+#include <math.h>
#include <stdlib.h>
+#include <string.h>
+#include "config.h"
#include "common.h"
+#include "globals.h"
#include "sprite.h"
+#include "rocks.h"
-void
-load_sprite(Sprite *sprite, char *filename)
-{
- struct base_sprite *spr = &sprite->sprite;
- spr->image = load_image(filename);
- if(!spr->image) return;
- if(!spr->shape) {
- spr->shape = malloc(sizeof(struct shape));
- if(!spr->shape) {
- fprintf(stderr, "load_sprite(): can't allocate shape structure.\n");
- exit(1);
- }
- get_shape(spr->image, spr->shape);
- }
-}
+SDL_Surface *load_image(char *filename);
+void load_ship(void);
-void
-get_shape(SDL_Surface *img, struct shape *s)
+// 2 sets of sprites, sorted by position
+static Sprite **sprites[2] = { NULL, NULL };
+
+// which set are we using?
+static int set = 0;
+
+// size of squares into which sprites are sorted.
+static int grid_size = 0;
+
+// screen size in grid squares.
+static int gw = 0, gh = 0;
+
+// lists of free sprites, by type.
+Sprite *free_sprites[N_TYPES];
+
+
+static void
+get_shape(Sprite *s)
{
int x, y;
uint16_t *px, transp;
uint32_t bits = 0, bit, *p;
- if(img->format->BytesPerPixel != 2) {
+ s->area = 0;
+ if(s->image->format->BytesPerPixel != 2) {
fprintf(stderr, "get_shape(): not a 16-bit image!\n");
exit(1);
}
- s->area = 0;
- s->w = img->w; s->h = img->h;
- s->mw = ((img->w+31)>>5);
- s->mask = malloc(4*s->mw*s->h);
+ s->w = s->image->w; s->h = s->image->h;
+ grid_size = max(grid_size, max(s->w, s->h));
+ s->mask_w = ((s->w+31)>>5);
+ s->mask = malloc(s->mask_w*s->h*sizeof(uint32_t));
if(!s->mask) {
fprintf(stderr, "get_shape(): can't allocate bitmask.\n");
exit(1);
}
- SDL_LockSurface(img);
- px = img->pixels;
- transp = img->format->colorkey;
+ SDL_LockSurface(s->image);
+ px = s->image->pixels;
+ transp = s->image->format->colorkey;
p = s->mask;
- for(y=0; y<img->h; y++) {
+ for(y=0; y<s->image->h; y++) {
bit = 0;
- for(x=0; x<img->w; x++) {
+ for(x=0; x<s->image->w; x++) {
if(!bit) { bits = 0; bit = 0x80000000; }
if(*px++ != transp) { bits |= bit; s->area++; }
bit >>= 1;
- if(!bit || x == img->w - 1) { *(p++) = bits; }
+ if(!bit || x == s->image->w - 1) { *(p++) = bits; }
}
- px = (uint16_t *) ((uint8_t *) px + img->pitch - 2*img->w);
+ px = (uint16_t *) ((uint8_t *) px + s->image->pitch - 2*s->image->w);
}
- SDL_UnlockSurface(img);
+ SDL_UnlockSurface(s->image);
}
-static int
-line_collide(int xov, struct shape *r, uint32_t *rbits, struct shape *s, uint32_t *sbits)
+
+void
+load_sprite(Sprite *s, char *filename)
{
- int lshift, n, i, ret = 0;
- uint32_t lbits;
- struct shape *st;
- uint32_t *bt;
+ s->image = load_image(filename);
+ if(s->image) get_shape(s);
+}
+
+
+static void
+load_sprites(void)
+{
+ load_ship();
+ load_rocks();
+}
- if(xov < 0) {
- st = r; r = s; s = st;
- bt = rbits; rbits = sbits; sbits = bt;
- xov = -xov;
+void
+init_sprites(void)
+{
+ load_sprites();
+
+ grid_size = grid_size * 3 / 2;
+ gw = (XSIZE-1 + 2*grid_size) / grid_size;
+ gh = (YSIZE-1 + 2*grid_size) / grid_size;
+
+ sprites[0] = malloc(2 * gw * gh * sizeof(Sprite *));
+ sprites[1] = (void *)sprites[0] + gw * gh * sizeof(Sprite *);
+ if(!sprites[0]) {
+ fprintf(stderr, "init_sprites(): can't allocate grid squares.\n");
+ exit(1);
}
