beta-0.89.2
[luatex.git] / source / libs / cairo / cairo-src / src / cairo-tor-scan-converter.c
blobb1b51872b19467d94a2fb0e431f0f1434697f747
1 /* -*- Mode: c; tab-width: 8; c-basic-offset: 4; indent-tabs-mode: t; -*- */
2 /* glitter-paths - polygon scan converter
4 * Copyright (c) 2008 M Joonas Pihlaja
5 * Copyright (c) 2007 David Turner
7 * Permission is hereby granted, free of charge, to any person
8 * obtaining a copy of this software and associated documentation
9 * files (the "Software"), to deal in the Software without
10 * restriction, including without limitation the rights to use,
11 * copy, modify, merge, publish, distribute, sublicense, and/or sell
12 * copies of the Software, and to permit persons to whom the
13 * Software is furnished to do so, subject to the following
14 * conditions:
16 * The above copyright notice and this permission notice shall be
17 * included in all copies or substantial portions of the Software.
19 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
20 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
21 * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
22 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
23 * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
24 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
25 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
26 * OTHER DEALINGS IN THE SOFTWARE.
28 /* This is the Glitter paths scan converter incorporated into cairo.
29 * The source is from commit 734c53237a867a773640bd5b64816249fa1730f8
30 * of
32 * http://gitweb.freedesktop.org/?p=users/joonas/glitter-paths
34 /* Glitter-paths is a stand alone polygon rasteriser derived from
35 * David Turner's reimplementation of Tor Anderssons's 15x17
36 * supersampling rasteriser from the Apparition graphics library. The
37 * main new feature here is cheaply choosing per-scan line between
38 * doing fully analytical coverage computation for an entire row at a
39 * time vs. using a supersampling approach.
41 * David Turner's code can be found at
43 * http://david.freetype.org/rasterizer-shootout/raster-comparison-20070813.tar.bz2
45 * In particular this file incorporates large parts of ftgrays_tor10.h
46 * from raster-comparison-20070813.tar.bz2
48 /* Overview
50 * A scan converter's basic purpose to take polygon edges and convert
51 * them into an RLE compressed A8 mask. This one works in two phases:
52 * gathering edges and generating spans.
54 * 1) As the user feeds the scan converter edges they are vertically
55 * clipped and bucketted into a _polygon_ data structure. The edges
56 * are also snapped from the user's coordinates to the subpixel grid
57 * coordinates used during scan conversion.
59 * user
60 * |
61 * | edges
62 * V
63 * polygon buckets
65 * 2) Generating spans works by performing a vertical sweep of pixel
66 * rows from top to bottom and maintaining an _active_list_ of edges
67 * that intersect the row. From the active list the fill rule
68 * determines which edges are the left and right edges of the start of
69 * each span, and their contribution is then accumulated into a pixel
70 * coverage list (_cell_list_) as coverage deltas. Once the coverage
71 * deltas of all edges are known we can form spans of constant pixel
72 * coverage by summing the deltas during a traversal of the cell list.
73 * At the end of a pixel row the cell list is sent to a coverage
74 * blitter for rendering to some target surface.
76 * The pixel coverages are computed by either supersampling the row
77 * and box filtering a mono rasterisation, or by computing the exact
78 * coverages of edges in the active list. The supersampling method is
79 * used whenever some edge starts or stops within the row or there are
80 * edge intersections in the row.
82 * polygon bucket for \
83 * current pixel row |
84 * | |
85 * | activate new edges | Repeat GRID_Y times if we
86 * V \ are supersampling this row,
87 * active list / or just once if we're computing
88 * | | analytical coverage.
89 * | coverage deltas |
90 * V |
91 * pixel coverage list /
92 * |
93 * V
94 * coverage blitter
96 #include "cairoint.h"
97 #include "cairo-spans-private.h"
98 #include "cairo-error-private.h"
100 #include <stdlib.h>
101 #include <string.h>
102 #include <limits.h>
103 #include <setjmp.h>
105 /*-------------------------------------------------------------------------
106 * cairo specific config
108 #define I static
110 /* Prefer cairo's status type. */
111 #define GLITTER_HAVE_STATUS_T 1
112 #define GLITTER_STATUS_SUCCESS CAIRO_STATUS_SUCCESS
113 #define GLITTER_STATUS_NO_MEMORY CAIRO_STATUS_NO_MEMORY
114 typedef cairo_status_t glitter_status_t;
116 /* The input coordinate scale and the rasterisation grid scales. */
117 #define GLITTER_INPUT_BITS CAIRO_FIXED_FRAC_BITS
118 #define GRID_X_BITS CAIRO_FIXED_FRAC_BITS
119 #define GRID_Y 15
121 /* Set glitter up to use a cairo span renderer to do the coverage
122 * blitting. */
123 struct pool;
124 struct cell_list;
126 /*-------------------------------------------------------------------------
127 * glitter-paths.h
130 /* "Input scaled" numbers are fixed precision reals with multiplier
131 * 2**GLITTER_INPUT_BITS. Input coordinates are given to glitter as
132 * pixel scaled numbers. These get converted to the internal grid
133 * scaled numbers as soon as possible. Internal overflow is possible
134 * if GRID_X/Y inside glitter-paths.c is larger than
135 * 1<<GLITTER_INPUT_BITS. */
136 #ifndef GLITTER_INPUT_BITS
137 # define GLITTER_INPUT_BITS 8
138 #endif
139 #define GLITTER_INPUT_SCALE (1<<GLITTER_INPUT_BITS)
140 typedef int glitter_input_scaled_t;
142 #if !GLITTER_HAVE_STATUS_T
143 typedef enum {
144 GLITTER_STATUS_SUCCESS = 0,
145 GLITTER_STATUS_NO_MEMORY
146 } glitter_status_t;
147 #endif
149 #ifndef I
150 # define I /*static*/
151 #endif
153 /* Opaque type for scan converting. */
154 typedef struct glitter_scan_converter glitter_scan_converter_t;
156 /* Reset a scan converter to accept polygon edges and set the clip box
157 * in pixels. Allocates O(ymax-ymin) bytes of memory. The clip box
158 * is set to integer pixel coordinates xmin <= x < xmax, ymin <= y <
159 * ymax. */
160 I glitter_status_t
161 glitter_scan_converter_reset(
162 glitter_scan_converter_t *converter,
163 int xmin, int ymin,
164 int xmax, int ymax);
166 /* Render the polygon in the scan converter to the given A8 format
167 * image raster. Only the pixels accessible as pixels[y*stride+x] for
168 * x,y inside the clip box are written to, where xmin <= x < xmax,
169 * ymin <= y < ymax. The image is assumed to be clear on input.
171 * If nonzero_fill is true then the interior of the polygon is
172 * computed with the non-zero fill rule. Otherwise the even-odd fill
173 * rule is used.
175 * The scan converter must be reset or destroyed after this call. */
177 /*-------------------------------------------------------------------------
178 * glitter-paths.c: Implementation internal types
180 #include <stdlib.h>
181 #include <string.h>
182 #include <limits.h>
184 /* All polygon coordinates are snapped onto a subsample grid. "Grid
185 * scaled" numbers are fixed precision reals with multiplier GRID_X or
186 * GRID_Y. */
187 typedef int grid_scaled_t;
188 typedef int grid_scaled_x_t;
189 typedef int grid_scaled_y_t;
191 /* Default x/y scale factors.
