Add new renderer and prepare for render overhaul

1 files changed, 811 insertions, 0 deletions

diff --git a/src/renderer_m0.c b/src/renderer_m0.c
new file mode 100644
index 0000000..8bd4263
--- /dev/null
+++ b/src/renderer_m0.c
@@ -0,0 +1,811 @@
+//
+// 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.
+//
+
+#include "renderer.h"
+#include "text.h"
+
+//
+// Parameters.
+//
+
+#define SUBPIXEL_LINES 1
+#define DEC_BIG_LUT 1
+#define FLIP_TYPE 3
+
+// Front/back buffers for double buffering.
+#define BUF_0 ((u32*)(MEM_VRAM))
+#define BUF_1 ((u32*)(MEM_VRAM + KB(20)))
+
+// Pointer to the backbuffer.
+static u32 *backbuf = BUF_1;
+
+// Tracking which tiles are "dirty" and need refreshing.
+static u32 dirty_tiles[21] = {0};
+
+// Position of the tilemap.
+#define TILE_MAP  ((u32*)(MEM_VRAM + KB(40)))
+
+// Charblock and screenblock for both render buffers.
+#define CB_0 0
+#define CB_1 1
+#define SB_0 20
+#define SB_1 22
+
+// 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 optimizations 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
+
+// Swap A and B values without a tmp variable.
+#define SWAP(A, B) (((A) ^= (B)), ((B) ^= (A)), ((A) ^= (B)))
+
+// Swap A and B values to make sure A <= B.
+#define MAYBE_SWAP(A,B) if ((A) > (B)) { SWAP(A,B); }
+
+//
+// Basic primitives.
+//
+
+static inline
+void
+redraw(void) {
+    for (size_t i = 0; i < 21; i++) {
+        dirty_tiles[i] = 0xFFFFFFFF;
+    }
+}
+
+IWRAM_CODE
+void screen_fill(u8 clr)  {
+    // We have to make sure we leave the last tile blank to use as alpha channel
+    // when moving the BG during double buffering.
+    dma_fill(backbuf, 0x11111111 * clr, KB(20) - 32, 3);
+    redraw();
+}
+
+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;
+    u32 *dst = &backbuf[start_row + (tile_x + tile_y * 32) * 8];
+
+    // Update backbuffer.
+    size_t shift = start_col * sizeof(u32);
+    u32 mask = 0xF << shift;
+    u32 row = clr << shift;
+    *dst = (*dst & ~mask) | row;
+    dirty_tiles[tile_y] |= 1 << tile_x;
+}
+
+IWRAM_CODE
+static inline
+void
+draw_hline(size_t x0, size_t x1, size_t y0, u8 clr) {
+    BOUNDCHECK_SCREEN(x0, y0);
+    BOUNDCHECK_SCREEN(x1, y0);
+    // Find row positions for the given x/y coordinates.
+    size_t tile_x0 = x0 / 8;
+    size_t tile_x1 = x1 / 8;
+    size_t tile_y = y0 / 8;
+    size_t start_col = x0 % 8;
+    size_t end_col = x1 % 8;
+    size_t start_row = y0 % 8;
+    u32 dirty = (1 << tile_x0) | (1 << tile_x1);
+
+    // 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 entire
+    //        row.
