Rev 3280 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
3280 | Serge | 1 | /* |
2 | * Copyright © 2012 Intel Corporation |
||
3 | * |
||
4 | * Permission is hereby granted, free of charge, to any person obtaining a |
||
5 | * copy of this software and associated documentation files (the "Software"), |
||
6 | * to deal in the Software without restriction, including without limitation |
||
7 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
||
8 | * and/or sell copies of the Software, and to permit persons to whom the |
||
9 | * Software is furnished to do so, subject to the following conditions: |
||
10 | * |
||
11 | * The above copyright notice and this permission notice (including the next |
||
12 | * paragraph) shall be included in all copies or substantial portions of the |
||
13 | * Software. |
||
14 | * |
||
15 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
||
16 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
||
17 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
||
18 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
||
19 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
||
20 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
||
21 | * SOFTWARE. |
||
22 | * |
||
23 | * Authors: |
||
24 | * Chris Wilson |
||
25 | * |
||
26 | */ |
||
27 | |||
28 | #ifdef HAVE_CONFIG_H |
||
29 | #include "config.h" |
||
30 | #endif |
||
31 | |||
32 | #include "sna.h" |
||
33 | #include "sna_render.h" |
||
34 | #include "sna_render_inline.h" |
||
35 | #include "gen4_vertex.h" |
||
36 | |||
4251 | Serge | 37 | #ifndef sse2 |
38 | #define sse2 |
||
39 | #endif |
||
40 | |||
3280 | Serge | 41 | void gen4_vertex_flush(struct sna *sna) |
42 | { |
||
43 | DBG(("%s[%x] = %d\n", __FUNCTION__, |
||
44 | 4*sna->render.vertex_offset, |
||
45 | sna->render.vertex_index - sna->render.vertex_start)); |
||
46 | |||
47 | assert(sna->render.vertex_offset); |
||
48 | assert(sna->render.vertex_index > sna->render.vertex_start); |
||
49 | |||
50 | sna->kgem.batch[sna->render.vertex_offset] = |
||
51 | sna->render.vertex_index - sna->render.vertex_start; |
||
52 | sna->render.vertex_offset = 0; |
||
53 | } |
||
54 | |||
55 | int gen4_vertex_finish(struct sna *sna) |
||
56 | { |
||
57 | struct kgem_bo *bo; |
||
58 | unsigned int i; |
||
59 | unsigned hint, size; |
||
60 | |||
61 | DBG(("%s: used=%d / %d\n", __FUNCTION__, |
||
62 | sna->render.vertex_used, sna->render.vertex_size)); |
||
63 | assert(sna->render.vertex_offset == 0); |
||
64 | assert(sna->render.vertex_used); |
||
65 | |||
66 | sna_vertex_wait__locked(&sna->render); |
||
67 | |||
68 | /* Note: we only need dword alignment (currently) */ |
||
69 | |||
70 | bo = sna->render.vbo; |
||
71 | if (bo) { |
||
72 | for (i = 0; i < sna->render.nvertex_reloc; i++) { |
||
73 | DBG(("%s: reloc[%d] = %d\n", __FUNCTION__, |
||
74 | i, sna->render.vertex_reloc[i])); |
||
75 | |||
76 | sna->kgem.batch[sna->render.vertex_reloc[i]] = |
||
77 | kgem_add_reloc(&sna->kgem, |
||
78 | sna->render.vertex_reloc[i], bo, |
||
79 | I915_GEM_DOMAIN_VERTEX << 16, |
||
80 | 0); |
||
81 | } |
||
82 | |||
83 | assert(!sna->render.active); |
||
84 | sna->render.nvertex_reloc = 0; |
||
85 | sna->render.vertex_used = 0; |
||
86 | sna->render.vertex_index = 0; |
||
87 | sna->render.vbo = NULL; |
||
88 | sna->render.vb_id = 0; |
||
89 | |||
90 | kgem_bo_destroy(&sna->kgem, bo); |
||
91 | } |
||
92 | |||
93 | hint = CREATE_GTT_MAP; |
||
94 | if (bo) |
||
95 | hint |= CREATE_CACHED | CREATE_NO_THROTTLE; |
||
96 | |||
97 | size = 256*1024; |
||
98 | assert(!sna->render.active); |
||
99 | sna->render.vertices = NULL; |
||
100 | sna->render.vbo = kgem_create_linear(&sna->kgem, size, hint); |
||
101 | while (sna->render.vbo == NULL && size > 16*1024) { |
||
102 | size /= 2; |
||
103 | sna->render.vbo = kgem_create_linear(&sna->kgem, size, hint); |
||
104 | } |
||
105 | if (sna->render.vbo == NULL) |
||
106 | sna->render.vbo = kgem_create_linear(&sna->kgem, |
||
107 | 256*1024, CREATE_GTT_MAP); |
||
108 | if (sna->render.vbo) |
||
109 | sna->render.vertices = kgem_bo_map(&sna->kgem, sna->render.vbo); |
||
110 | if (sna->render.vertices == NULL) { |
||
111 | if (sna->render.vbo) { |
||
112 | kgem_bo_destroy(&sna->kgem, sna->render.vbo); |
||
113 | sna->render.vbo = NULL; |
||
114 | } |
||
115 | sna->render.vertices = sna->render.vertex_data; |
||
116 | sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
||
117 | return 0; |
||
118 | } |
||
119 | |||
120 | if (sna->render.vertex_used) { |
||
121 | DBG(("%s: copying initial buffer x %d to handle=%d\n", |
||
122 | __FUNCTION__, |
||
123 | sna->render.vertex_used, |
||
124 | sna->render.vbo->handle)); |
||
125 | assert(sizeof(float)*sna->render.vertex_used <= |
||
126 | __kgem_bo_size(sna->render.vbo)); |
||
127 | memcpy(sna->render.vertices, |
||
128 | sna->render.vertex_data, |
||
129 | sizeof(float)*sna->render.vertex_used); |
||
130 | } |
||
131 | |||
132 | size = __kgem_bo_size(sna->render.vbo)/4; |
||
133 | if (size >= UINT16_MAX) |
||
134 | size = UINT16_MAX - 1; |
||
135 | |||
136 | DBG(("%s: create vbo handle=%d, size=%d\n", |
||
137 | __FUNCTION__, sna->render.vbo->handle, size)); |
||
138 | |||
139 | sna->render.vertex_size = size; |
||
140 | return sna->render.vertex_size - sna->render.vertex_used; |
||
141 | } |
||
142 | |||
143 | void gen4_vertex_close(struct sna *sna) |
||
144 | { |
||
145 | struct kgem_bo *bo, *free_bo = NULL; |
||
146 | unsigned int i, delta = 0; |
||
147 | |||
148 | assert(sna->render.vertex_offset == 0); |
||
149 | if (!sna->render.vb_id) |
||
150 | return; |
||
151 | |||
152 | DBG(("%s: used=%d, vbo active? %d, vb=%x, nreloc=%d\n", |
||
153 | __FUNCTION__, sna->render.vertex_used, sna->render.vbo ? sna->render.vbo->handle : 0, |
||
154 | sna->render.vb_id, sna->render.nvertex_reloc)); |
||
155 | |||
156 | assert(!sna->render.active); |
||
157 | |||
158 | bo = sna->render.vbo; |
||
159 | if (bo) { |
||
160 | if (sna->render.vertex_size - sna->render.vertex_used < 64) { |
||
161 | DBG(("%s: discarding vbo (full), handle=%d\n", __FUNCTION__, sna->render.vbo->handle)); |
||
162 | sna->render.vbo = NULL; |
||
163 | sna->render.vertices = sna->render.vertex_data; |
||
164 | sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
||
165 | free_bo = bo; |
||
166 | } else if (IS_CPU_MAP(bo->map) && !sna->kgem.has_llc) { |
||
167 | DBG(("%s: converting CPU map to GTT\n", __FUNCTION__)); |
||
168 | sna->render.vertices = |
||
169 | kgem_bo_map__gtt(&sna->kgem, sna->render.vbo); |
||
170 | if (sna->render.vertices == NULL) { |
||
171 | sna->render.vbo = NULL; |
||
172 | sna->render.vertices = sna->render.vertex_data; |
||
173 | sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
||
174 | free_bo = bo; |
||
175 | } |
||
176 | |||
177 | } |
||
178 | } else { |
||
179 | if (sna->kgem.nbatch + sna->render.vertex_used <= sna->kgem.surface) { |
||
180 | DBG(("%s: copy to batch: %d @ %d\n", __FUNCTION__, |
||
181 | sna->render.vertex_used, sna->kgem.nbatch)); |
||
182 | memcpy(sna->kgem.batch + sna->kgem.nbatch, |
||
183 | sna->render.vertex_data, |
||
184 | sna->render.vertex_used * 4); |
||
185 | delta = sna->kgem.nbatch * 4; |
||
186 | bo = NULL; |
||
187 | sna->kgem.nbatch += sna->render.vertex_used; |
||
188 | } else { |
||
189 | bo = kgem_create_linear(&sna->kgem, |
||
190 | 4*sna->render.vertex_used, |
||
191 | CREATE_NO_THROTTLE); |
||
192 | if (bo && !kgem_bo_write(&sna->kgem, bo, |
||
193 | sna->render.vertex_data, |
||
194 | 4*sna->render.vertex_used)) { |
||
195 | kgem_bo_destroy(&sna->kgem, bo); |
||
196 | bo = NULL; |
||
197 | } |
||
198 | DBG(("%s: new vbo: %d\n", __FUNCTION__, |
||
199 | sna->render.vertex_used)); |
||
200 | free_bo = bo; |
||
201 | } |
||
202 | } |
||
203 | |||
204 | assert(sna->render.nvertex_reloc); |
||
205 | for (i = 0; i < sna->render.nvertex_reloc; i++) { |
||
206 | DBG(("%s: reloc[%d] = %d\n", __FUNCTION__, |
||
207 | i, sna->render.vertex_reloc[i])); |
||
208 | |||
209 | sna->kgem.batch[sna->render.vertex_reloc[i]] = |
||
210 | kgem_add_reloc(&sna->kgem, |
||
211 | sna->render.vertex_reloc[i], bo, |
