2 * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
3 * Copyright 2009 Marek Olšák <maraeo@gmail.com>
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:
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
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. */
24 /* r300_emit: Functions for emitting state. */
26 #include "util/u_format.h"
27 #include "util/u_math.h"
28 #include "util/u_simple_list.h"
30 #include "r300_context.h"
32 #include "r300_emit.h"
34 #include "r300_screen.h"
35 #include "r300_screen_buffer.h"
38 void r300_emit_blend_state(struct r300_context
* r300
,
39 unsigned size
, void* state
)
41 struct r300_blend_state
* blend
= (struct r300_blend_state
*)state
;
42 struct pipe_framebuffer_state
* fb
=
43 (struct pipe_framebuffer_state
*)r300
->fb_state
.state
;
47 WRITE_CS_TABLE(blend
->cb
, size
);
49 WRITE_CS_TABLE(blend
->cb_no_readwrite
, size
);
53 void r300_emit_blend_color_state(struct r300_context
* r300
,
54 unsigned size
, void* state
)
56 struct r300_blend_color_state
* bc
= (struct r300_blend_color_state
*)state
;
59 WRITE_CS_TABLE(bc
->cb
, size
);
62 void r300_emit_clip_state(struct r300_context
* r300
,
63 unsigned size
, void* state
)
65 struct r300_clip_state
* clip
= (struct r300_clip_state
*)state
;
68 WRITE_CS_TABLE(clip
->cb
, size
);
71 void r300_emit_dsa_state(struct r300_context
* r300
, unsigned size
, void* state
)
73 struct r300_dsa_state
* dsa
= (struct r300_dsa_state
*)state
;
74 struct pipe_framebuffer_state
* fb
=
75 (struct pipe_framebuffer_state
*)r300
->fb_state
.state
;
79 WRITE_CS_TABLE(&dsa
->cb_begin
, size
);
81 WRITE_CS_TABLE(dsa
->cb_no_readwrite
, size
);
85 static const float * get_rc_constant_state(
86 struct r300_context
* r300
,
87 struct rc_constant
* constant
)
89 struct r300_textures_state
* texstate
= r300
->textures_state
.state
;
90 static float vec
[4] = { 0.0, 0.0, 0.0, 1.0 };
91 struct pipe_resource
*tex
;
93 assert(constant
->Type
== RC_CONSTANT_STATE
);
95 switch (constant
->u
.State
[0]) {
96 /* Factor for converting rectangle coords to
97 * normalized coords. Should only show up on non-r500. */
98 case RC_STATE_R300_TEXRECT_FACTOR
:
99 tex
= texstate
->sampler_views
[constant
->u
.State
[1]]->base
.texture
;
100 vec
[0] = 1.0 / tex
->width0
;
101 vec
[1] = 1.0 / tex
->height0
;
104 case RC_STATE_R300_VIEWPORT_SCALE
:
105 vec
[0] = r300
->viewport
.scale
[0];
106 vec
[1] = r300
->viewport
.scale
[1];
107 vec
[2] = r300
->viewport
.scale
[2];
110 case RC_STATE_R300_VIEWPORT_OFFSET
:
111 vec
[0] = r300
->viewport
.translate
[0];
112 vec
[1] = r300
->viewport
.translate
[1];
113 vec
[2] = r300
->viewport
.translate
[2];
117 fprintf(stderr
, "r300: Implementation error: "
118 "Unknown RC_CONSTANT type %d\n", constant
->u
.State
[0]);
121 /* This should either be (0, 0, 0, 1), which should be a relatively safe
122 * RGBA or STRQ value, or it could be one of the RC_CONSTANT_STATE
127 /* Convert a normal single-precision float into the 7.16 format
128 * used by the R300 fragment shader.
