#include <stdlib.h>
#include <string.h>
#include "DrawSpan.h"
#include "polygon_pack.h"
#include "task_texture.h"
#include "viewer_types.h"
#include "Func.h"
#include "sys.h"
#include "SchedTask.h"
#include "Tapestry.h"
#include "SpanPack.h"
#include <spu_intrinsics.h>
#include <math.h>

#if (__LITTLE_ENDIAN__)
#define LITTLEENDIAN 1
#else
#define LITTLEENDIAN 0
#endif

SchedDefineTask(DrawSpan);

#define TEX_LOAD1      0
#define TEX_LOAD2      1
#define SPAN_PACK_LOAD 2
#define FB_STORE       3

typedef struct {
    int **linebuf;
    float *zRow;
    TileListPtr tileList;
    int doneWrite;
} G, *Gptr; 

static    int** linebuf_init(SchedTask *smanager, int width, int height, int rgb);
static    float* zRow_init(SchedTask *smanager, int width, int height);
// static    TilePtr set_rgb(memaddr addr, int wait_tag);
// static    void set_rgbs(memaddr addr, uint32 *max_addr, int wait_tag);
static    uint32 get_rgb(int tx, int ty, TilePtr tile);
// static    TilePtr isAvailableTile(memaddr addr);
static    memaddr getTile(int tx, int ty, int tw, memaddr tex_addr_top);
static    int getTexBlock(int tx, int ty, int twidth);

static    void updateBuffer(Gptr g, float zpos, int rangex, int loaclx, int localy,
                            int tex_x, int tex_y, float normal_x, float nomral_x, float normal_z,
			    TilePtr tile, int world_x, int world_y, float world_z,SchedTask *smanager);

// static    void reboot(SpanPackPtr spack, int cur_span_x);

static    int drawDot1(SchedTask *smanager, Gptr g, SpanPtr span, int startx, int endx, int wait_tag);
//static    void drawDot2(SchedTask *smanager, SpanPtr span, int startx, int endx, int js, int wait_tag);
static    int drawLine1(SchedTask *smanager, Gptr g, SpanPtr span, int startx, int endx, int wait_tag);
// static    void drawLine2(SchedTask *smanager, SpanPtr span, int startx, int endx, int js, int wait_tag);

static    int infinity_light_calc(int color,float normal_x, float normal_y, float normal_z,
                                  SchedTask *smanager,int x, int y, float z, int world_x, int world_y, float world_z);



static inline void
normalize(vector float& v0, vector float& v1)
{
  float norm;
  vector float ret __attribute__((aligned(16))) = {0,0,0,0};

  ret = spu_mul(v0,v1);
  norm = (ret[0] + ret[1] + ret[2]);

  ret = (vector float)spu_splats(norm);
  ret = spu_rsqrte(ret);
  v0 = spu_mul(v1,ret);
}

static inline float
innerProduct1(vector float& v0, vector float& v1)
{

  vector float ret __attribute__((aligned(16))) = {0,0,0,0};
  float inner;
  ret = spu_mul(v0,v1);

  inner = (ret[0] + ret[1] + ret[2]);
  if (inner < 0) {
    inner = 0;
  }

  return inner;
}




/**
 * テクスチャは、TEXTURE_SPLIT_PIXEL^2 のブロックに分割する
 *
 * +---+---+---+---+---+---+
 * | 0 | 1 | 2 | 3 | 4 | 5 |
 * +---+---+---+---+---+---+
 * |   |   |   |   |   |11 |
 * +---+---+---+---+---+---+
 * |   |   |   |   |   |17 |
 * +---+---+---+---+---+---+
 * |   |   |   |   |   |23 |
 * +---+---+---+---+---+---+
 * |   |   |   |   |   |29 |
 * +---+---+---+---+---+---+
 * |   |   |   |   |   |35 |
 * +---+---+---+---+---+---+
 *
 * 一辺を TEXTURE_SPLIT とする
 * 各ブロックの数字がブロックIDとなる。
 */