||
212 | I915_GEM_DOMAIN_VERTEX << 16, |
||
213 | delta); |
||
214 | } |
||
215 | sna->render.nvertex_reloc = 0; |
||
216 | sna->render.vb_id = 0; |
||
217 | |||
218 | if (sna->render.vbo == NULL) { |
||
219 | assert(!sna->render.active); |
||
220 | sna->render.vertex_used = 0; |
||
221 | sna->render.vertex_index = 0; |
||
222 | assert(sna->render.vertices == sna->render.vertex_data); |
||
223 | assert(sna->render.vertex_size == ARRAY_SIZE(sna->render.vertex_data)); |
||
224 | } |
||
225 | |||
226 | if (free_bo) |
||
227 | kgem_bo_destroy(&sna->kgem, free_bo); |
||
228 | } |
||
229 | |||
4251 | Serge | 230 | /* specialised vertex emission routines */ |
231 | |||
232 | #define OUT_VERTEX(x,y) vertex_emit_2s(sna, x,y) /* XXX assert(!too_large(x, y)); */ |
||
233 | #define OUT_VERTEX_F(v) vertex_emit(sna, v) |
||
234 | |||
235 | force_inline static float |
||
236 | compute_linear(const struct sna_composite_channel *channel, |
||
237 | int16_t x, int16_t y) |
||
238 | { |
||
239 | return ((x+channel->offset[0]) * channel->u.linear.dx + |
||
240 | (y+channel->offset[1]) * channel->u.linear.dy + |
||
241 | channel->u.linear.offset); |
||
242 | } |
||
243 | |||
244 | sse2 inline static void |
||
245 | emit_texcoord(struct sna *sna, |
||
246 | const struct sna_composite_channel *channel, |
||
247 | int16_t x, int16_t y) |
||
248 | { |
||
249 | if (channel->is_solid) { |
||
250 | OUT_VERTEX_F(x); |
||
251 | return; |
||
252 | } |
||
253 | |||
254 | x += channel->offset[0]; |
||
255 | y += channel->offset[1]; |
||
256 | |||
257 | if (channel->is_affine) { |
||
258 | float s, t; |
||
259 | |||
260 | sna_get_transformed_coordinates(x, y, |
||
261 | channel->transform, |
||
262 | &s, &t); |
||
263 | OUT_VERTEX_F(s * channel->scale[0]); |
||
264 | OUT_VERTEX_F(t * channel->scale[1]); |
||
265 | } else { |
||
266 | float s, t, w; |
||
267 | |||
268 | sna_get_transformed_coordinates_3d(x, y, |
||
269 | channel->transform, |
||
270 | &s, &t, &w); |
||
271 | OUT_VERTEX_F(s * channel->scale[0]); |
||
272 | OUT_VERTEX_F(t * channel->scale[1]); |
||
273 | OUT_VERTEX_F(w); |
||
274 | } |
||
275 | } |
||
276 | |||
277 | sse2 force_inline static void |
||
278 | emit_vertex(struct sna *sna, |
||
279 | const struct sna_composite_op *op, |
||
280 | int16_t srcX, int16_t srcY, |
||
281 | int16_t mskX, int16_t mskY, |
||
282 | int16_t dstX, int16_t dstY) |
||
283 | { |
||
284 | OUT_VERTEX(dstX, dstY); |
||
285 | emit_texcoord(sna, &op->src, srcX, srcY); |
||
286 | } |
||
287 | |||
288 | sse2 fastcall static void |
||
289 | emit_primitive(struct sna *sna, |
||
290 | const struct sna_composite_op *op, |
||
291 | const struct sna_composite_rectangles *r) |
||
292 | { |
||
293 | emit_vertex(sna, op, |
||
294 | r->src.x + r->width, r->src.y + r->height, |
||
295 | r->mask.x + r->width, r->mask.y + r->height, |
||
296 | r->dst.x + r->width, r->dst.y + r->height); |
||
297 | emit_vertex(sna, op, |
||
298 | r->src.x, r->src.y + r->height, |
||
299 | r->mask.x, r->mask.y + r->height, |
||
300 | r->dst.x, r->dst.y + r->height); |
||
301 | emit_vertex(sna, op, |
||
302 | r->src.x, r->src.y, |
||
303 | r->mask.x, r->mask.y, |
||
304 | r->dst.x, r->dst.y); |
||
305 | } |
||
306 | |||
307 | sse2 force_inline static void |
||
308 | emit_vertex_mask(struct sna *sna, |
||
309 | const struct sna_composite_op *op, |
||
310 | int16_t srcX, int16_t srcY, |
||
311 | int16_t mskX, int16_t mskY, |
||
312 | int16_t dstX, int16_t dstY) |
||
313 | { |
||
314 | OUT_VERTEX(dstX, dstY); |
||
315 | emit_texcoord(sna, &op->src, srcX, srcY); |
||
316 | emit_texcoord(sna, &op->mask, mskX, mskY); |
||
317 | } |
||
318 | |||
319 | sse2 fastcall static void |
||
320 | emit_primitive_mask(struct sna *sna, |
||
321 | const struct sna_composite_op *op, |
||
322 | const struct sna_composite_rectangles *r) |
||
323 | { |
||
324 | emit_vertex_mask(sna, op, |
||
325 | r->src.x + r->width, r->src.y + r->height, |
||
326 | r->mask.x + r->width, r->mask.y + r->height, |
||
327 | r->dst.x + r->width, r->dst.y + r->height); |
||
328 | emit_vertex_mask(sna, op, |
||
329 | r->src.x, r->src.y + r->height, |
||
330 | r->mask.x, r->mask.y + r->height, |
||
331 | r->dst.x, r->dst.y + r->height); |
||
332 | emit_vertex_mask(sna, op, |
||
333 | r->src.x, r->src.y, |
||
334 | r->mask.x, r->mask.y, |
||
335 | r->dst.x, r->dst.y); |
||
336 | } |
||
337 | |||
338 | sse2 fastcall static void |
||
339 | emit_primitive_solid(struct sna *sna, |
||
340 | const struct sna_composite_op *op, |
||
341 | const struct sna_composite_rectangles *r) |
||
342 | { |
||
343 | float *v; |
||
344 | union { |
||
345 | struct sna_coordinate p; |
||
346 | float f; |
||
347 | } dst; |
||
348 | |||
349 | assert(op->floats_per_rect == 6); |
||
350 | assert((sna->render.vertex_used % 2) == 0); |
||
351 | v = sna->render.vertices + sna->render.vertex_used; |
||
352 | sna->render.vertex_used += 6; |
||
353 | assert(sna->render.vertex_used <= sna->render.vertex_size); |
||
354 | |||
355 | dst.p.x = r->dst.x + r->width; |
||
356 | dst.p.y = r->dst.y + r->height; |
||
357 | v[0] = dst.f; |
||
358 | dst.p.x = r->dst.x; |
||
359 | v[2] = dst.f; |
||
360 | dst.p.y = r->dst.y; |
||
361 | v[4] = dst.f; |
||
362 | |||
363 | v[5] = v[3] = v[1] = .5; |
||
364 | } |
||
365 | |||
366 | sse2 fastcall static void |
||
367 | emit_boxes_solid(const struct sna_composite_op *op, |
||
368 | const BoxRec *box, int nbox, |
||
369 | float *v) |
||
370 | { |
||
371 | do { |
||
372 | union { |
||
373 | struct sna_coordinate p; |
||
374 | float f; |
||
375 | } dst; |
||
376 | |||
377 | dst.p.x = box->x2; |
||
378 | dst.p.y = box->y2; |
||
379 | v[0] = dst.f; |
||
380 | dst.p.x = box->x1; |
||
381 | v[2] = dst.f; |
||
382 | dst.p.y = box->y1; |
||
383 | v[4] = dst.f; |
||
384 | |||
385 | v[5] = v[3] = v[1] = .5; |
||
386 | box++; |
||
387 | v += 6; |
||
388 | } while (--nbox); |
||
389 | } |
||
390 | |||
391 | sse2 fastcall static void |
||
392 | emit_primitive_linear(struct sna *sna, |
||
393 | const struct sna_composite_op *op, |
||
394 | const struct sna_composite_rectangles *r) |
||
395 | { |
||
396 | float *v; |
||
397 | union { |
||
398 | struct sna_coordinate p; |
||
399 | float f; |
||
400 | } dst; |
||
401 | |||
402 | assert(op->floats_per_rect == 6); |
||
403 | assert((sna->render.vertex_used % 2) == 0); |
||
404 | v = sna->render.vertices + sna->render.vertex_used; |
||
405 | sna->render.vertex_used += 6; |
||
406 | assert(sna->render.vertex_used <= sna->render.vertex_size); |
||
407 | |||
408 | dst.p.x = r->dst.x + r->width; |
||
409 | dst.p.y = r->dst.y + r->height; |
||
410 | v[0] = dst.f; |
||
411 | dst.p.x = r->dst.x; |
||
412 | v[2] = dst.f; |
||
413 | dst.p.y = r->dst.y; |
||
414 | v[4] = dst.f; |
||
415 | |||
416 | v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
||
417 | v[3] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
||
418 | v[5] = compute_linear(&op->src, r->src.x, r->src.y); |
||
419 | } |
||
420 | |||
421 | sse2 fastcall static void |
||
422 | emit_boxes_linear(const struct sna_composite_op *op, |
||
423 | const BoxRec *box, int nbox, |
||
424 | float *v) |
||
425 | { |
||
426 | union { |
||
427 | struct sna_coordinate p; |
||
428 | float f; |
||
429 | } dst; |
||
430 | |||
431 | do { |
||
432 | dst.p.x = box->x2; |
||
433 | dst.p.y = box->y2; |
||
434 | v[0] = dst.f; |
||
435 | dst.p.x = box->x1; |
||
436 | v[2] = dst.f; |
||
437 | dst.p.y = box->y1; |
||
438 | v[4] = dst.f; |
||
439 | |||
440 | v[1] = compute_linear(&op->src, box->x2, box->y2); |
||
441 | v[3] = compute_linear(&op->src, box->x1, box->y2); |
||
442 | v[5] = compute_linear(&op->src, box->x1, box->y1); |
||
443 | |||
444 | v += 6; |
||
445 | box++; |
||
446 | } while (--nbox); |
||
447 | } |
||
448 | |||
449 | sse2 fastcall static void |
||
450 | emit_primitive_identity_source(struct sna *sna, |
||
451 | const struct sna_composite_op *op, |
||
452 | const struct sna_composite_rectangles *r) |
||
453 | { |
||
454 | union { |
||
455 | struct sna_coordinate p; |
||
456 | float f; |
||
457 | } dst; |
||
458 | float *v; |
||
459 | |||
460 | assert(op->floats_per_rect == 9); |