192 * You can either define GRID_X/Y_BITS to get a power-of-two scale
193 * or define GRID_X/Y separately. */
194 #if !defined(GRID_X) && !defined(GRID_X_BITS)
195 # define GRID_X_BITS 8
196 #endif
197 #if !defined(GRID_Y) && !defined(GRID_Y_BITS)
198 # define GRID_Y 15
199 #endif
201 /* Use GRID_X/Y_BITS to define GRID_X/Y if they're available. */
202 #ifdef GRID_X_BITS
203 # define GRID_X (1 << GRID_X_BITS)
204 #endif
205 #ifdef GRID_Y_BITS
206 # define GRID_Y (1 << GRID_Y_BITS)
207 #endif
209 /* The GRID_X_TO_INT_FRAC macro splits a grid scaled coordinate into
210 * integer and fractional parts. The integer part is floored. */
211 #if defined(GRID_X_TO_INT_FRAC)
212 /* do nothing */
213 #elif defined(GRID_X_BITS)
214 # define GRID_X_TO_INT_FRAC(x, i, f) \
215 _GRID_TO_INT_FRAC_shift(x, i, f, GRID_X_BITS)
216 #else
217 # define GRID_X_TO_INT_FRAC(x, i, f) \
218 _GRID_TO_INT_FRAC_general(x, i, f, GRID_X)
219 #endif
221 #define _GRID_TO_INT_FRAC_general(t, i, f, m) do { \
222 (i) = (t) / (m); \
223 (f) = (t) % (m); \
224 if ((f) < 0) { \
225 --(i); \
226 (f) += (m); \
228 } while (0)
230 #define _GRID_TO_INT_FRAC_shift(t, i, f, b) do { \
231 (f) = (t) & ((1 << (b)) - 1); \
232 (i) = (t) >> (b); \
233 } while (0)
235 /* A grid area is a real in [0,1] scaled by 2*GRID_X*GRID_Y. We want
236 * to be able to represent exactly areas of subpixel trapezoids whose
237 * vertices are given in grid scaled coordinates. The scale factor
238 * comes from needing to accurately represent the area 0.5*dx*dy of a
239 * triangle with base dx and height dy in grid scaled numbers. */
240 #define GRID_XY (2*GRID_X*GRID_Y) /* Unit area on the grid. */
242 /* GRID_AREA_TO_ALPHA(area): map [0,GRID_XY] to [0,255]. */
243 #if GRID_XY == 510
244 # define GRID_AREA_TO_ALPHA(c) (((c)+1) >> 1)
245 #elif GRID_XY == 255
246 # define GRID_AREA_TO_ALPHA(c) (c)
247 #elif GRID_XY == 64
248 # define GRID_AREA_TO_ALPHA(c) (((c) << 2) | -(((c) & 0x40) >> 6))
249 #elif GRID_XY == 128
250 # define GRID_AREA_TO_ALPHA(c) ((((c) << 1) | -((c) >> 7)) & 255)
251 #elif GRID_XY == 256
252 # define GRID_AREA_TO_ALPHA(c) (((c) | -((c) >> 8)) & 255)
253 #elif GRID_XY == 15
254 # define GRID_AREA_TO_ALPHA(c) (((c) << 4) + (c))
255 #elif GRID_XY == 2*256*15
256 # define GRID_AREA_TO_ALPHA(c) (((c) + ((c)<<4) + 256) >> 9)
257 #else
258 # define GRID_AREA_TO_ALPHA(c) (((c)*255 + GRID_XY/2) / GRID_XY)
259 #endif
261 #define UNROLL3(x) x x x
263 struct quorem {
264 int32_t quo;
265 int64_t rem;
268 /* Header for a chunk of memory in a memory pool. */
269 struct _pool_chunk {
270 /* # bytes used in this chunk. */
271 size_t size;
273 /* # bytes total in this chunk */
274 size_t capacity;
276 /* Pointer to the previous chunk or %NULL if this is the sentinel
277 * chunk in the pool header. */
278 struct _pool_chunk *prev_chunk;
280 /* Actual data starts here. Well aligned even for 64 bit types. */
281 int64_t data;
284 /* The int64_t data member of _pool_chunk just exists to enforce alignment,
285 * it shouldn't be included in the allocated size for the struct. */
286 #define SIZEOF_POOL_CHUNK (sizeof(struct _pool_chunk) - sizeof(int64_t))
288 /* A memory pool. This is supposed to be embedded on the stack or
289 * within some other structure. It may optionally be followed by an
290 * embedded array from which requests are fulfilled until
291 * malloc needs to be called to allocate a first real chunk. */
292 struct pool {
293 /* Chunk we're allocating from. */
294 struct _pool_chunk *current;
296 jmp_buf *jmp;
298 /* Free list of previously allocated chunks. All have >= default
299 * capacity. */
300 struct _pool_chunk *first_free;
302 /* The default capacity of a chunk. */
303 size_t default_capacity;
305 /* Header for the sentinel chunk. Directly following the pool
306 * struct should be some space for embedded elements from which
307 * the sentinel chunk allocates from. This is expressed as a char
308 * array so that the 'int64_t data' member of _pool_chunk isn't
309 * included. This way embedding struct pool in other structs works
310 * without wasting space. */
311 char sentinel[SIZEOF_POOL_CHUNK];
314 /* A polygon edge. */
315 struct edge {
316 /* Next in y-bucket or active list. */
317 struct edge *next, *prev;
319 /* The clipped y of the top of the edge. */
320 grid_scaled_y_t ytop;
322 /* Number of subsample rows remaining to scan convert of this
323 * edge. */
324 grid_scaled_y_t height_left;
326 /* Original sign of the edge: +1 for downwards, -1 for upwards
327 * edges. */
328 int dir;
329 int cell;
331 /* Current x coordinate while the edge is on the active
332 * list. Initialised to the x coordinate of the top of the
333 * edge. The quotient is in grid_scaled_x_t units and the
334 * remainder is mod dy in grid_scaled_y_t units.*/
335 struct quorem x;
337 /* Advance of the current x when moving down a subsample line. */
338 struct quorem dxdy;
340 /* Advance of the current x when moving down a full pixel
341 * row. Only initialised when the height of the edge is large
342 * enough that there's a chance the edge could be stepped by a
343 * full row's worth of subsample rows at a time. */
344 struct quorem dxdy_full;
346 /* y2-y1 after orienting the edge downwards. */
347 int64_t dy;
350 #define EDGE_Y_BUCKET_INDEX(y, ymin) (((y) - (ymin))/GRID_Y)
352 /* A collection of sorted and vertically clipped edges of the polygon.
353 * Edges are moved from the polygon to an active list while scan
354 * converting. */
355 struct polygon {
356 /* The vertical clip extents. */
357 grid_scaled_y_t ymin, ymax;
359 /* Array of edges all starting in the same bucket. An edge is put
360 * into bucket EDGE_BUCKET_INDEX(edge->ytop, polygon->ymin) when
361 * it is added to the polygon. */
362 struct edge **y_buckets;
363 struct edge *y_buckets_embedded[64];
365 struct {
366 struct pool base[1];
367 struct edge embedded[32];
368 } edge_pool;
371 /* A cell records the effect on pixel coverage of polygon edges
372 * passing through a pixel. It contains two accumulators of pixel
373 * coverage.
375 * Consider the effects of a polygon edge on the coverage of a pixel
376 * it intersects and that of the following one. The coverage of the
377 * following pixel is the height of the edge multiplied by the width
378 * of the pixel, and the coverage of the pixel itself is the area of
379 * the trapezoid formed by the edge and the right side of the pixel.
381 * +-----------------------+-----------------------+
382 * | | |
383 * | | |
384 * |_______________________|_______________________|
385 * | \...................|.......................|\
386 * | \..................|.......................| |
387 * | \.................|.......................| |
388 * | \....covered.....|.......................| |
389 * | \....area.......|.......................| } covered height
390 * | \..............|.......................| |
391 * |uncovered\.............|.......................| |
392 * | area \............|.......................| |
393 * |___________\...........|.......................|/
394 * | | |
395 * | | |
396 * | | |
397 * +-----------------------+-----------------------+
399 * Since the coverage of the following pixel will always be a multiple
400 * of the width of the pixel, we can store the height of the covered
401 * area instead. The coverage of the pixel itself is the total
402 * coverage minus the area of the uncovered area to the left of the
403 * edge. As it's faster to compute the uncovered area we only store
404 * that and subtract it from the total coverage later when forming
405 * spans to blit.