/**
 * テクスチャの座標から、
 * テクスチャのどのブロックかを求める
 *
 * @param[in] tx X coordinates of texture
 * @param[in] tx Y coordinates of texture
 * @param[in] twidth  Width of texture
 * @return block ID
 */
static int
getTexBlock(int tx, int ty, int twidth)
{
     int blockX, blockY;

     blockX = tx / TEXTURE_SPLIT_PIXEL;
     blockY = ty / TEXTURE_SPLIT_PIXEL;

     return blockX + (twidth/TEXTURE_SPLIT_PIXEL)*blockY;
}

/**
 * block ID と、テクスチャの TOP address から
 * (tx,ty) で使われるテクスチャの Tile addres を求める
 *
 * @param[in] tx X coordinates of texture
 * @param[in] tx Y coordinates of texture
 * @param[in] tw Width of texture
 * @param[in] tex_addr_top (tx,ty) で使うテクスチャの先頭address
 * @return block ID
 */
static memaddr
getTile(int tx, int ty, int tw, memaddr tex_addr_top)
{
    int block = getTexBlock(tx, ty, tw);
    return tex_addr_top + block * TEXTURE_BLOCK_SIZE * sizeof(uint32);
}

/**
 * FrameBuffer に書き込む rgb の領域初期化
 *
 * @param width  Width of Buffer
 * @param height Height of Buffer
 * @param rgb    Initial value of RGB at Buffer
 * @return Buffer
 */
static int **
linebuf_init(SchedTask *smanager, int width, int height, int rgb)
{
    int **linebuf = (int**)smanager->allocate(height*sizeof(int*));
    for(int h = 0;h<height;h++) {
	int *buf = linebuf[h] = (int*)smanager->get_output(h);
	for (int i = 0; i < width; i++) {
	    buf[i] = rgb;
	}
    }
    return linebuf;
}

/**
 * Z-Buffer の初期化
 *
 * @param width  Width of Z-Buffer
 * @param height Height of Z-Buffer
 * @return Z-Buffer
 */
static float*
zRow_init(SchedTask *smanager, int width, int height)
{
    float *buf = (float*)smanager->allocate(sizeof(float)*width*height);
    float def = 65535.0f;

    for (int i = 0; i < width*height; i++) {
        buf[i] = def;
    }

    return buf;
}


static uint32
get_rgb(int tx, int ty, TilePtr tile)
{
    uint32 *data = (uint32 *)tile->data;
    return data[(TEXTURE_SPLIT_PIXEL)*ty+tx];
}



/**
 * zRow と Linebuf を更新する
 *
 * @param zpos     更新する pixel のZ座標
 * @param rangex   このタスクが処理する描画領域の x の長さ
 * @param x        pixel の、描画領域内での x 座標
 * @param y        〃 の、y 座標
 * @param tex_x    pixel が使用するテクスチャの、Tile (8x8) 内での x 座標
 * @param tex_y    〃 の y 座標
 * @param tex_addr テクスチャのアドレス(MainMemory)
 */
static void
updateBuffer(Gptr g, float zpos, int rangex, int localx, int localy, int tex_x, int tex_y,
	     float normal_x, float normal_y, float normal_z, TilePtr tile,
	     int world_x, int world_y, float world_z, SchedTask *smanager)
{

    uint32 color = get_rgb(tex_x, tex_y, tile);
    /*下位4bitを抽出*/
#if LITTLEENDIAN
    int alpha = color & 0x000000ff;
#else
    int alpha = color & 0xff000000;
#endif

    //完全に透けているか判断
    int flag = (alpha != 0 && normal_z < 0);
    int *light_sysswitch = (int*)smanager->global_get(LightSysSwitch);
    //smanager->printf("sys %d\n",light_sysswitch);

    if ( *light_sysswitch == 1 && flag) {
        color = infinity_light_calc(color,normal_x,normal_y,normal_z,
				    smanager,localx,localy,zpos,
				    world_x,world_y,world_z);
    }

    g->zRow[localx + (rangex*localy)] = zpos*flag + g->zRow[localx + (rangex*localy)]*(1-flag);
    int *point = &g->linebuf[localy][localx];
    *point =  color*flag + *point *(1-flag);