||
461 | assert((sna->render.vertex_used % 3) == 0); |
||
462 | v = sna->render.vertices + sna->render.vertex_used; |
||
463 | sna->render.vertex_used += 9; |
||
464 | |||
465 | dst.p.x = r->dst.x + r->width; |
||
466 | dst.p.y = r->dst.y + r->height; |
||
467 | v[0] = dst.f; |
||
468 | dst.p.x = r->dst.x; |
||
469 | v[3] = dst.f; |
||
470 | dst.p.y = r->dst.y; |
||
471 | v[6] = dst.f; |
||
472 | |||
473 | v[7] = v[4] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
||
474 | v[1] = v[4] + r->width * op->src.scale[0]; |
||
475 | |||
476 | v[8] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
||
477 | v[5] = v[2] = v[8] + r->height * op->src.scale[1]; |
||
478 | } |
||
479 | |||
480 | sse2 fastcall static void |
||
481 | emit_boxes_identity_source(const struct sna_composite_op *op, |
||
482 | const BoxRec *box, int nbox, |
||
483 | float *v) |
||
484 | { |
||
485 | do { |
||
486 | union { |
||
487 | struct sna_coordinate p; |
||
488 | float f; |
||
489 | } dst; |
||
490 | |||
491 | dst.p.x = box->x2; |
||
492 | dst.p.y = box->y2; |
||
493 | v[0] = dst.f; |
||
494 | dst.p.x = box->x1; |
||
495 | v[3] = dst.f; |
||
496 | dst.p.y = box->y1; |
||
497 | v[6] = dst.f; |
||
498 | |||
499 | v[7] = v[4] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
||
500 | v[1] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
||
501 | |||
502 | v[8] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
||
503 | v[2] = v[5] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
||
504 | |||
505 | v += 9; |
||
506 | box++; |
||
507 | } while (--nbox); |
||
508 | } |
||
509 | |||
510 | sse2 fastcall static void |
||
511 | emit_primitive_simple_source(struct sna *sna, |
||
512 | const struct sna_composite_op *op, |
||
513 | const struct sna_composite_rectangles *r) |
||
514 | { |
||
515 | float *v; |
||
516 | union { |
||
517 | struct sna_coordinate p; |
||
518 | float f; |
||
519 | } dst; |
||
520 | |||
521 | float xx = op->src.transform->matrix[0][0]; |
||
522 | float x0 = op->src.transform->matrix[0][2]; |
||
523 | float yy = op->src.transform->matrix[1][1]; |
||
524 | float y0 = op->src.transform->matrix[1][2]; |
||
525 | float sx = op->src.scale[0]; |
||
526 | float sy = op->src.scale[1]; |
||
527 | int16_t tx = op->src.offset[0]; |
||
528 | int16_t ty = op->src.offset[1]; |
||
529 | |||
530 | assert(op->floats_per_rect == 9); |
||
531 | assert((sna->render.vertex_used % 3) == 0); |
||
532 | v = sna->render.vertices + sna->render.vertex_used; |
||
533 | sna->render.vertex_used += 3*3; |
||
534 | |||
535 | dst.p.x = r->dst.x + r->width; |
||
536 | dst.p.y = r->dst.y + r->height; |
||
537 | v[0] = dst.f; |
||
538 | v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
||
539 | v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
||
540 | |||
541 | dst.p.x = r->dst.x; |
||
542 | v[3] = dst.f; |
||
543 | v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx; |
||
544 | |||
545 | dst.p.y = r->dst.y; |
||
546 | v[6] = dst.f; |
||
547 | v[8] = ((r->src.y + ty) * yy + y0) * sy; |
||
548 | } |
||
549 | |||
550 | sse2 fastcall static void |
||
551 | emit_boxes_simple_source(const struct sna_composite_op *op, |
||
552 | const BoxRec *box, int nbox, |
||
553 | float *v) |
||
554 | { |
||
555 | float xx = op->src.transform->matrix[0][0]; |
||
556 | float x0 = op->src.transform->matrix[0][2]; |
||
557 | float yy = op->src.transform->matrix[1][1]; |
||
558 | float y0 = op->src.transform->matrix[1][2]; |
||
559 | float sx = op->src.scale[0]; |
||
560 | float sy = op->src.scale[1]; |
||
561 | int16_t tx = op->src.offset[0]; |
||
562 | int16_t ty = op->src.offset[1]; |
||
563 | |||
564 | do { |
||
565 | union { |
||
566 | struct sna_coordinate p; |
||
567 | float f; |
||
568 | } dst; |
||
569 | |||
570 | dst.p.x = box->x2; |
||
571 | dst.p.y = box->y2; |
||
572 | v[0] = dst.f; |
||
573 | v[1] = ((box->x2 + tx) * xx + x0) * sx; |
||
574 | v[5] = v[2] = ((box->y2 + ty) * yy + y0) * sy; |
||
575 | |||
576 | dst.p.x = box->x1; |
||
577 | v[3] = dst.f; |
||
578 | v[7] = v[4] = ((box->x1 + tx) * xx + x0) * sx; |
||
579 | |||
580 | dst.p.y = box->y1; |
||
581 | v[6] = dst.f; |
||
582 | v[8] = ((box->y1 + ty) * yy + y0) * sy; |
||
583 | |||
584 | v += 9; |
||
585 | box++; |
||
586 | } while (--nbox); |
||
587 | } |
||
588 | |||
589 | sse2 fastcall static void |
||
590 | emit_primitive_affine_source(struct sna *sna, |
||
591 | const struct sna_composite_op *op, |
||
592 | const struct sna_composite_rectangles *r) |
||
593 | { |
||
594 | union { |
||
595 | struct sna_coordinate p; |
||
596 | float f; |
||
597 | } dst; |
||
598 | float *v; |
||
599 | |||
600 | assert(op->floats_per_rect == 9); |
||
601 | assert((sna->render.vertex_used % 3) == 0); |
||
602 | v = sna->render.vertices + sna->render.vertex_used; |
||
603 | sna->render.vertex_used += 9; |
||
604 | |||
605 | dst.p.x = r->dst.x + r->width; |
||
606 | dst.p.y = r->dst.y + r->height; |
||
607 | v[0] = dst.f; |
||
608 | _sna_get_transformed_scaled(op->src.offset[0] + r->src.x + r->width, |
||
609 | op->src.offset[1] + r->src.y + r->height, |
||
610 | op->src.transform, op->src.scale, |
||
611 | &v[1], &v[2]); |
||
612 | |||
613 | dst.p.x = r->dst.x; |
||
614 | v[3] = dst.f; |
||
615 | _sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
||
616 | op->src.offset[1] + r->src.y + r->height, |
||
617 | op->src.transform, op->src.scale, |
||
618 | &v[4], &v[5]); |
||
619 | |||
620 | dst.p.y = r->dst.y; |
||
621 | v[6] = dst.f; |
||
622 | _sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
||
623 | op->src.offset[1] + r->src.y, |
||
624 | op->src.transform, op->src.scale, |
||
625 | &v[7], &v[8]); |
||
626 | } |
||
627 | |||
628 | sse2 fastcall static void |
||
629 | emit_boxes_affine_source(const struct sna_composite_op *op, |
||
630 | const BoxRec *box, int nbox, |
||
631 | float *v) |
||
632 | { |
||
633 | do { |
||
634 | union { |
||
635 | struct sna_coordinate p; |
||
636 | float f; |
||
637 | } dst; |
||
638 | |||
639 | dst.p.x = box->x2; |
||
640 | dst.p.y = box->y2; |
||
641 | v[0] = dst.f; |
||
642 | _sna_get_transformed_scaled(op->src.offset[0] + box->x2, |
||
643 | op->src.offset[1] + box->y2, |
||
644 | op->src.transform, op->src.scale, |
||
645 | &v[1], &v[2]); |
||
646 | |||
647 | dst.p.x = box->x1; |
||
648 | v[3] = dst.f; |
||
649 | _sna_get_transformed_scaled(op->src.offset[0] + box->x1, |
||
650 | op->src.offset[1] + box->y2, |
||
651 | op->src.transform, op->src.scale, |
||
652 | &v[4], &v[5]); |
||
653 | |||
654 | dst.p.y = box->y1; |
||
655 | v[6] = dst.f; |
||
656 | _sna_get_transformed_scaled(op->src.offset[0] + box->x1, |
||
657 | op->src.offset[1] + box->y1, |
||
658 | op->src.transform, op->src.scale, |
||
659 | &v[7], &v[8]); |
||
660 | box++; |
||
661 | v += 9; |
||
662 | } while (--nbox); |
||
663 | } |
||
664 | |||
665 | sse2 fastcall static void |
||
666 | emit_primitive_identity_mask(struct sna *sna, |
||
667 | const struct sna_composite_op *op, |
||
668 | const struct sna_composite_rectangles *r) |
||
669 | { |
||
670 | union { |
||
671 | struct sna_coordinate p; |
||
672 | float f; |
||
673 | } dst; |
||
674 | float msk_x, msk_y; |
||
675 | float w, h; |
||
676 | float *v; |
||
677 | |||
678 | msk_x = r->mask.x + op->mask.offset[0]; |
||
679 | msk_y = r->mask.y + op->mask.offset[1]; |
||
680 | w = r->width; |
||
681 | h = r->height; |
||
682 | |||
683 | DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
||
684 | __FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
||
685 | |||
686 | assert(op->floats_per_rect == 12); |
||
687 | assert((sna->render.vertex_used % 4) == 0); |
||
688 | v = sna->render.vertices + sna->render.vertex_used; |
||
689 | sna->render.vertex_used += 12; |
||
690 | |||
691 | dst.p.x = r->dst.x + r->width; |
||
692 | dst.p.y = r->dst.y + r->height; |
||
693 | v[0] = dst.f; |
||
694 | v[2] = (msk_x + w) * op->mask.scale[0]; |
||
695 | v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
||
696 | |||
697 | dst.p.x = r->dst.x; |
||
698 | v[4] = dst.f; |
||
699 | v[10] = v[6] = msk_x * op->mask.scale[0]; |
||
700 | |||
