Merge remote-tracking branch 'mareko/r300g-draw-instanced' into pipe-video
[mesa.git] / src / gallium / drivers / r300 / r300_emit.c
1 /*
2 * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
3 * Copyright 2009 Marek Olšák <maraeo@gmail.com>
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * on the rights to use, copy, modify, merge, publish, distribute, sub
9 * license, and/or sell copies of the Software, and to permit persons to whom
10 * the Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
14 * Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
20 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
21 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
22 * USE OR OTHER DEALINGS IN THE SOFTWARE. */
23
24 /* r300_emit: Functions for emitting state. */
25
26 #include "util/u_format.h"
27 #include "util/u_math.h"
28 #include "util/u_mm.h"
29
30 #include "r300_context.h"
31 #include "r300_cb.h"
32 #include "r300_cs.h"
33 #include "r300_emit.h"
34 #include "r300_fs.h"
35 #include "r300_screen.h"
36 #include "r300_screen_buffer.h"
37 #include "r300_vs.h"
38
39 void r300_emit_blend_state(struct r300_context* r300,
40 unsigned size, void* state)
41 {
42 struct r300_blend_state* blend = (struct r300_blend_state*)state;
43 struct pipe_framebuffer_state* fb =
44 (struct pipe_framebuffer_state*)r300->fb_state.state;
45 CS_LOCALS(r300);
46
47 if (fb->nr_cbufs) {
48 if (fb->cbufs[0]->format == PIPE_FORMAT_R16G16B16A16_FLOAT)
49 WRITE_CS_TABLE(blend->cb_noclamp, size);
50 else
51 WRITE_CS_TABLE(blend->cb_clamp, size);
52 } else {
53 WRITE_CS_TABLE(blend->cb_no_readwrite, size);
54 }
55 }
56
57 void r300_emit_blend_color_state(struct r300_context* r300,
58 unsigned size, void* state)
59 {
60 struct r300_blend_color_state* bc = (struct r300_blend_color_state*)state;
61 CS_LOCALS(r300);
62
63 WRITE_CS_TABLE(bc->cb, size);
64 }
65
66 void r300_emit_clip_state(struct r300_context* r300,
67 unsigned size, void* state)
68 {
69 struct r300_clip_state* clip = (struct r300_clip_state*)state;
70 CS_LOCALS(r300);
71
72 WRITE_CS_TABLE(clip->cb, size);
73 }
74
75 void r300_emit_dsa_state(struct r300_context* r300, unsigned size, void* state)
76 {
77 struct r300_dsa_state* dsa = (struct r300_dsa_state*)state;
78 struct pipe_framebuffer_state* fb =
79 (struct pipe_framebuffer_state*)r300->fb_state.state;
80 CS_LOCALS(r300);
81
82 if (fb->zsbuf) {
83 if (fb->nr_cbufs && fb->cbufs[0]->format == PIPE_FORMAT_R16G16B16A16_FLOAT)
84 WRITE_CS_TABLE(&dsa->cb_begin_fp16, size);
85 else
86 WRITE_CS_TABLE(&dsa->cb_begin, size);
87 } else {
88 if (fb->nr_cbufs && fb->cbufs[0]->format == PIPE_FORMAT_R16G16B16A16_FLOAT)
89 WRITE_CS_TABLE(dsa->cb_fp16_zb_no_readwrite, size);
90 else
91 WRITE_CS_TABLE(dsa->cb_zb_no_readwrite, size);
92 }
93 }
94
95 static void get_rc_constant_state(
96 float vec[4],
97 struct r300_context * r300,
98 struct rc_constant * constant)
99 {
100 struct r300_textures_state* texstate = r300->textures_state.state;
101 struct r300_resource *tex;
102
103 assert(constant->Type == RC_CONSTANT_STATE);
104
105 /* vec should either be (0, 0, 0, 1), which should be a relatively safe
106 * RGBA or STRQ value, or it could be one of the RC_CONSTANT_STATE
107 * state factors. */
108
109 switch (constant->u.State[0]) {
110 /* Factor for converting rectangle coords to
111 * normalized coords. Should only show up on non-r500. */
112 case RC_STATE_R300_TEXRECT_FACTOR:
113 tex = r300_resource(texstate->sampler_views[constant->u.State[1]]->base.texture);
114 vec[0] = 1.0 / tex->tex.width0;
115 vec[1] = 1.0 / tex->tex.height0;
116 vec[2] = 0;
117 vec[3] = 1;
118 break;
119
120 case RC_STATE_R300_TEXSCALE_FACTOR:
121 tex = r300_resource(texstate->sampler_views[constant->u.State[1]]->base.texture);
122 /* Add a small number to the texture size to work around rounding errors in hw. */
123 vec[0] = tex->b.b.b.width0 / (tex->tex.width0 + 0.001f);
124 vec[1] = tex->b.b.b.height0 / (tex->tex.height0 + 0.001f);
125 vec[2] = tex->b.b.b.depth0 / (tex->tex.depth0 + 0.001f);
126 vec[3] = 1;
127 break;
128
129 case RC_STATE_R300_VIEWPORT_SCALE:
130 vec[0] = r300->viewport.scale[0];
131 vec[1] = r300->viewport.scale[1];
132 vec[2] = r300->viewport.scale[2];
133 vec[3] = 1;
134 break;
135
136 case RC_STATE_R300_VIEWPORT_OFFSET:
137 vec[0] = r300->viewport.translate[0];
138 vec[1] = r300->viewport.translate[1];
139 vec[2] = r300->viewport.translate[2];
140 vec[3] = 1;
141 break;
142
143 default:
144 fprintf(stderr, "r300: Implementation error: "
145 "Unknown RC_CONSTANT type %d\n", constant->u.State[0]);
146 vec[0] = 0;
147 vec[1] = 0;
148 vec[2] = 0;
149 vec[3] = 1;
150 }
151 }
152
153 /* Convert a normal single-precision float into the 7.16 format
154 * used by the R300 fragment shader.