+    size_t dtx = tile_x1 - tile_x0;
+    u32 *dst = &backbuf[start_row + (tile_x0 + tile_y * 32) * 8];
+    if (dtx < 1) {
+        size_t shift_left = start_col * 4;
+        size_t shift_right = (7 - end_col) * 4;
+        u32 mask = (0xFFFFFFFF >> shift_right) & (0xFFFFFFFF << shift_left);
+        u32 row = (0x11111111 * clr) & mask;
+        *dst = (*dst & ~mask) | row;
+    } else {
+        size_t shift_left = start_col * 4;
+        size_t shift_right = (7 - end_col) * 4;
+        u32 mask = 0xFFFFFFFF;
+        u32 row = 0x11111111 * clr;
+        *dst = (*dst & ~(mask << shift_left)) | (row << shift_left);
+        dst += 8;
+        for (size_t i = 1; i < dtx; i++) {
+            dirty |= (1 << (tile_x0 + i));
+            *dst = row;
+            dst += 8;
+        }
+        *dst = (*dst & ~(mask >> shift_right)) | (row >> shift_right);
+    }
+    dirty_tiles[tile_y] |= dirty;
+}
+
+IWRAM_CODE
+UNROLL_LOOPS
+static inline
+void
+draw_vline(size_t x0, size_t y0, size_t y1, u8 clr) {
+    BOUNDCHECK_SCREEN(x0, y0);
+    BOUNDCHECK_SCREEN(x0, y1);
+    size_t tile_x = x0 / 8;
+    size_t tile_y = y0 / 8;
+    size_t tile_y0 = y0 / 8;
+    size_t tile_y1 = y1 / 8;
+    size_t start_col = x0 % 8;
+    size_t start_row0 = y0 % 8;
+    size_t start_row1 = y1 % 8;
+
+    size_t shift_left = start_col * 4;
+    u32 dirty = (1 << tile_x);
+
+    u32 *dst = &backbuf[start_row0 + (tile_x + tile_y * 32) * 8];
+    u32 mask = 0x0000000F << shift_left;
+    u32 row = (0x11111111 * clr) & mask;
+    u32 dty = tile_y1 - tile_y0;
+    if (dty < 1) {
+        for (size_t i = 0; i <= (y1 - y0); i++, dst++) {
+            dst[0] = (dst[0] & ~mask) | row;
+        }
+    } else {
+        for (size_t i = 0; i < (8 - start_row0); i++, dst++) {
+            dst[0] = (dst[0] & ~mask) | row;
+        }
+        dst += 8 * 31;
+        for (size_t j = 1; j < dty; j++) {
+            dirty_tiles[tile_y0 + j] |= dirty;
+            for (size_t i = 0; i < 8; i++, dst++) {
+                dst[0] = (dst[0] & ~mask) | row;
+            }
+            dst += 8 * 31;
+        }
+        for (size_t i = 0; i <= start_row1; i++, dst++) {
+            dst[0] = (dst[0] & ~mask) | row;
+        }
+    }
+    dirty_tiles[tile_y0] |= dirty;
+    dirty_tiles[tile_y1] |= dirty;
+}
+
+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);
+    if (y0 == y1) {
+        MAYBE_SWAP(x0, x1);
+        draw_hline(x0, x1, y0, clr);
+    } else if (x0 == x1) {
+        MAYBE_SWAP(y0, y1);
+        draw_vline(x0, y0, y1, clr);
+    } else {
+        // Fixed Precision constants.
+        const int fp_bit = 6;
+        const int fp_one = FP_NUM(1, fp_bit);
+        const int fp_half = fp_one >> 1;
+
+        int dx = x0 > x1 ? x0 - x1 : x1 - x0;
+        int dy = y0 > y1 ? y0 - y1 : y1 - y0;
+
+        if ((dx >= dy && x0 > x1) || (dx < dy && y0 > y1)) {
+            SWAP(x0, x1);
+            SWAP(y0, y1);
+        }
+
+#if SUBPIXEL_LINES == 1
+        int dxf = (dx << fp_bit);
+        int dyf = (dy << fp_bit);
+        int frac_x = x0 > x1 ? FP_NUM(x0 - x1, fp_bit) : FP_NUM(x1 - x0, fp_bit);
+        int frac_y = y0 > y1 ? FP_NUM(y0 - y1, fp_bit) : FP_NUM(y1 - y0, fp_bit);
+        int x_step = x0 > x1 ? -1 : 1;
+        int y_step = y0 > y1 ? -1 : 1;
+        int distance = (frac_y - fp_one) * dx - (frac_x - fp_half) * dy;
+        if (dx >= dy) {
+            int step = dxf / dyf;
+            int remaining = dx;