701 | dst.p.y = r->dst.y; |
||
702 | v[8] = dst.f; |
||
703 | v[11] = msk_y * op->mask.scale[1]; |
||
704 | |||
705 | v[9] = v[5] = v[1] = .5; |
||
706 | } |
||
707 | |||
708 | sse2 fastcall static void |
||
709 | emit_boxes_identity_mask(const struct sna_composite_op *op, |
||
710 | const BoxRec *box, int nbox, |
||
711 | float *v) |
||
712 | { |
||
713 | float msk_x = op->mask.offset[0]; |
||
714 | float msk_y = op->mask.offset[1]; |
||
715 | |||
716 | do { |
||
717 | union { |
||
718 | struct sna_coordinate p; |
||
719 | float f; |
||
720 | } dst; |
||
721 | |||
722 | dst.p.x = box->x2; |
||
723 | dst.p.y = box->y2; |
||
724 | v[0] = dst.f; |
||
725 | v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
||
726 | v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
||
727 | |||
728 | dst.p.x = box->x1; |
||
729 | v[4] = dst.f; |
||
730 | v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
||
731 | |||
732 | dst.p.y = box->y1; |
||
733 | v[8] = dst.f; |
||
734 | v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
||
735 | |||
736 | v[9] = v[5] = v[1] = .5; |
||
737 | v += 12; |
||
738 | box++; |
||
739 | } while (--nbox); |
||
740 | } |
||
741 | |||
742 | sse2 fastcall static void |
||
743 | emit_primitive_linear_identity_mask(struct sna *sna, |
||
744 | const struct sna_composite_op *op, |
||
745 | const struct sna_composite_rectangles *r) |
||
746 | { |
||
747 | union { |
||
748 | struct sna_coordinate p; |
||
749 | float f; |
||
750 | } dst; |
||
751 | float msk_x, msk_y; |
||
752 | float w, h; |
||
753 | float *v; |
||
754 | |||
755 | msk_x = r->mask.x + op->mask.offset[0]; |
||
756 | msk_y = r->mask.y + op->mask.offset[1]; |
||
757 | w = r->width; |
||
758 | h = r->height; |
||
759 | |||
760 | DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
||
761 | __FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
||
762 | |||
763 | assert(op->floats_per_rect == 12); |
||
764 | assert((sna->render.vertex_used % 4) == 0); |
||
765 | v = sna->render.vertices + sna->render.vertex_used; |
||
766 | sna->render.vertex_used += 12; |
||
767 | |||
768 | dst.p.x = r->dst.x + r->width; |
||
769 | dst.p.y = r->dst.y + r->height; |
||
770 | v[0] = dst.f; |
||
771 | v[2] = (msk_x + w) * op->mask.scale[0]; |
||
772 | v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
||
773 | |||
774 | dst.p.x = r->dst.x; |
||
775 | v[4] = dst.f; |
||
776 | v[10] = v[6] = msk_x * op->mask.scale[0]; |
||
777 | |||
778 | dst.p.y = r->dst.y; |
||
779 | v[8] = dst.f; |
||
780 | v[11] = msk_y * op->mask.scale[1]; |
||
781 | |||
782 | v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
||
783 | v[5] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
||
784 | v[9] = compute_linear(&op->src, r->src.x, r->src.y); |
||
785 | } |
||
786 | |||
787 | sse2 fastcall static void |
||
788 | emit_boxes_linear_identity_mask(const struct sna_composite_op *op, |
||
789 | const BoxRec *box, int nbox, |
||
790 | float *v) |
||
791 | { |
||
792 | float msk_x = op->mask.offset[0]; |
||
793 | float msk_y = op->mask.offset[1]; |
||
794 | |||
795 | do { |
||
796 | union { |
||
797 | struct sna_coordinate p; |
||
798 | float f; |
||
799 | } dst; |
||
800 | |||
801 | dst.p.x = box->x2; |
||
802 | dst.p.y = box->y2; |
||
803 | v[0] = dst.f; |
||
804 | v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
||
805 | v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
||
806 | |||
807 | dst.p.x = box->x1; |
||
808 | v[4] = dst.f; |
||
809 | v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
||
810 | |||
811 | dst.p.y = box->y1; |
||
812 | v[8] = dst.f; |
||
813 | v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
||
814 | |||
815 | v[1] = compute_linear(&op->src, box->x2, box->y2); |
||
816 | v[5] = compute_linear(&op->src, box->x1, box->y2); |
||
817 | v[9] = compute_linear(&op->src, box->x1, box->y1); |
||
818 | |||
819 | v += 12; |
||
820 | box++; |
||
821 | } while (--nbox); |
||
822 | } |
||
823 | |||
824 | sse2 fastcall static void |
||
3280 | Serge | 825 | emit_primitive_identity_source_mask(struct sna *sna, |
826 | const struct sna_composite_op *op, |
||
827 | const struct sna_composite_rectangles *r) |
||
828 | { |
||
829 | union { |
||
830 | struct sna_coordinate p; |
||
831 | float f; |
||
832 | } dst; |
||
833 | float src_x, src_y; |
||
834 | float msk_x, msk_y; |
||
835 | float w, h; |
||
836 | float *v; |
||
837 | |||
838 | src_x = r->src.x + op->src.offset[0]; |
||
839 | src_y = r->src.y + op->src.offset[1]; |
||
840 | msk_x = r->mask.x + op->mask.offset[0]; |
||
841 | msk_y = r->mask.y + op->mask.offset[1]; |
||
842 | w = r->width; |
||
843 | h = r->height; |
||
844 | |||
845 | assert(op->floats_per_rect == 15); |
||
846 | assert((sna->render.vertex_used % 5) == 0); |
||
847 | v = sna->render.vertices + sna->render.vertex_used; |
||
848 | sna->render.vertex_used += 15; |
||
849 | |||
850 | dst.p.x = r->dst.x + r->width; |
||
851 | dst.p.y = r->dst.y + r->height; |
||
852 | v[0] = dst.f; |
||
853 | v[1] = (src_x + w) * op->src.scale[0]; |
||
854 | v[2] = (src_y + h) * op->src.scale[1]; |
||
855 | v[3] = (msk_x + w) * op->mask.scale[0]; |
||
856 | v[4] = (msk_y + h) * op->mask.scale[1]; |
||
857 | |||
858 | dst.p.x = r->dst.x; |
||
859 | v[5] = dst.f; |
||
860 | v[6] = src_x * op->src.scale[0]; |
||
861 | v[7] = v[2]; |
||
862 | v[8] = msk_x * op->mask.scale[0]; |
||
863 | v[9] = v[4]; |
||
864 | |||
865 | dst.p.y = r->dst.y; |
||
866 | v[10] = dst.f; |
||
867 | v[11] = v[6]; |
||
868 | v[12] = src_y * op->src.scale[1]; |
||
869 | v[13] = v[8]; |
||
870 | v[14] = msk_y * op->mask.scale[1]; |
||
871 | } |
||
872 | |||
4251 | Serge | 873 | sse2 fastcall static void |
874 | emit_primitive_simple_source_identity(struct sna *sna, |
||
875 | const struct sna_composite_op *op, |
||
876 | const struct sna_composite_rectangles *r) |
||
3280 | Serge | 877 | { |
4251 | Serge | 878 | float *v; |
879 | union { |
||
880 | struct sna_coordinate p; |
||
881 | float f; |
||
882 | } dst; |
||
883 | |||
884 | float xx = op->src.transform->matrix[0][0]; |
||
885 | float x0 = op->src.transform->matrix[0][2]; |
||
886 | float yy = op->src.transform->matrix[1][1]; |
||
887 | float y0 = op->src.transform->matrix[1][2]; |
||
888 | float sx = op->src.scale[0]; |
||
889 | float sy = op->src.scale[1]; |
||
890 | int16_t tx = op->src.offset[0]; |
||
891 | int16_t ty = op->src.offset[1]; |
||
892 | float msk_x = r->mask.x + op->mask.offset[0]; |
||
893 | float msk_y = r->mask.y + op->mask.offset[1]; |
||
894 | float w = r->width, h = r->height; |
||
895 | |||
896 | assert(op->floats_per_rect == 15); |
||
897 | assert((sna->render.vertex_used % 5) == 0); |
||
898 | v = sna->render.vertices + sna->render.vertex_used; |
||
899 | sna->render.vertex_used += 3*5; |
||
900 | |||
901 | dst.p.x = r->dst.x + r->width; |
||
902 | dst.p.y = r->dst.y + r->height; |
||
903 | v[0] = dst.f; |
||
904 | v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
||
905 | v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
||
906 | v[3] = (msk_x + w) * op->mask.scale[0]; |
||
907 | v[4] = (msk_y + h) * op->mask.scale[1]; |
||
908 | |||
909 | dst.p.x = r->dst.x; |
||
910 | v[5] = dst.f; |
||
911 | v[6] = ((r->src.x + tx) * xx + x0) * sx; |
||
912 | v[7] = v[2]; |
||
913 | v[8] = msk_x * op->mask.scale[0]; |
||
914 | v[9] = v[4]; |
||
915 | |||
916 | dst.p.y = r->dst.y; |
||
917 | v[10] = dst.f; |
||
918 | v[11] = v[6]; |
||
919 | v[12] = ((r->src.y + ty) * yy + y0) * sy; |
||
920 | v[13] = v[8]; |
||
921 | v[14] = msk_y * op->mask.scale[1]; |
||
922 | } |
||
923 | |||
924 | sse2 fastcall static void |
||
925 | emit_primitive_affine_source_identity(struct sna *sna, |
||
926 | const struct sna_composite_op *op, |
||
927 | const struct sna_composite_rectangles *r) |
||
928 | { |
||
929 | float *v; |
||
930 | union { |
||
931 | struct sna_coordinate p; |
||
932 | float f; |
||
933 | } dst; |
||
934 | float msk_x = r->mask.x + op->mask.offset[0]; |
||
935 | float msk_y = r->mask.y + op->mask.offset[1]; |
||
936 | float w = r->width, h = r->height; |
||
937 | |||
938 | assert(op->floats_per_rect == 15); |
||