130 uint32_t pack_float24(float f
)
138 uint32_t float24
= 0;
145 mantissa
= frexpf(f
, &exponent
);
149 float24
|= (1 << 23);
150 mantissa
= mantissa
* -1.0;
152 /* Handle exponent, bias of 63 */
154 float24
|= (exponent
<< 16);
155 /* Kill 7 LSB of mantissa */
156 float24
|= (u
.u
& 0x7FFFFF) >> 7;
161 void r300_emit_fs(struct r300_context
* r300
, unsigned size
, void *state
)
163 struct r300_fragment_shader
*fs
= r300_fs(r300
);
166 WRITE_CS_TABLE(fs
->shader
->cb_code
, fs
->shader
->cb_code_size
);
169 void r300_emit_fs_constants(struct r300_context
* r300
, unsigned size
, void *state
)
171 struct r300_fragment_shader
*fs
= r300_fs(r300
);
172 struct r300_constant_buffer
*buf
= (struct r300_constant_buffer
*)state
;
173 unsigned count
= fs
->shader
->externals_count
* 4;
180 OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X
, count
);
181 OUT_CS_TABLE(buf
->constants
, count
);
185 void r300_emit_fs_rc_constant_state(struct r300_context
* r300
, unsigned size
, void *state
)
187 struct r300_fragment_shader
*fs
= r300_fs(r300
);
188 struct rc_constant_list
*constants
= &fs
->shader
->code
.constants
;
190 unsigned count
= fs
->shader
->rc_state_count
;
191 unsigned first
= fs
->shader
->externals_count
;
192 unsigned end
= constants
->Count
;
201 for(i
= first
; i
< end
; ++i
) {
202 if (constants
->Constants
[i
].Type
== RC_CONSTANT_STATE
) {
204 get_rc_constant_state(r300
, &constants
->Constants
[i
]);
206 for (j
= 0; j
< 4; j
++)
207 cdata
[j
] = pack_float24(data
[j
]);
209 OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X
+ i
* 16, 4);
210 OUT_CS_TABLE(cdata
, 4);
216 void r500_emit_fs(struct r300_context
* r300
, unsigned size
, void *state
)
218 struct r300_fragment_shader
*fs
= r300_fs(r300
);
221 WRITE_CS_TABLE(fs
->shader
->cb_code
, fs
->shader
->cb_code_size
);
224 void r500_emit_fs_constants(struct r300_context
* r300
, unsigned size
, void *state
)
226 struct r300_fragment_shader
*fs
= r300_fs(r300
);
227 struct r300_constant_buffer
*buf
= (struct r300_constant_buffer
*)state
;
228 unsigned count
= fs
->shader
->externals_count
* 4;
235 OUT_CS_REG(R500_GA_US_VECTOR_INDEX
, R500_GA_US_VECTOR_INDEX_TYPE_CONST
);
236 OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA
, count
);
237 OUT_CS_TABLE(buf
->constants
, count
);
241 void r500_emit_fs_rc_constant_state(struct r300_context
* r300
, unsigned size
, void *state
)
243 struct r300_fragment_shader
*fs
= r300_fs(r300
);
244 struct rc_constant_list
*constants
= &fs
->shader
->code
.constants
;
246 unsigned count
= fs
->shader
->rc_state_count
;
247 unsigned first
= fs
->shader
->externals_count
;
248 unsigned end
= constants
->Count
;
255 for(i
= first
; i
< end
; ++i
) {
256 if (constants
->Constants
[i
].Type
== RC_CONSTANT_STATE
) {
258 get_rc_constant_state(r300
, &constants
->Constants
[i
]);
260 OUT_CS_REG(R500_GA_US_VECTOR_INDEX
,
261 R500_GA_US_VECTOR_INDEX_TYPE_CONST
|
262 (i
& R500_GA_US_VECTOR_INDEX_MASK
));
263 OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA
, 4);
264 OUT_CS_TABLE(data
, 4);
270 void r300_emit_fb_state(struct r300_context
* r300
, unsigned size
, void* state
)
272 struct pipe_framebuffer_state
* fb
= (struct pipe_framebuffer_state
*)state
;
273 struct r300_surface
* surf
;
279 /* Flush and free renderbuffer caches. */
280 OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT
,