}

/**
 * 長さが 1 の Span の描画 (要するに 1 pixel)
 *
 * @param span Span
 * @param startx 描画開始範囲
 * @param endx 描画終了範囲
 */
static int
drawDot1(SchedTask *smanager, Gptr g, SpanPtr span, int startx, int endx, int wait_tag)
{
    int rangex = endx - startx + 1;

    float normal_x = span->normal_x;
    float normal_y = span->normal_y;
    float normal_z = span->normal_z;


    /* span->x に対応する Texture の座標 (tex_xpos, tex_ypos) */
    int tex_xpos, tex_ypos;

    // span の始点に対応する Texture の座標 (tex1, tey1)
    float tex = span->tex_x1;
    float tey = span->tex_y1;

    // span の始点に対応する z 座標
    float zpos = span->start_z;

    /* Tile 内での座標 */
    int localx = getLocalX(span->x-1);
    int localy = getLocalY(span->y-1);

    /**
     * (tex_xpos, tex_ypos) の、Tile 内(上の図参照)での座標と
     * そのブロックのアドレス(MainMemory)
     */
    int tex_localx;
    int tex_localy;
    memaddr tex_addr;

    if (span->x < startx || endx < span->x) {
        return -1;
    }

    tex_xpos = (int)((span->tex_width-1) * tex);
    tex_ypos = (int)((span->tex_height-1) * tey);

    if (zpos < g->zRow[localx + (rangex*localy)]) {
        tex_addr = getTile(tex_xpos, tex_ypos,
                           span->tex_width, (memaddr)span->tex_addr);
        tex_localx = tex_xpos % TEXTURE_SPLIT_PIXEL;
        tex_localy = tex_ypos % TEXTURE_SPLIT_PIXEL;

	TilePtr tile = smanager->get_segment(tex_addr,g->tileList);
	smanager->wait_segment(tile);

        updateBuffer(g, zpos, rangex, localx, localy,
                     tex_localx, tex_localy,
                     normal_x,normal_y,normal_z,tile,
		     span->x, span->y, zpos, smanager);
    }

    return -1;
}

#if 0
static void
drawDot2(SchedTask *smanager, SpanPtr span, int startx, int end, int js, int wait_tag)
{
    //printf("%d\n", js);
}
#endif

/**
 * 長さが 1 より大きい Span の描画
 *
 * 本来の目的として、この関数(drawLine1) では
 *   : 既に SPE 上に Tile のある pixel だけ描画
 *   : それ以外は、ここで予め DMA load しておき、
 *   : drawLine2 で一気に描画する
 * ってものだったんだけど、どうも上手く行かなかったので
 * 今は drawLine1 で load -> wait -> rendering を全部やってます
 * (rendering といっても、rendering buffer に書き込むだけで
 *  まだ main memory (frame buffer) に dma store してるわけではない)
 *
 * @param span Span
 * @param startx 描画開始範囲
 * @param endx 描画終了範囲
 * @return 「span のどの位置まで rendering が終わったか」の x 座標
 */
static int
drawLine1(SchedTask *smanager, Gptr g, SpanPtr span, int startx, int endx, int wait_tag)
{
    int x = span->x;
    int rangex = endx - startx + 1;
    int x_len = span->length_x;

    float normal_x = span->normal_x;
    float normal_y = span->normal_y;
    float normal_z = span->normal_z;


    int js = (x < startx) ? startx - x : 0;
    int je = (x + x_len > endx) ? endx - x : x_len;

    /* span->x に対応する Texture の座標 (tex_xpos, tex_ypos) */
    int tex_xpos, tex_ypos;

    // span の始点に対応する座標 (tex1, tey1)
    float tex1 = span->tex_x1;
    float tey1 = span->tex_y1;

    // span の終点に対応する座標 (tex2, tey2)
    float tex2 = span->tex_x2;
    float tey2 = span->tex_y2;

    // span の始点、終点に対応する z 座標
    float zpos1 = span->start_z;
    float zpos2 = span->end_z;

    //spanを右から左に見ていくうちに、zが下がるのか、上がっていくのか。
    float z_inclination = (zpos2 - zpos1) / x_len;
    float world_z = zpos2;

