/* ScummVM - Graphic Adventure Engine * * ScummVM is the legal property of its developers, whose names * are too numerous to list here. Please refer to the COPYRIGHT * file distributed with this source distribution. * * 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. * * $URL$ * $Id$ * */ #include "sci/include/sciresource.h" #include "sci/include/engine.h" namespace Sci { /* Compute "velocity" vector (xStep,yStep)=(vx,vy) for a jump from (0,0) to (dx,dy), with gravity gy. The gravity is assumed to be non-negative. If this was ordinary continuous physics, we would compute the desired (floating point!) velocity vector (vx,vy) as follows, under the assumption that vx and vy are linearly correlated by some constant factor c, i.e. vy = c * vx: dx = t * vx dy = t * vy + gy * t^2 / 2 => dy = c * dx + gy * (dx/vx)^2 / 2 => |vx| = sqrt( gy * dx^2 / (2 * (dy - c * dx)) ) Here, the sign of vx must be chosen equal to the sign of dx, obviously. Clearly, this square root only makes sense in our context if the denominator is positive, or equivalently, (dy - c * dx) must be positive. For simplicity and by symmetry along the x-axis, we assume dx to be positive for all computations, and only adjust for its sign in the end. Switching the sign of c appropriately, we set tmp := (dy + c * dx) and compute c so that this term becomes positive. Remark #1: If the jump is straight up, i.e. dx == 0, then we should not assume the above linear correlation vy = c * vx of the velocities (as vx will be 0, but vy shouldn't be, unless we drop). Remark #2: We are actually in a discrete setup. The motion is computed iteratively: each iteration, we add vx and vy to the position, then add gy to vy. So the real formula is the following (where t is ideally close to an int): dx = t * vx dy = t * vy + gy * t*(t-1) / 2 But the solution resulting from that is a lot more complicated, so we use the above approximation instead. Still, what we compute in the end is of course not a real velocity anymore, but an integer approximation, used in an iterative stepping algorithm */ reg_t kSetJump(EngineState *s, int funct_nr, int argc, reg_t *argv) { // Input data reg_t object = argv[0]; int dx = SKPV(1); int dy = SKPV(2); int gy = SKPV(3); // Derived data int c; int tmp; int vx = 0; // x velocity int vy = 0; // y velocity int dxWasNegative = (dx < 0); dx = abs(dx); assert(gy >= 0); if (dx == 0) { // Upward jump. Value of c doesn't really matter c = 1; } else { // Compute a suitable value for c respectively tmp. // The important thing to consider here is that we want the resulting // *discrete* x/y velocities to be not-too-big integers, for a smooth // curve (i.e. we could just set vx=dx, vy=dy, and be done, but that // is hardly what you would call a parabolic jump, would ya? ;-). // // So, we make sure that 2.0*tmp will be bigger than dx (that way, // we ensure vx will be less than sqrt(gy * dx)). if (dx + dy < 0) { // dy is negative and |dy| > |dx| c = (2 * abs(dy)) / dx; //tmp = abs(dy); // ALMOST the resulting value, except for obvious rounding issues } else { // dy is either positive, or |dy| <= |dx| c = (dx * 3 / 2 - dy) / dx; // We force c to be strictly positive if (c < 1) c = 1; //tmp = dx * 3 / 2; // ALMOST the resulting value, except for obvious rounding issues // FIXME: Where is the 3 coming from? Maybe they hard/coded, by "accident", that usually gy=3 ? // Then this choice of will make t equal to roughly sqrt(dx) } } // POST: c >= 1 tmp = c * dx + dy; // POST: (dx != 0) ==> abs(tmp) > abs(dx) // POST: (dx != 0) ==> abs(tmp) ~>=~ abs(dy) SCIkdebug(SCIkBRESEN, "c: %d, tmp: %d\n", c, tmp); // Compute x step if (tmp != 0) vx = (int)(dx * sqrt(gy / (2.0 * tmp))); else vx = 0; // Restore the left/right direction: dx and vx