Merge remote branch 'origin/opengl-es-v2'
[mesa.git] / src / gallium / auxiliary / draw / draw_pt_fetch_emit.c
1 /**************************************************************************
2 *
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 /*
29 * Authors:
30 * Keith Whitwell <keith@tungstengraphics.com>
31 */
32
33 #include "util/u_memory.h"
34 #include "draw/draw_context.h"
35 #include "draw/draw_private.h"
36 #include "draw/draw_vbuf.h"
37 #include "draw/draw_vertex.h"
38 #include "draw/draw_pt.h"
39 #include "translate/translate.h"
40 #include "translate/translate_cache.h"
41
42 /* The simplest 'middle end' in the new vertex code.
43 *
44 * The responsibilities of a middle end are to:
45 * - perform vertex fetch using
46 * - draw vertex element/buffer state
47 * - a list of fetch indices we received as an input
48 * - run the vertex shader
49 * - cliptest,
50 * - clip coord calculation
51 * - viewport transformation
52 * - if necessary, run the primitive pipeline, passing it:
53 * - a linear array of vertex_header vertices constructed here
54 * - a set of draw indices we received as an input
55 * - otherwise, drive the hw backend,
56 * - allocate space for hardware format vertices
57 * - translate the vertex-shader output vertices to hw format
58 * - calling the backend draw functions.
59 *
60 * For convenience, we provide a helper function to drive the hardware
61 * backend given similar inputs to those required to run the pipeline.
62 *
63 * In the case of passthrough mode, many of these actions are disabled
64 * or noops, so we end up doing:
65 *
66 * - perform vertex fetch
67 * - drive the hw backend
68 *
69 * IE, basically just vertex fetch to post-vs-format vertices,
70 * followed by a call to the backend helper function.
71 */
72
73
74 struct fetch_emit_middle_end {
75 struct draw_pt_middle_end base;
76 struct draw_context *draw;
77
78 struct translate *translate;
79 const struct vertex_info *vinfo;
80
81 /* Cache point size somewhere it's address won't change:
82 */
83 float point_size;
84
85 struct translate_cache *cache;
86 };
87
88
89
90
91 static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
92 unsigned prim,
93 unsigned opt,
94 unsigned *max_vertices )
95 {
96 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
97 struct draw_context *draw = feme->draw;
98 const struct vertex_info *vinfo;
99 unsigned i, dst_offset;
100 boolean ok;
101 struct translate_key key;
102
103
104 ok = draw->render->set_primitive( draw->render,
105 prim );
106 if (!ok) {
107 assert(0);
108 return;
109 }
110
111 /* Must do this after set_primitive() above:
112 */
113 vinfo = feme->vinfo = draw->render->get_vertex_info(draw->render);
114
115
116
117 /* Transform from API vertices to HW vertices, skipping the
118 * pipeline_vertex intermediate step.
