Merge branch '7.8'
[mesa.git] / src / mesa / drivers / dri / i965 / brw_misc_state.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 "intel_batchbuffer.h"
35 #include "intel_regions.h"
36
37 #include "brw_context.h"
38 #include "brw_state.h"
39 #include "brw_defines.h"
40
41
42
43
44
45 /***********************************************************************
46 * Blend color
47 */
48
49 static void upload_blend_constant_color(struct brw_context *brw)
50 {
51 GLcontext *ctx = &brw->intel.ctx;
52 struct brw_blend_constant_color bcc;
53
54 memset(&bcc, 0, sizeof(bcc));
55 bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR;
56 bcc.header.length = sizeof(bcc)/4-2;
57 bcc.blend_constant_color[0] = ctx->Color.BlendColor[0];
58 bcc.blend_constant_color[1] = ctx->Color.BlendColor[1];
59 bcc.blend_constant_color[2] = ctx->Color.BlendColor[2];
60 bcc.blend_constant_color[3] = ctx->Color.BlendColor[3];
61
62 BRW_CACHED_BATCH_STRUCT(brw, &bcc);
63 }
64
65
66 const struct brw_tracked_state brw_blend_constant_color = {
67 .dirty = {
68 .mesa = _NEW_COLOR,
69 .brw = BRW_NEW_CONTEXT,
70 .cache = 0
71 },
72 .emit = upload_blend_constant_color
73 };
74
75 /* Constant single cliprect for framebuffer object or DRI2 drawing */
76 static void upload_drawing_rect(struct brw_context *brw)
77 {
78 struct intel_context *intel = &brw->intel;
79 GLcontext *ctx = &intel->ctx;
80
81 BEGIN_BATCH(4);
82 OUT_BATCH(_3DSTATE_DRAWRECT_INFO_I965);
83 OUT_BATCH(0); /* xmin, ymin */
84 OUT_BATCH(((ctx->DrawBuffer->Width - 1) & 0xffff) |
85 ((ctx->DrawBuffer->Height - 1) << 16));
86 OUT_BATCH(0);
87 ADVANCE_BATCH();
88 }
89
90 const struct brw_tracked_state brw_drawing_rect = {
91 .dirty = {
92 .mesa = _NEW_BUFFERS,
93 .brw = BRW_NEW_CONTEXT,
94 .cache = 0
95 },
96 .emit = upload_drawing_rect
97 };
98
99 static void prepare_binding_table_pointers(struct brw_context *brw)
100 {
101 brw_add_validated_bo(brw, brw->vs.bind_bo);
102 brw_add_validated_bo(brw, brw->wm.bind_bo);
103 }
104
105 /**
106 * Upload the binding table pointers, which point each stage's array of surface
107 * state pointers.
108 *
109 * The binding table pointers are relative to the surface state base address,
110 * which is 0.
111 */
112 static void upload_binding_table_pointers(struct brw_context *brw)
113 {
114 struct intel_context *intel = &brw->intel;
115
116 BEGIN_BATCH(6);
117 OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2));
118 if (brw->vs.bind_bo != NULL)
119 OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
120 else
121 OUT_BATCH(0);
122 OUT_BATCH(0); /* gs */
123 OUT_BATCH(0); /* clip */
124 OUT_BATCH(0); /* sf */
125 OUT_RELOC(brw->wm.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* wm/ps */
126 ADVANCE_BATCH();
127 }
128
129 const struct brw_tracked_state brw_binding_table_pointers = {
130 .dirty = {
131 .mesa = 0,
132 .brw = BRW_NEW_BATCH,
133 .cache = CACHE_NEW_SURF_BIND,
134 },
135 .prepare = prepare_binding_table_pointers,
136 .emit = upload_binding_table_pointers,
137 };
138
139 /**
140 * Upload the binding table pointers, which point each stage's array of surface
141 * state pointers.
142 *
143 * The binding table pointers are relative to the surface state base address,
144 * which is 0.
