Re: [PATCH V4] irq: Track the interrupt timings

From: Nicolas Pitre
Date: Fri Jun 10 2016 - 11:11:37 EST


On Fri, 10 Jun 2016, Thomas Gleixner wrote:

> Thanks,
>
> On Wed, 13 Apr 2016, Daniel Lezcano wrote:
> > + timings = this_cpu_ptr(desc->timings);
> > + now = local_clock();
> > + prev = timings->timestamp;
> > + timings->timestamp = now;
> > +
> > + /*
> > + * If it is the first interrupt of the series, we can't
> > + * compute an interval, just store the timestamp and exit.
> > + */
> > + if (unlikely(!prev))
> > + return;
>
> The delta will be large enough that you drop out in the check below. So you
> can spare that conditional.

Fair enough.

> > + diff = now - prev;
> > +
> > + /*
> > + * microsec (actually 1024th of a milisec) precision is good
> > + * enough for our purpose.
> > + */
> > + diff >>= 10;
>
> And that shift instruction is required because of the following?
>
> > * Otherwise we know the magnitude of diff is
> > + * well within 32 bits.
>
> AFAICT that's pointless. You are not saving anything because NSEC_PER_SEC is
> smaller than 2^32 and your 8 values are not going to overflow 64 bit in the
> sum.

Those values are squared later, so we really want 32 bits here.

> > + */
> > + if (unlikely(diff > USEC_PER_SEC)) {
> > + memset(timings, 0, sizeof(*timings));
> > + timings->timestamp = now;
>
> Redundant store.

We just trashed all our data with the memset so the current timestamp
needs to be restored.

> > + return;
> > + }
> > +
> > + /* The oldest value corresponds to the next index. */
> > + timings->w_index = (timings->w_index + 1) & IRQ_TIMINGS_MASK;
> > +
> > + /*
> > + * Remove the oldest value from the summing. If this is the
> > + * first time we go through this array slot, the previous
> > + * value will be zero and we won't substract anything from the
> > + * current sum. Hence this code relies on a zero-ed structure.
> > + */
> > + timings->sum -= timings->values[timings->w_index];
> > + timings->values[timings->w_index] = diff;
> > + timings->sum += diff;
>
> Now the real question is whether you really need all that math, checks and
> memsets in the irq hotpath. If you make the storage slightly larger then you
> can just store the values unconditionally in the circular buffer and do all
> the computational stuff when you really it.

Well... given that you need an IRQ everytime you come out of idle that
means there will always be more IRQs than entries into idle, so you're
probably right.


Nicolas