draw_pt_fetch_emit.c revision 488dd2c1912132fe7ee5e81b05fb64ba62a46098
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 "draw/draw_gs.h"
40#include "translate/translate.h"
41#include "translate/translate_cache.h"
42
43/* The simplest 'middle end' in the new vertex code.
44 *
45 * The responsibilities of a middle end are to:
46 *  - perform vertex fetch using
47 *       - draw vertex element/buffer state
48 *       - a list of fetch indices we received as an input
49 *  - run the vertex shader
50 *  - cliptest,
51 *  - clip coord calculation
52 *  - viewport transformation
53 *  - if necessary, run the primitive pipeline, passing it:
54 *       - a linear array of vertex_header vertices constructed here
55 *       - a set of draw indices we received as an input
56 *  - otherwise, drive the hw backend,
57 *       - allocate space for hardware format vertices
58 *       - translate the vertex-shader output vertices to hw format
59 *       - calling the backend draw functions.
60 *
61 * For convenience, we provide a helper function to drive the hardware
62 * backend given similar inputs to those required to run the pipeline.
63 *
64 * In the case of passthrough mode, many of these actions are disabled
65 * or noops, so we end up doing:
66 *
67 *  - perform vertex fetch
68 *  - drive the hw backend
69 *
70 * IE, basically just vertex fetch to post-vs-format vertices,
71 * followed by a call to the backend helper function.
72 */
73
74
75struct fetch_emit_middle_end {
76   struct draw_pt_middle_end base;
77   struct draw_context *draw;
78
79   struct translate *translate;
80   const struct vertex_info *vinfo;
81
82   /* Cache point size somewhere it's address won't change:
83    */
84   float point_size;
85
86   struct translate_cache *cache;
87};
88
89
90
91
92static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
93                                unsigned prim,
94				unsigned opt,
95                                unsigned *max_vertices )
96{
97   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
98   struct draw_context *draw = feme->draw;
99   const struct vertex_info *vinfo;
100   unsigned i, dst_offset;
101   struct translate_key key;
102
103   unsigned gs_out_prim = (draw->gs.geometry_shader ?
104                           draw->gs.geometry_shader->output_primitive :
105                           prim);
106
107
108
109   draw->render->set_primitive(draw->render, gs_out_prim);
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				  ~0);
173   }
174
175   feme->point_size = draw->rasterizer->point_size;
176
177   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
178      feme->translate->set_buffer(feme->translate,
179                                  i,
180                                  ((char *)draw->pt.user.vbuffer[i] +
181                                   draw->pt.vertex_buffer[i].buffer_offset),
182                                  draw->pt.vertex_buffer[i].stride,
183                                  draw->pt.max_index);
184   }
185
186   *max_vertices = (draw->render->max_vertex_buffer_bytes /
187                    (vinfo->size * 4));
188}
189
190
191
192
193
194static void fetch_emit_run( struct draw_pt_middle_end *middle,
195                            const unsigned *fetch_elts,
196                            unsigned fetch_count,
197                            const ushort *draw_elts,
198                            unsigned draw_count,
199                            unsigned prim_flags )
200{
201   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
202   struct draw_context *draw = feme->draw;
203   void *hw_verts;
204
205   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
206    */
207   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
208
209   draw->render->allocate_vertices( draw->render,
210                                    (ushort)feme->translate->key.output_stride,
211                                    (ushort)fetch_count );
212
213   hw_verts = draw->render->map_vertices( draw->render );
214   if (!hw_verts) {
215      assert(0);
216      return;
217   }
218
219
220   /* Single routine to fetch vertices and emit HW verts.
221    */
222   feme->translate->run_elts( feme->translate,
223			      fetch_elts,
224			      fetch_count,
225                              draw->instance_id,
226			      hw_verts );
227
228   if (0) {
229      unsigned i;
230      for (i = 0; i < fetch_count; i++) {
231         debug_printf("\n\nvertex %d:\n", i);
232         draw_dump_emitted_vertex( feme->vinfo,
233                                   (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
234      }
235   }
236
237   draw->render->unmap_vertices( draw->render,
238                                 0,
239                                 (ushort)(fetch_count - 1) );
240
241   /* XXX: Draw arrays path to avoid re-emitting index list again and
242    * again.
