Re: [PATCH v3 1/n] perf/core: addressing 4x slowdown during per-process profiling of STREAM benchmark on Intel Xeon Phi

From: Mark Rutland <mark.rutland@arm.com>
To: Andi Kleen <ak@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Kan Liang <kan.liang@intel.com>,
	Dmitri Prokhorov <Dmitry.Prohorov@intel.com>,
	Valery Cherepennikov <valery.cherepennikov@intel.com>,
	David Carrillo-Cisneros <davidcc@google.com>,
	Stephane Eranian <eranian@google.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3 1/n] perf/core: addressing 4x slowdown during per-process profiling of STREAM benchmark on Intel Xeon Phi
Date: Mon, 19 Jun 2017 16:09:39 +0100	[thread overview]
Message-ID: <20170619150939.GA4555@leverpostej> (raw)
In-Reply-To: <20170619145908.GA23705@tassilo.jf.intel.com>

On Mon, Jun 19, 2017 at 07:59:08AM -0700, Andi Kleen wrote:
> > > For comparison, can you give --per-thread a go prior to these patches
> > > being applied?
> > 
> > FWIW, I had a go with (an old) perf record on an arm64 system using
> > --per-thread, and I see that no samples are recorded, which seems like a
> > bug.
> > 
> > With --per-thread, the slwodown was ~20%, whereas with the defaults it
> > was > 400%.
> 
> I'm not sure what the point of the experiment is? It has to work
> with reasonable overead even without --per-thread.
> 
> FWIW Alexey already root caused the problem, so there's no need
> to restart the debugging.

Sure; we understand where that overhead is coming from, we have an idea
as to how to mitigate that, and we should try to make that work it we
can.

I was trying to get a feel for how that compares to what we can do
today. For other reasons (e.g. fd exhaustion), opening NR_CPUS * n
events might not be a great idea on systems with a huge number of CPUs.
We might want a heuristic in the perf tool regardless.

Thanks,
Mark.