draw_pt_fetch_shade_emit.c revision 994c33db875c6af2f5b535a7a173ff3bfa24fc32 
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   /* Temporaries:
54    */
55   const float *constants;
56   unsigned pitch[PIPE_MAX_ATTRIBS];
57   const ubyte *src[PIPE_MAX_ATTRIBS];
58   unsigned prim;
59
60   struct draw_vs_variant_key key;
61   struct draw_vs_variant *active;
62
63   const struct vertex_info *vinfo;
64};
65
66
67
68static void
69fse_prepare(struct draw_pt_middle_end *middle,
70            unsigned prim,
71            unsigned opt,
72            unsigned *max_vertices)
73{
74   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
75   struct draw_context *draw = fse->draw;
76   unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs;
77   const struct vertex_info *vinfo;
78   unsigned i;
79   unsigned nr_vbs = 0;
80
81   /* Can't support geometry shader on this path.
82    */
83   assert(!draw->gs.geometry_shader);
84
85   draw->render->set_primitive(draw->render, prim);
86
87   /* Must do this after set_primitive() above:
88    */
89   fse->vinfo = vinfo = draw->render->get_vertex_info(draw->render);
90
91   fse->key.output_stride = vinfo->size * 4;
92   fse->key.nr_outputs = vinfo->num_attribs;
93   fse->key.nr_inputs = num_vs_inputs;
94
95   fse->key.nr_elements = MAX2(fse->key.nr_outputs,     /* outputs - translate to hw format */
96                               fse->key.nr_inputs);     /* inputs - fetch from api format */
97
98   fse->key.viewport = !draw->identity_viewport;
99   fse->key.clip = draw->clip_xy || draw->clip_z || draw->clip_user;
100   fse->key.const_vbuffers = 0;
101
102   memset(fse->key.element, 0,
103          fse->key.nr_elements * sizeof(fse->key.element[0]));
104
105   for (i = 0; i < num_vs_inputs; i++) {
106      const struct pipe_vertex_element *src = &draw->pt.vertex_element[i];
107      fse->key.element[i].in.format = src->src_format;
108
109      /* Consider ignoring these, ie make generated programs
110       * independent of this state:
111       */
112      fse->key.element[i].in.buffer = src->vertex_buffer_index;
113      fse->key.element[i].in.offset = src->src_offset;
114      nr_vbs = MAX2(nr_vbs, src->vertex_buffer_index + 1);
115   }
116
117   for (i = 0; i < 5 && i < nr_vbs; i++) {
118      if (draw->pt.vertex_buffer[i].stride == 0)
119         fse->key.const_vbuffers |= (1<<i);
120   }
121
122   if (0) debug_printf("%s: lookup const_vbuffers: %x\n", __FUNCTION__, fse->key.const_vbuffers);
123
124   {
125      unsigned dst_offset = 0;
126
127      for (i = 0; i < vinfo->num_attribs; i++) {
128         unsigned emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
129
130         /* doesn't handle EMIT_OMIT */
131         assert(emit_sz != 0);
132
133         /* The elements in the key correspond to vertex shader output
134          * numbers, not to positions in the hw vertex description --
135          * that's handled by the output_offset field.
136          */
137         fse->key.element[i].out.format = vinfo->attrib[i].emit;
138         fse->key.element[i].out.vs_output = vinfo->attrib[i].src_index;
139         fse->key.element[i].out.offset = dst_offset;
140
141         dst_offset += emit_sz;
142         assert(fse->key.output_stride >= dst_offset);
143      }
144   }
145
146   fse->active = draw_vs_lookup_variant( draw->vs.vertex_shader,
147                                         &fse->key );
148
149   if (!fse->active) {
150      assert(0);
151      return ;
152   }
153
154   if (0) debug_printf("%s: found const_vbuffers: %x\n", __FUNCTION__,
155                       fse->active->key.const_vbuffers);
156
157   /* Now set buffer pointers:
158    */
159   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
160      fse->active->set_buffer( fse->active,
161                               i,
162                               ((const ubyte *) draw->pt.user.vbuffer[i] +
163                                draw->pt.vertex_buffer[i].buffer_offset),
164                              draw->pt.vertex_buffer[i].stride,
165                              draw->pt.max_index );
166   }
167
168   *max_vertices = (draw->render->max_vertex_buffer_bytes /
169                    (vinfo->size * 4));
170
171   /* Probably need to do this somewhere (or fix exec shader not to
172    * need it):
173    */
174   if (1) {
175      struct draw_vertex_shader *vs = draw->vs.vertex_shader;
176      vs->prepare(vs, draw);
177   }
178}
179
180
181
182static void
183fse_run_linear(struct draw_pt_middle_end *middle,
184               unsigned start,
185               unsigned count,
186               unsigned prim_flags)
187{
188   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
189   struct draw_context *draw = fse->draw;
190   char *hw_verts;
191
192   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
193    */
194   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
195
196   if (!draw->render->allocate_vertices( draw->render,
197                                         (ushort)fse->key.output_stride,
198                                         (ushort)count ))
199      goto fail;
200
201   hw_verts = draw->render->map_vertices( draw->render );
202   if (!hw_verts)
203      goto fail;
204
205   /* Single routine to fetch vertices, run shader and emit HW verts.