+            while (remaining > (step - 1)) {
+                distance += step * 2 * dyf;
+                if (distance >= 0) {
+                    draw_hline(x0, x0 + step - 1, y0, clr);
+                    x0 += x_step * step;
+                    remaining -= step;
+                } else {
+                    if (remaining < step) {
+                        break;
+                    }
+                    draw_hline(x0, x0 + step, y0, clr);
+                    distance += 2 * dyf;
+                    x0 += x_step * (step + 1);
+                    remaining -= step + 1;
+                }
+                distance -= 2 * dxf;
+                y0 += y_step;
+            }
+            if (remaining >= 0) {
+                draw_hline(x0, x0 + remaining, y0, clr);
+            }
+        } else {
+            int step = dyf / dxf;
+            int remaining = dy;
+            while (remaining > (step - 1)) {
+                    distance += step * 2 * dxf;
+                    if (distance >= 0) {
+                        draw_vline(x0, y0, y0 + step - 1, clr);
+                        y0 += y_step * step;
+                        remaining -= step;
+                    } else {
+                        draw_vline(x0, y0, y0 + step, clr);
+                        distance += 2 * dxf;
+                        y0 += y_step * (step + 1);
+                        remaining -= step + 1;
+                    }
+                    distance -= 2 * dyf;
+                    x0 += x_step;
+            }
+            if (remaining >= 0) {
+                draw_vline(x0, y0, y0 + remaining, clr);
+            }
+        }
+#else
+        int x_step = x0 > x1 ? -1 : 1;
+        int y_step = y0 > y1 ? -1 : 1;
+        if (dx >= dy) {
+            int diff = 2 * dy - dx;
+            for (int i = 0; i < dx + 1; i++) {
+                draw_pixel(x0, y0, clr);
+                if (diff >= 0) {
+                    diff -= 2 * dx;
+                    y0 += y_step;
+                }
+                diff += 2 * dy;
+                x0 += x_step;
+            }
+        } else {
+            int diff = 2 * dx - dy;
+            for (int i = 0; i < dy + 1; i++) {
+                draw_pixel(x0, y0, clr);
+                if (diff >= 0) {
+                    diff -= 2 * dy;
+                    x0 += x_step;
+                }
+                diff += 2 * dx;
+                y0 += y_step;
+            }
+        }
+#endif
+    }
+}
+
+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);
+    MAYBE_SWAP(x0, x1);
+    MAYBE_SWAP(y0, y1);
+
+    draw_hline(x0, x1, y0, clr);
+    draw_hline(x0, x1, y1, clr);
+    draw_vline(x0, y0, y1, clr);
+    draw_vline(x1, y0, y1, clr);
+}
+
+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);
+    MAYBE_SWAP(x0, x1);
+    MAYBE_SWAP(y0, y1);
+
+    // Special condition. If the screen is to be completely filled, use the DMA
+    // instead.
+    if (x0 == 0 && x1 >= (SCREEN_WIDTH - 1) && y0 == 0 && y1 >= (SCREEN_HEIGHT - 1)) {
+        screen_fill(clr);
+        return;
+    }
+
+    for (size_t y = y0; y <= y1; y++) {
+        draw_hline(x0, x1, y, clr);
+    }
+}
+
+//
+// Sprites (chr/icn).
+//
+
+#if DEC_BIG_LUT == 1
+static u32 dec_byte_flip_x[256] = {
+    0x00000000, 0x00000001, 0x00000010, 0x00000011, 0x00000100,
+    0x00000101, 0x00000110, 0x00000111, 0x00001000, 0x00001001,
+    0x00001010, 0x00001011, 0x00001100, 0x00001101, 0x00001110,
