Merge branch 'mesa_7_5_branch'
[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].input_format = input_format;
170 key.element[i].input_buffer = input_buffer;
171 key.element[i].input_offset = input_offset;
172 key.element[i].output_format = output_format;
173 key.element[i].output_offset = dst_offset;
174
175 dst_offset += emit_sz;
176 }
177
178 key.nr_elements = vinfo->num_attribs;
179 key.output_stride = vinfo->size * 4;
180
181 /* Don't bother with caching at this stage:
182 */
183 if (!feme->translate ||
184 translate_key_compare(&feme->translate->key, &key) != 0)
185 {
186 translate_key_sanitize(&key);
187 feme->translate = translate_cache_find(feme->cache,
188 &key);
189
190
191 feme->translate->set_buffer(feme->translate,
192 draw->pt.nr_vertex_buffers,
193 &feme->point_size,
194 0);
195 }
196
197 feme->point_size = draw->rasterizer->point_size;
198
199 for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
200 feme->translate->set_buffer(feme->translate,
201 i,
202 ((char *)draw->pt.user.vbuffer[i] +
203 draw->pt.vertex_buffer[i].buffer_offset),
204 draw->pt.vertex_buffer[i].stride );
205 }
206
207 *max_vertices = (draw->render->max_vertex_buffer_bytes /
208 (vinfo->size * 4));
209
210 /* Return an even number of verts.
211 * This prevents "parity" errors when splitting long triangle strips which
212 * can lead to front/back culling mix-ups.
213 * Every other triangle in a strip has an alternate front/back orientation
214 * so splitting at an odd position can cause the orientation of subsequent
215 * triangles to get reversed.
216 */
217 *max_vertices = *max_vertices & ~1;
218 }
219
220
221
222
223
224 static void fetch_emit_run( struct draw_pt_middle_end *middle,
225 const unsigned *fetch_elts,
226 unsigned fetch_count,
227 const ushort *draw_elts,
228 unsigned draw_count )
229 {
230 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
231 struct draw_context *draw = feme->draw;
232 void *hw_verts;
233
234 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
235 */
236 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
237
238 if (fetch_count >= UNDEFINED_VERTEX_ID) {
239 assert(0);
240 return;
241 }
242
243 draw->render->allocate_vertices( draw->render,
244 (ushort)feme->translate->key.output_stride,
245 (ushort)fetch_count );
246
247 hw_verts = draw->render->map_vertices( draw->render );
248 if (!hw_verts) {
249 assert(0);
250 return;
251 }
252
253
254 /* Single routine to fetch vertices and emit HW verts.
255 */
256 feme->translate->run_elts( feme->translate,
257 fetch_elts,
258 fetch_count,
259 hw_verts );
260
261 if (0) {
262 unsigned i;
263 for (i = 0; i < fetch_count; i++) {
264 debug_printf("\n\nvertex %d:\n", i);
265 draw_dump_emitted_vertex( feme->vinfo,
266 (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
267 }
268 }
269
270 draw->render->unmap_vertices( draw->render,
271 0,
272 (ushort)(fetch_count - 1) );
273
274 /* XXX: Draw arrays path to avoid re-emitting index list again and
275 * again.
276 */
277 draw->render->draw( draw->render,
278 draw_elts,
279 draw_count );
280
281 /* Done -- that was easy, wasn't it:
282 */
283 draw->render->release_vertices( draw->render );
284
285 }
286
287
288 static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
289 unsigned start,
290 unsigned count )
291 {
292 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
293 struct draw_context *draw = feme->draw;
294 void *hw_verts;
295
296 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
297 */
298 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
299
300 if (count >= UNDEFINED_VERTEX_ID)
301 goto fail;
302
303 if (!draw->render->allocate_vertices( draw->render,
304 (ushort)feme->translate->key.output_stride,
305 (ushort)count ))
306 goto fail;
307
308 hw_verts = draw->render->map_vertices( draw->render );
309 if (!hw_verts)
310 goto fail;
311
312 /* Single routine to fetch vertices and emit HW verts.
313 */
314 feme->translate->run( feme->translate,
315 start,
316 count,
317 hw_verts );
318
319 if (0) {
320 unsigned i;
321 for (i = 0; i < count; i++) {
322 debug_printf("\n\nvertex %d:\n", i);
323 draw_dump_emitted_vertex( feme->vinfo,
324 (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
325 }
326 }
327
328 draw->render->unmap_vertices( draw->render, 0, count - 1 );
329
330 /* XXX: Draw arrays path to avoid re-emitting index list again and
331 * again.
332 */
333 draw->render->draw_arrays( draw->render, 0, count );
334
335 /* Done -- that was easy, wasn't it:
336 */
337 draw->render->release_vertices( draw->render );
338 return;
339
340 fail:
341 assert(0);
342 return;
343 }
344
345
346 static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
347 unsigned start,
348 unsigned count,
349 const ushort *draw_elts,
350 unsigned draw_count )
351 {
352 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
353 struct draw_context *draw = feme->draw;
354 void *hw_verts;
355
356 /* XXX: need to flush to get prim_vbuf.c to release its allocation??
357 */
358 draw_do_flush( draw, DRAW_FLUSH_BACKEND );
359
360 if (count >= UNDEFINED_VERTEX_ID)
361 return FALSE;
362
363 if (!draw->render->allocate_vertices( draw->render,
364 (ushort)feme->translate->key.output_stride,
365 (ushort)count ))
366 return FALSE;
367
368 hw_verts = draw->render->map_vertices( draw->render );
369 if (!hw_verts)
370 return FALSE;
371
372 /* Single routine to fetch vertices and emit HW verts.
373 */
374 feme->translate->run( feme->translate,
375 start,
376 count,
377 hw_verts );
378
379 draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
380
381 /* XXX: Draw arrays path to avoid re-emitting index list again and
382 * again.
383 */
384 draw->render->draw( draw->render,
385 draw_elts,
386 draw_count );
387
388 /* Done -- that was easy, wasn't it:
389 */
390 draw->render->release_vertices( draw->render );
391
392 return TRUE;
393 }
394
395
396
397
398 static void fetch_emit_finish( struct draw_pt_middle_end *middle )
399 {
400 /* nothing to do */
401 }
402
403 static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
404 {
405 struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
406
407 if (feme->cache)
408 translate_cache_destroy(feme->cache);
409
410 FREE(middle);
411 }
412
413
414 struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw )
415 {
416 struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end );
417 if (fetch_emit == NULL)
418 return NULL;
419
420 fetch_emit->cache = translate_cache_create();
421 if (!fetch_emit->cache) {
422 FREE(fetch_emit);
423 return NULL;
424 }
425
426 fetch_emit->base.prepare = fetch_emit_prepare;
427 fetch_emit->base.run = fetch_emit_run;
428 fetch_emit->base.run_linear = fetch_emit_run_linear;
429 fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts;
430 fetch_emit->base.finish = fetch_emit_finish;
431 fetch_emit->base.destroy = fetch_emit_destroy;
432
433 fetch_emit->draw = draw;
434
435 return &fetch_emit->base;
436 }
437