Add initial screen fill implementation

Still need to add dirty tiles to it but thus far should be fine.

1 files changed, 59 insertions, 101 deletions

diff --git a/src/ppu.c b/src/ppu.c
index 3b159af..8e1710c 100644
--- a/src/ppu.c
+++ b/src/ppu.c
@@ -248,115 +248,73 @@ ppu_pixel(u32 *layer, u16 x, u16 y, u8 clr) {
 }
 
 IWRAM_CODE
+void clear_screen(u32 *layer, 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 in case we are using that.
+    dma_fill(layer, 0x11111111 * clr, KB(20) - 32, 3);
+}
+
+IWRAM_CODE
+static inline
 void
-ppu_rect(u32 *layer, size_t x0, size_t y0, size_t x1, size_t y1, u8 clr) {
+draw_hline(u32 *layer, size_t x0, size_t x1, size_t y0, u8 clr) {
     BOUNDCHECK_SCREEN(x0, y0);
-    BOUNDCHECK_SCREEN(x1, y1);
-
+    BOUNDCHECK_SCREEN(x1, y0);
     // 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 = &layer[start_row0 + (tile_x0 + tile_y0 * 32) * 8];
-    u32 *buf_bot = &layer[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;
+    size_t tile_y = y0 / 8;
+    size_t start_col = x0 % 8;
+    size_t end_col = x1 % 8;
+    size_t start_row = y0 % 8;
+
+    // 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 = &layer[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_col0 * 4;
-        size_t shift_right = (7 - start_col1) * 4;
-        u32 row_mask = 0xFFFFFFFF;
+        size_t shift_left = start_col * 4;
+        size_t shift_right = (7 - end_col) * 4;
+        u32 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;
+        *dst = (*dst & ~(mask << shift_left)) | (row << shift_left);
+        dst += 8;
+        for (size_t i = 1; i < dtx; i++) {
+            *dst = row;
+            dst += 8;
         }
+        *dst = (*dst & ~(mask >> shift_right)) | (row >> shift_right);
     }
 }
 
 IWRAM_CODE
 void
 screen_fill(u32 *layer, u16 x0, u16 y0, u16 x1, u16 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++) {
-        ppu_rect(layer, x0 + i, y0 + i, x1 - i, y1 - i, clr);
+    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)) {
+        clear_screen(layer, clr);
+        return;
+    }
+
+    // Drawline implementation.
+    for (size_t y = y0; y <= y1; y++) {
+        draw_hline(layer, x0, x1, y, clr);
     }
+    // TODO: dirty?
 }
 
 #if NEW_PPU == 0
@@ -616,15 +574,15 @@ ppu_2bpp(u32 *layer, u16 x, u16 y, u8 *sprite, u8 clr, u8 flip_x, u8 flip_y) {
     u32 *dst = &layer[start_row + (tile_x + tile_y * 32) * 8];
     u32 *lut = flip_x ? dec_byte_flip_x : dec_byte;
     if (clr == 1) {
-        u64 mask = ~((u64)0xFFFFFFFF << shift_left);
+        u32 mask = 0xFFFFFFFF;
         if (!flip_y) {
             for(size_t v = 0; v < 8; v++, dst++) {
                 if ((y + v) >= SCREEN_HEIGHT) break;
                 u8 ch1 = sprite[v];
                 u8 ch2 = sprite[v | 8];
                 u32 color = lut[ch1] | (lut[ch2] << 1);
-                dst[0] = (dst[0] & (mask << shift_left)) | color;
-                dst[8] = (dst[8] & (mask >> shift_right)) | (color >> shift_right);
+                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) dst += (32 - 1) * 8;
             }
         } else {
@@ -633,8 +591,8 @@ ppu_2bpp(u32 *layer, u16 x, u16 y, u8 *sprite, u8 clr, u8 flip_x, u8 flip_y) {
                 u8 ch1 = sprite[(7 - v)];
                 u8 ch2 = sprite[(7 - v) | 8];
                 u32 color = lut[ch1] | (lut[ch2] << 1);
-                dst[0] = (dst[0] & (mask << shift_left)) | color;
-                dst[8] = (dst[8] & (mask >> shift_right)) | (color >> shift_right);
+                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) dst += (32 - 1) * 8;
             }
         }
@@ -707,7 +665,7 @@ ppu_2bpp(u32 *layer, u16 x, u16 y, u8 *sprite, u8 clr, u8 flip_x, u8 flip_y) {
                 color = (clr1 * col1mask) |
                         (clr2 * col2mask) |
                         (clr3 * col3mask);
-                dst[0] = (dst[0] & (mask << shift_left)) | color;
+                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) dst += (32 - 1) * 8;
             }
@@ -727,7 +685,7 @@ ppu_2bpp(u32 *layer, u16 x, u16 y, u8 *sprite, u8 clr, u8 flip_x, u8 flip_y) {
                 color = (clr1 * col1mask) |
                         (clr2 * col2mask) |
                         (clr3 * col3mask);
-                dst[0] = (dst[0] & (mask << shift_left)) | color;
+                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) dst += (32 - 1) * 8;
             }