+    0x00001111, 0x00010000, 0x00010001, 0x00010010, 0x00010011,
+    0x00010100, 0x00010101, 0x00010110, 0x00010111, 0x00011000,
+    0x00011001, 0x00011010, 0x00011011, 0x00011100, 0x00011101,
+    0x00011110, 0x00011111, 0x00100000, 0x00100001, 0x00100010,
+    0x00100011, 0x00100100, 0x00100101, 0x00100110, 0x00100111,
+    0x00101000, 0x00101001, 0x00101010, 0x00101011, 0x00101100,
+    0x00101101, 0x00101110, 0x00101111, 0x00110000, 0x00110001,
+    0x00110010, 0x00110011, 0x00110100, 0x00110101, 0x00110110,
+    0x00110111, 0x00111000, 0x00111001, 0x00111010, 0x00111011,
+    0x00111100, 0x00111101, 0x00111110, 0x00111111, 0x01000000,
+    0x01000001, 0x01000010, 0x01000011, 0x01000100, 0x01000101,
+    0x01000110, 0x01000111, 0x01001000, 0x01001001, 0x01001010,
+    0x01001011, 0x01001100, 0x01001101, 0x01001110, 0x01001111,
+    0x01010000, 0x01010001, 0x01010010, 0x01010011, 0x01010100,
+    0x01010101, 0x01010110, 0x01010111, 0x01011000, 0x01011001,
+    0x01011010, 0x01011011, 0x01011100, 0x01011101, 0x01011110,
+    0x01011111, 0x01100000, 0x01100001, 0x01100010, 0x01100011,
+    0x01100100, 0x01100101, 0x01100110, 0x01100111, 0x01101000,
+    0x01101001, 0x01101010, 0x01101011, 0x01101100, 0x01101101,
+    0x01101110, 0x01101111, 0x01110000, 0x01110001, 0x01110010,
+    0x01110011, 0x01110100, 0x01110101, 0x01110110, 0x01110111,
+    0x01111000, 0x01111001, 0x01111010, 0x01111011, 0x01111100,
+    0x01111101, 0x01111110, 0x01111111, 0x10000000, 0x10000001,
+    0x10000010, 0x10000011, 0x10000100, 0x10000101, 0x10000110,
+    0x10000111, 0x10001000, 0x10001001, 0x10001010, 0x10001011,
+    0x10001100, 0x10001101, 0x10001110, 0x10001111, 0x10010000,
+    0x10010001, 0x10010010, 0x10010011, 0x10010100, 0x10010101,
+    0x10010110, 0x10010111, 0x10011000, 0x10011001, 0x10011010,
+    0x10011011, 0x10011100, 0x10011101, 0x10011110, 0x10011111,
+    0x10100000, 0x10100001, 0x10100010, 0x10100011, 0x10100100,
+    0x10100101, 0x10100110, 0x10100111, 0x10101000, 0x10101001,
+    0x10101010, 0x10101011, 0x10101100, 0x10101101, 0x10101110,
+    0x10101111, 0x10110000, 0x10110001, 0x10110010, 0x10110011,
+    0x10110100, 0x10110101, 0x10110110, 0x10110111, 0x10111000,
+    0x10111001, 0x10111010, 0x10111011, 0x10111100, 0x10111101,
+    0x10111110, 0x10111111, 0x11000000, 0x11000001, 0x11000010,
+    0x11000011, 0x11000100, 0x11000101, 0x11000110, 0x11000111,
+    0x11001000, 0x11001001, 0x11001010, 0x11001011, 0x11001100,
+    0x11001101, 0x11001110, 0x11001111, 0x11010000, 0x11010001,
+    0x11010010, 0x11010011, 0x11010100, 0x11010101, 0x11010110,
+    0x11010111, 0x11011000, 0x11011001, 0x11011010, 0x11011011,
+    0x11011100, 0x11011101, 0x11011110, 0x11011111, 0x11100000,
+    0x11100001, 0x11100010, 0x11100011, 0x11100100, 0x11100101,
+    0x11100110, 0x11100111, 0x11101000, 0x11101001, 0x11101010,
+    0x11101011, 0x11101100, 0x11101101, 0x11101110, 0x11101111,
+    0x11110000, 0x11110001, 0x11110010, 0x11110011, 0x11110100,
+    0x11110101, 0x11110110, 0x11110111, 0x11111000, 0x11111001,
+    0x11111010, 0x11111011, 0x11111100, 0x11111101, 0x11111110,
+    0x11111111
+};
+
+static u32 dec_byte[256] = {
+    0x00000000, 0x10000000, 0x01000000, 0x11000000, 0x00100000,
+    0x10100000, 0x01100000, 0x11100000, 0x00010000, 0x10010000,
+    0x01010000, 0x11010000, 0x00110000, 0x10110000, 0x01110000,
+    0x11110000, 0x00001000, 0x10001000, 0x01001000, 0x11001000,