281 R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS
|
282 R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D
);
283 OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT
,
284 R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE
|
285 R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE
);
287 /* NUM_MULTIWRITES replicates COLOR[0] to all colorbuffers, which is not
288 * what we usually want. */
289 if (r300
->screen
->caps
.is_r500
) {
290 OUT_CS_REG(R300_RB3D_CCTL
,
291 R300_RB3D_CCTL_INDEPENDENT_COLORFORMAT_ENABLE_ENABLE
);
293 OUT_CS_REG(R300_RB3D_CCTL
, 0);
296 /* Set up colorbuffers. */
297 for (i
= 0; i
< fb
->nr_cbufs
; i
++) {
298 surf
= r300_surface(fb
->cbufs
[i
]);
300 OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0
+ (4 * i
), 1);
301 OUT_CS_RELOC(surf
->buffer
, surf
->offset
, 0, surf
->domain
, 0);
303 OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0
+ (4 * i
), 1);
304 OUT_CS_RELOC(surf
->buffer
, surf
->pitch
, 0, surf
->domain
, 0);
306 OUT_CS_REG(R300_US_OUT_FMT_0
+ (4 * i
), surf
->format
);
309 OUT_CS_REG(R300_US_OUT_FMT_0
+ (4 * i
), R300_US_OUT_FMT_UNUSED
);
312 /* Set up a zbuffer. */
314 surf
= r300_surface(fb
->zsbuf
);
316 OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET
, 1);
317 OUT_CS_RELOC(surf
->buffer
, surf
->offset
, 0, surf
->domain
, 0);
319 OUT_CS_REG(R300_ZB_FORMAT
, surf
->format
);
321 OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH
, 1);
322 OUT_CS_RELOC(surf
->buffer
, surf
->pitch
, 0, surf
->domain
, 0);
325 OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL
, 2);
326 if (r300
->screen
->caps
.is_r500
) {
328 OUT_CS(((fb
->width
- 1) << R300_SCISSORS_X_SHIFT
) |
329 ((fb
->height
- 1) << R300_SCISSORS_Y_SHIFT
));
331 OUT_CS((1440 << R300_SCISSORS_X_SHIFT
) |
332 (1440 << R300_SCISSORS_Y_SHIFT
));
333 OUT_CS(((fb
->width
+ 1440-1) << R300_SCISSORS_X_SHIFT
) |
334 ((fb
->height
+ 1440-1) << R300_SCISSORS_Y_SHIFT
));
339 void r300_emit_query_start(struct r300_context
*r300
, unsigned size
, void*state
)
341 struct r300_query
*query
= r300
->query_current
;
348 if (r300
->screen
->caps
.family
== CHIP_FAMILY_RV530
) {
349 OUT_CS_REG(RV530_FG_ZBREG_DEST
, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL
);
351 OUT_CS_REG(R300_SU_REG_DEST
, R300_RASTER_PIPE_SELECT_ALL
);
353 OUT_CS_REG(R300_ZB_ZPASS_DATA
, 0);
355 query
->begin_emitted
= TRUE
;
356 query
->flushed
= FALSE
;
359 static void r300_emit_query_end_frag_pipes(struct r300_context
*r300
,
360 struct r300_query
*query
)
362 struct r300_capabilities
* caps
= &r300
->screen
->caps
;
363 struct r300_winsys_buffer
*buf
= r300
->query_current
->buffer
;
366 assert(caps
->num_frag_pipes
);
368 BEGIN_CS(6 * caps
->num_frag_pipes
+ 2);
369 /* I'm not so sure I like this switch, but it's hard to be elegant
370 * when there's so many special cases...
372 * So here's the basic idea. For each pipe, enable writes to it only,
373 * then put out the relocation for ZPASS_ADDR, taking into account a
374 * 4-byte offset for each pipe. RV380 and older are special; they have
375 * only two pipes, and the second pipe's enable is on bit 3, not bit 1,
376 * so there's a chipset cap for that. */
377 switch (caps
->num_frag_pipes
) {
380 OUT_CS_REG(R300_SU_REG_DEST
, 1 << 3);
381 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
382 OUT_CS_RELOC(buf
, (query
->num_results
+ 3) * 4,
383 0, query
->domain
, 0);
