Make point 2 in areas exclusive and simplify full_coverage.
This commit is contained in:
parent
3fad7de8db
commit
7a117f52ed
@ -138,6 +138,8 @@ bool displayio_group_get_area(displayio_group_t *self, displayio_buffer_transfor
|
|||||||
displayio_area_shift(area, -self->x * transform->scale, -self->y * transform->scale);
|
displayio_area_shift(area, -self->x * transform->scale, -self->y * transform->scale);
|
||||||
transform->scale *= self->scale;
|
transform->scale *= self->scale;
|
||||||
|
|
||||||
|
// Track if any of the layers finishes filling in the given area. We can ignore any remaining
|
||||||
|
// layers at that point.
|
||||||
bool full_coverage = false;
|
bool full_coverage = false;
|
||||||
for (int32_t i = self->size - 1; i >= 0 ; i--) {
|
for (int32_t i = self->size - 1; i >= 0 ; i--) {
|
||||||
mp_obj_t layer = self->children[i].native;
|
mp_obj_t layer = self->children[i].native;
|
||||||
|
@ -154,8 +154,8 @@ bool displayio_tilegrid_get_area(displayio_tilegrid_t *self, displayio_buffer_tr
|
|||||||
displayio_area_t scaled_area = {
|
displayio_area_t scaled_area = {
|
||||||
.x1 = self->area.x1 * transform->scale,
|
.x1 = self->area.x1 * transform->scale,
|
||||||
.y1 = self->area.y1 * transform->scale,
|
.y1 = self->area.y1 * transform->scale,
|
||||||
.x2 = (self->area.x2 + 1) * transform->scale - 1, // Second point is inclusive.
|
.x2 = self->area.x2 * transform->scale,
|
||||||
.y2 = (self->area.y2 + 1) * transform->scale - 1
|
.y2 = self->area.y2 * transform->scale
|
||||||
};
|
};
|
||||||
if (!displayio_area_compute_overlap(area, &scaled_area, &overlap)) {
|
if (!displayio_area_compute_overlap(area, &scaled_area, &overlap)) {
|
||||||
return false;
|
return false;
|
||||||
@ -169,19 +169,20 @@ bool displayio_tilegrid_get_area(displayio_tilegrid_t *self, displayio_buffer_tr
|
|||||||
}
|
}
|
||||||
uint16_t start = 0;
|
uint16_t start = 0;
|
||||||
if (transform->mirror_x) {
|
if (transform->mirror_x) {
|
||||||
start += (area->x2 - area->x1) * x_stride;
|
start += (area->x2 - area->x1 - 1) * x_stride;
|
||||||
x_stride *= -1;
|
x_stride *= -1;
|
||||||
}
|
}
|
||||||
if (transform->mirror_y) {
|
if (transform->mirror_y) {
|
||||||
start += (area->y2 - area->y1) * y_stride;
|
start += (area->y2 - area->y1 - 1) * y_stride;
|
||||||
y_stride *= -1;
|
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);
|
bool full_coverage = displayio_area_equal(area, &overlap);
|
||||||
|
|
||||||
// TODO(tannewt): Set full coverage to true if all pixels outside the overlap have already been
|
// TODO(tannewt): Skip coverage tracking if all pixels outside the overlap have already been
|
||||||
// set as well.
|
// set and our palette is all opaque.
|
||||||
bool always_full_coverage = false;
|
|
||||||
|
|
||||||
// TODO(tannewt): Check to see if the pixel_shader has any transparency. If it doesn't then we
|
// 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.
|
// can either return full coverage or bulk update the mask.
|
||||||
@ -191,17 +192,22 @@ bool displayio_tilegrid_get_area(displayio_tilegrid_t *self, displayio_buffer_tr
|
|||||||
}
|
}
|
||||||
int16_t x_shift = area->x1 - scaled_area.x1;
|
int16_t x_shift = area->x1 - scaled_area.x1;
|
||||||
int16_t y_shift = area->y1 - scaled_area.y1;
|
int16_t y_shift = area->y1 - scaled_area.y1;
|
||||||
for (; y <= overlap.y2 - scaled_area.y1; y++) {
|
for (; y < overlap.y2 - scaled_area.y1; y++) {
|
||||||
int16_t x = overlap.x1 - scaled_area.x1;
|
int16_t x = overlap.x1 - scaled_area.x1;
|
||||||
if (x < 0) {
|
if (x < 0) {
|
||||||
x = 0;
|
x = 0;
|
||||||
}
|
}
|
||||||
int16_t row_start = start + (y - y_shift) * y_stride;
|
int16_t row_start = start + (y - y_shift) * y_stride;
|
||||||
int16_t local_y = y / transform->scale;
|
int16_t local_y = y / transform->scale;
|
||||||
for (; x <= overlap.x2 - scaled_area.x1; x++) {
|
for (; x < overlap.x2 - scaled_area.x1; x++) {
|
||||||
// Compute the destination pixel in the buffer and mask based on the transformations.
|
// Compute the destination pixel in the buffer and mask based on the transformations.
|
||||||
uint16_t offset = row_start + (x - x_shift) * x_stride;
|
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.
