//
// This Mode 0 renderer provides a way of drawing directly to a framebuffer
// (similar to Mode 3 and 4) while retaining the flexibility of using other
// backgrounds if needed. It also performs double buffering to avoid tearing
// artifacts and tries to only draw tiles that changed on each frame.
//
// In addition to the frontbuffer (displayed on background 0), a tiled text
// layer is displayed on background 1, which can be used for application
// development or for debug information.
//
// These two layers occupy the first and second background charblocks, leaving
// the remaining two available for other background layers. There are 14KB of
// sprite memory available, since the backbuffer is located at the end of the
// VRAM, but if more space is needed it can be moved to the end of the BG
// charblocks instead.
//

#include "renderer.h"
#include "text.h"

// Keep track of which tiles need to be copied to the frontbuffer.
static u32 dirty_tiles[21] = {0};

// Boundchecks can be disable at compile time but this will not always improve
// the performance and can in fact make it worse. It is possible that this is
// due to some aliasing optimiztions but not sure at this moment.
#ifdef DISABLE_BOUNDCHECK_SCREEN
#define BOUNDCHECK_SCREEN(X,Y)
#else
#define BOUNDCHECK_SCREEN(X,Y) if ((X) >= SCREEN_WIDTH || (Y) >= SCREEN_HEIGHT) return;
#endif

IWRAM_CODE
void
draw_pixel(size_t x, size_t y, u8 clr) {
    BOUNDCHECK_SCREEN(x, y);

    // Find row position for the given x/y coordinates.
    size_t tile_x = x / 8;
    size_t tile_y = y / 8;
    size_t start_col = x % 8;
    size_t start_row = y % 8;
    size_t pos = start_row + (tile_x + tile_y * 32) * 8;

    // Update backbuffer.
    size_t shift = start_col * sizeof(u32);
    BACKBUF[pos] = (BACKBUF[pos] & ~(0xF << shift)) | clr << shift;

    // Mark tile as dirty.
    dirty_tiles[tile_y] |= 1 << tile_x;
}

IWRAM_CODE
void
draw_line(size_t x0, size_t y0, size_t x1, size_t y1, u8 clr) {
    BOUNDCHECK_SCREEN(x0, y0);
    BOUNDCHECK_SCREEN(x1, y1);

    // Find row positions for the given x/y coordinates.
    size_t tile_x0 = x0 / 8;
    size_t tile_y0 = y0 / 8;
    size_t tile_x1 = x1 / 8;
    size_t tile_y1 = y1 / 8;
    size_t start_col0 = x0 % 8;
    size_t start_col1 = x1 % 8;
    size_t start_row0 = y0 % 8;
    size_t start_row1 = y1 % 8;

    // Get a pointer to the backbuffer and the tile row.
    u32 *backbuffer = &BACKBUF[start_row0 + (tile_x0 + tile_y0 * 32) * 8];

    if (y0 == y1) {
        // Horizontal line. There are 3 cases:
        //     1. Lines fit on a single tile.
        //     2. Lines go through 2 tiles, both require partial row updates.
        //     3. Lines go through 3 or more tiles, first and last tiles use
        //        partial row updates, rows in the middle can write the.
        size_t dx = tile_x1 - tile_x0;
        if (dx < 1) {
            u32 row_mask = 0xFFFFFFFF;
            row_mask >>= (7 - start_col1 - dx) * 4;
            row_mask &= 0xFFFFFFFF << start_col0 * 4;
            u32 row = (0x11111111 * clr) & row_mask;
            backbuffer[0] = (backbuffer[0] & ~row_mask) | row;
            dirty_tiles[tile_y0] |= 1 << tile_x0;
        } else {
            size_t shift_left = start_col0 * 4;
            size_t shift_right = (7 - start_col1) * 4;
            u32 row_mask = 0xFFFFFFFF;
            u32 row = 0x11111111 * clr;
            backbuffer[0] = backbuffer[0] & ~(row_mask << shift_left);
            backbuffer[0] |= row << shift_left;
            dirty_tiles[tile_y0] |= 1 << tile_x0;
            for (size_t i = 1; i < dx; i++) {
                backbuffer[i * 8] = row;
                dirty_tiles[tile_y0] |= 1 << (tile_x0 + i);
            }
            backbuffer[dx * 8] = backbuffer[dx * 8] & ~(row_mask >> shift_right);
            backbuffer[dx * 8] |= row >> shift_right;
            dirty_tiles[tile_y0] |= 1 << (tile_x0 + dx);
        }
    } else if (x0 == x1) {
        // Vertical line. The cases are analogous to the horizontal ones.
        size_t dy = tile_y1 - tile_y0;
        u32 row_mask = 0xF << start_col0 * 4;
        u32 row_left = (0x11111111 * clr) & row_mask;
        if (dy < 1) {
            for (size_t i = 0; i <= y1 - y0; i++, backbuffer++) {
                backbuffer[0] = (backbuffer[0] & ~row_mask) | row_left;
            }
        } else {
            for (size_t i = 0; i < (8 - start_row0); i++, backbuffer++) {
                backbuffer[0] = (backbuffer[0] & ~row_mask) | row_left;
            }
            dirty_tiles[tile_y0] |= 1 << tile_x0;
            backbuffer += 8 * 31;
            for (size_t j = 1; j < dy; j++) {
                for (size_t i = 0; i < 8; i++, backbuffer++) {
                    backbuffer[0] = (backbuffer[0] & ~row_mask) | row_left;
                }
                backbuffer += 8 * 31;
                dirty_tiles[tile_y0 + j] |= 1 << tile_x0;
            }
            for (size_t i = 0; i <= start_row1; i++, backbuffer++) {
                backbuffer[0] = (backbuffer[0] & ~row_mask) | row_left;
            }
            dirty_tiles[tile_y1] |= 1 << tile_x0;
        }
    } else {
        // Diagonal line.
        int dx = x0 > x1 ? x0 - x1 : x1 - x0;
        int dy = y0 > y1 ? y1 - y0 : y0 - y1;
        int x_step = x0 < x1 ? 1 : -1;
        int y_step = y0 < y1 ? 1 : -1;
        int err = dx + dy;
        while (!(x0 == x1 && y0 == y1)) {
            draw_pixel(x0, y0, clr);
            int diff = 2 * err;
            if (diff >= dy) {
                err += dy;
                x0 += x_step;
            }
            if (diff <= dx) {
                err += dx;
                y0 += y_step;
            }
        }
    }
}