386 OUT_CS_REG(R300_SU_REG_DEST
, 1 << 2);
387 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
388 OUT_CS_RELOC(buf
, (query
->num_results
+ 2) * 4,
389 0, query
->domain
, 0);
392 /* As mentioned above, accomodate RV380 and older. */
393 OUT_CS_REG(R300_SU_REG_DEST
,
394 1 << (caps
->high_second_pipe
? 3 : 1));
395 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
396 OUT_CS_RELOC(buf
, (query
->num_results
+ 1) * 4,
397 0, query
->domain
, 0);
400 OUT_CS_REG(R300_SU_REG_DEST
, 1 << 0);
401 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
402 OUT_CS_RELOC(buf
, (query
->num_results
+ 0) * 4,
403 0, query
->domain
, 0);
406 fprintf(stderr
, "r300: Implementation error: Chipset reports %d"
407 " pixel pipes!\n", caps
->num_frag_pipes
);
411 /* And, finally, reset it to normal... */
412 OUT_CS_REG(R300_SU_REG_DEST
, 0xF);
416 static void rv530_emit_query_end_single_z(struct r300_context
*r300
,
417 struct r300_query
*query
)
419 struct r300_winsys_buffer
*buf
= r300
->query_current
->buffer
;
423 OUT_CS_REG(RV530_FG_ZBREG_DEST
, RV530_FG_ZBREG_DEST_PIPE_SELECT_0
);
424 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
425 OUT_CS_RELOC(buf
, query
->num_results
* 4, 0, query
->domain
, 0);
426 OUT_CS_REG(RV530_FG_ZBREG_DEST
, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL
);
430 static void rv530_emit_query_end_double_z(struct r300_context
*r300
,
431 struct r300_query
*query
)
433 struct r300_winsys_buffer
*buf
= r300
->query_current
->buffer
;
437 OUT_CS_REG(RV530_FG_ZBREG_DEST
, RV530_FG_ZBREG_DEST_PIPE_SELECT_0
);
438 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
439 OUT_CS_RELOC(buf
, (query
->num_results
+ 0) * 4, 0, query
->domain
, 0);
440 OUT_CS_REG(RV530_FG_ZBREG_DEST
, RV530_FG_ZBREG_DEST_PIPE_SELECT_1
);
441 OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR
, 1);
442 OUT_CS_RELOC(buf
, (query
->num_results
+ 1) * 4, 0, query
->domain
, 0);
443 OUT_CS_REG(RV530_FG_ZBREG_DEST
, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL
);
447 void r300_emit_query_end(struct r300_context
* r300
)
449 struct r300_capabilities
*caps
= &r300
->screen
->caps
;
450 struct r300_query
*query
= r300
->query_current
;
455 if (query
->begin_emitted
== FALSE
)
458 if (caps
->family
== CHIP_FAMILY_RV530
) {
459 if (caps
->num_z_pipes
== 2)
460 rv530_emit_query_end_double_z(r300
, query
);
462 rv530_emit_query_end_single_z(r300
, query
);
464 r300_emit_query_end_frag_pipes(r300
, query
);
466 query
->begin_emitted
= FALSE
;
467 query
->num_results
+= query
->num_pipes
;
469 /* XXX grab all the results and reset the counter. */
470 if (query
->num_results
>= query
->buffer_size
/ 4 - 4) {
471 query
->num_results
= (query
->buffer_size
/ 4) / 2;
472 fprintf(stderr
, "r300: Rewinding OQBO...\n");
476 void r300_emit_rs_state(struct r300_context
* r300
, unsigned size
, void* state
)
478 struct r300_rs_state
* rs
= (struct r300_rs_state
*)state
;
483 OUT_CS_REG(R300_VAP_CNTL_STATUS
, rs
->vap_control_status
);
485 OUT_CS_REG(R300_GB_AA_CONFIG
, rs
->antialiasing_config
);
487 OUT_CS_REG(R300_VAP_CNTL_STATUS
, rs
->vap_control_status
);
488 if (r300
->rws
->get_value(r300
->rws
, R300_VID_DRM_2_3_0
)) {
489 OUT_CS_REG_SEQ(R300_GB_MSPOS0
, 2);
490 OUT_CS(rs
->multisample_position_0
);
491 OUT_CS(rs
->multisample_position_1
);
493 OUT_CS_REG(R300_GB_AA_CONFIG
, rs
->antialiasing_config
);
494 OUT_CS_REG(R300_GA_POINT_SIZE