145 */
146 static void upload_gen6_binding_table_pointers(struct brw_context *brw)
147 {
148 struct intel_context *intel = &brw->intel;
149
150 BEGIN_BATCH(4);
151 OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 |
152 GEN6_BINDING_TABLE_MODIFY_VS |
153 GEN6_BINDING_TABLE_MODIFY_GS |
154 GEN6_BINDING_TABLE_MODIFY_PS |
155 (4 - 2));
156 if (brw->vs.bind_bo != NULL)
157 OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
158 else
159 OUT_BATCH(0);
160 OUT_BATCH(0); /* gs */
161 OUT_RELOC(brw->wm.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* wm/ps */
162 ADVANCE_BATCH();
163 }
164
165 const struct brw_tracked_state gen6_binding_table_pointers = {
166 .dirty = {
167 .mesa = 0,
168 .brw = BRW_NEW_BATCH,
169 .cache = CACHE_NEW_SURF_BIND,
170 },
171 .prepare = prepare_binding_table_pointers,
172 .emit = upload_gen6_binding_table_pointers,
173 };
174
175 /**
176 * Upload pointers to the per-stage state.
177 *
178 * The state pointers in this packet are all relative to the general state
179 * base address set by CMD_STATE_BASE_ADDRESS, which is 0.
180 */
181 static void upload_pipelined_state_pointers(struct brw_context *brw )
182 {
183 struct intel_context *intel = &brw->intel;
184
185 BEGIN_BATCH(7);
186 OUT_BATCH(CMD_PIPELINED_STATE_POINTERS << 16 | (7 - 2));
187 OUT_RELOC(brw->vs.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
188 if (brw->gs.prog_active)
189 OUT_RELOC(brw->gs.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1);
190 else
191 OUT_BATCH(0);
192 OUT_RELOC(brw->clip.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1);
193 OUT_RELOC(brw->sf.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
194 OUT_RELOC(brw->wm.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
195 OUT_RELOC(brw->cc.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
196 ADVANCE_BATCH();
197
198 brw->state.dirty.brw |= BRW_NEW_PSP;
199 }
200
201
202 static void prepare_psp_urb_cbs(struct brw_context *brw)
203 {
204 brw_add_validated_bo(brw, brw->vs.state_bo);
205 brw_add_validated_bo(brw, brw->gs.state_bo);
206 brw_add_validated_bo(brw, brw->clip.state_bo);
207 brw_add_validated_bo(brw, brw->sf.state_bo);
208 brw_add_validated_bo(brw, brw->wm.state_bo);
209 brw_add_validated_bo(brw, brw->cc.state_bo);
210 }
211
212 static void upload_psp_urb_cbs(struct brw_context *brw )
213 {
214 upload_pipelined_state_pointers(brw);
215 brw_upload_urb_fence(brw);
216 brw_upload_cs_urb_state(brw);
217 }
218
219 const struct brw_tracked_state brw_psp_urb_cbs = {
220 .dirty = {
221 .mesa = 0,
222 .brw = BRW_NEW_URB_FENCE | BRW_NEW_BATCH,
223 .cache = (CACHE_NEW_VS_UNIT |
224 CACHE_NEW_GS_UNIT |
225 CACHE_NEW_GS_PROG |
226 CACHE_NEW_CLIP_UNIT |
227 CACHE_NEW_SF_UNIT |
228 CACHE_NEW_WM_UNIT |
229 CACHE_NEW_CC_UNIT)
230 },
231 .prepare = prepare_psp_urb_cbs,
232 .emit = upload_psp_urb_cbs,
233 };
234
235 static void prepare_depthbuffer(struct brw_context *brw)
236 {
237 struct intel_region *region = brw->state.depth_region;
238
239 if (region != NULL)
240 brw_add_validated_bo(brw, region->buffer);
241 }
242
243 static void emit_depthbuffer(struct brw_context *brw)
244 {
245 struct intel_context *intel = &brw->intel;
246 struct intel_region *region = brw->state.depth_region;
247 unsigned int len;
248
249 if (intel->gen >= 6)
250 len = 7;
251 else if (intel->is_g4x || intel->gen == 5)
252 len = 6;
253 else
254 len = 5;
255
256 if (region == NULL) {
257 BEGIN_BATCH(len);
258 OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
259 OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
260 (BRW_SURFACE_NULL << 29));
261 OUT_BATCH(0);
262 OUT_BATCH(0);
263 OUT_BATCH(0);
264
265 if (intel->is_g4x || intel->gen >= 5)
266 OUT_BATCH(0);
267
268 if (intel->gen >= 6)
269 OUT_BATCH(0);