|
// Check the mask first to see if the pixel has already been set.
|
||||||
if ((mask[offset / 32] & (1 << (offset % 32))) != 0) {
|
if ((mask[offset / 32] & (1 << (offset % 32))) != 0) {
|
||||||
continue;
|
continue;
|
||||||
@ -229,22 +235,21 @@ bool displayio_tilegrid_get_area(displayio_tilegrid_t *self, displayio_buffer_tr
|
|||||||
return true;
|
return true;
|
||||||
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_palette_type)) {
|
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_palette_type)) {
|
||||||
if (!displayio_palette_get_color(self->pixel_shader, value, pixel)) {
|
if (!displayio_palette_get_color(self->pixel_shader, value, pixel)) {
|
||||||
// mark the pixel as transparent
|
// A pixel is transparent so we haven't fully covered the area ourselves.
|
||||||
full_coverage = false;
|
full_coverage = false;
|
||||||
} else if (!always_full_coverage) {
|
} else {
|
||||||
mask[offset / 32] |= 1 << (offset % 32);
|
mask[offset / 32] |= 1 << (offset % 32);
|
||||||
}
|
}
|
||||||
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_colorconverter_type)) {
|
} else if (MP_OBJ_IS_TYPE(self->pixel_shader, &displayio_colorconverter_type)) {
|
||||||
if (!common_hal_displayio_colorconverter_convert(self->pixel_shader, value, pixel)) {
|
if (!common_hal_displayio_colorconverter_convert(self->pixel_shader, value, pixel)) {
|
||||||
// mark the pixel as transparent
|
// A pixel is transparent so we haven't fully covered the area ourselves.
|
||||||
full_coverage = false;
|
full_coverage = false;
|
||||||
} else if (!always_full_coverage) {
|
} else {
|
||||||
mask[offset / 32] |= 1 << (offset % 32);
|
mask[offset / 32] |= 1 << (offset % 32);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
return full_coverage;
|
return full_coverage;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
@ -66,8 +66,8 @@ void displayio_refresh_displays(void) {
|
|||||||
displayio_area_t whole_screen = {
|
displayio_area_t whole_screen = {
|
||||||
.x1 = 0,
|
.x1 = 0,
|
||||||
.y1 = 0,
|
.y1 = 0,
|
||||||
.x2 = display->width - 1,
|
.x2 = display->width,
|
||||||
.y2 = display->height - 1
|
.y2 = display->height
|
||||||
};
|
};
|
||||||
if (display->transpose_xy) {
|
if (display->transpose_xy) {
|
||||||
swap(&whole_screen.x2, &whole_screen.y2);
|
swap(&whole_screen.x2, &whole_screen.y2);
|
||||||
@ -88,13 +88,13 @@ void displayio_refresh_displays(void) {
|
|||||||
displayio_area_t subrectangle = {
|
displayio_area_t subrectangle = {
|
||||||
.x1 = 0,
|
.x1 = 0,
|
||||||
.y1 = rows_per_buffer * j,
|
.y1 = rows_per_buffer * j,
|
||||||
.x2 = displayio_area_width(&whole_screen) - 1,
|
.x2 = displayio_area_width(&whole_screen),
|
||||||
.y2 = rows_per_buffer * (j + 1) - 1
|
.y2 = rows_per_buffer * (j + 1)
|
||||||
};
|
};
|
||||||
|
|
||||||
displayio_display_begin_transaction(display);
|
displayio_display_begin_transaction(display);
|
||||||
displayio_display_set_region_to_update(display, subrectangle.x1, subrectangle.y1,
|
displayio_display_set_region_to_update(display, subrectangle.x1, subrectangle.y1,
|
||||||
subrectangle.x2 + 1, subrectangle.y2 + 1);
|
subrectangle.x2, subrectangle.y2);
|
||||||
displayio_display_end_transaction(display);
|
displayio_display_end_transaction(display);
|
||||||
|
|
||||||
// Handle display mirroring and transpose.
|
// Handle display mirroring and transpose.
