[PATCH] block: convert ata_port_for_each_link to use a while loop to reduce code size

[PATCH] block: convert ata_port_for_each_link to use a while loop to reduce code size

[Richard Kennedy]

convert a for loop to a while loop in the ata_port_for_each_link macros
to reduce code size.

on x86_64 size reports reduction of text size in
	libata.ko : -272 bytes
	ahci.ko   : -44 bytes

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>

----
This patch is against 2.6.27 git head.

I've been running this patch on my AMD64 desktop machine for several
days & have not seen any problems.
regards
Richard


diff --git a/include/linux/libata.h b/include/linux/libata.h
index 947cf84..bbe3b9d 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -1283,12 +1283,12 @@ extern struct ata_link *__ata_port_next_link(struct ata_port *ap,
 					     bool dev_only);
 
 #define __ata_port_for_each_link(link, ap) \
-	for ((link) = __ata_port_next_link((ap), NULL, false); (link); \
-	     (link) = __ata_port_next_link((ap), (link), false))
+	(link) = NULL; \
+	while (((link) = __ata_port_next_link((ap), (link), false)))
 
 #define ata_port_for_each_link(link, ap) \
-	for ((link) = __ata_port_next_link((ap), NULL, true); (link); \
-	     (link) = __ata_port_next_link((ap), (link), true))
+	(link) = NULL; \
+	while (((link) = __ata_port_next_link((ap), (link), true)))
 
 #define ata_link_for_each_dev(dev, link) \
 	for ((dev) = (link)->device; \

Re: [PATCH] block: convert ata_port_for_each_link to use a while loop to reduce code size

[Alexander van Heukelum]

On Tue, 21 Oct 2008 21:24:32 +0100, "Richard Kennedy"
<richard@rsk.demon.co.uk> said:
> convert a for loop to a while loop in the ata_port_for_each_link macros
> to reduce code size.
> 
> on x86_64 size reports reduction of text size in
> 	libata.ko : -272 bytes
> 	ahci.ko   : -44 bytes
> 
> Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
> 
> ----
> This patch is against 2.6.27 git head.
> 
> I've been running this patch on my AMD64 desktop machine for several
> days & have not seen any problems.
> regards
> Richard
> 
> 
> diff --git a/include/linux/libata.h b/include/linux/libata.h
> index 947cf84..bbe3b9d 100644
> --- a/include/linux/libata.h
> +++ b/include/linux/libata.h
> @@ -1283,12 +1283,12 @@ extern struct ata_link
> *__ata_port_next_link(struct ata_port *ap,
>  					     bool dev_only);
>  
>  #define __ata_port_for_each_link(link, ap) \
> -	for ((link) = __ata_port_next_link((ap), NULL, false); (link); \
> -	     (link) = __ata_port_next_link((ap), (link), false))
> +	(link) = NULL; \
> +	while (((link) = __ata_port_next_link((ap), (link), false)))

Hi Richard,

Please imagine how this expands...

if (some_condition)
        __ata_port_for_each_link(link, ap)
                do_something(...)

But it is probably possible to get the same size reduction by
changing the for-loop like this (safety parentheses left out):

        for (link = NULL;
                link = __ata_port_next_link(ap, link, false), link; )

Some programmers think this is abuse of the for-construct, though ;).

Greetings,
    Alexander

>  #define ata_port_for_each_link(link, ap) \
> -	for ((link) = __ata_port_next_link((ap), NULL, true); (link); \
> -	     (link) = __ata_port_next_link((ap), (link), true))
> +	(link) = NULL; \
> +	while (((link) = __ata_port_next_link((ap), (link), true)))
>  
>  #define ata_link_for_each_dev(dev, link) \
>  	for ((dev) = (link)->device; \
-- 
  Alexander van Heukelum
  heukelum@fastmail.fm

-- 
http://www.fastmail.fm - mmm... Fastmail...

Re: [PATCH] block: convert ata_port_for_each_link to use a while loop to reduce code size

[Richard Kennedy]

On Wed, 2008-10-22 at 08:17 +0200, Alexander van Heukelum wrote:
> On Tue, 21 Oct 2008 21:24:32 +0100, "Richard Kennedy"
> <richard@rsk.demon.co.uk> said:
> > convert a for loop to a while loop in the ata_port_for_each_link macros
> > to reduce code size.
> > 
> > on x86_64 size reports reduction of text size in
> > 	libata.ko : -272 bytes
> > 	ahci.ko   : -44 bytes
> > 
> > Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
> > 
> > ----
> > This patch is against 2.6.27 git head.
> > 
> > I've been running this patch on my AMD64 desktop machine for several
> > days & have not seen any problems.
> > regards
> > Richard
> > 
> > 
> > diff --git a/include/linux/libata.h b/include/linux/libata.h
> > index 947cf84..bbe3b9d 100644
> > --- a/include/linux/libata.h
> > +++ b/include/linux/libata.h
> > @@ -1283,12 +1283,12 @@ extern struct ata_link
> > *__ata_port_next_link(struct ata_port *ap,
> >  					     bool dev_only);
> >  
> >  #define __ata_port_for_each_link(link, ap) \
> > -	for ((link) = __ata_port_next_link((ap), NULL, false); (link); \
> > -	     (link) = __ata_port_next_link((ap), (link), false))
> > +	(link) = NULL; \
> > +	while (((link) = __ata_port_next_link((ap), (link), false)))
> 
> Hi Richard,
> 
> Please imagine how this expands...
> 
> if (some_condition)
>         __ata_port_for_each_link(link, ap)
>                 do_something(...)
> 
> But it is probably possible to get the same size reduction by
> changing the for-loop like this (safety parentheses left out):
> 
>         for (link = NULL;
>                 link = __ata_port_next_link(ap, link, false), link; )
> 
> Some programmers think this is abuse of the for-construct, though ;).
> 
> Greetings,
>     Alexander
> 
> >  #define ata_port_for_each_link(link, ap) \
> > -	for ((link) = __ata_port_next_link((ap), NULL, true); (link); \
> > -	     (link) = __ata_port_next_link((ap), (link), true))
> > +	(link) = NULL; \
> > +	while (((link) = __ata_port_next_link((ap), (link), true)))
> >  
> >  #define ata_link_for_each_dev(dev, link) \
> >  	for ((dev) = (link)->device; \
> -- 
>   Alexander van Heukelum
>   heukelum@fastmail.fm

yes of course, that is the problem with this kind of macro :(

IMHO, this particular macro doesn't improve the overall code readability
so would be better being removed altogether. Whenever possible I much
prefer being able to see a for or while loop in the code rather than
hidden in a macro somewhere, but maybe that's just me ;)
regards
Richard