+ memset(sprites[0], 0, 2 * gw * gh * sizeof(Sprite *));
+ set = 0;
+}
+static inline Sprite **
+square(int x, int y, int set)
+{
+ int b = (x+grid_size)/grid_size + gw*((y+grid_size)/grid_size);
+ return &sprites[set][b];
+}
+
+void
+add_sprite(Sprite *s)
+{
+ insert_sprite(square(s->x, s->y, set), s);
+}
+
+void
+reset_sprites(void)
+{
+ int i;
- lshift = (r->w - xov) & 31;
- rbits += (r->w - xov) >> 5;
- n = (xov + 31) >> 5;
- for(i=0; i<n-1; i++) {
- lbits = *rbits++ << lshift;
- lbits |= *rbits >> (32 - lshift);
- if(lbits & *sbits++) ret = 1;
+ for(i=0; i<gw*gh; i++)
+ while(sprites[set][i]) {
+ Sprite *s = remove_sprite(&sprites[set][i]);
+ insert_sprite(&free_sprites[s->type], s);
+ s->flags = 0;
+ }
+}
+
+void
+move_sprite(Sprite *s)
+{
+ if(s->flags & MOVE) {
+ s->x += (s->dx - screendx)*t_frame;
+ s->y += (s->dy - screendy)*t_frame;
}
- lbits = *rbits << lshift;
- if(lbits & *sbits) ret = 1;
+}
- return ret;
+void
+sort_sprite(Sprite *s)
+{
+ // clip it, or sort it into the other set of sprites.
+ if(s->x + s->w < 0 || s->x >= XSIZE
+ || s->y + s->h < 0 || s->y >= YSIZE) {
+ insert_sprite(&free_sprites[s->type], s);
+ s->flags = 0;
+ } else insert_sprite(square(s->x, s->y, 1-set), s);
}
+void
+move_sprites(void)
+{
+ int sq;
+ Sprite **head;
+
+ // Move all the sprites
+ for(sq=0; sq<gw*gh; sq++) {
+ head=&sprites[set][sq];
+ while(*head) {
+ Sprite *s = remove_sprite(head);
+ move_sprite(s); sort_sprite(s);
+ }
+ }
+ set = 1-set; // switch to other set of sprites.
+}
+
+
static int
-mask_collide(int xov, int yov, struct shape *r, struct shape *s)
+line_collide(int xov, unsigned bit, uint32_t *amask, uint32_t *bmask)
{
- int y, ry, sy;
- uint32_t *rbits, *sbits;
+ int i, words = (xov-1) >> 5;
+ uint32_t abits;
+
+ for(i=0; i<words; i++) {
+ abits = *amask++ << bit;
+ abits |= *amask >> (32-bit);
+ if(abits & *bmask++) return true;
+ }
+ abits = *amask << bit;
+ if(abits & *bmask) return true;
+
+ return false;
+}
+
+static int
+mask_collide(int xov, int yov, Sprite *a, Sprite *b)
+{
+ int y;
+ int xoffset = a->w - xov;
+ int word = xoffset >> 5, bit = xoffset & 31;
+ uint32_t *amask = a->mask, *bmask = b->mask;
if(yov > 0) {
- ry = r->h - yov; sy = 0;
- rbits = r->mask + (r->h - yov) * r->mw;
- sbits = s->mask;
+ amask = a->mask + ((a->h - yov) * a->mask_w) + word;
+ bmask = b->mask;
} else {
- ry = 0; sy = s->h + yov;
- rbits = r->mask;
- sbits = s->mask + (s->h + yov) * s->mw;
+ yov = -yov;
+ amask = a->mask;
+ bmask = b->mask + ((b->h - yov) * b->mask_w) + word;
}
- for(y=0; y<abs(yov); y++) {
- if(line_collide(xov, r, rbits, s, sbits)) return 1;
- rbits += r->mw; sbits += s->mw;
+ for(y=0; y<yov; y++) {
+ if(line_collide(xov, bit, amask, bmask)) return 1;
+ amask += a->mask_w; bmask += b->mask_w;
}
return 0;
}
int
-collide(Sprite *r, Sprite *s)
+collide(Sprite *a, Sprite *b)
{
- struct shape *rs = r->sprite.shape;
- struct shape *ss = s->sprite.shape;
- int dx = s->sprite.x - r->sprite.x;
- int dy = s->sprite.y - r->sprite.y;
- int xov, yov;
+ int dx, dy, xov, yov;
+
+ if(!COLLIDES(a) || !COLLIDES(b)) return false;
+
+ if(b->x < a->x) { Sprite *tmp = a; a = b; b = tmp; }
- if(dx >= 0) xov = max(min(rs->w - dx, ss->w), 0);
- else xov = -max(min(ss->w + dx, rs->w), 0);
+ dx = b->x - a->x;
+ dy = b->y - a->y;
