From: "Rafael J. Wysocki" <rafael@kernel.org>
To: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@redhat.com>,
Peter Zijlstra <peterz@infradead.org>,
Borislav Petkov <bp@suse.de>, Len Brown <lenb@kernel.org>,
"Rafael J. Wysocki" <rjw@rjwysocki.net>,
Mel Gorman <mgorman@techsingularity.net>,
"the arch/x86 maintainers" <x86@kernel.org>,
Linux PM <linux-pm@vger.kernel.org>,
Viresh Kumar <viresh.kumar@linaro.org>,
Juri Lelli <juri.lelli@arm.com>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [RFC/RFT] [PATCH 05/10] cpufreq: intel_pstate: HWP boost performance on IO Wake
Date: Wed, 16 May 2018 11:45:32 +0200 [thread overview]
Message-ID: <CAJZ5v0jSfZyGoKJKHEQCZRgQtiNt+_D0_sY+K2p3nQ9xdg3qKg@mail.gmail.com> (raw)
In-Reply-To: <20180516044911.28797-6-srinivas.pandruvada@linux.intel.com>
On Wed, May 16, 2018 at 6:49 AM, Srinivas Pandruvada
<srinivas.pandruvada@linux.intel.com> wrote:
> When a task is woken up from IO wait, boost HWP prformance to max. This
> helps IO workloads on servers with per core P-states. But changing limits
> has extra over head of issuing new HWP Request MSR, which takes 1000+
> cycles. So this change limits setting HWP Request MSR. Also request can
> be for a remote CPU.
> Rate control in setting HWP Requests:
> - If the current performance is around P1, simply ignore IO flag.
> - Once set wait till hold time, till remove boost. While the boost
> is on, another IO flags is notified, it will prolong boost.
> - If the IO flags are notified multiple ticks apart, this may not be
> IO bound task. Othewise idle system gets periodic boosts for one
> IO wake.
>
> Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
> ---
> drivers/cpufreq/intel_pstate.c | 75 ++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 75 insertions(+)
>
> diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
> index e200887..d418265 100644
> --- a/drivers/cpufreq/intel_pstate.c
> +++ b/drivers/cpufreq/intel_pstate.c
> @@ -20,6 +20,7 @@
> #include <linux/tick.h>
> #include <linux/slab.h>
> #include <linux/sched/cpufreq.h>
> +#include <linux/sched/topology.h>
> #include <linux/list.h>
> #include <linux/cpu.h>
> #include <linux/cpufreq.h>
> @@ -224,6 +225,8 @@ struct global_params {
> * @hwp_req_cached: Cached value of the last HWP request MSR
> * @csd: A structure used to issue SMP async call, which
> * defines callback and arguments
> + * @hwp_boost_active: HWP performance is boosted on this CPU
> + * @last_io_update: Last time when IO wake flag was set
> *
> * This structure stores per CPU instance data for all CPUs.
> */
> @@ -258,6 +261,8 @@ struct cpudata {
> s16 epp_saved;
> u64 hwp_req_cached;
> call_single_data_t csd;
> + bool hwp_boost_active;
> + u64 last_io_update;
> };
>
> static struct cpudata **all_cpu_data;
> @@ -1421,10 +1426,80 @@ static void csd_init(struct cpudata *cpu)
> cpu->csd.info = cpu;
> }
>
> +/*
> + * Long hold time will keep high perf limits for long time,
> + * which negatively impacts perf/watt for some workloads,
> + * like specpower. 3ms is based on experiements on some
> + * workoads.
> + */
> +static int hwp_boost_hold_time_ms = 3;
> +
> +/* Default: This will roughly around P1 on SKX */
> +#define BOOST_PSTATE_THRESHOLD (SCHED_CAPACITY_SCALE / 2)
> +static int hwp_boost_pstate_threshold = BOOST_PSTATE_THRESHOLD;
> +
> +static inline bool intel_pstate_check_boost_threhold(struct cpudata *cpu)
> +{
> + /*
> + * If the last performance is above threshold, then return false,
> + * so that caller can ignore boosting.
> + */
> + if (arch_scale_freq_capacity(cpu->cpu) > hwp_boost_pstate_threshold)
> + return false;
> +
> + return true;
> +}
> +
> static inline void intel_pstate_update_util_hwp(struct update_util_data *data,
> u64 time, unsigned int flags)
> {
> + struct cpudata *cpu = container_of(data, struct cpudata, update_util);
> +
> + if (flags & SCHED_CPUFREQ_IOWAIT) {
> + /*
> + * Set iowait_boost flag and update time. Since IO WAIT flag
> + * is set all the time, we can't just conclude that there is
> + * some IO bound activity is scheduled on this CPU with just
> + * one occurrence. If we receive at least two in two
> + * consecutive ticks, then we start treating as IO. So
> + * there will be one tick latency.
> + */
> + if (time_before64(time, cpu->last_io_update + 2 * TICK_NSEC) &&
> + intel_pstate_check_boost_threhold(cpu))
> + cpu->iowait_boost = true;
> +
> + cpu->last_io_update = time;
> + cpu->last_update = time;
This is a shared data structure and it gets updated without
synchronization, unless I'm missing something.
How much does the cross-CPU case matter?
> + }
>
> + /*
> + * If the boost is active, we will remove it after timeout on local
> + * CPU only.