155 */
156 uint32_t pack_float24(float f)
157 {
158 union {
159 float fl;
160 uint32_t u;
161 } u;
162 float mantissa;
163 int exponent;
164 uint32_t float24 = 0;
165
166 if (f == 0.0)
167 return 0;
168
169 u.fl = f;
170
171 mantissa = frexpf(f, &exponent);
172
173 /* Handle -ve */
174 if (mantissa < 0) {
175 float24 |= (1 << 23);
176 mantissa = mantissa * -1.0;
177 }
178 /* Handle exponent, bias of 63 */
179 exponent += 62;
180 float24 |= (exponent << 16);
181 /* Kill 7 LSB of mantissa */
182 float24 |= (u.u & 0x7FFFFF) >> 7;
183
184 return float24;
185 }
186
187 void r300_emit_fs(struct r300_context* r300, unsigned size, void *state)
188 {
189 struct r300_fragment_shader *fs = r300_fs(r300);
190 CS_LOCALS(r300);
191
192 WRITE_CS_TABLE(fs->shader->cb_code, fs->shader->cb_code_size);
193 }
194
195 void r300_emit_fs_constants(struct r300_context* r300, unsigned size, void *state)
196 {
197 struct r300_fragment_shader *fs = r300_fs(r300);
198 struct r300_constant_buffer *buf = (struct r300_constant_buffer*)state;
199 unsigned count = fs->shader->externals_count;
200 unsigned i, j;
201 CS_LOCALS(r300);
202
203 if (count == 0)
204 return;
205
206 BEGIN_CS(size);
207 OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X, count * 4);
208 if (buf->remap_table){
209 for (i = 0; i < count; i++) {
210 float *data = (float*)&buf->ptr[buf->remap_table[i]*4];
211 for (j = 0; j < 4; j++)
212 OUT_CS(pack_float24(data[j]));
213 }
214 } else {
215 for (i = 0; i < count; i++)
216 for (j = 0; j < 4; j++)
217 OUT_CS(pack_float24(*(float*)&buf->ptr[i*4+j]));
218 }
219
220 END_CS;
221 }
222
223 void r300_emit_fs_rc_constant_state(struct r300_context* r300, unsigned size, void *state)
224 {
225 struct r300_fragment_shader *fs = r300_fs(r300);
226 struct rc_constant_list *constants = &fs->shader->code.constants;
227 unsigned i;
228 unsigned count = fs->shader->rc_state_count;
229 unsigned first = fs->shader->externals_count;
230 unsigned end = constants->Count;
231 unsigned j;
232 CS_LOCALS(r300);
233
234 if (count == 0)
235 return;
236
237 BEGIN_CS(size);
238 for(i = first; i < end; ++i) {
239 if (constants->Constants[i].Type == RC_CONSTANT_STATE) {
240 float data[4];
241
242 get_rc_constant_state(data, r300, &constants->Constants[i]);
243
244 OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X + i * 16, 4);
245 for (j = 0; j < 4; j++)
246 OUT_CS(pack_float24(data[j]));
247 }
248 }
249 END_CS;
250 }
251
252 void r500_emit_fs(struct r300_context* r300, unsigned size, void *state)
253 {
254 struct r300_fragment_shader *fs = r300_fs(r300);
255 CS_LOCALS(r300);
256
257 WRITE_CS_TABLE(fs->shader->cb_code, fs->shader->cb_code_size);
258 }
259
260 void r500_emit_fs_constants(struct r300_context* r300, unsigned size, void *state)
261 {
262 struct r300_fragment_shader *fs = r300_fs(r300);
263 struct r300_constant_buffer *buf = (struct r300_constant_buffer*)state;
264 unsigned count = fs->shader->externals_count;
265 CS_LOCALS(r300);
266
267 if (count == 0)
268 return;
269
270 BEGIN_CS(size);
271 OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST);
272 OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, count * 4);
273 if (buf->remap_table){
274 for (unsigned i = 0; i < count; i++) {
275 uint32_t *data = &buf->ptr[buf->remap_table[i]*4];
276 OUT_CS_TABLE(data, 4);
277 }
278 } else {
279 OUT_CS_TABLE(buf->ptr, count * 4);
280 }
281 END_CS;
282 }
283
284 void r500_emit_fs_rc_constant_state(struct r300_context* r300, unsigned size, void *state)
285 {
286 struct r300_fragment_shader *fs = r300_fs(r300);
287 struct rc_constant_list *constants = &fs->shader->code.constants;
288 unsigned i;
289 unsigned count = fs->shader->rc_state_count;
290 unsigned first = fs->shader->externals_count;
291 unsigned end = constants->Count;
292 CS_LOCALS(r300);
293
294 if (count == 0)
295 return;
296
297 BEGIN_CS(size);
298 for(i = first; i < end; ++i) {
299 if (constants->Constants[i].Type == RC_CONSTANT_STATE) {
300 float data[4];
301
302 get_rc_constant_state(data, r300, &constants->Constants[i]);
303
304 OUT_CS_REG(R500_GA_US_VECTOR_INDEX,
305 R500_GA_US_VECTOR_INDEX_TYPE_CONST |
306 (i & R500_GA_US_VECTOR_INDEX_MASK));
307 OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, 4);
308 OUT_CS_TABLE(data, 4);
309 }
310 }
311 END_CS;
312 }
313
314 void r300_emit_gpu_flush(struct r300_context *r300, unsigned size, void *state)
315 {
316 struct r300_gpu_flush *gpuflush = (struct r300_gpu_flush*)state;
317 struct pipe_framebuffer_state* fb =
318 (struct pipe_framebuffer_state*)r300->fb_state.state;
319 uint32_t height = fb->height;
320 uint32_t width = fb->width;
321 CS_LOCALS(r300);
322
323 if (r300->cbzb_clear) {
324 struct r300_surface *surf = r300_surface(fb->cbufs[0]);
325
326 height = surf->cbzb_height;
327 width = surf->cbzb_width;
328 }
329
330 DBG(r300, DBG_SCISSOR,
331 "r300: Scissor width: %i, height: %i, CBZB clear: %s\n",
332 width, height, r300->cbzb_clear ? "YES" : "NO");
333
334 BEGIN_CS(size);
335
336 /* Set up scissors.
