3/drivers/media/rc/mceusb.c	2012-09-01 18:53:32.053042123 +0200
> @@ -627,7 +627,7 @@ static void mceusb_dev_printdata(struct
>  			break;
>  		case MCE_RSP_EQIRCFS:
>  			period = DIV_ROUND_CLOSEST(
> -					(1 << data1 * 2) * (data2 + 1), 10);
> +					(1U << data1 * 2) * (data2 + 1), 10);
>  			if (!period)
>  				break;
>  			carrier = (1000 * 1000) / period;
> 
> 

Hmm... this generates the following warning with "W=1":

drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]

Perhaps it makes sense to use an optimized version for unsigned, or to
change the macro to take the data types into account.

Regards,
Mauro
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Guenter Roeck Sept. 18, 2012, 4:24 p.m. UTC | #3
On Tue, Sep 18, 2012 at 12:49:53PM -0300, Mauro Carvalho Chehab wrote:
> Em 01-09-2012 15:53, Jean Delvare escreveu:
> > DIV_ROUND_CLOSEST is faster if the compiler knows it will only be
> > dealing with unsigned dividends.
> > 
> > Signed-off-by: Jean Delvare <khali@linux-fr.org>
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Guenter Roeck <linux@roeck-us.net>
> > Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
> > ---
> >  drivers/media/rc/mceusb.c |    2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > --- linux-3.6-rc3.orig/drivers/media/rc/mceusb.c	2012-08-04 21:49:27.000000000 +0200
> > +++ linux-3.6-rc3/drivers/media/rc/mceusb.c	2012-09-01 18:53:32.053042123 +0200
> > @@ -627,7 +627,7 @@ static void mceusb_dev_printdata(struct
> >  			break;
> >  		case MCE_RSP_EQIRCFS:
> >  			period = DIV_ROUND_CLOSEST(
> > -					(1 << data1 * 2) * (data2 + 1), 10);
> > +					(1U << data1 * 2) * (data2 + 1), 10);
> >  			if (!period)
> >  				break;
> >  			carrier = (1000 * 1000) / period;
> > 
> > 
> 
> Hmm... this generates the following warning with "W=1":
> 
> drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
> drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
> 
> Perhaps it makes sense to use an optimized version for unsigned, or to
> change the macro to take the data types into account.
> 
DIV_ROUND_CLOSEST tests "((typeof(x))-1) >= 0" on purpose, so the compiler
can optimize the signed part of the macro away if the variable type is unsigned.
The test was borrowed from C99 code. Would be great if someone has an idea
how to tell the compiler not to create a warning for this test.

Guenter
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jean Delvare Sept. 18, 2012, 6:35 p.m. UTC | #4
Hi Mauro,

On Tue, 18 Sep 2012 12:49:53 -0300, Mauro Carvalho Chehab wrote:
> Em 01-09-2012 15:53, Jean Delvare escreveu:
> > DIV_ROUND_CLOSEST is faster if the compiler knows it will only be
> > dealing with unsigned dividends.
> > 
> > Signed-off-by: Jean Delvare <khali@linux-fr.org>
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Guenter Roeck <linux@roeck-us.net>
> > Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
> > ---
> >  drivers/media/rc/mceusb.c |    2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > --- linux-3.6-rc3.orig/drivers/media/rc/mceusb.c	2012-08-04 21:49:27.000000000 +0200
> > +++ linux-3.6-rc3/drivers/media/rc/mceusb.c	2012-09-01 18:53:32.053042123 +0200
> > @@ -627,7 +627,7 @@ static void mceusb_dev_printdata(struct
> >  			break;
> >  		case MCE_RSP_EQIRCFS:
> >  			period = DIV_ROUND_CLOSEST(
> > -					(1 << data1 * 2) * (data2 + 1), 10);
> > +					(1U << data1 * 2) * (data2 + 1), 10);
> >  			if (!period)
> >  				break;
> >  			carrier = (1000 * 1000) / period;
>
> Hmm... this generates the following warning with "W=1":
> 
> drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
> drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]

I doubt this is the only warning of that kind. There must be a reason
why -Wextra isn't enabled by default.

> Perhaps it makes sense to use an optimized version for unsigned, or to
> change the macro to take the data types into account.

This was discussed before, but Andrew said he preferred a single macro.
And I agree with him, having two macros would induce a risk of the
wrong one being called.

If you can come up with a variant of DIV_ROUND_CLOSEST which performs
the same and doesn't trigger the warning above, we'll be happy to see
it, but neither Guenter nor myself could come up with one.
Guenter Roeck Sept. 19, 2012, 3:46 a.m. UTC | #5
On Tue, Sep 18, 2012 at 08:35:09PM +0200, Jean Delvare wrote:
> Hi Mauro,
> 
> On Tue, 18 Sep 2012 12:49:53 -0300, Mauro Carvalho Chehab wrote:
> > Em 01-09-2012 15:53, Jean Delvare escreveu:
> > > DIV_ROUND_CLOSEST is faster if the compiler knows it will only be
> > > dealing with unsigned dividends.
> > > 
> > > Signed-off-by: Jean Delvare <khali@linux-fr.org>
> > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > Cc: Guenter Roeck <linux@roeck-us.net>
> > > Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
> > > ---
> > >  drivers/media/rc/mceusb.c |    2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > --- linux-3.6-rc3.orig/drivers/media/rc/mceusb.c	2012-08-04 21:49:27.000000000 +0200
> > > +++ linux-3.6-rc3/drivers/media/rc/mceusb.c	2012-09-01 18:53:32.053042123 +0200
> > > @@ -627,7 +627,7 @@ static void mceusb_dev_printdata(struct
> > >  			break;
> > >  		case MCE_RSP_EQIRCFS:
> > >  			period = DIV_ROUND_CLOSEST(
> > > -					(1 << data1 * 2) * (data2 + 1), 10);
> > > +					(1U << data1 * 2) * (data2 + 1), 10);
> > >  			if (!period)
> > >  				break;
> > >  			carrier = (1000 * 1000) / period;
> >
> > Hmm... this generates the following warning with "W=1":
> > 
> > drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
> > drivers/media/rc/mceusb.c:629:4: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
> 
> I doubt this is the only warning of that kind. There must be a reason
> why -Wextra isn't enabled by default.
> 
> > Perhaps it makes sense to use an optimized version for unsigned, or to
> > change the macro to take the data types into account.
> 
> This was discussed before, but Andrew said he preferred a single macro.
> And I agree with him, having two macros would induce a risk of the
> wrong one being called.
> 
> If you can come up with a variant of DIV_ROUND_CLOSEST which performs
> the same and doesn't trigger the warning above, we'll be happy to see
> it, but neither Guenter nor myself could come up with one.
> 
I did some more research, and I think I found a fix. I'll send out a patch
in a minute for people to try.

Guenter
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

--- linux-3.6-rc3.orig/drivers/media/rc/mceusb.c	2012-08-04 21:49:27.000000000 +0200
+++ linux-3.6-rc3/drivers/media/rc/mceusb.c	2012-09-01 18:53:32.053042123 +0200
@@ -627,7 +627,7 @@  static void mceusb_dev_printdata(struct
 			break;
 		case MCE_RSP_EQIRCFS:
 			period = DIV_ROUND_CLOSEST(
-					(1 << data1 * 2) * (data2 + 1), 10);
+					(1U << data1 * 2) * (data2 + 1), 10);
 			if (!period)
 				break;
 			carrier = (1000 * 1000) / period;