1 /**************************************************************************
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * 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, sub license, 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 portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
30 #include "main/glheader.h"
31 #include "main/mtypes.h"
32 #include "main/imports.h"
33 #include "main/macros.h"
34 #include "main/colormac.h"
36 #include "tnl/t_context.h"
37 #include "tnl/t_vertex.h"
39 #include "intel_batchbuffer.h"
40 #include "intel_tex.h"
41 #include "intel_regions.h"
42 #include "intel_tris.h"
45 #include "i915_context.h"
47 #include "glapi/glapi.h"
50 i915_render_prevalidate(struct intel_context
*intel
)
52 struct i915_context
*i915
= i915_context(&intel
->ctx
);
55 i915ValidateFragmentProgram(i915
);
59 i915_render_start(struct intel_context
*intel
)
65 i915_reduced_primitive_state(struct intel_context
*intel
, GLenum rprim
)
67 struct i915_context
*i915
= i915_context(&intel
->ctx
);
68 GLuint st1
= i915
->state
.Stipple
[I915_STPREG_ST1
];
73 case GL_QUADS
: /* from RASTERIZE(GL_QUADS) in t_dd_tritemp.h */
75 if (intel
->ctx
.Polygon
.StippleFlag
&& intel
->hw_stipple
)
84 i915
->intel
.reduced_primitive
= rprim
;
86 if (st1
!= i915
->state
.Stipple
[I915_STPREG_ST1
]) {
87 INTEL_FIREVERTICES(intel
);
89 I915_STATECHANGE(i915
, I915_UPLOAD_STIPPLE
);
90 i915
->state
.Stipple
[I915_STPREG_ST1
] = st1
;
95 /* Pull apart the vertex format registers and figure out how large a
96 * vertex is supposed to be.
99 i915_check_vertex_size(struct intel_context
*intel
, GLuint expected
)
101 struct i915_context
*i915
= i915_context(&intel
->ctx
);
102 int lis2
= i915
->current
->Ctx
[I915_CTXREG_LIS2
];
103 int lis4
= i915
->current
->Ctx
[I915_CTXREG_LIS4
];
106 switch (lis4
& S4_VFMT_XYZW_MASK
) {
120 fprintf(stderr
, "no xyzw specified\n");
124 if (lis4
& S4_VFMT_SPEC_FOG
)
126 if (lis4
& S4_VFMT_COLOR
)
128 if (lis4
& S4_VFMT_DEPTH_OFFSET
)
130 if (lis4
& S4_VFMT_POINT_WIDTH
)
132 if (lis4
& S4_VFMT_FOG_PARAM
)
135 for (i
= 0; i
< 8; i
++) {
136 switch (lis2
& S2_TEXCOORD_FMT0_MASK
) {
149 case TEXCOORDFMT_2D_16
:
152 case TEXCOORDFMT_4D_16
:
155 case TEXCOORDFMT_NOT_PRESENT
:
158 fprintf(stderr
, "bad texcoord fmt %d\n", i
);
161 lis2
>>= S2_TEXCOORD_FMT1_SHIFT
;
165 fprintf(stderr
, "vertex size mismatch %d/%d\n", sz
, expected
);
167 return sz
== expected
;
172 i915_emit_invarient_state(struct intel_context
*intel
)
176 BEGIN_BATCH(200, IGNORE_CLIPRECTS
);
178 OUT_BATCH(_3DSTATE_AA_CMD
|
179 AA_LINE_ECAAR_WIDTH_ENABLE
|
180 AA_LINE_ECAAR_WIDTH_1_0
|
181 AA_LINE_REGION_WIDTH_ENABLE
| AA_LINE_REGION_WIDTH_1_0
);
183 OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD
);
186 OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD
);
189 OUT_BATCH(_3DSTATE_DFLT_Z_CMD
);
192 /* Don't support texture crossbar yet */
193 OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS
|
198 CSB_TCB(4, 4) | CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7));
200 OUT_BATCH(_3DSTATE_RASTER_RULES_CMD
|
201 ENABLE_POINT_RASTER_RULE
|
202 OGL_POINT_RASTER_RULE
|
203 ENABLE_LINE_STRIP_PROVOKE_VRTX
|
204 ENABLE_TRI_FAN_PROVOKE_VRTX
|
205 LINE_STRIP_PROVOKE_VRTX(1) |
206 TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D
| TEXKILL_4D
);
208 /* Need to initialize this to zero.
