2 * Mesa 3-D graphics library
4 * Copyright (C) 2013 LunarG, Inc.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
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 NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
25 * Chia-I Wu <olv@lunarg.com>
28 #include "util/u_dual_blend.h"
29 #include "intel_reg.h"
31 #include "ilo_common.h"
32 #include "ilo_context.h"
34 #include "ilo_gpe_gen7.h"
35 #include "ilo_shader.h"
36 #include "ilo_state.h"
37 #include "ilo_3d_pipeline.h"
38 #include "ilo_3d_pipeline_gen6.h"
39 #include "ilo_3d_pipeline_gen7.h"
42 gen7_wa_pipe_control_cs_stall(struct ilo_3d_pipeline
*p
,
43 bool change_multisample_state
,
44 bool change_depth_state
)
46 struct intel_bo
*bo
= NULL
;
47 uint32_t dw1
= PIPE_CONTROL_CS_STALL
;
49 assert(p
->dev
->gen
== ILO_GEN(7));
52 if (p
->state
.has_gen6_wa_pipe_control
)
54 p
->state
.has_gen6_wa_pipe_control
= true;
57 * From the Ivy Bridge PRM, volume 2 part 1, page 258:
59 * "Due to an HW issue driver needs to send a pipe control with stall
60 * when ever there is state change in depth bias related state"
62 * From the Ivy Bridge PRM, volume 2 part 1, page 292:
64 * "A PIPE_CONTOL command with the CS Stall bit set must be programmed
65 * in the ring after this instruction
66 * (3DSTATE_PUSH_CONSTANT_ALLOC_PS)."
68 * From the Ivy Bridge PRM, volume 2 part 1, page 304:
70 * "Driver must ierarchi that all the caches in the depth pipe are
71 * flushed before this command (3DSTATE_MULTISAMPLE) is parsed. This
72 * requires driver to send a PIPE_CONTROL with a CS stall along with a
73 * Depth Flush prior to this command.
75 * From the Ivy Bridge PRM, volume 2 part 1, page 315:
77 * "Driver must send a least one PIPE_CONTROL command with CS Stall and
78 * a post sync operation prior to the group of depth
79 * commands(3DSTATE_DEPTH_BUFFER, 3DSTATE_CLEAR_PARAMS,
80 * 3DSTATE_STENCIL_BUFFER, and 3DSTATE_HIER_DEPTH_BUFFER)."
83 if (change_multisample_state
)
84 dw1
|= PIPE_CONTROL_DEPTH_CACHE_FLUSH
;
86 if (change_depth_state
) {
87 dw1
|= PIPE_CONTROL_WRITE_IMMEDIATE
;
88 bo
= p
->workaround_bo
;
91 gen6_emit_PIPE_CONTROL(p
->dev
, dw1
, bo
, 0, false, p
->cp
);
95 gen7_wa_pipe_control_vs_depth_stall(struct ilo_3d_pipeline
*p
)
97 assert(p
->dev
->gen
== ILO_GEN(7));
100 * From the Ivy Bridge PRM, volume 2 part 1, page 106:
102 * "A PIPE_CONTROL with Post-Sync Operation set to 1h and a depth stall
103 * needs to be sent just prior to any 3DSTATE_VS, 3DSTATE_URB_VS,
104 * 3DSTATE_CONSTANT_VS, 3DSTATE_BINDING_TABLE_POINTER_VS,
105 * 3DSTATE_SAMPLER_STATE_POINTER_VS command. Only one PIPE_CONTROL
106 * needs to be sent before any combination of VS associated 3DSTATE."
