Re: [patch] x86, perf_counter, bts: add bts to perf_counter

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Fri, 2009-08-07 at 09:22 +0200, Markus Metzger wrote:
> Implement a performance counter with:
>     attr.type           = PERF_TYPE_HARDWARE
>     attr.config         = PERF_COUNT_HW_BRANCH_INSTRUCTIONS
>     attr.sample_period  = 1
> 
> using branch trace store (BTS) on x86 hardware, if available.
> 
> The from and to address for each branch can be sampled using:
>     PERF_SAMPLE_IP      for the from address
>     PERF_SAMPLE_ADDR    for the to address


The patch seems to forget one fundamental aspect, it doesn't provide a
counter.

The 'fake' fixed purpose counter we create only provides event samples,
but it does not provide a counter abstraction, so people who'd expect
read() to return the current count of things will get 0.

We could fix this by keeping a software counter along with it that
counts the number of BTS entries processed so far.