Merge remote-tracking branch 'origin/master' into pipe-video
[mesa.git] / src / mesa / drivers / dri / i965 / brw_state_upload.c
1 /*
2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
5
6 Permission is hereby granted, free of charge, to any person obtaining
7 a 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, sublicense, 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
16 portions of the Software.
17
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25
26 **********************************************************************/
27 /*
28 * Authors:
29 * Keith Whitwell <keith@tungstengraphics.com>
30 */
31
32
33
34 #include "brw_context.h"
35 #include "brw_state.h"
36 #include "intel_batchbuffer.h"
37 #include "intel_buffers.h"
38
39 /* This is used to initialize brw->state.atoms[]. We could use this
40 * list directly except for a single atom, brw_constant_buffer, which
41 * has a .dirty value which changes according to the parameters of the
42 * current fragment and vertex programs, and so cannot be a static
43 * value.
44 */
45 static const struct brw_tracked_state *gen4_atoms[] =
46 {
47 &brw_check_fallback,
48
49 &brw_wm_input_sizes,
50 &brw_vs_prog,
51 &brw_gs_prog,
52 &brw_clip_prog,
53 &brw_sf_prog,
54 &brw_wm_prog,
55
56 /* Once all the programs are done, we know how large urb entry
57 * sizes need to be and can decide if we need to change the urb
58 * layout.
59 */
60 &brw_curbe_offsets,
61 &brw_recalculate_urb_fence,
62
63 &brw_cc_vp,
64 &brw_cc_unit,
65
66 &brw_vs_constants, /* Before vs_surfaces and constant_buffer */
67 &brw_wm_constants, /* Before wm_surfaces and constant_buffer */
68
69 &brw_vs_surfaces, /* must do before unit */
70 &brw_wm_constant_surface, /* must do before wm surfaces/bind bo */
71 &brw_wm_surfaces, /* must do before samplers and unit */
72 &brw_wm_binding_table,
73 &brw_wm_samplers,
74
75 &brw_wm_unit,
76 &brw_sf_vp,
77 &brw_sf_unit,
78 &brw_vs_unit, /* always required, enabled or not */
79 &brw_clip_unit,
80 &brw_gs_unit,
81
82 /* Command packets:
83 */
84 &brw_invarient_state,
85 &brw_state_base_address,
86
87 &brw_binding_table_pointers,
88 &brw_blend_constant_color,
89
90 &brw_depthbuffer,
91
92 &brw_polygon_stipple,
93 &brw_polygon_stipple_offset,
94
95 &brw_line_stipple,
96 &brw_aa_line_parameters,
97
98 &brw_psp_urb_cbs,
99
100 &brw_drawing_rect,
101 &brw_indices,
102 &brw_index_buffer,
103 &brw_vertices,
104
105 &brw_constant_buffer
106 };
107
108 static const struct brw_tracked_state *gen6_atoms[] =
109 {
110 &brw_check_fallback,
111
112 &brw_wm_input_sizes,
113 &brw_vs_prog,
114 &brw_gs_prog,
115 &brw_wm_prog,
116
117 &gen6_clip_vp,
118 &gen6_sf_vp,
119
120 /* Command packets: */
121 &brw_invarient_state,
122
123 /* must do before binding table pointers, cc state ptrs */
124 &brw_state_base_address,
125
126 &brw_cc_vp,
127 &gen6_viewport_state, /* must do after *_vp stages */
128
129 &gen6_urb,
130 &gen6_blend_state, /* must do before cc unit */
131 &gen6_color_calc_state, /* must do before cc unit */
132 &gen6_depth_stencil_state, /* must do before cc unit */
133 &gen6_cc_state_pointers,
134
135 &brw_vs_constants, /* Before vs_surfaces and constant_buffer */
136 &brw_wm_constants, /* Before wm_surfaces and constant_buffer */