+    0x00101000, 0x10101000, 0x01101000, 0x11101000, 0x00011000,
+    0x10011000, 0x01011000, 0x11011000, 0x00111000, 0x10111000,
+    0x01111000, 0x11111000, 0x00000100, 0x10000100, 0x01000100,
+    0x11000100, 0x00100100, 0x10100100, 0x01100100, 0x11100100,
+    0x00010100, 0x10010100, 0x01010100, 0x11010100, 0x00110100,
+    0x10110100, 0x01110100, 0x11110100, 0x00001100, 0x10001100,
+    0x01001100, 0x11001100, 0x00101100, 0x10101100, 0x01101100,
+    0x11101100, 0x00011100, 0x10011100, 0x01011100, 0x11011100,
+    0x00111100, 0x10111100, 0x01111100, 0x11111100, 0x00000010,
+    0x10000010, 0x01000010, 0x11000010, 0x00100010, 0x10100010,
+    0x01100010, 0x11100010, 0x00010010, 0x10010010, 0x01010010,
+    0x11010010, 0x00110010, 0x10110010, 0x01110010, 0x11110010,
+    0x00001010, 0x10001010, 0x01001010, 0x11001010, 0x00101010,
+    0x10101010, 0x01101010, 0x11101010, 0x00011010, 0x10011010,
+    0x01011010, 0x11011010, 0x00111010, 0x10111010, 0x01111010,
+    0x11111010, 0x00000110, 0x10000110, 0x01000110, 0x11000110,
+    0x00100110, 0x10100110, 0x01100110, 0x11100110, 0x00010110,
+    0x10010110, 0x01010110, 0x11010110, 0x00110110, 0x10110110,
+    0x01110110, 0x11110110, 0x00001110, 0x10001110, 0x01001110,
+    0x11001110, 0x00101110, 0x10101110, 0x01101110, 0x11101110,
+    0x00011110, 0x10011110, 0x01011110, 0x11011110, 0x00111110,
+    0x10111110, 0x01111110, 0x11111110, 0x00000001, 0x10000001,
+    0x01000001, 0x11000001, 0x00100001, 0x10100001, 0x01100001,
+    0x11100001, 0x00010001, 0x10010001, 0x01010001, 0x11010001,
+    0x00110001, 0x10110001, 0x01110001, 0x11110001, 0x00001001,
+    0x10001001, 0x01001001, 0x11001001, 0x00101001, 0x10101001,
+    0x01101001, 0x11101001, 0x00011001, 0x10011001, 0x01011001,
+    0x11011001, 0x00111001, 0x10111001, 0x01111001, 0x11111001,
+    0x00000101, 0x10000101, 0x01000101, 0x11000101, 0x00100101,
+    0x10100101, 0x01100101, 0x11100101, 0x00010101, 0x10010101,
+    0x01010101, 0x11010101, 0x00110101, 0x10110101, 0x01110101,
+    0x11110101, 0x00001101, 0x10001101, 0x01001101, 0x11001101,
+    0x00101101, 0x10101101, 0x01101101, 0x11101101, 0x00011101,
+    0x10011101, 0x01011101, 0x11011101, 0x00111101, 0x10111101,
+    0x01111101, 0x11111101, 0x00000011, 0x10000011, 0x01000011,
+    0x11000011, 0x00100011, 0x10100011, 0x01100011, 0x11100011,
+    0x00010011, 0x10010011, 0x01010011, 0x11010011, 0x00110011,
+    0x10110011, 0x01110011, 0x11110011, 0x00001011, 0x10001011,
+    0x01001011, 0x11001011, 0x00101011, 0x10101011, 0x01101011,
+    0x11101011, 0x00011011, 0x10011011, 0x01011011, 0x11011011,
+    0x00111011, 0x10111011, 0x01111011, 0x11111011, 0x00000111,
+    0x10000111, 0x01000111, 0x11000111, 0x00100111, 0x10100111,
+    0x01100111, 0x11100111, 0x00010111, 0x10010111, 0x01010111,
+    0x11010111, 0x00110111, 0x10110111, 0x01110111, 0x11110111,
+    0x00001111, 0x10001111, 0x01001111, 0x11001111, 0x00101111,
+    0x10101111, 0x01101111, 0x11101111, 0x00011111, 0x10011111,
+    0x01011111, 0x11011111, 0x00111111, 0x10111111, 0x01111111,
+    0x11111111
+};
+
+IWRAM_CODE
+static inline
+u32
+decode_1bpp(u8 row, u8 flip_x) {
+    if (flip_x) {
+        return dec_byte_flip_x[row];
+    }
+    return dec_byte[row];
+}
+#else
+static u16 dec_nibble[] = {
+    0x0000, 0x1000, 0x0100, 0x1100,
+    0x0010, 0x1010, 0x0110, 0x1110,
+    0x0001, 0x1001, 0x0101, 0x1101,
+    0x0011, 0x1011, 0x0111, 0x1111,
+};
+