- if(dy >= 0) yov = max(min(rs->h - dy, ss->h), 0);
- else yov = -max(min(ss->h + dy, rs->h), 0);
+ xov = max(min(a->w - dx, b->w), 0);
+
+ if(dy >= 0) yov = max(min(a->h - dy, b->h), 0);
+ else yov = -max(min(b->h - -dy, a->h), 0);
if(xov == 0 || yov == 0) return false;
- else return mask_collide(xov, yov, rs, ss);
+ else return mask_collide(xov, yov, a, b);
+}
+
+void
+collide_with_list(Sprite *s, Sprite *list)
+{
+ for(; list; list=list->next)
+ if(collide(s, list)) do_collision(s, list);
+}
+
+void
+collisions(void)
+{
+ int i, end = gw*gh;
+ Sprite *s;
+ for(i=0; i<end; i++) {
+ for(s=sprites[set][i]; s; s=s->next) {
+ collide_with_list(s, s->next);
+ if(i+1 < end) collide_with_list(s, sprites[set][i+1]);
+ if(i+gw < end) collide_with_list(s, sprites[set][i+gw]);
+ if(i+gw+1 < end) collide_with_list(s, sprites[set][i+gw+1]);
+ }
+ }
+}
+
+Sprite *
+hit_in_square(Sprite *r, Sprite *s)
+{
+ for(; r; r=r->next)
+ if(collide(r, s)) break;
+ return r;
+}
+
+Sprite *
+collides(Sprite *s)
+{
+ int l, r, t, b;
+ Sprite **sq;
+ Sprite *c;
+
+ l = (s->x + grid_size) / grid_size;
+ r = (s->x + s->w + grid_size) / grid_size;
+ t = (s->y + grid_size) / grid_size;
+ b = (s->y + s->h + grid_size) / grid_size;
+ sq = &sprites[set][l + t*gw];
+
+ if((c = hit_in_square(*sq, s))) return c;
+ if(l > 0 && (c = hit_in_square(*(sq-1), s))) return c;
+ if(t > 0 && (c = hit_in_square(*(sq-gw), s))) return c;
+ if(l > 0 && t > 0 && (c = hit_in_square(*(sq-1-gw), s))) return c;
+
+ if(r > l) {
+ if((c = hit_in_square(*(sq+1), s))) return c;
+ if(t > 0 && hit_in_square(*(sq+1-gw), s)) return c;
+ }
+ if(b > t) {
+ if((c = hit_in_square(*(sq+gw), s))) return c;
+ if(l > 0 && (c = hit_in_square(*(sq-1+gw), s))) return c;
+ }
+ if(r > l && b > t && (c = hit_in_square(*(sq+1+gw), s))) return c;
+ return NULL;
}
int
-pixel_collide(unsigned int xoff, unsigned int yoff, struct shape *r)
+pixel_collide(Sprite *s, int x, int y)
{
uint32_t pmask;
+
+ if(!COLLIDES(s)) return false;
- if(xoff >= r->w || yoff >= r->h) return 0;
+ if(x < s->x || y < s->y || x >= s->x + s->w || y >= s->y + s->h) return 0;
+
+ x -= s->x; y -= s->y;
+ pmask = 0x80000000 >> (x&0x1f);
+ return s->mask[(y*s->mask_w) + (x>>5)] & pmask;
+}
+
+Sprite *
+pixel_hit_in_square(Sprite *r, float x, float y)
+{
+ for(; r; r=r->next) {
+ if(COLLIDES(r) && pixel_collide(r, x, y)) return r;
+ }
+ return 0;
+}
+
+Sprite *
+pixel_collides(float x, float y)
+{
+ int l, t;
+ Sprite **sq;
+ Sprite *ret;
+
+ l = (x + grid_size) / grid_size; t = (y + grid_size) / grid_size;
+ sq = &sprites[set][l + t*gw];
+ if((ret = pixel_hit_in_square(*sq, x, y))) return ret;
+ if(l > 0 && (ret = pixel_hit_in_square(*(sq-1), x, y))) return ret;
+ if(t > 0 && (ret = pixel_hit_in_square(*(sq-gw), x, y))) return ret;
+ if(l > 0 && t > 0 && (ret = pixel_hit_in_square(*(sq-1-gw), x, y))) return ret;
+ return 0;
+}
+
+
+float
+sprite_mass(Sprite *s)
+{
+ if(s->type == SHIP) return s->area;
+ else if(s->type == ROCK) return 3 * s->area;
+ else return 0;
+}
+
+/*
+ * BOUNCE THEORY
+ *
+ * ****************** In 1 Dimension *****************
+ *
+ * For now we will imagine bouncing A and B off each other in 1 dimension (along
+ * a line). We can safely save the other dimension for later.
