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
[] =
117 /* Once all the programs are done, we know how large urb entry
118 * sizes need to be and can decide if we need to change the urb
122 &brw_recalculate_urb_fence
,
127 &brw_vs_surfaces
, /* must do before unit */
128 &brw_wm_constant_surface
, /* must do before wm surfaces/bind bo */
129 &brw_wm_surfaces
, /* must do before samplers and unit */
135 &brw_vs_unit
, /* always required, enabled or not */
141 &brw_invarient_state
,
142 &brw_state_base_address
,
144 &brw_binding_table_pointers
,
145 &brw_blend_constant_color
,
149 &brw_polygon_stipple
,
150 &brw_polygon_stipple_offset
,
153 &brw_aa_line_parameters
,
166 void brw_init_state( struct brw_context
*brw
)
168 brw_init_caches(brw
);
172 void brw_destroy_state( struct brw_context
*brw
)
174 brw_destroy_caches(brw
);
175 brw_destroy_batch_cache(brw
);
178 /***********************************************************************
181 static GLboolean
check_state( const struct brw_state_flags
*a
,
182 const struct brw_state_flags
*b
)
184 return ((a
->mesa
& b
->mesa
) ||
186 (a
->cache
& b
->cache
));
189 static void accumulate_state( struct brw_state_flags
*a
,
190 const struct brw_state_flags
*b
)
194 a
->cache
|= b
->cache
;
198 static void xor_states( struct brw_state_flags
*result
,
199 const struct brw_state_flags
*a
,
200 const struct brw_state_flags
*b
)
202 result
->mesa
= a
->mesa
^ b
->mesa
;
203 result
->brw
= a
->brw
^ b
->brw
;
204 result
->cache
= a
->cache
^ b
->cache
;
208 brw_clear_validated_bos(struct brw_context
*brw
)
212 /* Clear the last round of validated bos */
213 for (i
= 0; i
< brw
->state
.validated_bo_count
; i
++) {
214 dri_bo_unreference(brw
->state
.validated_bos
[i
]);
215 brw
->state
.validated_bos
[i
] = NULL
;
217 brw
->state
.validated_bo_count
= 0;
220 struct dirty_bit_map
{
226 #define DEFINE_BIT(name) {name, #name, 0}
228 static struct dirty_bit_map mesa_bits
[] = {
229 DEFINE_BIT(_NEW_MODELVIEW
),
230 DEFINE_BIT(_NEW_PROJECTION
),
231 DEFINE_BIT(_NEW_TEXTURE_MATRIX
),
232 DEFINE_BIT(_NEW_COLOR_MATRIX
),
233 DEFINE_BIT(_NEW_ACCUM
),
234 DEFINE_BIT(_NEW_COLOR
),
235 DEFINE_BIT(_NEW_DEPTH
),
236 DEFINE_BIT(_NEW_EVAL
),
237 DEFINE_BIT(_NEW_FOG
),
238 DEFINE_BIT(_NEW_HINT
),
239 DEFINE_BIT(_NEW_LIGHT
),
240 DEFINE_BIT(_NEW_LINE
),
241 DEFINE_BIT(_NEW_PIXEL
),
242 DEFINE_BIT(_NEW_POINT
),
243 DEFINE_BIT(_NEW_POLYGON
),
244 DEFINE_BIT(_NEW_POLYGONSTIPPLE
),
245 DEFINE_BIT(_NEW_SCISSOR
),
246 DEFINE_BIT(_NEW_STENCIL
),
247 DEFINE_BIT(_NEW_TEXTURE
),
248 DEFINE_BIT(_NEW_TRANSFORM
),
249 DEFINE_BIT(_NEW_VIEWPORT
),
250 DEFINE_BIT(_NEW_PACKUNPACK
),
251 DEFINE_BIT(_NEW_ARRAY
),
252 DEFINE_BIT(_NEW_RENDERMODE
),
253 DEFINE_BIT(_NEW_BUFFERS
),
254 DEFINE_BIT(_NEW_MULTISAMPLE
),
255 DEFINE_BIT(_NEW_TRACK_MATRIX