210 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1
| I1_LOAD_S(3) | (0));
214 OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD
| DISABLE_SCISSOR_RECT
);
216 OUT_BATCH(_3DSTATE_SCISSOR_RECT_0_CMD
);
220 OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE
);
222 OUT_BATCH(_3DSTATE_LOAD_INDIRECT
| 0); /* disable indirect state */
226 /* Don't support twosided stencil yet */
227 OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS
| BFO_ENABLE_STENCIL_TWO_SIDE
| 0);
234 #define emit(intel, state, size ) \
235 intel_batchbuffer_data(intel->batch, state, size, IGNORE_CLIPRECTS )
238 get_dirty(struct i915_hw_state
*state
)
242 /* Workaround the multitex hang - if one texture unit state is
243 * modified, emit all texture units.
245 dirty
= state
->active
& ~state
->emitted
;
246 if (dirty
& I915_UPLOAD_TEX_ALL
)
247 state
->emitted
&= ~I915_UPLOAD_TEX_ALL
;
248 dirty
= state
->active
& ~state
->emitted
;
254 get_state_size(struct i915_hw_state
*state
)
256 GLuint dirty
= get_dirty(state
);
260 if (dirty
& I915_UPLOAD_INVARIENT
)
263 if (dirty
& I915_UPLOAD_CTX
)
264 sz
+= sizeof(state
->Ctx
);
266 if (dirty
& I915_UPLOAD_BUFFERS
)
267 sz
+= sizeof(state
->Buffer
);
269 if (dirty
& I915_UPLOAD_STIPPLE
)
270 sz
+= sizeof(state
->Stipple
);
272 if (dirty
& I915_UPLOAD_FOG
)
273 sz
+= sizeof(state
->Fog
);
275 if (dirty
& I915_UPLOAD_TEX_ALL
) {
277 for (i
= 0; i
< I915_TEX_UNITS
; i
++)
278 if (dirty
& I915_UPLOAD_TEX(i
))
281 sz
+= (2 + nr
* 3) * sizeof(GLuint
) * 2;
284 if (dirty
& I915_UPLOAD_CONSTANTS
)
285 sz
+= state
->ConstantSize
* sizeof(GLuint
);
287 if (dirty
& I915_UPLOAD_PROGRAM
)
288 sz
+= state
->ProgramSize
* sizeof(GLuint
);
293 /* Push the state into the sarea and/or texture memory.
296 i915_emit_state(struct intel_context
*intel
)
298 struct i915_context
*i915
= i915_context(&intel
->ctx
);
299 struct i915_hw_state
*state
= i915
->current
;
300 int i
, count
, aper_count
;
302 dri_bo
*aper_array
[3 + I915_TEX_UNITS
];
303 GET_CURRENT_CONTEXT(ctx
);
306 /* We don't hold the lock at this point, so want to make sure that
307 * there won't be a buffer wrap between the state emits and the primitive
310 * It might be better to talk about explicit places where
311 * scheduling is allowed, rather than assume that it is whenever a
312 * batchbuffer fills up.
314 * Set the space as LOOP_CLIPRECTS now, since that's what our primitives
315 * will be emitted under.
317 intel_batchbuffer_require_space(intel
->batch
,
318 get_state_size(state
) + INTEL_PRIM_EMIT_SIZE
,
323 dirty
= get_dirty(state
);
325 aper_array
[aper_count
++] = intel
->batch
->buf
;
326 if (dirty
& I915_UPLOAD_BUFFERS
) {
327 aper_array
[aper_count
++] = state
->draw_region
->buffer
;
328 if (state
->depth_region
)
329 aper_array
[aper_count
++] = state
->depth_region
->buffer
;
332 if (dirty
& I915_UPLOAD_TEX_ALL
) {
333 for (i
= 0; i
< I915_TEX_UNITS
; i
++) {
334 if (dirty
& I915_UPLOAD_TEX(i
)) {
335 if (state
->tex_buffer
[i
]) {
336 aper_array
[aper_count
++] = state
->tex_buffer
[i
];
342 if (dri_bufmgr_check_aperture_space(aper_array
, aper_count
)) {
345 intel_batchbuffer_flush(intel
->batch
);
348 _mesa_error(ctx
, GL_OUT_OF_MEMORY
, "i915 emit state");
353 /* work out list of buffers to emit */
355 /* Do this here as we may have flushed the batchbuffer above,
356 * causing more state to be dirty!
