/*
 * The PIC32 twin's bare-metal fixture: no library, no framework, registers only, so that what
 * it states is about the machine and nothing between. Built with clang for mipsel (MIPS32r2)
 * by build.sh; runs on the PIC32MX3xx/4xx and 5xx/6xx/7xx parts (their ports, Timer2, UART1 and
 * interrupt controller registers sit at the same addresses; DS61143H Table 4-x, DS60001156J).
 *
 *   UART1 at 115200: the lines below;
 *   RG6: toggled every 20000 core-timer counts by polling Count -- 1 kHz from Count's
 *        SYSCLK/2 = 40 MHz (PIC32 FRM 2.12.3);
 *   RF0: toggled by the core timer's interrupt every 1 ms, priority 7 on the shadow register
 *        set -- 500 Hz;
 *   RD1: toggled by Timer2's interrupt, priority 3 on the normal set, PBCLK/8/40000 = 250 Hz
 *        periods -- 125 Hz, until the second phase.
 *
 * After DONE, a second phase checks four rules no measurement above depends on: the core timer's
 * handler on the shadow set read the interrupted code's stack pointer with RDPGPR (the
 * previous set's, not its own); a request is taken only above Status.IPL, not at it (IPL 3
 * holds Timer2's priority-3 request, IPL 2 lets it in); OSCCON.PBDIV /2, written after the
 * SYSKEY unlock, halves the clock Timer2 counts; a 16-bit timer started above its period
 * register counts on to 0xFFFF and wraps to 0 (FRM 14.3.3).
 *
 * SPDX-License-Identifier: Apache-2.0
 */
#include <stdint.h>

#define SFR(a) (*(volatile uint32_t *)(a))
#define CLR 1
#define SET 2
#define INV 3
#define REG(a, op) SFR((a) + 4 * (op))

#define U1MODE  0xBF806000u
#define U1STA   0xBF806010u
#define U1TXREG 0xBF806020u
#define U1BRG   0xBF806040u
#define TRISD   0xBF8860C0u
#define PORTD   0xBF8860D0u
#define LATD    0xBF8860E0u
#define TRISF   0xBF886140u
#define LATF    0xBF886160u
#define TRISG   0xBF886180u
#define LATG    0xBF8861A0u
#define CNCON   0xBF8861C0u
#define CNPUE   0xBF8861E0u
#define T2CON   0xBF800800u
#define TMR2    0xBF800810u
#define PR2     0xBF800820u
#define INTCON  0xBF881000u
#define INTSTAT 0xBF881010u
#define IFS0    0xBF881030u
#define IEC0    0xBF881060u
#define IPC0    0xBF881090u
#define IPC2    0xBF8810B0u
#define DEVID   0xBF80F220u
#define SYSKEY  0xBF80F230u
#define OSCCON  0xBF80F000u

static inline uint32_t mfc0_count(void)
{
    uint32_t v;
    __asm__ volatile("mfc0 %0, $9" : "=r"(v));
    return v;
}

static inline uint32_t mfc0_srsctl(void)
{
    uint32_t v;
    __asm__ volatile("mfc0 %0, $12, 2" : "=r"(v));
    return v;
}

static inline void mtc0_compare(uint32_t v)
{
    __asm__ volatile("mtc0 %0, $11" : : "r"(v));
}

static inline uint32_t mfc0_status(void)
{
    uint32_t v;
    __asm__ volatile("mfc0 %0, $12" : "=r"(v));
    return v;
}

static inline void mtc0_status(uint32_t v)
{
    __asm__ volatile("mtc0 %0, $12\n\tehb" : : "r"(v));
}

