draw_pt_fetch_shade_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
34#include "util/u_math.h"
35#include "util/u_memory.h"
36#include "draw/draw_context.h"
37#include "draw/draw_private.h"
38#include "draw/draw_vbuf.h"
39#include "draw/draw_vertex.h"
40#include "draw/draw_pt.h"
41#include "draw/draw_vs.h"
42
43
44struct fetch_shade_emit;
45
46
47/* Prototype fetch, shade, emit-hw-verts all in one go.
48 */
49struct fetch_shade_emit {
50   struct draw_pt_middle_end base;
51   struct draw_context *draw;
52
53
54   /* Temporaries:
55    */
56   const float *constants;
57   unsigned pitch[PIPE_MAX_ATTRIBS];
58   const ubyte *src[PIPE_MAX_ATTRIBS];
59   unsigned prim;
60
61   struct draw_vs_variant_key key;
62   struct draw_vs_variant *active;
63
64
65   const struct vertex_info *vinfo;
66};
67
68
69
70static void fse_prepare( struct draw_pt_middle_end *middle,
71                         unsigned prim,
72                         unsigned opt,
73                         unsigned *max_vertices )
74{
75   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
76   struct draw_context *draw = fse->draw;
77   unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs;
78   const struct vertex_info *vinfo;
79   unsigned i;
80   unsigned nr_vbs = 0;
81
82   /* Can't support geometry shader on this path.
83    */
84   assert(!draw->gs.geometry_shader);
85
86   draw->render->set_primitive(draw->render, prim);
87
88   /* Must do this after set_primitive() above:
89    */
90   fse->vinfo = vinfo = draw->render->get_vertex_info(draw->render);
91
92
93   fse->key.output_stride = vinfo->size * 4;
94   fse->key.nr_outputs = vinfo->num_attribs;
95   fse->key.nr_inputs = num_vs_inputs;
96
97   fse->key.nr_elements = MAX2(fse->key.nr_outputs,     /* outputs - translate to hw format */
98                               fse->key.nr_inputs);     /* inputs - fetch from api format */
99
100   fse->key.viewport = !draw->identity_viewport;
101   fse->key.clip = draw->clip_xy || draw->clip_z || draw->clip_user;
102   fse->key.const_vbuffers = 0;
103
104   memset(fse->key.element, 0,
105          fse->key.nr_elements * sizeof(fse->key.element[0]));
106
107   for (i = 0; i < num_vs_inputs; i++) {
108      const struct pipe_vertex_element *src = &draw->pt.vertex_element[i];
109      fse->key.element[i].in.format = src->src_format;
110
111      /* Consider ignoring these, ie make generated programs
112       * independent of this state:
113       */
114      fse->key.element[i].in.buffer = src->vertex_buffer_index;
115      fse->key.element[i].in.offset = src->src_offset;
116      nr_vbs = MAX2(nr_vbs, src->vertex_buffer_index + 1);
117   }
118
119   for (i = 0; i < 5 && i < nr_vbs; i++) {
120      if (draw->pt.vertex_buffer[i].stride == 0)
121         fse->key.const_vbuffers |= (1<<i);
122   }
123
124   if (0) debug_printf("%s: lookup const_vbuffers: %x\n", __FUNCTION__, fse->key.const_vbuffers);
125
126   {
127      unsigned dst_offset = 0;
128
129      for (i = 0; i < vinfo->num_attribs; i++) {
130         unsigned emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
131
132         /* doesn't handle EMIT_OMIT */
133         assert(emit_sz != 0);
134
135         /* The elements in the key correspond to vertex shader output
136          * numbers, not to positions in the hw vertex description --
137          * that's handled by the output_offset field.
138          */
139         fse->key.element[i].out.format = vinfo->attrib[i].emit;
140         fse->key.element[i].out.vs_output = vinfo->attrib[i].src_index;
141         fse->key.element[i].out.offset = dst_offset;
142
143         dst_offset += emit_sz;
144         assert(fse->key.output_stride >= dst_offset);
145      }
146   }
147
148
149   fse->active = draw_vs_lookup_variant( draw->vs.vertex_shader,
150                                         &fse->key );
151
152   if (!fse->active) {
153      assert(0);
154      return ;
155   }
156
157   if (0) debug_printf("%s: found const_vbuffers: %x\n", __FUNCTION__,
158                       fse->active->key.const_vbuffers);
159
160   /* Now set buffer pointers:
161    */
162   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
163      fse->active->set_buffer( fse->active,
164                               i,
165                               ((const ubyte *) draw->pt.user.vbuffer[i] +
166                                draw->pt.vertex_buffer[i].buffer_offset),
167                              draw->pt.vertex_buffer[i].stride,
168                              draw->pt.max_index );
169   }
170
171   *max_vertices = (draw->render->max_vertex_buffer_bytes /
172                    (vinfo->size * 4));
173
174   /* Probably need to do this somewhere (or fix exec shader not to
175    * need it):
176    */
177   if (1) {
178      struct draw_vertex_shader *vs = draw->vs.vertex_shader;
179      vs->prepare(vs, draw);
180   }
181}
182
183
184
185static void fse_run_linear( struct draw_pt_middle_end *middle,
186                            unsigned start,
187                            unsigned count,
188                            unsigned prim_flags )
189{
190   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
191   struct draw_context *draw = fse->draw;
192   char *hw_verts;
193
194   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
195    */
196   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
197
198   if (!draw->render->allocate_vertices( draw->render,
199                                         (ushort)fse->key.output_stride,
200                                         (ushort)count ))
201      goto fail;
202
203   hw_verts = draw->render->map_vertices( draw->render );
204   if (!hw_verts)
205      goto fail;
206
207   /* Single routine to fetch vertices, run shader and emit HW verts.