108 gen6_emit_PIPE_CONTROL(p
->dev
,
109 PIPE_CONTROL_DEPTH_STALL
|
110 PIPE_CONTROL_WRITE_IMMEDIATE
,
111 p
->workaround_bo
, 0, false, p
->cp
);
115 gen7_wa_pipe_control_wm_depth_stall(struct ilo_3d_pipeline
*p
,
116 bool change_depth_buffer
)
118 assert(p
->dev
->gen
== ILO_GEN(7));
121 * From the Ivy Bridge PRM, volume 2 part 1, page 276:
123 * "The driver must make sure a PIPE_CONTROL with the Depth Stall
124 * Enable bit set after all the following states are programmed:
127 * * 3DSTATE_VIEWPORT_STATE_POINTERS_CC
128 * * 3DSTATE_CONSTANT_PS
129 * * 3DSTATE_BINDING_TABLE_POINTERS_PS
130 * * 3DSTATE_SAMPLER_STATE_POINTERS_PS
131 * * 3DSTATE_CC_STATE_POINTERS
132 * * 3DSTATE_BLEND_STATE_POINTERS
133 * * 3DSTATE_DEPTH_STENCIL_STATE_POINTERS"
135 * From the Ivy Bridge PRM, volume 2 part 1, page 315:
137 * "Restriction: Prior to changing Depth/Stencil Buffer state (i.e.,
138 * any combination of 3DSTATE_DEPTH_BUFFER, 3DSTATE_CLEAR_PARAMS,
139 * 3DSTATE_STENCIL_BUFFER, 3DSTATE_HIER_DEPTH_BUFFER) SW must first
140 * issue a pipelined depth stall (PIPE_CONTROL with Depth Stall bit
141 * set), followed by a pipelined depth cache flush (PIPE_CONTROL with
142 * Depth Flush Bit set, followed by another pipelined depth stall
143 * (PIPE_CONTROL with Depth Stall Bit set), unless SW can otherwise
144 * guarantee that the pipeline from WM onwards is already flushed
145 * (e.g., via a preceding MI_FLUSH)."
147 gen6_emit_PIPE_CONTROL(p
->dev
,
148 PIPE_CONTROL_DEPTH_STALL
,
149 NULL
, 0, false, p
->cp
);
151 if (!change_depth_buffer
)
154 gen6_emit_PIPE_CONTROL(p
->dev
,
155 PIPE_CONTROL_DEPTH_CACHE_FLUSH
,
156 NULL
, 0, false, p
->cp
);
158 gen6_emit_PIPE_CONTROL(p
->dev
,
159 PIPE_CONTROL_DEPTH_STALL
,
160 NULL
, 0, false, p
->cp
);
164 gen7_wa_pipe_control_wm_max_threads_stall(struct ilo_3d_pipeline
*p
)
166 assert(p
->dev
->gen
== ILO_GEN(7));
169 * From the Ivy Bridge PRM, volume 2 part 1, page 286:
171 * "If this field (Maximum Number of Threads in 3DSTATE_WM) is changed
172 * between 3DPRIMITIVE commands, a PIPE_CONTROL command with Stall at
173 * Pixel Scoreboard set is required to be issued."
175 gen6_emit_PIPE_CONTROL(p
->dev
,
176 PIPE_CONTROL_STALL_AT_SCOREBOARD
,
177 NULL
, 0, false, p
->cp
);
181 #define DIRTY(state) (session->pipe_dirty & ILO_DIRTY_ ## state)
184 gen7_pipeline_common_urb(struct ilo_3d_pipeline
*p
,
185 const struct ilo_context
*ilo
,
186 struct gen6_pipeline_session
*session
)
188 /* 3DSTATE_URB_{VS,GS,HS,DS} */
189 if (DIRTY(VE
) || DIRTY(VS
)) {
190 /* the first 16KB are reserved for VS and PS PCBs */
191 const int offset
= 16 * 1024;
192 int vs_entry_size
, vs_total_size
;
194 vs_entry_size
= (ilo
->vs
) ?
195 ilo_shader_get_kernel_param(ilo
->vs
, ILO_KERNEL_OUTPUT_COUNT
) : 0;
198 * From the Ivy Bridge PRM, volume 2 part 1, page 35:
200 * "Programming Restriction: As the VS URB entry serves as both the
201 * per-vertex input and output of the VS shader, the VS URB
202 * Allocation Size must be sized to the maximum of the vertex input
203 * and output structures."