137 &gen6_vs_constants, /* Before vs_state */
138 &gen6_wm_constants, /* Before wm_state */
139
140 &brw_vs_surfaces, /* must do before unit */
141 &brw_wm_constant_surface, /* must do before wm surfaces/bind bo */
142 &brw_wm_surfaces, /* must do before samplers and unit */
143 &brw_wm_binding_table,
144
145 &brw_wm_samplers,
146 &gen6_sampler_state,
147
148 &gen6_vs_state,
149 &gen6_gs_state,
150 &gen6_clip_state,
151 &gen6_sf_state,
152 &gen6_wm_state,
153
154 &gen6_scissor_state,
155
156 &gen6_binding_table_pointers,
157
158 &brw_depthbuffer,
159
160 &brw_polygon_stipple,
161 &brw_polygon_stipple_offset,
162
163 &brw_line_stipple,
164 &brw_aa_line_parameters,
165
166 &brw_drawing_rect,
167
168 &brw_indices,
169 &brw_index_buffer,
170 &brw_vertices,
171 };
172
173 void brw_init_state( struct brw_context *brw )
174 {
175 const struct brw_tracked_state **atoms;
176 int num_atoms;
177
178 brw_init_caches(brw);
179
180 if (brw->intel.gen >= 6) {
181 atoms = gen6_atoms;
182 num_atoms = ARRAY_SIZE(gen6_atoms);
183 } else {
184 atoms = gen4_atoms;
185 num_atoms = ARRAY_SIZE(gen4_atoms);
186 }
187
188 while (num_atoms--) {
189 assert((*atoms)->dirty.mesa |
190 (*atoms)->dirty.brw |
191 (*atoms)->dirty.cache);
192
193 if ((*atoms)->prepare)
194 brw->prepare_atoms[brw->num_prepare_atoms++] = **atoms;
195 if ((*atoms)->emit)
196 brw->emit_atoms[brw->num_emit_atoms++] = **atoms;
197 atoms++;
198 }
199 assert(brw->num_emit_atoms <= ARRAY_SIZE(brw->emit_atoms));
200 assert(brw->num_prepare_atoms <= ARRAY_SIZE(brw->prepare_atoms));
201 }
202
203
204 void brw_destroy_state( struct brw_context *brw )
205 {
206 brw_destroy_caches(brw);
207 }
208
209 /***********************************************************************
210 */
211
212 static GLuint check_state( const struct brw_state_flags *a,
213 const struct brw_state_flags *b )
214 {
215 return ((a->mesa & b->mesa) |
216 (a->brw & b->brw) |
217 (a->cache & b->cache)) != 0;
218 }
219
220 static void accumulate_state( struct brw_state_flags *a,
221 const struct brw_state_flags *b )
222 {
223 a->mesa |= b->mesa;
224 a->brw |= b->brw;
225 a->cache |= b->cache;
226 }
227
228
229 static void xor_states( struct brw_state_flags *result,
230 const struct brw_state_flags *a,
231 const struct brw_state_flags *b )
232 {
233 result->mesa = a->mesa ^ b->mesa;
234 result->brw = a->brw ^ b->brw;
235 result->cache = a->cache ^ b->cache;
236 }
237
238 void
239 brw_clear_validated_bos(struct brw_context *brw)
240 {
241 int i;
242
243 /* Clear the last round of validated bos */
244 for (i = 0; i < brw->state.validated_bo_count; i++) {
245 drm_intel_bo_unreference(brw->state.validated_bos[i]);
246 brw->state.validated_bos[i] = NULL;
247 }
248 brw->state.validated_bo_count = 0;
249 }
250
251 struct dirty_bit_map {
252 uint32_t bit;
253 char *name;
254 uint32_t count;
255 };
256
257 #define DEFINE_BIT(name) {name, #name, 0}
258
259 static struct dirty_bit_map mesa_bits[] = {
260 DEFINE_BIT(_NEW_MODELVIEW),
261 DEFINE_BIT(_NEW_PROJECTION),
262 DEFINE_BIT(_NEW_TEXTURE_MATRIX),
263 DEFINE_BIT(_NEW_COLOR),
264 DEFINE_BIT(_NEW_DEPTH),
265 DEFINE_BIT(_NEW_EVAL),
266 DEFINE_BIT(_NEW_FOG),
267 DEFINE_BIT(_NEW_HINT),
268 DEFINE_BIT(_NEW_LIGHT),
269 DEFINE_BIT(_NEW_LINE),
270 DEFINE_BIT(_NEW_PIXEL),
271 DEFINE_BIT(_NEW_POINT),