270
271 ADVANCE_BATCH();
272 } else {
273 unsigned int format;
274
275 switch (region->cpp) {
276 case 2:
277 format = BRW_DEPTHFORMAT_D16_UNORM;
278 break;
279 case 4:
280 if (intel->depth_buffer_is_float)
281 format = BRW_DEPTHFORMAT_D32_FLOAT;
282 else
283 format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT;
284 break;
285 default:
286 assert(0);
287 return;
288 }
289
290 assert(region->tiling != I915_TILING_X);
291 if (IS_GEN6(intel->intelScreen->deviceID))
292 assert(region->tiling != I915_TILING_NONE);
293
294 BEGIN_BATCH(len);
295 OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
296 OUT_BATCH(((region->pitch * region->cpp) - 1) |
297 (format << 18) |
298 (BRW_TILEWALK_YMAJOR << 26) |
299 ((region->tiling != I915_TILING_NONE) << 27) |
300 (BRW_SURFACE_2D << 29));
301 OUT_RELOC(region->buffer,
302 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
303 0);
304 OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) |
305 ((region->pitch - 1) << 6) |
306 ((region->height - 1) << 19));
307 OUT_BATCH(0);
308
309 if (intel->is_g4x || intel->gen >= 5)
310 OUT_BATCH(0);
311
312 if (intel->gen >= 6)
313 OUT_BATCH(0);
314
315 ADVANCE_BATCH();
316 }
317
318 /* Initialize it for safety. */
319 if (intel->gen >= 6) {
320 BEGIN_BATCH(2);
321 OUT_BATCH(CMD_3D_CLEAR_PARAMS << 16 | (2 - 2));
322 OUT_BATCH(0);
323 ADVANCE_BATCH();
324 }
325 }
326
327 const struct brw_tracked_state brw_depthbuffer = {
328 .dirty = {
329 .mesa = 0,
330 .brw = BRW_NEW_DEPTH_BUFFER | BRW_NEW_BATCH,
331 .cache = 0,
332 },
333 .prepare = prepare_depthbuffer,
334 .emit = emit_depthbuffer,
335 };
336
337
338
339 /***********************************************************************
340 * Polygon stipple packet
341 */
342
343 static void upload_polygon_stipple(struct brw_context *brw)
344 {
345 GLcontext *ctx = &brw->intel.ctx;
346 struct brw_polygon_stipple bps;
347 GLuint i;
348
349 memset(&bps, 0, sizeof(bps));
350 bps.header.opcode = CMD_POLY_STIPPLE_PATTERN;
351 bps.header.length = sizeof(bps)/4-2;
352
353 /* Polygon stipple is provided in OpenGL order, i.e. bottom
354 * row first. If we're rendering to a window (i.e. the
355 * default frame buffer object, 0), then we need to invert
356 * it to match our pixel layout. But if we're rendering
357 * to a FBO (i.e. any named frame buffer object), we *don't*
358 * need to invert - we already match the layout.
359 */
360 if (ctx->DrawBuffer->Name == 0) {
361 for (i = 0; i < 32; i++)
362 bps.stipple[i] = ctx->PolygonStipple[31 - i]; /* invert */
363 }
364 else {
365 for (i = 0; i < 32; i++)
366 bps.stipple[i] = ctx->PolygonStipple[i]; /* don't invert */
367 }
368
369 BRW_CACHED_BATCH_STRUCT(brw, &bps);
370 }
371
372 const struct brw_tracked_state brw_polygon_stipple = {
373 .dirty = {
374 .mesa = _NEW_POLYGONSTIPPLE,
375 .brw = BRW_NEW_CONTEXT,
376 .cache = 0
377 },
378 .emit = upload_polygon_stipple
379 };
380
381
382 /***********************************************************************
383 * Polygon stipple offset packet
384 */
385
386 static void upload_polygon_stipple_offset(struct brw_context *brw)
387 {
388 GLcontext *ctx = &brw->intel.ctx;
389 struct brw_polygon_stipple_offset bpso;
390
391 memset(&bpso, 0, sizeof(bpso));
392 bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET;
393 bpso.header.length = sizeof(bpso)/4-2;
394
395 /* If we're drawing to a system window (ctx->DrawBuffer->Name == 0),
396 * we have to invert the Y axis in order to match the OpenGL
397 * pixel coordinate system, and our offset must be matched
398 * to the window position. If we're drawing to a FBO
399 * (ctx->DrawBuffer->Name != 0), then our native pixel coordinate
400 * system works just fine, and there's no window system to
401 * worry about.
