r300g: fix reference counting when translating indices
[mesa.git] / src / gallium / drivers / 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 "brw_debug.h"
35 #include "brw_batchbuffer.h"
36 #include "brw_context.h"
37 #include "brw_state.h"
38 #include "brw_defines.h"
39 #include "brw_screen.h"
40 #include "brw_pipe_rast.h"
41
42
43
44
45
46 /***********************************************************************
47 * Blend color
48 */
49
50 static int upload_blend_constant_color(struct brw_context *brw)
51 {
52 BRW_CACHED_BATCH_STRUCT(brw, &brw->curr.bcc);
53 return 0;
54 }
55
56
57 const struct brw_tracked_state brw_blend_constant_color = {
58 .dirty = {
59 .mesa = PIPE_NEW_BLEND_COLOR,
60 .brw = 0,
61 .cache = 0
62 },
63 .emit = upload_blend_constant_color
64 };
65
66 /***********************************************************************
67 * Drawing rectangle - framebuffer dimensions
68 */
69 static int upload_drawing_rect(struct brw_context *brw)
70 {
71 BEGIN_BATCH(4, NO_LOOP_CLIPRECTS);
72 OUT_BATCH(_3DSTATE_DRAWRECT_INFO_I965);
73 OUT_BATCH(0);
74 OUT_BATCH(((brw->curr.fb.width - 1) & 0xffff) |
75 ((brw->curr.fb.height - 1) << 16));
76 OUT_BATCH(0);
77 ADVANCE_BATCH();
78 return 0;
79 }
80
81 const struct brw_tracked_state brw_drawing_rect = {
82 .dirty = {
83 .mesa = PIPE_NEW_FRAMEBUFFER_DIMENSIONS,
84 .brw = 0,
85 .cache = 0
86 },
87 .emit = upload_drawing_rect
88 };
89
90
91 /***********************************************************************
92 * Binding table pointers
93 */
94
95 static int prepare_binding_table_pointers(struct brw_context *brw)
96 {
97 brw_add_validated_bo(brw, brw->vs.bind_bo);
98 brw_add_validated_bo(brw, brw->wm.bind_bo);
99 return 0;
100 }
101
102 /**
103 * Upload the binding table pointers, which point each stage's array of surface
104 * state pointers.
105 *
106 * The binding table pointers are relative to the surface state base address,
107 * which is 0.
108 */
109 static int upload_binding_table_pointers(struct brw_context *brw)
110 {
111 BEGIN_BATCH(6, IGNORE_CLIPRECTS);
112 OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2));
113 if (brw->vs.bind_bo != NULL)
114 OUT_RELOC(brw->vs.bind_bo,
115 BRW_USAGE_SAMPLER,
116 0); /* vs */
117 else
118 OUT_BATCH(0);
119 OUT_BATCH(0); /* gs */
120 OUT_BATCH(0); /* clip */
121 OUT_BATCH(0); /* sf */
122 OUT_RELOC(brw->wm.bind_bo,
123 BRW_USAGE_SAMPLER,
124 0); /* wm/ps */
125 ADVANCE_BATCH();
126 return 0;
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 /**********************************************************************
141 * Upload pointers to the per-stage state.
142 *
143 * The state pointers in this packet are all relative to the general state
144 * base address set by CMD_STATE_BASE_ADDRESS, which is 0.
