ZH
games/alis / src / Alis / fn / rrq3d_3do.php · commit 9c722cf
<?phpnamespace App\Alis;// Port of alis/src/render/rrq3d_3do.c (MIT, Olivier Huguenot, Vadim Kindl).// Robinson's Requiem 3DO renderer: land tiles, billboards and sprites as 3DO cels, drawn into the// true-colour logic buffer (Img::$rgblogic: 3 bytes per pixel B, G, R, stride width * 3, see image_v6.php).// Struct layouts: see R3doVars. Every static function of the C file is prefixed with r3do_ (no collision// with rrq3d.c / deus3d.c names).// PORT: colours are 0xRRGGBB (the C ORs an opaque alpha 0xff000000 in: dropped).// PORT: the C float (single precision) math runs in PHP doubles: sub-pixel differences possible.const R3DO_CELMAX = 0x140;const R3DO_TILEMAX = 0x400;const R3DO_FARMUL = 2;const R3DO_CELCAP = R3DO_CELMAX * R3DO_FARMUL * R3DO_FARMUL;const R3DO_TILECAP = R3DO_TILEMAX * R3DO_FARMUL * R3DO_FARMUL;const R3DO_BUCKETS = 0x280;const R3DO_TEXMAX = 128;const R3DO_TEXBYTES = 0x80000;// '--far' doubles the view distance, 3DOs cel and tile limits grow with the areafunction r3do_viewfar(int $scr): int{    return mul32(xread32($scr + 0x146), Img::$ddrawdist ? R3DO_FARMUL : 1);}function r3do_celmax(): int{    return Img::$ddrawdist ? R3DO_CELCAP : R3DO_CELMAX;}function r3do_tilemax(): int{    return Img::$ddrawdist ? R3DO_TILECAP : R3DO_TILEMAX;}// C mulw(a, b) = (s32)((u32)a * (u32)b) -> mul32(); C sdiv(d, n) = d ? n / d : 0 -> cdiv(n, d)function r3do_tables(): int{    if (R3doVars::$tables_done) {        return 1;    }    R3doVars::$tables_done = 1;    // rotation recurrence: cos() is an fcos the 68040/060 lack (Line-F under plain TOS)    $costab = [];    $c = 1.0;    $s = 0.0;    $C = 0.99999882345170187925;    $S = 0.00153398018628476550;   // cos, sin(pi / 2048)    for ($i = 0; $i < 0x1000; $i++) {        $costab[$i] = (int) (4096 * $c);        $t = $c * $C - $s * $S;        $s = $s * $C + $c * $S;        $c = $t;    }    $costab[2048] = -4095;    $costab[2532] = -3017;    $recip = [0 => 0];    for ($i = 1; $i < 0x8000; $i++) {        $recip[$i] = intdiv(0x40000, $i);    }    R3doVars::$costab = $costab;    R3doVars::$recip = $recip;    return 1;}function r3do_rot(array &$m, int $a, int $b, int $ang): void{    if ($ang == 0) {        return;    }    $c = R3doVars::$costab[$ang & 0xfff];    $s = R3doVars::$costab[($ang + 0xc00) & 0xfff];    for ($i = 0; $i < 3; $i++) {        $x = $m[$a * 3 + $i];        $y = $m[$b * 3 + $i];        $m[$a * 3 + $i] = s16(($x * $c - $y * $s) >> 12);        $m[$b * 3 + $i] = s16(($y * $c + $x * $s) >> 12);    }    $x = $m[9 + $a];    $y = $m[9 + $b];    $m[9 + $a] = s32(mul32($x, $c) - mul32($y, $s)) >> 12;    $m[9 + $b] = s32(mul32($y, $c) + mul32($x, $s)) >> 12;}function r3do_cammat(int $scr, int $inv): array{    $m = [0x1000, 0, 0, 0, 0x1000, 0, 0, 0, 0x1000, 0, 0, 0];    $ax = xread16($scr + 0x18);    $az = xread16($scr + 0x1c);    if (!$inv) {        r3do_rot($m, 1, 2, $ax >> 4);        r3do_rot($m, 0, 1, $az >> 4);        for ($i = 0; $i < 3; $i++) {            $m[9 + $i] = s32($m[9 + $i] + xread32($scr + 0xc + $i * 4));        }        return $m;    }    for ($i = 0; $i < 3; $i++) {        $m[9 + $i] = s32($m[9 + $i] - xread32($scr + 0xc + $i * 4));    }    r3do_rot($m, 0, 1, (-$az) >> 4);    r3do_rot($m, 1, 2, (-$ax) >> 4);    return $m;}// s16 * s32 products are exact in a double (32-bit PHP) and in 64-bit ints: s32() of the sum wraps like Cfunction r3do_xform(array $m, array &$v): void{    $x = $v[0];    $y = $v[1];    $z = $v[2];    for ($r = 0; $r < 3; $r++) {        $v[$r] = s32($m[9 + $r] + (s32($m[$r * 3] * $x + $m[$r * 3 + 1] * $y + $m[$r * 3 + 2] * $z) >> 12));    }}/** Returns the pt3d [x, z, y, valid]. */function r3do_project(array $v, int $x, int $y, int $z): array{    $d = s32($v[10] + (s32($x * $v[3] + $y * $v[4] + $z * $v[5]) >> 12));    $px = s32($v[9] + (s32($x * $v[0] + $y * $v[1] + $z * $v[2]) >> 12));    $py = s32($v[11] + (s32($x * $v[6] + $y * $v[7] + $z * $v[8]) >> 12));    if ($d >= $v[16]) {        $r = R3doVars::$recip[$d < 0x8000 ? $d : 0x7fff] ?? 0;        return [            s32($v[14] + (mul32(mul32($r, $v[12]), $px) >> 18)),            $d,            s32($v[15] - (mul32(mul32($r, $v[13]), $py) >> 18)),            1,        ];    }    $div = $d < 1 ? $v[16] : $d;    $ox = s32($v[14] + cdiv(mul32($v[12], $px), $div));    $oy = s32($v[15] - cdiv(mul32($v[13], $py), $div));    if ($oy > 300) {        $oy = 300;    }    if ($oy < -0x32) {        $oy = -0x32;    }    if ($ox > 0x280) {        $ox = 0x280;    }    if ($ox < -0x140) {        $ox = -0x140;    }    return [$ox, $d, $oy, 1];}function r3do_fogpixc(int $scr, int $depth, int $shade): int{    $fade = 0x100;    $mode = xread16($scr + 0x94) & 0xffff;    if ($mode != 100 && $mode != 101) {        return 0x05000500;    }    $fog = Alis::$fog3do > 0xff ? 