243    */
244   draw->render->draw_elements( draw->render,
245                                draw_elts,
246                                draw_count );
247
248   /* Done -- that was easy, wasn't it:
249    */
250   draw->render->release_vertices( draw->render );
251
252}
253
254
255static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
256                                   unsigned start,
257                                   unsigned count,
258                                   unsigned prim_flags )
259{
260   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
261   struct draw_context *draw = feme->draw;
262   void *hw_verts;
263
264   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
265    */
266   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
267
268   if (!draw->render->allocate_vertices( draw->render,
269                                         (ushort)feme->translate->key.output_stride,
270                                         (ushort)count ))
271      goto fail;
272
273   hw_verts = draw->render->map_vertices( draw->render );
274   if (!hw_verts)
275      goto fail;
276
277   /* Single routine to fetch vertices and emit HW verts.
278    */
279   feme->translate->run( feme->translate,
280                         start,
281                         count,
282                         draw->instance_id,
283                         hw_verts );
284
285   if (0) {
286      unsigned i;
287      for (i = 0; i < count; i++) {
288         debug_printf("\n\nvertex %d:\n", i);
289         draw_dump_emitted_vertex( feme->vinfo,
290                                   (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
291      }
292   }
293
294   draw->render->unmap_vertices( draw->render, 0, count - 1 );
295
296   /* XXX: Draw arrays path to avoid re-emitting index list again and
297    * again.
298    */
299   draw->render->draw_arrays( draw->render, 0, count );
300
301   /* Done -- that was easy, wasn't it:
302    */
303   draw->render->release_vertices( draw->render );
304   return;
305
306fail:
307   assert(0);
308   return;
309}
310
311
312static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
313                                        unsigned start,
314                                        unsigned count,
315                                        const ushort *draw_elts,
316                                        unsigned draw_count,
317                                        unsigned prim_flags )
318{
319   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
320   struct draw_context *draw = feme->draw;
321   void *hw_verts;
322
323   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
324    */
325   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
326
327   if (!draw->render->allocate_vertices( draw->render,
328                                         (ushort)feme->translate->key.output_stride,
329                                         (ushort)count ))
330      return FALSE;
331
332   hw_verts = draw->render->map_vertices( draw->render );
333   if (!hw_verts)
334      return FALSE;
335
336   /* Single routine to fetch vertices and emit HW verts.
337    */
338   feme->translate->run( feme->translate,
339                         start,
340                         count,
341                         draw->instance_id,
342                         hw_verts );
343
344   draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
345
346   /* XXX: Draw arrays path to avoid re-emitting index list again and
347    * again.
348    */
349   draw->render->draw_elements( draw->render,
350                                draw_elts,
351                                draw_count );
352
353   /* Done -- that was easy, wasn't it:
354    */
355   draw->render->release_vertices( draw->render );
356
357   return TRUE;
358}
359
360
361
362
363static void fetch_emit_finish( struct draw_pt_middle_end *middle )
364{
365   /* nothing to do */
366}
367
368static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
369{
370   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
371
372   if (feme->cache)
373      translate_cache_destroy(feme->cache);
374
375   FREE(middle);
376}
377
378
379struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw )
380{
381   struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end );
382   if (fetch_emit == NULL)
383      return NULL;
384
385   fetch_emit->cache = translate_cache_create();
386   if (!fetch_emit->cache) {
387      FREE(fetch_emit);
388      return NULL;
389   }
390
391   fetch_emit->base.prepare    = fetch_emit_prepare;
392   fetch_emit->base.run        = fetch_emit_run;
393   fetch_emit->base.run_linear = fetch_emit_run_linear;
394   fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts;
395   fetch_emit->base.finish     = fetch_emit_finish;
396   fetch_emit->base.destroy    = fetch_emit_destroy;
397
398   fetch_emit->draw = draw;
399
400   return &fetch_emit->base;
401}
402
403