+static u16 dec_nibble_flip_x[] = {
+    0x0000, 0x0001, 0x0010, 0x0011,
+    0x0100, 0x0101, 0x0110, 0x0111,
+    0x1000, 0x1001, 0x1010, 0x1011,
+    0x1100, 0x1101, 0x1110, 0x1111,
+};
+
+IWRAM_CODE
+static inline
+u32
+decode_1bpp(u8 row, u8 flip_x) {
+    if (flip_x) {
+        u16 *lut = dec_nibble_flip_x;
+        return (u32)lut[(row >> 4) & 0xF] << 16 | (u32)lut[(row >> 0) & 0xF];
+    }
+    u16 *lut = dec_nibble;
+    return (u32)lut[(row >> 0) & 0xF] << 16 | (u32)lut[(row >> 4) & 0xF];
+}
+#endif
+
+IWRAM_CODE
+UNROLL_LOOPS
+void
+draw_chr(size_t x, size_t y, u8 *sprite, u8 clr, u8 flip_x, u8 flip_y) {
+    BOUNDCHECK_SCREEN(x, y);
+    size_t tile_x0 = x / 8;
+    size_t tile_x1 = (x + 7) / 8;
+    size_t tile_y = y / 8;
+    size_t start_col = x % 8;
+    size_t start_row = y % 8;
+    size_t shift_left = start_col * 4;
+    size_t shift_right = (8 - start_col) * 4;
+    u32 dirty = (1 << tile_x0) | (1 << tile_x1);
+    u32 *dst = &backbuf[start_row + (tile_x0 + tile_y * 32) * 8];
+#if DEC_BIG_LUT
+    u32 *lut = flip_x ? dec_byte_flip_x : dec_byte;
+#endif
+    if (!flip_y) {
+        for(size_t v = 0; v < 8; v++, dst++) {
+            if ((y + v) >= SCREEN_HEIGHT) break;
+            u8 ch1 = sprite[v + 0];
+            u8 ch2 = sprite[v + 8];
+#if DEC_BIG_LUT
+            u32 clr_a = lut[ch1];
+            u32 clr_b = lut[ch2];
+#else
+            u32 clr_a = decode_1bpp(ch1, flip_x);
+            u32 clr_b = decode_1bpp(ch2, flip_x);
+#endif
+            u32 mask_a = (clr_a * 0xF);
+            u32 mask_b = (clr_b * 0xF);
+            u32 mask = (mask_a | mask_b);
+            u32 color;
+            if (clr == 0) {
+                color = clr_a + (clr_b << 1);
+            } else if (clr == 15) {
+                color = 0;
+            } else {
+                color = (clr_a | clr_b) * clr;
+            }
+            dst[0] = (dst[0] & ~(mask << shift_left))  | (color << shift_left);
+            dst[8] = (dst[8] & ~(mask >> shift_right)) | (color >> shift_right);
+            if ((start_row + v)  == 7) {
+                dirty_tiles[tile_y + 1] |= dirty;
+                dst += (32 - 1) * 8;
+            }
+        }
+    } else {
+        for(size_t v = 0; v < 8; v++, dst++) {
+            if ((y + v) >= SCREEN_HEIGHT) break;
+            u8 ch1 = sprite[(7 - v) + 0];
+            u8 ch2 = sprite[(7 - v) + 8];
+#if DEC_BIG_LUT
+            u32 clr_a = lut[ch1];
+            u32 clr_b = lut[ch2];
+#else
+            u32 clr_a = decode_1bpp(ch1, flip_x);
+            u32 clr_b = decode_1bpp(ch2, flip_x);
+#endif
+            u32 mask_a = (clr_a * 0xF);
+            u32 mask_b = (clr_b * 0xF);
+            u32 mask = (mask_a | mask_b);
+            u32 color;
+            if (clr == 0) {
+                color = clr_a + (clr_b << 1);
+            } else if (clr == 15) {
+                color = 0;
+            } else {
+                color = (clr_a | clr_b) * clr;
+            }
+            dst[0] = (dst[0] & ~(mask << shift_left))  | (color << shift_left);
+            dst[8] = (dst[8] & ~(mask >> shift_right)) | (color >> shift_right);
+            if ((start_row + v)  == 7) {
+                dirty_tiles[tile_y + 1] |= dirty;
+                dst += (32 - 1) * 8;
+            }
+        }
+    }
+    dirty_tiles[tile_y] |= dirty;
+}
+
+IWRAM_CODE
+UNROLL_LOOPS
+void
+draw_icn(size_t x, size_t y, u8 *sprite, u8 clr, u8 flip_x, u8 flip_y) {
+    BOUNDCHECK_SCREEN(x, y);