939 | assert((sna->render.vertex_used % 5) == 0); |
||
940 | v = sna->render.vertices + sna->render.vertex_used; |
||
941 | sna->render.vertex_used += 3*5; |
||
942 | |||
943 | dst.p.x = r->dst.x + r->width; |
||
944 | dst.p.y = r->dst.y + r->height; |
||
945 | v[0] = dst.f; |
||
946 | _sna_get_transformed_scaled(op->src.offset[0] + r->src.x + r->width, |
||
947 | op->src.offset[1] + r->src.y + r->height, |
||
948 | op->src.transform, op->src.scale, |
||
949 | &v[1], &v[2]); |
||
950 | v[3] = (msk_x + w) * op->mask.scale[0]; |
||
951 | v[4] = (msk_y + h) * op->mask.scale[1]; |
||
952 | |||
953 | dst.p.x = r->dst.x; |
||
954 | v[5] = dst.f; |
||
955 | _sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
||
956 | op->src.offset[1] + r->src.y + r->height, |
||
957 | op->src.transform, op->src.scale, |
||
958 | &v[6], &v[7]); |
||
959 | v[8] = msk_x * op->mask.scale[0]; |
||
960 | v[9] = v[4]; |
||
961 | |||
962 | dst.p.y = r->dst.y; |
||
963 | v[10] = dst.f; |
||
964 | _sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
||
965 | op->src.offset[1] + r->src.y, |
||
966 | op->src.transform, op->src.scale, |
||
967 | &v[11], &v[12]); |
||
968 | v[13] = v[8]; |
||
969 | v[14] = msk_y * op->mask.scale[1]; |
||
970 | } |
||
971 | |||
972 | /* SSE4_2 */ |
||
973 | #if defined(sse4_2) |
||
974 | |||
975 | sse4_2 fastcall static void |
||
976 | emit_primitive_linear__sse4_2(struct sna *sna, |
||
977 | const struct sna_composite_op *op, |
||
978 | const struct sna_composite_rectangles *r) |
||
979 | { |
||
980 | float *v; |
||
981 | union { |
||
982 | struct sna_coordinate p; |
||
983 | float f; |
||
984 | } dst; |
||
985 | |||
986 | assert(op->floats_per_rect == 6); |
||
987 | assert((sna->render.vertex_used % 2) == 0); |
||
988 | v = sna->render.vertices + sna->render.vertex_used; |
||
989 | sna->render.vertex_used += 6; |
||
990 | assert(sna->render.vertex_used <= sna->render.vertex_size); |
||
991 | |||
992 | dst.p.x = r->dst.x + r->width; |
||
993 | dst.p.y = r->dst.y + r->height; |
||
994 | v[0] = dst.f; |
||
995 | dst.p.x = r->dst.x; |
||
996 | v[2] = dst.f; |
||
997 | dst.p.y = r->dst.y; |
||
998 | v[4] = dst.f; |
||
999 | |||
1000 | v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
||
1001 | v[3] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
||
1002 | v[5] = compute_linear(&op->src, r->src.x, r->src.y); |
||
1003 | } |
||
1004 | |||
1005 | sse4_2 fastcall static void |
||
1006 | emit_boxes_linear__sse4_2(const struct sna_composite_op *op, |
||
1007 | const BoxRec *box, int nbox, |
||
1008 | float *v) |
||
1009 | { |
||
1010 | union { |
||
1011 | struct sna_coordinate p; |
||
1012 | float f; |
||
1013 | } dst; |
||
1014 | |||
1015 | do { |
||
1016 | dst.p.x = box->x2; |
||
1017 | dst.p.y = box->y2; |
||
1018 | v[0] = dst.f; |
||
1019 | dst.p.x = box->x1; |
||
1020 | v[2] = dst.f; |
||
1021 | dst.p.y = box->y1; |
||
1022 | v[4] = dst.f; |
||
1023 | |||
1024 | v[1] = compute_linear(&op->src, box->x2, box->y2); |
||
1025 | v[3] = compute_linear(&op->src, box->x1, box->y2); |
||
1026 | v[5] = compute_linear(&op->src, box->x1, box->y1); |
||
1027 | |||
1028 | v += 6; |
||
1029 | box++; |
||
1030 | } while (--nbox); |
||
1031 | } |
||
1032 | |||
1033 | sse4_2 fastcall static void |
||
1034 | emit_primitive_identity_source__sse4_2(struct sna *sna, |
||
1035 | const struct sna_composite_op *op, |
||
1036 | const struct sna_composite_rectangles *r) |
||
1037 | { |
||
1038 | union { |
||
1039 | struct sna_coordinate p; |
||
1040 | float f; |
||
1041 | } dst; |
||
1042 | float *v; |
||
1043 | |||
1044 | assert(op->floats_per_rect == 9); |
||
1045 | assert((sna->render.vertex_used % 3) == 0); |
||
1046 | v = sna->render.vertices + sna->render.vertex_used; |
||
1047 | sna->render.vertex_used += 9; |
||
1048 | |||
1049 | dst.p.x = r->dst.x + r->width; |
||
1050 | dst.p.y = r->dst.y + r->height; |
||
1051 | v[0] = dst.f; |
||
1052 | dst.p.x = r->dst.x; |
||
1053 | v[3] = dst.f; |
||
1054 | dst.p.y = r->dst.y; |
||
1055 | v[6] = dst.f; |
||
1056 | |||
1057 | v[7] = v[4] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
||
1058 | v[1] = v[4] + r->width * op->src.scale[0]; |
||
1059 | |||
1060 | v[8] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
||
1061 | v[5] = v[2] = v[8] + r->height * op->src.scale[1]; |
||
1062 | } |
||
1063 | |||
1064 | sse4_2 fastcall static void |
||
1065 | emit_boxes_identity_source__sse4_2(const struct sna_composite_op *op, |
||
1066 | const BoxRec *box, int nbox, |
||
1067 | float *v) |
||
1068 | { |
||
1069 | do { |
||
1070 | union { |
||
1071 | struct sna_coordinate p; |
||
1072 | float f; |
||
1073 | } dst; |
||
1074 | |||
1075 | dst.p.x = box->x2; |
||
1076 | dst.p.y = box->y2; |
||
1077 | v[0] = dst.f; |
||
1078 | dst.p.x = box->x1; |
||
1079 | v[3] = dst.f; |
||
1080 | dst.p.y = box->y1; |
||
1081 | v[6] = dst.f; |
||
1082 | |||
1083 | v[7] = v[4] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
||
1084 | v[1] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
||
1085 | |||
1086 | v[8] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
||
1087 | v[2] = v[5] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
||
1088 | |||
1089 | v += 9; |
||
1090 | box++; |
||
1091 | } while (--nbox); |
||
1092 | } |
||
1093 | |||
1094 | sse4_2 fastcall static void |
||
1095 | emit_primitive_simple_source__sse4_2(struct sna *sna, |
||
1096 | const struct sna_composite_op *op, |
||
1097 | const struct sna_composite_rectangles *r) |
||
1098 | { |
||
1099 | float *v; |
||
1100 | union { |
||
1101 | struct sna_coordinate p; |
||
1102 | float f; |
||
1103 | } dst; |
||
1104 | |||
1105 | float xx = op->src.transform->matrix[0][0]; |
||
1106 | float x0 = op->src.transform->matrix[0][2]; |
||
1107 | float yy = op->src.transform->matrix[1][1]; |
||
1108 | float y0 = op->src.transform->matrix[1][2]; |
||
1109 | float sx = op->src.scale[0]; |
||
1110 | float sy = op->src.scale[1]; |
||
1111 | int16_t tx = op->src.offset[0]; |
||
1112 | int16_t ty = op->src.offset[1]; |
||
1113 | |||
1114 | assert(op->floats_per_rect == 9); |
||
1115 | assert((sna->render.vertex_used % 3) == 0); |
||
1116 | v = sna->render.vertices + sna->render.vertex_used; |
||
1117 | sna->render.vertex_used += 3*3; |
||
1118 | |||
1119 | dst.p.x = r->dst.x + r->width; |
||
1120 | dst.p.y = r->dst.y + r->height; |
||
1121 | v[0] = dst.f; |
||
1122 | v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
||
1123 | v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
||
1124 | |||
1125 | dst.p.x = r->dst.x; |
||
1126 | v[3] = dst.f; |
||
1127 | v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx; |
||
1128 | |||
1129 | dst.p.y = r->dst.y; |
||
1130 | v[6] = dst.f; |
||
1131 | v[8] = ((r->src.y + ty) * yy + y0) * sy; |
||
1132 | } |
||
1133 | |||
1134 | sse4_2 fastcall static void |
||
1135 | emit_boxes_simple_source__sse4_2(const struct sna_composite_op *op, |
||
1136 | const BoxRec *box, int nbox, |
||
1137 | float *v) |
||
1138 | { |
||
1139 | float xx = op->src.transform->matrix[0][0]; |
||
1140 | float x0 = op->src.transform->matrix[0][2]; |
||
1141 | float yy = op->src.transform->matrix[1][1]; |
||
1142 | float y0 = op->src.transform->matrix[1][2]; |
||
1143 | float sx = op->src.scale[0]; |
||
1144 | float sy = op->src.scale[1]; |
||
1145 | int16_t tx = op->src.offset[0]; |
||
1146 | int16_t ty = op->src.offset[1]; |
||
1147 | |||
1148 | do { |
||
1149 | union { |
||
1150 | struct sna_coordinate p; |
||
1151 | float f; |
||
1152 | } dst; |
||
1153 | |||
1154 | dst.p.x = box->x2; |
||
1155 | dst.p.y = box->y2; |
||
1156 | v[0] = dst.f; |
||
1157 | v[1] = ((box->x2 + tx) * xx + x0) * sx; |
||
1158 | v[5] = v[2] = ((box->y2 + ty) * yy + y0) * sy; |
||
1159 | |||
1160 | dst.p.x = box->x1; |
||
1161 | v[3] = dst.f; |
||
1162 | v[7] = v[4] = ((box->x1 + tx) * xx + x0) * sx; |
||
1163 | |||
1164 | dst.p.y = box->y1; |
||
1165 | v[6] = dst.f; |
||
1166 | v[8] = ((box->y1 + ty) * yy + y0) * sy; |
||
1167 | |||
1168 | v += 9; |
||
1169 | box++; |
||
1170 | } while (--nbox); |
||
1171 | } |
||
1172 | |||
1173 | sse4_2 fastcall static void |
||
