Re: [Xen-devel] [PATCH 1/4] xen/p2m: Optimize the get_phys_to_machine function a bit.

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Mon, Jul 09, 2012 at 04:41:24PM +0100, David Vrabel wrote:
> On 03/07/12 16:40, Konrad Rzeszutek Wilk wrote:
> > Running perf tells me that the check for p2m_identity
> > is done a lot - more than it should.
> 
> This patch doesn't change the how often the check is done.  Is this just
> bad wording of the description?

Its bad wording plus it really does not change the assmebler code - I thought
it did but since there is only one "if" case it does not matter how
much I put unlikely in front of it - it is the only branch condition.

Perhaps if I make the code be put in a variable (register) that would
optimize this a bit more...


> > 
> > Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> > ---
> >  arch/x86/xen/p2m.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
> > index 64effdc..161bcdc 100644
> > --- a/arch/x86/xen/p2m.c
> > +++ b/arch/x86/xen/p2m.c
> > @@ -406,7 +406,7 @@ unsigned long get_phys_to_machine(unsigned long pfn)
> >  	 * and in p2m_*missing, so returning the INVALID_P2M_ENTRY
> >  	 * would be wrong.
> >  	 */
> > -	if (p2m_top[topidx][mididx] == p2m_identity)
> > +	if (unlikely(p2m_top[topidx][mididx] == p2m_identity))
> >  		return IDENTITY_FRAME(pfn);
> >  
> >  	return p2m_top[topidx][mididx][idx];