/* Busy-wait n counts of Count (40 MHz). */
static void wait_counts(uint32_t n)
{
    uint32_t t = mfc0_count();

    while (mfc0_count() - t < n) {
    }
}

static void putc_(char c)
{
    while (SFR(U1STA) & (1u << 9)) {       /* UTXBF */
    }
    SFR(U1TXREG) = (uint8_t)c;
}

static void puts_(const char *s)
{
    while (*s) {
        putc_(*s++);
    }
}

static void put_hex(uint32_t v)
{
    puts_("0x");
    for (int i = 28; i >= 0; i -= 4) {
        putc_("0123456789abcdef"[(v >> i) & 0xf]);
    }
}

static void put_dec(uint32_t v)
{
    char b[11];
    int n = 0;

    do {
        b[n++] = '0' + v % 10;
        v /= 10;
    } while (v);
    while (n) {
        putc_(b[--n]);
    }
}

static volatile uint32_t ct_ticks, ct_srs, t2_ticks, t2_srs, ct_next;
/* The stack pointer the core timer's handler took from the previous set (start.S). */
volatile uint32_t ct_seen_sp;

void ct_handler(void)
{
    if (!ct_ticks) {
        ct_srs = mfc0_srsctl();
    }
    ct_ticks++;
    ct_next += 40000;                       /* 1 ms at 40 MHz */
    mtc0_compare(ct_next);
    REG(IFS0, CLR) = 1u << 0;               /* CTIF */
    REG(LATF, INV) = 1u << 0;               /* RF0 */
}

void t2_handler(void)
{
    if (!t2_ticks) {
        t2_srs = mfc0_srsctl();
    }
    t2_ticks++;
    REG(IFS0, CLR) = 1u << 8;               /* T2IF */
    REG(LATD, INV) = 1u << 1;               /* RD1 */
}

void exc_handler(void)
{
    uint32_t cause, epc;

    __asm__ volatile("mfc0 %0, $13" : "=r"(cause));
    __asm__ volatile("mfc0 %0, $14" : "=r"(epc));
    puts_("EXC cause=");
    put_hex(cause);
    puts_(" epc=");
    put_hex(epc);
    puts_("\n");
    for (;;) {
    }
}

/* Timer2 at 1:1 from `start` for 40 counts of Count; what TMR2 reads then. */
static uint32_t t2_count_from(uint32_t start)
{
    uint32_t tmr;

    SFR(T2CON) = 0;
    SFR(PR2) = 100;
    SFR(TMR2) = start;
    SFR(T2CON) = 1u << 15;
    wait_counts(40);
    tmr = SFR(TMR2);
    SFR(T2CON) = 0;
    return tmr;
}

/* The second phase (see the top of the file). */
static void phase2(void)
{
    uint32_t sp, st, before, held, ran, tmr;

    __asm__ volatile("move %0, $sp" : "=r"(sp));
    puts_("SRS rdpgpr sp ok=");
    put_dec(ct_seen_sp <= sp && sp - ct_seen_sp < 1024);
    puts_("\n");

    /* Status.IPL is bits 15:10 with an external interrupt controller (MIPS32 PRA, EIC mode). */
    st = mfc0_status();
    before = t2_ticks;
    mtc0_status((st & ~(0x3fu << 10)) | (3u << 10));
    wait_counts(400000);                    /* 10 ms: two and a half Timer2 periods */
    held = t2_ticks == before;
    before = t2_ticks;
    mtc0_status((st & ~(0x3fu << 10)) | (2u << 10));
    wait_counts(400000);
    ran = t2_ticks >= before + 2;
    mtc0_status(st);
    puts_("IPL3 t2 held=");
    put_dec(held);
    puts_(" IPL2 t2 ran=");
    put_dec(ran);
    puts_("\n");