358 dirty
= get_dirty(state
);
359 state
->emitted
|= dirty
;
360 assert(get_dirty(state
) == 0);
362 if (INTEL_DEBUG
& DEBUG_STATE
)
363 fprintf(stderr
, "%s dirty: %x\n", __FUNCTION__
, dirty
);
365 if (dirty
& I915_UPLOAD_INVARIENT
) {
366 if (INTEL_DEBUG
& DEBUG_STATE
)
367 fprintf(stderr
, "I915_UPLOAD_INVARIENT:\n");
368 i915_emit_invarient_state(intel
);
371 if (dirty
& I915_UPLOAD_CTX
) {
372 if (INTEL_DEBUG
& DEBUG_STATE
)
373 fprintf(stderr
, "I915_UPLOAD_CTX:\n");
375 emit(intel
, state
->Ctx
, sizeof(state
->Ctx
));
378 if (dirty
& I915_UPLOAD_BUFFERS
) {
379 if (INTEL_DEBUG
& DEBUG_STATE
)
380 fprintf(stderr
, "I915_UPLOAD_BUFFERS:\n");
381 BEGIN_BATCH(I915_DEST_SETUP_SIZE
+ 2, IGNORE_CLIPRECTS
);
382 OUT_BATCH(state
->Buffer
[I915_DESTREG_CBUFADDR0
]);
383 OUT_BATCH(state
->Buffer
[I915_DESTREG_CBUFADDR1
]);
384 OUT_RELOC(state
->draw_region
->buffer
,
385 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
386 state
->draw_region
->draw_offset
);
388 if (state
->depth_region
) {
389 OUT_BATCH(state
->Buffer
[I915_DESTREG_DBUFADDR0
]);
390 OUT_BATCH(state
->Buffer
[I915_DESTREG_DBUFADDR1
]);
391 OUT_RELOC(state
->depth_region
->buffer
,
392 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
393 state
->depth_region
->draw_offset
);
396 OUT_BATCH(state
->Buffer
[I915_DESTREG_DV0
]);
397 OUT_BATCH(state
->Buffer
[I915_DESTREG_DV1
]);
398 OUT_BATCH(state
->Buffer
[I915_DESTREG_SENABLE
]);
399 OUT_BATCH(state
->Buffer
[I915_DESTREG_SR0
]);
400 OUT_BATCH(state
->Buffer
[I915_DESTREG_SR1
]);
401 OUT_BATCH(state
->Buffer
[I915_DESTREG_SR2
]);
405 if (dirty
& I915_UPLOAD_STIPPLE
) {
406 if (INTEL_DEBUG
& DEBUG_STATE
)
407 fprintf(stderr
, "I915_UPLOAD_STIPPLE:\n");
408 emit(intel
, state
->Stipple
, sizeof(state
->Stipple
));
411 if (dirty
& I915_UPLOAD_FOG
) {
412 if (INTEL_DEBUG
& DEBUG_STATE
)
413 fprintf(stderr
, "I915_UPLOAD_FOG:\n");
414 emit(intel
, state
->Fog
, sizeof(state
->Fog
));
417 /* Combine all the dirty texture state into a single command to
418 * avoid lockups on I915 hardware.