should have the same sign. if (dxWasNegative) vx = -vx; if ((dy < 0) && (vx == 0)) { // Special case: If this was a jump (almost) straight upward, i.e. dy < 0 (upward), // and vx == 0 (i.e. no horizontal movement, at least not after rounding), then we // compute vy directly. // For this, we drop the assumption on the linear correlation of vx and vy (obviously). // FIXME: This choice of vy makes t roughly (2+sqrt(2))/gy * sqrt(dy); // so if gy==3, then t is roughly sqrt(dy)... vy = (int)sqrt((double)gy * abs(2 * dy)) + 1; } else { // As stated above, the vertical direction is correlated to the horizontal by the // (non-zero) factor c. // Strictly speaking, we should probably be using the value of vx *before* rounding // it to an integer... Ah well vy = c * vx; } // Always force vy to be upwards vy = -abs(vy); SCIkdebug(SCIkBRESEN, "SetJump for object at "PREG"\n", PRINT_REG(object)); SCIkdebug(SCIkBRESEN, "xStep: %d, yStep: %d\n", vx, vy); PUT_SEL32V(object, xStep, vx); PUT_SEL32V(object, yStep, vy); return s->r_acc; } #define _K_BRESEN_AXIS_X 0 #define _K_BRESEN_AXIS_Y 1 void initialize_bresen(EngineState *s, int funct_nr, int argc, reg_t *argv, reg_t mover, int step_factor, int deltax, int deltay) { reg_t client = GET_SEL32(mover, client); int stepx = GET_SEL32SV(client, xStep) * step_factor; int stepy = GET_SEL32SV(client, yStep) * step_factor; int numsteps_x = stepx ? (abs(deltax) + stepx - 1) / stepx : 0; int numsteps_y = stepy ? (abs(deltay) + stepy - 1) / stepy : 0; int bdi, i1; int numsteps; int deltax_step; int deltay_step; if (numsteps_x > numsteps_y) { numsteps = numsteps_x; deltax_step = (deltax < 0) ? -stepx : stepx; deltay_step = numsteps ? deltay / numsteps : deltay; } else { // numsteps_x <= numsteps_y numsteps = numsteps_y; deltay_step = (deltay < 0) ? -stepy : stepy; deltax_step = numsteps ? deltax / numsteps : deltax; } /* if (abs(deltax) > abs(deltay)) {*/ // Bresenham on y if (numsteps_y < numsteps_x) { PUT_SEL32V(mover, b_xAxis, _K_BRESEN_AXIS_Y); PUT_SEL32V(mover, b_incr, (deltay < 0) ? -1 : 1); //i1 = 2 * (abs(deltay) - abs(deltay_step * numsteps)) * abs(deltax_step); //bdi = -abs(deltax); i1 = 2 * (abs(deltay) - abs(deltay_step * (numsteps - 1))) * abs(deltax_step); bdi = -abs(deltax); } else { // Bresenham on x PUT_SEL32V(mover, b_xAxis, _K_BRESEN_AXIS_X); PUT_SEL32V(mover, b_incr, (deltax < 0) ? -1 : 1); //i1= 2 * (abs(deltax) - abs(deltax_step * numsteps)) * abs(deltay_step); //bdi = -abs(deltay); i1 = 2 * (abs(deltax) - abs(deltax_step * (numsteps - 1))) * abs(deltay_step); bdi = -abs(deltay); } PUT_SEL32V(mover, dx, deltax_step); PUT_SEL32V(mover, dy, deltay_step); SCIkdebug(SCIkBRESEN, "Init bresen for mover "PREG": d=(%d,%d)\n", PRINT_REG(mover), deltax, deltay); SCIkdebug(SCIkBRESEN, " steps=%d, mv=(%d, %d), i1= %d, i2=%d\n", numsteps, deltax_step, deltay_step, i1, bdi*2); //PUT_SEL32V(mover, b_movCnt, numsteps); // Needed for HQ1/Ogre? PUT_SEL32V(mover, b_di, bdi); PUT_SEL32V(mover, b_i1, i1); PUT_SEL32V(mover, b_i2, bdi * 2); } reg_t kInitBresen(EngineState *s, int funct_nr, int argc, reg_t *argv) { reg_t mover = argv[0]; reg_t client = GET_SEL32(mover, client); int deltax = GET_SEL32SV(mover, x) - GET_SEL32SV(client, x); int deltay = GET_SEL32SV(mover, y) - GET_SEL32SV(client, y); initialize_bresen(s, funct_nr, argc, argv, mover, KP_UINT(KP_ALT(1, make_reg(0, 1))), deltax, deltay); return s->r_acc; } #define MOVING_ON_X (((axis == _K_BRESEN_AXIS_X)&&bi1) || dx) #define MOVING_ON_Y (((axis == _K_BRESEN_AXIS_Y)&&bi1) || dy) static enum { IGNORE_MOVECNT, INCREMENT_MOVECNT, UNINITIALIZED } handle_movecnt = UNINITIALIZED; int parse_reg_t(EngineState *s, const char *str, reg_t *dest); static int checksum_bytes(byte *data, int size) { int result = 0; int i; for (i = 0; i < size; i++) { result += *data; data++; } return result; } static void bresenham_autodetect(EngineState *s) { reg_t motion_class; if (!parse_reg_t(s, "?Motion", &motion_class)) { object_t *obj = obj_get(s, motion_class); reg_t fptr; byte *buf; if (obj == NULL) { warning("bresenham_autodetect failed"); handle_movecnt = INCREMENT_MOVECNT; // Most games do this, so best guess return; } if (lookup_selector(s, motion_class, s->selector_map.doit, NULL, &fptr) != SELECTOR_METHOD) { warning("bresenham_autodetect failed"); handle_movecnt = INCREMENT_MOVECNT; // Most games do this, so best guess return; } buf = s->seg_manager.heap[fptr.segment]->data.script.buf + fptr.offset; handle_movecnt = (SCI_VERSION_MAJOR(s->version) == 0 || checksum_bytes(buf, 8) == 0x216) ? INCREMENT_MOVECNT : IGNORE_MOVECNT; sciprintf("b-moveCnt action based on checksum: %s\n", handle_movecnt == IGNORE_MOVECNT ? "ignore" : "increment"); } else { warning("bresenham_autodetect failed"); handle_movecnt = INCREMENT_MOVECNT; // Most games do this, so best guess } } reg_t kDoBresen(EngineState *s, int funct_nr, int argc, reg_t *argv) { reg_t mover = argv[0]; reg_t client = GET_SEL32(mover, client); int x = GET_SEL32SV(client, x); int y = GET_SEL32SV(client, y); int oldx, oldy, destx, desty, dx, dy, bdi, bi1, bi2, movcnt, bdelta, axis; word signal = GET_SEL32V(client, signal); int completed = 0; int max_movcnt = GET_SEL32V(client, moveSpeed); if (SCI_VERSION_MAJOR(s->version) > 0) signal &= ~_K_VIEW_SIG_FLAG_HIT_OBSTACLE; if (handle_movecnt == UNINITIALIZED) bresenham_autodetect(s); PUT_SEL32(client, signal, make_reg(0, signal)); // This is a NOP for SCI0 oldx = x; oldy = y; destx = GET_SEL32SV(mover, x); desty = GET_SEL32SV(mover, y); dx = GET_SEL32SV(mover, dx); dy = GET_SEL32SV(mover, dy); bdi = GET_SEL32SV(mover, b_di); bi1 = GET_SEL32SV(mover, b_i1); bi2 = GET_SEL32SV(mover, b_i2); movcnt = GET_SEL32V(mover, b_movCnt); bdelta = GET_SEL32SV(mover, b_incr); axis = GET_SEL32SV(mover, b_xAxis); //sciprintf("movecnt %d, move speed %d\n", movcnt, max_movcnt); if (handle_movecnt) { if (max_movcnt > movcnt) { ++movcnt; PUT_SEL32V(mover, b_movCnt, movcnt); // Needed for HQ1/Ogre? return NULL_REG; } else { movcnt = 0; PUT_SEL32V(mover, b_movCnt, movcnt); // Needed for HQ1/Ogre? } } if ((bdi += bi1) > 0) { bdi += bi2; if (axis == _K_BRESEN_AXIS_X) dx += bdelta; else dy += bdelta; } PUT_SEL32V(mover, b_di, bdi); x += dx; y += dy; if ((MOVING_ON_X && (((x < destx) && (oldx >= destx)) // Moving left, exceeded? || ((x > destx) && (oldx <= destx)) // Moving right, exceeded? || ((x == destx) && (abs(dx) > abs(dy))) // Moving fast, reached? // Treat this last case specially- when doing sub-pixel movements // on the other axis, we could still be far away from the destination )) || (MOVING_ON_Y && (((y < desty) && (oldy >= desty)) /* Moving upwards, exceeded? */ || ((y > desty) && (oldy <= desty)) /* Moving downwards, exceeded? */ || ((y == desty) && (abs(dy) >= abs(dx))) /* Moving fast, reached? */ ))) { // Whew... in short: If we have reached or passed our target position x = destx; y = desty; completed = 1; SCIkdebug(SCIkBRESEN, "Finished mover "PREG"\n", PRINT_REG(mover)); } PUT_SEL32V(client, x, x); PUT_SEL32V(client, y, y); SCIkdebug(SCIkBRESEN, "New data: (x,y)=(%d,%d), di=%d\n", x, y, bdi); if (s->version >= SCI_VERSION_FTU_INVERSE_CANBEHERE) invoke_selector(INV_SEL(client, cantBeHere, 0), 0); else invoke_selector(INV_SEL(client, canBeHere, 0), 0); s->r_acc = not_register(s, s->r_acc); if (!s->r_acc.offset) { // Contains the return value signal = GET_SEL32V(client, signal); PUT_SEL32V(client, x, oldx); PUT_SEL32V(client, y, oldy); PUT_SEL32V(client, signal, (signal | _K_VIEW_SIG_FLAG_HIT_OBSTACLE)); SCIkdebug(SCIkBRESEN, "Finished