[PATCH] mm/slab.c, kernel <2.4.20>

[PATCH] mm/slab.c, kernel <2.4.20>

[Shangc]

/**
 * --readme--
 * make a better initial value for i, which will
reduce the number of 
while loop.
 * the change can be proved mathematically:
 *
 * i*size + L1_CACHE_ALIGN(base+i*extra) <= wastage
 * ==> i*size + base + i*extra <= wastage
 * ==> solve about equation about i
 * ==> i <= (wastage - base) / (size + extra)
 *
 * Chen Shang
 * shangcs@yahoo.com
 */

--- slab.c	2003-02-08 04:26:50.000000000 -0500
+++ slab.c	2003-02-08 04:26:14.000000000 -0500
@@ -397,7 +397,7 @@
 		base = sizeof(slab_t);
 		extra = sizeof(kmem_bufctl_t);
 	}
-	i = 0;
+	i = (wastage - base) / (size + extra);
 	while (i*size + L1_CACHE_ALIGN(base+i*extra) <=
wastage)
 		i++;
 	if (i > 0)

__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com

Re: [PATCH] mm/slab.c, kernel <2.4.20>

[Marc-Christian Petersen]

[-- Attachment #1: Type: text/plain, Size: 462 bytes --]

On Wednesday 08 January 2003 23:03, Shangc wrote:

Hi Shangc,

> --- slab.c	2003-02-08 04:26:50.000000000 -0500
> +++ slab.c	2003-02-08 04:26:14.000000000 -0500
> @@ -397,7 +397,7 @@
>  		base = sizeof(slab_t);
>  		extra = sizeof(kmem_bufctl_t);
>  	}
> -	i = 0;
> +	i = (wastage - base) / (size + extra);
>  	while (i*size + L1_CACHE_ALIGN(base+i*extra) <=> wastage)
>  		i++;
>  	if (i > 0)
if you use this you may also want this.

ciao, Marc

[-- Attachment #2: slab-other.patch --]
[-- Type: text/x-diff, Size: 434 bytes --]

diff -Naurp a/mm/slab.c b/mm/slab.c
--- a/mm/slab.c	Wed Jul 17 12:25:04 2002
+++ b/mm/slab.c	Wed Jul 17 12:25:04 2002
@@ -399,10 +399,10 @@
 		base = sizeof(slab_t);
 		extra = sizeof(kmem_bufctl_t);
 	}
-	i = 0;
+       i = (wastage - base)/(size + extra);
 	while (i*size + L1_CACHE_ALIGN(base+i*extra) <= wastage)
 		i++;
-	if (i > 0)
+       while (i*size + L1_CACHE_ALIGN(base+i*extra) > wastage)
 		i--;
 
 	if (i > SLAB_LIMIT)

Re: [PATCH] mm/slab.c, kernel <2.4.20>

[Shangc]

thank you for reply. However, I am not very sure to be
vary agree with you.

after assigned the calculated initial value for "i",
it is very close to the break value for the next while
loop, actually, it is only one loop at most, so "i"
never goes to be 
i*size + L1_CACHE_ALIGN(base+i*extra) >  wastage +
L1_CACHE_BYTES

so, we do not need change the line 399 of slab.c.

thanks,

chen

--- Marc-Christian Petersen <m.c.p@wolk-project.de>
wrote:
> On Wednesday 08 January 2003 23:03, Shangc wrote:
> 
> Hi Shangc,
> 
> > --- slab.c	2003-02-08 04:26:50.000000000 -0500
> > +++ slab.c	2003-02-08 04:26:14.000000000 -0500
> > @@ -397,7 +397,7 @@
> >  		base = sizeof(slab_t);
> >  		extra = sizeof(kmem_bufctl_t);
> >  	}
> > -	i = 0;
> > +	i = (wastage - base) / (size + extra);
> >  	while (i*size + L1_CACHE_ALIGN(base+i*extra) <=>
> wastage)
> >  		i++;
> >  	if (i > 0)
> if you use this you may also want this.
> 
> ciao, Marc> diff -Naurp a/mm/slab.c b/mm/slab.c
> --- a/mm/slab.c	Wed Jul 17 12:25:04 2002
> +++ b/mm/slab.c	Wed Jul 17 12:25:04 2002
> @@ -399,10 +399,10 @@
>  		base = sizeof(slab_t);
>  		extra = sizeof(kmem_bufctl_t);
>  	}
> -	i = 0;
> +       i = (wastage - base)/(size + extra);
>  	while (i*size + L1_CACHE_ALIGN(base+i*extra) <=
> wastage)
>  		i++;
> -	if (i > 0)
> +       while (i*size + L1_CACHE_ALIGN(base+i*extra)
> > wastage)
>  		i--;
>  
>  	if (i > SLAB_LIMIT)
> 


__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com