|
||||||
@ -102,22 +102,22 @@ void displayio_refresh_displays(void) {
|
|||||||
displayio_buffer_transform_t transform;
|
displayio_buffer_transform_t transform;
|
||||||
if (display->mirror_x) {
|
if (display->mirror_x) {
|
||||||
uint16_t width = displayio_area_width(&whole_screen);
|
uint16_t width = displayio_area_width(&whole_screen);
|
||||||
transformed_subrectangle.x1 = width - subrectangle.x2 - 1;
|
transformed_subrectangle.x1 = width - subrectangle.x2;
|
||||||
transformed_subrectangle.x2 = width - subrectangle.x1 - 1;
|
transformed_subrectangle.x2 = width - subrectangle.x1;
|
||||||
} else {
|
} else {
|
||||||
transformed_subrectangle.x1 = subrectangle.x1;
|
transformed_subrectangle.x1 = subrectangle.x1;
|
||||||
transformed_subrectangle.x2 = subrectangle.x2;
|
transformed_subrectangle.x2 = subrectangle.x2;
|
||||||
}
|
}
|
||||||
if (display->mirror_y != display->transpose_xy) {
|
if (display->mirror_y != display->transpose_xy) {
|
||||||
uint16_t height = displayio_area_height(&whole_screen);
|
uint16_t height = displayio_area_height(&whole_screen);
|
||||||
transformed_subrectangle.y1 = height - subrectangle.y2 - 1;
|
transformed_subrectangle.y1 = height - subrectangle.y2;
|
||||||
transformed_subrectangle.y2 = height - subrectangle.y1 - 1;
|
transformed_subrectangle.y2 = height - subrectangle.y1;
|
||||||
} else {
|
} else {
|
||||||
transformed_subrectangle.y1 = subrectangle.y1;
|
transformed_subrectangle.y1 = subrectangle.y1;
|
||||||
transformed_subrectangle.y2 = subrectangle.y2;
|
transformed_subrectangle.y2 = subrectangle.y2;
|
||||||
}
|
}
|
||||||
transform.width = transformed_subrectangle.x2 - transformed_subrectangle.x1 + 1;
|
transform.width = transformed_subrectangle.x2 - transformed_subrectangle.x1;
|
||||||
transform.height = transformed_subrectangle.y2 - transformed_subrectangle.y1 + 1;
|
transform.height = transformed_subrectangle.y2 - transformed_subrectangle.y1;
|
||||||
if (display->transpose_xy) {
|
if (display->transpose_xy) {
|
||||||
int16_t y1 = transformed_subrectangle.y1;
|
int16_t y1 = transformed_subrectangle.y1;
|
||||||
int16_t y2 = transformed_subrectangle.y2;
|
int16_t y2 = transformed_subrectangle.y2;
|
||||||
@ -139,8 +139,8 @@ void displayio_refresh_displays(void) {
|
|||||||
if (!full_coverage) {
|
if (!full_coverage) {
|
||||||
uint32_t index = 0;
|
uint32_t index = 0;
|
||||||
uint32_t current_mask = 0;
|
uint32_t current_mask = 0;
|
||||||
for (int16_t y = subrectangle.y1; y <= subrectangle.y2; y++) {
|
for (int16_t y = subrectangle.y1; y < subrectangle.y2; y++) {
|
||||||
for (int16_t x = subrectangle.x1; x <= subrectangle.x2; x++) {
|
for (int16_t x = subrectangle.x1; x < subrectangle.x2; x++) {
|
||||||
if (index % 32 == 0) {
|
if (index % 32 == 0) {
|
||||||
current_mask = mask[index / 32];
|
current_mask = mask[index / 32];
|
||||||
}
|
}
|
||||||
@ -268,11 +268,11 @@ bool displayio_area_compute_overlap(const displayio_area_t* a,
|
|||||||
}
|
}
|
||||||
|
|
||||||
uint16_t displayio_area_width(const displayio_area_t* area) {
|
uint16_t displayio_area_width(const displayio_area_t* area) {
|
||||||
return area->x2 - area->x1 + 1;
|
return area->x2 - area->x1;
|
||||||
}
|
}
|
||||||
|
|
||||||
uint16_t displayio_area_height(const displayio_area_t* area) {
|
uint16_t displayio_area_height(const displayio_area_t* area) {
|
||||||
return area->y2 - area->y1 + 1;
|
return area->y2 - area->y1;
|
||||||
}
|
}
|
||||||
|
|
||||||
uint32_t displayio_area_size(const displayio_area_t* area) {
|
uint32_t displayio_area_size(const displayio_area_t* area) {
|
||||||
|
@ -32,7 +32,7 @@
|
|||||||
typedef struct {
|
typedef struct {
|
||||||
int16_t x1;
|
int16_t x1;
|
||||||
int16_t y1;
|
int16_t y1;
|
||||||
int16_t x2; // Second point is inclusive.
|
int16_t x2; // Second point is exclusive.
|
||||||
int16_t y2;
|
int16_t y2;
|
||||||
} displayio_area_t;
|
} displayio_area_t;
|
||||||
|
|
||||||
|
@ -70,8 +70,8 @@ void supervisor_start_terminal(uint16_t width_px, uint16_t height_px) {
|
|||||||
|
|
||||||
grid->width_in_tiles = width_in_tiles;
|
grid->width_in_tiles = width_in_tiles;
|
||||||
grid->height_in_tiles = height_in_tiles;
|
grid->height_in_tiles = height_in_tiles;
|
||||||
grid->area.x2 = grid->area.x1 + width_in_tiles * grid->tile_width - 1;
|
grid->area.x2 = grid->area.x1 + width_in_tiles * grid->tile_width;
|
||||||
grid->area.y2 = grid->area.y1 + height_in_tiles * grid->tile_height - 1;
|
grid->area.y2 = grid->area.y1 + height_in_tiles * grid->tile_height;
|
||||||
grid->tiles = tiles;
|
grid->tiles = tiles;
|
||||||
|
|
||||||
supervisor_terminal.cursor_x = 0;
|
supervisor_terminal.cursor_x = 0;
|
||||||
|
Loading…
Reference in New Issue
Block a user