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 #include "draw/draw_context.h"
26 #include "util/u_math.h"
27 #include "util/u_memory.h"
29 #include "r300_context.h"
31 #include "r300_hyperz.h"
32 #include "r300_screen.h"
33 #include "r300_shader_semantics.h"
34 #include "r300_state.h"
35 #include "r300_state_derived.h"
36 #include "r300_state_inlines.h"
39 /* r300_state_derived: Various bits of state which are dependent upon
40 * currently bound CSO data. */
42 enum r300_rs_swizzle
{
49 static void r300_draw_emit_attrib(struct r300_context
* r300
,
50 enum attrib_emit emit
,
51 enum interp_mode interp
,
54 struct r300_vertex_shader
* vs
= r300
->vs_state
.state
;
55 struct tgsi_shader_info
* info
= &vs
->info
;
58 output
= draw_find_shader_output(r300
->draw
,
59 info
->output_semantic_name
[index
],
60 info
->output_semantic_index
[index
]);
61 draw_emit_vertex_attr(&r300
->vertex_info
, emit
, interp
, output
);
64 static void r300_draw_emit_all_attribs(struct r300_context
* r300
)
66 struct r300_vertex_shader
* vs
= r300
->vs_state
.state
;
67 struct r300_shader_semantics
* vs_outputs
= &vs
->outputs
;
71 if (vs_outputs
->pos
!= ATTR_UNUSED
) {
72 r300_draw_emit_attrib(r300
, EMIT_4F
, INTERP_PERSPECTIVE
,
79 if (vs_outputs
->psize
!= ATTR_UNUSED
) {
80 r300_draw_emit_attrib(r300
, EMIT_1F_PSIZE
, INTERP_POS
,
85 for (i
= 0; i
< ATTR_COLOR_COUNT
; i
++) {
86 if (vs_outputs
->color
[i
] != ATTR_UNUSED
) {
87 r300_draw_emit_attrib(r300
, EMIT_4F
, INTERP_LINEAR
,
88 vs_outputs
->color
[i
]);
92 /* XXX Back-face colors. */
94 /* Texture coordinates. */
95 /* Only 8 generic vertex attributes can be used. If there are more,
96 * they won't be rasterized. */
98 for (i
= 0; i
< ATTR_GENERIC_COUNT
&& gen_count
< 8; i
++) {
99 if (vs_outputs
->generic
[i
] != ATTR_UNUSED
) {
100 r300_draw_emit_attrib(r300
, EMIT_4F
, INTERP_PERSPECTIVE
,
101 vs_outputs
->generic
[i
]);
106 /* Fog coordinates. */
107 if (gen_count
< 8 && vs_outputs
->fog
!= ATTR_UNUSED
) {
108 r300_draw_emit_attrib(r300
, EMIT_4F
, INTERP_PERSPECTIVE
,
114 /* Update the PSC tables for SW TCL, using Draw. */
115 static void r300_swtcl_vertex_psc(struct r300_context
*r300
)
117 struct r300_vertex_stream_state
*vstream
= r300
->vertex_stream_state
.state
;
118 struct r300_vertex_shader
* vs
= r300
->vs_state
.state
;
119 struct vertex_info
* vinfo
= &r300
->vertex_info
;
120 uint16_t type
, swizzle
;
121 enum pipe_format format
;
122 unsigned i
, attrib_count
;
123 int* vs_output_tab
= vs
->stream_loc_notcl
;
126 memset(vstream
, 0, sizeof(struct r300_vertex_stream_state
));
128 /* For each Draw attribute, route it to the fragment shader according
129 * to the vs_output_tab. */
130 attrib_count
= vinfo
->num_attribs
;
131 DBG(r300
, DBG_DRAW
, "r300: attrib count: %d\n", attrib_count
);
132 for (i
= 0; i
< attrib_count
; i
++) {
133 DBG(r300
, DBG_DRAW
, "r300: attrib: offset %d, interp %d, size %d,"
134 " vs_output_tab %d\n", vinfo
->attrib
[i
].src_index
,
135 vinfo
->attrib
[i
].interp_mode
, vinfo
->attrib
[i
].emit
,
138 /* Make sure we have a proper destination for our attribute. */
139 assert(vs_output_tab