205 if (vs_entry_size
< ilo
->ve
->count
)
206 vs_entry_size
= ilo
->ve
->count
;
208 vs_entry_size
*= sizeof(float) * 4;
209 vs_total_size
= ilo
->dev
->urb_size
- offset
;
211 gen7_wa_pipe_control_vs_depth_stall(p
);
213 gen7_emit_3DSTATE_URB_VS(p
->dev
,
214 offset
, vs_total_size
, vs_entry_size
, p
->cp
);
216 gen7_emit_3DSTATE_URB_GS(p
->dev
, offset
, 0, 0, p
->cp
);
217 gen7_emit_3DSTATE_URB_HS(p
->dev
, offset
, 0, 0, p
->cp
);
218 gen7_emit_3DSTATE_URB_DS(p
->dev
, offset
, 0, 0, p
->cp
);
223 gen7_pipeline_common_pcb_alloc(struct ilo_3d_pipeline
*p
,
224 const struct ilo_context
*ilo
,
225 struct gen6_pipeline_session
*session
)
227 /* 3DSTATE_PUSH_CONSTANT_ALLOC_{VS,PS} */
228 if (session
->hw_ctx_changed
) {
230 * push constant buffers are only allowed to take up at most the first
233 gen7_emit_3DSTATE_PUSH_CONSTANT_ALLOC_VS(p
->dev
,
236 gen7_emit_3DSTATE_PUSH_CONSTANT_ALLOC_PS(p
->dev
,
239 gen7_wa_pipe_control_cs_stall(p
, true, true);
244 gen7_pipeline_common_pointers_1(struct ilo_3d_pipeline
*p
,
245 const struct ilo_context
*ilo
,
246 struct gen6_pipeline_session
*session
)
248 /* 3DSTATE_VIEWPORT_STATE_POINTERS_{CC,SF_CLIP} */
249 if (session
->viewport_state_changed
) {
250 gen7_emit_3DSTATE_VIEWPORT_STATE_POINTERS_CC(p
->dev
,
251 p
->state
.CC_VIEWPORT
, p
->cp
);
253 gen7_emit_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP(p
->dev
,
254 p
->state
.SF_CLIP_VIEWPORT
, p
->cp
);
259 gen7_pipeline_common_pointers_2(struct ilo_3d_pipeline
*p
,
260 const struct ilo_context
*ilo
,
261 struct gen6_pipeline_session
*session
)
263 /* 3DSTATE_BLEND_STATE_POINTERS */
264 if (session
->cc_state_blend_changed
) {
265 gen7_emit_3DSTATE_BLEND_STATE_POINTERS(p
->dev
,
266 p
->state
.BLEND_STATE
, p
->cp
);
269 /* 3DSTATE_CC_STATE_POINTERS */
270 if (session
->cc_state_cc_changed
) {
271 gen7_emit_3DSTATE_CC_STATE_POINTERS(p
->dev
,
272 p
->state
.COLOR_CALC_STATE
, p
->cp
);
275 /* 3DSTATE_DEPTH_STENCIL_STATE_POINTERS */
276 if (session
->cc_state_dsa_changed
) {
277 gen7_emit_3DSTATE_DEPTH_STENCIL_STATE_POINTERS(p
->dev
,
278 p
->state
.DEPTH_STENCIL_STATE
, p
->cp
);
283 gen7_pipeline_vs(struct ilo_3d_pipeline
*p
,
284 const struct ilo_context
*ilo
,
285 struct gen6_pipeline_session
*session
)
287 const bool emit_3dstate_binding_table
= session
->binding_table_vs_changed
;
288 const bool emit_3dstate_sampler_state
= session
->sampler_state_vs_changed
;
289 /* see gen6_pipeline_vs() */
290 const bool emit_3dstate_constant_vs
= session
->pcb_state_vs_changed
;
291 const bool emit_3dstate_vs
= (DIRTY(VS
) || DIRTY(SAMPLER_VS
) ||
292 session
->kernel_bo_changed
);
294 /* emit depth stall before any of the VS commands */
295 if (emit_3dstate_binding_table
|| emit_3dstate_sampler_state
||
296 emit_3dstate_constant_vs
|| emit_3dstate_vs
