Re: [PATCH, RFC, tip/core/rcu] scalable classic RCU implementation

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Wed, Aug 27, 2008 at 01:23:28PM -0700, Josh Triplett wrote:
> On Wed, 2008-08-27 at 11:34 -0700, Paul E. McKenney wrote:
> > On Tue, Aug 26, 2008 at 05:38:36PM -0700, Josh Triplett wrote:
> > > On Tue, 2008-08-26 at 09:05 -0700, Paul E. McKenney wrote:
> > > > On Mon, Aug 25, 2008 at 03:02:30PM -0700, Josh Triplett wrote:
> > > > > On Fri, 2008-08-22 at 18:53 -0700, Paul E. McKenney wrote:
> > > > > > On Fri, Aug 22, 2008 at 04:29:32PM -0700, Josh Triplett wrote:
> > > > > > > On Thu, 2008-08-21 at 16:43 -0700, Paul E. McKenney wrote:
> > > > > > > > @@ -658,14 +806,19 @@ int rcu_needs_cpu(int cpu)
> > > > > > > >  	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
> > > > > > > >  	struct rcu_data *rdp_bh = &per_cpu(rcu_bh_data, cpu);
> > > > > > > > 
> > > > > > > > -	return !!rdp->nxtlist || !!rdp_bh->nxtlist || rcu_pending(cpu);
> > > > > > > > +	return !!*rdp->nxttail[RCU_DONE_TAIL] ||
> > > > > > > > +	       !!*rdp_bh->nxttail[RCU_DONE_TAIL] ||
> > > > > > > > +	       rcu_pending(cpu);
> > > > > > > 
> > > > > > > !! seems unnecessary here.
> > > > > > 
> > > > > > Someone once told me why this was necessary, but I forget.  It was in the
> > > > > > original, and I didn't put it there.  Some weirdness about conversion
> > > > > > to 32-bit integer when the lower 32 bits of the pointer was zero or
> > > > > > some such.  So if your pointer value was 0x100000000, for example,
> > > > > > so that conversion to int gives zero.
> > > > > 
> > > > > Good point!  That doesn't apply if you use ||, though.  If you just did
> > > > > "return somepointer" that could potentially cause the problem you
> > > > > describe.  In any case, it can't *hurt* to have it; GCC should do the
> > > > > sane thing.
> > > > 
> > > > OK.  I will review this towards the end, leaving it there to remind me
> > > > in the meantime.
> > > > 
> > > > So, would I need the !! on the left-hand operand of the first || due
> > > > to short-circuiting?
> > > 
> > > No.  || will always return 1 or 0.  You only need the !! if you want to
> > > directly return the boolean value of a potentially 64-bit pointer.
> > 
> > Even if one argument of || is long and the other int or some fool thing
> > like that?  (What, me paranoid???)
> 
> What, you don't know exactly how C behaves in every strange corner
> case? ;)

I used to, back when identifiers were only guaranteed to be
differentiated by their first 8 characters (6 or 7 if extern).  ;-)

> || always produces a result of type int, and it compares each of its two
> arguments to 0 independently; to the best of my knowledge the size of
> those arguments never matters.

I suppose I should read the spec.

							Thanx, Paul