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 **************************************************************************/
28 /** @file intel_tris.c
30 * This file contains functions for managing the vertex buffer and emitting
34 #include "main/glheader.h"
35 #include "main/context.h"
36 #include "main/macros.h"
37 #include "main/enums.h"
38 #include "main/texobj.h"
39 #include "main/state.h"
41 #include "main/fbobject.h"
43 #include "swrast/swrast.h"
44 #include "swrast_setup/swrast_setup.h"
45 #include "tnl/t_context.h"
46 #include "tnl/t_pipeline.h"
47 #include "tnl/t_vertex.h"
49 #include "intel_screen.h"
50 #include "intel_context.h"
51 #include "intel_tris.h"
52 #include "intel_batchbuffer.h"
53 #include "intel_buffers.h"
54 #include "intel_reg.h"
55 #include "intel_span.h"
56 #include "i830_context.h"
58 #include "i915_context.h"
60 static void intelRenderPrimitive(struct gl_context
* ctx
, GLenum prim
);
61 static void intelRasterPrimitive(struct gl_context
* ctx
, GLenum rprim
,
65 intel_flush_inline_primitive(struct intel_context
*intel
)
67 GLuint used
= intel
->batch
.used
- intel
->prim
.start_ptr
;
69 assert(intel
->prim
.primitive
!= ~0);
76 intel
->batch
.map
[intel
->prim
.start_ptr
] =
77 _3DPRIMITIVE
| intel
->prim
.primitive
| (used
- 2);
82 intel
->batch
.used
= intel
->prim
.start_ptr
;
85 intel
->prim
.primitive
= ~0;
86 intel
->prim
.start_ptr
= 0;
87 intel
->prim
.flush
= 0;
90 static void intel_start_inline(struct intel_context
*intel
, uint32_t prim
)
94 intel
->vtbl
.emit_state(intel
);
96 intel
->no_batch_wrap
= true;
98 /*printf("%s *", __progname);*/
100 /* Emit a slot which will be filled with the inline primitive
105 intel
->prim
.start_ptr
= intel
->batch
.used
;
106 intel
->prim
.primitive
= prim
;
107 intel
->prim
.flush
= intel_flush_inline_primitive
;
112 intel
->no_batch_wrap
= false;
116 static void intel_wrap_inline(struct intel_context
*intel
)
118 GLuint prim
= intel
->prim
.primitive
;
120 intel_flush_inline_primitive(intel
);
121 intel_batchbuffer_flush(intel
);
122 intel_start_inline(intel
, prim
); /* ??? */
125 static GLuint
*intel_extend_inline(struct intel_context
*intel
, GLuint dwords
)
129 assert(intel
->prim
.flush
== intel_flush_inline_primitive
);
131 if (intel_batchbuffer_space(intel
) < dwords
* sizeof(GLuint
))
132 intel_wrap_inline(intel
);
136 intel
->vtbl
.assert_not_dirty(intel
);
138 ptr
= intel
->batch
.map
+ intel
->batch
.used
;
139 intel
->batch
.used
+= dwords
;
144 /** Sets the primitive type for a primitive sequence, flushing as needed. */
145 void intel_set_prim(struct intel_context
*intel
, uint32_t prim
)
147 /* if we have no VBOs */
149 if (intel
->intelScreen
->no_vbo
) {
150 intel_start_inline(intel
, prim
);
153 if (prim
!= intel
->prim
.primitive
) {
154 INTEL_FIREVERTICES(intel
);
155 intel
->prim
.primitive
= prim
;
159 /** Returns mapped VB space for the given number of vertices */
160 uint32_t *intel_get_prim_space(struct intel_context
*intel
, unsigned int count
)
164 if (intel
->intelScreen
->no_vbo
) {
165 return intel_extend_inline(intel
, count
* intel
->vertex_size
);
168 /* Check for space in the existing VB */
169 if (intel
->prim
.vb_bo
== NULL
||
170 (intel
->prim
.current_offset
+
171 count
* intel
->vertex_size
* 4) > INTEL_VB_SIZE
||
172 (intel
->prim
.count
+ count
) >= (1 << 16)) {
173 /* Flush existing prim if any */
174 INTEL_FIREVERTICES(intel
);
176 intel_finish_vb(intel
);
179 if (intel
->prim
.vb
== NULL
)
180 intel
->prim
.vb
= malloc(INTEL_VB_SIZE
);
181 intel
->prim
.vb_bo
= drm_intel_bo_alloc(intel
->bufmgr
, "vb",
183 intel
->prim
.start_offset
= 0;
184 intel
->prim
.current_offset
= 0;
187 intel
->prim
.flush
= intel_flush_prim
;
189 addr
= (uint32_t *)(intel
->prim
.vb
+ intel
->prim
.current_offset
);
190 intel
->prim
.current_offset
+= intel
->vertex_size
* 4 * count
;
191 intel
->prim
.count
+= count
;
196 /** Dispatches the accumulated primitive to the batchbuffer. */
197 void intel_flush_prim(struct intel_context
*intel
)
199 drm_intel_bo
*aper_array
[2];
201 unsigned int offset
, count
;
204 /* Must be called after an intel_start_prim. */
205 assert(intel
->prim
.primitive
!= ~0);
207 if (intel
->prim
.count
== 0)
210 /* Clear the current prims out of the context state so that a batch flush
211 * flush triggered by emit_state doesn't loop back to flush_prim again.