IWRAM_CODE
void
draw_rect(size_t x0, size_t y0, size_t x1, size_t y1, u8 clr) {
    BOUNDCHECK_SCREEN(x0, y0);
    BOUNDCHECK_SCREEN(x1, y1);

    // Find row positions for the given x/y coordinates.
    size_t tile_x0 = x0 / 8;
    size_t tile_y0 = y0 / 8;
    size_t tile_x1 = x1 / 8;
    size_t tile_y1 = y1 / 8;
    size_t start_col0 = x0 % 8;
    size_t start_col1 = x1 % 8;
    size_t start_row0 = y0 % 8;
    size_t start_row1 = y1 % 8;

    // Get a pointer to the backbuffer and the tile row.
    u32 *buf_top = &BACKBUF[start_row0 + (tile_x0 + tile_y0 * 32) * 8];
    u32 *buf_bot = &BACKBUF[start_row1 + (tile_x0 + tile_y1 * 32) * 8];

    size_t dx = tile_x1 - tile_x0;
    size_t dy = tile_y1 - tile_y0;

    // We can update two lines at a time, which is faster than calling draw_line
    // four times.
    if (dx < 1) {
        u32 row_mask = 0xFFFFFFFF;
        row_mask >>= (7 - start_col1 - dx) * 4;
        row_mask &= 0xFFFFFFFF << start_col0 * 4;
        u32 row = (0x11111111 * clr) & row_mask;
        buf_top[0] = (buf_top[0] & ~row_mask) | row;
        buf_bot[0] = (buf_bot[0] & ~row_mask) | row;
        dirty_tiles[tile_y0] |= 1 << tile_x0;
        dirty_tiles[tile_y1] |= 1 << tile_x0;
    } else {
        size_t shift_left = start_col0 * 4;
        size_t shift_right = (7 - start_col1) * 4;
        u32 row_mask = 0xFFFFFFFF;
        u32 row = 0x11111111 * clr;
        buf_top[0] = buf_top[0] & ~(row_mask << shift_left);
        buf_top[0] |= row << shift_left;
        buf_bot[0] = buf_bot[0] & ~(row_mask << shift_left);
        buf_bot[0] |= row << shift_left;
        dirty_tiles[tile_y0] |= 1 << tile_x0;
        dirty_tiles[tile_y1] |= 1 << tile_x0;
        for (size_t i = 1; i < dx; i++) {
            buf_top[i * 8] = row;
            buf_bot[i * 8] = row;
            dirty_tiles[tile_y0] |= 1 << (tile_x0 + i);
            dirty_tiles[tile_y1] |= 1 << (tile_x0 + i);
        }
        buf_top[dx * 8] = buf_top[dx * 8] & ~(row_mask >> shift_right);
        buf_top[dx * 8] |= row >> shift_right;
        buf_bot[dx * 8] = buf_bot[dx * 8] & ~(row_mask >> shift_right);
        buf_bot[dx * 8] |= row >> shift_right;
        dirty_tiles[tile_y0] |= 1 << (tile_x0 + dx);
        dirty_tiles[tile_y1] |= 1 << (tile_x0 + dx);
    }
    u32 row_mask_left = 0xF << start_col0 * 4;
    u32 row_mask_right = 0xF << start_col1 * 4;
    u32 row_left = (0x11111111 * clr) & row_mask_left;
    u32 row_right = (0x11111111 * clr) & row_mask_right;
    if (dy < 1) {
        for (size_t i = 1; i < y1 - y0; i++, buf_top++) {
            buf_top[1] = buf_top[1] & ~row_mask_left;
            buf_top[1] |= row_left;
            buf_top[1 + 8 * dx] = buf_top[1 + 8 * dx] & ~row_mask_right;
            buf_top[1 + 8 * dx] |= row_right;
        }
    } else {
        for (size_t i = 1; i < (8 - start_row0); i++, buf_top++) {
            buf_top[1] = buf_top[1] & ~row_mask_left;
            buf_top[1] |= row_left;
            buf_top[1 + 8 * dx] = buf_top[1 + 8 * dx] & ~row_mask_right;
            buf_top[1 + 8 * dx] |= row_right;
        }
        buf_top += 8 * 31;
        for (size_t j = 1; j < dy; j++) {
            for (size_t i = 0; i < 8; i++, buf_top++) {
                buf_top[1] = buf_top[1] & ~row_mask_left;
                buf_top[1] |= row_left;
                buf_top[1 + 8 * dx] = buf_top[1 + 8 * dx] & ~row_mask_right;
                buf_top[1 + 8 * dx] |= row_right;
            }
            buf_top += 8 * 31;
            dirty_tiles[tile_y0 + j] |= 1 << tile_x0;
            dirty_tiles[tile_y0 + j] |= 1 << (tile_x0 + dx);
        }
        for (size_t i = 0; i < start_row1; i++, buf_top++) {
            buf_top[1] = buf_top[1] & ~row_mask_left;
            buf_top[1] |= row_left;
            buf_top[1 + 8 * dx] = buf_top[1 + 8 * dx] & ~row_mask_right;
            buf_top[1 + 8 * dx] |= row_right;
        }
    }
}

