2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
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:
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.
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.
26 **********************************************************************/
29 * Keith Whitwell <keith@tungstengraphics.com>
34 #include "brw_context.h"
35 #include "brw_state.h"
36 #include "intel_batchbuffer.h"
37 #include "intel_buffers.h"
38 #include "intel_chipset.h"
40 /* This is used to initialize brw->state.atoms[]. We could use this
41 * list directly except for a single atom, brw_constant_buffer, which
42 * has a .dirty value which changes according to the parameters of the
43 * current fragment and vertex programs, and so cannot be a static
46 static const struct brw_tracked_state
*gen4_atoms
[] =
57 /* Once all the programs are done, we know how large urb entry
58 * sizes need to be and can decide if we need to change the urb
62 &brw_recalculate_urb_fence
,
67 &brw_vs_surfaces
, /* must do before unit */
68 &brw_wm_constant_surface
, /* must do before wm surfaces/bind bo */
69 &brw_wm_surfaces
, /* must do before samplers and unit */
75 &brw_vs_unit
, /* always required, enabled or not */
82 &brw_state_base_address
,
84 &brw_binding_table_pointers
,
85 &brw_blend_constant_color
,
90 &brw_polygon_stipple_offset
,
93 &brw_aa_line_parameters
,
105 const struct brw_tracked_state
*gen6_atoms
[] =
115 /* Once all the programs are done, we know how large urb entry
116 * sizes need to be and can decide if we need to change the urb
124 &gen6_viewport_state
, /* must do after *_vp stages */
127 &gen6_blend_state
, /* must do before cc unit */
128 &gen6_color_calc_state
, /* must do before cc unit */
129 &gen6_depth_stencil_state
, /* must do before cc unit */
130 &gen6_cc_state_pointers
,
132 &brw_vs_surfaces
, /* must do before unit */
133 &brw_wm_constant_surface
, /* must do before wm surfaces/bind bo */
134 &brw_wm_surfaces
, /* must do before samplers and unit */
151 &brw_invarient_state
,
154 &brw_state_base_address
,
157 &brw_binding_table_pointers
,
158 &brw_blend_constant_color
,
164 &brw_polygon_stipple
,
165 &brw_polygon_stipple_offset
,
168 &brw_aa_line_parameters
,
180 void brw_init_state( struct brw_context
*brw
)
182 brw_init_caches(brw
);
186 void brw_destroy_state( struct brw_context
*brw
)
188 brw_destroy_caches(brw
);
189 brw_destroy_batch_cache(brw
);
192 /***********************************************************************
195 static GLboolean
check_state( const struct brw_state_flags
*a
,
196 const struct brw_state_flags
*b
)
198 return ((a
->mesa
& b
->mesa
) ||
200 (a
->cache
& b
->cache
));
203 static void accumulate_state( struct brw_state_flags
*a
,
204 const struct brw_state_flags
*b
)
208 a
->cache
|= b
->cache
;
212 static void xor_states( struct brw_state_flags
*result
,
213 const struct brw_state_flags
*a
,
214 const struct brw_state_flags
*b
)
216 result
->mesa
= a
->mesa
^ b
->mesa
;
217 result
->brw
= a
->brw
^ b
->brw
;
218 result
->cache
= a
->cache
^ b
->cache
;
222 brw_clear_validated_bos(struct brw_context
*brw
)
226 /* Clear the last round of validated bos */
227 for (i
= 0; i
< brw
->state
.validated_bo_count
; i
++) {
228 dri_bo_unreference(brw
->state
.validated_bos
[i
]);
229 brw
->state
.validated_bos