),
256 DEFINE_BIT(_NEW_PROGRAM
),
257 DEFINE_BIT(_NEW_PROGRAM_CONSTANTS
),
261 static struct dirty_bit_map brw_bits
[] = {
262 DEFINE_BIT(BRW_NEW_URB_FENCE
),
263 DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM
),
264 DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM
),
265 DEFINE_BIT(BRW_NEW_INPUT_DIMENSIONS
),
266 DEFINE_BIT(BRW_NEW_CURBE_OFFSETS
),
267 DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE
),
268 DEFINE_BIT(BRW_NEW_PRIMITIVE
),
269 DEFINE_BIT(BRW_NEW_CONTEXT
),
270 DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS
),
271 DEFINE_BIT(BRW_NEW_PSP
),
272 DEFINE_BIT(BRW_NEW_INDICES
),
273 DEFINE_BIT(BRW_NEW_INDEX_BUFFER
),
274 DEFINE_BIT(BRW_NEW_VERTICES
),
275 DEFINE_BIT(BRW_NEW_BATCH
),
276 DEFINE_BIT(BRW_NEW_DEPTH_BUFFER
),
280 static struct dirty_bit_map cache_bits
[] = {
281 DEFINE_BIT(CACHE_NEW_CC_VP
),
282 DEFINE_BIT(CACHE_NEW_CC_UNIT
),
283 DEFINE_BIT(CACHE_NEW_WM_PROG
),
284 DEFINE_BIT(CACHE_NEW_SAMPLER_DEFAULT_COLOR
),
285 DEFINE_BIT(CACHE_NEW_SAMPLER
),
286 DEFINE_BIT(CACHE_NEW_WM_UNIT
),
287 DEFINE_BIT(CACHE_NEW_SF_PROG
),
288 DEFINE_BIT(CACHE_NEW_SF_VP
),
289 DEFINE_BIT(CACHE_NEW_SF_UNIT
),
290 DEFINE_BIT(CACHE_NEW_VS_UNIT
),
291 DEFINE_BIT(CACHE_NEW_VS_PROG
),
292 DEFINE_BIT(CACHE_NEW_GS_UNIT
),
293 DEFINE_BIT(CACHE_NEW_GS_PROG
),
294 DEFINE_BIT(CACHE_NEW_CLIP_VP
),
295 DEFINE_BIT(CACHE_NEW_CLIP_UNIT
),
296 DEFINE_BIT(CACHE_NEW_CLIP_PROG
),
297 DEFINE_BIT(CACHE_NEW_SURFACE
),
298 DEFINE_BIT(CACHE_NEW_SURF_BIND
),
304 brw_update_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
308 for (i
= 0; i
< 32; i
++) {
309 if (bit_map
[i
].bit
== 0)
312 if (bit_map
[i
].bit
& bits
)
318 brw_print_dirty_count(struct dirty_bit_map
*bit_map
, int32_t bits
)
322 for (i
= 0; i
< 32; i
++) {
323 if (bit_map
[i
].bit
== 0)
326 fprintf(stderr
, "0x%08x: %12d (%s)\n",
327 bit_map
[i
].bit
, bit_map
[i
].count
, bit_map
[i
].name
);
331 /***********************************************************************
334 void brw_validate_state( struct brw_context
*brw
)
336 GLcontext
*ctx
= &brw
->intel
.ctx
;
337 struct intel_context
*intel
= &brw
->intel
;
338 struct brw_state_flags
*state
= &brw
->state
.dirty
;
340 const struct brw_tracked_state
**atoms
;
343 brw_clear_validated_bos(brw
);
345 state
->mesa
|= brw
->intel
.NewGLState
;
346 brw
->intel
.NewGLState
= 0;
348 brw_add_validated_bo(brw
, intel
->batch
->buf
);
350 if (IS_GEN6(intel
->intelScreen
->deviceID
)) {
352 num_atoms
= ARRAY_SIZE(gen6_atoms
);
355 num_atoms
= ARRAY_SIZE(gen4_atoms
);
358 if (brw
->emit_state_always
) {
364 if (brw
->fragment_program
!= ctx
->FragmentProgram
._Current
) {
365 brw
->fragment_program
= ctx
->FragmentProgram
._Current
;
366 brw
->state
.dirty
.brw
|= BRW_NEW_FRAGMENT_PROGRAM
;
369 if (brw
->vertex_program