420 if (dirty
& I915_UPLOAD_TEX_ALL
) {
423 for (i
= 0; i
< I915_TEX_UNITS
; i
++)
424 if (dirty
& I915_UPLOAD_TEX(i
))
427 BEGIN_BATCH(2 + nr
* 3, IGNORE_CLIPRECTS
);
428 OUT_BATCH(_3DSTATE_MAP_STATE
| (3 * nr
));
429 OUT_BATCH((dirty
& I915_UPLOAD_TEX_ALL
) >> I915_UPLOAD_TEX_0_SHIFT
);
430 for (i
= 0; i
< I915_TEX_UNITS
; i
++)
431 if (dirty
& I915_UPLOAD_TEX(i
)) {
433 if (state
->tex_buffer
[i
]) {
434 OUT_RELOC(state
->tex_buffer
[i
],
435 I915_GEM_DOMAIN_SAMPLER
, 0,
436 state
->tex_offset
[i
]);
438 else if (state
== &i915
->meta
) {
443 OUT_BATCH(state
->tex_offset
[i
]);
446 OUT_BATCH(state
->Tex
[i
][I915_TEXREG_MS3
]);
447 OUT_BATCH(state
->Tex
[i
][I915_TEXREG_MS4
]);
451 BEGIN_BATCH(2 + nr
* 3, IGNORE_CLIPRECTS
);
452 OUT_BATCH(_3DSTATE_SAMPLER_STATE
| (3 * nr
));
453 OUT_BATCH((dirty
& I915_UPLOAD_TEX_ALL
) >> I915_UPLOAD_TEX_0_SHIFT
);
454 for (i
= 0; i
< I915_TEX_UNITS
; i
++)
455 if (dirty
& I915_UPLOAD_TEX(i
)) {
456 OUT_BATCH(state
->Tex
[i
][I915_TEXREG_SS2
]);
457 OUT_BATCH(state
->Tex
[i
][I915_TEXREG_SS3
]);
458 OUT_BATCH(state
->Tex
[i
][I915_TEXREG_SS4
]);
463 if (dirty
& I915_UPLOAD_CONSTANTS
) {
464 if (INTEL_DEBUG
& DEBUG_STATE
)
465 fprintf(stderr
, "I915_UPLOAD_CONSTANTS:\n");
466 emit(intel
, state
->Constant
, state
->ConstantSize
* sizeof(GLuint
));
469 if (dirty
& I915_UPLOAD_PROGRAM
) {
470 if (state
->ProgramSize
) {
471 if (INTEL_DEBUG
& DEBUG_STATE
)
472 fprintf(stderr
, "I915_UPLOAD_PROGRAM:\n");
474 assert((state
->Program
[0] & 0x1ff) + 2 == state
->ProgramSize
);
476 emit(intel
, state
->Program
, state
->ProgramSize
* sizeof(GLuint
));
477 if (INTEL_DEBUG
& DEBUG_STATE
)
478 i915_disassemble_program(state
->Program
, state
->ProgramSize
);
482 intel
->batch
->dirty_state
&= ~dirty
;
483 assert(get_dirty(state
) == 0);
484 assert((intel
->batch
->dirty_state
& (1<<1)) == 0);
488 i915_destroy_context(struct intel_context
*intel
)
491 struct i915_context
*i915
= i915_context(&intel
->ctx
);
493 intel_region_release(&i915
->state
.draw_region
);
494 intel_region_release(&i915
->state
.depth_region
);
495 intel_region_release(&i915
->meta
.draw_region
);
496 intel_region_release(&i915
->meta
.depth_region
);
497 intel_region_release(&i915
->initial
.draw_region
);
498 intel_region_release(&i915
->initial
.depth_region
);
500 for (i
= 0; i
< I915_TEX_UNITS
; i
++) {
501 if (i915
->state
.tex_buffer
[i
] != NULL
) {
502 dri_bo_unreference(i915
->state
.tex_buffer
[i
]);
503 i915
->state
.tex_buffer
[i
] = NULL
;
507 _tnl_free_vertices(&intel
->ctx
);
512 * Set the drawing regions for the color and depth/stencil buffers.
513 * This involves setting the pitch, cpp and buffer ID/location.
514 * Also set pixel format for color and Z rendering
515 * Used for setting both regular and meta state.