407 * The heights and areas are signed, with left edges of the polygon
408 * having positive sign and right edges having negative sign. When
409 * two edges intersect they swap their left/rightness so their
410 * contribution above and below the intersection point must be
411 * computed separately. */
412 struct cell {
413 struct cell *next;
414 int x;
415 int16_t uncovered_area;
416 int16_t covered_height;
419 /* A cell list represents the scan line sparsely as cells ordered by
420 * ascending x. It is geared towards scanning the cells in order
421 * using an internal cursor. */
422 struct cell_list {
423 /* Sentinel nodes */
424 struct cell head, tail;
426 /* Cursor state for iterating through the cell list. */
427 struct cell *cursor, *rewind;
429 /* Cells in the cell list are owned by the cell list and are
430 * allocated from this pool. */
431 struct {
432 struct pool base[1];
433 struct cell embedded[32];
434 } cell_pool;
437 struct cell_pair {
438 struct cell *cell1;
439 struct cell *cell2;
442 /* The active list contains edges in the current scan line ordered by
443 * the x-coordinate of the intercept of the edge and the scan line. */
444 struct active_list {
445 /* Leftmost edge on the current scan line. */
446 struct edge head, tail;
448 /* A lower bound on the height of the active edges is used to
449 * estimate how soon some active edge ends. We can't advance the
450 * scan conversion by a full pixel row if an edge ends somewhere
451 * within it. */
452 grid_scaled_y_t min_height;
453 int is_vertical;
456 struct glitter_scan_converter {
457 struct polygon polygon[1];
458 struct active_list active[1];
459 struct cell_list coverages[1];
461 cairo_half_open_span_t *spans;
462 cairo_half_open_span_t spans_embedded[64];
464 /* Clip box. */
465 grid_scaled_x_t xmin, xmax;
466 grid_scaled_y_t ymin, ymax;
469 static struct _pool_chunk *
470 _pool_chunk_init(
471 struct _pool_chunk *p,
472 struct _pool_chunk *prev_chunk,
473 size_t capacity)
475 p->prev_chunk = prev_chunk;
476 p->size = 0;
477 p->capacity = capacity;
478 return p;
481 static struct _pool_chunk *
482 _pool_chunk_create(struct pool *pool, size_t size)
484 struct _pool_chunk *p;
486 p = malloc(SIZEOF_POOL_CHUNK + size);
487 if (unlikely (NULL == p))
488 longjmp (*pool->jmp, _cairo_error (CAIRO_STATUS_NO_MEMORY));
490 return _pool_chunk_init(p, pool->current, size);
493 static void
494 pool_init(struct pool *pool,
495 jmp_buf *jmp,
496 size_t default_capacity,
497 size_t embedded_capacity)
499 pool->jmp = jmp;
500 pool->current = (void*) pool->sentinel;
501 pool->first_free = NULL;
502 pool->default_capacity = default_capacity;
503 _pool_chunk_init(pool->current, NULL, embedded_capacity);
506 static void
507 pool_fini(struct pool *pool)
509 struct _pool_chunk *p = pool->current;
510 do {
511 while (NULL != p) {
512 struct _pool_chunk *prev = p->prev_chunk;
513 if (p != (void *) pool->sentinel)
514 free(p);
515 p = prev;
517 p = pool->first_free;
518 pool->first_free = NULL;
519 } while (NULL != p);
522 /* Satisfy an allocation by first allocating a new large enough chunk
523 * and adding it to the head of the pool's chunk list. This function
524 * is called as a fallback if pool_alloc() couldn't do a quick
525 * allocation from the current chunk in the pool. */
526 static void *
527 _pool_alloc_from_new_chunk(
528 struct pool *pool,
529 size_t size)
531 struct _pool_chunk *chunk;
532 void *obj;
533 size_t capacity;
535 /* If the allocation is smaller than the default chunk size then
536 * try getting a chunk off the free list. Force alloc of a new
537 * chunk for large requests. */
538 capacity = size;
539 chunk = NULL;
540 if (size < pool->default_capacity) {
541 capacity = pool->default_capacity;
542 chunk = pool->first_free;
543 if (chunk) {
544 pool->first_free = chunk->prev_chunk;
545 _pool_chunk_init(chunk, pool->current, chunk->capacity);
549 if (NULL == chunk)
550 chunk = _pool_chunk_create (pool, capacity);
551 pool->current = chunk;
553 obj = ((unsigned char*)&chunk->data + chunk->size);
554 chunk->size += size;
555 return obj;
558 /* Allocate size bytes from the pool. The first allocated address
559 * returned from a pool is aligned to 8 bytes. Subsequent
560 * addresses will maintain alignment as long as multiples of 8 are
561 * allocated. Returns the address of a new memory area or %NULL on
562 * allocation failures. The pool retains ownership of the returned
563 * memory. */
564 inline static void *
565 pool_alloc (struct pool *pool, size_t size)
567 struct _pool_chunk *chunk = pool->current;
569 if (size <= chunk->capacity - chunk->size) {
570 void *obj = ((unsigned char*)&chunk->data + chunk->size);
571 chunk->size += size;
572 return obj;
573 } else {
574 return _pool_alloc_from_new_chunk(pool, size);
578 /* Relinquish all pool_alloced memory back to the pool. */
579 static void
580 pool_reset (struct pool *pool)
582 /* Transfer all used chunks to the chunk free list. */
583 struct _pool_chunk *chunk = pool->current;
584 if (chunk != (void *) pool->sentinel) {
585 while (chunk->prev_chunk != (void *) pool->sentinel) {
586 chunk = chunk->prev_chunk;
588 chunk->prev_chunk = pool->first_free;
589 pool->first_free = pool->current;
591 /* Reset the sentinel as the current chunk. */
592 pool->current = (void *) pool->sentinel;
593 pool->current->size = 0;
596 /* Rewinds the cell list's cursor to the beginning. After rewinding
597 * we're good to cell_list_find() the cell any x coordinate. */
598 inline static void
599 cell_list_rewind (struct cell_list *cells)
601 cells->cursor = &cells->head;
604 inline static void
605 cell_list_maybe_rewind (struct cell_list *cells, int x)
607 if (x < cells->cursor->x) {
608 cells->cursor = cells->rewind;
609 if (x < cells->cursor->x)
610 cells->cursor = &cells->head;
614 inline static void
615 cell_list_set_rewind (struct cell_list *cells)
617 cells->rewind = cells->cursor;
620 static void
621 cell_list_init(struct cell_list *cells, jmp_buf *jmp)
623 pool_init(cells->cell_pool.base, jmp,
624 256*sizeof(struct cell),
625 sizeof(cells->cell_pool.embedded));
626 cells->tail.next = NULL;
627 cells->tail.x = INT_MAX;
628 cells->head.x = INT_MIN;
629 cells->head.next = &cells->tail;