)
297 gen7_wa_pipe_control_vs_depth_stall(p
);
299 /* 3DSTATE_BINDING_TABLE_POINTERS_VS */
300 if (emit_3dstate_binding_table
) {
301 gen7_emit_3DSTATE_BINDING_TABLE_POINTERS_VS(p
->dev
,
302 p
->state
.vs
.BINDING_TABLE_STATE
, p
->cp
);
305 /* 3DSTATE_SAMPLER_STATE_POINTERS_VS */
306 if (emit_3dstate_sampler_state
) {
307 gen7_emit_3DSTATE_SAMPLER_STATE_POINTERS_VS(p
->dev
,
308 p
->state
.vs
.SAMPLER_STATE
, p
->cp
);
311 /* 3DSTATE_CONSTANT_VS */
312 if (emit_3dstate_constant_vs
) {
313 gen7_emit_3DSTATE_CONSTANT_VS(p
->dev
,
314 &p
->state
.vs
.PUSH_CONSTANT_BUFFER
,
315 &p
->state
.vs
.PUSH_CONSTANT_BUFFER_size
,
320 if (emit_3dstate_vs
) {
321 const int num_samplers
= ilo
->sampler
[PIPE_SHADER_VERTEX
].count
;
323 gen6_emit_3DSTATE_VS(p
->dev
, ilo
->vs
, num_samplers
, p
->cp
);
328 gen7_pipeline_hs(struct ilo_3d_pipeline
*p
,
329 const struct ilo_context
*ilo
,
330 struct gen6_pipeline_session
*session
)
332 /* 3DSTATE_CONSTANT_HS and 3DSTATE_HS */
333 if (session
->hw_ctx_changed
) {
334 gen7_emit_3DSTATE_CONSTANT_HS(p
->dev
, 0, 0, 0, p
->cp
);
335 gen7_emit_3DSTATE_HS(p
->dev
, NULL
, 0, p
->cp
);
338 /* 3DSTATE_BINDING_TABLE_POINTERS_HS */
339 if (session
->hw_ctx_changed
)
340 gen7_emit_3DSTATE_BINDING_TABLE_POINTERS_HS(p
->dev
, 0, p
->cp
);
344 gen7_pipeline_te(struct ilo_3d_pipeline
*p
,
345 const struct ilo_context
*ilo
,
346 struct gen6_pipeline_session
*session
)
349 if (session
->hw_ctx_changed
)
350 gen7_emit_3DSTATE_TE(p
->dev
, p
->cp
);
354 gen7_pipeline_ds(struct ilo_3d_pipeline
*p
,
355 const struct ilo_context
*ilo
,
356 struct gen6_pipeline_session
*session
)
358 /* 3DSTATE_CONSTANT_DS and 3DSTATE_DS */
359 if (session
->hw_ctx_changed
) {
360 gen7_emit_3DSTATE_CONSTANT_DS(p
->dev
, 0, 0, 0, p
->cp
);
361 gen7_emit_3DSTATE_DS(p
->dev
, NULL
, 0, p
->cp
);
364 /* 3DSTATE_BINDING_TABLE_POINTERS_DS */
365 if (session
->hw_ctx_changed
)
366 gen7_emit_3DSTATE_BINDING_TABLE_POINTERS_DS(p
->dev
, 0, p
->cp
);
371 gen7_pipeline_gs(struct ilo_3d_pipeline
*p
,
372 const struct ilo_context
*ilo
,
373 struct gen6_pipeline_session
*session
)
375 /* 3DSTATE_CONSTANT_GS and 3DSTATE_GS */
376 if (session
->hw_ctx_changed
) {
377 gen7_emit_3DSTATE_CONSTANT_GS(p
->dev
, 0, 0, 0, p
->cp
);
378 gen7_emit_3DSTATE_GS(p
->dev
, NULL
, 0, p
->cp
);
381 /* 3DSTATE_BINDING_TABLE_POINTERS_GS */
382 if (session
->binding_table_gs_changed
) {
383 gen7_emit_3DSTATE_BINDING_TABLE_POINTERS_GS(p
->dev
,
384 p
->state
.gs
.BINDING_TABLE_STATE
, p
->cp
);
389 gen7_pipeline_sol(struct ilo_3d_pipeline
*p
,
390 const struct ilo_context
*ilo
,
391 struct gen6_pipeline_session
*session
)
393 const struct pipe_stream_output_info
*so_info
;
394 const struct ilo_shader_state
*shader
;
395 bool dirty_sh
= false;
399 dirty_sh
= DIRTY(GS
);
403 dirty_sh