206    * Clipping is done elsewhere -- either by the API or on hardware,
207    * or for some other reason not required...
208    */
209   fse->active->run_linear( fse->active,
210                            start, count,
211                            hw_verts );
212
213   if (0) {
214      unsigned i;
215      for (i = 0; i < count; i++) {
216         debug_printf("\n\n%s vertex %d: (stride %d, offset %d)\n", __FUNCTION__, i,
217                      fse->key.output_stride,
218                      fse->key.output_stride * i);
219
220         draw_dump_emitted_vertex( fse->vinfo,
221                                   (const uint8_t *)hw_verts + fse->key.output_stride * i );
222      }
223   }
224
225   draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
226
227   /* Draw arrays path to avoid re-emitting index list again and
228    * again.
229    */
230   draw->render->draw_arrays( draw->render,
231                              0,
232                              count );
233
234   draw->render->release_vertices( draw->render );
235
236   return;
237
238fail:
239   assert(0);
240   return;
241}
242
243
244static void
245fse_run(struct draw_pt_middle_end *middle,
246        const unsigned *fetch_elts,
247        unsigned fetch_count,
248        const ushort *draw_elts,
249        unsigned draw_count,
250        unsigned prim_flags )
251{
252   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
253   struct draw_context *draw = fse->draw;
254   void *hw_verts;
255
256   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
257    */
258   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
259
260   if (!draw->render->allocate_vertices( draw->render,
261                                         (ushort)fse->key.output_stride,
262                                         (ushort)fetch_count ))
263      goto fail;
264
265   hw_verts = draw->render->map_vertices( draw->render );
266   if (!hw_verts)
267      goto fail;
268
269   /* Single routine to fetch vertices, run shader and emit HW verts.
270    */
271   fse->active->run_elts( fse->active,
272                          fetch_elts,
273                          fetch_count,
274                          hw_verts );
275
276   if (0) {
277      unsigned i;
278      for (i = 0; i < fetch_count; i++) {
279         debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i);
280         draw_dump_emitted_vertex( fse->vinfo,
281                                   (const uint8_t *)hw_verts +
282                                   fse->key.output_stride * i );
283      }
284   }
285
286   draw->render->unmap_vertices( draw->render, 0, (ushort)(fetch_count - 1) );
287
288   draw->render->draw_elements( draw->render,
289                                draw_elts,
290                                draw_count );
291
292   draw->render->release_vertices( draw->render );
293   return;
294
295fail:
296   assert(0);
297   return;
298}
299
300
301
302static boolean
303fse_run_linear_elts(struct draw_pt_middle_end *middle,
304                    unsigned start,
305                    unsigned count,
306                    const ushort *draw_elts,
307                    unsigned draw_count,
308                    unsigned prim_flags)
309{
310   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
311   struct draw_context *draw = fse->draw;
312   char *hw_verts;
313
314   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
315    */
316   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
317
318   if (!draw->render->allocate_vertices( draw->render,
319                                         (ushort)fse->key.output_stride,
320                                         (ushort)count ))
321      return FALSE;
322
323   hw_verts = draw->render->map_vertices( draw->render );
324   if (!hw_verts)
325      return FALSE;
326
327   /* Single routine to fetch vertices, run shader and emit HW verts.
328    * Clipping is done elsewhere -- either by the API or on hardware,
329    * or for some other reason not required...
330    */
331   fse->active->run_linear( fse->active,
332                            start, count,
333                            hw_verts );
334
335   draw->render->draw_elements( draw->render,
336                                draw_elts,
337                                draw_count );
338
339   draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
340
341   draw->render->release_vertices( draw->render );
342
343   return TRUE;
344}
345
346
347
348static void
349fse_finish(struct draw_pt_middle_end *middle)
350{
351}
352
353
354static void
355fse_destroy(struct draw_pt_middle_end *middle)
356{
357   FREE(middle);
358}
359
360
361struct draw_pt_middle_end *
362draw_pt_middle_fse(struct draw_context *draw)
363{
364   struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit);
365   if (!fse)
366      return NULL;
367
368   fse->base.prepare = fse_prepare;
369   fse->base.run = fse_run;
370   fse->base.run_linear = fse_run_linear;
371   fse->base.run_linear_elts = fse_run_linear_elts;
372   fse->base.finish = fse_finish;
373   fse->base.destroy = fse_destroy;
374   fse->draw = draw;
375
376   return &fse->base;
377}
378