630 cell_list_rewind (cells);
633 static void
634 cell_list_fini(struct cell_list *cells)
636 pool_fini (cells->cell_pool.base);
639 /* Empty the cell list. This is called at the start of every pixel
640 * row. */
641 inline static void
642 cell_list_reset (struct cell_list *cells)
644 cell_list_rewind (cells);
645 cells->head.next = &cells->tail;
646 pool_reset (cells->cell_pool.base);
649 inline static struct cell *
650 cell_list_alloc (struct cell_list *cells,
651 struct cell *tail,
652 int x)
654 struct cell *cell;
656 cell = pool_alloc (cells->cell_pool.base, sizeof (struct cell));
657 cell->next = tail->next;
658 tail->next = cell;
659 cell->x = x;
660 *(uint32_t *)&cell->uncovered_area = 0;
662 return cell;
665 /* Find a cell at the given x-coordinate. Returns %NULL if a new cell
666 * needed to be allocated but couldn't be. Cells must be found with
667 * non-decreasing x-coordinate until the cell list is rewound using
668 * cell_list_rewind(). Ownership of the returned cell is retained by
669 * the cell list. */
670 inline static struct cell *
671 cell_list_find (struct cell_list *cells, int x)
673 struct cell *tail = cells->cursor;
675 if (tail->x == x)
676 return tail;
678 while (1) {
679 UNROLL3({
680 if (tail->next->x > x)
681 break;
682 tail = tail->next;
686 if (tail->x != x)
687 tail = cell_list_alloc (cells, tail, x);
688 return cells->cursor = tail;
692 /* Find two cells at x1 and x2. This is exactly equivalent
693 * to
695 * pair.cell1 = cell_list_find(cells, x1);
696 * pair.cell2 = cell_list_find(cells, x2);
698 * except with less function call overhead. */
699 inline static struct cell_pair
700 cell_list_find_pair(struct cell_list *cells, int x1, int x2)
702 struct cell_pair pair;
704 pair.cell1 = cells->cursor;
705 while (1) {
706 UNROLL3({
707 if (pair.cell1->next->x > x1)
708 break;
709 pair.cell1 = pair.cell1->next;
712 if (pair.cell1->x != x1)
713 pair.cell1 = cell_list_alloc (cells, pair.cell1, x1);
715 pair.cell2 = pair.cell1;
716 while (1) {
717 UNROLL3({
718 if (pair.cell2->next->x > x2)
719 break;
720 pair.cell2 = pair.cell2->next;
723 if (pair.cell2->x != x2)
724 pair.cell2 = cell_list_alloc (cells, pair.cell2, x2);
726 cells->cursor = pair.cell2;
727 return pair;
730 /* Add a subpixel span covering [x1, x2) to the coverage cells. */
731 inline static void
732 cell_list_add_subspan(struct cell_list *cells,
733 grid_scaled_x_t x1,
734 grid_scaled_x_t x2)
736 int ix1, fx1;
737 int ix2, fx2;
739 if (x1 == x2)
740 return;
742 GRID_X_TO_INT_FRAC(x1, ix1, fx1);
743 GRID_X_TO_INT_FRAC(x2, ix2, fx2);
745 if (ix1 != ix2) {
746 struct cell_pair p;
747 p = cell_list_find_pair(cells, ix1, ix2);
748 p.cell1->uncovered_area += 2*fx1;
749 ++p.cell1->covered_height;
750 p.cell2->uncovered_area -= 2*fx2;
751 --p.cell2->covered_height;
752 } else {
753 struct cell *cell = cell_list_find(cells, ix1);
754 cell->uncovered_area += 2*(fx1-fx2);
758 inline static void full_step (struct edge *e)
760 if (e->dy == 0)
761 return;
763 e->x.quo += e->dxdy_full.quo;
764 e->x.rem += e->dxdy_full.rem;
765 if (e->x.rem < 0) {
766 e->x.quo--;
767 e->x.rem += e->dy;
768 } else if (e->x.rem >= e->dy) {
769 ++e->x.quo;
770 e->x.rem -= e->dy;
773 e->cell = e->x.quo + (e->x.rem >= e->dy/2);
777 /* Adds the analytical coverage of an edge crossing the current pixel
778 * row to the coverage cells and advances the edge's x position to the
779 * following row.
781 * This function is only called when we know that during this pixel row:
783 * 1) The relative order of all edges on the active list doesn't
784 * change. In particular, no edges intersect within this row to pixel
785 * precision.
787 * 2) No new edges start in this row.
789 * 3) No existing edges end mid-row.
791 * This function depends on being called with all edges from the
792 * active list in the order they appear on the list (i.e. with
793 * non-decreasing x-coordinate.) */
794 static void
795 cell_list_render_edge(struct cell_list *cells,
796 struct edge *edge,
797 int sign)
799 struct quorem x1, x2;
800 grid_scaled_x_t fx1, fx2;
801 int ix1, ix2;
803 x1 = edge->x;
804 full_step (edge);
805 x2 = edge->x;
807 /* Step back from the sample location (half-subrow) to the pixel origin */
808 if (edge->dy) {
809 x1.quo -= edge->dxdy.quo / 2;
810 x1.rem -= edge->dxdy.rem / 2;
811 if (x1.rem < 0) {
812 --x1.quo;
813 x1.rem += edge->dy;
814 } else if (x1.rem >= edge->dy) {
815 ++x1.quo;
816 x1.rem -= edge->dy;
819 x2.quo -= edge->dxdy.quo / 2;
820 x2.rem -= edge->dxdy.rem / 2;
821 if (x2.rem < 0) {
822 --x2.quo;
823 x2.rem += edge->dy;
824 } else if (x2.rem >= edge->dy) {
825 ++x2.quo;
826 x2.rem -= edge->dy;
830 GRID_X_TO_INT_FRAC(x1.quo, ix1, fx1);
831 GRID_X_TO_INT_FRAC(x2.quo, ix2, fx2);
833 cell_list_maybe_rewind(cells, MIN(ix1, ix2));
835 /* Edge is entirely within a column? */
836 if (ix1 == ix2) {
837 /* We always know that ix1 is >= the cell list cursor in this
838 * case due to the no-intersections precondition. */
839 struct cell *cell = cell_list_find(cells, ix1);
840 cell->covered_height += sign*GRID_Y;
841 cell->uncovered_area += sign*(fx1 + fx2)*GRID_Y;
842 return;
845 /* Orient the edge left-to-right. */
846 if (ix2 < ix1) {
847 struct quorem tx;
848 int t;
850 t = ix1;
851 ix1 = ix2;
852 ix2 = t;
854 t = fx1;
855 fx1 = fx2;
856 fx2 = t;
858 tx = x1;
859 x1 = x2;
860 x2 = tx;
863 /* Add coverage for all pixels [ix1,ix2] on this row crossed
864 * by the edge. */
866 struct cell_pair pair;
867 struct quorem y;
868 int64_t tmp, dx;
869 int y_last;
871 dx = (x2.quo - x1.quo) * edge->dy + (x2.rem - x1.rem);
873 tmp = (ix1 + 1) * GRID_X * edge->dy;
874 tmp -= x1.quo * edge->dy + x1.rem;
875 tmp *= GRID_Y;
877 y.quo = tmp / dx;
878 y.rem = tmp % dx;
880 /* When rendering a previous edge on the active list we may
881 * advance the cell list cursor past the leftmost pixel of the
882 * current edge even though the two edges don't intersect.
883 * e.g. consider two edges going down and rightwards:
885 * --\_+---\_+-----+-----+----
886 * \_ \_ | |
887 * | \_ | \_ | |
888 * | \_| \_| |
889 * | \_ \_ |
890 * ----+-----+-\---+-\---+----
892 * The left edge touches cells past the starting cell of the
893 * right edge. Fortunately such cases are rare.