[i
] = NULL
;
231 brw
->state
.validated_bo_count
= 0;
234 struct dirty_bit_map
{
240 #define DEFINE_BIT(name) {name, #name, 0}
242 static struct dirty_bit_map mesa_bits
[] = {
243 DEFINE_BIT(_NEW_MODELVIEW
),
244 DEFINE_BIT(_NEW_PROJECTION
),
245 DEFINE_BIT(_NEW_TEXTURE_MATRIX
),
246 DEFINE_BIT(_NEW_COLOR_MATRIX
),
247 DEFINE_BIT(_NEW_ACCUM
),
248 DEFINE_BIT(_NEW_COLOR
),
249 DEFINE_BIT(_NEW_DEPTH
),
250 DEFINE_BIT(_NEW_EVAL
),
251 DEFINE_BIT(_NEW_FOG
),
252 DEFINE_BIT(_NEW_HINT
),
253 DEFINE_BIT(_NEW_LIGHT
),
254 DEFINE_BIT(_NEW_LINE
),
255 DEFINE_BIT(_NEW_PIXEL
),
256 DEFINE_BIT(_NEW_POINT
),
257 DEFINE_BIT(_NEW_POLYGON
),
258 DEFINE_BIT(_NEW_POLYGONSTIPPLE
),
259 DEFINE_BIT(_NEW_SCISSOR
),
260 DEFINE_BIT(_NEW_STENCIL
),
261 DEFINE_BIT(_NEW_TEXTURE
),
262 DEFINE_BIT(_NEW_TRANSFORM
),
263 DEFINE_BIT(_NEW_VIEWPORT
),
264 DEFINE_BIT(_NEW_PACKUNPACK
),
265 DEFINE_BIT(_NEW_ARRAY
),
266 DEFINE_BIT(_NEW_RENDERMODE
),
267 DEFINE_BIT(_NEW_BUFFERS
),
268 DEFINE_BIT(_NEW_MULTISAMPLE
),
269 DEFINE_BIT(_NEW_TRACK_MATRIX
),
270 DEFINE_BIT(_NEW_PROGRAM
),
271 DEFINE_BIT(_NEW_PROGRAM_CONSTANTS
),
275 static struct dirty_bit_map brw_bits
[] = {
276 DEFINE_BIT(BRW_NEW_URB_FENCE
),
277 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM
),
278 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM
),
279 DEFINE_BIT(BRW_NEW_INPUT_DIMENSIONS
),
280 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS
),
281 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE
),
282 DEFINE_BIT(BRW_NEW_PRIMITIVE
),
283 DEFINE_BIT(BRW_NEW_CONTEXT
),
284 DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS
),
285 DEFINE_BIT(BRW_NEW_PSP
),
286 DEFINE_BIT(BRW_NEW_INDICES
),
287 DEFINE_BIT(BRW_NEW_INDEX_BUFFER
),
288 DEFINE_BIT(BRW_NEW_VERTICES
),
289 DEFINE_BIT(BRW_NEW_BATCH
),
290 DEFINE_BIT(BRW_NEW_DEPTH_BUFFER
),
294 static struct dirty_bit_map cache_bits
[] = {
295 DEFINE_BIT(CACHE_NEW_BLEND_STATE
),
296 DEFINE_BIT(CACHE_NEW_CC_VP
),
297 DEFINE_BIT(CACHE_NEW_CC_UNIT
),
298 DEFINE_BIT(CACHE_NEW_WM_PROG
),
299 DEFINE_BIT(CACHE_NEW_SAMPLER_DEFAULT_COLOR
),
300 DEFINE_BIT(CACHE_NEW_SAMPLER
),
301 DEFINE_BIT(CACHE_NEW_WM_UNIT
),
302 DEFINE_BIT(CACHE_NEW_SF_PROG
),
303 DEFINE_BIT(CACHE_NEW_SF_VP
),
304 DEFINE_BIT(CACHE_NEW_SF_UNIT
),
305 DEFINE_BIT(CACHE_NEW_VS_UNIT
),
306 DEFINE_BIT(CACHE_NEW_VS_PROG
),
307 DEFINE_BIT(CACHE_NEW_GS_UNIT
),
308 DEFINE_BIT(CACHE_NEW_GS_PROG
),
309 DEFINE_BIT(CACHE_NEW_CLIP_VP
),
310 DEFINE_BIT(CACHE_NEW_CLIP_UNIT
),
311 DEFINE_BIT(CACHE_NEW_CLIP_PROG
),
312 DEFINE_BIT(CACHE_NEW_SURFACE
),
313 DEFINE_BIT(CACHE_NEW_SURF_BIND
),
319 brw_update_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
323 for (i
= 0; i
< 32; i
++) {
324 if (bit_map
[i
].bit
== 0)
327 if (bit_map
[i
].bit
& bits
)
333 brw_print_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
337 for (i
= 0; i
< 32; i
++) {
338 if (bit_map
[i
].bit
== 0)
341 fprintf(stderr
, "0x%08x: %12d (%s)\n",
342 bit_map
[i
].bit
, bit_map
[i
].count
, bit_map
[i
].name
);
346 /***********************************************************************
349 void brw_validate_state( struct brw_context
*brw
)
351 GLcontext
*ctx
= &brw
->intel
.ctx
;
352 struct intel_context
*intel
= &brw
->intel
;
353 struct brw_state_flags
*state
= &brw
->state
.dirty
;
355 const struct brw_tracked_state
**atoms
;
358 brw_clear_validated_bos(brw