, rs
->point_size
);
495 OUT_CS_REG_SEQ(R300_GA_POINT_MINMAX
, 2);
496 OUT_CS(rs
->point_minmax
);
497 OUT_CS(rs
->line_control
);
499 if (rs
->polygon_offset_enable
) {
500 scale
= rs
->depth_scale
* 12;
501 offset
= rs
->depth_offset
;
503 switch (r300
->zbuffer_bpp
) {
512 OUT_CS_REG_SEQ(R300_SU_POLY_OFFSET_FRONT_SCALE
, 4);
519 OUT_CS_REG_SEQ(R300_SU_POLY_OFFSET_ENABLE
, 2);
520 OUT_CS(rs
->polygon_offset_enable
);
521 OUT_CS(rs
->cull_mode
);
522 OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG
, rs
->line_stipple_config
);
523 OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE
, rs
->line_stipple_value
);
524 OUT_CS_REG(R300_GA_POLY_MODE
, rs
->polygon_mode
);
525 OUT_CS_REG(R300_SC_CLIP_RULE
, rs
->clip_rule
);
526 OUT_CS_REG(R300_GB_ENABLE
, rs
->stuffing_enable
);
527 OUT_CS_REG_SEQ(R300_GA_POINT_S0
, 4);
528 OUT_CS_32F(rs
->point_texcoord_left
);
529 OUT_CS_32F(rs
->point_texcoord_bottom
);
530 OUT_CS_32F(rs
->point_texcoord_right
);
531 OUT_CS_32F(rs
->point_texcoord_top
);
535 void r300_emit_rs_block_state(struct r300_context
* r300
,
536 unsigned size
, void* state
)
538 struct r300_rs_block
* rs
= (struct r300_rs_block
*)state
;
540 /* It's the same for both INST and IP tables */
541 unsigned count
= (rs
->inst_count
& R300_RS_INST_COUNT_MASK
) + 1;
544 if (SCREEN_DBG_ON(r300
->screen
, DBG_DRAW
)) {
545 r500_dump_rs_block(rs
);
548 DBG(r300
, DBG_DRAW
, "r300: RS emit:\n");
551 OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL
, 2);
552 OUT_CS(rs
->vap_vtx_state_cntl
);
553 OUT_CS(rs
->vap_vsm_vtx_assm
);
554 OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0
, 2);
555 OUT_CS(rs
->vap_out_vtx_fmt
[0]);
556 OUT_CS(rs
->vap_out_vtx_fmt
[1]);
558 if (r300
->screen
->caps
.is_r500
) {
559 OUT_CS_REG_SEQ(R500_RS_IP_0
, count
);
561 OUT_CS_REG_SEQ(R300_RS_IP_0
, count
);
563 OUT_CS_TABLE(rs
->ip
, count
);
564 for (i
= 0; i
< count
; i
++) {
565 DBG(r300
, DBG_DRAW
, " : ip %d: 0x%08x\n", i
, rs
->ip
[i
]);
568 OUT_CS_REG_SEQ(R300_RS_COUNT
, 2);
570 OUT_CS(rs
->inst_count
);
572 if (r300
->screen
->caps
.is_r500
) {
573 OUT_CS_REG_SEQ(R500_RS_INST_0
, count
);
575 OUT_CS_REG_SEQ(R300_RS_INST_0
, count
);
577 OUT_CS_TABLE(rs
->inst
, count
);
578 for (i
= 0; i
< count
; i
++) {
579 DBG(r300
, DBG_DRAW
, " : inst %d: 0x%08x\n", i
, rs
->inst
[i
]);
582 DBG(r300
, DBG_DRAW
, " : count: 0x%08x inst_count: 0x%08x\n",
583 rs
->count
, rs
->inst_count
);
588 void r300_emit_scissor_state(struct r300_context
* r300
,
589 unsigned size
, void* state
)
591 struct pipe_scissor_state
* scissor
= (struct pipe_scissor_state
*)state
;
595 OUT_CS_REG_SEQ(R300_SC_CLIPRECT_TL_0
, 2);
596 if (r300
->screen
->caps
.is_r500
) {
597 OUT_CS((scissor
->minx
<< R300_CLIPRECT_X_SHIFT
) |
598 (scissor
->miny
<< R300_CLIPRECT_Y_SHIFT
));
599 OUT_CS(((scissor
->maxx
- 1) << R300_CLIPRECT_X_SHIFT
) |
600 ((scissor
->maxy
- 1) << R300_CLIPRECT_Y_SHIFT
));
602 OUT_CS(((scissor
->minx
+ 1440) << R300_CLIPRECT_X_SHIFT
) |
603 ((scissor
->miny
+ 1440) << R300_CLIPRECT_Y_SHIFT
));
604 OUT_CS(((scissor
->maxx
+ 1440-1) << R300_CLIPRECT_X_SHIFT
) |
605 ((scissor
->maxy
+ 1440-1) << R300_CLIPRECT_Y_SHIFT
));
610 void r300_emit_textures_state(struct r300_context
*r300
,
611 unsigned size
, void *state
)
613 struct r300_textures_state
*allstate
= (struct r300_textures_state
*)state
;
614 struct r300_texture_sampler_state
*texstate