896 pair = cell_list_find_pair(cells, ix1, ix1+1);
897 pair.cell1->uncovered_area += sign*y.quo*(GRID_X + fx1);
898 pair.cell1->covered_height += sign*y.quo;
899 y_last = y.quo;
901 if (ix1+1 < ix2) {
902 struct cell *cell = pair.cell2;
903 struct quorem dydx_full;
905 dydx_full.quo = GRID_Y * GRID_X * edge->dy / dx;
906 dydx_full.rem = GRID_Y * GRID_X * edge->dy % dx;
908 ++ix1;
909 do {
910 y.quo += dydx_full.quo;
911 y.rem += dydx_full.rem;
912 if (y.rem >= dx) {
913 y.quo++;
914 y.rem -= dx;
917 cell->uncovered_area += sign*(y.quo - y_last)*GRID_X;
918 cell->covered_height += sign*(y.quo - y_last);
919 y_last = y.quo;
921 ++ix1;
922 cell = cell_list_find(cells, ix1);
923 } while (ix1 != ix2);
925 pair.cell2 = cell;
927 pair.cell2->uncovered_area += sign*(GRID_Y - y_last)*fx2;
928 pair.cell2->covered_height += sign*(GRID_Y - y_last);
932 static void
933 polygon_init (struct polygon *polygon, jmp_buf *jmp)
935 polygon->ymin = polygon->ymax = 0;
936 polygon->y_buckets = polygon->y_buckets_embedded;
937 pool_init (polygon->edge_pool.base, jmp,
938 8192 - sizeof (struct _pool_chunk),
939 sizeof (polygon->edge_pool.embedded));
942 static void
943 polygon_fini (struct polygon *polygon)
945 if (polygon->y_buckets != polygon->y_buckets_embedded)
946 free (polygon->y_buckets);
948 pool_fini (polygon->edge_pool.base);
951 /* Empties the polygon of all edges. The polygon is then prepared to
952 * receive new edges and clip them to the vertical range
953 * [ymin,ymax). */
954 static glitter_status_t
955 polygon_reset (struct polygon *polygon,
956 grid_scaled_y_t ymin,
957 grid_scaled_y_t ymax)
959 unsigned h = ymax - ymin;
960 unsigned num_buckets = EDGE_Y_BUCKET_INDEX(ymax + GRID_Y-1, ymin);
962 pool_reset(polygon->edge_pool.base);
964 if (unlikely (h > 0x7FFFFFFFU - GRID_Y))
965 goto bail_no_mem; /* even if you could, you wouldn't want to. */
967 if (polygon->y_buckets != polygon->y_buckets_embedded)
968 free (polygon->y_buckets);
970 polygon->y_buckets = polygon->y_buckets_embedded;
971 if (num_buckets > ARRAY_LENGTH (polygon->y_buckets_embedded)) {
972 polygon->y_buckets = _cairo_malloc_ab (num_buckets,
973 sizeof (struct edge *));
974 if (unlikely (NULL == polygon->y_buckets))
975 goto bail_no_mem;
977 memset (polygon->y_buckets, 0, num_buckets * sizeof (struct edge *));
979 polygon->ymin = ymin;
980 polygon->ymax = ymax;
981 return GLITTER_STATUS_SUCCESS;
983 bail_no_mem:
984 polygon->ymin = 0;
985 polygon->ymax = 0;
986 return GLITTER_STATUS_NO_MEMORY;
989 static void
990 _polygon_insert_edge_into_its_y_bucket(struct polygon *polygon,
991 struct edge *e)
993 unsigned ix = EDGE_Y_BUCKET_INDEX(e->ytop, polygon->ymin);
994 struct edge **ptail = &polygon->y_buckets[ix];
995 e->next = *ptail;
996 *ptail = e;
999 static void
1000 active_list_reset (struct active_list *active)
1002 active->head.height_left = INT_MAX;
1003 active->head.dy = 0;
1004 active->head.cell = INT_MIN;
1005 active->head.prev = NULL;
1006 active->head.next = &active->tail;
1007 active->tail.prev = &active->head;
1008 active->tail.next = NULL;
1009 active->tail.cell = INT_MAX;
1010 active->tail.height_left = INT_MAX;
1011 active->tail.dy = 0;
1012 active->min_height = 0;
1013 active->is_vertical = 1;
1016 static void
1017 active_list_init(struct active_list *active)
1019 active_list_reset(active);
1023 * Merge two sorted edge lists.
1024 * Input:
1025 * - head_a: The head of the first list.
1026 * - head_b: The head of the second list; head_b cannot be NULL.
1027 * Output:
1028 * Returns the head of the merged list.
1030 * Implementation notes:
1031 * To make it fast (in particular, to reduce to an insertion sort whenever
1032 * one of the two input lists only has a single element) we iterate through
1033 * a list until its head becomes greater than the head of the other list,
1034 * then we switch their roles. As soon as one of the two lists is empty, we
1035 * just attach the other one to the current list and exit.
1036 * Writes to memory are only needed to "switch" lists (as it also requires
1037 * attaching to the output list the list which we will be iterating next) and
1038 * to attach the last non-empty list.
1040 static struct edge *
1041 merge_sorted_edges (struct edge *head_a, struct edge *head_b)
1043 struct edge *head, **next, *prev;
1044 int32_t x;
1046 prev = head_a->prev;
1047 next = &head;
1048 if (head_a->cell <= head_b->cell) {
1049 head = head_a;
1050 } else {
1051 head = head_b;
1052 head_b->prev = prev;
1053 goto start_with_b;
1056 do {
1057 x = head_b->cell;
1058 while (head_a != NULL && head_a->cell <= x) {
1059 prev = head_a;
1060 next = &head_a->next;
1061 head_a = head_a->next;
1064 head_b->prev = prev;
1065 *next = head_b;
1066 if (head_a == NULL)
1067 return head;
1069 start_with_b:
1070 x = head_a->cell;
1071 while (head_b != NULL && head_b->cell <= x) {
1072 prev = head_b;
1073 next = &head_b->next;
1074 head_b = head_b->next;
1077 head_a->prev = prev;
1078 *next = head_a;
1079 if (head_b == NULL)
1080 return head;
1081 } while (1);
1085 * Sort (part of) a list.
1086 * Input:
1087 * - list: The list to be sorted; list cannot be NULL.
1088 * - limit: Recursion limit.
1089 * Output:
1090 * - head_out: The head of the sorted list containing the first 2^(level+1) elements of the
1091 * input list; if the input list has fewer elements, head_out be a sorted list
1092 * containing all the elements of the input list.
1093 * Returns the head of the list of unprocessed elements (NULL if the sorted list contains
1094 * all the elements of the input list).
1096 * Implementation notes:
1097 * Special case single element list, unroll/inline the sorting of the first two elements.
1098 * Some tail recursion is used since we iterate on the bottom-up solution of the problem
1099 * (we start with a small sorted list and keep merging other lists of the same size to it).