213 vb_bo
= intel
->prim
.vb_bo
;
214 drm_intel_bo_reference(vb_bo
);
215 count
= intel
->prim
.count
;
216 intel
->prim
.count
= 0;
217 offset
= intel
->prim
.start_offset
;
218 intel
->prim
.start_offset
= intel
->prim
.current_offset
;
220 intel
->prim
.current_offset
= intel
->prim
.start_offset
= ALIGN(intel
->prim
.start_offset
, 128);
221 intel
->prim
.flush
= NULL
;
223 intel
->vtbl
.emit_state(intel
);
225 aper_array
[0] = intel
->batch
.bo
;
226 aper_array
[1] = vb_bo
;
227 if (dri_bufmgr_check_aperture_space(aper_array
, 2)) {
228 intel_batchbuffer_flush(intel
);
229 intel
->vtbl
.emit_state(intel
);
232 /* Ensure that we don't start a new batch for the following emit, which
233 * depends on the state just emitted. emit_state should be making sure we
234 * have the space for this.
236 intel
->no_batch_wrap
= true;
238 if (intel
->always_flush_cache
) {
239 intel_batchbuffer_emit_mi_flush(intel
);
243 printf("emitting %d..%d=%d vertices size %d\n", offset
,
244 intel
->prim
.current_offset
, count
,
245 intel
->vertex_size
* 4);
248 if (intel
->gen
>= 3) {
249 struct i915_context
*i915
= i915_context(&intel
->ctx
);
250 unsigned int cmd
= 0, len
= 0;
252 if (vb_bo
!= i915
->current_vb_bo
) {
257 if (intel
->vertex_size
!= i915
->current_vertex_size
) {
266 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1
| cmd
| (len
- 2));
267 if (vb_bo
!= i915
->current_vb_bo
) {
268 OUT_RELOC(vb_bo
, I915_GEM_DOMAIN_VERTEX
, 0, 0);
269 i915
->current_vb_bo
= vb_bo
;
271 if (intel
->vertex_size
!= i915
->current_vertex_size
) {
272 OUT_BATCH((intel
->vertex_size
<< S1_VERTEX_WIDTH_SHIFT
) |
273 (intel
->vertex_size
<< S1_VERTEX_PITCH_SHIFT
));
274 i915
->current_vertex_size
= intel
->vertex_size
;
276 OUT_BATCH(_3DPRIMITIVE
|
278 PRIM_INDIRECT_SEQUENTIAL
|
279 intel
->prim
.primitive
|
281 OUT_BATCH(offset
/ (intel
->vertex_size
* 4));
284 struct i830_context
*i830
= i830_context(&intel
->ctx
);
287 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1
|
288 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
290 assert((offset
& ~S0_VB_OFFSET_MASK_830
) == 0);
291 OUT_RELOC(vb_bo
, I915_GEM_DOMAIN_VERTEX
, 0,
292 offset
| (intel
->vertex_size
<< S0_VB_PITCH_SHIFT_830
) |
295 * This is somewhat unfortunate -- VB width is tied up with
296 * vertex format data that we've already uploaded through
297 * _3DSTATE_VFT[01]_CMD. We may want to replace emits of VFT state with
298 * STATE_IMMEDIATE_1 like this to avoid duplication.