;
615 struct r300_texture
*tex
;
620 OUT_CS_REG(R300_TX_ENABLE
, allstate
->tx_enable
);
622 for (i
= 0; i
< allstate
->count
; i
++) {
623 if ((1 << i
) & allstate
->tx_enable
) {
624 texstate
= &allstate
->regs
[i
];
625 tex
= r300_texture(allstate
->sampler_views
[i
]->base
.texture
);
627 OUT_CS_REG(R300_TX_FILTER0_0
+ (i
* 4), texstate
->filter0
);
628 OUT_CS_REG(R300_TX_FILTER1_0
+ (i
* 4), texstate
->filter1
);
629 OUT_CS_REG(R300_TX_BORDER_COLOR_0
+ (i
* 4),
630 texstate
->border_color
);
632 OUT_CS_REG(R300_TX_FORMAT0_0
+ (i
* 4), texstate
->format
.format0
);
633 OUT_CS_REG(R300_TX_FORMAT1_0
+ (i
* 4), texstate
->format
.format1
);
634 OUT_CS_REG(R300_TX_FORMAT2_0
+ (i
* 4), texstate
->format
.format2
);
636 OUT_CS_REG_SEQ(R300_TX_OFFSET_0
+ (i
* 4), 1);
637 OUT_CS_TEX_RELOC(tex
, texstate
->format
.tile_config
, tex
->domain
,
644 void r300_emit_aos(struct r300_context
* r300
, int offset
, boolean indexed
)
646 struct pipe_vertex_buffer
*vb1
, *vb2
, *vbuf
= r300
->vertex_buffer
;
647 struct pipe_vertex_element
*velem
= r300
->velems
->velem
;
648 struct r300_buffer
*buf
;
650 unsigned *hw_format_size
= r300
->velems
->hw_format_size
;
651 unsigned size1
, size2
, aos_count
= r300
->velems
->count
;
652 unsigned packet_size
= (aos_count
* 3 + 1) / 2;
655 BEGIN_CS(2 + packet_size
+ aos_count
* 2);
656 OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR
, packet_size
);
657 OUT_CS(aos_count
| (!indexed
? R300_VC_FORCE_PREFETCH
: 0));
659 for (i
= 0; i
< aos_count
- 1; i
+= 2) {
660 vb1
= &vbuf
[velem
[i
].vertex_buffer_index
];
661 vb2
= &vbuf
[velem
[i
+1].vertex_buffer_index
];
662 size1
= hw_format_size
[i
];
663 size2
= hw_format_size
[i
+1];
665 OUT_CS(R300_VBPNTR_SIZE0(size1
) | R300_VBPNTR_STRIDE0(vb1
->stride
) |
666 R300_VBPNTR_SIZE1(size2
) | R300_VBPNTR_STRIDE1(vb2
->stride
));
667 OUT_CS(vb1
->buffer_offset
+ velem
[i
].src_offset
+ offset
* vb1
->stride
);
668 OUT_CS(vb2
->buffer_offset
+ velem
[i
+1].src_offset
+ offset
* vb2
->stride
);
672 vb1
= &vbuf
[velem
[i
].vertex_buffer_index
];
673 size1
= hw_format_size
[i
];
675 OUT_CS(R300_VBPNTR_SIZE0(size1
) | R300_VBPNTR_STRIDE0(vb1
->stride
));
676 OUT_CS(vb1
->buffer_offset
+ velem
[i
].src_offset
+ offset
* vb1
->stride
);
679 for (i
= 0; i
< aos_count
; i
++) {
680 buf
= r300_buffer(vbuf
[velem
[i
].vertex_buffer_index
].buffer
);
681 OUT_CS_BUF_RELOC_NO_OFFSET(&buf
->b
.b
, buf
->domain
, 0, 0);
686 void r300_emit_aos_swtcl(struct r300_context
*r300
, boolean indexed
)
690 DBG(r300
, DBG_DRAW
, "r300: Preparing vertex buffer %p for render, "
691 "vertex size %d\n", r300
->vbo
,
692 r300
->vertex_info
.size
);
693 /* Set the pointer to our vertex buffer. The emitted values are this:
694 * PACKET3 [3D_LOAD_VBPNTR]
696 * FORMAT [size | stride << 8]
697 * OFFSET [offset into BO]
698 * VBPNTR [relocated BO]
701 OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR
, 3);
702 OUT_CS(1 | (!indexed
? R300_VC_FORCE_PREFETCH
: 0));
703 OUT_CS(r300
->vertex_info
.size
|
704 (r300
->vertex_info
.size
<< 8));
705 OUT_CS(r300
->vbo_offset
);
706 OUT_CS_BUF_RELOC(r300
->vbo
, 0, r300_buffer(r300
->vbo
)->domain
, 0, 0);
710 void r300_emit_vertex_stream_state(struct r300_context
* r300
,
711 unsigned size
, void* state
)
713 struct r300_vertex_stream_state
*streams
=
714 (struct r300_vertex_stream_state
*)state
;
718 DBG(r300
, DBG_DRAW
, "r300: PSC emit:\n");