145 */
146 static int upload_pipelined_state_pointers(struct brw_context *brw )
147 {
148 BEGIN_BATCH(7, IGNORE_CLIPRECTS);
149 OUT_BATCH(CMD_PIPELINED_STATE_POINTERS << 16 | (7 - 2));
150 OUT_RELOC(brw->vs.state_bo,
151 BRW_USAGE_STATE,
152 0);
153 if (brw->gs.prog_active)
154 OUT_RELOC(brw->gs.state_bo,
155 BRW_USAGE_STATE,
156 1);
157 else
158 OUT_BATCH(0);
159 OUT_RELOC(brw->clip.state_bo,
160 BRW_USAGE_STATE,
161 1);
162 OUT_RELOC(brw->sf.state_bo,
163 BRW_USAGE_STATE,
164 0);
165 OUT_RELOC(brw->wm.state_bo,
166 BRW_USAGE_STATE,
167 0);
168 OUT_RELOC(brw->cc.state_bo,
169 BRW_USAGE_STATE,
170 0);
171 ADVANCE_BATCH();
172
173 brw->state.dirty.brw |= BRW_NEW_PSP;
174 return 0;
175 }
176
177
178 static int prepare_psp_urb_cbs(struct brw_context *brw)
179 {
180 brw_add_validated_bo(brw, brw->vs.state_bo);
181 brw_add_validated_bo(brw, brw->gs.state_bo);
182 brw_add_validated_bo(brw, brw->clip.state_bo);
183 brw_add_validated_bo(brw, brw->sf.state_bo);
184 brw_add_validated_bo(brw, brw->wm.state_bo);
185 brw_add_validated_bo(brw, brw->cc.state_bo);
186 return 0;
187 }
188
189 static int upload_psp_urb_cbs(struct brw_context *brw )
190 {
191 int ret;
192
193 ret = upload_pipelined_state_pointers(brw);
194 if (ret)
195 return ret;
196
197 ret = brw_upload_urb_fence(brw);
198 if (ret)
199 return ret;
200
201 ret = brw_upload_cs_urb_state(brw);
202 if (ret)
203 return ret;
204
205 return 0;
206 }
207
208 const struct brw_tracked_state brw_psp_urb_cbs = {
209 .dirty = {
210 .mesa = 0,
211 .brw = BRW_NEW_URB_FENCE | BRW_NEW_BATCH,
212 .cache = (CACHE_NEW_VS_UNIT |
213 CACHE_NEW_GS_UNIT |
214 CACHE_NEW_GS_PROG |
215 CACHE_NEW_CLIP_UNIT |
216 CACHE_NEW_SF_UNIT |
217 CACHE_NEW_WM_UNIT |
218 CACHE_NEW_CC_UNIT)
219 },
220 .prepare = prepare_psp_urb_cbs,
221 .emit = upload_psp_urb_cbs,
222 };
223
224
225 /***********************************************************************
226 * Depth buffer
227 */
228
229 static int prepare_depthbuffer(struct brw_context *brw)
230 {
231 struct pipe_surface *zsbuf = brw->curr.fb.zsbuf;
232
233 if (zsbuf)
234 brw_add_validated_bo(brw, brw_surface(zsbuf)->bo);
235
236 return 0;
237 }
238
239 static int emit_depthbuffer(struct brw_context *brw)
240 {
241 struct pipe_surface *surface = brw->curr.fb.zsbuf;
242 unsigned int len = (brw->is_g4x || brw->gen == 5) ? 6 : 5;
243
244 if (surface == NULL) {
245 BEGIN_BATCH(len, IGNORE_CLIPRECTS);
246 OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
247 OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
248 (BRW_SURFACE_NULL << 29));
249 OUT_BATCH(0);
250 OUT_BATCH(0);
251 OUT_BATCH(0);
252
253 if (brw->is_g4x || brw->gen == 5)
254 OUT_BATCH(0);
255
256 ADVANCE_BATCH();
257 } else {
258 struct brw_winsys_buffer *bo;
259 unsigned int format;
260 unsigned int pitch;
261 unsigned int cpp;
262
263 switch (surface->format) {
264 case PIPE_FORMAT_Z16_UNORM:
265 format = BRW_DEPTHFORMAT_D16_UNORM;
266 cpp = 2;
267 break;
268 case PIPE_FORMAT_Z24X8_UNORM:
269 case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
270 format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT;
271 cpp = 4;
272 break;
273 case PIPE_FORMAT_Z32_FLOAT:
274 format = BRW_DEPTHFORMAT_D32_FLOAT;
275 cpp = 4;
276 break;
277 default:
278 assert(0);
279 return PIPE_ERROR_BAD_INPUT;
280 }
281
282 bo = brw_surface(surface)->bo;
283 pitch = brw_surface(surface)->pitch;
284
285 BEGIN_BATCH(len, IGNORE_CLIPRECTS);
286 OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
287 OUT_BATCH(((pitch * cpp) - 1) |
288 (format << 18) |
289 (BRW_TILEWALK_YMAJOR << 26) |
290 /* always linear ?