= DIRTY(VS
);
406 so_info
= ilo_shader_get_kernel_so_info(shader
);
408 gen6_pipeline_update_max_svbi(p
, ilo
, session
);
410 /* 3DSTATE_SO_BUFFER */
411 if ((DIRTY(SO
) || dirty_sh
|| session
->batch_bo_changed
) &&
415 for (i
= 0; i
< ilo
->so
.count
; i
++) {
416 const int stride
= so_info
->stride
[i
] * 4; /* in bytes */
419 /* reset HW write offsets and offset buffer base */
420 if (!p
->cp
->render_ctx
) {
421 ilo_cp_set_one_off_flags(p
->cp
, INTEL_EXEC_GEN7_SOL_RESET
);
422 base
+= p
->state
.so_num_vertices
* stride
;
425 gen7_emit_3DSTATE_SO_BUFFER(p
->dev
, i
, base
, stride
,
426 ilo
->so
.states
[i
], p
->cp
);
430 gen7_emit_3DSTATE_SO_BUFFER(p
->dev
, i
, 0, 0, NULL
, p
->cp
);
433 /* 3DSTATE_SO_DECL_LIST */
434 if (dirty_sh
&& ilo
->so
.enabled
)
435 gen7_emit_3DSTATE_SO_DECL_LIST(p
->dev
, so_info
, p
->cp
);
437 /* 3DSTATE_STREAMOUT */
438 if (DIRTY(SO
) || DIRTY(RASTERIZER
) || dirty_sh
) {
439 const unsigned buffer_mask
= (1 << ilo
->so
.count
) - 1;
440 const int output_count
= ilo_shader_get_kernel_param(shader
,
441 ILO_KERNEL_OUTPUT_COUNT
);
443 gen7_emit_3DSTATE_STREAMOUT(p
->dev
, buffer_mask
, output_count
,
444 ilo
->rasterizer
->state
.rasterizer_discard
, p
->cp
);
449 gen7_pipeline_sf(struct ilo_3d_pipeline
*p
,
450 const struct ilo_context
*ilo
,
451 struct gen6_pipeline_session
*session
)
454 if (DIRTY(RASTERIZER
) || DIRTY(FS
))
455 gen7_emit_3DSTATE_SBE(p
->dev
, ilo
->rasterizer
, ilo
->fs
, ilo
->cp
);
458 if (DIRTY(RASTERIZER
) || DIRTY(FB
)) {
459 gen7_wa_pipe_control_cs_stall(p
, true, true);
460 gen7_emit_3DSTATE_SF(p
->dev
, ilo
->rasterizer
, ilo
->fb
.state
.zsbuf
, p
->cp
);
465 gen7_pipeline_wm(struct ilo_3d_pipeline
*p
,
466 const struct ilo_context
*ilo
,
467 struct gen6_pipeline_session
*session
)
470 if (DIRTY(FS
) || DIRTY(BLEND
) || DIRTY(DSA
) || DIRTY(RASTERIZER
)) {
471 const bool cc_may_kill
= (ilo
->dsa
->dw_alpha
||
472 ilo
->blend
->alpha_to_coverage
);
474 if (p
->dev
->gen
== ILO_GEN(7) && session
->hw_ctx_changed
)
475 gen7_wa_pipe_control_wm_max_threads_stall(p
);
477 gen7_emit_3DSTATE_WM(p
->dev
, ilo
->fs
,
478 ilo
->rasterizer
, cc_may_kill
, p
->cp
);
481 /* 3DSTATE_BINDING_TABLE_POINTERS_PS */
482 if (session
->binding_table_fs_changed
) {
483 gen7_emit_3DSTATE_BINDING_TABLE_POINTERS_PS(p
->dev
,
484 p
->state
.wm
.BINDING_TABLE_STATE
, p
->cp
);
487 /* 3DSTATE_SAMPLER_STATE_POINTERS_PS */
488 if (session
->sampler_state_fs_changed
) {
489 gen7_emit_3DSTATE_SAMPLER_STATE_POINTERS_PS(p
->dev
,
490 p
->state
.wm
.SAMPLER_STATE
, p
->cp
);
493 /* 3DSTATE_CONSTANT_PS */
494 if (session
->pcb_state_fs_changed
) {
495 gen7_emit_3DSTATE_CONSTANT_PS(p
->dev
,
496 &p
->state
.wm
.PUSH_CONSTANT_BUFFER
,
497 &p
->state
.wm
.PUSH_CONSTANT_BUFFER_size
,
502 if (DIRTY(FS
) || DIRTY(SAMPLER_FS