    /* PBCLK /2: Timer2 at 1:1 counts 2000 ticks in 2000 counts of Count (40 MHz), not 4000. */
    REG(IEC0, CLR) = 1u << 8;
    SFR(T2CON) = 0;
    SFR(TMR2) = 0;
    SFR(PR2) = 0xffff;
    SFR(SYSKEY) = 0;
    SFR(SYSKEY) = 0xAA996655u;
    SFR(SYSKEY) = 0x556699AAu;
    REG(OSCCON, CLR) = 3u << 19;
    REG(OSCCON, SET) = 1u << 19;            /* PBDIV 01: /2 */
    SFR(SYSKEY) = 0;
    SFR(T2CON) = 1u << 15;                  /* ON, 1:1 */
    wait_counts(2000);
    tmr = SFR(TMR2);
    SFR(T2CON) = 0;
    puts_("PBDIV2 tmr2/100=");
    put_dec((tmr + 50) / 100);
    puts_("\n");

    /* Started at 0xFFF0, above PR2 = 100, it takes sixteen ticks to the wrap and goes on from 0:
     * it reads sixteen less than the same wait started from 0 (give or take the tick the two
     * starts may fall either side of). */
    before = t2_count_from(0);
    tmr = t2_count_from(0xfff0);
    puts_("T2 above PR wraps ok=");
    put_dec(before - tmr >= 15 && before - tmr <= 17);
    puts_("\nPHASE2 DONE\n");
}

int main(void)
{
    uint32_t next, nopull, pullup;

    /* 115200 at PBCLK 80 MHz: BRG = 80e6 / (16 * 115200) - 1 = 42 (FRM 21.3). */
    SFR(U1BRG) = 42;
    SFR(U1STA) = (1u << 10) | (1u << 12);  /* UTXEN, URXEN */
    SFR(U1MODE) = 1u << 15;                /* ON, 8N1 */
    puts_("PIC32 bare-metal: start\n");
    puts_("DEVID=");
    put_hex(SFR(DEVID));
    puts_(" OSCCON=");
    put_hex(SFR(OSCCON));
    puts_("\n");

    /* An input: RD5 (CN14) with nothing on it, then with its weak pull-up (FRM 12.2.6). */
    nopull = (SFR(PORTD) >> 5) & 1;
    REG(CNPUE, SET) = 1u << 14;
    pullup = (SFR(PORTD) >> 5) & 1;
    REG(CNPUE, CLR) = 1u << 14;
    puts_("IN RD5 nopull=");
    put_dec(nopull);
    puts_(" pullup=");
    put_dec(pullup);
    puts_("\n");

    REG(TRISG, CLR) = 1u << 6;
    REG(TRISF, CLR) = 1u << 0;
    REG(TRISD, CLR) = 1u << 1;

    /* Multi-vector mode; the core timer at priority 7 (IPC0<4:2>), Timer2 at 3 (IPC2<4:2>). */
    REG(INTCON, SET) = 1u << 12;
    SFR(IPC0) = 7u << 2;
    SFR(IPC2) = 3u << 2;
    ct_next = mfc0_count() + 40000;
    mtc0_compare(ct_next);
    SFR(T2CON) = 0;
    SFR(TMR2) = 0;
    SFR(PR2) = 39999;
    SFR(T2CON) = (1u << 15) | (3u << 4);  /* ON, 1:8 */
    REG(IFS0, CLR) = (1u << 0) | (1u << 8);
    REG(IEC0, SET) = (1u << 0) | (1u << 8);
    __asm__ volatile("ei; ehb");

    next = mfc0_count();
    for (;;) {
        next += 20000;
        while ((int32_t)(mfc0_count() - next) < 0) {
        }
        REG(LATG, INV) = 1u << 6;           /* RG6 */
        if (ct_ticks == 200) {
            puts_("CT ticks=200 css=");
            put_dec(ct_srs & 0xf);
            puts_(" pss=");
            put_dec((ct_srs >> 6) & 0xf);
            puts_(" T2 ticks=");
            put_dec(t2_ticks);
            puts_(" css=");
            put_dec(t2_srs & 0xf);
            puts_(" main css=");
            put_dec(mfc0_srsctl() & 0xf);
            puts_("\nDONE\n");
            ct_ticks++;                     /* say it once */
            phase2();
        }
    }
}