1174 | emit_primitive_identity_mask__sse4_2(struct sna *sna, |
||
1175 | const struct sna_composite_op *op, |
||
1176 | const struct sna_composite_rectangles *r) |
||
1177 | { |
||
1178 | union { |
||
1179 | struct sna_coordinate p; |
||
1180 | float f; |
||
1181 | } dst; |
||
1182 | float msk_x, msk_y; |
||
1183 | float w, h; |
||
1184 | float *v; |
||
1185 | |||
1186 | msk_x = r->mask.x + op->mask.offset[0]; |
||
1187 | msk_y = r->mask.y + op->mask.offset[1]; |
||
1188 | w = r->width; |
||
1189 | h = r->height; |
||
1190 | |||
1191 | DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
||
1192 | __FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
||
1193 | |||
1194 | assert(op->floats_per_rect == 12); |
||
1195 | assert((sna->render.vertex_used % 4) == 0); |
||
1196 | v = sna->render.vertices + sna->render.vertex_used; |
||
1197 | sna->render.vertex_used += 12; |
||
1198 | |||
1199 | dst.p.x = r->dst.x + r->width; |
||
1200 | dst.p.y = r->dst.y + r->height; |
||
1201 | v[0] = dst.f; |
||
1202 | v[2] = (msk_x + w) * op->mask.scale[0]; |
||
1203 | v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
||
1204 | |||
1205 | dst.p.x = r->dst.x; |
||
1206 | v[4] = dst.f; |
||
1207 | v[10] = v[6] = msk_x * op->mask.scale[0]; |
||
1208 | |||
1209 | dst.p.y = r->dst.y; |
||
1210 | v[8] = dst.f; |
||
1211 | v[11] = msk_y * op->mask.scale[1]; |
||
1212 | |||
1213 | v[9] = v[5] = v[1] = .5; |
||
1214 | } |
||
1215 | |||
1216 | sse4_2 fastcall static void |
||
1217 | emit_boxes_identity_mask__sse4_2(const struct sna_composite_op *op, |
||
1218 | const BoxRec *box, int nbox, |
||
1219 | float *v) |
||
1220 | { |
||
1221 | float msk_x = op->mask.offset[0]; |
||
1222 | float msk_y = op->mask.offset[1]; |
||
1223 | |||
1224 | do { |
||
1225 | union { |
||
1226 | struct sna_coordinate p; |
||
1227 | float f; |
||
1228 | } dst; |
||
1229 | |||
1230 | dst.p.x = box->x2; |
||
1231 | dst.p.y = box->y2; |
||
1232 | v[0] = dst.f; |
||
1233 | v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
||
1234 | v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
||
1235 | |||
1236 | dst.p.x = box->x1; |
||
1237 | v[4] = dst.f; |
||
1238 | v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
||
1239 | |||
1240 | dst.p.y = box->y1; |
||
1241 | v[8] = dst.f; |
||
1242 | v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
||
1243 | |||
1244 | v[9] = v[5] = v[1] = .5; |
||
1245 | v += 12; |
||
1246 | box++; |
||
1247 | } while (--nbox); |
||
1248 | } |
||
1249 | |||
1250 | sse4_2 fastcall static void |
||
1251 | emit_primitive_linear_identity_mask__sse4_2(struct sna *sna, |
||
1252 | const struct sna_composite_op *op, |
||
1253 | const struct sna_composite_rectangles *r) |
||
1254 | { |
||
1255 | union { |
||
1256 | struct sna_coordinate p; |
||
1257 | float f; |
||
1258 | } dst; |
||
1259 | float msk_x, msk_y; |
||
1260 | float w, h; |
||
1261 | float *v; |
||
1262 | |||
1263 | msk_x = r->mask.x + op->mask.offset[0]; |
||
1264 | msk_y = r->mask.y + op->mask.offset[1]; |
||
1265 | w = r->width; |
||
1266 | h = r->height; |
||
1267 | |||
1268 | DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
||
1269 | __FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
||
1270 | |||
1271 | assert(op->floats_per_rect == 12); |
||
1272 | assert((sna->render.vertex_used % 4) == 0); |
||
1273 | v = sna->render.vertices + sna->render.vertex_used; |
||
1274 | sna->render.vertex_used += 12; |
||
1275 | |||
1276 | dst.p.x = r->dst.x + r->width; |
||
1277 | dst.p.y = r->dst.y + r->height; |
||
1278 | v[0] = dst.f; |
||
1279 | v[2] = (msk_x + w) * op->mask.scale[0]; |
||
1280 | v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
||
1281 | |||
1282 | dst.p.x = r->dst.x; |
||
1283 | v[4] = dst.f; |
||
1284 | v[10] = v[6] = msk_x * op->mask.scale[0]; |
||
1285 | |||
1286 | dst.p.y = r->dst.y; |
||
1287 | v[8] = dst.f; |
||
1288 | v[11] = msk_y * op->mask.scale[1]; |
||
1289 | |||
1290 | v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
||
1291 | v[5] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
||
1292 | v[9] = compute_linear(&op->src, r->src.x, r->src.y); |
||
1293 | } |
||
1294 | |||
1295 | sse4_2 fastcall static void |
||
1296 | emit_boxes_linear_identity_mask__sse4_2(const struct sna_composite_op *op, |
||
1297 | const BoxRec *box, int nbox, |
||
1298 | float *v) |
||
1299 | { |
||
1300 | float msk_x = op->mask.offset[0]; |
||
1301 | float msk_y = op->mask.offset[1]; |
||
1302 | |||
1303 | do { |
||
1304 | union { |
||
1305 | struct sna_coordinate p; |
||
1306 | float f; |
||
1307 | } dst; |
||
1308 | |||
1309 | dst.p.x = box->x2; |
||
1310 | dst.p.y = box->y2; |
||
1311 | v[0] = dst.f; |
||
1312 | v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
||
1313 | v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
||
1314 | |||
1315 | dst.p.x = box->x1; |
||
1316 | v[4] = dst.f; |
||
1317 | v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
||
1318 | |||
1319 | dst.p.y = box->y1; |
||
1320 | v[8] = dst.f; |
||
1321 | v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
||
1322 | |||
1323 | v[1] = compute_linear(&op->src, box->x2, box->y2); |
||
1324 | v[5] = compute_linear(&op->src, box->x1, box->y2); |
||
1325 | v[9] = compute_linear(&op->src, box->x1, box->y1); |
||
1326 | |||
1327 | v += 12; |
||
1328 | box++; |
||
1329 | } while (--nbox); |
||
1330 | } |
||
1331 | |||
1332 | #endif |
||
1333 | |||
1334 | /* AVX2 */ |
||
1335 | #if defined(avx2) |
||
1336 | |||
1337 | avx2 fastcall static void |
||
1338 | emit_primitive_linear__avx2(struct sna *sna, |
||
1339 | const struct sna_composite_op *op, |
||
1340 | const struct sna_composite_rectangles *r) |
||
1341 | { |
||
1342 | float *v; |
||
1343 | union { |
||
1344 | struct sna_coordinate p; |
||
1345 | float f; |
||
1346 | } dst; |
||
1347 | |||
1348 | assert(op->floats_per_rect == 6); |
||
1349 | assert((sna->render.vertex_used % 2) == 0); |
||
1350 | v = sna->render.vertices + sna->render.vertex_used; |
||
1351 | sna->render.vertex_used += 6; |
||
1352 | assert(sna->render.vertex_used <= sna->render.vertex_size); |
||
1353 | |||
1354 | dst.p.x = r->dst.x + r->width; |
||
1355 | dst.p.y = r->dst.y + r->height; |
||
1356 | v[0] = dst.f; |
||
1357 | dst.p.x = r->dst.x; |
||
1358 | v[2] = dst.f; |
||
1359 | dst.p.y = r->dst.y; |
||
1360 | v[4] = dst.f; |
||
1361 | |||
1362 | v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
||
1363 | v[3] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
||
1364 | v[5] = compute_linear(&op->src, r->src.x, r->src.y); |
||
1365 | } |
||
1366 | |||
1367 | avx2 fastcall static void |
||
1368 | emit_boxes_linear__avx2(const struct sna_composite_op *op, |
||
1369 | const BoxRec *box, int nbox, |
||
1370 | float *v) |
||
1371 | { |
||
1372 | union { |
||
1373 | struct sna_coordinate p; |
||
1374 | float f; |
||
1375 | } dst; |
||
1376 | |||
1377 | do { |
||
1378 | dst.p.x = box->x2; |
||
1379 | dst.p.y = box->y2; |
||
1380 | v[0] = dst.f; |
||
1381 | dst.p.x = box->x1; |
||
1382 | v[2] = dst.f; |
||
1383 | dst.p.y = box->y1; |
||
1384 | v[4] = dst.f; |
||
1385 | |||
1386 | v[1] = compute_linear(&op->src, box->x2, box->y2); |
||
1387 | v[3] = compute_linear(&op->src, box->x1, box->y2); |
||
1388 | v[5] = compute_linear(&op->src, box->x1, box->y1); |
||
1389 | |||
1390 | v += 6; |
||
1391 | box++; |
||
1392 | } while (--nbox); |
||
1393 | } |
||
1394 | |||
1395 | avx2 fastcall static void |
||
1396 | emit_primitive_identity_source__avx2(struct sna *sna, |
||
1397 | const struct sna_composite_op *op, |
||
1398 | const struct sna_composite_rectangles *r) |
||
1399 | { |
||
1400 | union { |
||
1401 | struct sna_coordinate p; |
||
1402 | float f; |
||
1403 | } dst; |
||
1404 | float *v; |
||
1405 | |||
1406 | assert(op->floats_per_rect == 9); |
||
1407 | assert((sna->render.vertex_used % 3) == 0); |
||
1408 | v = sna->render.vertices + sna->render.vertex_used; |
||
1409 | sna->render.vertex_used += 9; |
||
1410 | |||
1411 | dst.p.x = r->dst.x + r->width; |
||
1412 | dst.p.y = r->dst.y + r->height; |
||
1413 | v[0] = dst.f; |
||
1414 | dst.p.x = r->dst.x; |
||