0xff : Alis::$fog3do;    $far = r3do_viewfar($scr);          // u32    if (ult($far, $depth)) {        $depth = $far;    }    $f = shr32(mul32($depth, R3doVars::$recip[ult($far, 0x8000) ? $far : 0x7fff]), 4);    $f = shr32(mul32($f, $f), 5);    $nf = s32(0x800000 - $f);    $g = shr32(mul32($fade, s32(shr32(mul32($f, $fog), 23) + (mul32(0x28 - $shade, $nf) >> 23))), 8);    if (ult(0xff, $g)) {        $g = 0xff;    }    $k = mul32($fade, shr32($nf, 15));    $kk = shr32($k, 8);    $pt = R3doVars::pixctab;    if (ule($kk, 0x80)) {        $pix = $pt[shr32($k, 12)];    } elseif (ule($kk, 0x100)) {        $pix = $pt[(($kk - 0x80) >> 5) + 8];    } elseif (ult($kk, 0x140)) {        $pix = $pt[13];    } else {        $pix = $pt[14];    }    return (($g >> 3) << 1) | (($g >> 3) << 17) | $pix | 0x400040;}/** C newcel(): PORT: the corners p and an accw override are passed in (the C callers fill them after). */function r3do_newcel(int $b, int $img, int $pixc, int $depth, int $owner, int $aux, array $p, int $accw = -1): void{    R3doVars::$ncel++;    if ($b > R3doVars::$maxbucket) {        R3doVars::$maxbucket = $b;    }    R3doVars::$bucket[$b][] = [        'img' => $img,        'p' => $p,        'pixc' => $pixc,        'accw' => $accw >= 0 ? $accw : (xread8($img + 0x1b) & 8),        'depth' => $depth,        'owner' => $owner,        'aux' => $aux,    ];}function r3do_billboard(int $img, array $p, int $scale, int $scr, int $yoff, int $tag): void{    if (R3doVars::$ncel >= r3do_celmax()) {        return;    }    if ($scale > 0x280000) {        $scale = 0x200000;    }    $b = $p[1] >> 5;    if ($b < 0 || $b >= R3DO_BUCKETS) {        return;    }    $yoff = s16($yoff);    $pixc = (xread8($img + 0x1b) & 4) ? 0x05000500 : r3do_fogpixc($scr, $b << 5, 0);    $w = mul32($scale, xread16($img + 0x10) & 0xffff) >> 19;    $top = mul32($scale, (xread16($img + 0x12) & 0xffff) + $yoff) >> 18;    $bot = mul32($scale, $yoff) >> 18;    $x = $p[0];    $y = $p[2];    r3do_newcel($b, $img, $pixc, $p[1], $tag, $yoff, [        s32($x - $w), s32($y - $top), s32($x + $w), s32($y - $top),        s32($x + $w), s32($y - $bot), s32($x - $w), s32($y - $bot),    ]);}function r3do_texquad(array $a, array $b, array $c, array $d, int $img, int $shade, int $scr, int $minh): void{    if (R3doVars::$ncel >= r3do_celmax()) {        return;    }    $z = $a[1];    if ($z < $b[1]) {        $z = $b[1];    }    if ($z < $c[1]) {        $z = $c[1];    }    if ($z < $d[1]) {        $z = $d[1];    }    if ($z < 0 || ($z >> 5) >= R3DO_BUCKETS) {        return;    }    $pixc = (xread8($img + 0x1b) & 4) ? 0x05000500 : r3do_fogpixc($scr, ($z >> 5) << 5, $shade);    r3do_newcel($z >> 5, $img, $pixc, $z, $scr, ($minh & 0xff) << 16,        [$a[0], $a[2], $d[0], $d[2], $c[0], $c[2], $b[0], $b[2]]);}function r3do_mirror(int $i, int $n): int{    return $i < 0 ? -$i : ($i >= $n ? 2 * $n - 2 - $i : $i);}function r3do_sprcel(int $img, array $p, int $scale, int $scr, int $e, int $hx, int $hy): void{    if (R3doVars::$ncel >= r3do_celmax()) {        return;    }    if ($scale > 0x280000) {        $scale = 0x200000;    }    $b = $p[1] >> 5;    if ($b < 0 || $b >= R3DO_BUCKETS) {        return;    }    $pixc = (xread8($img + 0x1b) & 4) ? 0x05000500 : r3do_fogpixc($scr, $b << 5, 0);    $ox = xread16($img + 0x14) - $hx;    $oy = xread16($img + 0x16) - $hy;    $w = xread16($img + 0x10) & 0xffff;    $h = xread16($img + 0x12) & 0xffff;    $l = s32($p[0] - (mul32($scale, $ox) >> 18));    $r = s32($p[0] + (mul32($scale, $w - $ox) >> 18));    $t = s32($p[2] - (mul32($scale, $oy) >> 18));    $bt = s32($p[2] + (mul32($scale, $h - $oy) >> 18));    $accw = -1;    if (xread8($e + 0x25) & 1) {        $k = $l;        $l = $r;        $r = $k;        $accw = 1;    }    r3do_newcel($b, $img, $pixc, $p[1], $e, 0, [$l, $t, $r, $t, $r, $bt, $l, $bt], $accw);}function r3do_mat7778(int $e0, int $cam): array{    $m = [0x1000, 0, 0, 0, 0x1000, 0, 0, 0, 0x1000, 0, 0, 0];    for ($e = $e0; $e; $e = xread32($e + 8)) {        $f = xread8($e + 0x25);        r3do_rot($m, 1, 2, xread16((($f & 2) && $cam ? $cam : $e) + 0x18) >> 4);        r3do_rot($m, 2, 0, xread16((($f & 4) && $cam ? $cam : $e) + 0x1a) >> 4);        r3do_rot($m, 0, 1, xread16((($f & 8) && $cam ? $cam : $e) + 0x1c) >> 4);        for ($i = 0; $i < 3; $i++) {            $m[9 + $i] = s32($m[9 + $i] + xread32($e + 0xc + $i * 4));        }    }    $chain = [];    $n = 0;    for ($c = $cam; $c && $n < 256; $c = xread32($c + 8)) {        $chain[$n++] = $c;    }    $f = $e0 ? xread8($e0 + 0x25) : 0;    while ($n--) {        $c = $chain[$n];        if (!($f & 0x30)) {            for ($i = 0; $i < 3; $i++) {                $m[9 + $i] = s32($m[9 + $i] - xread32($c + 0xc + $i * 4));            }            r3do_rot($m, 0, 1, (-xread16($c + 0x1c)) >> 4);        }        if (!($f & 0x20)) {            r3do_rot($m, 2, 0, (-xread16($c + 0x1a)) >> 4);        }        if (!($f & 0x10)) {            r3do_rot($m, 1, 2, (-xread16($c + 0x18)) >> 4);        }    }    return $m;}function r3do_elemmat(int $e): array{    return r3do_mat7778($e, 0);}// move an entity by d turned by itd own anglesfunction r3do_polar(int $vram, int $dx, int $dy, int $dz): void{    if (!r3do_tables()) {        return;    }    $m = r3do_mat7778($vram, xread32($vram + 8));    for ($r = 0; $r < 3; $r++) {        $v = s32($m[$r * 3] * $dx + $m[$r * 3 + 1] * $dy + $m[$r * 3 + 2] * $dz) >> 12;        xwrite32($vram + 0xc + $r * 4, s32($m[9 + $r] + $v + ($v < 0 ? 1 : 0)));    }}function r3do_landmap(int $scr, int &$L, int &$H): int{    $owner = xread16($scr + 0xc4);    $vram = $owner ? xread32(Alis::$atent + ($owner & 0xffff)) : 0;    if (!