518 i915_state_draw_region(struct intel_context
*intel
,
519 struct i915_hw_state
*state
,
520 struct intel_region
*color_region
,
521 struct intel_region
*depth_region
)
523 struct i915_context
*i915
= i915_context(&intel
->ctx
);
526 ASSERT(state
== &i915
->state
|| state
== &i915
->meta
);
528 if (state
->draw_region
!= color_region
) {
529 intel_region_release(&state
->draw_region
);
530 intel_region_reference(&state
->draw_region
, color_region
);
532 if (state
->depth_region
!= depth_region
) {
533 intel_region_release(&state
->depth_region
);
534 intel_region_reference(&state
->depth_region
, depth_region
);
538 * Set stride/cpp values
541 state
->Buffer
[I915_DESTREG_CBUFADDR0
] = _3DSTATE_BUF_INFO_CMD
;
542 state
->Buffer
[I915_DESTREG_CBUFADDR1
] =
543 (BUF_3D_ID_COLOR_BACK
|
544 BUF_3D_PITCH(color_region
->pitch
* color_region
->cpp
) |
549 state
->Buffer
[I915_DESTREG_DBUFADDR0
] = _3DSTATE_BUF_INFO_CMD
;
550 state
->Buffer
[I915_DESTREG_DBUFADDR1
] =
552 BUF_3D_PITCH(depth_region
->pitch
* depth_region
->cpp
) |
557 * Compute/set I915_DESTREG_DV1 value
559 value
= (DSTORG_HORT_BIAS(0x8) | /* .5 */
560 DSTORG_VERT_BIAS(0x8) | /* .5 */
561 LOD_PRECLAMP_OGL
| TEX_DEFAULT_COLOR_OGL
);
562 if (color_region
&& color_region
->cpp
== 4) {
566 value
|= (DITHER_FULL_ALWAYS
| DV_PF_565
);
568 if (depth_region
&& depth_region
->cpp
== 4) {
569 value
|= DEPTH_FRMT_24_FIXED_8_OTHER
;
572 value
|= DEPTH_FRMT_16_FIXED
;
574 state
->Buffer
[I915_DESTREG_DV1
] = value
;
576 I915_STATECHANGE(i915
, I915_UPLOAD_BUFFERS
);
581 i915_set_draw_region(struct intel_context
*intel
,
582 struct intel_region
*color_regions
[],
583 struct intel_region
*depth_region
,
586 struct i915_context
*i915
= i915_context(&intel
->ctx
);
587 i915_state_draw_region(intel
, &i915
->state
, color_regions
[0], depth_region
);
593 i915_new_batch(struct intel_context
*intel
)
595 struct i915_context
*i915
= i915_context(&intel
->ctx
);
597 /* Mark all state as needing to be emitted when starting a new batchbuffer.
598 * Using hardware contexts would be an alternative, but they have some
599 * difficulties associated with them (physical address requirements).
601 i915
->state
.emitted
= 0;
603 /* Check that we didn't just wrap our batchbuffer at a bad time. */
604 assert(!intel
->no_batch_wrap
);
610 return MI_FLUSH
| FLUSH_MAP_CACHE
;
614 i915_assert_not_dirty( struct intel_context
*intel
)
616 struct i915_context
*i915
= i915_context(&intel
->ctx
);
617 struct i915_hw_state
*state
= i915
->current
;
618 GLuint dirty
= get_dirty(state
);
623 i915_note_unlock( struct intel_context
*intel
)
630 i915InitVtbl(struct i915_context
*i915
)
632 i915
->intel
.vtbl
.check_vertex_size
= i915_check_vertex_size
;
633 i915
->intel
.vtbl
.destroy
= i915_destroy_context
;
634 i915
->intel
.vtbl
.emit_state
= i915_emit_state
;
635 i915
->intel
.vtbl
.new_batch
= i915_new_batch
;
636 i915
->intel
.vtbl
.reduced_primitive_state
= i915_reduced_primitive_state
;
637 i915
->intel
.vtbl
.render_start
= i915_render_start
;
638 i915
->intel
.vtbl
.render_prevalidate
= i915_render_prevalidate
;
639 i915
->intel
.vtbl
.set_draw_region
= i915_set_draw_region
;
640 i915
->intel
.vtbl
.update_texture_state
= i915UpdateTextureState
;
641 i915
->intel
.vtbl
.flush_cmd
= i915_flush_cmd
;
642 i915
->intel
.vtbl
.assert_not_dirty
= i915_assert_not_dirty
;
643 i915
->intel
.vtbl
.note_unlock
= i915_note_unlock
;
644 i915
->intel
.vtbl
.finish_batch
= intel_finish_vb
;