272 DEFINE_BIT(_NEW_POLYGON),
273 DEFINE_BIT(_NEW_POLYGONSTIPPLE),
274 DEFINE_BIT(_NEW_SCISSOR),
275 DEFINE_BIT(_NEW_STENCIL),
276 DEFINE_BIT(_NEW_TEXTURE),
277 DEFINE_BIT(_NEW_TRANSFORM),
278 DEFINE_BIT(_NEW_VIEWPORT),
279 DEFINE_BIT(_NEW_PACKUNPACK),
280 DEFINE_BIT(_NEW_ARRAY),
281 DEFINE_BIT(_NEW_RENDERMODE),
282 DEFINE_BIT(_NEW_BUFFERS),
283 DEFINE_BIT(_NEW_MULTISAMPLE),
284 DEFINE_BIT(_NEW_TRACK_MATRIX),
285 DEFINE_BIT(_NEW_PROGRAM),
286 DEFINE_BIT(_NEW_PROGRAM_CONSTANTS),
287 {0, 0, 0}
288 };
289
290 static struct dirty_bit_map brw_bits[] = {
291 DEFINE_BIT(BRW_NEW_URB_FENCE),
292 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM),
293 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM),
294 DEFINE_BIT(BRW_NEW_INPUT_DIMENSIONS),
295 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS),
296 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE),
297 DEFINE_BIT(BRW_NEW_PRIMITIVE),
298 DEFINE_BIT(BRW_NEW_CONTEXT),
299 DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS),
300 DEFINE_BIT(BRW_NEW_PSP),
301 DEFINE_BIT(BRW_NEW_WM_SURFACES),
302 DEFINE_BIT(BRW_NEW_BINDING_TABLE),
303 DEFINE_BIT(BRW_NEW_INDICES),
304 DEFINE_BIT(BRW_NEW_INDEX_BUFFER),
305 DEFINE_BIT(BRW_NEW_VERTICES),
306 DEFINE_BIT(BRW_NEW_BATCH),
307 DEFINE_BIT(BRW_NEW_DEPTH_BUFFER),
308 DEFINE_BIT(BRW_NEW_NR_WM_SURFACES),
309 DEFINE_BIT(BRW_NEW_NR_VS_SURFACES),
310 DEFINE_BIT(BRW_NEW_VS_CONSTBUF),
311 DEFINE_BIT(BRW_NEW_WM_CONSTBUF),
312 {0, 0, 0}
313 };
314
315 static struct dirty_bit_map cache_bits[] = {
316 DEFINE_BIT(CACHE_NEW_BLEND_STATE),
317 DEFINE_BIT(CACHE_NEW_CC_VP),
318 DEFINE_BIT(CACHE_NEW_CC_UNIT),
319 DEFINE_BIT(CACHE_NEW_WM_PROG),
320 DEFINE_BIT(CACHE_NEW_SAMPLER),
321 DEFINE_BIT(CACHE_NEW_WM_UNIT),
322 DEFINE_BIT(CACHE_NEW_SF_PROG),
323 DEFINE_BIT(CACHE_NEW_SF_VP),
324 DEFINE_BIT(CACHE_NEW_SF_UNIT),
325 DEFINE_BIT(CACHE_NEW_VS_UNIT),
326 DEFINE_BIT(CACHE_NEW_VS_PROG),
327 DEFINE_BIT(CACHE_NEW_GS_UNIT),
328 DEFINE_BIT(CACHE_NEW_GS_PROG),
329 DEFINE_BIT(CACHE_NEW_CLIP_VP),
330 DEFINE_BIT(CACHE_NEW_CLIP_UNIT),
331 DEFINE_BIT(CACHE_NEW_CLIP_PROG),
332 {0, 0, 0}
333 };
334
335
336 static void
337 brw_update_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
338 {
339 int i;
340
341 for (i = 0; i < 32; i++) {
342 if (bit_map[i].bit == 0)
343 return;
344
345 if (bit_map[i].bit & bits)
346 bit_map[i].count++;
347 }
348 }
349
350 static void
351 brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
352 {
353 int i;
354
355 for (i = 0; i < 32; i++) {
356 if (bit_map[i].bit == 0)
357 return;
358
359 fprintf(stderr, "0x%08x: %12d (%s)\n",
360 bit_map[i].bit, bit_map[i].count, bit_map[i].name);
361 }
362 }
363
364 /***********************************************************************
365 * Emit all state:
366 */
367 void brw_validate_state( struct brw_context *brw )
368 {
369 struct gl_context *ctx = &brw->intel.ctx;
370 struct intel_context *intel = &brw->intel;
371 struct brw_state_flags *state = &brw->state.dirty;
372 const struct brw_tracked_state *atoms = brw->prepare_atoms;
373 int num_atoms = brw->num_prepare_atoms;
374 GLuint i;
375
376 brw_clear_validated_bos(brw);
377
378 state->mesa |= brw->intel.NewGLState;
379 brw->intel.NewGLState = 0;
380
381 brw_add_validated_bo(brw, intel->batch.bo);
382