IWRAM_CODE
void
draw_filled_rect(size_t x0, size_t y0, size_t x1, size_t y1, u8 clr) {
    BOUNDCHECK_SCREEN(x0, y0);
    BOUNDCHECK_SCREEN(x1, y1);

    size_t dx = x1 - x0;
    size_t dy = y1 - y0;
    size_t n_rect = MIN(dx, dy);
    n_rect = n_rect / 2 + 1;
    for (size_t i = 0; i < n_rect; i++) {
        draw_rect(x0 + i, y0 + i, x1 - i, y1 - i, clr);
    }
}

void
clear_screen(void) {
    dma_fill(FRONTBUF, 0, KB(20), 3);
}

IWRAM_CODE
void
flip_buffer(void) {
    // Copy dirty tiles from the backbuffer to the frontbuffer.
    Tile *dst = FRONTBUF;
    Tile *src = BACKBUF;
    for (size_t j = 0; j < 20; ++j) {
        if (dirty_tiles[j] == 0) {
            continue;
        }
        for (size_t i = 0, k = 1; i < 30; ++i, k <<= 1) {
            if (dirty_tiles[j] & k) {
                dst[i + j * 32] = src[i + j * 32];
            }
        }
        dirty_tiles[j] = 0;
    }
}

void
renderer_init(void) {
    // Initialize display mode and bg palette.
    DISP_CTRL = DISP_MODE_0 | DISP_BG_0 | DISP_BG_1 | DISP_OBJ;

    // Initialize backgrounds.
    BG_CTRL(0) = BG_CHARBLOCK(0) | BG_SCREENBLOCK(FRONTBUF_SB) | BG_PRIORITY(1);
    BG_CTRL(1) = BG_CHARBLOCK(1) | BG_SCREENBLOCK(FONT_SB) | BG_PRIORITY(0);

    // Use DMA to clear front and back buffers as well as the font memory map.
    dma_fill(FRONTBUF, 0, KB(20), 3);
    dma_fill(FRONTBUF_TILEMAP, 0, KB(2), 3);
    dma_fill(BACKBUF, 0, KB(20), 3);
    dma_fill(FONT_DATA, 0, KB(8), 3);
    dma_fill(FONT_TILEMAP, (FONT_OFFSET << 16) | FONT_OFFSET, KB(2), 3);

    // Initialize default palette.
    PAL_BUFFER_BG[0] = COLOR_BLACK;
    PAL_BUFFER_BG[1] = COLOR_WHITE;
    PAL_BUFFER_BG[2] = COLOR_RED;
    PAL_BUFFER_BG[3] = COLOR_BLUE;
    PAL_BUFFER_BG[4] = COLOR_CYAN;
    PAL_BUFFER_BG[5] = COLOR_GREY;

    // Initialize background memory map for frontbuffer and font backgorund.
    for (size_t i = 0; i < 32 * 20; ++i) {
        FRONTBUF_TILEMAP[i] = i;
    }

    // Initialize text engine.
    txt_init(FONT_DATA, FONT_TILEMAP, FONT_OFFSET);
}