[apple/xnu.git] / osfmk / kern / sched_average.c

/*
 * Copyright (c) 2000-2007 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 *
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 *
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 *
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 */
/*
 * @OSF_COPYRIGHT@
 */
/*
 * Mach Operating System
 * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University
 * All Rights Reserved.
 *
 * Permission to use, copy, modify and distribute this software and its
 * documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 *
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
 * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 *
 * Carnegie Mellon requests users of this software to return to
 *
 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 *
 * any improvements or extensions that they make and grant Carnegie Mellon
 * the rights to redistribute these changes.
 */
/*
 */
/*
 *      Author: Avadis Tevanian, Jr.
 *      Date:   1986
 *
 *      Compute various averages.
 */

#include <mach/mach_types.h>

#include <kern/sched.h>
#include <kern/assert.h>
#include <kern/processor.h>
#include <kern/thread.h>
#if CONFIG_TELEMETRY
#include <kern/telemetry.h>
#endif
#include <kern/zalloc_internal.h>

#include <sys/kdebug.h>

uint32_t        avenrun[3] = {0, 0, 0};
uint32_t        mach_factor[3] = {0, 0, 0};

uint32_t        sched_load_average, sched_mach_factor;

#if defined(CONFIG_SCHED_TIMESHARE_CORE)
/*
 * Values are scaled by LOAD_SCALE, defined in processor_info.h
 */
#define base(n)         ((n) << SCHED_TICK_SHIFT)
#define frac(n)         (((base(n) - 1) * LOAD_SCALE) / base(n))

static uint32_t         fract[3] = {
        frac(5),                /* 5 second average */
        frac(30),               /* 30 second average */
        frac(60),               /* 1 minute average */
};

#undef base
#undef frac

#endif /* CONFIG_SCHED_TIMESHARE_CORE */

static unsigned int             sched_nrun;

typedef void    (*sched_avg_comp_t)(
        void                    *param);

static struct sched_average {
        sched_avg_comp_t        comp;
        void                    *param;
        int                     period; /* in seconds */
        uint64_t                deadline;
} sched_average[] = {
        { compute_averunnable, &sched_nrun, 5, 0 },
        { compute_stack_target, NULL, 5, 1 },
        { compute_pageout_gc_throttle, NULL, 1, 0 },
        { compute_pmap_gc_throttle, NULL, 60, 0 },
        { compute_zone_working_set_size, NULL, ZONE_WSS_UPDATE_PERIOD, 0 },
#if CONFIG_TELEMETRY
        { compute_telemetry, NULL, 1, 0 },
#endif
        { NULL, NULL, 0, 0 }
};

typedef struct sched_average    *sched_average_t;

/*
 * Scheduler load calculation algorithm
 *
 * The scheduler load values provide an estimate of the number of runnable
 * timeshare threads in the system at various priority bands. The load
 * ultimately affects the priority shifts applied to all threads in a band
 * causing them to timeshare with other threads in the system. The load is
 * maintained in buckets, with each bucket corresponding to a priority band.
 *
 * Each runnable thread on the system contributes its load to its priority
 * band and to the bands above it. The contribution of a thread to the bands
 * above it is not strictly 1:1 and is weighted based on the priority band
 * of the thread. The rules of thread load contribution to each of its higher
 * bands are as follows:
 *
 * - DF threads: Upto (2 * NCPUs) threads
 * - UT threads: Upto NCPUs threads
 * - BG threads: Upto 1 thread
 *
 * To calculate the load values, the various run buckets are sampled (every
 * sched_load_compute_interval_abs) and the weighted contributions of the the
 * lower bucket threads are added. The resultant value is plugged into an
 * exponentially weighted moving average formula:
 *      new-load = alpha * old-load + (1 - alpha) * run-bucket-sample-count
 *      (where, alpha < 1)
 * The calculations for the scheduler load are done using fixpoint math with
 * a scale factor of 16 to avoid expensive divides and floating point
 * operations. The final load values are a smooth curve representative of
 * the actual number of runnable threads in a priority band.
 */

/* Maintains the current (scaled for fixpoint) load in various buckets */
uint32_t sched_load[TH_BUCKET_MAX];

/*
 * Alpha factor for the EWMA alogrithm. The current values are chosen as
 * 6:10 ("old load":"new samples") to make sure the scheduler reacts fast
 * enough to changing system load but does not see too many spikes from bursty
 * activity. The current values ensure that the scheduler would converge
 * to the latest load in 2-3 sched_load_compute_interval_abs intervals
 * (which amounts to ~30-45ms with current values).
 */
#define SCHED_LOAD_EWMA_ALPHA_OLD      6
#define SCHED_LOAD_EWMA_ALPHA_NEW      10
#define SCHED_LOAD_EWMA_ALPHA_SHIFT    4
static_assert((SCHED_LOAD_EWMA_ALPHA_OLD + SCHED_LOAD_EWMA_ALPHA_NEW) == (1ul << SCHED_LOAD_EWMA_ALPHA_SHIFT));