208    * Clipping is done elsewhere -- either by the API or on hardware,
209    * or for some other reason not required...
210    */
211   fse->active->run_linear( fse->active,
212                            start, count,
213                            hw_verts );
214
215
216   if (0) {
217      unsigned i;
218      for (i = 0; i < count; i++) {
219         debug_printf("\n\n%s vertex %d: (stride %d, offset %d)\n", __FUNCTION__, i,
220                      fse->key.output_stride,
221                      fse->key.output_stride * i);
222
223         draw_dump_emitted_vertex( fse->vinfo,
224                                   (const uint8_t *)hw_verts + fse->key.output_stride * i );
225      }
226   }
227
228   draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
229
230   /* Draw arrays path to avoid re-emitting index list again and
231    * again.
232    */
233   draw->render->draw_arrays( draw->render,
234                              0,
235                              count );
236
237
238   draw->render->release_vertices( draw->render );
239
240   return;
241
242fail:
243   assert(0);
244   return;
245}
246
247
248static void
249fse_run(struct draw_pt_middle_end *middle,
250        const unsigned *fetch_elts,
251        unsigned fetch_count,
252        const ushort *draw_elts,
253        unsigned draw_count,
254        unsigned prim_flags )
255{
256   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
257   struct draw_context *draw = fse->draw;
258   void *hw_verts;
259
260   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
261    */
262   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
263
264   if (!draw->render->allocate_vertices( draw->render,
265                                         (ushort)fse->key.output_stride,
266                                         (ushort)fetch_count ))
267      goto fail;
268
269   hw_verts = draw->render->map_vertices( draw->render );
270   if (!hw_verts)
271      goto fail;
272
273
274   /* Single routine to fetch vertices, run shader and emit HW verts.
275    */
276   fse->active->run_elts( fse->active,
277                          fetch_elts,
278                          fetch_count,
279                          hw_verts );
280
281
282   if (0) {
283      unsigned i;
284      for (i = 0; i < fetch_count; i++) {
285         debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i);
286         draw_dump_emitted_vertex( fse->vinfo,
287                                   (const uint8_t *)hw_verts +
288                                   fse->key.output_stride * i );
289      }
290   }
291
292   draw->render->unmap_vertices( draw->render, 0, (ushort)(fetch_count - 1) );
293
294   draw->render->draw_elements( draw->render,
295                                draw_elts,
296                                draw_count );
297
298
299   draw->render->release_vertices( draw->render );
300   return;
301
302fail:
303   assert(0);
304   return;
305}
306
307
308
309static boolean fse_run_linear_elts( struct draw_pt_middle_end *middle,
310                                 unsigned start,
311                                 unsigned count,
312                                 const ushort *draw_elts,
313                                 unsigned draw_count,
314                                 unsigned prim_flags )
315{
316   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
317   struct draw_context *draw = fse->draw;
318   char *hw_verts;
319
320   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
321    */
322   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
323
324   if (!draw->render->allocate_vertices( draw->render,
325                                         (ushort)fse->key.output_stride,
326                                         (ushort)count ))
327      return FALSE;
328
329   hw_verts = draw->render->map_vertices( draw->render );
330   if (!hw_verts)
331      return FALSE;
332
333   /* Single routine to fetch vertices, run shader and emit HW verts.
334    * Clipping is done elsewhere -- either by the API or on hardware,
335    * or for some other reason not required...
336    */
337   fse->active->run_linear( fse->active,
338                            start, count,
339                            hw_verts );
340
341
342   draw->render->draw_elements( draw->render,
343                                draw_elts,
344                                draw_count );
345
346
347   draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
348
349   draw->render->release_vertices( draw->render );
350
351   return TRUE;
352}
353
354
355
356static void fse_finish( struct draw_pt_middle_end *middle )
357{
358}
359
360
361static void
362fse_destroy( struct draw_pt_middle_end *middle )
363{
364   FREE(middle);
365}
366
367struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw )
368{
369   struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit);
370   if (!fse)
371      return NULL;
372
373   fse->base.prepare = fse_prepare;
374   fse->base.run = fse_run;
375   fse->base.run_linear = fse_run_linear;
376   fse->base.run_linear_elts = fse_run_linear_elts;
377   fse->base.finish = fse_finish;
378   fse->base.destroy = fse_destroy;
379   fse->draw = draw;
380
381   return &fse->base;
382}
383