1101 static struct edge *
1102 sort_edges (struct edge *list,
1103 unsigned int level,
1104 struct edge **head_out)
1106 struct edge *head_other, *remaining;
1107 unsigned int i;
1109 head_other = list->next;
1111 if (head_other == NULL) {
1112 *head_out = list;
1113 return NULL;
1116 remaining = head_other->next;
1117 if (list->cell <= head_other->cell) {
1118 *head_out = list;
1119 head_other->next = NULL;
1120 } else {
1121 *head_out = head_other;
1122 head_other->prev = list->prev;
1123 head_other->next = list;
1124 list->prev = head_other;
1125 list->next = NULL;
1128 for (i = 0; i < level && remaining; i++) {
1129 remaining = sort_edges (remaining, i, &head_other);
1130 *head_out = merge_sorted_edges (*head_out, head_other);
1133 return remaining;
1136 static struct edge *
1137 merge_unsorted_edges (struct edge *head, struct edge *unsorted)
1139 sort_edges (unsorted, UINT_MAX, &unsorted);
1140 return merge_sorted_edges (head, unsorted);
1143 /* Test if the edges on the active list can be safely advanced by a
1144 * full row without intersections or any edges ending. */
1145 inline static int
1146 can_do_full_row (struct active_list *active)
1148 const struct edge *e;
1149 int prev_x = INT_MIN;
1151 /* Recomputes the minimum height of all edges on the active
1152 * list if we have been dropping edges. */
1153 if (active->min_height <= 0) {
1154 int min_height = INT_MAX;
1155 int is_vertical = 1;
1157 e = active->head.next;
1158 while (NULL != e) {
1159 if (e->height_left < min_height)
1160 min_height = e->height_left;
1161 is_vertical &= e->dy == 0;
1162 e = e->next;
1165 active->is_vertical = is_vertical;
1166 active->min_height = min_height;
1169 if (active->min_height < GRID_Y)
1170 return 0;
1172 /* Check for intersections as no edges end during the next row. */
1173 for (e = active->head.next; e != &active->tail; e = e->next) {
1174 int cell;
1176 if (e->dy) {
1177 struct quorem x = e->x;
1178 x.quo += e->dxdy_full.quo;
1179 x.rem += e->dxdy_full.rem;
1180 if (x.rem < 0) {
1181 x.quo--;
1182 x.rem += e->dy;
1183 } else if (x.rem >= e->dy) {
1184 x.quo++;
1185 x.rem -= e->dy;
1187 cell = x.quo + (x.rem >= e->dy/2);
1188 } else
1189 cell = e->cell;
1191 if (cell < prev_x)
1192 return 0;
1194 prev_x = cell;
1197 return 1;
1200 /* Merges edges on the given subpixel row from the polygon to the
1201 * active_list. */
1202 inline static void
1203 active_list_merge_edges_from_bucket(struct active_list *active,
1204 struct edge *edges)
1206 active->head.next = merge_unsorted_edges (active->head.next, edges);
1209 inline static int
1210 polygon_fill_buckets (struct active_list *active,
1211 struct edge *edge,
1212 int y,
1213 struct edge **buckets)
1215 grid_scaled_y_t min_height = active->min_height;
1216 int is_vertical = active->is_vertical;
1217 int max_suby = 0;
1219 while (edge) {
1220 struct edge *next = edge->next;
1221 int suby = edge->ytop - y;
1222 if (buckets[suby])
1223 buckets[suby]->prev = edge;
1224 edge->next = buckets[suby];
1225 edge->prev = NULL;
1226 buckets[suby] = edge;
1227 if (edge->height_left < min_height)
1228 min_height = edge->height_left;
1229 is_vertical &= edge->dy == 0;
1230 edge = next;
1231 if (suby > max_suby)
1232 max_suby = suby;
1235 active->is_vertical = is_vertical;
1236 active->min_height = min_height;
1238 return max_suby;
1241 static void step (struct edge *edge)
1243 if (edge->dy == 0)
1244 return;
1246 edge->x.quo += edge->dxdy.quo;
1247 edge->x.rem += edge->dxdy.rem;
1248 if (edge->x.rem < 0) {
1249 --edge->x.quo;
1250 edge->x.rem += edge->dy;
1251 } else if (edge->x.rem >= edge->dy) {
1252 ++edge->x.quo;
1253 edge->x.rem -= edge->dy;
1256 edge->cell = edge->x.quo + (edge->x.rem >= edge->dy/2);
1259 inline static void
1260 sub_row (struct active_list *active,
1261 struct cell_list *coverages,
1262 unsigned int mask)
1264 struct edge *edge = active->head.next;
1265 int xstart = INT_MIN, prev_x = INT_MIN;
1266 int winding = 0;
1268 cell_list_rewind (coverages);
1270 while (&active->tail != edge) {
1271 struct edge *next = edge->next;
1272 int xend = edge->cell;
1274 if (--edge->height_left) {
1275 step (edge);
1277 if (edge->cell < prev_x) {
1278 struct edge *pos = edge->prev;
1279 pos->next = next;
1280 next->prev = pos;
1281 do {
1282 pos = pos->prev;
1283 } while (edge->cell < pos->cell);
1284 pos->next->prev = edge;
1285 edge->next = pos->next;
1286 edge->prev = pos;
1287 pos->next = edge;
1288 } else
1289 prev_x = edge->cell;
1290 active->min_height = -1;
1291 } else {
1292 edge->prev->next = next;
1293 next->prev = edge->prev;
1296 winding += edge->dir;
1297 if ((winding & mask) == 0) {
1298 if (next->cell != xend) {
1299 cell_list_add_subspan (coverages, xstart, xend);
1300 xstart = INT_MIN;
1302 } else if (xstart == INT_MIN)
1303 xstart = xend;
1305 edge = next;
1309 inline static void dec (struct active_list *a, struct edge *e, int h)
1311 e->height_left -= h;
1312 if (e->height_left == 0) {
1313 e->prev->next = e->next;
1314 e->next->prev = e->prev;
1315 a->min_height = -1;
1319 static void
1320 full_row (struct active_list *active,
1321 struct cell_list *coverages,
1322 unsigned int mask)
1324 struct edge *left = active->head.next;
1326 while (&active->tail != left) {
1327 struct edge *right;
1328 int winding;
1330 dec (active, left, GRID_Y);
1332 winding = left->dir;
1333 right = left->next;
1334 do {
1335 dec (active, right, GRID_Y);
1337 winding += right->dir;
1338 if ((winding & mask) == 0 && right->next->cell != right->cell)
1339 break;
1341 full_step (right);
1343 right = right->next;
1344 } while (1);
1346 cell_list_set_rewind (coverages);
1347 cell_list_render_edge (coverages, left, +1);
1348 cell_list_render_edge (coverages, right, -1);
1350 left = right->next;
1354 static void
1355 _glitter_scan_converter_init(glitter_scan_converter_t *converter, jmp_buf *jmp)
1357 polygon_init(converter->polygon, jmp);
1358 active_list_init(converter->active);
1359 cell_list_init(converter->coverages, jmp);
1360 converter->xmin=0;
1361 converter->ymin=0;
1362 converter->xmax=0;
1363 converter->ymax=0;
1366 static void
1367 _glitter_scan_converter_fini(glitter_scan_converter_t *self)
1369 if (self->spans != self->spans_embedded)
1370 free (self->spans);
1372 polygon_fini(self->polygon);
1373 cell_list_fini(self->coverages);
1375 self->xmin=0;
1376 self->ymin=0;
1377 self->xmax=0;
1378 self->ymax=0;
1381 static grid_scaled_t
1382 int_to_grid_scaled(int i, int scale)
1384 /* Clamp to max/min representable scaled number. */
1385 if (i >= 0) {
1386 if (i >= INT_MAX/scale)
1387 i = INT_MAX/scale;
1389 else {
1390 if (i <= INT_MIN/scale)
1391 i = INT_MIN/scale;
1393 return i*scale;
1396 #define int_to_grid_scaled_x(x) int_to_grid_scaled((x), GRID_X)
1397 #define int_to_grid_scaled_y(x) int_to_grid_scaled((x), GRID_Y)
1399 I glitter_status_t
1400 glitter_scan_converter_reset(
1401 glitter_scan_converter_t *converter,
1402 int xmin, int ymin,
1403 int xmax, int ymax)
1405 glitter_status_t status;
1406 int max_num_spans;
1408 converter->xmin = 0; converter->xmax = 0;
1409 converter->ymin = 0; converter->ymax = 0;
1411 max_num_spans = xmax - xmin + 1;
1413 if (max_num_spans > ARRAY_LENGTH(converter->spans_embedded)) {
1414 converter->spans = _cairo_malloc_ab (max_num_spans,
1415 sizeof (cairo_half_open_span_t));
1416 if (unlikely (converter->spans == NULL))
1417 return _cairo_error (CAIRO_STATUS_NO_MEMORY);
1418 } else
1419 converter->spans = converter->spans_embedded;
1421 xmin = int_to_grid_scaled_x(xmin);
1422 ymin = int_to_grid_scaled_y(ymin);
1423 xmax = int_to_grid_scaled_x(xmax);
1424 ymax = int_to_grid_scaled_y(ymax);
1426 active_list_reset(converter->active);
1427 cell_list_reset(converter->coverages);
1428 status = polygon_reset(converter->polygon, ymin, ymax);
1429 if (status)
1430 return status;
1432 converter->xmin = xmin;
1433 converter->xmax = xmax;
1434 converter->ymin = ymin;
1435 converter->ymax = ymax;
1436 return GLITTER_STATUS_SUCCESS;
1439 /* INPUT_TO_GRID_X/Y (in_coord, out_grid_scaled, grid_scale)
1440 * These macros convert an input coordinate in the client's
1441 * device space to the rasterisation grid.
1443 /* Gah.. this bit of ugly defines INPUT_TO_GRID_X/Y so as to use
1444 * shifts if possible, and something saneish if not.