300 OUT_BATCH((i830
->state
.Ctx
[I830_CTXREG_VF
] & VFT0_TEX_COUNT_MASK
) >>
301 VFT0_TEX_COUNT_SHIFT
<< S2_TEX_COUNT_SHIFT_830
|
302 (i830
->state
.Ctx
[I830_CTXREG_VF2
] << 16) |
303 intel
->vertex_size
<< S2_VERTEX_0_WIDTH_SHIFT_830
);
305 OUT_BATCH(_3DPRIMITIVE
|
307 PRIM_INDIRECT_SEQUENTIAL
|
308 intel
->prim
.primitive
|
310 OUT_BATCH(0); /* Beginning vertex index */
314 if (intel
->always_flush_cache
) {
315 intel_batchbuffer_emit_mi_flush(intel
);
318 intel
->no_batch_wrap
= false;
320 drm_intel_bo_unreference(vb_bo
);
324 * Uploads the locally-accumulated VB into the buffer object.
326 * This avoids us thrashing the cachelines in and out as the buffer gets
327 * filled, dispatched, then reused as the hardware completes rendering from it,
328 * and also lets us clflush less if we dispatch with a partially-filled VB.
330 * This is called normally from get_space when we're finishing a BO, but also
331 * at batch flush time so that we don't try accessing the contents of a
332 * just-dispatched buffer.
334 void intel_finish_vb(struct intel_context
*intel
)
336 if (intel
->prim
.vb_bo
== NULL
)
339 drm_intel_bo_subdata(intel
->prim
.vb_bo
, 0, intel
->prim
.start_offset
,
341 drm_intel_bo_unreference(intel
->prim
.vb_bo
);
342 intel
->prim
.vb_bo
= NULL
;
345 /***********************************************************************
346 * Emit primitives as inline vertices *
347 ***********************************************************************/
350 #define COPY_DWORDS( j, vb, vertsize, v ) \
353 __asm__ __volatile__( "rep ; movsl" \
354 : "=%c" (j), "=D" (vb), "=S" (__tmp) \
360 #define COPY_DWORDS( j, vb, vertsize, v ) \
362 for ( j = 0 ; j < vertsize ; j++ ) { \
363 vb[j] = ((GLuint *)v)[j]; \
370 intel_draw_quad(struct intel_context
*intel
,
372 intelVertexPtr v1
, intelVertexPtr v2
, intelVertexPtr v3
)
374 GLuint vertsize
= intel
->vertex_size
;
375 GLuint
*vb
= intel_get_prim_space(intel
, 6);
378 COPY_DWORDS(j
, vb
, vertsize
, v0
);
379 COPY_DWORDS(j
, vb
, vertsize
, v1
);
381 /* If smooth shading, draw like a trifan which gives better
382 * rasterization. Otherwise draw as two triangles with provoking
383 * vertex in third position as required for flat shading.
385 if (intel
->ctx
.Light
.ShadeModel
== GL_FLAT
) {
386 COPY_DWORDS(j
, vb
, vertsize
, v3
);
387 COPY_DWORDS(j
, vb
, vertsize
, v1
);
390 COPY_DWORDS(j
, vb
, vertsize
, v2
);
391 COPY_DWORDS(j
, vb
, vertsize
, v0
);
394 COPY_DWORDS(j
, vb
, vertsize
, v2
);
395 COPY_DWORDS(j
, vb
, vertsize
, v3
);
399 intel_draw_triangle(struct intel_context
*intel
,
400 intelVertexPtr v0
, intelVertexPtr v1
, intelVertexPtr v2
)
402 GLuint vertsize
= intel
->vertex_size
;
403 GLuint
*vb
= intel_get_prim_space(intel
, 3);
406 COPY_DWORDS(j
, vb
, vertsize
, v0
);
407 COPY_DWORDS(j
, vb
, vertsize
, v1
);
408 COPY_DWORDS(j
, vb
, vertsize
, v2
);
413 intel_draw_line(struct intel_context
*intel
,
414 intelVertexPtr v0
, intelVertexPtr v1
)
416 GLuint vertsize
= intel
->vertex_size
;
417 GLuint
*vb
= intel_get_prim_space(intel
, 2);
420 COPY_DWORDS(j
, vb
, vertsize
, v0
);
421 COPY_DWORDS(j
, vb
, vertsize
, v1
);
426 intel_draw_point(struct intel_context
*intel
, intelVertexPtr v0
)
428 GLuint vertsize
= intel
->vertex_size
;
429 GLuint
*vb
= intel_get_prim_space(intel
, 1);
432 /* Adjust for sub pixel position -- still required for conform. */
433 *(float *) &vb
[0] = v0
->v
.x
;
434 *(float *) &vb
[1] = v0
->v
.y
;
435 for (j
= 2; j
< vertsize
; j
++)
441 /***********************************************************************
442 * Fixup for ARB_point_parameters *
443 ***********************************************************************/
445 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
446 * represented in the fragment program InputsRead field.