/* For fixpoint EWMA, roundup the load to make it converge */
#define SCHED_LOAD_EWMA_ROUNDUP(load)   (((load) & (1ul << (SCHED_LOAD_EWMA_ALPHA_SHIFT - 1))) != 0)

/* Macro to convert scaled sched load to a real load value */
#define SCHED_LOAD_EWMA_UNSCALE(load)   (((load) >> SCHED_LOAD_EWMA_ALPHA_SHIFT) + SCHED_LOAD_EWMA_ROUNDUP(load))

/*
 * Routine to capture the latest runnable counts and update sched_load (only used for non-clutch schedulers)
 */
void
compute_sched_load(void)
{
        /*
         * Retrieve a snapshot of the current run counts.
         *
         * Why not a bcopy()? Because we need atomic word-sized reads of sched_run_buckets,
         * not byte-by-byte copy.
         */
        uint32_t ncpus = processor_avail_count;
        uint32_t load_now[TH_BUCKET_MAX];

        load_now[TH_BUCKET_RUN]      = os_atomic_load(&sched_run_buckets[TH_BUCKET_RUN], relaxed);
        load_now[TH_BUCKET_FIXPRI]   = os_atomic_load(&sched_run_buckets[TH_BUCKET_FIXPRI], relaxed);
        load_now[TH_BUCKET_SHARE_FG] = os_atomic_load(&sched_run_buckets[TH_BUCKET_SHARE_FG], relaxed);
        load_now[TH_BUCKET_SHARE_DF] = os_atomic_load(&sched_run_buckets[TH_BUCKET_SHARE_DF], relaxed);
        load_now[TH_BUCKET_SHARE_UT] = os_atomic_load(&sched_run_buckets[TH_BUCKET_SHARE_UT], relaxed);
        load_now[TH_BUCKET_SHARE_BG] = os_atomic_load(&sched_run_buckets[TH_BUCKET_SHARE_BG], relaxed);

        assert(load_now[TH_BUCKET_RUN] >= 0);
        assert(load_now[TH_BUCKET_FIXPRI] >= 0);

        uint32_t nthreads = load_now[TH_BUCKET_RUN];
        uint32_t nfixpri  = load_now[TH_BUCKET_FIXPRI];

        KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
            MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_LOAD) | DBG_FUNC_NONE,
            load_now[TH_BUCKET_FIXPRI], (load_now[TH_BUCKET_SHARE_FG] + load_now[TH_BUCKET_SHARE_DF]),
            load_now[TH_BUCKET_SHARE_BG], load_now[TH_BUCKET_SHARE_UT], 0);

        /*
         * Compute the timeshare priority conversion factor based on loading.
         * Because our counters may be incremented and accessed
         * concurrently with respect to each other, we may have
         * windows where the invariant (nthreads - nfixpri) == (fg + df + bg + ut)
         * is broken, so truncate values in these cases.
         */
        uint32_t timeshare_threads = (nthreads - nfixpri);
        for (uint32_t i = TH_BUCKET_SHARE_FG; i <= TH_BUCKET_SHARE_BG; i++) {
                if (load_now[i] > timeshare_threads) {
                        load_now[i] = timeshare_threads;
                }
        }

        /*
         * Default threads contribute up to (NCPUS * 2) of load to FG threads
         */
        if (load_now[TH_BUCKET_SHARE_DF] <= (ncpus * 2)) {
                load_now[TH_BUCKET_SHARE_FG] += load_now[TH_BUCKET_SHARE_DF];
        } else {
                load_now[TH_BUCKET_SHARE_FG] += (ncpus * 2);
        }

        /*
         * Utility threads contribute up to NCPUS of load to FG & DF threads
         */
        if (load_now[TH_BUCKET_SHARE_UT] <= ncpus) {
                load_now[TH_BUCKET_SHARE_FG] += load_now[TH_BUCKET_SHARE_UT];
                load_now[TH_BUCKET_SHARE_DF] += load_now[TH_BUCKET_SHARE_UT];
        } else {
                load_now[TH_BUCKET_SHARE_FG] += ncpus;
                load_now[TH_BUCKET_SHARE_DF] += ncpus;
        }

        /*
         * BG threads contribute up to 1 thread worth of load to FG, DF and UT threads
         */
        if (load_now[TH_BUCKET_SHARE_BG] > 0) {
                load_now[TH_BUCKET_SHARE_FG] += 1;
                load_now[TH_BUCKET_SHARE_DF] += 1;
                load_now[TH_BUCKET_SHARE_UT] += 1;
        }