$vram) {        return 0;    }    $L = s32($vram + xread16($scr + 0xc6));    if (xread32($L - 0xc00) == 0) {        return 0;    }    $H = $L;    if (xread8($L - 2) & 0x80) {        $H = s32(xread32($L) + xread32(xread32($L)));    }    return 1;}// probe box against the land of mode-5 screenfunction r3do_maptest(int $scr, int $X, int $Y, array $mn, array $mx, int $walking, int &$mat): int{    $L = 0;    $H = 0;    if (!r3do_landmap($scr, $L, $H)) {        return 0;    }    $S = $L - 0xc00;    $xs = xread16($L - 0x3c0);    $ys = xread16($L - 0x3be);    $hs = xread16($L - 0x3b8);    $nx = xread16($L - 0x3cc);    $ny = xread16($L - 0x3ca);    $ix = $X >> $xs;    $iy = $Y >> $ys;    $cx = [$ix, $ix + 1, $ix + 1, $ix];    $cy = [$iy, $iy, $iy + 1, $iy + 1];    $c = [];    $avg = 0;    for ($i = 0; $i < 4; $i++) {        $c[$i] = xread16($H + (r3do_mirror($cx[$i], $nx) * $ny + r3do_mirror($cy[$i], $ny)) * 2) & 0xffff;        $avg += $c[$i] & 0xff;    }    $avg >>= 2;    if ($walking) {        $slot = $S + (($c[0] >> 8) & 0x3f) * 0x14;        $img = s32($slot + xread32($slot));        $x0 = mul32($ix, 1 << $xs);        $x1 = mul32($ix + 1, 1 << $xs);        $y0 = mul32($iy, 1 << $xs);        $y1 = mul32($iy + 1, 1 << $xs);        if (($x0 < $mx[0] && $x1 > $mn[0] && $y0 < $mx[1] && $y1 > $mn[1] && $mx[2] > 0 && mul32($avg, 1 << $hs) > $mn[2]) || $mx[2] < 0 || $mn[2] < 0) {            $mat = (xread16($img + 8) & 0xffff) | shl32(xread16($img + 0xa) & 0xffff, 16);            return -3;        }    }    for ($i = 0; $i < 4; $i++) {        $slot = $S + (($c[$i] >> 8) & 0x3f) * 0x14;        if (!xread32($slot + 0xc) || xread8($slot + 9) != 2) {            continue;        }        $obj = s32($slot + 0xc + xread32($slot + 0xc));        $r = xread16($obj + 0x10) & 0xffff;        $ht = xread16($obj + 0x12) & 0xffff;        $ox = mul32($cx[$i], 1 << $xs);        $oy = mul32($cy[$i], 1 << $xs);        $oz = mul32($c[$i] & 0xff, 1 << $hs);        if ($ox - 2 * $r <= $mx[0] && $ox + 2 * $r >= $mn[0] && $oy - 2 * $r <= $mx[1] && $oy + 2 * $r >= $mn[1] && $oz <= $mx[2] && $oz + $ht >= $mn[2]) {            $mat = (xread16($obj + 8) & 0xffff) | shl32(xread16($obj + 0xa) & 0xffff, 16);            return xread32($slot + 0x10);        }    }    return 0;}function r3do_cellat(int $vram, int $dx, int $dy, int $dz, int $L, int $mode): int{    $H = $L;    if (xread8($L - 2) & 0x80) {        $H = s32(xread32($L) + xread32(xread32($L)));    }    $xs = xread16($L - 0x3c0);    $ys = xread16($L - 0x3be);    $nx = xread16($L - 0x3cc);    $ny = xread16($L - 0x3ca);    r3do_polar($vram, $dx, $dy, $dz);    $X = 0;    $Y = 0;    for ($e = $vram; $e; $e = xread32($e + 8)) {        $X = s32($X + xread32($e + 0xc));        $Y = s32($Y + xread32($e + 0x10));    }    $ix = $X >> $xs;    $iy = $Y >> $ys;    $m0 = r3do_mirror($ix, $nx);    $n0 = r3do_mirror($iy, $ny);    $m1 = r3do_mirror($ix + 1, $nx);    $n1 = r3do_mirror($iy + 1, $ny);    $best = xread16($H + ($m0 * $ny + $n0) * 2) & 0xffff;    if ($mode == 2) {        return $best;    }    $k = [        xread16($H + ($m1 * $ny + $n0) * 2) & 0xffff,        xread16($H + ($m1 * $ny + $n1) * 2) & 0xffff,        xread16($H + ($m0 * $ny + $n1) * 2) & 0xffff,    ];    for ($i = 0; $i < 3; $i++) {        if ($mode == 0 ? ($k[$i] & 0xff) > ($best & 0xff) : ($k[$i] & 0xff) < ($best & 0xff)) {            $best = $k[$i];        }    }    return $best;}function r3do_sprites(int $scr, array $v, int $x0, int $x1, int $y0, int $y1): void{    $node = xread32($scr + 0x5e);    if (!$node) {        return;    }    for ($e = xread32($node + 0x38); $e;) {        $next = xread32($e + 0x38);        if (xread8($e + 0x22) == 0x32 && !(xread8($e + 0x24) & 0x40)) {            $m = r3do_elemmat($e);            if ($m[9] < $x1 && $x0 < $m[9] && $m[10] < $y1 && $y0 < $m[10]) {                $p = r3do_project($v, $m[9], $m[10], $m[11]);                $az = 0;                for ($k = $e; $k; $k = xread32($k + 8)) {                    $az = s16($az + xread16($k + 0x1c));                }                $n = s8(xread8($e + 0x2e));                $vi = 0;                if ($n) {                    $vi = ((($n * (((($az - xread16($scr + 0x1c)) & 0xffff) + 0x8000) & 0xffff) + 0x8000) >> 16) * 4) & 0xffff;                    if ($vi >= $n * 4) {                        $vi = 0;                    }                }                if (xread16($e + 0x1e) == 0) {                    xwrite16($e + 0x1e, 0x100);                }                $img = 0;                $mv = xread32($e + 0x50);                $hx = 0;                $hy = 0;                $skip = false;                if (!$mv) {                    $r = s32(xread32($e + 0x3c) + $vi);                    $img = s32($r + xread32($r));                } else {                    $r = s32($mv + $vi);                    $t6 = s32($r + xread32($r));                    if (!ult($t6, s32(Alis::$platform->ram_sz - 0x20)) || xread8($t6) != 6) {                        $skip = true;                    } else {                        $k = xread16($t6 + 0xe);                        $img = s32($r + $k * 4 + xread32($r + $k * 4));                        xwrite8($e + 0x25, xread8($t6 + 0xc) == 1 ? xread8($e + 0x25) | 1 : xread8($e + 0x25) & 0xfe);                        xwrite16($e + 0x54, xread16($t6 + 0x12));                        xwrite16($e + 0x56, xread16($t6 + 0x1a));                        $hx = xread16($e + 0x54);                        $hy = xread16($e + 0x56);                    }                }                if (!$skip && $img) {                    r3do_sprcel($img, $p, cdiv(mul32(mul32(xread16($e + 0x1e), $v[13]), 0x400), $p[1]), $scr, $e, $hx, $hy);                }            }        }        // skip:        if ($next == $e) {            break;        }        $e = $next;    }}function r3do_land(int $scr): void{    if (!r3do_tables()) {        return;    }    R3doVars::$bucket = [];    R3doVars::$ncel = 0;    R3doVars::$maxbucket = -1;    R3doVars::$drewland = 0;    $owner = xread16($scr + 0xc4);    if ($owner == 0) {        return;    }    $vram = xread32(Alis::$atent + ($owner & 0xffff));    if ($vram == 0) {        return;    }    $L = s32($vram + xread16($scr + 0xc6));    $slots = $L - 0xc00;    if (xread32($slots) == 0) {        return;    }    $H = $L;    if (xread8($L - 2) & 0x80) {        $H = s32(xread32($L) + xread32(xread32($L)));    }    $sh = s8(xread8($L - 0x3bf));    $cc = xread16($L - 0x3cc);    $rc = xread16($L - 0x3ca);    $tilemax = r3do_tilemax();    $cache = [];                        // pt3d per tile corner (C: valid = 0 for every entry)    $tocam = r3do_cammat($scr, 0);    $v = r3do_cammat($scr, 1);    $v[12] = xread16($scr + 0x48);              // fx    $v[13] = xread16($scr + 0x4c);              // fy    $v[14] = xread16($scr + 0x62) & 0xffff;     // cx    $v[15] = xread16($scr + 0x64) & 0xffff;     // cy    $v[16] = xread32($scr + 0x142);             // near    $far = r3do_viewfar($scr);    // view triangle (eye, far right, far left) in world space    $half = cdiv(mul32($v[14], $far), $v[12]);    $p1 = [$half, $far, 0];    $p2 = [s32(-$half), $far, 0];    $p3 = [0, 0, 0];    r3do_xform($tocam, $p1);    r3do_xform($tocam, $p2);    r3do_xform($tocam, $p3);    $ymin = min($p1[1], $p2[1], $p3[1]);    $ymax = max($p1[1], $p2[1], $p3[1]);    $xmin = min($p1[0], $p2[0], $p3[0]);    $xmax = max($p1[0], $p2[0], $p3[0]);    $d1x = s32($p1[0] - $p3[0]);    $d1y = s32($p1[1] - $p3[1]);    $d2x = s32($p2[0] - $p3[0]);    $d2y = s32($p2[1] - $p3[1]);    $d3x = s32($p2[0] - $p1[0]);    $d3y = s32($p2[1] - $p1[1]);    // per tile row, columns triangle crosses at the sampled lines    $r0 = $ymin >> $sh;    $r1 = $ymax >> $sh;    $eyerow = $p3[1] >> $sh;    $e4 = $xmin >> $sh;    $e8 = $xmax >> $sh;    if ($r1 - $r0 + 1 > $tilemax) {        return;    }    $spans = [];    $ex1 = 0;    $ex2 = 0;    $ex3 = 0;    $prevlo = 0;    $prevhi = 0;    $first = 1;    for ($r = $r0; $r <= $r1; $r++) {        for ($pass = ($r == $eyerow) ? 0 : 1; $pass < 2; $pass++) {            $y = $pass == 0 ? $p3[1] : ($r == $r1 ? shl32($r, $sh) : shl32($r + 1, $sh));            if ($d1y) {                $ex1 = s32(cdiv(mul32($d1x, s32($y - $p1[1])), $d1y) + $p1[0]);            }            if ($d2y) {                $ex2 = s32(cdiv(mul32($d2x, s32($y - $p2[1])), $d2y) + $p2[0]);            }            if ($d3y) {                $ex3 = s32(cdiv(mul32($d3x, s32($y - $p2[1])), $d3y) + $p2[0]);            }            $ok1 = $ex1 >= $xmin && $ex1 <= $xmax && $d1y;            $ok2 = $ex2 >= $xmin && $ex2 <= $xmax && $d2y;            $ok3 = $ex3 >= $xmin && $ex3 <= $xmax && $d3y;            $a = $ex2;            $b = $ex3;            if ($ok1 && !$ok2) {                $a = $ex1;            } elseif ($ok1) {                $a = $ex1;                $b = $ex2;            }            $lo = $a < $b ? $a : $b;            $hi = $a < $b ? $b : $a;            if ($ok1 && $ok2 && $ok3) {                if ($ex3 < $lo) {                    $lo = $ex3;                }                if ($ex3 > $hi) {                    $hi = $ex3;                }            }            $lo >>= $sh;            $hi >>= $sh;            if ($lo == $hi) {                $hi++;                $lo--;            }            $rawlo = $lo;            $rawhi = $hi;            if (!$first) {                if ($hi < $prevhi) {                    $hi = $prevhi;                }                if ($prevlo < $lo) {                    $lo = $prevlo;                }            }            $first = 0;            $spans[$r - $r0] = [$lo, $hi];            if ($lo < $e4) {                $e4 = $lo;            }            if ($hi > $e8) {                $e8 = $hi;            }            $prevlo = $rawlo;            $prevhi = $rawhi;        }    }    $f0 = $e8 - $e4 + 2;    if ($f0 * ($r1 + 2 - $r0) > $tilemax) {        return;    }    R3doVars::$drewland = 1;    r3do_sprites($scr, $v, shl32($e4 - 1, $sh), shl32($e8 + 1, $sh), shl32($r0 - 1, $sh), shl32($r1 + 1, $sh));    $camx = xread32($scr + 0xc);    $camy = xread32($scr + 0x10);    $camz = xread32($scr + 0x14);    $clx1 = xread16($scr + 0x66) & 0xffff;    $cly1 = xread16($scr + 0x68) & 0xffff;    $clx2 = xread16($scr + 0x6a) & 0xffff;    $cly2 = xread16($scr + 0x6c) & 0xffff;    $bscale = mul32(xread16($scr + 0x4c), 0x2a800);    for ($r = $r0; $r <= $r1; $r++) {        $rowa = $f0 * ($r - $r0);        $rowb = $f0 * ($r + 1 - $r0);        $rm0 = r3do_mirror($r, $rc);        $rm1 = r3do_mirror($r + 1, $rc);        $y0 = shl32($r, $sh);        $y1 = shl32($r + 1, $sh);        [$clo, $chi] = $spans[$r - $r0];        for ($col = $clo; $col <= $chi; $col++) {            $i00 = $col - $e4 + $rowa + 1;            $i10 = $col + 1 - $e4 + $rowa + 1;            $i11 = $col + 1 - $e4 + $rowb + 1;            $i01 = $col - $e4 + $rowb + 1;            $cm0 = r3do_mirror($col, $cc);            $cm1 = r3do_mirror($col + 1, $cc);            $c00 = xread16($H + ($rc * $cm0 + $rm0) * 2) & 0xffff;            $c10 = xread16($H + ($rc * $cm1 + $rm0) * 2) & 0xffff;            $c11 = xread16($H + ($rc * $cm1 + $rm1) * 2) & 0xffff;            $c01 = xread16($H + ($rc * $cm0 + $rm1) * 2) & 0xffff;            $h00 = $c00 & 0xff;            $h10 = $c10 & 0xff;            $h11 = $c11 & 0xff;            $h01 = $c01 & 0xff;            $x0 = shl32($col, $sh);            $x1 = shl32($col + 1, $sh);            $slot = $slots + (($c00 >> 8) & 0x3f) * 0x14;            $obj = xread32($slot + 0xc) ? s32($slot + 0xc + xread32($slot + 0xc)) : 0;            $front = s32(mul32(s32($y0 - $camy), $h00 - $h01) + mul32($h00 - $h10, s32($x0 - $camx)) + shl32(s32($h00 - $camz), $sh)) < 0                || s32(mul32($h01 - $h11, s32($x1 - $camx)) + mul32($h10 - $h11, s32($y1 - $camy)) + shl32(s32($h11 - $camz), $sh)) < 0;            $q00 = $cache[$i00] ?? ($cache[$i00] = r3do_project($v, $x0, $y0, $h00));            if (!