!= ctx
->VertexProgram
._Current
) {
370 brw
->vertex_program
= ctx
->VertexProgram
._Current
;
371 brw
->state
.dirty
.brw
|= BRW_NEW_VERTEX_PROGRAM
;
374 if (state
->mesa
== 0 &&
379 if (brw
->state
.dirty
.brw
& BRW_NEW_CONTEXT
)
380 brw_clear_batch_cache(brw
);
382 brw
->intel
.Fallback
= GL_FALSE
; /* boolean, not bitfield */
384 /* do prepare stage for all atoms */
385 for (i
= 0; i
< num_atoms
; i
++) {
386 const struct brw_tracked_state
*atom
= atoms
[i
];
388 if (brw
->intel
.Fallback
)
391 if (check_state(state
, &atom
->dirty
)) {
398 intel_check_front_buffer_rendering(intel
);
400 /* Make sure that the textures which are referenced by the current
401 * brw fragment program are actually present/valid.
402 * If this fails, we can experience GPU lock-ups.
405 const struct brw_fragment_program
*fp
;
406 fp
= brw_fragment_program_const(brw
->fragment_program
);
408 assert((fp
->tex_units_used
& ctx
->Texture
._EnabledUnits
)
409 == fp
->tex_units_used
);
415 void brw_upload_state(struct brw_context
*brw
)
417 struct intel_context
*intel
= &brw
->intel
;
418 struct brw_state_flags
*state
= &brw
->state
.dirty
;
420 static int dirty_count
= 0;
421 const struct brw_tracked_state
**atoms
;
424 if (IS_GEN6(intel
->intelScreen
->deviceID
)) {
426 num_atoms
= ARRAY_SIZE(gen6_atoms
);
429 num_atoms
= ARRAY_SIZE(gen4_atoms
);
432 brw_clear_validated_bos(brw
);
435 /* Debug version which enforces various sanity checks on the
436 * state flags which are generated and checked to help ensure
437 * state atoms are ordered correctly in the list.
439 struct brw_state_flags examined
, prev
;
440 memset(&examined
, 0, sizeof(examined
));
443 for (i
= 0; i
< num_atoms
; i
++) {
444 const struct brw_tracked_state
*atom
= atoms
[i
];
445 struct brw_state_flags generated
;
447 assert(atom
->dirty
.mesa
||
451 if (brw
->intel
.Fallback
)
454 if (check_state(state
, &atom
->dirty
)) {
460 accumulate_state(&examined
, &atom
->dirty
);
462 /* generated = (prev ^ state)
463 * if (examined & generated)
466 xor_states(&generated
, &prev
, state
);
467 assert(!check_state(&examined
, &generated
));
472 for (i
= 0; i
< num_atoms
; i
++) {
473 const struct brw_tracked_state
*atom
= atoms
[i
];
475 if (brw
->intel
.Fallback
)
478 if (check_state(state
, &atom
->dirty
)) {
486 if (INTEL_DEBUG
& DEBUG_STATE
) {
487 brw_update_dirty_count(mesa_bits
, state
->mesa
);
488 brw_update_dirty_count(brw_bits
, state
->brw
);
489 brw_update_dirty_count(cache_bits
, state
->cache
);
490 if (dirty_count
++ % 1000 == 0) {
491 brw_print_dirty_count(mesa_bits
, state
->mesa
);
492 brw_print_dirty_count(brw_bits
, state
->brw
);
493 brw_print_dirty_count(cache_bits
, state
->cache
);
494 fprintf(stderr
, "\n");
498 if (!brw
->intel
.Fallback
)
499 memset(state
, 0, sizeof(*state
));