1446 #if !defined(INPUT_TO_GRID_Y) && defined(GRID_Y_BITS) && GRID_Y_BITS <= GLITTER_INPUT_BITS
1447 # define INPUT_TO_GRID_Y(in, out) (out) = (in) >> (GLITTER_INPUT_BITS - GRID_Y_BITS)
1448 #else
1449 # define INPUT_TO_GRID_Y(in, out) INPUT_TO_GRID_general(in, out, GRID_Y)
1450 #endif
1452 #if !defined(INPUT_TO_GRID_X) && defined(GRID_X_BITS) && GRID_X_BITS <= GLITTER_INPUT_BITS
1453 # define INPUT_TO_GRID_X(in, out) (out) = (in) >> (GLITTER_INPUT_BITS - GRID_X_BITS)
1454 #else
1455 # define INPUT_TO_GRID_X(in, out) INPUT_TO_GRID_general(in, out, GRID_X)
1456 #endif
1458 #define INPUT_TO_GRID_general(in, out, grid_scale) do { \
1459 long long tmp__ = (long long)(grid_scale) * (in); \
1460 tmp__ += 1 << (GLITTER_INPUT_BITS-1); \
1461 tmp__ >>= GLITTER_INPUT_BITS; \
1462 (out) = tmp__; \
1463 } while (0)
1465 inline static void
1466 polygon_add_edge (struct polygon *polygon,
1467 const cairo_edge_t *edge)
1469 struct edge *e;
1470 grid_scaled_y_t ytop, ybot;
1471 const cairo_point_t *p1, *p2;
1473 INPUT_TO_GRID_Y (edge->top, ytop);
1474 if (ytop < polygon->ymin)
1475 ytop = polygon->ymin;
1477 INPUT_TO_GRID_Y (edge->bottom, ybot);
1478 if (ybot > polygon->ymax)
1479 ybot = polygon->ymax;
1481 if (ybot <= ytop)
1482 return;
1484 e = pool_alloc (polygon->edge_pool.base, sizeof (struct edge));
1486 e->ytop = ytop;
1487 e->height_left = ybot - ytop;
1488 if (edge->line.p2.y > edge->line.p1.y) {
1489 e->dir = edge->dir;
1490 p1 = &edge->line.p1;
1491 p2 = &edge->line.p2;
1492 } else {
1493 e->dir = -edge->dir;
1494 p1 = &edge->line.p2;
1495 p2 = &edge->line.p1;
1498 if (p2->x == p1->x) {
1499 e->cell = p1->x;
1500 e->x.quo = p1->x;
1501 e->x.rem = 0;
1502 e->dxdy.quo = e->dxdy.rem = 0;
1503 e->dxdy_full.quo = e->dxdy_full.rem = 0;
1504 e->dy = 0;
1505 } else {
1506 int64_t Ex, Ey, tmp;
1508 Ex = (int64_t)(p2->x - p1->x) * GRID_X;
1509 Ey = (int64_t)(p2->y - p1->y) * GRID_Y * (2 << GLITTER_INPUT_BITS);
1511 e->dxdy.quo = Ex * (2 << GLITTER_INPUT_BITS) / Ey;
1512 e->dxdy.rem = Ex * (2 << GLITTER_INPUT_BITS) % Ey;
1514 tmp = (int64_t)(2*ytop + 1) << GLITTER_INPUT_BITS;
1515 tmp -= (int64_t)p1->y * GRID_Y * 2;
1516 tmp *= Ex;
1517 e->x.quo = tmp / Ey;
1518 e->x.rem = tmp % Ey;
1520 #if GRID_X_BITS == GLITTER_INPUT_BITS
1521 e->x.quo += p1->x;
1522 #else
1523 tmp = (int64_t)p1->x * GRID_X;
1524 e->x.quo += tmp >> GLITTER_INPUT_BITS;
1525 e->x.rem += ((tmp & ((1 << GLITTER_INPUT_BITS) - 1)) * Ey) / (1 << GLITTER_INPUT_BITS);
1526 #endif
1528 if (e->x.rem < 0) {
1529 e->x.quo--;
1530 e->x.rem += Ey;
1531 } else if (e->x.rem >= Ey) {
1532 e->x.quo++;
1533 e->x.rem -= Ey;
1536 if (e->height_left >= GRID_Y) {
1537 tmp = Ex * (2 * GRID_Y << GLITTER_INPUT_BITS);
1538 e->dxdy_full.quo = tmp / Ey;
1539 e->dxdy_full.rem = tmp % Ey;
1540 } else
1541 e->dxdy_full.quo = e->dxdy_full.rem = 0;
1543 e->cell = e->x.quo + (e->x.rem >= Ey/2);
1544 e->dy = Ey;
1547 _polygon_insert_edge_into_its_y_bucket (polygon, e);
1550 /* Add a new polygon edge from pixel (x1,y1) to (x2,y2) to the scan
1551 * converter. The coordinates represent pixel positions scaled by
1552 * 2**GLITTER_PIXEL_BITS. If this function fails then the scan
1553 * converter should be reset or destroyed. Dir must be +1 or -1,
1554 * with the latter reversing the orientation of the edge. */
1555 I void
1556 glitter_scan_converter_add_edge (glitter_scan_converter_t *converter,
1557 const cairo_edge_t *edge)
1559 polygon_add_edge (converter->polygon, edge);
1562 static void
1563 step_edges (struct active_list *active, int count)
1565 struct edge *edge;
1567 count *= GRID_Y;
1568 for (edge = active->head.next; edge != &active->tail; edge = edge->next) {
1569 edge->height_left -= count;
1570 if (! edge->height_left) {
1571 edge->prev->next = edge->next;
1572 edge->next->prev = edge->prev;
1573 active->min_height = -1;
1578 static glitter_status_t
1579 blit_a8 (struct cell_list *cells,
1580 cairo_span_renderer_t *renderer,
1581 cairo_half_open_span_t *spans,
1582 int y, int height,
1583 int xmin, int xmax)
1585 struct cell *cell = cells->head.next;
1586 int prev_x = xmin, last_x = -1;
1587 int16_t cover = 0, last_cover = 0;
1588 unsigned num_spans;
1590 if (cell == &cells->tail)
1591 return CAIRO_STATUS_SUCCESS;
1593 /* Skip cells to the left of the clip region. */
1594 while (cell->x < xmin) {
1595 cover += cell->covered_height;
1596 cell = cell->next;
1598 cover *= GRID_X*2;
1600 /* Form the spans from the coverages and areas. */
1601 num_spans = 0;
1602 for (; cell->x < xmax; cell = cell->next) {
1603 int x = cell->x;
1604 int16_t area;
1606 if (x > prev_x && cover != last_cover) {
1607 spans[num_spans].x = prev_x;
1608 spans[num_spans].coverage = GRID_AREA_TO_ALPHA (cover);
1609 last_cover = cover;
1610 last_x = prev_x;
1611 ++num_spans;
1614 cover += cell->covered_height*GRID_X*2;
1615 area = cover - cell->uncovered_area;
1617 if (area != last_cover) {
1618 spans[num_spans].x = x;
1619 spans[num_spans].coverage = GRID_AREA_TO_ALPHA (area);
1620 last_cover = area;
1621 last_x = x;
1622 ++num_spans;
1625 prev_x = x+1;
1628 if (prev_x <= xmax && cover != last_cover) {
1629 spans[num_spans].x = prev_x;
1630 spans[num_spans].coverage = GRID_AREA_TO_ALPHA (cover);
1631 last_cover = cover;
1632 last_x = prev_x;
1633 ++num_spans;
1636 if (last_x < xmax && last_cover) {
1637 spans[num_spans].x = xmax;
1638 spans[num_spans].coverage = 0;
1639 ++num_spans;
1642 /* Dump them into the renderer. */
1643 return renderer->render_rows (renderer, y, height, spans, num_spans);
1646 #define GRID_AREA_TO_A1(A) ((GRID_AREA_TO_ALPHA (A) > 127) ? 255 : 0)
1647 static glitter_status_t
1648 blit_a1 (struct cell_list *cells,
1649 cairo_span_renderer_t *renderer,
1650 cairo_half_open_span_t *spans,
1651 int y, int height,
1652 int xmin, int xmax)
1654 struct cell *cell = cells->head.next;
1655 int prev_x = xmin, last_x = -1;
1656 int16_t cover = 0;
1657 uint8_t coverage, last_cover = 0;
1658 unsigned num_spans;
1660 if (cell == &cells->tail)
1661 return CAIRO_STATUS_SUCCESS;