) || DIRTY(BLEND
) ||
503 session
->kernel_bo_changed
) {
504 const int num_samplers
= ilo
->sampler
[PIPE_SHADER_FRAGMENT
].count
;
505 const bool dual_blend
= ilo
->blend
->dual_blend
;
507 gen7_emit_3DSTATE_PS(p
->dev
, ilo
->fs
, num_samplers
, dual_blend
, p
->cp
);
510 /* 3DSTATE_SCISSOR_STATE_POINTERS */
511 if (session
->scissor_state_changed
) {
512 gen6_emit_3DSTATE_SCISSOR_STATE_POINTERS(p
->dev
,
513 p
->state
.SCISSOR_RECT
, p
->cp
);
516 /* XXX what is the best way to know if this workaround is needed? */
518 const bool emit_3dstate_ps
=
519 (DIRTY(FS
) || DIRTY(SAMPLER_FS
) || DIRTY(BLEND
));
520 const bool emit_3dstate_depth_buffer
=
521 (DIRTY(FB
) || DIRTY(DSA
) || session
->state_bo_changed
);
523 if (emit_3dstate_ps
||
524 emit_3dstate_depth_buffer
||
525 session
->pcb_state_fs_changed
||
526 session
->viewport_state_changed
||
527 session
->binding_table_fs_changed
||
528 session
->sampler_state_fs_changed
||
529 session
->cc_state_cc_changed
||
530 session
->cc_state_blend_changed
||
531 session
->cc_state_dsa_changed
)
532 gen7_wa_pipe_control_wm_depth_stall(p
, emit_3dstate_depth_buffer
);
535 /* 3DSTATE_DEPTH_BUFFER and 3DSTATE_CLEAR_PARAMS */
536 if (DIRTY(FB
) || session
->batch_bo_changed
) {
537 const struct ilo_zs_surface
*zs
;
539 if (ilo
->fb
.state
.zsbuf
) {
540 const struct ilo_surface_cso
*surface
=
541 (const struct ilo_surface_cso
*) ilo
->fb
.state
.zsbuf
;
543 assert(!surface
->is_rt
);
547 zs
= &ilo
->fb
.null_zs
;
550 gen6_emit_3DSTATE_DEPTH_BUFFER(p
->dev
, zs
, p
->cp
);
551 gen6_emit_3DSTATE_HIER_DEPTH_BUFFER(p
->dev
, zs
, p
->cp
);
552 gen6_emit_3DSTATE_STENCIL_BUFFER(p
->dev
, zs
, p
->cp
);
555 gen7_emit_3DSTATE_CLEAR_PARAMS(p
->dev
, 0, p
->cp
);
560 gen7_pipeline_wm_multisample(struct ilo_3d_pipeline
*p
,
561 const struct ilo_context
*ilo
,
562 struct gen6_pipeline_session
*session
)
564 /* 3DSTATE_MULTISAMPLE and 3DSTATE_SAMPLE_MASK */
565 if (DIRTY(SAMPLE_MASK
) || DIRTY(FB
)) {
566 const uint32_t *packed_sample_pos
;
568 gen7_wa_pipe_control_cs_stall(p
, true, true);
571 (ilo
->fb
.num_samples
> 4) ? p
->packed_sample_position_8x
:
572 (ilo
->fb
.num_samples
> 1) ? &p
->packed_sample_position_4x
:
573 &p
->packed_sample_position_1x
;
575 gen6_emit_3DSTATE_MULTISAMPLE(p
->dev
,
576 ilo
->fb
.num_samples
, packed_sample_pos
,
577 ilo
->rasterizer
->state
.half_pixel_center
, p
->cp
);
579 gen7_emit_3DSTATE_SAMPLE_MASK(p
->dev
,
580 (ilo
->fb
.num_samples
> 1) ? ilo
->sample_mask
: 0x1,
581 ilo
->fb
.num_samples
, p
->cp
);
586 gen7_pipeline_vf_draw(struct ilo_3d_pipeline
*p
,
587 const struct ilo_context
*ilo
,
588 struct gen6_pipeline_session
*session
)
591 gen7_emit_3DPRIMITIVE(p
->dev
, ilo
->draw
, &ilo
->ib
, false, p
->cp
);
592 p
->state
.has_gen6_wa_pipe_control
= false;