$front) {                if ($obj) {                    r3do_billboard($obj, $q00, cdiv($bscale, $q00[1]), $scr, xread16($slot + 0xa), xread32($slot + 0x10));                }                continue;            }            $q10 = $cache[$i10] ?? ($cache[$i10] = r3do_project($v, $x1, $y0, $h10));            $q11 = $cache[$i11] ?? ($cache[$i11] = r3do_project($v, $x1, $y1, $h11));            $q01 = $cache[$i01] ?? ($cache[$i01] = r3do_project($v, $x0, $y1, $h01));            if ($obj) {                r3do_billboard($obj, $q00, cdiv($bscale, $q00[1]), $scr, xread16($slot + 0xa), xread32($slot + 0x10));            }            if (($q00[2] < $cly1 && $q10[2] < $cly1 && $q11[2] < $cly1 && $q01[2] < $cly1)                || ($q00[2] > $cly2 && $q10[2] > $cly2 && $q11[2] > $cly2 && $q01[2] > $cly2)                || ($q00[0] < $clx1 && $q10[0] < $clx1 && $q11[0] < $clx1 && $q01[0] < $clx1)                || ($q00[0] > $clx2 && $q10[0] > $clx2 && $q11[0] > $clx2 && $q01[0] > $clx2)) {                continue;            }            if ($q00[0] == $q10[0] && $q11[0] == $q01[0] && $q00[2] == $q10[2] && $q11[2] == $q01[2]) {                continue;            }            if (xread32($slot)) {                $shade = (($c00 >> 14) + ($c10 >> 14) + ($c11 >> 14) + ($c01 >> 14)) >> 1;                $minh = min($h00, $h10, $h11, $h01);                r3do_texquad($q00, $q10, $q11, $q01, s32($slot + xread32($slot)), $shade * 10 - 0x14, $scr, $minh);            }        }    }}function r3do_texsig(int $img, int $len): string{    if (ult($len, 16) || ult(Alis::$platform->ram_sz, s32($img + $len))) {        $len = 16;    }    return substr(Alis::$mem, $img, 32) . substr(Alis::$mem, $img + $len - 16, 16);}function r3do_texflush(): void{    R3doVars::$tex = [];    R3doVars::$ntex = 0;    R3doVars::$texbytes = 0;}/** Decoded cel pixels (u16 int array, index y * w + x) or null. */function r3do_texget(int $img, int &$w, int &$h): ?array{    $len = read32($img + 4);    $sig = r3do_texsig($img, $len);    $slot = -1;    $ntex = R3doVars::$ntex;    for ($i = 0; $i < $ntex; $i++) {        $t = R3doVars::$tex[$i];        if ($t['img'] != $img) {            continue;        }        if ($t['len'] == $len && $t['sig'] === $sig) {            $w = $t['w'];            $h = $t['h'];            return $t['px'];        }        // stale: same address, different cel        R3doVars::$texbytes -= $t['w'] * $t['h'] * 2;        R3doVars::$tex[$i]['px'] = null;        $slot = $i;        break;    }    $px = cel_decode($img, $w, $h);    if ($px === null) {        if ($slot >= 0) {            R3doVars::$ntex--;            R3doVars::$tex[$slot] = R3doVars::$tex[R3doVars::$ntex];            unset(R3doVars::$tex[R3doVars::$ntex]);        }        return null;    }    $sz = $w * $h * 2;    if ($slot < 0 && (R3doVars::$ntex >= R3DO_TEXMAX || R3doVars::$texbytes + $sz > R3DO_TEXBYTES)) {        r3do_texflush();    }    if ($slot < 0) {        $slot = R3doVars::$ntex++;    }    R3doVars::$tex[$slot] = ['img' => $img, 'len' => $len, 'sig' => $sig, 'w' => $w, 'h' => $h, 'px' => $px];    R3doVars::$texbytes += $sz;    return $px;}function r3do_rgb5(int $r, int $g, int $b): int{    return ((($r << 3) | ($r >> 2)) << 16) | ((($g << 3) | ($g >> 2)) << 8) | (($b << 3) | ($b >> 2));}function r3do_edge(array $p, int $a, int $b): float{    return hypot((float) ($p[$b] - $p[$a]), (float) ($p[$b + 1] - $p[$a + 1]));}// MapCel corners as p = a + e*u + f*v + g*u*v, plus the ~1.5 pixel overlap per texture axis// (to keep neighbouring quads from showing seams)function r3do_celsetup(array $c): array{    $p = $c['p'];    $ax = (float) $p[0];    $ay = (float) $p[1];    $ex = $p[2] - $ax;    $ey = $p[3] - $ay;    $fx = $p[6] - $ax;    $fy = $p[7] - $ay;    $gx = $ax - $p[2] + $p[4] - $p[6];    $gy = $ay - $p[3] + $p[5] - $p[7];    return [        'ax' => $ax, 'ay' => $ay, 'ex' => $ex, 'ey' => $ey, 'fx' => $fx, 'fy' => $fy, 'gx' => $gx, 'gy' => $gy,        'k2' => $gx * $fy - $gy * $fx,        'eu' => 1.5 / max(min(r3do_edge($p, 0, 2), r3do_edge($p, 6, 4)), 1.0),        'ev' => 1.5 / max(min(r3do_edge($p, 0, 6), r3do_edge($p, 2, 4)), 1.0),    ];}function r3do_celbox(array $c, int &$x1, int &$y1, int &$x2, int &$y2): void{    $p = $c['p'];    $x1 = $x2 = $p[0];    $y1 = $y2 = $p[1];    for ($i = 2; $i < 8; $i += 2) {        if ($p[$i] < $x1) {            $x1 = $p[$i];        }        if ($p[$i] > $x2) {            $x2 = $p[$i];        }        if ($p[$i + 1] < $y1) {            $y1 = $p[$i + 1];        }        if ($p[$i + 1] > $y2) {            $y2 = $p[$i + 1];        }    }}/** C (s32) of a float: truncation, 0 for NaN / out of range (C: undefined). */function r3do_ftoi(float $f): int{    if (is_nan($f) || $f >= 2147483647.0 || $f <= -2147483648.0) {        return $f > 0 ? 0x7fffffff : ($f < 0 ? -0x7fffffff - 1 : 0);    }    return (int) $f;}// texel index under pixel (xx, yy), or -1function r3do_texel(float $u, float $v, int $w, int $h): int{    $tx = r3do_ftoi($u * $w);    $ty = r3do_ftoi($v * $h);    $tx = $tx < 0 ? 0 : ($tx >= $w ? $w - 1 : $tx);    $ty = $ty < 0 ? 