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 "intel_batchbuffer.h"
35 #include "intel_regions.h"
37 #include "brw_context.h"
38 #include "brw_state.h"
39 #include "brw_defines.h"
45 /***********************************************************************
49 static void upload_blend_constant_color(struct brw_context
*brw
)
51 GLcontext
*ctx
= &brw
->intel
.ctx
;
52 struct brw_blend_constant_color bcc
;
54 memset(&bcc
, 0, sizeof(bcc
));
55 bcc
.header
.opcode
= CMD_BLEND_CONSTANT_COLOR
;
56 bcc
.header
.length
= sizeof(bcc
)/4-2;
57 bcc
.blend_constant_color
[0] = ctx
->Color
.BlendColor
[0];
58 bcc
.blend_constant_color
[1] = ctx
->Color
.BlendColor
[1];
59 bcc
.blend_constant_color
[2] = ctx
->Color
.BlendColor
[2];
60 bcc
.blend_constant_color
[3] = ctx
->Color
.BlendColor
[3];
62 BRW_CACHED_BATCH_STRUCT(brw
, &bcc
);
66 const struct brw_tracked_state brw_blend_constant_color
= {
72 .emit
= upload_blend_constant_color
75 /* Constant single cliprect for framebuffer object or DRI2 drawing */
76 static void upload_drawing_rect(struct brw_context
*brw
)
78 struct intel_context
*intel
= &brw
->intel
;
79 GLcontext
*ctx
= &intel
->ctx
;
81 if (!intel
->constant_cliprect
)
84 BEGIN_BATCH(4, NO_LOOP_CLIPRECTS
);
85 OUT_BATCH(_3DSTATE_DRAWRECT_INFO_I965
);
86 OUT_BATCH(0); /* xmin, ymin */
87 OUT_BATCH(((ctx
->DrawBuffer
->Width
- 1) & 0xffff) |
88 ((ctx
->DrawBuffer
->Height
- 1) << 16));
93 const struct brw_tracked_state brw_drawing_rect
= {
99 .emit
= upload_drawing_rect
102 static void prepare_binding_table_pointers(struct brw_context
*brw
)
104 brw_add_validated_bo(brw
, brw
->vs
.bind_bo
);
105 brw_add_validated_bo(brw
, brw
->wm
.bind_bo
);
109 * Upload the binding table pointers, which point each stage's array of surface
112 * The binding table pointers are relative to the surface state base address,
115 static void upload_binding_table_pointers(struct brw_context
*brw
)
117 struct intel_context
*intel
= &brw
->intel
;
119 BEGIN_BATCH(6, IGNORE_CLIPRECTS
);
120 OUT_BATCH(CMD_BINDING_TABLE_PTRS
<< 16 | (6 - 2));
121 if (brw
->vs
.bind_bo
!= NULL
)
122 OUT_RELOC(brw
->vs
.bind_bo
, I915_GEM_DOMAIN_SAMPLER
, 0, 0); /* vs */
125 OUT_BATCH(0); /* gs */
126 OUT_BATCH(0); /* clip */
127 OUT_BATCH(0); /* sf */
128 OUT_RELOC(brw
->wm
.bind_bo
, I915_GEM_DOMAIN_SAMPLER
, 0, 0); /* wm/ps */
132 const struct brw_tracked_state brw_binding_table_pointers
= {
135 .brw
= BRW_NEW_BATCH
,
136 .cache
= CACHE_NEW_SURF_BIND
,
138 .prepare
= prepare_binding_table_pointers
,
139 .emit
= upload_binding_table_pointers
,
144 * Upload pointers to the per-stage state.
146 * The state pointers in this packet are all relative to the general state
147 * base address set by CMD_STATE_BASE_ADDRESS, which is 0.