    // Tile 内での座標
    int localx, localy = getLocalY(span->y-1);

    int ret = je+1;

    //for (int j = js; j <= je; j++) {
    for (int j = je; j >= js; j--) {
        float tex_x, tex_y, tex_z;

	world_z -= z_inclination;

        localx = getLocalX(x-1+j);

        tex_z = zpos1*(x_len-1-j)/(x_len-1) + zpos2*j/(x_len-1);

        tex_x = tex1*(x_len-1-j)/(x_len-1) + tex2*j/(x_len-1);
        tex_y = tey1*(x_len-1-j)/(x_len-1) + tey2*j/(x_len-1);
        if (tex_x > 1) tex_x = 1;
        if (tex_x < 0) tex_x = 0;
        if (tex_y > 1) tex_y = 1;
        if (tex_y < 0) tex_y = 0;
        tex_xpos = (int)((span->tex_width-1) * tex_x);
        tex_ypos = (int)((span->tex_height-1) * tex_y);

        if (tex_z < g->zRow[localx + (rangex*localy)]) {
            // (tex_xpos, tex_ypos) の、Tile 内(上の図参照)での座標と
            // そのブロックのアドレス(MainMemory)
            memaddr tex_addr;
            int tex_localx;
            int tex_localy;

            tex_addr = getTile(tex_xpos, tex_ypos,
                               span->tex_width, (memaddr)span->tex_addr);
            tex_localx = tex_xpos % TEXTURE_SPLIT_PIXEL;
            tex_localy = tex_ypos % TEXTURE_SPLIT_PIXEL;
	    TilePtr tile = smanager->get_segment(tex_addr,g->tileList);
	    smanager->wait_segment(tile);

            updateBuffer(g, tex_z, rangex, localx, localy,
                         tex_localx, tex_localy,
                         normal_x, normal_y, normal_z, tile,
			 span->x+j, span->y, world_z, smanager);
        }
    }

    return ret;
}

static int
infinity_light_calc(int color,float normal_x, float normal_y,
		    float normal_z, SchedTask *smanager, int x, int y, float z,
		    int world_x, int world_y, float world_z)

{

    unsigned char rgb[4];

    // 引数で受け取った color の rgb 情報の抜き出し
#if LITTLEENDIAN
    rgb[0] = (color & 0xff000000) >> 24;
    rgb[1] = (color & 0x00ff0000) >> 16;
    rgb[2] = (color & 0x0000ff00) >> 8;
    rgb[3] = (color & 0x000000ff);
#else
    rgb[3] = (color & 0xff000000) >> 24;
    rgb[2] = (color & 0x00ff0000) >> 16;
    rgb[1] = (color & 0x0000ff00) >> 8;
    rgb[0] = (color & 0x000000ff);
#endif


    vector float *light_xyz = (vector float*)smanager->global_get(Light);

    vector float v_rgb __attribute__((aligned(16))) = {(float)rgb[0],(float)rgb[1],(float)rgb[2],0};
    vector float normal_vector __attribute__((aligned(16))) = {normal_x,normal_y,normal_z,0};
    vector float light_vector __attribute__((aligned(16))) = {0,0,0,0};
    vector float v_inner __attribute__((aligned(16)));
    vector float v_world[4] __attribute__((aligned(16))) = {{world_x, world_y, -world_z, 0},
                                                            {world_x, world_y, -world_z, 0},
                                                            {world_x, world_y, -world_z, 0},
                                                            {world_x, world_y, -world_z, 0}};


    int light_rgb;
    float inner_product;

    normalize(normal_vector, normal_vector);

    vector float vtmp_rgb __attribute__((aligned(16))) = {0,0,0,0};

    int light_num = 4;
    int *light_switch = (int*)smanager->global_get(LightSwitch);

    for (int i = 0; i < light_num; i++) {

      if (light_switch[i] == 1) {
          light_vector = spu_sub(v_world[i],light_xyz[i]);
	  normalize(light_vector, light_vector);
	  inner_product = innerProduct1(normal_vector,light_vector);
	  v_inner = spu_splats(inner_product);
	  vtmp_rgb = spu_madd(v_rgb,v_inner,vtmp_rgb);
      }