[i
] != -1);
141 format
= draw_translate_vinfo_format(vinfo
->attrib
[i
].emit
);
143 /* Obtain the type of data in this attribute. */
144 type
= r300_translate_vertex_data_type(format
) |
145 vs_output_tab
[i
] << R300_DST_VEC_LOC_SHIFT
;
147 /* Obtain the swizzle for this attribute. Note that the default
148 * swizzle in the hardware is not XYZW! */
149 swizzle
= r300_translate_vertex_data_swizzle(format
);
151 /* Add the attribute to the PSC table. */
153 vstream
->vap_prog_stream_cntl
[i
>> 1] |= type
<< 16;
154 vstream
->vap_prog_stream_cntl_ext
[i
>> 1] |= swizzle
<< 16;
156 vstream
->vap_prog_stream_cntl
[i
>> 1] |= type
;
157 vstream
->vap_prog_stream_cntl_ext
[i
>> 1] |= swizzle
;
161 /* Set the last vector in the PSC. */
165 vstream
->vap_prog_stream_cntl
[i
>> 1] |=
166 (R300_LAST_VEC
<< (i
& 1 ? 16 : 0));
168 vstream
->count
= (i
>> 1) + 1;
169 r300
->vertex_stream_state
.dirty
= TRUE
;
170 r300
->vertex_stream_state
.size
= (1 + vstream
->count
) * 2;
173 static void r300_rs_col(struct r300_rs_block
* rs
, int id
, int ptr
,
174 enum r300_rs_swizzle swiz
)
176 rs
->ip
[id
] |= R300_RS_COL_PTR(ptr
);
177 if (swiz
== SWIZ_0001
) {
178 rs
->ip
[id
] |= R300_RS_COL_FMT(R300_RS_COL_FMT_0001
);
180 rs
->ip
[id
] |= R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA
);
182 rs
->inst
[id
] |= R300_RS_INST_COL_ID(id
);
185 static void r300_rs_col_write(struct r300_rs_block
* rs
, int id
, int fp_offset
)
187 rs
->inst
[id
] |= R300_RS_INST_COL_CN_WRITE
|
188 R300_RS_INST_COL_ADDR(fp_offset
);
191 static void r300_rs_tex(struct r300_rs_block
* rs
, int id
, int ptr
,
192 enum r300_rs_swizzle swiz
)
194 if (swiz
== SWIZ_X001
) {
195 rs
->ip
[id
] |= R300_RS_TEX_PTR(ptr
*4) |
196 R300_RS_SEL_S(R300_RS_SEL_C0
) |
197 R300_RS_SEL_T(R300_RS_SEL_K0
) |
198 R300_RS_SEL_R(R300_RS_SEL_K0
) |
199 R300_RS_SEL_Q(R300_RS_SEL_K1
);
200 } else if (swiz
== SWIZ_XY01
) {
201 rs
->ip
[id
] |= R300_RS_TEX_PTR(ptr
*4) |
202 R300_RS_SEL_S(R300_RS_SEL_C0
) |
203 R300_RS_SEL_T(R300_RS_SEL_C1
) |
204 R300_RS_SEL_R(R300_RS_SEL_K0
) |
205 R300_RS_SEL_Q(R300_RS_SEL_K1
);
207 rs
->ip
[id
] |= R300_RS_TEX_PTR(ptr
*4) |
208 R300_RS_SEL_S(R300_RS_SEL_C0
) |
209 R300_RS_SEL_T(R300_RS_SEL_C1
) |
210 R300_RS_SEL_R(R300_RS_SEL_C2
) |
211 R300_RS_SEL_Q(R300_RS_SEL_C3
);
213 rs
->inst
[id
] |= R300_RS_INST_TEX_ID(id
);
216 static void r300_rs_tex_write(struct r300_rs_block
* rs
, int id
, int fp_offset
)
218 rs
->inst
[id
] |= R300_RS_INST_TEX_CN_WRITE
|
219 R300_RS_INST_TEX_ADDR(fp_offset
);
222 static void r500_rs_col(struct r300_rs_block
* rs
, int id
, int ptr
,
223 enum r300_rs_swizzle swiz
)
225 rs
->ip
[id
] |= R500_RS_COL_PTR(ptr
);
226 if (swiz
== SWIZ_0001
) {
227 rs
->ip
[id
] |= R500_RS_COL_FMT(R300_RS_COL_FMT_0001
);
229 rs
->ip
[id
] |= R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA
);
231 rs
->inst
[id
] |= R500_RS_INST_COL_ID(id
);
234 static void r500_rs_col_write(struct r300_rs_block
* rs
, int id
, int fp_offset
)
236 rs
->inst
[id
] |= R500_RS_INST_COL_CN_WRITE
|
237 R500_RS_INST_COL_ADDR(fp_offset
);
240 static void r500_rs_tex(struct r300_rs_block
* rs