> + */
> + if (cpu->hwp_boost_active) {
> + if (smp_processor_id() == cpu->cpu) {
> + bool expired;
> +
> + expired = time_after64(time, cpu->last_update +
> + (hwp_boost_hold_time_ms * NSEC_PER_MSEC));
> + if (expired) {
> + intel_pstate_hwp_boost_down(cpu);
> + cpu->hwp_boost_active = false;
> + cpu->iowait_boost = false;
> + }
> + }
> + return;
> + }
> +
> + cpu->last_update = time;
> +
> + if (cpu->iowait_boost) {
> + cpu->hwp_boost_active = true;
> + if (smp_processor_id() == cpu->cpu)
> + intel_pstate_hwp_boost_up(cpu);
> + else
> + smp_call_function_single_async(cpu->cpu, &cpu->csd);
> + }
> }
>
> static inline void intel_pstate_calc_avg_perf(struct cpudata *cpu)
> --
> 2.9.5
>
next prev parent reply other threads:[~2018-05-16 10:02 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-05-16 4:49 [RFC/RFT] [PATCH 00/10] Intel_pstate: HWP Dynamic performance boost Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 01/10] x86,sched: Add support for frequency invariance Srinivas Pandruvada
2018-05-16 9:56 ` Peter Zijlstra
2018-05-16 4:49 ` [RFC/RFT] [PATCH 02/10] cpufreq: intel_pstate: Conditional frequency invariant accounting Srinivas Pandruvada
2018-05-16 7:16 ` Peter Zijlstra
2018-05-16 7:29 ` Peter Zijlstra
2018-05-16 9:07 ` Rafael J. Wysocki
2018-05-16 17:32 ` Srinivas Pandruvada
2018-05-16 15:19 ` Juri Lelli
2018-05-16 15:47 ` Peter Zijlstra
2018-05-16 16:31 ` Juri Lelli
2018-05-17 10:59 ` Juri Lelli
2018-05-17 15:04 ` Juri Lelli
2018-05-17 15:41 ` Srinivas Pandruvada
2018-05-17 16:16 ` Peter Zijlstra
2018-05-17 16:42 ` Srinivas Pandruvada
2018-05-17 16:56 ` Rafael J. Wysocki
2018-05-17 18:28 ` Peter Zijlstra
2018-05-18 7:36 ` Rafael J. Wysocki
2018-05-18 10:57 ` Patrick Bellasi
2018-05-18 11:29 ` Peter Zijlstra
2018-05-18 13:33 ` Patrick Bellasi
2018-05-30 16:57 ` Patrick Bellasi
2018-05-18 14:09 ` Valentin Schneider
2018-05-16 15:58 ` Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 03/10] cpufreq: intel_pstate: Utility functions to boost HWP performance limits Srinivas Pandruvada
2018-05-16 7:22 ` Peter Zijlstra
2018-05-16 9:15 ` Rafael J. Wysocki
2018-05-16 10:43 ` Peter Zijlstra
2018-05-16 15:39 ` Srinivas Pandruvada
2018-05-16 15:41 ` Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 04/10] cpufreq: intel_pstate: Add update_util_hook for HWP Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 05/10] cpufreq: intel_pstate: HWP boost performance on IO Wake Srinivas Pandruvada
2018-05-16 7:37 ` Peter Zijlstra
2018-05-16 17:55 ` Srinivas Pandruvada
2018-05-17 8:19 ` Peter Zijlstra
2018-05-16 9:45 ` Rafael J. Wysocki [this message]
2018-05-16 19:28 ` Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 06/10] cpufreq / sched: Add interface to get utilization values Srinivas Pandruvada
2018-05-16 6:40 ` Viresh Kumar
2018-05-16 22:25 ` Srinivas Pandruvada
2018-05-16 8:11 ` Peter Zijlstra
2018-05-16 22:40 ` Srinivas Pandruvada
2018-05-17 7:50 ` Peter Zijlstra
2018-05-16 4:49 ` [RFC/RFT] [PATCH 07/10] cpufreq: intel_pstate: HWP boost performance on busy task migrate Srinivas Pandruvada
2018-05-16 9:49 ` Rafael J. Wysocki
2018-05-16 20:59 ` Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 08/10] cpufreq: intel_pstate: Dyanmically update busy pct Srinivas Pandruvada
2018-05-16 7:43 ` Peter Zijlstra
2018-05-16 7:47 ` Peter Zijlstra
2018-05-16 4:49 ` [RFC/RFT] [PATCH 09/10] cpufreq: intel_pstate: New sysfs entry to control HWP boost Srinivas Pandruvada
2018-05-16 4:49 ` [RFC/RFT] [PATCH 10/10] cpufreq: intel_pstate: enable boost for SKX Srinivas Pandruvada
2018-05-16 7:49 ` Peter Zijlstra
2018-05-16 15:46 ` Srinivas Pandruvada
2018-05-16 15:54 ` Peter Zijlstra
2018-05-17 0:52 ` Srinivas Pandruvada
2018-05-16 6:49 ` [RFC/RFT] [PATCH 00/10] Intel_pstate: HWP Dynamic performance boost Juri Lelli
2018-05-16 15:43 ` Srinivas Pandruvada
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAJZ5v0jSfZyGoKJKHEQCZRgQtiNt+_D0_sY+K2p3nQ9xdg3qKg@mail.gmail.com \
--to=rafael@kernel.org \
--cc=bp@suse.de \
--cc=juri.lelli@arm.com \
--cc=lenb@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pm@vger.kernel.org \
--cc=mgorman@techsingularity.net \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=rjw@rjwysocki.net \
--cc=srinivas.pandruvada@linux.intel.com \
--cc=tglx@linutronix.de \
--cc=viresh.kumar@linaro.org \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).