Merge remote branch 'origin/7.8'
[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 output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
133 emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
134
135 if (vinfo->attrib[i].emit == EMIT_OMIT)
136 continue;
137
138 if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
139 input_format = PIPE_FORMAT_R32_FLOAT;
140 input_buffer = draw->pt.nr_vertex_buffers;
141 input_offset = 0;
142 }
143
144 key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
145 key.element[i].input_format = input_format;
146 key.element[i].input_buffer = input_buffer;
147 key.element[i].input_offset = input_offset;
148 key.element[i].instance_divisor = src->instance_divisor;
149 key.element[i].output_format = output_format;
150 key.element[i].output_offset = dst_offset;
151
152 dst_offset += emit_sz;
153 }
154
155 key.nr_elements = vinfo->num_attribs;
156 key.output_stride = vinfo->size * 4;
157
158 /* Don't bother with caching at this stage:
159 */
160 if (!feme->translate ||
161 translate_key_compare(&feme->translate->key, &key) != 0)
162 {
163 translate_key_sanitize(&key);
164 feme->translate = translate_cache_find(feme->cache,
165 &key);
166
167
168 feme->translate->set_buffer(feme->translate,
169 draw->pt.nr_vertex_buffers,
170 &feme->point_size,
171 0);
172 }
173
174 feme->point_size = draw->rasterizer->point_size;
175
176 for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
177 feme->translate->set_buffer(feme->translate,
178 i,
179 ((char *)draw->pt.user.vbuffer[i] +
180 draw->pt.vertex_buffer[i].buffer_offset),
181 draw->pt.vertex_buffer[i].stride );
182 }
183
184 *max_vertices = (draw->render->max_vertex_buffer_bytes /
185 (vinfo->size * 4));
186
187 /* Return an even number of verts.
188 * This prevents "parity" errors when splitting long triangle strips which
189 * can lead to front/back culling mix-ups.
190 * Every other triangle in a strip has an alternate front/back orientation
191 * so splitting at an odd position can cause the orientation of subsequent
192 * triangles to get reversed.
193 */
194 *max_vertices = *max_vertices & ~1;
195 }
196
197
198
199
200
201 static void fetch_emit_run( struct draw_pt_middle_end *middle,
202 const unsigned *fetch_elts,
203 unsigned fetch_count,
204 const ushort *draw_elts,
205 unsigned draw_count )
206 {
207 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
208 struct draw_context *draw = feme->draw;
209 void *hw_verts;
210
211 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
212 */
213 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
214
215 if (fetch_count >= UNDEFINED_VERTEX_ID) {
216 assert(0);
217 return;
218 }
219
220 draw->render->allocate_vertices( draw->render,
221 (ushort)feme->translate->key.output_stride,
222 (ushort)fetch_count );
223
224 hw_verts = draw->render->map_vertices( draw->render );
225 if (!hw_verts) {
226 assert(0);
227 return;
228 }
229
230
231 /* Single routine to fetch vertices and emit HW verts.
232 */
233 feme->translate->run_elts( feme->translate,
234 fetch_elts,
235 fetch_count,
236 draw->instance_id,
237 hw_verts );
238
239 if (0) {
240 unsigned i;
241 for (i = 0; i < fetch_count; i++) {
242 debug_printf("\n\nvertex %d:\n", i);
243 draw_dump_emitted_vertex( feme->vinfo,
244 (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
245 }
246 }
247
248 draw->render->unmap_vertices( draw->render,
249 0,
250 (ushort)(fetch_count - 1) );
251
252 /* XXX: Draw arrays path to avoid re-emitting index list again and
253 * again.
254 */
255 draw->render->draw( draw->render,
256 draw_elts,
257 draw_count );
258
259 /* Done -- that was easy, wasn't it:
260 */
261 draw->render->release_vertices( draw->render );
262
263 }
264
265
266 static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
267 unsigned start,
268 unsigned count )
269 {
270 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
271 struct draw_context *draw = feme->draw;
272 void *hw_verts;
273
274 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
275 */
276 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
277
278 if (count >= UNDEFINED_VERTEX_ID)
279 goto fail;
280
281 if (!draw->render->allocate_vertices( draw->render,
282 (ushort)feme->translate->key.output_stride,
283 (ushort)count ))
284 goto fail;
285
286 hw_verts = draw->render->map_vertices( draw->render );
287 if (!hw_verts)
288 goto fail;
289
290 /* Single routine to fetch vertices and emit HW verts.
291 */
292 feme->translate->run( feme->translate,
293 start,
294 count,
295 draw->instance_id,
296 hw_verts );
297
298 if (0) {
299 unsigned i;
300 for (i = 0; i < count; i++) {
301 debug_printf("\n\nvertex %d:\n", i);
302 draw_dump_emitted_vertex( feme->vinfo,
303 (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
304 }
305 }
306
307 draw->render->unmap_vertices( draw->render, 0, count - 1 );
308
309 /* XXX: Draw arrays path to avoid re-emitting index list again and
310 * again.
311 */
312 draw->render->draw_arrays( draw->render, 0, count );
313
314 /* Done -- that was easy, wasn't it:
315 */
316 draw->render->release_vertices( draw->render );
317 return;
318
319 fail:
320 assert(0);
321 return;
322 }
323
324
325 static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
326 unsigned start,
327 unsigned count,
328 const ushort *draw_elts,
329 unsigned draw_count )
330 {
331 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
332 struct draw_context *draw = feme->draw;
333 void *hw_verts;
334
335 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
336 */
337 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
338
339 if (count >= UNDEFINED_VERTEX_ID)
340 return FALSE;
341
342 if (!draw->render->allocate_vertices( draw->render,
343 (ushort)feme->translate->key.output_stride,
344 (ushort)count ))
345 return FALSE;
346
347 hw_verts = draw->render->map_vertices( draw->render );
348 if (!hw_verts)
349 return FALSE;
350
351 /* Single routine to fetch vertices and emit HW verts.
352 */
353 feme->translate->run( feme->translate,
354 start,
355 count,
356 draw->instance_id,
357 hw_verts );
358
359 draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
360
361 /* XXX: Draw arrays path to avoid re-emitting index list again and
362 * again.
363 */
364 draw->render->draw( draw->render,
365 draw_elts,
366 draw_count );
367
368 /* Done -- that was easy, wasn't it:
369 */
370 draw->render->release_vertices( draw->render );
371
372 return TRUE;
373 }
374
375
376
377
378 static void fetch_emit_finish( struct draw_pt_middle_end *middle )
379 {
380 /* nothing to do */
381 }
382
383 static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
384 {
385 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
386
387 if (feme->cache)
388 translate_cache_destroy(feme->cache);
389
390 FREE(middle);
391 }
392
393
394 struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw )
395 {
396 struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end );
397 if (fetch_emit == NULL)
398 return NULL;
399
400 fetch_emit->cache = translate_cache_create();
401 if (!fetch_emit->cache) {
402 FREE(fetch_emit);
403 return NULL;
404 }
405
406 fetch_emit->base.prepare = fetch_emit_prepare;
407 fetch_emit->base.run = fetch_emit_run;
408 fetch_emit->base.run_linear = fetch_emit_run_linear;
409 fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts;
410 fetch_emit->base.finish = fetch_emit_finish;
411 fetch_emit->base.destroy = fetch_emit_destroy;
412
413 fetch_emit->draw = draw;
414
415 return &fetch_emit->base;
416 }
417