149 static void upload_pipelined_state_pointers(struct brw_context
*brw
)
151 struct intel_context
*intel
= &brw
->intel
;
153 BEGIN_BATCH(7, IGNORE_CLIPRECTS
);
154 OUT_BATCH(CMD_PIPELINED_STATE_POINTERS
<< 16 | (7 - 2));
155 OUT_RELOC(brw
->vs
.state_bo
, I915_GEM_DOMAIN_INSTRUCTION
, 0, 0);
156 if (brw
->gs
.prog_active
)
157 OUT_RELOC(brw
->gs
.state_bo
, I915_GEM_DOMAIN_INSTRUCTION
, 0, 1);
160 OUT_RELOC(brw
->clip
.state_bo
, I915_GEM_DOMAIN_INSTRUCTION
, 0, 1);
161 OUT_RELOC(brw
->sf
.state_bo
, I915_GEM_DOMAIN_INSTRUCTION
, 0, 0);
162 OUT_RELOC(brw
->wm
.state_bo
, I915_GEM_DOMAIN_INSTRUCTION
, 0, 0);
163 OUT_RELOC(brw
->cc
.state_bo
, I915_GEM_DOMAIN_INSTRUCTION
, 0, 0);
166 brw
->state
.dirty
.brw
|= BRW_NEW_PSP
;
170 static void prepare_psp_urb_cbs(struct brw_context
*brw
)
172 brw_add_validated_bo(brw
, brw
->vs
.state_bo
);
173 brw_add_validated_bo(brw
, brw
->gs
.state_bo
);
174 brw_add_validated_bo(brw
, brw
->clip
.state_bo
);
175 brw_add_validated_bo(brw
, brw
->wm
.state_bo
);
176 brw_add_validated_bo(brw
, brw
->cc
.state_bo
);
179 static void upload_psp_urb_cbs(struct brw_context
*brw
)
181 upload_pipelined_state_pointers(brw
);
182 brw_upload_urb_fence(brw
);
183 brw_upload_cs_urb_state(brw
);
186 const struct brw_tracked_state brw_psp_urb_cbs
= {
189 .brw
= BRW_NEW_URB_FENCE
| BRW_NEW_BATCH
,
190 .cache
= (CACHE_NEW_VS_UNIT
|
193 CACHE_NEW_CLIP_UNIT
|
198 .prepare
= prepare_psp_urb_cbs
,
199 .emit
= upload_psp_urb_cbs
,
202 static void prepare_depthbuffer(struct brw_context
*brw
)
204 struct intel_region
*region
= brw
->state
.depth_region
;
207 brw_add_validated_bo(brw
, region
->buffer
);
210 static void emit_depthbuffer(struct brw_context
*brw
)
212 struct intel_context
*intel
= &brw
->intel
;
213 struct intel_region
*region
= brw
->state
.depth_region
;
214 unsigned int len
= BRW_IS_G4X(brw
) ? 6 : 5;
216 if (region
== NULL
) {
217 BEGIN_BATCH(len
, IGNORE_CLIPRECTS
);
218 OUT_BATCH(CMD_DEPTH_BUFFER
<< 16 | (len
- 2));
219 OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT
<< 18) |
220 (BRW_SURFACE_NULL
<< 29));
232 switch (region
->cpp
) {
234 format
= BRW_DEPTHFORMAT_D16_UNORM
;
237 if (intel
->depth_buffer_is_float
)
238 format
= BRW_DEPTHFORMAT_D32_FLOAT
;
240 format
= BRW_DEPTHFORMAT_D24_UNORM_S8_UINT
;
247 assert(region
->tiling
!= I915_TILING_X
);
249 BEGIN_BATCH(len
, IGNORE_CLIPRECTS
);
250 OUT_BATCH(CMD_DEPTH_BUFFER
<< 16 | (len
- 2));
251 OUT_BATCH(((region
->pitch
* region
->cpp
) - 1) |
253 (BRW_TILEWALK_YMAJOR
<< 26) |
254 ((region
->tiling
!= I915_TILING_NONE
) << 27) |
255 (BRW_SURFACE_2D
<< 29));
256 OUT_RELOC(region
->buffer
,
257 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
259 OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW
<< 1) |
260 ((region
->pitch
- 1) << 6) |
261 ((region
->height
- 1) << 19));
271 const struct brw_tracked_state brw_depthbuffer
= {
274 .brw
= BRW_NEW_DEPTH_BUFFER
| BRW_NEW_BATCH
,
277 .prepare
= prepare_depthbuffer
,
278 .emit
= emit_depthbuffer
,
283 /***********************************************************************
284 * Polygon stipple packet
287 static void upload_polygon_stipple(struct brw_context
*brw
)
289 GLcontext
*ctx
= &brw
->intel
.ctx
;
290 struct brw_polygon_stipple bps
;
293 memset(&bps
, 0, sizeof(bps
));
294 bps
.header
.opcode
= CMD_POLY_STIPPLE_PATTERN
;
295 bps
.header
.length
= sizeof(bps
)/4-2;
297 /* Polygon stipple is provided in OpenGL order, i.e. bottom
298 * row first. If we're rendering to a window (i.e. the
299 * default frame buffer object, 0), then we need to invert
300 * it to match our pixel layout. But if we're rendering
301 * to a FBO (i.e. any named frame buffer object), we *don't*
302 * need to invert - we already match the layout.