1415 | v[3] = dst.f; |
||
1416 | dst.p.y = r->dst.y; |
||
1417 | v[6] = dst.f; |
||
1418 | |||
1419 | v[7] = v[4] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
||
1420 | v[1] = v[4] + r->width * op->src.scale[0]; |
||
1421 | |||
1422 | v[8] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
||
1423 | v[5] = v[2] = v[8] + r->height * op->src.scale[1]; |
||
1424 | } |
||
1425 | |||
1426 | avx2 fastcall static void |
||
1427 | emit_boxes_identity_source__avx2(const struct sna_composite_op *op, |
||
1428 | const BoxRec *box, int nbox, |
||
1429 | float *v) |
||
1430 | { |
||
1431 | do { |
||
1432 | union { |
||
1433 | struct sna_coordinate p; |
||
1434 | float f; |
||
1435 | } dst; |
||
1436 | |||
1437 | dst.p.x = box->x2; |
||
1438 | dst.p.y = box->y2; |
||
1439 | v[0] = dst.f; |
||
1440 | dst.p.x = box->x1; |
||
1441 | v[3] = dst.f; |
||
1442 | dst.p.y = box->y1; |
||
1443 | v[6] = dst.f; |
||
1444 | |||
1445 | v[7] = v[4] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
||
1446 | v[1] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
||
1447 | |||
1448 | v[8] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
||
1449 | v[2] = v[5] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
||
1450 | |||
1451 | v += 9; |
||
1452 | box++; |
||
1453 | } while (--nbox); |
||
1454 | } |
||
1455 | |||
1456 | avx2 fastcall static void |
||
1457 | emit_primitive_simple_source__avx2(struct sna *sna, |
||
1458 | const struct sna_composite_op *op, |
||
1459 | const struct sna_composite_rectangles *r) |
||
1460 | { |
||
1461 | float *v; |
||
1462 | union { |
||
1463 | struct sna_coordinate p; |
||
1464 | float f; |
||
1465 | } dst; |
||
1466 | |||
1467 | float xx = op->src.transform->matrix[0][0]; |
||
1468 | float x0 = op->src.transform->matrix[0][2]; |
||
1469 | float yy = op->src.transform->matrix[1][1]; |
||
1470 | float y0 = op->src.transform->matrix[1][2]; |
||
1471 | float sx = op->src.scale[0]; |
||
1472 | float sy = op->src.scale[1]; |
||
1473 | int16_t tx = op->src.offset[0]; |
||
1474 | int16_t ty = op->src.offset[1]; |
||
1475 | |||
1476 | assert(op->floats_per_rect == 9); |
||
1477 | assert((sna->render.vertex_used % 3) == 0); |
||
1478 | v = sna->render.vertices + sna->render.vertex_used; |
||
1479 | sna->render.vertex_used += 3*3; |
||
1480 | |||
1481 | dst.p.x = r->dst.x + r->width; |
||
1482 | dst.p.y = r->dst.y + r->height; |
||
1483 | v[0] = dst.f; |
||
1484 | v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
||
1485 | v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
||
1486 | |||
1487 | dst.p.x = r->dst.x; |
||
1488 | v[3] = dst.f; |
||
1489 | v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx; |
||
1490 | |||
1491 | dst.p.y = r->dst.y; |
||
1492 | v[6] = dst.f; |
||
1493 | v[8] = ((r->src.y + ty) * yy + y0) * sy; |
||
1494 | } |
||
1495 | |||
1496 | avx2 fastcall static void |
||
1497 | emit_boxes_simple_source__avx2(const struct sna_composite_op *op, |
||
1498 | const BoxRec *box, int nbox, |
||
1499 | float *v) |
||
1500 | { |
||
1501 | float xx = op->src.transform->matrix[0][0]; |
||
1502 | float x0 = op->src.transform->matrix[0][2]; |
||
1503 | float yy = op->src.transform->matrix[1][1]; |
||
1504 | float y0 = op->src.transform->matrix[1][2]; |
||
1505 | float sx = op->src.scale[0]; |
||
1506 | float sy = op->src.scale[1]; |
||
1507 | int16_t tx = op->src.offset[0]; |
||
1508 | int16_t ty = op->src.offset[1]; |
||
1509 | |||
1510 | do { |
||
1511 | union { |
||
1512 | struct sna_coordinate p; |
||
1513 | float f; |
||
1514 | } dst; |
||
1515 | |||
1516 | dst.p.x = box->x2; |
||
1517 | dst.p.y = box->y2; |
||
1518 | v[0] = dst.f; |
||
1519 | v[1] = ((box->x2 + tx) * xx + x0) * sx; |
||
1520 | v[5] = v[2] = ((box->y2 + ty) * yy + y0) * sy; |
||
1521 | |||
1522 | dst.p.x = box->x1; |
||
1523 | v[3] = dst.f; |
||
1524 | v[7] = v[4] = ((box->x1 + tx) * xx + x0) * sx; |
||
1525 | |||
1526 | dst.p.y = box->y1; |
||
1527 | v[6] = dst.f; |
||
1528 | v[8] = ((box->y1 + ty) * yy + y0) * sy; |
||
1529 | |||
1530 | v += 9; |
||
1531 | box++; |
||
1532 | } while (--nbox); |
||
1533 | } |
||
1534 | |||
1535 | avx2 fastcall static void |
||
1536 | emit_primitive_identity_mask__avx2(struct sna *sna, |
||
1537 | const struct sna_composite_op *op, |
||
1538 | const struct sna_composite_rectangles *r) |
||
1539 | { |
||
1540 | union { |
||
1541 | struct sna_coordinate p; |
||
1542 | float f; |
||
1543 | } dst; |
||
1544 | float msk_x, msk_y; |
||
1545 | float w, h; |
||
1546 | float *v; |
||
1547 | |||
1548 | msk_x = r->mask.x + op->mask.offset[0]; |
||
1549 | msk_y = r->mask.y + op->mask.offset[1]; |
||
1550 | w = r->width; |
||
1551 | h = r->height; |
||
1552 | |||
1553 | DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
||
1554 | __FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
||
1555 | |||
1556 | assert(op->floats_per_rect == 12); |
||
1557 | assert((sna->render.vertex_used % 4) == 0); |
||
1558 | v = sna->render.vertices + sna->render.vertex_used; |
||
1559 | sna->render.vertex_used += 12; |
||
1560 | |||
1561 | dst.p.x = r->dst.x + r->width; |
||
1562 | dst.p.y = r->dst.y + r->height; |
||
1563 | v[0] = dst.f; |
||
1564 | v[2] = (msk_x + w) * op->mask.scale[0]; |
||
1565 | v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
||
1566 | |||
1567 | dst.p.x = r->dst.x; |
||
1568 | v[4] = dst.f; |
||
1569 | v[10] = v[6] = msk_x * op->mask.scale[0]; |
||
1570 | |||
1571 | dst.p.y = r->dst.y; |
||
1572 | v[8] = dst.f; |
||
1573 | v[11] = msk_y * op->mask.scale[1]; |
||
1574 | |||
1575 | v[9] = v[5] = v[1] = .5; |
||
1576 | } |
||
1577 | |||
1578 | avx2 fastcall static void |
||
1579 | emit_boxes_identity_mask__avx2(const struct sna_composite_op *op, |
||
1580 | const BoxRec *box, int nbox, |
||
1581 | float *v) |
||
1582 | { |
||
1583 | float msk_x = op->mask.offset[0]; |
||
1584 | float msk_y = op->mask.offset[1]; |
||
1585 | |||
1586 | do { |
||
1587 | union { |
||
1588 | struct sna_coordinate p; |
||
1589 | float f; |
||
1590 | } dst; |
||
1591 | |||
1592 | dst.p.x = box->x2; |
||
1593 | dst.p.y = box->y2; |
||
1594 | v[0] = dst.f; |
||
1595 | v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
||
1596 | v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
||
1597 | |||
1598 | dst.p.x = box->x1; |
||
1599 | v[4] = dst.f; |
||
1600 | v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
||
1601 | |||
1602 | dst.p.y = box->y1; |
||
1603 | v[8] = dst.f; |
||
1604 | v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
||
1605 | |||
1606 | v[9] = v[5] = v[1] = .5; |
||
1607 | v += 12; |
||
1608 | box++; |
||
1609 | } while (--nbox); |
||
1610 | } |
||
1611 | |||
1612 | avx2 fastcall static void |
||
1613 | emit_primitive_linear_identity_mask__avx2(struct sna *sna, |
||
1614 | const struct sna_composite_op *op, |
||
1615 | const struct sna_composite_rectangles *r) |
||
1616 | { |
||
1617 | union { |
||
1618 | struct sna_coordinate p; |
||
1619 | float f; |
||
1620 | } dst; |
||
1621 | float msk_x, msk_y; |
||
1622 | float w, h; |
||
1623 | float *v; |
||
1624 | |||
1625 | msk_x = r->mask.x + op->mask.offset[0]; |
||
1626 | msk_y = r->mask.y + op->mask.offset[1]; |
||
1627 | w = r->width; |
||
1628 | h = r->height; |
||
1629 | |||
1630 | DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
||
1631 | __FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
||
1632 | |||
1633 | assert(op->floats_per_rect == 12); |
||
1634 | assert((sna->render.vertex_used % 4) == 0); |
||
1635 | v = sna->render.vertices + sna->render.vertex_used; |
||
1636 | sna->render.vertex_used += 12; |
||
1637 | |||
1638 | dst.p.x = r->dst.x + r->width; |
||
1639 | dst.p.y = r->dst.y + r->height; |
||
1640 | v[0] = dst.f; |
||
1641 | v[2] = (msk_x + w) * op->mask.scale[0]; |
||
1642 | v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
||
1643 | |||
1644 | dst.p.x = r->dst.x; |
||
1645 | v[4] = dst.f; |
||
1646 | v[10] = v[6] = msk_x * op->mask.scale[0]; |
||
1647 | |||
1648 | dst.p.y = r->dst.y; |
||
1649 | v[8] = dst.f; |
||
1650 | v[11] = msk_y * op->mask.scale[1]; |
||
1651 | |||
1652 | v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