1663 /* Skip cells to the left of the clip region. */
1664 while (cell->x < xmin) {
1665 cover += cell->covered_height;
1666 cell = cell->next;
1668 cover *= GRID_X*2;
1670 /* Form the spans from the coverages and areas. */
1671 num_spans = 0;
1672 for (; cell->x < xmax; cell = cell->next) {
1673 int x = cell->x;
1674 int16_t area;
1676 coverage = GRID_AREA_TO_A1 (cover);
1677 if (x > prev_x && coverage != last_cover) {
1678 last_x = spans[num_spans].x = prev_x;
1679 last_cover = spans[num_spans].coverage = coverage;
1680 ++num_spans;
1683 cover += cell->covered_height*GRID_X*2;
1684 area = cover - cell->uncovered_area;
1686 coverage = GRID_AREA_TO_A1 (area);
1687 if (coverage != last_cover) {
1688 last_x = spans[num_spans].x = x;
1689 last_cover = spans[num_spans].coverage = coverage;
1690 ++num_spans;
1693 prev_x = x+1;
1696 coverage = GRID_AREA_TO_A1 (cover);
1697 if (prev_x <= xmax && coverage != last_cover) {
1698 last_x = spans[num_spans].x = prev_x;
1699 last_cover = spans[num_spans].coverage = coverage;
1700 ++num_spans;
1703 if (last_x < xmax && last_cover) {
1704 spans[num_spans].x = xmax;
1705 spans[num_spans].coverage = 0;
1706 ++num_spans;
1708 if (num_spans == 1)
1709 return CAIRO_STATUS_SUCCESS;
1711 /* Dump them into the renderer. */
1712 return renderer->render_rows (renderer, y, height, spans, num_spans);
1716 I void
1717 glitter_scan_converter_render(glitter_scan_converter_t *converter,
1718 unsigned int winding_mask,
1719 int antialias,
1720 cairo_span_renderer_t *renderer)
1722 int i, j;
1723 int ymax_i = converter->ymax / GRID_Y;
1724 int ymin_i = converter->ymin / GRID_Y;
1725 int xmin_i, xmax_i;
1726 int h = ymax_i - ymin_i;
1727 struct polygon *polygon = converter->polygon;
1728 struct cell_list *coverages = converter->coverages;
1729 struct active_list *active = converter->active;
1730 struct edge *buckets[GRID_Y] = { 0 };
1732 xmin_i = converter->xmin / GRID_X;
1733 xmax_i = converter->xmax / GRID_X;
1734 if (xmin_i >= xmax_i)
1735 return;
1737 /* Render each pixel row. */
1738 for (i = 0; i < h; i = j) {
1739 int do_full_row = 0;
1741 j = i + 1;
1743 /* Determine if we can ignore this row or use the full pixel
1744 * stepper. */
1745 if (polygon_fill_buckets (active,
1746 polygon->y_buckets[i],
1747 (i+ymin_i)*GRID_Y,
1748 buckets) == 0) {
1749 if (buckets[0]) {
1750 active_list_merge_edges_from_bucket (active, buckets[0]);
1751 buckets[0] = NULL;
1754 if (active->head.next == &active->tail) {
1755 active->min_height = INT_MAX;
1756 active->is_vertical = 1;
1757 for (; j < h && ! polygon->y_buckets[j]; j++)
1759 continue;
1762 do_full_row = can_do_full_row (active);
1765 if (do_full_row) {
1766 /* Step by a full pixel row's worth. */
1767 full_row (active, coverages, winding_mask);
1769 if (active->is_vertical) {
1770 while (j < h &&
1771 polygon->y_buckets[j] == NULL &&
1772 active->min_height >= 2*GRID_Y)
1774 active->min_height -= GRID_Y;
1775 j++;
1777 if (j != i + 1)
1778 step_edges (active, j - (i + 1));
1780 } else {
1781 int sub;
1783 /* Subsample this row. */
1784 for (sub = 0; sub < GRID_Y; sub++) {
1785 if (buckets[sub]) {
1786 active_list_merge_edges_from_bucket (active, buckets[sub]);
1787 buckets[sub] = NULL;
1789 sub_row (active, coverages, winding_mask);
1793 if (antialias)
1794 blit_a8 (coverages, renderer, converter->spans,
1795 i+ymin_i, j-i, xmin_i, xmax_i);
1796 else
1797 blit_a1 (coverages, renderer, converter->spans,
1798 i+ymin_i, j-i, xmin_i, xmax_i);
1799 cell_list_reset (coverages);
1801 active->min_height -= GRID_Y;
1805 struct _cairo_tor_scan_converter {
1806 cairo_scan_converter_t base;
1808 glitter_scan_converter_t converter[1];
1809 cairo_fill_rule_t fill_rule;
1810 cairo_antialias_t antialias;
1812 jmp_buf jmp;
1815 typedef struct _cairo_tor_scan_converter cairo_tor_scan_converter_t;
1817 static void
1818 _cairo_tor_scan_converter_destroy (void *converter)
1820 cairo_tor_scan_converter_t *self = converter;
1821 if (self == NULL) {
1822 return;
1824 _glitter_scan_converter_fini (self->converter);
1825 free(self);
1828 cairo_status_t
1829 _cairo_tor_scan_converter_add_polygon (void *converter,
1830 const cairo_polygon_t *polygon)
1832 cairo_tor_scan_converter_t *self = converter;
1833 int i;
1835 #if 0
1836 FILE *file = fopen ("polygon.txt", "w");
1837 _cairo_debug_print_polygon (file, polygon);
1838 fclose (file);
1839 #endif
1841 for (i = 0; i < polygon->num_edges; i++)
1842 glitter_scan_converter_add_edge (self->converter, &polygon->edges[i]);
1844 return CAIRO_STATUS_SUCCESS;
1847 static cairo_status_t
1848 _cairo_tor_scan_converter_generate (void *converter,
1849 cairo_span_renderer_t *renderer)
1851 cairo_tor_scan_converter_t *self = converter;
1852 cairo_status_t status;
1854 if ((status = setjmp (self->jmp)))
1855 return _cairo_scan_converter_set_error (self, _cairo_error (status));
1857 glitter_scan_converter_render (self->converter,
1858 self->fill_rule == CAIRO_FILL_RULE_WINDING ? ~0 : 1,
1859 self->antialias != CAIRO_ANTIALIAS_NONE,
1860 renderer);
1861 return CAIRO_STATUS_SUCCESS;
1864 cairo_scan_converter_t *
1865 _cairo_tor_scan_converter_create (int xmin,
1866 int ymin,
1867 int xmax,
1868 int ymax,
1869 cairo_fill_rule_t fill_rule,
1870 cairo_antialias_t antialias)
1872 cairo_tor_scan_converter_t *self;
1873 cairo_status_t status;
1875 self = malloc (sizeof(struct _cairo_tor_scan_converter));
1876 if (unlikely (self == NULL)) {
1877 status = _cairo_error (CAIRO_STATUS_NO_MEMORY);
1878 goto bail_nomem;
1881 self->base.destroy = _cairo_tor_scan_converter_destroy;
1882 self->base.generate = _cairo_tor_scan_converter_generate;
1884 _glitter_scan_converter_init (self->converter, &self->jmp);
1885 status = glitter_scan_converter_reset (self->converter,
1886 xmin, ymin, xmax, ymax);
1887 if (unlikely (status))
1888 goto bail;
1890 self->fill_rule = fill_rule;
1891 self->antialias = antialias;
1893 return &self->base;
1895 bail:
1896 self->base.destroy(&self->base);
1897 bail_nomem:
1898 return _cairo_scan_converter_create_in_error (status);