304 if (ctx
->DrawBuffer
->Name
== 0) {
305 for (i
= 0; i
< 32; i
++)
306 bps
.stipple
[i
] = ctx
->PolygonStipple
[31 - i
]; /* invert */
309 for (i
= 0; i
< 32; i
++)
310 bps
.stipple
[i
] = ctx
->PolygonStipple
[i
]; /* don't invert */
313 BRW_CACHED_BATCH_STRUCT(brw
, &bps
);
316 const struct brw_tracked_state brw_polygon_stipple
= {
318 .mesa
= _NEW_POLYGONSTIPPLE
,
322 .emit
= upload_polygon_stipple
326 /***********************************************************************
327 * Polygon stipple offset packet
330 static void upload_polygon_stipple_offset(struct brw_context
*brw
)
332 __DRIdrawablePrivate
*dPriv
= brw
->intel
.driDrawable
;
333 struct brw_polygon_stipple_offset bpso
;
335 memset(&bpso
, 0, sizeof(bpso
));
336 bpso
.header
.opcode
= CMD_POLY_STIPPLE_OFFSET
;
337 bpso
.header
.length
= sizeof(bpso
)/4-2;
339 /* If we're drawing to a system window (ctx->DrawBuffer->Name == 0),
340 * we have to invert the Y axis in order to match the OpenGL
341 * pixel coordinate system, and our offset must be matched
342 * to the window position. If we're drawing to a FBO
343 * (ctx->DrawBuffer->Name != 0), then our native pixel coordinate
344 * system works just fine, and there's no window system to
347 if (brw
->intel
.ctx
.DrawBuffer
->Name
== 0) {
348 bpso
.bits0
.x_offset
= (32 - (dPriv
->x
& 31)) & 31;
349 bpso
.bits0
.y_offset
= (32 - ((dPriv
->y
+ dPriv
->h
) & 31)) & 31;
352 bpso
.bits0
.y_offset
= 0;
353 bpso
.bits0
.x_offset
= 0;
356 BRW_CACHED_BATCH_STRUCT(brw
, &bpso
);
359 #define _NEW_WINDOW_POS 0x40000000
361 const struct brw_tracked_state brw_polygon_stipple_offset
= {
363 .mesa
= _NEW_WINDOW_POS
,
367 .emit
= upload_polygon_stipple_offset
370 /**********************************************************************
373 static void upload_aa_line_parameters(struct brw_context
*brw
)
375 struct brw_aa_line_parameters balp
;
377 if (!BRW_IS_G4X(brw
))
380 /* use legacy aa line coverage computation */
381 memset(&balp
, 0, sizeof(balp
));
382 balp
.header
.opcode
= CMD_AA_LINE_PARAMETERS
;
383 balp
.header
.length
= sizeof(balp
) / 4 - 2;
385 BRW_CACHED_BATCH_STRUCT(brw
, &balp
);
388 const struct brw_tracked_state brw_aa_line_parameters
= {
391 .brw
= BRW_NEW_CONTEXT
,
394 .emit
= upload_aa_line_parameters
397 /***********************************************************************
398 * Line stipple packet
401 static void upload_line_stipple(struct brw_context
*brw
)
403 GLcontext
*ctx
= &brw
->intel
.ctx
;
404 struct brw_line_stipple bls
;
408 memset(&bls
, 0, sizeof(bls
));
409 bls
.header
.opcode
= CMD_LINE_STIPPLE_PATTERN
;
410 bls
.header
.length
= sizeof(bls
)/4 - 2;
412 bls
.bits0
.pattern
= ctx
->Line
.StipplePattern
;
413 bls
