circuitpython/shared-module/displayio/TileGrid.c

277 lines
11 KiB
C

/*
* This file is part of the Micro Python project, http://micropython.org/
*
* The MIT License (MIT)
*
* Copyright (c) 2018 Scott Shawcroft for Adafruit Industries
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "shared-bindings/displayio/TileGrid.h"
#include "shared-bindings/displayio/Bitmap.h"
#include "shared-bindings/displayio/ColorConverter.h"
#include "shared-bindings/displayio/OnDiskBitmap.h"
#include "shared-bindings/displayio/Palette.h"
#include "shared-bindings/displayio/Shape.h"
void common_hal_displayio_tilegrid_construct(displayio_tilegrid_t *self, mp_obj_t bitmap,
uint16_t bitmap_width_in_tiles,
mp_obj_t pixel_shader, uint16_t width, uint16_t height,
uint16_t tile_width, uint16_t tile_height, uint16_t x, uint16_t y, uint8_t default_tile) {
uint32_t total_tiles = width * height;
// Sprites will only have one tile so save a little memory by inlining values in the pointer.
uint8_t inline_tiles = sizeof(uint8_t*);
if (total_tiles <= inline_tiles) {
self->tiles = 0;
// Pack values into the pointer since there are only a few.
for (uint32_t i = 0; i < inline_tiles; i++) {
((uint8_t*) &self->tiles)[i] = default_tile;
}
self->inline_tiles = true;
} else {
self->tiles = (uint8_t*) m_malloc(total_tiles, false);
for (uint32_t i = 0; i < total_tiles; i++) {
self->tiles[i] = default_tile;
}
self->inline_tiles = false;
}
self->bitmap_width_in_tiles = bitmap_width_in_tiles;
self->width_in_tiles = width;
self->height_in_tiles = height;
self->area.x1 = x;
self->area.y1 = y;
self->area.x2 = x + width * tile_width;
self->area.y2 = y + height * tile_height;
self->tile_width = tile_width;
self->tile_height = tile_height;
self->bitmap = bitmap;
self->pixel_shader = pixel_shader;
}
mp_int_t common_hal_displayio_tilegrid_get_x(displayio_tilegrid_t *self) {
return self->area.x1;
}
void common_hal_displayio_tilegrid_set_x(displayio_tilegrid_t *self, mp_int_t x) {
if (self->area.x1 == x) {
return;
}
self->needs_refresh = true;
self->area.x2 += (self->area.x1 - x);
self->area.x1 = x;
}
mp_int_t common_hal_displayio_tilegrid_get_y(displayio_tilegrid_t *self) {
return self->area.y1;
}
void common_hal_displayio_tilegrid_set_y(displayio_tilegrid_t *self, mp_int_t y) {
if (self->area.y1 == y) {
return;
}
self->needs_refresh = true;
self->area.y2 += (self->area.y1 - y);
self->area.y1 = y;
}
mp_obj_t common_hal_displayio_tilegrid_get_pixel_shader(displayio_tilegrid_t *self) {
return self->pixel_shader;
}
void common_hal_displayio_tilegrid_set_pixel_shader(displayio_tilegrid_t *self, mp_obj_t pixel_shader) {
self->pixel_shader = pixel_shader;
self->needs_refresh = true;
}
uint16_t common_hal_displayio_tilegrid_get_width(displayio_tilegrid_t *self) {
return self->width_in_tiles;
}
uint16_t common_hal_displayio_tilegrid_get_height(displayio_tilegrid_t *self) {
return self->height_in_tiles;
}
uint8_t common_hal_displayio_tilegrid_get_tile(displayio_tilegrid_t *self, uint16_t x, uint16_t y) {
uint8_t* tiles = self->tiles;
if (self->inline_tiles) {
tiles = (uint8_t*) &self->tiles;
}
if (tiles == NULL) {
return 0;
}
return tiles[y * self->width_in_tiles + x];
}
void common_hal_displayio_tilegrid_set_tile(displayio_tilegrid_t *self, uint16_t x, uint16_t y, uint8_t tile_index) {
uint8_t* tiles = self->tiles;
if (self->inline_tiles) {
tiles = (uint8_t*) &self->tiles;
}
if (tiles == NULL) {
return;
}
tiles[y * self->width_in_tiles + x] = tile_index;
self->needs_refresh = true;
}
void common_hal_displayio_tilegrid_set_top_left(displayio_tilegrid_t *self, uint16_t x, uint16_t y) {
self->top_left_x = x;
self->top_left_y = y;
self->needs_refresh = true;
}
bool displayio_tilegrid_get_area(displayio_tilegrid_t *self, displayio_buffer_transform_t* transform, displayio_area_t* area, uint32_t* mask, uint32_t *buffer) {
// If no tiles are present we have no impact.