337 * By writing to the SC registers, SC & US assert idle. */
338 OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2);
339 if (r300->screen->caps.is_r500) {
340 OUT_CS(0);
341 OUT_CS(((width - 1) << R300_SCISSORS_X_SHIFT) |
342 ((height - 1) << R300_SCISSORS_Y_SHIFT));
343 } else {
344 OUT_CS((1440 << R300_SCISSORS_X_SHIFT) |
345 (1440 << R300_SCISSORS_Y_SHIFT));
346 OUT_CS(((width + 1440-1) << R300_SCISSORS_X_SHIFT) |
347 ((height + 1440-1) << R300_SCISSORS_Y_SHIFT));
348 }
349
350 /* Flush CB & ZB caches and wait until the 3D engine is idle and clean. */
351 OUT_CS_TABLE(gpuflush->cb_flush_clean, 6);
352 END_CS;
353 }
354
355 void r300_emit_aa_state(struct r300_context *r300, unsigned size, void *state)
356 {
357 struct r300_aa_state *aa = (struct r300_aa_state*)state;
358 CS_LOCALS(r300);
359
360 BEGIN_CS(size);
361 OUT_CS_REG(R300_GB_AA_CONFIG, aa->aa_config);
362
363 if (aa->dest) {
364 OUT_CS_REG(R300_RB3D_AARESOLVE_OFFSET, aa->dest->offset);
365 OUT_CS_RELOC(aa->dest);
366 OUT_CS_REG(R300_RB3D_AARESOLVE_PITCH, aa->dest->pitch);
367 }
368
369 OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, aa->aaresolve_ctl);
370 END_CS;
371 }
372
373 void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
374 {
375 struct pipe_framebuffer_state* fb = (struct pipe_framebuffer_state*)state;
376 struct r300_surface* surf;
377 unsigned i;
378 uint32_t rb3d_cctl = 0;
379
380 CS_LOCALS(r300);
381
382 BEGIN_CS(size);
383
384 /* NUM_MULTIWRITES replicates COLOR[0] to all colorbuffers, which is not
385 * what we usually want. */
386 if (r300->screen->caps.is_r500) {
387 rb3d_cctl = R300_RB3D_CCTL_INDEPENDENT_COLORFORMAT_ENABLE_ENABLE;
388 }
389 if (fb->nr_cbufs && r300->fb_multiwrite) {
390 rb3d_cctl |= R300_RB3D_CCTL_NUM_MULTIWRITES(fb->nr_cbufs);
391 }
392
393 OUT_CS_REG(R300_RB3D_CCTL, rb3d_cctl);
394
395 /* Set up colorbuffers. */
396 for (i = 0; i < fb->nr_cbufs; i++) {
397 surf = r300_surface(fb->cbufs[i]);
398
399 OUT_CS_REG(R300_RB3D_COLOROFFSET0 + (4 * i), surf->offset);
400 OUT_CS_RELOC(surf);
401
402 OUT_CS_REG(R300_RB3D_COLORPITCH0 + (4 * i), surf->pitch);
403 OUT_CS_RELOC(surf);
404 }
405
406 /* Set up the ZB part of the CBZB clear. */
407 if (r300->cbzb_clear) {
408 surf = r300_surface(fb->cbufs[0]);
409
410 OUT_CS_REG(R300_ZB_FORMAT, surf->cbzb_format);
411
412 OUT_CS_REG(R300_ZB_DEPTHOFFSET, surf->cbzb_midpoint_offset);
413 OUT_CS_RELOC(surf);
414
415 OUT_CS_REG(R300_ZB_DEPTHPITCH, surf->cbzb_pitch);
416 OUT_CS_RELOC(surf);
417
418 DBG(r300, DBG_CBZB,
419 "CBZB clearing cbuf %08x %08x\n", surf->cbzb_format,
420 surf->cbzb_pitch);
421 }
422 /* Set up a zbuffer. */
423 else if (fb->zsbuf) {
424 surf = r300_surface(fb->zsbuf);
425
426 OUT_CS_REG(R300_ZB_FORMAT, surf->format);
427
428 OUT_CS_REG(R300_ZB_DEPTHOFFSET, surf->offset);
429 OUT_CS_RELOC(surf);
430
431 OUT_CS_REG(R300_ZB_DEPTHPITCH, surf->pitch);
432 OUT_CS_RELOC(surf);
433
434 if (r300->hyperz_enabled) {
435 /* HiZ RAM. */
436 OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0);
437 OUT_CS_REG(R300_ZB_HIZ_PITCH, surf->pitch_hiz);
438 /* Z Mask RAM. (compressed zbuffer) */
439 OUT_CS_REG(R300_ZB_ZMASK_OFFSET, 0);
440 OUT_CS_REG(R300_ZB_ZMASK_PITCH, surf->pitch_zmask);
441 }
442 }
443
444 END_CS;
445 }
446
447 void r300_emit_hyperz_state(struct r300_context *r300,
448 unsigned size, void *state)
449 {
450 struct r300_hyperz_state *z = state;
451 CS_LOCALS(r300);
452
453 if (z->flush)
454 WRITE_CS_TABLE(&z->cb_flush_begin, size);
455 else
456 WRITE_CS_TABLE(&z->cb_begin, size - 2);
457 }
458
459 void r300_emit_hyperz_end(struct r300_context *r300)
460 {
461 struct r300_hyperz_state z =
462 *(struct r300_hyperz_state*)r300->hyperz_state.state;
463
464 z.flush = 1;
465 z.zb_bw_cntl = 0;
466 z.zb_depthclearvalue = 0;
467 z.sc_hyperz = R300_SC_HYPERZ_ADJ_2;
468 z.gb_z_peq_config = 0;
469
470 r300_emit_hyperz_state(r300, r300->hyperz_state.size, &z);
471 }
472
473 void r300_emit_fb_state_pipelined(struct r300_context *r300,
474 unsigned size, void *state)
475 {
476 struct pipe_framebuffer_state* fb =
477 (struct pipe_framebuffer_state*)r300->fb_state.state;
478 unsigned i, num_cbufs = fb->nr_cbufs;
479 unsigned mspos0, mspos1;
480 CS_LOCALS(r300);
481
482 /* If we use the multiwrite feature, the colorbuffers 2,3,4 must be
483 * marked as UNUSED in the US block. */
484 if (r300->fb_multiwrite) {
485 num_cbufs = MIN2(num_cbufs, 1);
486 }
487
488 BEGIN_CS(size);
489
490 /* Colorbuffer format in the US block.
491 * (must be written after unpipelined regs) */
492 OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4);
493 for (i = 0; i < num_cbufs; i++) {
494 OUT_CS(r300_surface(fb->cbufs[i])->format);
495 }
496 for (; i < 1; i++) {
497 OUT_CS(R300_US_OUT_FMT_C4_8 |
498 R300_C0_SEL_B | R300_C1_SEL_G |
499 R300_C2_SEL_R | R300_C3_SEL_A);
500 }
501 for (; i < 4; i++) {
502 OUT_CS(R300_US_OUT_FMT_UNUSED);
503 }
504
505 /* Multisampling. Depends on framebuffer sample count.