.bits1
.repeat_count
= ctx
->Line
.StippleFactor
;
415 tmp
= 1.0 / (GLfloat
) ctx
->Line
.StippleFactor
;
416 tmpi
= tmp
* (1<<13);
419 bls
.bits1
.inverse_repeat_count
= tmpi
;
421 BRW_CACHED_BATCH_STRUCT(brw
, &bls
);
424 const struct brw_tracked_state brw_line_stipple
= {
430 .emit
= upload_line_stipple
434 /***********************************************************************
435 * Misc invarient state packets
438 static void upload_invarient_state( struct brw_context
*brw
)
441 /* 0x61040000 Pipeline Select */
442 /* PipelineSelect : 0 */
443 struct brw_pipeline_select ps
;
445 memset(&ps
, 0, sizeof(ps
));
446 ps
.header
.opcode
= CMD_PIPELINE_SELECT(brw
);
447 ps
.header
.pipeline_select
= 0;
448 BRW_BATCH_STRUCT(brw
, &ps
);
452 struct brw_global_depth_offset_clamp gdo
;
453 memset(&gdo
, 0, sizeof(gdo
));
455 /* Disable depth offset clamping.
457 gdo
.header
.opcode
= CMD_GLOBAL_DEPTH_OFFSET_CLAMP
;
458 gdo
.header
.length
= sizeof(gdo
)/4 - 2;
459 gdo
.depth_offset_clamp
= 0.0;
461 BRW_BATCH_STRUCT(brw
, &gdo
);
465 /* 0x61020000 State Instruction Pointer */
467 struct brw_system_instruction_pointer sip
;
468 memset(&sip
, 0, sizeof(sip
));
470 sip
.header
.opcode
= CMD_STATE_INSN_POINTER
;
471 sip
.header
.length
= 0;
473 sip
.bits0
.system_instruction_pointer
= 0;
474 BRW_BATCH_STRUCT(brw
, &sip
);
479 struct brw_vf_statistics vfs
;
480 memset(&vfs
, 0, sizeof(vfs
));
482 vfs
.opcode
= CMD_VF_STATISTICS(brw
);
483 if (INTEL_DEBUG
& DEBUG_STATS
)
484 vfs
.statistics_enable
= 1;
486 BRW_BATCH_STRUCT(brw
, &vfs
);
490 const struct brw_tracked_state brw_invarient_state
= {
493 .brw
= BRW_NEW_CONTEXT
,
496 .emit
= upload_invarient_state
500 * Define the base addresses which some state is referenced from.
502 * This allows us to avoid having to emit relocations in many places for
503 * cached state, and instead emit pointers inside of large, mostly-static
504 * state pools. This comes at the expense of memory, and more expensive cache
507 static void upload_state_base_address( struct brw_context
*brw
)
509 struct intel_context
*intel
= &brw
->intel
;
511 /* Output the structure (brw_state_base_address) directly to the
512 * batchbuffer, so we can emit relocations inline.
514 BEGIN_BATCH(6, IGNORE_CLIPRECTS
);
515 OUT_BATCH(CMD_STATE_BASE_ADDRESS
<< 16 | (6 - 2));
516 OUT_BATCH(1); /* General state base address */
517 OUT_BATCH(1); /* Surface state base address */
518 OUT_BATCH(1); /* Indirect object base address */
519 OUT_BATCH(1); /* General state upper bound */
520 OUT_BATCH(1); /* Indirect object upper bound */
524 const struct brw_tracked_state brw_state_base_address
= {
527 .brw
= BRW_NEW_CONTEXT
,
530 .emit
= upload_state_base_address