        /*
         * The conversion factor consists of two components:
         * a fixed value based on the absolute time unit (sched_fixed_shift),
         * and a dynamic portion based on load (sched_load_shifts).
         *
         * Zero load results in a out of range shift count.
         */

        for (uint32_t i = TH_BUCKET_SHARE_FG; i <= TH_BUCKET_SHARE_BG; i++) {
                uint32_t bucket_load = 0;

                if (load_now[i] > ncpus) {
                        /* Normalize the load to number of CPUs */
                        if (ncpus > 1) {
                                bucket_load = load_now[i] / ncpus;
                        } else {
                                bucket_load = load_now[i];
                        }

                        if (bucket_load > MAX_LOAD) {
                                bucket_load = MAX_LOAD;
                        }
                }
                /* Plug the load values into the EWMA algorithm to calculate (scaled for fixpoint) sched_load */
                sched_load[i] = (sched_load[i] * SCHED_LOAD_EWMA_ALPHA_OLD) + ((bucket_load << SCHED_LOAD_EWMA_ALPHA_SHIFT) * SCHED_LOAD_EWMA_ALPHA_NEW);
                sched_load[i] = sched_load[i] >> SCHED_LOAD_EWMA_ALPHA_SHIFT;
        }

        KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
            MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_LOAD_EFFECTIVE) | DBG_FUNC_NONE,
            SCHED_LOAD_EWMA_UNSCALE(sched_load[TH_BUCKET_SHARE_FG]), SCHED_LOAD_EWMA_UNSCALE(sched_load[TH_BUCKET_SHARE_DF]),
            SCHED_LOAD_EWMA_UNSCALE(sched_load[TH_BUCKET_SHARE_UT]), SCHED_LOAD_EWMA_UNSCALE(sched_load[TH_BUCKET_SHARE_BG]), 0);
}

void
compute_averages(uint64_t stdelta)
{
        uint32_t nthreads = os_atomic_load(&sched_run_buckets[TH_BUCKET_RUN], relaxed) - 1;
        uint32_t ncpus = processor_avail_count;

        /* Update the global pri_shifts based on the latest values */
        for (uint32_t i = TH_BUCKET_SHARE_FG; i <= TH_BUCKET_SHARE_BG; i++) {
                uint32_t bucket_load = SCHED_LOAD_EWMA_UNSCALE(sched_load[i]);
                uint32_t shift = sched_fixed_shift - sched_load_shifts[bucket_load];

                if (shift > SCHED_PRI_SHIFT_MAX) {
                        sched_pri_shifts[i] = INT8_MAX;
                } else {
                        sched_pri_shifts[i] = shift;
                }
        }

        /*
         * Sample total running threads for the load average calculation.
         */
        sched_nrun = nthreads;

        /*
         * Load average and mach factor calculations for
         * those which ask about these things.
         */
        uint32_t average_now = nthreads * LOAD_SCALE;
        uint32_t factor_now;

        if (nthreads > ncpus) {
                factor_now = (ncpus * LOAD_SCALE) / (nthreads + 1);
        } else {
                factor_now = (ncpus - nthreads) * LOAD_SCALE;
        }

        /*
         * For those statistics that formerly relied on being recomputed
         * on timer ticks, advance by the approximate number of corresponding
         * elapsed intervals, thus compensating for potential idle intervals.
         */
        for (uint32_t index = 0; index < stdelta; index++) {
                sched_mach_factor = ((sched_mach_factor << 2) + factor_now) / 5;
                sched_load_average = ((sched_load_average << 2) + average_now) / 5;
        }

        /*
         * Compute old-style Mach load averages.
         */
        for (uint32_t index = 0; index < stdelta; index++) {
                for (uint32_t i = 0; i < 3; i++) {
                        mach_factor[i] = ((mach_factor[i] * fract[i]) +
                            (factor_now * (LOAD_SCALE - fract[i]))) / LOAD_SCALE;

                        avenrun[i] = ((avenrun[i] * fract[i]) +
                            (average_now * (LOAD_SCALE - fract[i]))) / LOAD_SCALE;
                }
        }

        /*
         * Compute averages in other components.
         */
        uint64_t abstime = mach_absolute_time();

        for (sched_average_t avg = sched_average; avg->comp != NULL; ++avg) {
                if (abstime >= avg->deadline) {
                        uint64_t period_abs = (avg->period * sched_one_second_interval);
                        uint64_t ninvokes = 1;

                        ninvokes += (abstime - avg->deadline) / period_abs;
                        ninvokes = MIN(ninvokes, SCHED_TICK_MAX_DELTA);

                        for (uint32_t index = 0; index < ninvokes; index++) {
                                (*avg->comp)(avg->param);
                        }
                        avg->deadline = abstime + period_abs;
                }
        }
}