506 * These are pipelined regs and as such cannot be moved
507 * to the AA state. */
508 mspos0 = 0x66666666;
509 mspos1 = 0x6666666;
510
511 if (fb->nr_cbufs && fb->cbufs[0]->texture->nr_samples > 1) {
512 /* Subsample placement. These may not be optimal. */
513 switch (fb->cbufs[0]->texture->nr_samples) {
514 case 2:
515 mspos0 = 0x33996633;
516 mspos1 = 0x6666663;
517 break;
518 case 3:
519 mspos0 = 0x33936933;
520 mspos1 = 0x6666663;
521 break;
522 case 4:
523 mspos0 = 0x33939933;
524 mspos1 = 0x3966663;
525 break;
526 case 6:
527 mspos0 = 0x22a2aa22;
528 mspos1 = 0x2a65672;
529 break;
530 default:
531 debug_printf("r300: Bad number of multisamples!\n");
532 }
533 }
534
535 OUT_CS_REG_SEQ(R300_GB_MSPOS0, 2);
536 OUT_CS(mspos0);
537 OUT_CS(mspos1);
538 END_CS;
539 }
540
541 void r300_emit_query_start(struct r300_context *r300, unsigned size, void*state)
542 {
543 struct r300_query *query = r300->query_current;
544 CS_LOCALS(r300);
545
546 if (!query)
547 return;
548
549 BEGIN_CS(size);
550 if (r300->screen->caps.family == CHIP_FAMILY_RV530) {
551 OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
552 } else {
553 OUT_CS_REG(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_ALL);
554 }
555 OUT_CS_REG(R300_ZB_ZPASS_DATA, 0);
556 END_CS;
557 query->begin_emitted = TRUE;
558 }
559
560 static void r300_emit_query_end_frag_pipes(struct r300_context *r300,
561 struct r300_query *query)
562 {
563 struct r300_capabilities* caps = &r300->screen->caps;
564 CS_LOCALS(r300);
565
566 assert(caps->num_frag_pipes);
567
568 BEGIN_CS(6 * caps->num_frag_pipes + 2);
569 /* I'm not so sure I like this switch, but it's hard to be elegant
570 * when there's so many special cases...
571 *
572 * So here's the basic idea. For each pipe, enable writes to it only,
573 * then put out the relocation for ZPASS_ADDR, taking into account a
574 * 4-byte offset for each pipe. RV380 and older are special; they have
575 * only two pipes, and the second pipe's enable is on bit 3, not bit 1,
576 * so there's a chipset cap for that. */
577 switch (caps->num_frag_pipes) {
578 case 4:
579 /* pipe 3 only */
580 OUT_CS_REG(R300_SU_REG_DEST, 1 << 3);
581 OUT_CS_REG(R300_ZB_ZPASS_ADDR, (query->num_results + 3) * 4);
582 OUT_CS_RELOC(r300->query_current);
583 case 3:
584 /* pipe 2 only */
585 OUT_CS_REG(R300_SU_REG_DEST, 1 << 2);
586 OUT_CS_REG(R300_ZB_ZPASS_ADDR, (query->num_results + 2) * 4);
587 OUT_CS_RELOC(r300->query_current);
588 case 2:
589 /* pipe 1 only */
590 /* As mentioned above, accomodate RV380 and older. */
591 OUT_CS_REG(R300_SU_REG_DEST,
592 1 << (caps->high_second_pipe ? 3 : 1));
593 OUT_CS_REG(R300_ZB_ZPASS_ADDR, (query->num_results + 1) * 4);
594 OUT_CS_RELOC(r300->query_current);
595 case 1:
596 /* pipe 0 only */
597 OUT_CS_REG(R300_SU_REG_DEST, 1 << 0);
598 OUT_CS_REG(R300_ZB_ZPASS_ADDR, (query->num_results + 0) * 4);
599 OUT_CS_RELOC(r300->query_current);
600 break;
601 default:
602 fprintf(stderr, "r300: Implementation error: Chipset reports %d"
603 " pixel pipes!\n", caps->num_frag_pipes);
604 abort();
605 }
606
607 /* And, finally, reset it to normal... */
608 OUT_CS_REG(R300_SU_REG_DEST, 0xF);
609 END_CS;
610 }
611
612 static void rv530_emit_query_end_single_z(struct r300_context *r300,
613 struct r300_query *query)
614 {
615 CS_LOCALS(r300);
616
617 BEGIN_CS(8);
618 OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
619 OUT_CS_REG(R300_ZB_ZPASS_ADDR, query->num_results * 4);
620 OUT_CS_RELOC(r300->query_current);
621 OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
622 END_CS;
623 }
624
625 static void rv530_emit_query_end_double_z(struct r300_context *r300,
626 struct r300_query *query)
627 {
628 CS_LOCALS(r300);
629
630 BEGIN_CS(14);
631 OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
632 OUT_CS_REG(R300_ZB_ZPASS_ADDR, (query->num_results + 0) * 4);
633 OUT_CS_RELOC(r300->query_current);
634 OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_1);
635 OUT_CS_REG(R300_ZB_ZPASS_ADDR, (query->num_results + 1) * 4);
636 OUT_CS_RELOC(r300->query_current);
637 OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
638 END_CS;
639 }
640
641 void r300_emit_query_end(struct r300_context* r300)
642 {
643 struct r300_capabilities *caps = &r300->screen->caps;
644 struct r300_query *query = r300->query_current;
645
646 if (!query)
647 return;
648
649 if (query->begin_emitted == FALSE)
650 return;
651
652 if (caps->family == CHIP_FAMILY_RV530) {
653 if (caps->num_z_pipes == 2)
654 rv530_emit_query_end_double_z(r300, query);
655 else
656 rv530_emit_query_end_single_z(r300, query);
657 } else
658 r300_emit_query_end_frag_pipes(r300, query);
659
660 query->begin_emitted = FALSE;
661 query->num_results += query->num_pipes;
662
663 /* XXX grab all the results and reset the counter. */
664 if (query->num_results >= query->buffer_size / 4 - 4) {
665 query->num_results = (query->buffer_size / 4) / 2;
666 fprintf(stderr, "r300: Rewinding OQBO...\n");
667 }
668 }
669
670 void r300_emit_invariant_state(struct r300_context *r300,
671 unsigned size, void *state)
672 {
673 CS_LOCALS(r300);
674 WRITE_CS_TABLE(state, size);
675 }
676
677 void r300_emit_rs_state(struct r300_context* r300, unsigned size, void* state)
678 {
679 struct r300_rs_state* rs = state;