596 gen7_pipeline_commands(struct ilo_3d_pipeline
*p
,
597 const struct ilo_context
*ilo
,
598 struct gen6_pipeline_session
*session
)
601 * We try to keep the order of the commands match, as closely as possible,
602 * that of the classic i965 driver. It allows us to compare the command
605 gen6_pipeline_common_select(p
, ilo
, session
);
606 gen6_pipeline_common_sip(p
, ilo
, session
);
607 gen6_pipeline_vf_statistics(p
, ilo
, session
);
608 gen7_pipeline_common_pcb_alloc(p
, ilo
, session
);
609 gen6_pipeline_common_base_address(p
, ilo
, session
);
610 gen7_pipeline_common_pointers_1(p
, ilo
, session
);
611 gen7_pipeline_common_urb(p
, ilo
, session
);
612 gen7_pipeline_common_pointers_2(p
, ilo
, session
);
613 gen7_pipeline_wm_multisample(p
, ilo
, session
);
614 gen7_pipeline_gs(p
, ilo
, session
);
615 gen7_pipeline_hs(p
, ilo
, session
);
616 gen7_pipeline_te(p
, ilo
, session
);
617 gen7_pipeline_ds(p
, ilo
, session
);
618 gen7_pipeline_vs(p
, ilo
, session
);
619 gen7_pipeline_sol(p
, ilo
, session
);
620 gen6_pipeline_clip(p
, ilo
, session
);
621 gen7_pipeline_sf(p
, ilo
, session
);
622 gen7_pipeline_wm(p
, ilo
, session
);
623 gen6_pipeline_wm_raster(p
, ilo
, session
);
624 gen6_pipeline_sf_rect(p
, ilo
, session
);
625 gen6_pipeline_vf(p
, ilo
, session
);
626 gen7_pipeline_vf_draw(p
, ilo
, session
);
630 ilo_3d_pipeline_emit_draw_gen7(struct ilo_3d_pipeline
*p
,
631 const struct ilo_context
*ilo
)
633 struct gen6_pipeline_session session
;
635 gen6_pipeline_prepare(p
, ilo
, &session
);
637 session
.emit_draw_states
= gen6_pipeline_states
;
638 session
.emit_draw_commands
= gen7_pipeline_commands
;
640 gen6_pipeline_draw(p
, ilo
, &session
);
641 gen6_pipeline_end(p
, ilo
, &session
);
645 gen7_pipeline_estimate_commands(const struct ilo_3d_pipeline
*p
,
646 const struct ilo_context
*ilo
)
649 enum ilo_gpe_gen7_command cmd
;
654 for (cmd
= 0; cmd
< ILO_GPE_GEN7_COMMAND_COUNT
; cmd
++) {
658 case ILO_GPE_GEN7_PIPE_CONTROL
:
659 /* for the workaround */
661 /* another one after 3DSTATE_URB */
663 /* and another one after 3DSTATE_CONSTANT_VS */
666 case ILO_GPE_GEN7_3DSTATE_VERTEX_BUFFERS
:
669 case ILO_GPE_GEN7_3DSTATE_VERTEX_ELEMENTS
:
672 case ILO_GPE_GEN7_MEDIA_VFE_STATE
:
673 case ILO_GPE_GEN7_MEDIA_CURBE_LOAD
:
674 case ILO_GPE_GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD
:
675 case ILO_GPE_GEN7_MEDIA_STATE_FLUSH
:
676 case ILO_GPE_GEN7_GPGPU_WALKER
:
686 size
+= ilo_gpe_gen7_estimate_command_size(p
->dev
,
695 gen7_pipeline_estimate_states(const struct ilo_3d_pipeline
*p
,
696 const struct ilo_context
*ilo
)
698 static int static_size
;
699 int shader_type
, count
, size
;
703 enum ilo_gpe_gen7_state state
;
705 } static_states
[] = {
707 { ILO_GPE_GEN7_SF_CLIP_VIEWPORT
, 1 },
708 { ILO_GPE_GEN7_CC_VIEWPORT
, 1 },