, int id
, int ptr
,
241 enum r300_rs_swizzle swiz
)
243 int rs_tex_comp
= ptr
*4;
245 if (swiz
== SWIZ_X001
) {
246 rs
->ip
[id
] |= R500_RS_SEL_S(rs_tex_comp
) |
247 R500_RS_SEL_T(R500_RS_IP_PTR_K0
) |
248 R500_RS_SEL_R(R500_RS_IP_PTR_K0
) |
249 R500_RS_SEL_Q(R500_RS_IP_PTR_K1
);
250 } else if (swiz
== SWIZ_XY01
) {
251 rs
->ip
[id
] |= R500_RS_SEL_S(rs_tex_comp
) |
252 R500_RS_SEL_T(rs_tex_comp
+ 1) |
253 R500_RS_SEL_R(R500_RS_IP_PTR_K0
) |
254 R500_RS_SEL_Q(R500_RS_IP_PTR_K1
);
256 rs
->ip
[id
] |= R500_RS_SEL_S(rs_tex_comp
) |
257 R500_RS_SEL_T(rs_tex_comp
+ 1) |
258 R500_RS_SEL_R(rs_tex_comp
+ 2) |
259 R500_RS_SEL_Q(rs_tex_comp
+ 3);
261 rs
->inst
[id
] |= R500_RS_INST_TEX_ID(id
);
264 static void r500_rs_tex_write(struct r300_rs_block
* rs
, int id
, int fp_offset
)
266 rs
->inst
[id
] |= R500_RS_INST_TEX_CN_WRITE
|
267 R500_RS_INST_TEX_ADDR(fp_offset
);
270 /* Set up the RS block.
272 * This is the part of the chipset that actually does the rasterization
273 * of vertices into fragments. This is also the part of the chipset that
274 * locks up if any part of it is even slightly wrong. */
275 static void r300_update_rs_block(struct r300_context
*r300
)
277 struct r300_vertex_shader
*vs
= r300
->vs_state
.state
;
278 struct r300_shader_semantics
*vs_outputs
= &vs
->outputs
;
279 struct r300_shader_semantics
*fs_inputs
= &r300_fs(r300
)->shader
->inputs
;
280 struct r300_rs_block rs
= { { 0 } };
281 int i
, col_count
= 0, tex_count
= 0, fp_offset
= 0, count
;
282 void (*rX00_rs_col
)(struct r300_rs_block
*, int, int, enum r300_rs_swizzle
);
283 void (*rX00_rs_col_write
)(struct r300_rs_block
*, int, int);
284 void (*rX00_rs_tex
)(struct r300_rs_block
*, int, int, enum r300_rs_swizzle
);
285 void (*rX00_rs_tex_write
)(struct r300_rs_block
*, int, int);
286 boolean any_bcolor_used
= vs_outputs
->bcolor
[0] != ATTR_UNUSED
||
287 vs_outputs
->bcolor
[1] != ATTR_UNUSED
;
289 if (r300
->screen
->caps
.is_r500
) {
290 rX00_rs_col
= r500_rs_col
;
291 rX00_rs_col_write
= r500_rs_col_write
;
292 rX00_rs_tex
= r500_rs_tex
;
293 rX00_rs_tex_write
= r500_rs_tex_write
;
295 rX00_rs_col
= r300_rs_col
;
296 rX00_rs_col_write
= r300_rs_col_write
;
297 rX00_rs_tex
= r300_rs_tex
;
298 rX00_rs_tex_write
= r300_rs_tex_write
;
301 /* Rasterize colors. */
302 for (i
= 0; i
< ATTR_COLOR_COUNT
; i
++) {
303 if (vs_outputs
->color
[i
] != ATTR_UNUSED
|| any_bcolor_used
||
304 vs_outputs
->color
[1] != ATTR_UNUSED
) {
305 /* Always rasterize if it's written by the VS,
306 * otherwise it locks up. */
307 rX00_rs_col(&rs
, col_count
, col_count
, SWIZ_XYZW
);
309 /* Write it to the FS input register if it's used by the FS. */
310 if (fs_inputs
->color
[i
] != ATTR_UNUSED
) {
311 rX00_rs_col_write(&rs
, col_count
, fp_offset
);
315 "r300: Rasterized color %i written to FS.\n", i
);
317 DBG(r300
, DBG_RS
, "r300: Rasterized color %i unused.\n", i
);
321 /* Skip the FS input register, leave it uninitialized. */
322 /* If we try to set it to (0,0,0,1), it will lock up. */
323 if (fs_inputs
->color
[i
] != ATTR_UNUSED
) {
326 DBG(r300
, DBG_RS
, "r300: FS input color %i unassigned%s.\n",