291 ((surface->layout != PIPE_SURFACE_LAYOUT_LINEAR) << 27) |*/
292 (BRW_SURFACE_2D << 29));
293 OUT_RELOC(bo,
294 BRW_USAGE_DEPTH_BUFFER,
295 brw_surface(surface)->offset);
296 OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) |
297 ((pitch - 1) << 6) |
298 ((surface->height - 1) << 19));
299 OUT_BATCH(0);
300
301 if (brw->is_g4x || brw->gen == 5)
302 OUT_BATCH(0);
303
304 ADVANCE_BATCH();
305 }
306
307 return 0;
308 }
309
310 const struct brw_tracked_state brw_depthbuffer = {
311 .dirty = {
312 .mesa = PIPE_NEW_DEPTH_BUFFER,
313 .brw = BRW_NEW_BATCH,
314 .cache = 0,
315 },
316 .prepare = prepare_depthbuffer,
317 .emit = emit_depthbuffer,
318 };
319
320
321
322 /***********************************************************************
323 * Polygon stipple packet
324 */
325
326 static int upload_polygon_stipple(struct brw_context *brw)
327 {
328 BRW_CACHED_BATCH_STRUCT(brw, &brw->curr.bps);
329 return 0;
330 }
331
332 const struct brw_tracked_state brw_polygon_stipple = {
333 .dirty = {
334 .mesa = PIPE_NEW_POLYGON_STIPPLE,
335 .brw = 0,
336 .cache = 0
337 },
338 .emit = upload_polygon_stipple
339 };
340
341
342 /***********************************************************************
343 * Line stipple packet
344 */
345
346 static int upload_line_stipple(struct brw_context *brw)
347 {
348 const struct brw_line_stipple *bls = &brw->curr.rast->bls;
349 if (bls->header.opcode) {
350 BRW_CACHED_BATCH_STRUCT(brw, bls);
351 }
352 return 0;
353 }
354
355 const struct brw_tracked_state brw_line_stipple = {
356 .dirty = {
357 .mesa = PIPE_NEW_RAST,
358 .brw = 0,
359 .cache = 0
360 },
361 .emit = upload_line_stipple
362 };
363
364
365 /***********************************************************************
366 * Misc invariant state packets
367 */
368
369 static int upload_invariant_state( struct brw_context *brw )
370 {
371 {
372 /* 0x61040000 Pipeline Select */
373 /* PipelineSelect : 0 */
374 struct brw_pipeline_select ps;
375
376 memset(&ps, 0, sizeof(ps));
377 if (brw->is_g4x || brw->gen == 5)
378 ps.header.opcode = CMD_PIPELINE_SELECT_GM45;
379 else
380 ps.header.opcode = CMD_PIPELINE_SELECT_965;
381 ps.header.pipeline_select = 0;
382 BRW_BATCH_STRUCT(brw, &ps);
383 }
384
385 {
386 struct brw_global_depth_offset_clamp gdo;
387 memset(&gdo, 0, sizeof(gdo));
388
389 /* Disable depth offset clamping.