);
360 state
->mesa
|= brw
->intel
.NewGLState
;
361 brw
->intel
.NewGLState
= 0;
363 brw_add_validated_bo(brw
, intel
->batch
->buf
);
365 if (IS_GEN6(intel
->intelScreen
->deviceID
)) {
367 num_atoms
= ARRAY_SIZE(gen6_atoms
);
370 num_atoms
= ARRAY_SIZE(gen4_atoms
);
373 if (brw
->emit_state_always
) {
379 if (brw
->fragment_program
!= ctx
->FragmentProgram
._Current
) {
380 brw
->fragment_program
= ctx
->FragmentProgram
._Current
;
381 brw
->state
.dirty
.brw
|= BRW_NEW_FRAGMENT_PROGRAM
;
384 if (brw
->vertex_program
!= ctx
->VertexProgram
._Current
) {
385 brw
->vertex_program
= ctx
->VertexProgram
._Current
;
386 brw
->state
.dirty
.brw
|= BRW_NEW_VERTEX_PROGRAM
;
389 if (state
->mesa
== 0 &&
394 if (brw
->state
.dirty
.brw
& BRW_NEW_CONTEXT
)
395 brw_clear_batch_cache(brw
);
397 brw
->intel
.Fallback
= GL_FALSE
; /* boolean, not bitfield */
399 /* do prepare stage for all atoms */
400 for (i
= 0; i
< num_atoms
; i
++) {
401 const struct brw_tracked_state
*atom
= atoms
[i
];
403 if (brw
->intel
.Fallback
)
406 if (check_state(state
, &atom
->dirty
)) {
413 intel_check_front_buffer_rendering(intel
);
415 /* Make sure that the textures which are referenced by the current
416 * brw fragment program are actually present/valid.
417 * If this fails, we can experience GPU lock-ups.
420 const struct brw_fragment_program
*fp
;
421 fp
= brw_fragment_program_const(brw
->fragment_program
);
423 assert((fp
->tex_units_used
& ctx
->Texture
._EnabledUnits
)
424 == fp
->tex_units_used
);
430 void brw_upload_state(struct brw_context
*brw
)
432 struct intel_context
*intel
= &brw
->intel
;
433 struct brw_state_flags
*state
= &brw
->state
.dirty
;
435 static int dirty_count
= 0;
436 const struct brw_tracked_state
**atoms
;
439 if (IS_GEN6(intel
->intelScreen
->deviceID
)) {
441 num_atoms
= ARRAY_SIZE(gen6_atoms
);
444 num_atoms
= ARRAY_SIZE(gen4_atoms
);
447 brw_clear_validated_bos(brw
);
450 /* Debug version which enforces various sanity checks on the
451 * state flags which are generated and checked to help ensure
452 * state atoms are ordered correctly in the list.
454 struct brw_state_flags examined
, prev
;
455 memset(&examined
, 0, sizeof(examined
));
458 for (i
= 0; i
< num_atoms
; i
++) {
459 const struct brw_tracked_state
*atom
= atoms
[i
];
460 struct brw_state_flags generated
;
462 assert(atom
->dirty
.mesa
||
466 if (brw
->intel
.Fallback
)
469 if (check_state(state
, &atom
->dirty
)) {
475 accumulate_state(&examined
, &atom
->dirty
);
477 /* generated = (prev ^ state)
478 * if (examined & generated)
481 xor_states(&generated
, &prev
, state
);
482 assert(!check_state(&examined
, &generated
));
487 for (i
= 0; i
< num_atoms
; i
++) {
488 const struct brw_tracked_state
*atom
= atoms
[i
];
490 if (brw
->intel
.Fallback
)
493 if (check_state(state
, &atom
->dirty
)) {
501 if (INTEL_DEBUG
& DEBUG_STATE
) {
502 brw_update_dirty_count(mesa_bits
, state
->mesa
);
503 brw_update_dirty_count(brw_bits
, state
->brw
);
504 brw_update_dirty_count(cache_bits
, state
->cache
);
505 if (dirty_count
++ % 1000 == 0) {
506 brw_print_dirty_count(mesa_bits
, state
->mesa
);
507 brw_print_dirty_count(brw_bits
, state
->brw
);
508 brw_print_dirty_count(cache_bits
, state
->cache
);
509 fprintf(stderr
, "\n");
513 if (!brw
->intel
.Fallback
)
514 memset(state
, 0, sizeof(*state
));