/** \file   archdep_tick.c
 * \brief   Relating to the management of time.
 *
 * \author  David Hogan <david.q.hogan@gmail.com>
 */

/*
 * This file is part of VICE, the Versatile Commodore Emulator.
 * See README for copyright notice.
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
 *  02111-1307  USA.
 *
 */

#include "vice.h"

#include "archdep_defs.h"

#ifdef WINDOWS_COMPILE
#   include <windows.h>
#elif defined(HAVE_NANOSLEEP)
#   include <time.h>
#else
#   include <unistd.h>
#   include <errno.h>
#   include <sys/time.h>
#endif
#ifdef MACOS_COMPILE
#   include <mach/mach.h>
#   include <mach/mach_time.h>
#endif
#include <stdio.h>

#include "archdep_tick.h"


#ifndef MIN
#define MIN(a, b)  (((a) < (b)) ? (a) : (b))
#endif

/* #define OVERSLEEP_COMPENSATION */

/* ------------------------------------------------------------------------- */

#ifdef WINDOWS_COMPILE
static LARGE_INTEGER timer_frequency;
static HANDLE wait_timer;
#elif defined(MACOS_COMPILE)
static mach_timebase_info_data_t timebase_info;
#endif

void tick_init(void)
{
#ifdef WINDOWS_COMPILE
    QueryPerformanceFrequency(&timer_frequency);

    wait_timer = CreateWaitableTimer(NULL, TRUE, NULL);

#elif defined(MACOS_COMPILE)
    mach_timebase_info(&timebase_info);

#endif
}

tick_t tick_per_second(void)
{
    return TICK_PER_SECOND;
}

#ifdef WINDOWS_COMPILE
tick_t tick_now(void)
{
    LARGE_INTEGER time_now;

    QueryPerformanceCounter(&time_now);

    return (tick_t)(time_now.QuadPart / ((double)timer_frequency.QuadPart / TICK_PER_SECOND));
}

#elif defined(MACOS_COMPILE)
tick_t tick_now(void)
{
    return NANO_TO_TICK(mach_absolute_time() * timebase_info.numer / timebase_info.denom);
}

#else
tick_t tick_now(void)
{
    struct timespec now;

#if defined(LINUX_COMPILE)
    clock_gettime(CLOCK_MONOTONIC_RAW, &now);
#elif defined(FREEBSD_COMPILE)
    clock_gettime(CLOCK_MONOTONIC_PRECISE, &now);
#else
    clock_gettime(CLOCK_MONOTONIC, &now);
#endif

    return NANO_TO_TICK(((uint64_t)NANO_PER_SECOND * now.tv_sec) + now.tv_nsec);
}
#endif

#ifdef WINDOWS_COMPILE
static inline void sleep_impl(tick_t sleep_ticks)
{
    LARGE_INTEGER timeout;

    timeout.QuadPart = 0LL - ((NANO_PER_SECOND / 100) * ((double)sleep_ticks / TICK_PER_SECOND));

    SetWaitableTimer(wait_timer, &timeout, 0, NULL, NULL, 0);
    WaitForSingleObject(wait_timer, INFINITE);
}

#elif defined(HAVE_NANOSLEEP)
static inline void sleep_impl(tick_t sleep_ticks)
{
    struct timespec ts;
    uint64_t nanos = TICK_TO_NANO(sleep_ticks);

    if (nanos < NANO_PER_SECOND) {
        ts.tv_sec = 0;
        ts.tv_nsec = nanos;
    } else {
        ts.tv_sec = nanos / NANO_PER_SECOND;
        ts.tv_nsec = nanos % NANO_PER_SECOND;
    }

    nanosleep(&ts, NULL);
}

#else
static inline void sleep_impl(tick_t sleep_ticks)
{
    if (usleep(TICK_TO_MICRO(sleep_ticks)) == -EINVAL) {
        usleep(MICRO_PER_SECOND - 1);
    }
}
#endif

/* Sleep a number of timer units. */
void tick_sleep(tick_t sleep_ticks)
{
#ifdef OVERSLEEP_COMPENSATION
#   define OVERSLEEP_PUSH_FACTOR (0.99)

    tick_t before_yield_tick = tick_now(); /* do this asap. */
    tick_t adjusted_sleep_ticks;

    static double rolling_oversleep_ticks;
    tick_t slept_ticks;
    tick_t oversleep_ticks;

    /*
     * Oversleep compensation. Systems will typically sleep for longer than
     * requested. This code measures a rolling average amount of oversleep
     * and uses it to request a shorter sleep, followed by a hotloop up to
     * the originally requested time. This yields more accurate timing at
     * a small cost to CPU usage.
     *
     * Disabled by default for now.
     */

    /* Push the oversleep compensation towards zero */
    rolling_oversleep_ticks *= OVERSLEEP_PUSH_FACTOR;

    /* Sleep, if we're not expecting to oversleep by more than the requested amount */
    if (sleep_ticks > rolling_oversleep_ticks) {
        adjusted_sleep_ticks = sleep_ticks - rolling_oversleep_ticks;

        sleep_impl(adjusted_sleep_ticks);

        slept_ticks = tick_now_delta(before_yield_tick);
        if (slept_ticks > adjusted_sleep_ticks) {
            /* Then we overslept */
            oversleep_ticks = slept_ticks - adjusted_sleep_ticks;
            rolling_oversleep_ticks += (double)oversleep_ticks * (1.0 - OVERSLEEP_PUSH_FACTOR);
        }
    }

    while (tick_now_delta(before_yield_tick) < sleep_ticks) {
        /* hot loop to catch up */
    }

#else /*#ifdef OVERSLEEP_COMPENSATION */

    sleep_impl(sleep_ticks);

#endif
}

tick_t tick_now_after(tick_t previous_tick)
{
    /*
     * Fark, high performance counters, called from different threads / cpus, can be off by 1 tick.
     *
     *    "When you compare performance counter results that are acquired from different
     *     threads, consider values that differ by +/- 1 tick to have an ambiguous ordering.
     *     If the time stamps are taken from the same thread, this +/- 1 tick uncertainty
     *     doesn't apply. In this context, the term tick refers to a period of time equal
     *     to 1 divided by (the frequency of the performance counter obtained from
     *     QueryPerformanceFrequency)."
     *
     * https://docs.microsoft.com/en-us/windows/win32/sysinfo/acquiring-high-resolution-time-stamps#guidance-for-acquiring-time-stamps
     */

    tick_t after = tick_now();

    if (after == previous_tick - 1) {
        after = previous_tick;
    }

    return after;
}

tick_t tick_now_delta(tick_t previous_tick)
{
    return tick_now_after(previous_tick) - previous_tick;
}