390 */
391 gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP;
392 gdo.header.length = sizeof(gdo)/4 - 2;
393 gdo.depth_offset_clamp = 0.0;
394
395 BRW_BATCH_STRUCT(brw, &gdo);
396 }
397
398
399 /* 0x61020000 State Instruction Pointer */
400 {
401 struct brw_system_instruction_pointer sip;
402 memset(&sip, 0, sizeof(sip));
403
404 sip.header.opcode = CMD_STATE_INSN_POINTER;
405 sip.header.length = 0;
406 sip.bits0.pad = 0;
407 sip.bits0.system_instruction_pointer = 0;
408 BRW_BATCH_STRUCT(brw, &sip);
409 }
410
411 /* VF Statistics */
412 {
413 struct brw_vf_statistics vfs;
414 memset(&vfs, 0, sizeof(vfs));
415
416 if (brw->is_g4x || brw->gen == 5)
417 vfs.opcode = CMD_VF_STATISTICS_GM45;
418 else
419 vfs.opcode = CMD_VF_STATISTICS_965;
420
421 if (BRW_DEBUG & DEBUG_STATS)
422 vfs.statistics_enable = 1;
423
424 BRW_BATCH_STRUCT(brw, &vfs);
425 }
426
427 if (!(brw->gen == 4))
428 {
429 struct brw_aa_line_parameters balp;
430
431 /* use legacy aa line coverage computation */
432 memset(&balp, 0, sizeof(balp));
433 balp.header.opcode = CMD_AA_LINE_PARAMETERS;
434 balp.header.length = sizeof(balp) / 4 - 2;
435
436 BRW_BATCH_STRUCT(brw, &balp);
437 }
438
439 {
440 struct brw_polygon_stipple_offset bpso;
441
442 /* This is invariant state in gallium:
443 */
444 memset(&bpso, 0, sizeof(bpso));
445 bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET;
446 bpso.header.length = sizeof(bpso)/4-2;
447 bpso.bits0.y_offset = 0;
448 bpso.bits0.x_offset = 0;
449
450 BRW_BATCH_STRUCT(brw, &bpso);
451 }
452
453 return 0;
454 }
455
456 const struct brw_tracked_state brw_invariant_state = {
457 .dirty = {
458 .mesa = 0,
459 .brw = BRW_NEW_CONTEXT,
460 .cache = 0
461 },
462 .emit = upload_invariant_state
463 };
464
465
466 /***********************************************************************
467 * State base address
468 */
469
470 /**
471 * Define the base addresses which some state is referenced from.
472 *
473 * This allows us to avoid having to emit relocations in many places for
474 * cached state, and instead emit pointers inside of large, mostly-static
475 * state pools. This comes at the expense of memory, and more expensive cache
476 * misses.
477 */
478 static int upload_state_base_address( struct brw_context *brw )
479 {
480 /* Output the structure (brw_state_base_address) directly to the
481 * batchbuffer, so we can emit relocations inline.
482 */
483 if (brw->gen == 5) {
484 BEGIN_BATCH(8, IGNORE_CLIPRECTS);
485 OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (8 - 2));
486 OUT_BATCH(1); /* General state base address */
487 OUT_BATCH(1); /* Surface state base address */
488 OUT_BATCH(1); /* Indirect object base address */
489 OUT_BATCH(1); /* Instruction base address */
490 OUT_BATCH(1); /* General state upper bound */
491 OUT_BATCH(1); /* Indirect object upper bound */
492 OUT_BATCH(1); /* Instruction access upper bound */
493 ADVANCE_BATCH();
494 } else {
495 BEGIN_BATCH(6, IGNORE_CLIPRECTS);
496 OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
497 OUT_BATCH(1); /* General state base address */
498 OUT_BATCH(1); /* Surface state base address */
499 OUT_BATCH(1); /* Indirect object base address */
500 OUT_BATCH(1); /* General state upper bound */
501 OUT_BATCH(1); /* Indirect object upper bound */
502 ADVANCE_BATCH();
503 }
504 return 0;
505 }
506
507 const struct brw_tracked_state brw_state_base_address = {
508 .dirty = {
509 .mesa = 0,
510 .brw = BRW_NEW_CONTEXT,
511 .cache = 0,
512 },
513 .emit = upload_state_base_address
514 };