402 */
403 if (brw->intel.ctx.DrawBuffer->Name == 0) {
404 bpso.bits0.x_offset = 0;
405 bpso.bits0.y_offset = (32 - (ctx->DrawBuffer->Height & 31)) & 31;
406 }
407 else {
408 bpso.bits0.y_offset = 0;
409 bpso.bits0.x_offset = 0;
410 }
411
412 BRW_CACHED_BATCH_STRUCT(brw, &bpso);
413 }
414
415 #define _NEW_WINDOW_POS 0x40000000
416
417 const struct brw_tracked_state brw_polygon_stipple_offset = {
418 .dirty = {
419 .mesa = _NEW_WINDOW_POS,
420 .brw = BRW_NEW_CONTEXT,
421 .cache = 0
422 },
423 .emit = upload_polygon_stipple_offset
424 };
425
426 /**********************************************************************
427 * AA Line parameters
428 */
429 static void upload_aa_line_parameters(struct brw_context *brw)
430 {
431 struct brw_aa_line_parameters balp;
432
433 if (!brw->has_aa_line_parameters)
434 return;
435
436 /* use legacy aa line coverage computation */
437 memset(&balp, 0, sizeof(balp));
438 balp.header.opcode = CMD_AA_LINE_PARAMETERS;
439 balp.header.length = sizeof(balp) / 4 - 2;
440
441 BRW_CACHED_BATCH_STRUCT(brw, &balp);
442 }
443
444 const struct brw_tracked_state brw_aa_line_parameters = {
445 .dirty = {
446 .mesa = 0,
447 .brw = BRW_NEW_CONTEXT,
448 .cache = 0
449 },
450 .emit = upload_aa_line_parameters
451 };
452
453 /***********************************************************************
454 * Line stipple packet
455 */
456
457 static void upload_line_stipple(struct brw_context *brw)
458 {
459 GLcontext *ctx = &brw->intel.ctx;
460 struct brw_line_stipple bls;
461 GLfloat tmp;
462 GLint tmpi;
463
464 memset(&bls, 0, sizeof(bls));
465 bls.header.opcode = CMD_LINE_STIPPLE_PATTERN;
466 bls.header.length = sizeof(bls)/4 - 2;
467
468 bls.bits0.pattern = ctx->Line.StipplePattern;
469 bls.bits1.repeat_count = ctx->Line.StippleFactor;
470
471 tmp = 1.0 / (GLfloat) ctx->Line.StippleFactor;
472 tmpi = tmp * (1<<13);
473
474
475 bls.bits1.inverse_repeat_count = tmpi;
476
477 BRW_CACHED_BATCH_STRUCT(brw, &bls);
478 }
479
480 const struct brw_tracked_state brw_line_stipple = {
481 .dirty = {
482 .mesa = _NEW_LINE,
483 .brw = BRW_NEW_CONTEXT,
484 .cache = 0
485 },
486 .emit = upload_line_stipple
487 };
488
489
490 /***********************************************************************
491 * Misc invarient state packets
492 */
493
494 static void upload_invarient_state( struct brw_context *brw )
495 {
496 struct intel_context *intel = &brw->intel;
497
498 {
499 /* 0x61040000 Pipeline Select */
500 /* PipelineSelect : 0 */
501 struct brw_pipeline_select ps;
502
503 memset(&ps, 0, sizeof(ps));
504 ps.header.opcode = brw->CMD_PIPELINE_SELECT;
505 ps.header.pipeline_select = 0;
506 BRW_BATCH_STRUCT(brw, &ps);
507 }
508
509 if (intel->gen < 6) {
510 struct brw_global_depth_offset_clamp gdo;
511 memset(&gdo, 0, sizeof(gdo));
512
513 /* Disable depth offset clamping.