+    size_t tile_x0 = x / 8;
+    size_t tile_x1 = (x + 7) / 8;
+    size_t tile_y = y / 8;
+    size_t start_col = x % 8;
+    size_t start_row = y % 8;
+    size_t shift_left = start_col * 4;
+    size_t shift_right = (8 - start_col) * 4;
+    u32 dirty = (1 << tile_x0) | (1 << tile_x1);
+    u32 *dst = &backbuf[start_row + (tile_x0 + tile_y * 32) * 8];
+#if DEC_BIG_LUT
+    u32 *lut = flip_x ? dec_byte_flip_x : dec_byte;
+#endif
+    if (!flip_y) {
+        for(size_t v = 0; v < 8; v++, dst++) {
+            if ((y + v) >= SCREEN_HEIGHT) break;
+            u8 ch1 = sprite[v + 0];
+#if DEC_BIG_LUT
+            u32 color = lut[ch1];
+#else
+            u32 color = decode_1bpp(ch1, flip_x);
+#endif
+            u32 mask = (color * 0xF);
+            color *= clr;
+            dst[0] = (dst[0] & ~(mask << shift_left))  | (color << shift_left);
+            dst[8] = (dst[8] & ~(mask >> shift_right)) | (color >> shift_right);
+            if ((start_row + v)  == 7) {
+                dirty_tiles[tile_y + 1] |= dirty;
+                dst += (32 - 1) * 8;
+            }
+        }
+    } else {
+        for(size_t v = 0; v < 8; v++, dst++) {
+            if ((y + v) >= SCREEN_HEIGHT) break;
+            u8 ch1 = sprite[(7 - v) + 0];
+#if DEC_BIG_LUT
+            u32 color = lut[ch1];
+#else
+            u32 color = decode_1bpp(ch1, flip_x);
+#endif
+            u32 mask = (color * 0xF);
+            color *= clr;
+            dst[0] = (dst[0] & ~(mask << shift_left))  | (color << shift_left);
+            dst[8] = (dst[8] & ~(mask >> shift_right)) | (color >> shift_right);
+            if ((start_row + v)  == 7) {
+                dirty_tiles[tile_y + 1] |= dirty;
+                dst += (32 - 1) * 8;
+            }
+        }
+    }
+    dirty_tiles[tile_y] |= dirty;
+}
+
+//
+// Flipping buffers/copying memory.
+//
+
+IWRAM_CODE
+void
+flip_buffer(void) {
+// Mode 0: double buffering without dirty tiles.
+#if FLIP_TYPE == 0
+    if (backbuf == BUF_0) {
+        backbuf = BUF_1;
+        BG_H_SCROLL_0 = 0;
+        BG_H_SCROLL_1 = -240;
+    } else {
+        backbuf = BUF_0;
+        BG_H_SCROLL_0 = -240;
+        BG_H_SCROLL_1 = 0;
+    }
+
+// Mode 1: single buffer, copy the dirty lines from backbuffer (BUF_1) to
+// frontbuffer (BUF_0) using the DMA.
+#elif FLIP_TYPE == 1
+    u32 *front = BUF_0;
+    u32 *back = BUF_1;
+    BG_H_SCROLL_0 = 0;
+    BG_H_SCROLL_1 = -240;
+    for (size_t j = 0; j < 20; ++j) {
+        if (dirty_tiles[j] == 0) {
+            continue;
+        }
+        u32 offset = j * 32 * 8;
+        dma_copy(front + offset, back + offset, (30 * 8 * 4), 3);
+        dirty_tiles[j] = 0;
+    }
+
+// Mode 2: single buffer, copy the dirty tiles from backbuffer (BUF_1) to
+// frontbuffer (BUF_0).
+#elif FLIP_TYPE == 2
+    u32 *front = BUF_0;
+    u32 *back = BUF_1;
+    BG_H_SCROLL_0 = 0;
+    BG_H_SCROLL_1 = -240;
+    for (size_t j = 0; j < 20; ++j) {
+        if (dirty_tiles[j] == 0) {
+            continue;
+        }
+        size_t k = 1;
+        for (size_t i = 0; i < 30; ++i, k <<= 1) {
+            if (dirty_tiles[j] & k) {
+                Tile *mem_front = front;
+                Tile *mem_back = back;
+                mem_front[i + j * 32] = mem_back[i + j * 32];
+            }
+        }
+        dirty_tiles[j] = 0;
+    }
+
+// Mode 3: Double buffering with dirty line, copying the dirty lines if needed
+// after flipping buffers with the DMA.