383 if (brw->emit_state_always) {
384 state->mesa |= ~0;
385 state->brw |= ~0;
386 state->cache |= ~0;
387 }
388
389 if (brw->fragment_program != ctx->FragmentProgram._Current) {
390 brw->fragment_program = ctx->FragmentProgram._Current;
391 brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
392 }
393
394 if (brw->vertex_program != ctx->VertexProgram._Current) {
395 brw->vertex_program = ctx->VertexProgram._Current;
396 brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
397 }
398
399 if ((state->mesa | state->cache | state->brw) == 0)
400 return;
401
402 brw->intel.Fallback = GL_FALSE; /* boolean, not bitfield */
403
404 /* do prepare stage for all atoms */
405 for (i = 0; i < num_atoms; i++) {
406 const struct brw_tracked_state *atom = &atoms[i];
407
408 if (check_state(state, &atom->dirty)) {
409 atom->prepare(brw);
410
411 if (brw->intel.Fallback)
412 break;
413 }
414 }
415
416 intel_check_front_buffer_rendering(intel);
417
418 /* Make sure that the textures which are referenced by the current
419 * brw fragment program are actually present/valid.
420 * If this fails, we can experience GPU lock-ups.
421 */
422 {
423 const struct brw_fragment_program *fp;
424 fp = brw_fragment_program_const(brw->fragment_program);
425 if (fp) {
426 assert((fp->tex_units_used & ctx->Texture._EnabledUnits)
427 == fp->tex_units_used);
428 }
429 }
430 }
431
432
433 void brw_upload_state(struct brw_context *brw)
434 {
435 struct brw_state_flags *state = &brw->state.dirty;
436 const struct brw_tracked_state *atoms = brw->emit_atoms;
437 int num_atoms = brw->num_emit_atoms;
438 int i;
439 static int dirty_count = 0;
440
441 brw_clear_validated_bos(brw);
442
443 if (unlikely(INTEL_DEBUG)) {
444 /* Debug version which enforces various sanity checks on the
445 * state flags which are generated and checked to help ensure
446 * state atoms are ordered correctly in the list.
447 */
448 struct brw_state_flags examined, prev;
449 memset(&examined, 0, sizeof(examined));
450 prev = *state;
451
452 for (i = 0; i < num_atoms; i++) {
453 const struct brw_tracked_state *atom = &atoms[i];
454 struct brw_state_flags generated;
455
456 if (brw->intel.Fallback)
457 break;
458
459 if (check_state(state, &atom->dirty)) {
460 atom->emit(brw);
461 }
462
463 accumulate_state(&examined, &atom->dirty);
464
465 /* generated = (prev ^ state)
466 * if (examined & generated)
467 * fail;
468 */
469 xor_states(&generated, &prev, state);
470 assert(!check_state(&examined, &generated));
471 prev = *state;
472 }
473 }
474 else {
475 for (i = 0; i < num_atoms; i++) {
476 const struct brw_tracked_state *atom = &atoms[i];
477
478 if (brw->intel.Fallback)
479 break;
480
481 if (check_state(state, &atom->dirty)) {
482 atom->emit(brw);
483 }
484 }
485 }
486
487 if (unlikely(INTEL_DEBUG & DEBUG_STATE)) {
488 brw_update_dirty_count(mesa_bits, state->mesa);
489 brw_update_dirty_count(brw_bits, state->brw);
490 brw_update_dirty_count(cache_bits, state->cache);
491 if (dirty_count++ % 1000 == 0) {
492 brw_print_dirty_count(mesa_bits, state->mesa);
493 brw_print_dirty_count(brw_bits, state->brw);
494 brw_print_dirty_count(cache_bits, state->cache);
495 fprintf(stderr, "\n");
496 }
497 }
498
499 if (!brw->intel.Fallback)
500 memset(state, 0, sizeof(*state));
501 }