721 OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0
, streams
->count
);
722 OUT_CS_TABLE(streams
->vap_prog_stream_cntl
, streams
->count
);
723 for (i
= 0; i
< streams
->count
; i
++) {
724 DBG(r300
, DBG_DRAW
, " : prog_stream_cntl%d: 0x%08x\n", i
,
725 streams
->vap_prog_stream_cntl
[i
]);
727 OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0
, streams
->count
);
728 OUT_CS_TABLE(streams
->vap_prog_stream_cntl_ext
, streams
->count
);
729 for (i
= 0; i
< streams
->count
; i
++) {
730 DBG(r300
, DBG_DRAW
, " : prog_stream_cntl_ext%d: 0x%08x\n", i
,
731 streams
->vap_prog_stream_cntl_ext
[i
]);
736 void r300_emit_pvs_flush(struct r300_context
* r300
, unsigned size
, void* state
)
741 OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG
, 0x0);
745 void r300_emit_vs_state(struct r300_context
* r300
, unsigned size
, void* state
)
747 struct r300_vertex_shader
* vs
= (struct r300_vertex_shader
*)state
;
748 struct r300_vertex_program_code
* code
= &vs
->code
;
749 struct r300_screen
* r300screen
= r300
->screen
;
750 unsigned instruction_count
= code
->length
/ 4;
753 unsigned vtx_mem_size
= r300screen
->caps
.is_r500
? 128 : 72;
754 unsigned input_count
= MAX2(util_bitcount(code
->InputsRead
), 1);
755 unsigned output_count
= MAX2(util_bitcount(code
->OutputsWritten
), 1);
756 unsigned temp_count
= MAX2(code
->num_temporaries
, 1);
758 unsigned pvs_num_slots
= MIN3(vtx_mem_size
/ input_count
,
759 vtx_mem_size
/ output_count
, 10);
760 unsigned pvs_num_controllers
= MIN2(vtx_mem_size
/ temp_count
, 6);
762 unsigned imm_first
= vs
->externals_count
;
763 unsigned imm_end
= vs
->code
.constants
.Count
;
764 unsigned imm_count
= vs
->immediates_count
;
769 /* R300_VAP_PVS_CODE_CNTL_0
770 * R300_VAP_PVS_CONST_CNTL
771 * R300_VAP_PVS_CODE_CNTL_1
772 * See the r5xx docs for instructions on how to use these. */
773 OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0
, 3);
774 OUT_CS(R300_PVS_FIRST_INST(0) |
775 R300_PVS_XYZW_VALID_INST(instruction_count
- 1) |
776 R300_PVS_LAST_INST(instruction_count
- 1));
777 OUT_CS(R300_PVS_MAX_CONST_ADDR(code
->constants
.Count
- 1));
778 OUT_CS(instruction_count
- 1);
780 OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG
, 0);
781 OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA
, code
->length
);
782 OUT_CS_TABLE(code
->body
.d
, code
->length
);
784 OUT_CS_REG(R300_VAP_CNTL
, R300_PVS_NUM_SLOTS(pvs_num_slots
) |
785 R300_PVS_NUM_CNTLRS(pvs_num_controllers
) |
786 R300_PVS_NUM_FPUS(r300screen
->caps
.num_vert_fpus
) |
787 R300_PVS_VF_MAX_VTX_NUM(12) |
788 (r300screen
->caps
.is_r500
? R500_TCL_STATE_OPTIMIZATION
: 0));
790 /* Emit immediates. */
792 OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG
,
793 (r300
->screen
->caps
.is_r500
?
794 R500_PVS_CONST_START
: R300_PVS_CONST_START
) +
796 OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA
, imm_count
* 4);
797 for (i
= imm_first
; i
< imm_end
; i
++) {
798 const float *data
= vs
->code
.constants
.Constants
[i
].u
.Immediate
;
799 OUT_CS_TABLE(data
, 4);
805 void r300_emit_vs_constants(struct r300_context
* r300
,
806 unsigned size
, void *state
)
809 ((struct r300_vertex_shader
*)r300
->vs_state
.state
)->externals_count
;
810 struct r300_constant_buffer
*buf
= (struct r300_constant_buffer
*)state
;
817 OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG
,
818 (r300
->screen
->caps
.is_r500
?