+#elif FLIP_TYPE == 3
+    bool should_flip = false;
+    for (size_t j = 0; j < 20; ++j) {
+        if (dirty_tiles[j] == 0) {
+            continue;
+        }
+        should_flip = true;
+        break;
+    }
+    if (!should_flip) {
+        return;
+    }
+    u32 *frontbuf = backbuf;
+    if (backbuf == BUF_0) {
+        backbuf = BUF_1;
+        BG_H_SCROLL_0 = 0;
+        BG_H_SCROLL_1 = -240;
+    } else {
+        backbuf = BUF_0;
+        BG_H_SCROLL_0 = -240;
+        BG_H_SCROLL_1 = 0;
+    }
+    for (size_t j = 0; j < 20; ++j) {
+        if (dirty_tiles[j] == 0) {
+            continue;
+        }
+        u32 offset = j * 32 * 8;
+        dma_copy(backbuf + offset, frontbuf + offset, (30 * 8 * 4), 3);
+        dirty_tiles[j] = 0;
+    }
+
+// Mode 4: Double buffering with dirty tiles, copying the dirty tiles if needed
+// after flipping buffers.
+#elif FLIP_TYPE == 4
+    bool should_flip = false;
+    for (size_t j = 0; j < 20; ++j) {
+        if (dirty_tiles[j] == 0) {
+            continue;
+        }
+        should_flip = true;
+        break;
+    }
+    if (!should_flip) {
+        return;
+    }
+    u32 *frontbuf = backbuf;
+    if (backbuf == BUF_0) {
+        backbuf = BUF_1;
+        BG_H_SCROLL_0 = 0;
+        BG_H_SCROLL_1 = -240;
+    } else {
+        backbuf = BUF_0;
+        BG_H_SCROLL_0 = -240;
+        BG_H_SCROLL_1 = 0;
+    }
+    for (size_t j = 0; j < 20; ++j) {
+        if (dirty_tiles[j] == 0) {
+            continue;
+        }
+        size_t k = 1;
+        for (size_t i = 0; i < 30; ++i, k <<= 1) {
+            if (dirty_tiles[j] & k) {
+                Tile *mem_front = frontbuf;
+                Tile *mem_back = backbuf;
+                mem_back[i + j * 32] = mem_front[i + j * 32];
+            }
+        }
+        dirty_tiles[j] = 0;
+    }
+#endif
+}
+
+//
+// Text rendering.
+//
+
+#include "font.h"
+
+// Font rendering function for the text engine.
+void
+txt_drawc(char c, size_t x, size_t y, u8 clr) {
+    u8 *tile = font_icn;
+    draw_icn(x, y, tile + 8 * c, clr, 1, 0);
+}
+
+//
+// Initialization.
+//
+
+void
+renderer_init(void) {
+    // Initialize display mode and bg palette.
+    DISP_CTRL = DISP_MODE_0 | DISP_BG_0 | DISP_BG_1;
+
+    // Clear VRAM.
+    dma_fill((u32*)MEM_VRAM, 0, KB(96), 3);
+
+    // Initialize backgrounds.
+    BG_CTRL(0) = BG_CHARBLOCK(CB_0) | BG_SCREENBLOCK(SB_0) | BG_PRIORITY(0) | BG_SIZE(1);
+    BG_CTRL(1) = BG_CHARBLOCK(CB_1) | BG_SCREENBLOCK(SB_1) | BG_PRIORITY(1) | BG_SIZE(1);
+
+    // Initialize background memory map for the render buffers. The backgrounds
+    // are 64x32 each, with the second screenblock pointing to a zeroed tile.
+    // This makes it so while scrolling the backgrounds to the second screen we
+    // effectively disabling them. Thanks to this we can perform double
+    // buffering with mode 0 rendering.
+    u16 *mem_map_fg = SCREENBLOCK_MEM[SB_0];
+    u16 *mem_map_fg_blank = SCREENBLOCK_MEM[SB_0 + 1];
+    u16 *mem_map_bg = SCREENBLOCK_MEM[SB_1];
+    u16 *mem_map_bg_blank = SCREENBLOCK_MEM[SB_1 + 1];
+    for (size_t i = 0; i < 32 * 20; ++i) {
+        mem_map_fg[i] = i;
+        mem_map_fg_blank[i] = 32 * 20 - 1;
+        mem_map_bg[i] = i + 32 * 4;
+        mem_map_bg_blank[i] = (32 * 20 - 1) +  32 * 4;
+    }
+
+    // Setup initial background state.
+    BG_H_SCROLL_0 = -240;
+    BG_H_SCROLL_1 = -240;
+
+    // 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;
+    PAL_BUFFER_BG[6] = COLOR_WHITE;
+    PAL_BUFFER_BG[7] = COLOR_GREEN;
+    PAL_BUFFER_BG[8] = COLOR_PURPLE;
+
+    // Initialize text engine.
+    txt_init(txt_drawc);
+}