Re: Avoiding external fragmentation with a placement policy Version 11

[Mel Gorman <mel@csn.ul.ie>]

On Wed, 25 May 2005, Mike Kravetz wrote:

> On Sun, May 22, 2005 at 09:05:07PM +0100, Mel Gorman wrote:
> >  /*
> > + * Calculate the size of the zone->usemap
> > + */
> > +static unsigned long __init usemap_size(unsigned long zonesize) {
> > +	unsigned long usemapsize;
> > +
> > +	/* - Number of MAX_ORDER blocks in the zone */
> > +	usemapsize = (zonesize + (1 << (MAX_ORDER-1))) >> (MAX_ORDER-1);
> > +
> > +	/* - BITS_PER_ALLOC_TYPE bits to record what type of block it is */
> > +	usemapsize = (usemapsize * BITS_PER_ALLOC_TYPE + (sizeof(unsigned long)*8)) / 8;
> > +
> > +	return L1_CACHE_ALIGN(usemapsize);
> > +}
>
> In the first calculation, I think you are trying to 'round up'.  If this
> is the case, then I believe the calculation should be:
>
> usemapsize = (zonesize + ((1 << (MAX_ORDER-1)) - 1) >> (MAX_ORDER-1);
>

You're right. This excessively large calculation is left-over from a Magic
Bug That Wouldn't Go Away at the time. I thought I might be overflowing
the usemap (even though I couldn't be but I was desperate for explanations
at the time). The second calculation does not need "+ sizeof(unsigned
long)"

This wastes a few bytes but should not cause problems. I'll fix it and
release a version against -rc5 on Monday (The delay is because I don't
have access to a test environment right now)

-- 
Mel Gorman
Part-time Phd Student                          Java Applications Developer
University of Limerick                         IBM Dublin Software Lab
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>