mover "PREG" by collision\n", PRINT_REG(mover)); completed = 1; } if (SCI_VERSION_MAJOR(s->version) > 0) if (completed) invoke_selector(INV_SEL(mover, moveDone, 0), 0); return make_reg(0, completed); } extern void _k_dirloop(reg_t obj, word angle, EngineState *s, int funct_nr, int argc, reg_t *argv); int is_heap_object(EngineState *s, reg_t pos); extern int get_angle(int xrel, int yrel); reg_t kDoAvoider(EngineState *s, int funct_nr, int argc, reg_t *argv) { reg_t avoider = argv[0]; reg_t client, looper, mover; int angle; int dx, dy; int destx, desty; s->r_acc = make_reg(0, -1); if (!is_heap_object(s, avoider)) { warning("DoAvoider() where avoider "PREG" is not an object", PRINT_REG(avoider)); return NULL_REG; } client = GET_SEL32(avoider, client); if (!is_heap_object(s, client)) { warning("DoAvoider() where client "PREG" is not an object", PRINT_REG(client)); return NULL_REG; } looper = GET_SEL32(client, looper); mover = GET_SEL32(client, mover); if (!is_heap_object(s, mover)) { if (mover.segment) { warning("DoAvoider() where mover "PREG" is not an object", PRINT_REG(mover)); } return s->r_acc; } destx = GET_SEL32V(mover, x); desty = GET_SEL32V(mover, y); SCIkdebug(SCIkBRESEN, "Doing avoider %04x (dest=%d,%d)\n", avoider, destx, desty); if (invoke_selector(INV_SEL(mover, doit, 1) , 0)) { error("Mover "PREG" of avoider "PREG" doesn't have a doit() funcselector\n", PRINT_REG(mover), PRINT_REG(avoider)); return NULL_REG; } mover = GET_SEL32(client, mover); if (!mover.segment) // Mover has been disposed? return s->r_acc; // Return gracefully. if (invoke_selector(INV_SEL(client, isBlocked, 1) , 0)) { error("Client "PREG" of avoider "PREG" doesn't" " have an isBlocked() funcselector\n", PRINT_REG(client), PRINT_REG(avoider)); return NULL_REG; } dx = destx - GET_SEL32V(client, x); dy = desty - GET_SEL32V(client, y); angle = get_angle(dx, dy); SCIkdebug(SCIkBRESEN, "Movement (%d,%d), angle %d is %sblocked\n", dx, dy, angle, (s->r_acc.offset) ? " " : "not "); if (s->r_acc.offset) { // isBlocked() returned non-zero int rotation = (rand() & 1) ? 45 : (360 - 45); // Clockwise/counterclockwise int oldx = GET_SEL32V(client, x); int oldy = GET_SEL32V(client, y); int xstep = GET_SEL32V(client, xStep); int ystep = GET_SEL32V(client, yStep); int moves; SCIkdebug(SCIkBRESEN, " avoider "PREG"\n", PRINT_REG(avoider)); for (moves = 0; moves < 8; moves++) { int move_x = (int)(sin(angle * PI / 180.0) * (xstep)); int move_y = (int)(-cos(angle * PI / 180.0) * (ystep)); PUT_SEL32V(client, x, oldx + move_x); PUT_SEL32V(client, y, oldy + move_y); SCIkdebug(SCIkBRESEN, "Pos (%d,%d): Trying angle %d; delta=(%d,%d)\n", oldx, oldy, angle, move_x, move_y); if (invoke_selector(INV_SEL(client, canBeHere, 1) , 0)) { error("Client "PREG" of avoider "PREG" doesn't" " have a canBeHere() funcselector\n", PRINT_REG(client), PRINT_REG(avoider)); return NULL_REG; } PUT_SEL32V(client, x, oldx); PUT_SEL32V(client, y, oldy); if (s->r_acc.offset) { // We can be here SCIkdebug(SCIkBRESEN, "Success\n"); PUT_SEL32V(client, heading, angle); return make_reg(0, angle); } angle += rotation; if (angle > 360) angle -= 360; } warning("DoAvoider failed for avoider "PREG"", PRINT_REG(avoider)); } else { int heading = GET_SEL32V(client, heading); if (heading == -1) return s->r_acc; // No change PUT_SEL32V(client, heading, angle); s->r_acc = make_reg(0, angle); if (looper.segment) { if (invoke_selector(INV_SEL(looper, doit, 1), 2, angle, client)) { error("Looper "PREG" of avoider "PREG" doesn't" " have a doit() funcselector\n", PRINT_REG(looper), PRINT_REG(avoider)); } else return s->r_acc; } else { // No looper? Fall back to DirLoop _k_dirloop(client, (word)angle, s, funct_nr, argc, argv); } } return s->r_acc; } } // End of namespace Sci