332 /* Rasterize texture coordinates. */
333 for (i
= 0; i
< ATTR_GENERIC_COUNT
; i
++) {
334 bool sprite_coord
= !!(r300
->sprite_coord_enable
& (1 << i
));
336 if (vs_outputs
->generic
[i
] != ATTR_UNUSED
|| sprite_coord
) {
337 /* Always rasterize if it's written by the VS,
338 * otherwise it locks up. */
339 rX00_rs_tex(&rs
, tex_count
, tex_count
,
340 sprite_coord
? SWIZ_XY01
: SWIZ_XYZW
);
342 /* Write it to the FS input register if it's used by the FS. */
343 if (fs_inputs
->generic
[i
] != ATTR_UNUSED
) {
344 rX00_rs_tex_write(&rs
, tex_count
, fp_offset
);
348 "r300: Rasterized generic %i written to FS%s.\n",
349 i
, sprite_coord
? " (sprite coord)" : "");
352 "r300: Rasterized generic %i unused%s.\n",
353 i
, sprite_coord
? " (sprite coord)" : "");
357 /* Skip the FS input register, leave it uninitialized. */
358 /* If we try to set it to (0,0,0,1), it will lock up. */
359 if (fs_inputs
->generic
[i
] != ATTR_UNUSED
) {
362 DBG(r300
, DBG_RS
, "r300: FS input generic %i unassigned%s.\n",
363 i
, sprite_coord
? " (sprite coord)" : "");
368 /* Rasterize fog coordinates. */
369 if (vs_outputs
->fog
!= ATTR_UNUSED
) {
370 /* Always rasterize if it's written by the VS,
371 * otherwise it locks up. */
372 rX00_rs_tex(&rs
, tex_count
, tex_count
, SWIZ_X001
);
374 /* Write it to the FS input register if it's used by the FS. */
375 if (fs_inputs
->fog
!= ATTR_UNUSED
) {
376 rX00_rs_tex_write(&rs
, tex_count
, fp_offset
);
379 DBG(r300
, DBG_RS
, "r300: Rasterized fog written to FS.\n");
381 DBG(r300
, DBG_RS
, "r300: Rasterized fog unused.\n");
385 /* Skip the FS input register, leave it uninitialized. */
386 /* If we try to set it to (0,0,0,1), it will lock up. */
387 if (fs_inputs
->fog
!= ATTR_UNUSED
) {
390 DBG(r300
, DBG_RS
, "r300: FS input fog unassigned.\n");
394 /* Rasterize WPOS. */
395 /* If the FS doesn't need it, it's not written by the VS. */
396 if (vs_outputs
->wpos
!= ATTR_UNUSED
&& fs_inputs
->wpos
!= ATTR_UNUSED
) {
397 rX00_rs_tex(&rs
, tex_count
, tex_count
, SWIZ_XYZW
);
398 rX00_rs_tex_write(&rs
, tex_count
, fp_offset
);
400 DBG(r300
, DBG_RS
, "r300: Rasterized WPOS written to FS.\n");
406 /* Rasterize at least one color, or bad things happen. */
407 if (col_count
== 0 && tex_count
== 0) {
408 rX00_rs_col(&rs
, 0, 0, SWIZ_0001
);
411 DBG(r300
, DBG_RS
, "r300: Rasterized color 0 to prevent lockups.\n");
414 DBG(r300
, DBG_RS
, "r300: --- Rasterizer status ---: colors: %i, "
415 "generics: %i.\n", col_count
, tex_count
);
417 rs
.count
= (tex_count
*4) | (col_count
<< R300_IC_COUNT_SHIFT
) |
420 count
= MAX3(col_count
, tex_count
, 1);
421 rs
.inst_count
= count
- 1;
423 /* Now, after all that, see if we actually need to update the state. */
424 if (memcmp(r300
->rs_block_state
.state
, &rs
, sizeof(struct r300_rs_block
))) {
425 memcpy(r300
->rs_block_state
.state
, &rs
, sizeof(struct r300_rs_block
));
426 r300
->rs_block_state
.size
= 5 + count
*2;
430 static void r300_merge_textures_and_samplers(struct r300_context
* r300
)
432 struct r300_textures_state
*state
=
433 (struct r300_textures_state
*)r300
->textures_state
.state
;