680 CS_LOCALS(r300);
681
682 BEGIN_CS(size);
683 OUT_CS_TABLE(rs->cb_main, RS_STATE_MAIN_SIZE);
684 if (rs->polygon_offset_enable) {
685 if (r300->zbuffer_bpp == 16) {
686 OUT_CS_TABLE(rs->cb_poly_offset_zb16, 5);
687 } else {
688 OUT_CS_TABLE(rs->cb_poly_offset_zb24, 5);
689 }
690 }
691 END_CS;
692 }
693
694 void r300_emit_rs_block_state(struct r300_context* r300,
695 unsigned size, void* state)
696 {
697 struct r300_rs_block* rs = (struct r300_rs_block*)state;
698 unsigned i;
699 /* It's the same for both INST and IP tables */
700 unsigned count = (rs->inst_count & R300_RS_INST_COUNT_MASK) + 1;
701 CS_LOCALS(r300);
702
703 if (DBG_ON(r300, DBG_RS_BLOCK)) {
704 r500_dump_rs_block(rs);
705
706 fprintf(stderr, "r300: RS emit:\n");
707
708 for (i = 0; i < count; i++)
709 fprintf(stderr, " : ip %d: 0x%08x\n", i, rs->ip[i]);
710
711 for (i = 0; i < count; i++)
712 fprintf(stderr, " : inst %d: 0x%08x\n", i, rs->inst[i]);
713
714 fprintf(stderr, " : count: 0x%08x inst_count: 0x%08x\n",
715 rs->count, rs->inst_count);
716 }
717
718 BEGIN_CS(size);
719 OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL, 2);
720 OUT_CS(rs->vap_vtx_state_cntl);
721 OUT_CS(rs->vap_vsm_vtx_assm);
722 OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
723 OUT_CS(rs->vap_out_vtx_fmt[0]);
724 OUT_CS(rs->vap_out_vtx_fmt[1]);
725 OUT_CS_REG_SEQ(R300_GB_ENABLE, 1);
726 OUT_CS(rs->gb_enable);
727
728 if (r300->screen->caps.is_r500) {
729 OUT_CS_REG_SEQ(R500_RS_IP_0, count);
730 } else {
731 OUT_CS_REG_SEQ(R300_RS_IP_0, count);
732 }
733 OUT_CS_TABLE(rs->ip, count);
734
735 OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
736 OUT_CS(rs->count);
737 OUT_CS(rs->inst_count);
738
739 if (r300->screen->caps.is_r500) {
740 OUT_CS_REG_SEQ(R500_RS_INST_0, count);
741 } else {
742 OUT_CS_REG_SEQ(R300_RS_INST_0, count);
743 }
744 OUT_CS_TABLE(rs->inst, count);
745 END_CS;
746 }
747
748 void r300_emit_scissor_state(struct r300_context* r300,
749 unsigned size, void* state)
750 {
751 struct pipe_scissor_state* scissor = (struct pipe_scissor_state*)state;
752 CS_LOCALS(r300);
753
754 BEGIN_CS(size);
755 OUT_CS_REG_SEQ(R300_SC_CLIPRECT_TL_0, 2);
756 if (r300->screen->caps.is_r500) {
757 OUT_CS((scissor->minx << R300_CLIPRECT_X_SHIFT) |
758 (scissor->miny << R300_CLIPRECT_Y_SHIFT));
759 OUT_CS(((scissor->maxx - 1) << R300_CLIPRECT_X_SHIFT) |
760 ((scissor->maxy - 1) << R300_CLIPRECT_Y_SHIFT));
761 } else {
762 OUT_CS(((scissor->minx + 1440) << R300_CLIPRECT_X_SHIFT) |
763 ((scissor->miny + 1440) << R300_CLIPRECT_Y_SHIFT));
764 OUT_CS(((scissor->maxx + 1440-1) << R300_CLIPRECT_X_SHIFT) |
765 ((scissor->maxy + 1440-1) << R300_CLIPRECT_Y_SHIFT));
766 }
767 END_CS;
768 }
769
770 void r300_emit_textures_state(struct r300_context *r300,
771 unsigned size, void *state)
772 {
773 struct r300_textures_state *allstate = (struct r300_textures_state*)state;
774 struct r300_texture_sampler_state *texstate;
775 struct r300_resource *tex;
776 unsigned i;
777 boolean has_us_format = r300->screen->caps.has_us_format;
778 CS_LOCALS(r300);
779
780 BEGIN_CS(size);
781 OUT_CS_REG(R300_TX_ENABLE, allstate->tx_enable);
782
783 for (i = 0; i < allstate->count; i++) {
784 if ((1 << i) & allstate->tx_enable) {
785 texstate = &allstate->regs[i];
786 tex = r300_resource(allstate->sampler_views[i]->base.texture);
787
788 OUT_CS_REG(R300_TX_FILTER0_0 + (i * 4), texstate->filter0);
789 OUT_CS_REG(R300_TX_FILTER1_0 + (i * 4), texstate->filter1);
790 OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (i * 4),
791 texstate->border_color);
792
793 OUT_CS_REG(R300_TX_FORMAT0_0 + (i * 4), texstate->format.format0);
794 OUT_CS_REG(R300_TX_FORMAT1_0 + (i * 4), texstate->format.format1);
795 OUT_CS_REG(R300_TX_FORMAT2_0 + (i * 4), texstate->format.format2);
796
797 OUT_CS_REG(R300_TX_OFFSET_0 + (i * 4), texstate->format.tile_config);
798 OUT_CS_RELOC(tex);
799
800 if (has_us_format) {
801 OUT_CS_REG(R500_US_FORMAT0_0 + (i * 4),
802 texstate->format.us_format0);
803 }
804 }
805 }
806 END_CS;
807 }
808
809 void r300_emit_vertex_arrays(struct r300_context* r300, int offset,
810 boolean indexed, int instance_id)
811 {
812 struct pipe_vertex_buffer *vbuf = r300->vbuf_mgr->vertex_buffer;
813 struct pipe_resource **valid_vbuf = r300->vbuf_mgr->real_vertex_buffer;
814 struct pipe_vertex_element *velem = r300->velems->velem;
815 struct r300_resource *buf;
816 int i;
817 unsigned vertex_array_count = r300->velems->count;
818 unsigned real_vertex_array_count = vertex_array_count +
819 (vertex_array_count == 16 || instance_id == -1 ? 0 : 1);
820 unsigned packet_size = (real_vertex_array_count * 3 + 1) / 2;
821 struct pipe_vertex_buffer *vb1, *vb2;
822 unsigned *hw_format_size = r300->velems->format_size;
823 unsigned size1, size2, offset1, offset2, stride1, stride2;
824 CS_LOCALS(r300);
825
826 BEGIN_CS(2 + packet_size + real_vertex_array_count * 2);
827 OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size);
828 OUT_CS(real_vertex_array_count | (!indexed ? R300_VC_FORCE_PREFETCH : 0));
829
830 if (instance_id == -1) {
831 /* Non-instanced arrays. This ignores instance_divisor and instance_id. */
832 for (i = 0; i < vertex_array_count - 1; i += 2) {
833 vb1 = &vbuf[velem[i].vertex_buffer_index];
834 vb2 = &vbuf[velem[i+1].vertex_buffer_index];