||
1653 | v[5] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
||
1654 | v[9] = compute_linear(&op->src, r->src.x, r->src.y); |
||
1655 | } |
||
1656 | |||
1657 | avx2 fastcall static void |
||
1658 | emit_boxes_linear_identity_mask__avx2(const struct sna_composite_op *op, |
||
1659 | const BoxRec *box, int nbox, |
||
1660 | float *v) |
||
1661 | { |
||
1662 | float msk_x = op->mask.offset[0]; |
||
1663 | float msk_y = op->mask.offset[1]; |
||
1664 | |||
1665 | do { |
||
1666 | union { |
||
1667 | struct sna_coordinate p; |
||
1668 | float f; |
||
1669 | } dst; |
||
1670 | |||
1671 | dst.p.x = box->x2; |
||
1672 | dst.p.y = box->y2; |
||
1673 | v[0] = dst.f; |
||
1674 | v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
||
1675 | v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
||
1676 | |||
1677 | dst.p.x = box->x1; |
||
1678 | v[4] = dst.f; |
||
1679 | v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
||
1680 | |||
1681 | dst.p.y = box->y1; |
||
1682 | v[8] = dst.f; |
||
1683 | v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
||
1684 | |||
1685 | v[1] = compute_linear(&op->src, box->x2, box->y2); |
||
1686 | v[5] = compute_linear(&op->src, box->x1, box->y2); |
||
1687 | v[9] = compute_linear(&op->src, box->x1, box->y1); |
||
1688 | |||
1689 | v += 12; |
||
1690 | box++; |
||
1691 | } while (--nbox); |
||
1692 | } |
||
1693 | |||
1694 | #endif |
||
1695 | |||
1696 | unsigned gen4_choose_composite_emitter(struct sna *sna, struct sna_composite_op *tmp) |
||
1697 | { |
||
3280 | Serge | 1698 | unsigned vb; |
1699 | |||
1700 | if (tmp->mask.bo) { |
||
1701 | if (tmp->mask.transform == NULL) { |
||
1702 | if (tmp->src.is_solid) { |
||
1703 | DBG(("%s: solid, identity mask\n", __FUNCTION__)); |
||
4251 | Serge | 1704 | #if defined(avx2) |
1705 | if (sna->cpu_features & AVX2) { |
||
1706 | tmp->prim_emit = emit_primitive_identity_mask__avx2; |
||
1707 | } else |
||
1708 | #endif |
||
1709 | #if defined(sse4_2) |
||
1710 | if (sna->cpu_features & SSE4_2) { |
||
1711 | tmp->prim_emit = emit_primitive_identity_mask__sse4_2; |
||
1712 | } else |
||
1713 | #endif |
||
1714 | { |
||
1715 | tmp->prim_emit = emit_primitive_identity_mask; |
||
1716 | } |
||
1717 | tmp->floats_per_vertex = 4; |
||
1718 | vb = 1 | 2 << 2; |
||
3280 | Serge | 1719 | } else if (tmp->src.is_linear) { |
1720 | DBG(("%s: linear, identity mask\n", __FUNCTION__)); |
||
4251 | Serge | 1721 | #if defined(avx2) |
1722 | if (sna->cpu_features & AVX2) { |
||
1723 | tmp->prim_emit = emit_primitive_linear_identity_mask__avx2; |
||
1724 | } else |
||
1725 | #endif |
||
1726 | #if defined(sse4_2) |
||
1727 | if (sna->cpu_features & SSE4_2) { |
||
1728 | tmp->prim_emit = emit_primitive_linear_identity_mask__sse4_2; |
||
1729 | } else |
||
1730 | #endif |
||
1731 | { |
||
1732 | tmp->prim_emit = emit_primitive_linear_identity_mask; |
||
1733 | } |
||
1734 | tmp->floats_per_vertex = 4; |
||
1735 | vb = 1 | 2 << 2; |
||
3280 | Serge | 1736 | } else if (tmp->src.transform == NULL) { |
1737 | DBG(("%s: identity source, identity mask\n", __FUNCTION__)); |
||
1738 | tmp->prim_emit = emit_primitive_identity_source_mask; |
||
1739 | tmp->floats_per_vertex = 5; |
||
1740 | vb = 2 << 2 | 2; |
||
1741 | } else if (tmp->src.is_affine) { |
||
4251 | Serge | 1742 | tmp->src.scale[0] /= tmp->src.transform->matrix[2][2]; |
1743 | tmp->src.scale[1] /= tmp->src.transform->matrix[2][2]; |
||
1744 | if (!sna_affine_transform_is_rotation(tmp->src.transform)) { |
||
1745 | DBG(("%s: simple src, identity mask\n", __FUNCTION__)); |
||
1746 | tmp->prim_emit = emit_primitive_simple_source_identity; |
||
1747 | } else { |
||
1748 | DBG(("%s: affine src, identity mask\n", __FUNCTION__)); |
||
1749 | tmp->prim_emit = emit_primitive_affine_source_identity; |
||
1750 | } |
||
1751 | tmp->floats_per_vertex = 5; |
||
1752 | vb = 2 << 2 | 2; |
||
3280 | Serge | 1753 | } else { |
1754 | DBG(("%s: projective source, identity mask\n", __FUNCTION__)); |
||
4251 | Serge | 1755 | tmp->prim_emit = emit_primitive_mask; |
1756 | tmp->floats_per_vertex = 6; |
||
1757 | vb = 2 << 2 | 3; |
||
3280 | Serge | 1758 | } |
1759 | } else { |
||
4251 | Serge | 1760 | tmp->prim_emit = emit_primitive_mask; |
1761 | tmp->floats_per_vertex = 1; |
||
1762 | vb = 0; |
||
1763 | if (tmp->mask.is_solid) { |
||
1764 | tmp->floats_per_vertex += 1; |
||
1765 | vb |= 1 << 2; |
||
1766 | } else if (tmp->mask.is_affine) { |
||
1767 | tmp->floats_per_vertex += 2; |
||
1768 | vb |= 2 << 2; |
||
1769 | }else { |
||
1770 | tmp->floats_per_vertex += 3; |
||
1771 | vb |= 3 << 2; |
||
1772 | } |
||
1773 | if (tmp->src.is_solid) { |
||
1774 | tmp->floats_per_vertex += 1; |
||
1775 | vb |= 1; |
||
1776 | } else if (tmp->src.is_affine) { |
||
1777 | tmp->floats_per_vertex += 2; |
||
1778 | vb |= 2 ; |
||
1779 | }else { |
||
1780 | tmp->floats_per_vertex += 3; |
||
1781 | vb |= 3; |
||
1782 | } |
||
3280 | Serge | 1783 | DBG(("%s: general mask: floats-per-vertex=%d, vb=%x\n", |
1784 | __FUNCTION__,tmp->floats_per_vertex, vb)); |
||
1785 | } |
||
1786 | } else { |
||
4251 | Serge | 1787 | #if 0 |
1788 | if (tmp->src.is_solid) { |
||
1789 | DBG(("%s: solid, no mask\n", __FUNCTION__)); |
||
1790 | tmp->prim_emit = emit_primitive_solid; |
||
1791 | if (tmp->src.is_opaque && tmp->op == PictOpOver) |
||
1792 | tmp->op = PictOpSrc; |
||
1793 | tmp->floats_per_vertex = 2; |
||
1794 | vb = 1; |
||
1795 | } else if (tmp->src.is_linear) { |
||
1796 | DBG(("%s: linear, no mask\n", __FUNCTION__)); |
||
1797 | #if defined(avx2) |
||
1798 | if (sna->cpu_features & AVX2) { |
||
1799 | tmp->prim_emit = emit_primitive_linear__avx2; |
||
1800 | } else |
||
1801 | #endif |
||
1802 | #if defined(sse4_2) |
||
1803 | if (sna->cpu_features & SSE4_2) { |
||
1804 | tmp->prim_emit = emit_primitive_linear__sse4_2; |
||
1805 | } else |
||
1806 | #endif |
||
1807 | { |
||
1808 | tmp->prim_emit = emit_primitive_linear; |
||
1809 | } |
||
1810 | tmp->floats_per_vertex = 2; |
||
1811 | vb = 1; |
||
1812 | } else if (tmp->src.transform == NULL) { |
||
1813 | DBG(("%s: identity src, no mask\n", __FUNCTION__)); |
||
1814 | #if defined(avx2) |
||
1815 | if (sna->cpu_features & AVX2) { |
||
1816 | tmp->prim_emit = emit_primitive_identity_source__avx2; |
||
1817 | } else |
||
1818 | #endif |
||
1819 | #if defined(sse4_2) |
||
1820 | if (sna->cpu_features & SSE4_2) { |
||
1821 | tmp->prim_emit = emit_primitive_identity_source__sse4_2; |
||
1822 | } else |
||
1823 | #endif |
||
1824 | { |
||
1825 | tmp->prim_emit = emit_primitive_identity_source; |
||
1826 | } |
||
1827 | tmp->floats_per_vertex = 3; |
||
1828 | vb = 2; |
||
1829 | } else if (tmp->src.is_affine) { |
||
1830 | tmp->src.scale[0] /= tmp->src.transform->matrix[2][2]; |
||
1831 | tmp->src.scale[1] /= tmp->src.transform->matrix[2][2]; |
||
1832 | if (!sna_affine_transform_is_rotation(tmp->src.transform)) { |
||
1833 | DBG(("%s: simple src, no mask\n", __FUNCTION__)); |
||
1834 | #if defined(avx2) |
||
1835 | if (sna->cpu_features & AVX2) { |
||
1836 | tmp->prim_emit = emit_primitive_simple_source__avx2; |
||
1837 | } else |
||
1838 | #endif |
||
1839 | #if defined(sse4_2) |
||
1840 | if (sna->cpu_features & SSE4_2) { |
||
1841 | tmp->prim_emit = emit_primitive_simple_source__sse4_2; |
||
1842 | } else |
||
1843 | #endif |
||
1844 | { |
||
1845 | tmp->prim_emit = emit_primitive_simple_source; |
||
1846 | } |
||
1847 | } else { |
||
1848 | DBG(("%s: affine src, no mask\n", __FUNCTION__)); |
||
1849 | tmp->prim_emit = emit_primitive_affine_source; |
||
1850 | } |
||
1851 | tmp->floats_per_vertex = 3; |
||
1852 | vb = 2; |
||
1853 | } else { |
||
1854 | DBG(("%s: projective src, no mask\n", __FUNCTION__)); |
||
1855 | assert(!tmp->src.is_solid); |
||
1856 | tmp->prim_emit = emit_primitive; |
||
1857 | tmp->floats_per_vertex = 4; |
||
1858 | vb = 3; |
||
1859 | } |
||
1860 | #endif |
||
3280 | Serge | 1861 | } |
1862 | tmp->floats_per_rect = 3 * tmp->floats_per_vertex; |
||
1863 | |||
1864 | return vb; |
||
1865 | }><>><>><>><>><>><>><>><>=>=>=>=>><>>=>>= |
||
4251 | Serge | 1866 | |
1867 |