119 */
120 dst_offset = 0;
121 memset(&key, 0, sizeof(key));
122
123 for (i = 0; i < vinfo->num_attribs; i++) {
124 const struct pipe_vertex_element *src = &draw->pt.vertex_element[vinfo->attrib[i].src_index];
125
126 unsigned emit_sz = 0;
127 unsigned input_format = src->src_format;
128 unsigned input_buffer = src->vertex_buffer_index;
129 unsigned input_offset = src->src_offset;
130 unsigned output_format;
131
132 switch (vinfo->attrib[i].emit) {
133 case EMIT_4UB:
134 output_format = PIPE_FORMAT_R8G8B8A8_UNORM;
135 emit_sz = 4 * sizeof(unsigned char);
136 break;
137 case EMIT_4F:
138 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
139 emit_sz = 4 * sizeof(float);
140 break;
141 case EMIT_3F:
142 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
143 emit_sz = 3 * sizeof(float);
144 break;
145 case EMIT_2F:
146 output_format = PIPE_FORMAT_R32G32_FLOAT;
147 emit_sz = 2 * sizeof(float);
148 break;
149 case EMIT_1F:
150 output_format = PIPE_FORMAT_R32_FLOAT;
151 emit_sz = 1 * sizeof(float);
152 break;
153 case EMIT_1F_PSIZE:
154 input_format = PIPE_FORMAT_R32_FLOAT;
155 input_buffer = draw->pt.nr_vertex_buffers;
156 input_offset = 0;
157 output_format = PIPE_FORMAT_R32_FLOAT;
158 emit_sz = 1 * sizeof(float);
159 break;
160 case EMIT_OMIT:
161 continue;
162 default:
163 assert(0);
164 output_format = PIPE_FORMAT_NONE;
165 emit_sz = 0;
166 continue;
167 }
168
169 key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
170 key.element[i].input_format = input_format;
171 key.element[i].input_buffer = input_buffer;
172 key.element[i].input_offset = input_offset;
173 key.element[i].instance_divisor = src->instance_divisor;
174 key.element[i].output_format = output_format;
175 key.element[i].output_offset = dst_offset;
176
177 dst_offset += emit_sz;
178 }
179
180 key.nr_elements = vinfo->num_attribs;
181 key.output_stride = vinfo->size * 4;
182
183 /* Don't bother with caching at this stage:
184 */
185 if (!feme->translate ||
186 translate_key_compare(&feme->translate->key, &key) != 0)
187 {
188 translate_key_sanitize(&key);
189 feme->translate = translate_cache_find(feme->cache,
190 &key);
191
192
193 feme->translate->set_buffer(feme->translate,
194 draw->pt.nr_vertex_buffers,
195 &feme->point_size,
196 0);
197 }
198
199 feme->point_size = draw->rasterizer->point_size;
200
201 for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
202 feme->translate->set_buffer(feme->translate,
203 i,
204 ((char *)draw->pt.user.vbuffer[i] +
205 draw->pt.vertex_buffer[i].buffer_offset),
206 draw->pt.vertex_buffer[i].stride );
207 }
208
209 *max_vertices = (draw->render->max_vertex_buffer_bytes /
210 (vinfo->size * 4));
211
212 /* Return an even number of verts.
213 * This prevents "parity" errors when splitting long triangle strips which
214 * can lead to front/back culling mix-ups.
215 * Every other triangle in a strip has an alternate front/back orientation
216 * so splitting at an odd position can cause the orientation of subsequent
217 * triangles to get reversed.
218 */
219 *max_vertices = *max_vertices & ~1;
220 }
221
222
223
224
225
226 static void fetch_emit_run( struct draw_pt_middle_end *middle,
227 const unsigned *fetch_elts,
228 unsigned fetch_count,
229 const ushort *draw_elts,
230 unsigned draw_count )
231 {
232 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
233 struct draw_context *draw = feme->draw;
234 void *hw_verts;
235
236 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
237 */
238 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
239
240 if (fetch_count >= UNDEFINED_VERTEX_ID) {
241 assert(0);
242 return;
243 }
244
245 draw->render->allocate_vertices( draw->render,
246 (ushort)feme->translate->key.output_stride,
247 (ushort)fetch_count );
248
249 hw_verts = draw->render->map_vertices( draw->render );
250 if (!hw_verts) {
251 assert(0);
252 return;
253 }
254
255
256 /* Single routine to fetch vertices and emit HW verts.
257 */
258 feme->translate->run_elts( feme->translate,
259 fetch_elts,
260 fetch_count,
261 draw->instance_id,
262 hw_verts );
263
264 if (0) {
265 unsigned i;
266 for (i = 0; i < fetch_count; i++) {
267 debug_printf("\n\nvertex %d:\n", i);
268 draw_dump_emitted_vertex( feme->vinfo,
269 (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
270 }
271 }
272
273 draw->render->unmap_vertices( draw->render,
274 0,
275 (ushort)(fetch_count - 1) );
276
277 /* XXX: Draw arrays path to avoid re-emitting index list again and
278 * again.