434 struct r300_texture_sampler_state
*texstate
;
435 struct r300_sampler_state
*sampler
;
436 struct r300_sampler_view
*view
;
437 struct r300_texture
*tex
;
438 unsigned min_level
, max_level
, i
, size
;
439 unsigned count
= MIN2(state
->sampler_view_count
,
440 state
->sampler_state_count
);
442 state
->tx_enable
= 0;
446 for (i
= 0; i
< count
; i
++) {
447 if (state
->sampler_views
[i
] && state
->sampler_states
[i
]) {
448 state
->tx_enable
|= 1 << i
;
450 view
= state
->sampler_views
[i
];
451 tex
= r300_texture(view
->base
.texture
);
452 sampler
= state
->sampler_states
[i
];
454 texstate
= &state
->regs
[i
];
455 texstate
->format
= view
->format
;
456 texstate
->filter0
= sampler
->filter0
;
457 texstate
->filter1
= sampler
->filter1
;
458 texstate
->border_color
= sampler
->border_color
;
460 /* to emulate 1D textures through 2D ones correctly */
461 if (tex
->b
.b
.target
== PIPE_TEXTURE_1D
) {
462 texstate
->filter0
&= ~R300_TX_WRAP_T_MASK
;
463 texstate
->filter0
|= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE
);
466 if (tex
->uses_pitch
) {
467 /* NPOT textures don't support mip filter, unfortunately.
468 * This prevents incorrect rendering. */
469 texstate
->filter0
&= ~R300_TX_MIN_FILTER_MIP_MASK
;
471 /* Mask out the mirrored flag. */
472 if (texstate
->filter0
& R300_TX_WRAP_S(R300_TX_MIRRORED
)) {
473 texstate
->filter0
&= ~R300_TX_WRAP_S(R300_TX_MIRRORED
);
475 if (texstate
->filter0
& R300_TX_WRAP_T(R300_TX_MIRRORED
)) {
476 texstate
->filter0
&= ~R300_TX_WRAP_T(R300_TX_MIRRORED
);
479 /* Change repeat to clamp-to-edge.
480 * (the repeat bit has a value of 0, no masking needed). */
481 if ((texstate
->filter0
& R300_TX_WRAP_S_MASK
) ==
482 R300_TX_WRAP_S(R300_TX_REPEAT
)) {
483 texstate
->filter0
|= R300_TX_WRAP_S(R300_TX_CLAMP_TO_EDGE
);
485 if ((texstate
->filter0
& R300_TX_WRAP_T_MASK
) ==
486 R300_TX_WRAP_T(R300_TX_REPEAT
)) {
487 texstate
->filter0
|= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE
);
490 /* determine min/max levels */
491 /* the MAX_MIP level is the largest (finest) one */
492 max_level
= MIN3(sampler
->max_lod
+ view
->base
.first_level
,
493 tex
->b
.b
.last_level
, view
->base
.last_level
);
494 min_level
= MIN2(sampler
->min_lod
+ view
->base
.first_level
,
496 texstate
->format
.format0
|= R300_TX_NUM_LEVELS(max_level
);
497 texstate
->filter0
|= R300_TX_MAX_MIP_LEVEL(min_level
);
500 texstate
->filter0
|= i
<< 28;
507 r300
->textures_state
.size
= size
;
509 /* Pick a fragment shader based on either the texture compare state
510 * or the uses_pitch flag. */
511 if (r300
->fs
.state
&& count
) {
512 if (r300_pick_fragment_shader(r300
)) {
513 r300_mark_fs_code_dirty(r300
);
518 void r300_update_derived_state(struct r300_context
* r300
)
520 if (r300
->textures_state
.dirty
) {
521 r300_merge_textures_and_samplers(r300
);
524 if (r300
->rs_block_state
.dirty
) {
525 r300_update_rs_block(r300
);
529 memset(&r300
->vertex_info
, 0, sizeof(struct vertex_info
));
530 r300_draw_emit_all_attribs(r300
);
531 draw_compute_vertex_size(&r300
->vertex_info
);
532 r300_swtcl_vertex_psc(r300
);
535 r300_update_hyperz_state(r300
);