835 size1 = hw_format_size[i];
836 size2 = hw_format_size[i+1];
837
838 OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride) |
839 R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(vb2->stride));
840 OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride);
841 OUT_CS(vb2->buffer_offset + velem[i+1].src_offset + offset * vb2->stride);
842 }
843
844 if (vertex_array_count & 1) {
845 vb1 = &vbuf[velem[i].vertex_buffer_index];
846 size1 = hw_format_size[i];
847
848 OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride));
849 OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride);
850 }
851
852 for (i = 0; i < vertex_array_count; i++) {
853 buf = r300_resource(valid_vbuf[velem[i].vertex_buffer_index]);
854 OUT_CS_RELOC(buf);
855 }
856 } else {
857 /* Instanced arrays. */
858 for (i = 0; i < vertex_array_count - 1; i += 2) {
859 vb1 = &vbuf[velem[i].vertex_buffer_index];
860 vb2 = &vbuf[velem[i+1].vertex_buffer_index];
861 size1 = hw_format_size[i];
862 size2 = hw_format_size[i+1];
863
864 if (velem[i].instance_divisor) {
865 stride1 = 0;
866 offset1 = vb1->buffer_offset + velem[i].src_offset +
867 (instance_id / velem[i].instance_divisor) * vb1->stride;
868 } else {
869 stride1 = vb1->stride;
870 offset1 = vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride;
871 }
872 if (velem[i+1].instance_divisor) {
873 stride2 = 0;
874 offset2 = vb2->buffer_offset + velem[i+1].src_offset +
875 (instance_id / velem[i+1].instance_divisor) * vb2->stride;
876 } else {
877 stride2 = vb2->stride;
878 offset2 = vb2->buffer_offset + velem[i+1].src_offset + offset * vb2->stride;
879 }
880
881 OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(stride1) |
882 R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(stride2));
883 OUT_CS(offset1);
884 OUT_CS(offset2);
885 }
886
887 if (vertex_array_count & 1) {
888 vb1 = &vbuf[velem[i].vertex_buffer_index];
889 size1 = hw_format_size[i];
890
891 if (velem[i].instance_divisor) {
892 stride1 = 0;
893 offset1 = vb1->buffer_offset + velem[i].src_offset +
894 (instance_id / velem[i].instance_divisor) * vb1->stride;
895 } else {
896 stride1 = vb1->stride;
897 offset1 = vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride;
898 }
899
900 /* Insert vertex buffer containing InstanceID. */
901 if (vertex_array_count < 16) {
902 OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(stride1) |
903 R300_VBPNTR_SIZE1(4));
904 OUT_CS(offset1);
905 OUT_CS(4 * instance_id);
906 } else {
907 OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(stride1));
908 OUT_CS(offset1);
909 }
910 } else if (vertex_array_count < 16) {
911 /* Insert vertex buffer containing InstanceID. */
912 OUT_CS(R300_VBPNTR_SIZE0(4));
913 OUT_CS(4 * instance_id);
914 }
915
916 for (i = 0; i < vertex_array_count; i++) {
917 buf = r300_resource(valid_vbuf[velem[i].vertex_buffer_index]);
918 OUT_CS_RELOC(buf);
919 }
920 if (vertex_array_count < 16)
921 OUT_CS_RELOC(r300->vb_instanceid);
922 }
923 END_CS;
924 }
925
926 void r300_emit_vertex_arrays_swtcl(struct r300_context *r300, boolean indexed)
927 {
928 CS_LOCALS(r300);
929
930 DBG(r300, DBG_SWTCL, "r300: Preparing vertex buffer %p for render, "
931 "vertex size %d\n", r300->vbo,
932 r300->vertex_info.size);
933 /* Set the pointer to our vertex buffer. The emitted values are this:
934 * PACKET3 [3D_LOAD_VBPNTR]
935 * COUNT [1]
936 * FORMAT [size | stride << 8]
937 * OFFSET [offset into BO]
938 * VBPNTR [relocated BO]
939 */
940 BEGIN_CS(7);
941 OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
942 OUT_CS(1 | (!indexed ? R300_VC_FORCE_PREFETCH : 0));
943 OUT_CS(r300->vertex_info.size |
944 (r300->vertex_info.size << 8));
945 OUT_CS(r300->draw_vbo_offset);
946 OUT_CS(0);
947 OUT_CS_RELOC(r300_resource(r300->vbo));
948 END_CS;
949 }
950
951 void r300_emit_vertex_stream_state(struct r300_context* r300,
952 unsigned size, void* state)
953 {
954 struct r300_vertex_element_state *velems =
955 (struct r300_vertex_element_state*)state;
956 struct r300_vertex_stream_state *streams;
957 unsigned i;
958 CS_LOCALS(r300);
959
960 if (r300->screen->caps.has_tcl && r300->instancing_enabled) {
961 streams = &velems->vertex_stream_instanced;
962 } else {
963 streams = &velems->vertex_stream;
964 }
965
966 if (DBG_ON(r300, DBG_PSC)) {
967 fprintf(stderr, "r300: PSC emit:\n");
968
969 for (i = 0; i < streams->count; i++) {
970 fprintf(stderr, " : prog_stream_cntl%d: 0x%08x\n", i,
971 streams->vap_prog_stream_cntl[i]);
972 }
973
974 for (i = 0; i < streams->count; i++) {
975 fprintf(stderr, " : prog_stream_cntl_ext%d: 0x%08x\n", i,
976 streams->vap_prog_stream_cntl_ext[i]);
977 }
978 }
979
980 BEGIN_CS((1 + streams->count) * 2);
981 OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, streams->count);
982 OUT_CS_TABLE(streams->vap_prog_stream_cntl, streams->count);
983 OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, streams->count);
984 OUT_CS_TABLE(streams->vap_prog_stream_cntl_ext, streams->count);
985 END_CS;
986 }
987
988 void r300_emit_pvs_flush(struct r300_context* r300, unsigned size, void* state)
989 {
990 CS_LOCALS(r300);
991
992 BEGIN_CS(size);
993 OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
994 END_CS;
995 }
996