819 R500_PVS_CONST_START
: R300_PVS_CONST_START
));
820 OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA
, count
* 4);
821 OUT_CS_TABLE(buf
->constants
, count
* 4);
825 void r300_emit_viewport_state(struct r300_context
* r300
,
826 unsigned size
, void* state
)
828 struct r300_viewport_state
* viewport
= (struct r300_viewport_state
*)state
;
832 OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE
, 6);
833 OUT_CS_TABLE(&viewport
->xscale
, 6);
834 OUT_CS_REG(R300_VAP_VTE_CNTL
, viewport
->vte_control
);
838 void r300_emit_ztop_state(struct r300_context
* r300
,
839 unsigned size
, void* state
)
841 struct r300_ztop_state
* ztop
= (struct r300_ztop_state
*)state
;
845 OUT_CS_REG(R300_ZB_ZTOP
, ztop
->z_buffer_top
);
849 void r300_emit_texture_cache_inval(struct r300_context
* r300
, unsigned size
, void* state
)
854 OUT_CS_REG(R300_TX_INVALTAGS
, 0);
858 void r300_emit_buffer_validate(struct r300_context
*r300
,
859 boolean do_validate_vertex_buffers
,
860 struct pipe_resource
*index_buffer
)
862 struct pipe_framebuffer_state
* fb
=
863 (struct pipe_framebuffer_state
*)r300
->fb_state
.state
;
864 struct r300_textures_state
*texstate
=
865 (struct r300_textures_state
*)r300
->textures_state
.state
;
866 struct r300_texture
* tex
;
867 struct pipe_vertex_buffer
*vbuf
= r300
->vertex_buffer
;
868 struct pipe_vertex_element
*velem
= r300
->velems
->velem
;
869 struct pipe_resource
*pbuf
;
871 boolean invalid
= FALSE
;
873 /* upload buffers first */
874 if (r300
->screen
->caps
.has_tcl
&& r300
->any_user_vbs
) {
875 r300_upload_user_buffers(r300
);
876 r300
->any_user_vbs
= false;
880 r300
->rws
->reset_bos(r300
->rws
);
883 /* Color buffers... */
884 for (i
= 0; i
< fb
->nr_cbufs
; i
++) {
885 tex
= r300_texture(fb
->cbufs
[i
]->texture
);
886 assert(tex
&& tex
->buffer
&& "cbuf is marked, but NULL!");
887 if (!r300_add_texture(r300
->rws
, tex
, 0, tex
->domain
)) {
888 r300
->context
.flush(&r300
->context
, 0, NULL
);
892 /* ...depth buffer... */
894 tex
= r300_texture(fb
->zsbuf
->texture
);
895 assert(tex
&& tex
->buffer
&& "zsbuf is marked, but NULL!");
896 if (!r300_add_texture(r300
->rws
, tex
,
898 r300
->context
.flush(&r300
->context
, 0, NULL
);
903 for (i
= 0; i
< texstate
->count
; i
++) {
904 if (!(texstate
->tx_enable
& (1 << i
))) {
908 tex
= r300_texture(texstate
->sampler_views
[i
]->base
.texture
);
909 if (!r300_add_texture(r300
->rws
, tex
, tex
->domain
, 0)) {
910 r300
->context
.flush(&r300
->context
, 0, NULL
);
914 /* ...occlusion query buffer... */
915 if (r300
->query_current
) {
916 if (!r300
->rws
->add_buffer(r300
->rws
, r300
->query_current
->buffer
,
917 0, r300
->query_current
->domain
)) {
918 r300
->context
.flush(&r300
->context
, 0, NULL
);
922 /* ...vertex buffer for SWTCL path... */
924 if (!r300_add_buffer(r300
->rws
, r300
->vbo
,
925 r300_buffer(r300
->vbo
)->domain
, 0)) {
926 r300
->context
.flush(&r300
->context
, 0, NULL
);
930 /* ...vertex buffers for HWTCL path... */
931 if (do_validate_vertex_buffers
) {
932 for (i
= 0; i
< r300
->velems
->count
; i
++) {
933 pbuf
= vbuf
[velem
[i
].vertex_buffer_index
].buffer
;
935 if (!r300_add_buffer(r300
->rws
, pbuf
,
936 r300_buffer(pbuf
)->domain
, 0)) {
937 r300
->context
.flush(&r300
->context
, 0, NULL
);
942 /* ...and index buffer for HWTCL path. */
944 if (!r300_add_buffer(r300
->rws
, index_buffer
,
945 r300_buffer(index_buffer
)->domain
, 0)) {
946 r300
->context
.flush(&r300
->context
, 0, NULL
);
950 if (!r300
->rws
->validate(r300
->rws
)) {
951 r300
->context
.flush(&r300
->context
, 0, NULL
);
954 fprintf(stderr
, "r300: Stuck in validation loop, gonna quit now.\n");
962 unsigned r300_get_num_dirty_dwords(struct r300_context
*r300
)
964 struct r300_atom
* atom
;
967 foreach(atom
, &r300
->atom_list
) {
969 dwords
+= atom
->size
;
973 /* let's reserve some more, just in case */
979 /* Emit all dirty state. */
980 void r300_emit_dirty_state(struct r300_context
* r300
)
982 struct r300_atom
* atom
;
984 foreach(atom
, &r300
->atom_list
) {
986 atom
->emit(r300
, atom
->size
, atom
->state
);
987 if (SCREEN_DBG_ON(r300
->screen
, DBG_STATS
)) {