710 { ILO_GPE_GEN7_COLOR_CALC_STATE
, 1 },
711 { ILO_GPE_GEN7_BLEND_STATE
, ILO_MAX_DRAW_BUFFERS
},
712 { ILO_GPE_GEN7_DEPTH_STENCIL_STATE
, 1 },
714 { ILO_GPE_GEN7_SCISSOR_RECT
, 1 },
715 /* binding table (vs, gs, fs) */
716 { ILO_GPE_GEN7_BINDING_TABLE_STATE
, ILO_MAX_VS_SURFACES
},
717 { ILO_GPE_GEN7_BINDING_TABLE_STATE
, ILO_MAX_GS_SURFACES
},
718 { ILO_GPE_GEN7_BINDING_TABLE_STATE
, ILO_MAX_WM_SURFACES
},
722 for (i
= 0; i
< Elements(static_states
); i
++) {
723 static_size
+= ilo_gpe_gen7_estimate_state_size(p
->dev
,
724 static_states
[i
].state
,
725 static_states
[i
].count
);
732 * render targets (fs)
733 * sampler views (vs, fs)
734 * constant buffers (vs, fs)
736 count
= ilo
->fb
.state
.nr_cbufs
;
737 for (shader_type
= 0; shader_type
< PIPE_SHADER_TYPES
; shader_type
++) {
738 count
+= ilo
->view
[shader_type
].count
;
739 count
+= util_bitcount(ilo
->cbuf
[shader_type
].enabled_mask
);
743 size
+= ilo_gpe_gen7_estimate_state_size(p
->dev
,
744 ILO_GPE_GEN7_SURFACE_STATE
, count
);
747 /* samplers (vs, fs) */
748 for (shader_type
= 0; shader_type
< PIPE_SHADER_TYPES
; shader_type
++) {
749 count
= ilo
->sampler
[shader_type
].count
;
751 size
+= ilo_gpe_gen7_estimate_state_size(p
->dev
,
752 ILO_GPE_GEN7_SAMPLER_BORDER_COLOR_STATE
, count
);
753 size
+= ilo_gpe_gen7_estimate_state_size(p
->dev
,
754 ILO_GPE_GEN7_SAMPLER_STATE
, count
);
760 const int cbuf0_size
=
761 ilo_shader_get_kernel_param(ilo
->vs
, ILO_KERNEL_PCB_CBUF0_SIZE
);
763 ilo_shader_get_kernel_param(ilo
->vs
, ILO_KERNEL_VS_PCB_UCP_SIZE
);
765 size
+= ilo_gpe_gen7_estimate_state_size(p
->dev
,
766 ILO_GPE_GEN7_PUSH_CONSTANT_BUFFER
, cbuf0_size
+ ucp_size
);
771 const int cbuf0_size
=
772 ilo_shader_get_kernel_param(ilo
->fs
, ILO_KERNEL_PCB_CBUF0_SIZE
);
774 size
+= ilo_gpe_gen7_estimate_state_size(p
->dev
,
775 ILO_GPE_GEN7_PUSH_CONSTANT_BUFFER
, cbuf0_size
);
782 ilo_3d_pipeline_estimate_size_gen7(struct ilo_3d_pipeline
*p
,
783 enum ilo_3d_pipeline_action action
,
789 case ILO_3D_PIPELINE_DRAW
:
791 const struct ilo_context
*ilo
= arg
;
793 size
= gen7_pipeline_estimate_commands(p
, ilo
) +
794 gen7_pipeline_estimate_states(p
, ilo
);
797 case ILO_3D_PIPELINE_FLUSH
:
798 case ILO_3D_PIPELINE_WRITE_TIMESTAMP
:
799 case ILO_3D_PIPELINE_WRITE_DEPTH_COUNT
:
800 size
= ilo_gpe_gen7_estimate_command_size(p
->dev
,
801 ILO_GPE_GEN7_PIPE_CONTROL
, 1);
804 assert(!"unknown 3D pipeline action");
813 ilo_3d_pipeline_init_gen7(struct ilo_3d_pipeline
*p
)
815 p
->estimate_size
= ilo_3d_pipeline_estimate_size_gen7
;
816 p
->emit_draw
= ilo_3d_pipeline_emit_draw_gen7
;
817 p
->emit_flush
= ilo_3d_pipeline_emit_flush_gen6
;
818 p
->emit_write_timestamp
= ilo_3d_pipeline_emit_write_timestamp_gen6
;
819 p
->emit_write_depth_count
= ilo_3d_pipeline_emit_write_depth_count_gen6
;