997 void r300_emit_vap_invariant_state(struct r300_context *r300,
998 unsigned size, void *state)
999 {
1000 CS_LOCALS(r300);
1001 WRITE_CS_TABLE(state, size);
1002 }
1003
1004 void r300_emit_vs_state(struct r300_context* r300, unsigned size, void* state)
1005 {
1006 struct r300_vertex_shader* vs = (struct r300_vertex_shader*)state;
1007 struct r300_vertex_program_code* code = &vs->code;
1008 struct r300_screen* r300screen = r300->screen;
1009 unsigned instruction_count = code->length / 4;
1010
1011 unsigned vtx_mem_size = r300screen->caps.is_r500 ? 128 : 72;
1012 unsigned input_count = MAX2(util_bitcount(code->InputsRead), 1);
1013 unsigned output_count = MAX2(util_bitcount(code->OutputsWritten), 1);
1014 unsigned temp_count = MAX2(code->num_temporaries, 1);
1015
1016 unsigned pvs_num_slots = MIN3(vtx_mem_size / input_count,
1017 vtx_mem_size / output_count, 10);
1018 unsigned pvs_num_controllers = MIN2(vtx_mem_size / temp_count, 5);
1019
1020 CS_LOCALS(r300);
1021
1022 BEGIN_CS(size);
1023
1024 /* R300_VAP_PVS_CODE_CNTL_0
1025 * R300_VAP_PVS_CONST_CNTL
1026 * R300_VAP_PVS_CODE_CNTL_1
1027 * See the r5xx docs for instructions on how to use these. */
1028 OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, R300_PVS_FIRST_INST(0) |
1029 R300_PVS_XYZW_VALID_INST(instruction_count - 1) |
1030 R300_PVS_LAST_INST(instruction_count - 1));
1031 OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, instruction_count - 1);
1032
1033 OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0);
1034 OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->length);
1035 OUT_CS_TABLE(code->body.d, code->length);
1036
1037 OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(pvs_num_slots) |
1038 R300_PVS_NUM_CNTLRS(pvs_num_controllers) |
1039 R300_PVS_NUM_FPUS(r300screen->caps.num_vert_fpus) |
1040 R300_PVS_VF_MAX_VTX_NUM(12) |
1041 (r300screen->caps.is_r500 ? R500_TCL_STATE_OPTIMIZATION : 0));
1042
1043 /* Emit flow control instructions. */
1044 if (code->num_fc_ops) {
1045
1046 OUT_CS_REG(R300_VAP_PVS_FLOW_CNTL_OPC, code->fc_ops);
1047 if (r300screen->caps.is_r500) {
1048 OUT_CS_REG_SEQ(R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0, code->num_fc_ops * 2);
1049 OUT_CS_TABLE(code->fc_op_addrs.r500, code->num_fc_ops * 2);
1050 } else {
1051 OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_ADDRS_0, code->num_fc_ops);
1052 OUT_CS_TABLE(code->fc_op_addrs.r300, code->num_fc_ops);
1053 }
1054 OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0, code->num_fc_ops);
1055 OUT_CS_TABLE(code->fc_loop_index, code->num_fc_ops);
1056 }
1057
1058 END_CS;
1059 }
1060
1061 void r300_emit_vs_constants(struct r300_context* r300,
1062 unsigned size, void *state)
1063 {
1064 unsigned count =
1065 ((struct r300_vertex_shader*)r300->vs_state.state)->externals_count;
1066 struct r300_constant_buffer *buf = (struct r300_constant_buffer*)state;
1067 struct r300_vertex_shader *vs = (struct r300_vertex_shader*)r300->vs_state.state;
1068 unsigned i;
1069 int imm_first = vs->externals_count;
1070 int imm_end = vs->code.constants.Count;
1071 int imm_count = vs->immediates_count;
1072 CS_LOCALS(r300);
1073
1074 BEGIN_CS(size);
1075 OUT_CS_REG(R300_VAP_PVS_CONST_CNTL,
1076 R300_PVS_CONST_BASE_OFFSET(buf->buffer_base) |
1077 R300_PVS_MAX_CONST_ADDR(MAX2(imm_end - 1, 0)));
1078 if (vs->externals_count) {
1079 OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
1080 (r300->screen->caps.is_r500 ?
1081 R500_PVS_CONST_START : R300_PVS_CONST_START) + buf->buffer_base);
1082 OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, count * 4);
1083 if (buf->remap_table){
1084 for (i = 0; i < count; i++) {
1085 uint32_t *data = &buf->ptr[buf->remap_table[i]*4];
1086 OUT_CS_TABLE(data, 4);
1087 }
1088 } else {
1089 OUT_CS_TABLE(buf->ptr, count * 4);
1090 }
1091 }
1092
1093 /* Emit immediates. */
1094 if (imm_count) {
1095 OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
1096 (r300->screen->caps.is_r500 ?
1097 R500_PVS_CONST_START : R300_PVS_CONST_START) +
1098 buf->buffer_base + imm_first);
1099 OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, imm_count * 4);
1100 for (i = imm_first; i < imm_end; i++) {
1101 const float *data = vs->code.constants.Constants[i].u.Immediate;
1102 OUT_CS_TABLE(data, 4);
1103 }
1104 }
1105 END_CS;
1106 }
1107
1108 void r300_emit_viewport_state(struct r300_context* r300,
1109 unsigned size, void* state)
1110 {
1111 struct r300_viewport_state* viewport = (struct r300_viewport_state*)state;
1112 CS_LOCALS(r300);
1113
1114 BEGIN_CS(size);
1115 OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
1116 OUT_CS_TABLE(&viewport->xscale, 6);
1117 OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
1118 END_CS;
1119 }
1120
1121 void r300_emit_hiz_clear(struct r300_context *r300, unsigned size, void *state)
1122 {
1123 struct pipe_framebuffer_state *fb =
1124 (struct pipe_framebuffer_state*)r300->fb_state.state;
1125 struct r300_resource* tex;
1126 CS_LOCALS(r300);
1127
1128 tex = r300_resource(fb->zsbuf->texture);
1129
1130 BEGIN_CS(size);
1131 OUT_CS_PKT3(R300_PACKET3_3D_CLEAR_HIZ, 2);
1132 OUT_CS(0);
1133 OUT_CS(tex->tex.hiz_dwords[fb->zsbuf->u.tex.level]);
1134 OUT_CS(r300->hiz_clear_value);
1135 END_CS;
1136
1137 /* Mark the current zbuffer's hiz ram as in use. */