0 : ($ty >= $h ? $h - 1 : $ty);    return $ty * $w + $tx;}// barycentric weights of (px,py) for b and c in triangle abc (points = offsets ia, ib, ic into p)function r3do_bary(float $px, float $py, array $p, int $ia, int $ib, int $ic, float &$lb, float &$lc): float{    $a0 = $p[$ia];    $a1 = $p[$ia + 1];    $b0 = $p[$ib];    $b1 = $p[$ib + 1];    $c0 = $p[$ic];    $c1 = $p[$ic + 1];    $d = (float) ($b0 - $a0) * ($c1 - $a1) - (float) ($c0 - $a0) * ($b1 - $a1);    if (abs($d) < 1e-6) {        return 0.0;    }    $lb = (($px - $a0) * ($c1 - $a1) - ($c0 - $a0) * ($py - $a1)) / $d;    $lc = (($b0 - $a0) * ($py - $a1) - ($px - $a0) * ($b1 - $a1)) / $d;    return $d;}// pixel within e screen pixels of triangle abc (weights lb, lc; d = signed area x2)function r3do_intri(array $p, int $ia, int $ib, int $ic, float $d, float $lb, float $lc, float $e): bool{    $k = $e / abs($d);    $la = 1 - $lb - $lc;    return $la >= -$k * hypot($p[$ic] - $p[$ib], $p[$ic + 1] - $p[$ib + 1])        && $lb >= -$k * hypot($p[$ic] - $p[$ia], $p[$ic + 1] - $p[$ia + 1])        && $lc >= -$k * hypot($p[$ib] - $p[$ia], $p[$ib + 1] - $p[$ia + 1]);}// non-convex quad leaves pixels with no bilinear solution, which 3DOs forward mapping still covers: map those through the quad's two triangles insteadfunction r3do_celtri(array $c, float $px, float $py, int $w, int $h, float $e): int{    $p = $c['p'];    $lb = 0.0;    $lc = 0.0;    $d = r3do_bary($px, $py, $p, 0, 2, 4, $lb, $lc);    if ($d != 0 && ($d > 0 || $c['accw']) && r3do_intri($p, 0, 2, 4, $d, $lb, $lc, $e)) {        return r3do_texel($lb + $lc, $lc, $w, $h);    }    $d = r3do_bary($px, $py, $p, 0, 4, 6, $lb, $lc);    if ($d != 0 && ($d > 0 || $c['accw']) && r3do_intri($p, 0, 4, 6, $d, $lb, $lc, $e)) {        return r3do_texel($lb, $lb + $lc, $w, $h);    }    return -1;}// texel index under pixel (xx, yy), or -1function r3do_celtexel(array $c, array $m, int $w, int $h, int $xx, int $yy): int{    $hx = $xx + 0.5 - $m['ax'];    $hy = $yy + 0.5 - $m['ay'];    $k1 = $m['ex'] * $m['fy'] - $m['ey'] * $m['fx'] + $hx * $m['gy'] - $hy * $m['gx'];    $k0 = $hx * $m['ey'] - $hy * $m['ex'];    if (abs($m['k2']) < 1e-4) {        if (abs($k1) < 1e-6) {            return r3do_celtri($c, $xx + 0.5, $yy + 0.5, $w, $h, 1.5);        }        $v = -$k0 / $k1;    } else {        $disc = $k1 * $k1 - 4 * $k0 * $m['k2'];        if ($disc < 0) {            return r3do_celtri($c, $xx + 0.5, $yy + 0.5, $w, $h, 1.5);        }        $disc = sqrt($disc);        $v = (-$k1 - $disc) / (2 * $m['k2']);        if ($v < -$m['ev'] || $v > 1 + $m['ev']) {            $v = (-$k1 + $disc) / (2 * $m['k2']);        }    }    $dx = $m['ex'] + $m['gx'] * $v;    $dy = $m['ey'] + $m['gy'] * $v;    if (abs($dx) > abs($dy)) {        $u = $dx != 0 ? ($hx - $m['fx'] * $v) / $dx : NAN;    } else {        $u = $dy != 0 ? ($hy - $m['fy'] * $v) / $dy : NAN;   // C: float division by 0 = inf / nan    }    if (!($u >= -$m['eu'] && $u <= 1 + $m['eu'] && $v >= -$m['ev'] && $v <= 1 + $m['ev'])) {        return r3do_celtri($c, $xx + 0.5, $yy + 0.5, $w, $h, 1.5);    }    $wind = ($m['ex'] + $m['gx'] * $v) * ($m['fy'] + $m['gy'] * $u) - ($m['ey'] + $m['gy'] * $v) * ($m['fx'] + $m['gx'] * $u);    if ($wind < 0 && !$c['accw']) {        return -1;    }    return r3do_texel($u, $v, $w, $h);}// 3DO PIXC per channel: min(31, c * mf / df + av) (halved if bit 0); df is a power of twofunction r3do_pixc_lut(int $pixc): void{    if ($pixc == R3doVars::$pixlutc) {        return;    }    static $dsh = [4, 1, 2, 3];    R3doVars::$pixlutc = $pixc;    R3doVars::$pixstr = [];    $mf = (($pixc >> 10) & 7) + 1;    $sh = $dsh[($pixc >> 8) & 3];    $av = (($pixc >> 6) & 3) == 1 ? ($pixc >> 1) & 0x1f : 0;    $lut = [];    for ($i = 0; $i < 32; $i++) {        $v = (($i * $mf) >> $sh) + $av;        if ($pixc & 1) {            $v >>= 1;        }        if ($v > 31) {            $v = 31;        }        $lut[$i] = ($v << 3) | ($v >> 2);   // C: pixlut[0] = e << 16 (red), [1] = e << 8, [2] = e    }    R3doVars::$pixlut = $lut;}/** The 3 bytes B, G, R of texel t through the current pixlut (cached in R3doVars::$pixstr). */function r3do_pixstr(int $t): string{    $lut = R3doVars::$pixlut;    return R3doVars::$pixstr[$t] = chr($lut[$t & 0x1f]) . chr($lut[($t >> 5) & 0x1f]) . chr($lut[($t >> 10) & 0x1f]);}// one quad edge stepped per scanline: x in 24.8, u/v in texels 16.16 (celedge = [y0, y1, x, dx, u, du, v, dv])function r3do_celedge_set(array &$e, int $ax, int $ay, int $bx, int $by, int $ua, int $va, int $ub, int $vb, int $ystart): int{    if ($ay == $by) {        return 0;    }    if ($ay > $by) {        [$ax, $bx] = [$bx, $ax];        [$ay, $by] = [$by, $ay];        [$ua, $ub] = [$ub, $ua];        [$va, $vb] = [$vb, $va];    }    $dy = $by - $ay;    $e[0] = $ay;    $e[1] = $by;    // C: (s32)(((s64)(b0 - a0) << 8) / dy): a float on 32-bit PHP if it overflows, truncated like C    $num = ($bx - $ax) * 256;    $dx = s32(is_int($num) ? intdiv($num, $dy) : (int) ($num / $dy));    $e[3] = $dx;    $du = intdiv(s32($ub - $ua), $dy);    $dv = intdiv(s32($vb - $va), $dy);    $e[5] = $du;    $e[7] = $dv;    // first scanline centre at or below the top vertex, or the clip top    $y = $ystart > $ay ? $ystart : $ay;    $k = $y - $ay;    $e[2] = s32(shl32($ax, 8) + $dx * $k + intdiv($dx, 2));    $e[4] = s32($ua + $du * $k + intdiv($du, 2));    $e[6] = s32($va + $dv * $k + intdiv($dv, 2));    return 1;}// one textured span [xs, xe) on row y, clipped to [cx1, cx2]; the ends (overlap pixels) are// clamped, the rest runs unclamped// PORT: celpix() is inlined (clamped texel fetch, then the 3 bytes B,G,R written in place).function r3do_celspan(array $px, int $w, int $h, int $y, int $xs, int $xe, int $u, int $v, int $du, int $dv, int $cx1, int $cx2): void{    $U = $w << 16;    $V = $h << 16;    if ($xs < $cx1) {        $u = s32($u + $du * ($cx1 - $xs));        $v = s32($v + $dv * ($cx1 - $xs));        $xs = $cx1;    }    if ($xe > $cx2 + 1) {        $xe = $cx2 + 1;    }    $n = $xe - $xs;    if ($n <= 0) {        return;    }    $d = Img::$rgblogic + ($y * Alis::$platform->width + $xs) * 3;    $m = &Alis::$mem;    $map = &R3doVars::$pixstr;    $wm1 = $w - 1;    $hm1 = $h - 1;    // clamped ends    for ($k = 0; $k < 2 && $n > 0 && ($u < 0 || $u >= $U || $v < 0 || $v >= $V); $k++, $n--, $d += 3, $u = s32($u + $du), $v = s32($v + $dv)) {        $tu = $u < 0 ? 