    }

    vector unsigned int v_flag __attribute__((aligned(16)));
    vector float max_rgb __attribute__((aligned(16))) = (vector float)spu_splats((float)255.0f);

    v_flag = spu_cmpgt(max_rgb,vtmp_rgb);
    vtmp_rgb = spu_sel(max_rgb,vtmp_rgb,v_flag);

    vector unsigned int last_rgb __attribute__((aligned(16)));
    last_rgb = spu_convtu(vtmp_rgb,0);


    //計算した rgb を light_rgb にまとめる。
#if LITTLEENDIAN
    light_rgb = (last_rgb[0] << 24) + (last_rgb[1] << 16) + (last_rgb[2] << 8) + (last_rgb[3]);
#else
    light_rgb = (last_rgb[3] << 24) + (last_rgb[2] << 16) + (last_rgb[1] << 8) + (last_rgb[0]);
#endif

    return light_rgb;
}


static int
run(SchedTask *smanager, void *rbuf, void *wbuf)
{

  if ((int)smanager->get_param(5) == 0) {

    int rangex_start = (long)smanager->get_param(2);
    int rangex_end   = (long)smanager->get_param(3);

    int rangex = rangex_end - rangex_start + 1;
    int rangey = (long)smanager->get_param(4);

    for(int h = 0; h < rangey; h++) {
	int *buf = (int*)smanager->get_output(h);
	for (int i = 0; i < rangex; i++) {
	    buf[i] = 0;
	}
    }

    return 0;

  }

    G g0;
    Gptr g = &g0;

    SpanPackPtr spack = (SpanPackPtr)smanager->get_input(0);
    SpanPackPtr next_spack = (SpanPackPtr)smanager->allocate(sizeof(SpanPack));
    SpanPackPtr free_spack = next_spack; // next_spack の free() 用
    Span *span;

    Span nop_span;
    nop_span.length_x = 1;

    // uint32 display   = (long)smanager->get_param(0);
    // int screen_width = (long)smanager->get_param(1);
    int rangex_start = (long)smanager->get_param(2);
    int rangex_end   = (long)smanager->get_param(3);

    // このタスクが担当する x の範囲
    int rangex = rangex_end - rangex_start + 1;

    // y の範囲
    int rangey = (long)smanager->get_param(4);
    g->tileList = (TileListPtr)smanager->global_get(GLOBAL_TILE_LIST);

    g->zRow = zRow_init(smanager, rangex, rangey);
    g->linebuf = linebuf_init(smanager, rangex, rangey, 0);

    int tl_tag[2] = {TEX_LOAD1, TEX_LOAD2};
    int tl_tag_flg1 = 0;
    int tl_tag_flg2 = 1;

    do {
        /**
         * SpanPack->next が存在する場合、
         * 現在の SpanPack を処理してる間に
         * 次の SpanPack の DMA 転送を行う
         */
        if (spack->next != NULL) {
            smanager->dma_load(next_spack, (memaddr)spack->next,
                               sizeof(SpanPack), SPAN_PACK_LOAD);
        } else {
            next_spack = NULL;
        }

        SpanPtr resume_span = &nop_span;

        for (int t = 0; t < spack->info.size; t++) {
            SpanPtr next_span;

            span = &spack->span[t];

	    if (span->length_x != 1) {
	      drawLine1(
	      	    smanager, g,
                    span, rangex_start, rangex_end, tl_tag[tl_tag_flg1]);
	    } else {
	      drawDot1(
	          smanager, g,
	           span, rangex_start, rangex_end, tl_tag[tl_tag_flg1]);
            }
            next_span = span;

            resume_span = next_span;

            tl_tag_flg1 ^= 1;
            tl_tag_flg2 ^= 1;
        }

        smanager->dma_wait(SPAN_PACK_LOAD);

        SpanPackPtr tmp_spack = spack;
        spack = next_spack;
        next_spack = tmp_spack;
    } while (spack);


    free(g->zRow);
    free(g->linebuf);

//FINISH:
    /**
     * goto FINISH; の時は reboot なので
     * linebuf, zRow は free() しない
     */

    free(free_spack);


    return 0;
}

/* end */