Re: [PATCH 2/9] perf/x86/intel: Basic support for metrics counters

From: Peter Zijlstra
Date: Wed May 29 2019 - 03:32:26 EST


On Tue, May 28, 2019 at 02:21:49PM -0400, Liang, Kan wrote:
>
>
> On 5/28/2019 8:15 AM, Peter Zijlstra wrote:
> > On Tue, May 21, 2019 at 02:40:48PM -0700, kan.liang@xxxxxxxxxxxxxxx wrote:
> > > +/*
> > > + * We model PERF_METRICS as more magic fixed-mode PMCs, one for each metric
> > > + * and another for the whole slots counter
> > > + *
> > > + * Internally they all map to Fixed Ctr 3 (SLOTS), and allocate PERF_METRICS
> > > + * as an extra_reg. PERF_METRICS has no own configuration, but we fill in
> > > + * the configuration of FxCtr3 to enforce that all the shared users of SLOTS
> > > + * have the same configuration.
> > > + */
> > > +#define INTEL_PMC_IDX_FIXED_METRIC_BASE (INTEL_PMC_IDX_FIXED + 17)
> > > +#define INTEL_PMC_IDX_TD_RETIRING (INTEL_PMC_IDX_FIXED_METRIC_BASE + 0)
> > > +#define INTEL_PMC_IDX_TD_BAD_SPEC (INTEL_PMC_IDX_FIXED_METRIC_BASE + 1)
> > > +#define INTEL_PMC_IDX_TD_FE_BOUND (INTEL_PMC_IDX_FIXED_METRIC_BASE + 2)
> > > +#define INTEL_PMC_IDX_TD_BE_BOUND (INTEL_PMC_IDX_FIXED_METRIC_BASE + 3)
> > > +#define INTEL_PMC_MSK_ANY_SLOTS ((0xfull << INTEL_PMC_IDX_FIXED_METRIC_BASE) | \
> > > + INTEL_PMC_MSK_FIXED_SLOTS)
> > > +static inline bool is_metric_idx(int idx)
> > > +{
> > > + return idx >= INTEL_PMC_IDX_FIXED_METRIC_BASE && idx <= INTEL_PMC_IDX_TD_BE_BOUND;
> > > +}
> >
> > Something like:
> >
> > return (idx >> INTEL_PMC_IDX_FIXED_METRIC_BASE) & 0xf;
> >
> > might be faster code... (if it wasn't for 64bit literals being a pain,
> > it could be a simple test instruction).
> >
>
> is_metric_idx() is not a mask. It's to check if the idx between 49 and 52.

blergh, indeed. Check that it compiles into a single branch though, if
it gets that wrong it needs hand holding.

One way would be to make these 48-51 and put BTS as 52, and then you do
have a mask.

Another way would be to write it as:

(unsigned)(idx - INTEL_PMC_IDX_FIXED_METRIC_BASE) < 4;