0 : ($u >= $U ? $wm1 : $u >> 16);        $tv = $v < 0 ? 0 : ($v >= $V ? $hm1 : $v >> 16);        $t = $px[$tv * $w + $tu];        if ($t) {            $s = $map[$t] ?? r3do_pixstr($t);            $m[$d] = $s[0];            $m[$d + 1] = $s[1];            $m[$d + 2] = $s[2];        }    }    for ($k = 0; $k < 2 && $n > 0; $k++, $n--) {        $ue = s32($u + $du * ($n - 1));        $ve = s32($v + $dv * ($n - 1));        if (!($ue < 0 || $ue >= $U || $ve < 0 || $ve >= $V)) {            break;        }        $tu = $ue < 0 ? 0 : ($ue >= $U ? $wm1 : $ue >> 16);        $tv = $ve < 0 ? 0 : ($ve >= $V ? $hm1 : $ve >> 16);        $t = $px[$tv * $w + $tu];        if ($t) {            $s = $map[$t] ?? r3do_pixstr($t);            $o = $d + ($n - 1) * 3;            $m[$o] = $s[0];            $m[$o + 1] = $s[1];            $m[$o + 2] = $s[2];        }    }    if ($n <= 0) {        return;    }    $ue = s32($u + $du * ($n - 1));    $ve = s32($v + $dv * ($n - 1));    if ($u < 0 || $u >= $U || $v < 0 || $v >= $V || $ue < 0 || $ue >= $U || $ve < 0 || $ve >= $V) {        for (; $n > 0; $n--, $d += 3, $u = s32($u + $du), $v = s32($v + $dv)) {            $tu = $u < 0 ? 0 : ($u >= $U ? $wm1 : $u >> 16);            $tv = $v < 0 ? 0 : ($v >= $V ? $hm1 : $v >> 16);            $t = $px[$tv * $w + $tu];            if ($t) {                $s = $map[$t] ?? r3do_pixstr($t);                $m[$d] = $s[0];                $m[$d + 1] = $s[1];                $m[$d + 2] = $s[2];            }        }        return;    }    // both ends inside: every u/v of the span stays in [0, U) / [0, V), no wrap    for (; $n > 0; $n--, $d += 3, $u += $du, $v += $dv) {        $t = $px[($v >> 16) * $w + ($u >> 16)];        if ($t) {            $s = $map[$t] ?? r3do_pixstr($t);            $m[$d] = $s[0];            $m[$d + 1] = $s[1];            $m[$d + 2] = $s[2];        }    }}// span between edges l and r (any order) on row y, plus the overlap row above/below the celfunction r3do_celrow(array $l, array $r, int $y): void{    if ($l[2] > $r[2]) {        [$l, $r] = [$r, $l];    }    // 1 pixel overlap all round hides T-junction cracks between terrain quads (old 1.5 px margin)    $xs = (($l[2] - 128 + 255) >> 8) - 1;    $xe = (($r[2] - 128 + 255) >> 8) + 1;    $span = ($r[2] - $l[2]) >> 4;    $du = $span > 0 ? intdiv(shl32(s32($r[4] - $l[4]), 4), $span) : 0;    $dv = $span > 0 ? intdiv(shl32(s32($r[6] - $l[6]), 4), $span) : 0;    $f = ($xs << 8) + 128 - $l[2];    $u = s32($l[4] + ($du >> 8) * $f);    $v = s32($l[6] + ($dv >> 8) * $f);    $px = R3doVars::$cc_px;    $w = R3doVars::$cc_w;    $h = R3doVars::$cc_h;    $cx1 = R3doVars::$cc_cx1;    $cx2 = R3doVars::$cc_cx2;    r3do_celspan($px, $w, $h, $y, $xs, $xe, $u, $v, $du, $dv, $cx1, $cx2);    if ($y == R3doVars::$cc_ymin && $y - 1 >= R3doVars::$cc_cy1 && R3doVars::$cc_ymin == R3doVars::$cc_ytop) {        r3do_celspan($px, $w, $h, $y - 1, $xs, $xe, $u, $v, $du, $dv, $cx1, $cx2);    }    if ($y == R3doVars::$cc_ymax - 1 && $y + 1 <= R3doVars::$cc_cy2 && R3doVars::$cc_ymax == R3doVars::$cc_ybot) {        r3do_celspan($px, $w, $h, $y + 1, $xs, $xe, $u, $v, $du, $dv, $cx1, $cx2);    }}function r3do_celstep(array &$e): void{    $e[2] = s32($e[2] + $e[3]);    $e[4] = s32($e[4] + $e[5]);    $e[6] = s32($e[6] + $e[7]);}// edge from corner i to corner n, starting at row y; 0 if horizontalfunction r3do_celedge_at(array &$e, array $p, int $i, int $n, int $y): int{    static $celcu = [0, 1, 1, 0], $celcv = [0, 0, 1, 1];    $U = (R3doVars::$cc_w << 16) - 1;    $V = (R3doVars::$cc_h << 16) - 1;    return r3do_celedge_set($e, $p[$i * 2], $p[$i * 2 + 1], $p[$n * 2], $p[$n * 2 + 1],        $celcu[$i] ? $U : 0, $celcv[$i] ? $V : 0, $celcu[$n] ? $U : 0, $celcv[$n] ? $V : 0, $y);}// convex quad: one left and one right chain from the top cornerfunction r3do_celconvex(array $p): void{    $top = 0;    for ($i = 1; $i < 4; $i++) {        if ($p[$i * 2 + 1] < $p[$top * 2 + 1]) {            $top = $i;        }    }    $a = [0, 0, 0, 0, 0, 0, 0, 0];    $b = [0, 0, 0, 0, 0, 0, 0, 0];    $ai = $top;    $bi = $top;    $y = R3doVars::$cc_ymin;    $ymax = R3doVars::$cc_ymax;    $a[1] = $b[1] = $y - 0x8000;    while ($y < $ymax) {        // next edge on each chain once the current one ends (horizontal ones are skipped)        for ($k = 0; $k < 4 && $y >= $a[1]; $k++) {            $n = ($ai + 3) & 3;            if (!r3do_celedge_at($a, $p, $ai, $n, $y) || $a[1] <= $y) {                $a[1] = $y - 0x8000;            }            $ai = $n;        }        for ($k = 0; $k < 4 && $y >= $b[1]; $k++) {            $n = ($bi + 1) & 3;            if (!r3do_celedge_at($b, $p, $bi, $n, $y) || $b[1] <= $y) {                $b[1] = $y - 0x8000;            }            $bi = $n;        }        if ($y >= $a[1] || $y >= $b[1]) {            return;        }        $end = $a[1] < $b[1] ? $a[1] : $b[1];        if ($end > $ymax) {            $end = $ymax;        }        for (; $y < $end; $y++) {            r3do_celrow($a, $b, $y);            r3do_celstep($a);            r3do_celstep($b);        }    }}// concave or bowtie quad: even-odd over all four edgesfunction r3do_celgeneral(array $p): void{    $e = [];    $ne = 0;    $ymin = R3doVars::$cc_ymin;    for ($i = 0; $i < 4; $i++) {        $t = [0, 0, 0, 0, 0, 0, 0, 0];        if (r3do_celedge_at($t, $p, $i, ($i + 1) & 3, $ymin)) {            $e[$ne++] = $t;        }    }    for ($y = $ymin; $y < R3doVars::$cc_ymax; $y++) {        $x = [];        $nx = 0;        for ($i = 0; $i < $ne; $i++) {            if ($y < $e[$i][0] || $y >= $e[$i][1]) {                continue;            }            $k = $nx++;            while ($k > 0 && $e[$x[$k - 1]][2] > $e[$i][2]) {                $x[$k] = $x[$k - 1];                $k--;            }            $x[$k] = $i;        }        for ($k = 0; $k + 1 < $nx; $k += 2) {            r3do_celrow($e[$x[$k]], $e[$x[$k + 1]], $y);        }        for ($i = 0; $i < $ne; $i++) {            if ($y >= $e[$i][0] && $y < $e[$i][1]) {                r3do_celstep($e[$i]);            }        }    }}// 3DO MapCel quad, scanline filled; u/v exact on the edges, linear across each span. Pixel// centres decide coverage, so shared edges leave no seams.function r3do_drawcel(array $c, int $cx1, int $cy1, int $cx2, int $cy2): void{    $w = 0;    $h = 0;    $px = r3do_texget($c['img'], $w, $h);    if ($px === null) {        return;    }    $p = $c['p'];    $area = 0;    $ymin = $p[1];    $ymax = $p[1];    $pos = 0;    $neg = 0;    for ($i = 0; $i < 4; $i++) {        $a = $i * 2;        $b = (($i + 1) & 3) * 2;        $d = (($i + 2) & 3) * 2;        $area += $p[$a] * $p[$b + 1] - $p[$b] * $p[$a + 1];        $z = ($p[$b] - $p[$a]) * ($p[$d + 1] - $p[$b + 1]) - ($p[$b + 1] - $p[$a + 1]) * ($p[$d] - $p[$b]);        $pos |= (int) ($z > 0);        $neg |= (int) ($z < 0);        if ($p[$a + 1] < $ymin) {            $ymin = $p[$a + 1];        }        if ($p[$a + 1] > $ymax) {            $ymax = $p[$a + 1];        }    }    // PORT: area is a C s32 sum (wraps); here exact (int or float): only its sign / zero test matters    if ($area == 0 || ($area < 0 && !$c['accw'])) {        return;    }    R3doVars::$cc_ytop = $ymin;    R3doVars::$cc_ybot = $ymax;    R3doVars::$cc_ymin = $ymin < $cy1 ? $cy1 : $ymin;    R3doVars::$cc_ymax = $ymax > $cy2 + 1 ? $cy2 + 1 : $ymax;    if (R3doVars::$cc_ymin >= R3doVars::$cc_ymax) {        return;    }    R3doVars::$cc_px = $px;    R3doVars::$cc_w = $w;    R3doVars::$cc_h = $h;    R3doVars::$cc_cx1 = $cx1;    R3doVars::$cc_cy1 = $cy1;    R3doVars::$cc_cx2 = $cx2;    R3doVars::$cc_cy2 = $cy2;    r3do_pixc_lut($c['pixc']);    // (celrows[]: row pointers for the C asm span loop, not needed here)    if ($pos && $neg) {        r3do_celgeneral($p);    } else {        r3do_celconvex($p);    }    R3doVars::$cc_px = [];}/** C r3do_pick(scr, x, y, s32 w3[3], u32 *aux, u32 *img): returns the owner of the cel under (x, y). */function r3do_pick(int $scr, int $x, int $y, array &$w3, int &$aux, int &$img): int{    $best = null;    $bw = 0x7fffffff;    $bh = 0x7fffffff;    $hit = false;    for ($b = 0; $b <= R3doVars::$maxbucket && !$hit; $b++) {        $list = R3doVars::$bucket[$b] ?? [];        for ($i = count($list) - 1; $i >= 0; $i--) {            $c = $list[$i];            $x1 = $y1 = $x2 = $y2 = 0;            r3do_celbox($c, $x1, $y1, $x2, $y2);            if ($x < $x1 || $x > $x2 || $y < $y1 || $y > $y2) {                continue;            }            $w = 0;            $h = 0;            $px = r3do_texget($c['img'], $w, $h);            $m = r3do_celsetup($c);            $t = $px !== null ? r3do_celtexel($c, $m, $w, $h, $x, $y) : -1;            $hit = $t >= 0 && $px[$t];            if ($hit && ($best === null || $c['depth'] < $best['depth'] || ($c['depth'] == $best['depth'] && ($x2 - $x1 < $bw || $y2 - $y1 < $bh)))) {                $best = $c;                $bw = $x2 - $x1;                $bh = $y2 - $y1;            }        }    }    if ($best === null) {        return 0;    }    $d = $best['depth'];    $v = [        cdiv(mul32($d, $x - (xread16($scr + 0x62) & 0xffff)), xread16($scr + 0x48)),        $d,        cdiv(mul32($d, (xread16($scr + 0x64) & 0xffff) - $y), xread16($scr + 0x4c)),    ];    $m = r3do_cammat($scr, 0);    r3do_xform($m, $v);    if ($best['aux'] >> 16) {        $v[2] = $best['aux'] >> 16;    }    $w3 = $v;    $aux = $best['aux'];    $img = $best['img'];    return $best['owner'];}// nothing to draw since the last r3do_landfunction r3do_empty(): int{    return (int) (!R3doVars::$drewland && R3doVars::$maxbucket < 0);}// PORT: one write of the whole block (rows of fill + the untouched bytes between them): mem_write// copies the memory string once instead of once per row.function r3do_fogfill(int $col, int $x1, int $y1, int $x2, int $y2): void{    if ($x2 < $x1 || $y2 < $y1) {        return;    }    $W3 = Alis::$platform->width * 3;    $n3 = ($x2 - $x1 + 1) * 3;    $fill = str_repeat(chr($col & 0xff) . chr(($col >> 8) & 0xff) . chr(($col >> 16) & 0xff), $x2 - $x1 + 1);    $start = Img::$rgblogic + $y1 * $W3 + $x1 * 3;    $gap = $W3 - $n3;    $s = '';    for ($yy = $y1; $yy <= $y2; $yy++) {        $s .= $fill;        if ($yy < $y2 && $gap > 0) {            $s .= substr(Alis::$mem, $start + ($yy - $y1) * $W3 + $n3, $gap);        }    }    mem_write($start, $s);}function r3do_flush(int $scr): void{    if (!Img::$rgblogic || !r3do_tables()) {        return;    }    $W = Alis::$platform->width;    $H = Alis::$platform->height;    $x1 = xread16($scr + 0x66);    $y1 = xread16($scr + 0x68);    $x2 = xread16($scr + 0x6a);    $y2 = xread16($scr + 0x6c);    if ($x1 < 0) {        $x1 = 0;    }    if ($y1 < 0) {        $y1 = 0;    }    if ($x2 >= $W) {        $x2 = $W - 1;    }    if ($y2 >= $H) {        $y2 = $H - 1;    }    if (R3doVars::$drewland) {        $g = (Alis::$fog3do > 0xff ? 0xff : Alis::$fog3do) >> 3;        r3do_fogfill(r3do_rgb5($g, $g, $g), $x1, $y1, $x2, $y2);    }    for ($b = R3doVars::$maxbucket; $b >= 0; $b--) {        $list = R3doVars::$bucket[$b] ?? [];        for ($i = count($list) - 1; $i >= 0; $i--) {            r3do_drawcel($list[$i], $x1, $y1, $x2, $y2);        }    }}