1138 r300->hiz_in_use = TRUE;
1139 r300->hiz_func = HIZ_FUNC_NONE;
1140 r300_mark_atom_dirty(r300, &r300->hyperz_state);
1141 }
1142
1143 void r300_emit_zmask_clear(struct r300_context *r300, unsigned size, void *state)
1144 {
1145 struct pipe_framebuffer_state *fb =
1146 (struct pipe_framebuffer_state*)r300->fb_state.state;
1147 struct r300_resource *tex;
1148 CS_LOCALS(r300);
1149
1150 tex = r300_resource(fb->zsbuf->texture);
1151
1152 BEGIN_CS(size);
1153 OUT_CS_PKT3(R300_PACKET3_3D_CLEAR_ZMASK, 2);
1154 OUT_CS(0);
1155 OUT_CS(tex->tex.zmask_dwords[fb->zsbuf->u.tex.level]);
1156 OUT_CS(0);
1157 END_CS;
1158
1159 /* Mark the current zbuffer's zmask as in use. */
1160 r300->zmask_in_use = TRUE;
1161 r300_mark_atom_dirty(r300, &r300->hyperz_state);
1162 }
1163
1164 void r300_emit_ztop_state(struct r300_context* r300,
1165 unsigned size, void* state)
1166 {
1167 struct r300_ztop_state* ztop = (struct r300_ztop_state*)state;
1168 CS_LOCALS(r300);
1169
1170 BEGIN_CS(size);
1171 OUT_CS_REG(R300_ZB_ZTOP, ztop->z_buffer_top);
1172 END_CS;
1173 }
1174
1175 void r300_emit_texture_cache_inval(struct r300_context* r300, unsigned size, void* state)
1176 {
1177 CS_LOCALS(r300);
1178
1179 BEGIN_CS(size);
1180 OUT_CS_REG(R300_TX_INVALTAGS, 0);
1181 END_CS;
1182 }
1183
1184 boolean r300_emit_buffer_validate(struct r300_context *r300,
1185 boolean do_validate_vertex_buffers,
1186 struct pipe_resource *index_buffer)
1187 {
1188 struct pipe_framebuffer_state *fb =
1189 (struct pipe_framebuffer_state*)r300->fb_state.state;
1190 struct r300_textures_state *texstate =
1191 (struct r300_textures_state*)r300->textures_state.state;
1192 struct r300_resource *tex;
1193 unsigned i;
1194 boolean flushed = FALSE;
1195
1196 validate:
1197 if (r300->fb_state.dirty) {
1198 /* Color buffers... */
1199 for (i = 0; i < fb->nr_cbufs; i++) {
1200 tex = r300_resource(fb->cbufs[i]->texture);
1201 assert(tex && tex->buf && "cbuf is marked, but NULL!");
1202 r300->rws->cs_add_reloc(r300->cs, tex->cs_buf, 0,
1203 r300_surface(fb->cbufs[i])->domain);
1204 }
1205 /* ...depth buffer... */
1206 if (fb->zsbuf) {
1207 tex = r300_resource(fb->zsbuf->texture);
1208 assert(tex && tex->buf && "zsbuf is marked, but NULL!");
1209 r300->rws->cs_add_reloc(r300->cs, tex->cs_buf, 0,
1210 r300_surface(fb->zsbuf)->domain);
1211 }
1212 }
1213 if (r300->textures_state.dirty) {
1214 /* ...textures... */
1215 for (i = 0; i < texstate->count; i++) {
1216 if (!(texstate->tx_enable & (1 << i))) {
1217 continue;
1218 }
1219
1220 tex = r300_resource(texstate->sampler_views[i]->base.texture);
1221 r300->rws->cs_add_reloc(r300->cs, tex->cs_buf, tex->domain, 0);
1222 }
1223 }
1224 /* ...occlusion query buffer... */
1225 if (r300->query_current)
1226 r300->rws->cs_add_reloc(r300->cs, r300->query_current->cs_buf,
1227 0, r300->query_current->domain);
1228 /* ...vertex buffer for SWTCL path... */
1229 if (r300->vbo)
1230 r300->rws->cs_add_reloc(r300->cs, r300_resource(r300->vbo)->cs_buf,
1231 r300_resource(r300->vbo)->domain, 0);
1232 /* ...vertex buffers for HWTCL path... */
1233 if (do_validate_vertex_buffers && r300->vertex_arrays_dirty) {
1234 struct pipe_resource **buf = r300->vbuf_mgr->real_vertex_buffer;
1235 struct pipe_resource **last = r300->vbuf_mgr->real_vertex_buffer +
1236 r300->vbuf_mgr->nr_real_vertex_buffers;
1237 for (; buf != last; buf++) {
1238 if (!*buf)
1239 continue;
1240
1241 r300->rws->cs_add_reloc(r300->cs, r300_resource(*buf)->cs_buf,
1242 r300_resource(*buf)->domain, 0);
1243 }
1244 if (r300->instancing_enabled) {
1245 r300->rws->cs_add_reloc(r300->cs, r300->vb_instanceid->cs_buf,
1246 r300->vb_instanceid->domain, 0);
1247 }
1248 }
1249 /* ...and index buffer for HWTCL path. */
1250 if (index_buffer)
1251 r300->rws->cs_add_reloc(r300->cs, r300_resource(index_buffer)->cs_buf,
1252 r300_resource(index_buffer)->domain, 0);
1253
1254 /* Now do the validation. */
1255 if (!r300->rws->cs_validate(r300->cs)) {
1256 /* Ooops, an infinite loop, give up. */
1257 if (flushed)
1258 return FALSE;
1259
1260 r300_flush(&r300->context, RADEON_FLUSH_ASYNC, NULL);
1261 flushed = TRUE;
1262 goto validate;
1263 }
1264
1265 return TRUE;
1266 }
1267
1268 unsigned r300_get_num_dirty_dwords(struct r300_context *r300)
1269 {
1270 struct r300_atom* atom;
1271 unsigned dwords = 0;
1272
1273 foreach_dirty_atom(r300, atom) {
1274 if (atom->dirty) {
1275 dwords += atom->size;
1276 }
1277 }
1278
1279 /* let's reserve some more, just in case */
1280 dwords += 32;
1281
1282 return dwords;
1283 }
1284
1285 unsigned r300_get_num_cs_end_dwords(struct r300_context *r300)
1286 {
1287 unsigned dwords = 0;
1288
1289 /* Emitted in flush. */
1290 dwords += 26; /* emit_query_end */
1291 dwords += r300->hyperz_state.size + 2; /* emit_hyperz_end + zcache flush */
1292 if (r300->screen->caps.is_r500)
1293 dwords += 2;
1294
1295 return dwords;
1296 }
1297
1298 /* Emit all dirty state. */
1299 void r300_emit_dirty_state(struct r300_context* r300)
1300 {
1301 struct r300_atom *atom;
1302
1303 foreach_dirty_atom(r300, atom) {
1304 if (atom->dirty) {
1305 atom->emit(r300, atom->size, atom->state);
1306 atom->dirty = FALSE;
1307 }
1308 }
1309
1310 r300->first_dirty = NULL;
1311 r300->last_dirty = NULL;
1312 r300->dirty_hw++;
1313 }