279 */
280 draw->render->draw( draw->render,
281 draw_elts,
282 draw_count );
283
284 /* Done -- that was easy, wasn't it:
285 */
286 draw->render->release_vertices( draw->render );
287
288 }
289
290
291 static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
292 unsigned start,
293 unsigned count )
294 {
295 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
296 struct draw_context *draw = feme->draw;
297 void *hw_verts;
298
299 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
300 */
301 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
302
303 if (count >= UNDEFINED_VERTEX_ID)
304 goto fail;
305
306 if (!draw->render->allocate_vertices( draw->render,
307 (ushort)feme->translate->key.output_stride,
308 (ushort)count ))
309 goto fail;
310
311 hw_verts = draw->render->map_vertices( draw->render );
312 if (!hw_verts)
313 goto fail;
314
315 /* Single routine to fetch vertices and emit HW verts.
316 */
317 feme->translate->run( feme->translate,
318 start,
319 count,
320 draw->instance_id,
321 hw_verts );
322
323 if (0) {
324 unsigned i;
325 for (i = 0; i < count; i++) {
326 debug_printf("\n\nvertex %d:\n", i);
327 draw_dump_emitted_vertex( feme->vinfo,
328 (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
329 }
330 }
331
332 draw->render->unmap_vertices( draw->render, 0, count - 1 );
333
334 /* XXX: Draw arrays path to avoid re-emitting index list again and
335 * again.
336 */
337 draw->render->draw_arrays( draw->render, 0, count );
338
339 /* Done -- that was easy, wasn't it:
340 */
341 draw->render->release_vertices( draw->render );
342 return;
343
344 fail:
345 assert(0);
346 return;
347 }
348
349
350 static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
351 unsigned start,
352 unsigned count,
353 const ushort *draw_elts,
354 unsigned draw_count )
355 {
356 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
357 struct draw_context *draw = feme->draw;
358 void *hw_verts;
359
360 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
361 */
362 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
363
364 if (count >= UNDEFINED_VERTEX_ID)
365 return FALSE;
366
367 if (!draw->render->allocate_vertices( draw->render,
368 (ushort)feme->translate->key.output_stride,
369 (ushort)count ))
370 return FALSE;
371
372 hw_verts = draw->render->map_vertices( draw->render );
373 if (!hw_verts)
374 return FALSE;
375
376 /* Single routine to fetch vertices and emit HW verts.
377 */
378 feme->translate->run( feme->translate,
379 start,
380 count,
381 draw->instance_id,
382 hw_verts );
383
384 draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
385
386 /* XXX: Draw arrays path to avoid re-emitting index list again and
387 * again.
388 */
389 draw->render->draw( draw->render,
390 draw_elts,
391 draw_count );
392
393 /* Done -- that was easy, wasn't it:
394 */
395 draw->render->release_vertices( draw->render );
396
397 return TRUE;
398 }
399
400
401
402
403 static void fetch_emit_finish( struct draw_pt_middle_end *middle )
404 {
405 /* nothing to do */
406 }
407
408 static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
409 {
410 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
411
412 if (feme->cache)
413 translate_cache_destroy(feme->cache);
414
415 FREE(middle);
416 }
417
418
419 struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw )
420 {
421 struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end );
422 if (fetch_emit == NULL)
423 return NULL;
424
425 fetch_emit->cache = translate_cache_create();
426 if (!fetch_emit->cache) {
427 FREE(fetch_emit);
428 return NULL;
429 }
430
431 fetch_emit->base.prepare = fetch_emit_prepare;
432 fetch_emit->base.run = fetch_emit_run;
433 fetch_emit->base.run_linear = fetch_emit_run_linear;
434 fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts;
435 fetch_emit->base.finish = fetch_emit_finish;
436 fetch_emit->base.destroy = fetch_emit_destroy;
437
438 fetch_emit->draw = draw;
439
440 return &fetch_emit->base;
441 }
442