514 */
515 gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP;
516 gdo.header.length = sizeof(gdo)/4 - 2;
517 gdo.depth_offset_clamp = 0.0;
518
519 BRW_BATCH_STRUCT(brw, &gdo);
520 }
521
522 intel_batchbuffer_emit_mi_flush(intel->batch);
523
524 if (intel->gen >= 6) {
525 int i;
526
527 BEGIN_BATCH(3);
528 OUT_BATCH(CMD_3D_MULTISAMPLE << 16 | (3 - 2));
529 OUT_BATCH(MS_PIXEL_LOCATION_CENTER |
530 MS_NUMSAMPLES_1);
531 OUT_BATCH(0); /* positions for 4/8-sample */
532 ADVANCE_BATCH();
533
534 BEGIN_BATCH(2);
535 OUT_BATCH(CMD_3D_SAMPLE_MASK << 16 | (2 - 2));
536 OUT_BATCH(1);
537 ADVANCE_BATCH();
538
539 for (i = 0; i < 4; i++) {
540 BEGIN_BATCH(4);
541 OUT_BATCH(CMD_GS_SVB_INDEX << 16 | (4 - 2));
542 OUT_BATCH(i << SVB_INDEX_SHIFT);
543 OUT_BATCH(0);
544 OUT_BATCH(0xffffffff);
545 ADVANCE_BATCH();
546 }
547 }
548
549 /* 0x61020000 State Instruction Pointer */
550 {
551 struct brw_system_instruction_pointer sip;
552 memset(&sip, 0, sizeof(sip));
553
554 sip.header.opcode = CMD_STATE_INSN_POINTER;
555 sip.header.length = 0;
556 sip.bits0.pad = 0;
557 sip.bits0.system_instruction_pointer = 0;
558 BRW_BATCH_STRUCT(brw, &sip);
559 }
560
561
562 {
563 struct brw_vf_statistics vfs;
564 memset(&vfs, 0, sizeof(vfs));
565
566 vfs.opcode = brw->CMD_VF_STATISTICS;
567 if (INTEL_DEBUG & DEBUG_STATS)
568 vfs.statistics_enable = 1;
569
570 BRW_BATCH_STRUCT(brw, &vfs);
571 }
572 }
573
574 const struct brw_tracked_state brw_invarient_state = {
575 .dirty = {
576 .mesa = 0,
577 .brw = BRW_NEW_CONTEXT,
578 .cache = 0
579 },
580 .emit = upload_invarient_state
581 };
582
583 /**
584 * Define the base addresses which some state is referenced from.
585 *
586 * This allows us to avoid having to emit relocations in many places for
587 * cached state, and instead emit pointers inside of large, mostly-static
588 * state pools. This comes at the expense of memory, and more expensive cache
589 * misses.
590 */
591 static void upload_state_base_address( struct brw_context *brw )
592 {
593 struct intel_context *intel = &brw->intel;
594
595 /* Output the structure (brw_state_base_address) directly to the
596 * batchbuffer, so we can emit relocations inline.
597 */
598 if (intel->gen >= 6) {
599 BEGIN_BATCH(10);
600 OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (10 - 2));
601 OUT_BATCH(1); /* General state base address */
602 OUT_BATCH(1); /* Surface state base address */
603 OUT_BATCH(1); /* Dynamic state base address */
604 OUT_BATCH(1); /* Indirect object base address */
605 OUT_BATCH(1); /* Instruction base address */
606 OUT_BATCH(1); /* General state upper bound */
607 OUT_BATCH(1); /* Dynamic state upper bound */
608 OUT_BATCH(1); /* Indirect object upper bound */
609 OUT_BATCH(1); /* Instruction access upper bound */
610 ADVANCE_BATCH();
611 } else if (intel->gen == 5) {
612 BEGIN_BATCH(8);
613 OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (8 - 2));
614 OUT_BATCH(1); /* General state base address */
615 OUT_BATCH(1); /* Surface state base address */
616 OUT_BATCH(1); /* Indirect object base address */
617 OUT_BATCH(1); /* Instruction base address */
618 OUT_BATCH(1); /* General state upper bound */
619 OUT_BATCH(1); /* Indirect object upper bound */
620 OUT_BATCH(1); /* Instruction access upper bound */
621 ADVANCE_BATCH();
622 } else {
623 BEGIN_BATCH(6);
624 OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
625 OUT_BATCH(1); /* General state base address */
626 OUT_BATCH(1); /* Surface state base address */
627 OUT_BATCH(1); /* Indirect object base address */
628 OUT_BATCH(1); /* General state upper bound */
629 OUT_BATCH(1); /* Indirect object upper bound */
630 ADVANCE_BATCH();
631 }
632 }
633
634 const struct brw_tracked_state brw_state_base_address = {
635 .dirty = {
636 .mesa = 0,
637 .brw = BRW_NEW_CONTEXT,
638 .cache = 0,
639 },
640 .emit = upload_state_base_address
641 };