uint8_t* tiles = self->tiles;
if (self->inline_tiles) {
tiles = (uint8_t*) &self->tiles;
}
if (tiles == NULL) {
return false;
}
displayio_area_t overlap;
displayio_area_t scaled_area = {
.x1 = self->area.x1 * transform->scale,
.y1 = self->area.y1 * transform->scale,
.x2 = self->area.x2 * transform->scale,
.y2 = self->area.y2 * transform->scale
};
if (!displayio_area_compute_overlap(area, &scaled_area, &overlap)) {
return false;
}
int16_t x_stride = 1;
int16_t y_stride = displayio_area_width(area);
if (transform->transpose_xy) {
x_stride = displayio_area_height(area);
y_stride = 1;
}
uint16_t start = 0;
if (transform->mirror_x) {
start += (area->x2 - area->x1 - 1) * x_stride;
x_stride *= -1;
}
if (transform->mirror_y) {
start += (area->y2 - area->y1 - 1) * y_stride;
y_stride *= -1;
}
// Track if this layer finishes filling in the given area. We can ignore any remaining
// layers at that point.
bool full_coverage = displayio_area_equal(area, &overlap);
// TODO(tannewt): Skip coverage tracking if all pixels outside the overlap have already been
// set and our palette is all opaque.
// TODO(tannewt): Check to see if the pixel_shader has any transparency. If it doesn't then we
// can either return full coverage or bulk update the mask.
int16_t y = overlap.y1 - scaled_area.y1;
if (y < 0) {
y = 0;
}
int16_t x_shift = area->x1 - scaled_area.x1;
int16_t y_shift = area->y1 - scaled_area.y1;
for (; y < overlap.y2 - scaled_area.y1; y++) {
int16_t x = overlap.x1 - scaled_area.x1;
if (x < 0) {
x = 0;
}
int16_t row_start = start + (y - y_shift) * y_stride;
int16_t local_y = y / transform->scale;
for (; x < overlap.x2 - scaled_area.x1; x++) {
// Compute the destination pixel in the buffer and mask based on the transformations.
uint16_t offset = row_start + (x - x_shift) * x_stride;
// This is super useful for debugging out range accesses. Uncomment to use.
// if (offset < 0 || offset >= displayio_area_size(area)) {
// asm("bkpt");
// }
// Check the mask first to see if the pixel has already been set.
if ((mask[offset / 32] & (1 << (offset % 32))) != 0) {
continue;
}
int16_t local_x = x / transform->scale;
uint16_t tile_location = ((local_y / self->tile_height + self->top_left_y) % self->height_in_tiles) * self->width_in_tiles + (local_x / self->tile_width + self->top_left_x) % self->width_in_tiles;
uint8_t tile = tiles[tile_location];
uint16_t tile_x = (tile % self->bitmap_width_in_tiles) * self->tile_width + local_x % self->tile_width;
uint16_t tile_y = (tile / self->bitmap_width_in_tiles) * self->tile_height + local_y % self->tile_height;
uint32_t value = 0;
// We always want to read bitmap pixels by row first and then transpose into the destination
// buffer because most bitmaps are row associated.
if (MP_OBJ_IS_TYPE(self->bitmap, &displayio_bitmap_type)) {
value = common_hal_displayio_bitmap_get_pixel(self->bitmap, tile_x, tile_y);
} else if (MP_OBJ_IS_TYPE(self->bitmap, &displayio_shape_type)) {
value = common_hal_displayio_shape_get_pixel(self->bitmap, tile_x, tile_y);
} else if (MP_OBJ_IS_TYPE(self->bitmap, &displayio_ondiskbitmap_type)) {
value = common_hal_displayio_ondiskbitmap_get_pixel(self->bitmap, tile_x, tile_y);
}
uint16_t* pixel = ((uint16_t*) buffer) + offset;
if (self->pixel_shader == mp_const_none) {
*pixel = value;
return true;
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_palette_type)) {
if (!displayio_palette_get_color(self->pixel_shader, value, pixel)) {
// A pixel is transparent so we haven't fully covered the area ourselves.
full_coverage = false;
} else {
mask[offset / 32] |= 1 << (offset % 32);
}
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_colorconverter_type)) {
if (!common_hal_displayio_colorconverter_convert(self->pixel_shader, value, pixel)) {
// A pixel is transparent so we haven't fully covered the area ourselves.
full_coverage = false;
} else {
mask[offset / 32] |= 1 << (offset % 32);
}
}
}
}
return full_coverage;
}
bool displayio_tilegrid_needs_refresh(displayio_tilegrid_t *self) {
if (self->needs_refresh) {
return true;
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_palette_type)) {
return displayio_palette_needs_refresh(self->pixel_shader);
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_colorconverter_type)) {
return displayio_colorconverter_needs_refresh(self->pixel_shader);
}
return false;
}
void displayio_tilegrid_finish_refresh(displayio_tilegrid_t *self) {
self->needs_refresh = false;
if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_palette_type)) {
displayio_palette_finish_refresh(self->pixel_shader);
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_colorconverter_type)) {
displayio_colorconverter_finish_refresh(self->pixel_shader);
}
// TODO(tannewt): We could double buffer changes to position and move them over here.
// That way they won't change during a refresh and tear.
}