+ *
+ * A and B are the same weight, and are both traveling 1m/sec, to collide right
+ * at the origin. With perfect bounciness, their full momentum is reversed.
+ *
+ * If we cut the weight of A down by half, then the center of our colision will
+ * drift towards A (the speeds of A and B are not simply reversed as in our last
+ * example.) However, there is always a place between A and B on the line (I'll
+ * call it x) such that the speeds of A and B relative to x, are simply
+ * reversed. Thus we can find the new speed for A like so:
+ *
+ * new A = x -(A - x)
+ *
+ * new B = x -(B - x)
+ *
+ * or, simply:
+ *
+ * new A = 2x - A
+ *
+ * new B = 2x - B
+ *
+ *
+ * this point x is the sort of center of momentum. If, instead of bouncing, A
+ * and B just globbed together, x would be center of the new glob.
+ *
+ * x is the point where there's an equal amount of force coming in from both
+ * sides. ie the weighted average of the speeds of A and B.
+ *
+ * average force = (A force + B force) / total mass
+ *
+ * x.speed = (a.speed * a.mass + b.speed * b.mass) / (a.mass + b.mas)
+ *
+ * then we apply the formula above for calculating the new A and B.
+ *
+ *
+ *
+ *
+ * ****************** In 2 Dimensions *****************
+ *
+ * OK, that's how we do it in 1D. Now we need to deal with 2D.
+ *
+ * Imagine (or draw) the two balls just as they are bouncing off each other.
+ * Imagine drawing a line through the centers of the balls. The balls are
+ * exerting force on each other only along this axis. So if we rotate
+ * everything, we can do our earlier 1D math along this line.
+ *
+ * It doesn't matter what direction the balls are going in, they only exert
+ * force on each other along this line. What we will do is to compute the part
+ * of the balls' momentum that is going along this line, and bounce it according
+ * to our math above. The other part is unaffected by the bounce, and we can
+ * just leave it alone.
+ *
+ * To get this component of the balls' momentum, we can use the dot product.
+ *
+ * dot(U, V) = length(U) * length(V) * cos(angle between U and V)
+ *
+ * If U is a length 1 vector, then dot(U, V) is the length of the component of V
+ * in the direction of U. So the components of V are:
+ *
+ * U * dot(U, V) parallel to U
+ *
+ * V - U * dot(U, V) perpendicular to U
+ *
+ * To do the actual bounce, we compute the unit vector between the center of the
+ * two balls, compute the components of the balls' speeds along this vector (A
+ * and B), and then bounce them according to the math above:
+ *
+ * new A = 2x - A
+ *
+ * new B = 2x - B
+ *
+ * But we rewrite it in relative terms:
+ *
+ * new A = A + 2(x-A)
+ *
+ * new B = B + 2(x-B)
+ */
+
+void
+bounce(Sprite *a, Sprite *b)
+{
+ float x, y, n; // (x, y) is unit vector from a to b.
+ float va, vb; // va, vb are balls' speeds along (x, y)
+ float ma, mb; // ma, mb are the balls' masses.
+ float vc; // vc is the "center of momentum"
+
+ // (x, y) is unit vector pointing from A's center to B's center.
+ x = (b->x + b->w / 2) - (a->x + a->w / 2);
+ y = (b->y + b->h / 2) - (a->y + a->h / 2);
+ n = sqrt(x*x + y*y); x /= n; y /= n;
+
+ // velocities along (x, y)
+ va = x*a->dx + y*a->dy;
+ vb = x*b->dx + y*b->dy;
+ if(vb-va > 0) return; // don't bounce if we're already moving away.
+
+ // get masses and compute "center" speed
+ ma = sprite_mass(a); mb = sprite_mass(b);
+ vc = (va*ma + vb*mb) / (ma+mb);
- pmask = 0x80000000 >> (xoff&0x1f);
- return r->mask[(yoff*r->mw) + (xoff>>5)] & pmask;
+ // bounce off the center speed.
+ a->dx += 2*x*(vc-va); a->dy += 2*y*(vc-va);
+ b->dx += 2*x*(vc-vb); b->dy += 2*y*(vc-vb);
}
Got questions, comments, patches, etc.? Contact Jason Woofenden