449 intel_atten_point(struct intel_context
*intel
, intelVertexPtr v0
)
451 struct gl_context
*ctx
= &intel
->ctx
;
452 GLfloat psz
[4], col
[4], restore_psz
, restore_alpha
;
454 _tnl_get_attr(ctx
, v0
, _TNL_ATTRIB_POINTSIZE
, psz
);
455 _tnl_get_attr(ctx
, v0
, _TNL_ATTRIB_COLOR0
, col
);
457 restore_psz
= psz
[0];
458 restore_alpha
= col
[3];
460 if (psz
[0] >= ctx
->Point
.Threshold
) {
461 psz
[0] = MIN2(psz
[0], ctx
->Point
.MaxSize
);
464 GLfloat dsize
= psz
[0] / ctx
->Point
.Threshold
;
465 psz
[0] = MAX2(ctx
->Point
.Threshold
, ctx
->Point
.MinSize
);
466 col
[3] *= dsize
* dsize
;
472 if (restore_psz
!= psz
[0] || restore_alpha
!= col
[3]) {
473 _tnl_set_attr(ctx
, v0
, _TNL_ATTRIB_POINTSIZE
, psz
);
474 _tnl_set_attr(ctx
, v0
, _TNL_ATTRIB_COLOR0
, col
);
476 intel_draw_point(intel
, v0
);
478 psz
[0] = restore_psz
;
479 col
[3] = restore_alpha
;
481 _tnl_set_attr(ctx
, v0
, _TNL_ATTRIB_POINTSIZE
, psz
);
482 _tnl_set_attr(ctx
, v0
, _TNL_ATTRIB_COLOR0
, col
);
485 intel_draw_point(intel
, v0
);
492 /***********************************************************************
493 * Fixup for I915 WPOS texture coordinate *
494 ***********************************************************************/
497 intel_emit_fragcoord(struct intel_context
*intel
, intelVertexPtr v
)
499 struct gl_context
*ctx
= &intel
->ctx
;
500 struct gl_framebuffer
*fb
= ctx
->DrawBuffer
;
501 GLuint offset
= intel
->wpos_offset
;
502 float *vertex_position
= (float *)v
;
503 float *fragcoord
= (float *)((char *)v
+ offset
);
505 fragcoord
[0] = vertex_position
[0];
507 if (_mesa_is_user_fbo(fb
))
508 fragcoord
[1] = vertex_position
[1];
510 fragcoord
[1] = fb
->Height
- vertex_position
[1];
512 fragcoord
[2] = vertex_position
[2];
513 fragcoord
[3] = vertex_position
[3];
517 intel_wpos_triangle(struct intel_context
*intel
,
518 intelVertexPtr v0
, intelVertexPtr v1
, intelVertexPtr v2
)
520 intel_emit_fragcoord(intel
, v0
);
521 intel_emit_fragcoord(intel
, v1
);
522 intel_emit_fragcoord(intel
, v2
);
524 intel_draw_triangle(intel
, v0
, v1
, v2
);
529 intel_wpos_line(struct intel_context
*intel
,
530 intelVertexPtr v0
, intelVertexPtr v1
)
532 intel_emit_fragcoord(intel
, v0
);
533 intel_emit_fragcoord(intel
, v1
);
534 intel_draw_line(intel
, v0
, v1
);
539 intel_wpos_point(struct intel_context
*intel
, intelVertexPtr v0
)
541 intel_emit_fragcoord(intel
, v0
);
542 intel_draw_point(intel
, v0
);
550 /***********************************************************************
551 * Macros for t_dd_tritmp.h to draw basic primitives *
552 ***********************************************************************/
554 #define TRI( a, b, c ) \
557 intel->draw_tri( intel, a, b, c ); \
559 intel_draw_triangle( intel, a, b, c ); \
562 #define QUAD( a, b, c, d ) \
565 intel->draw_tri( intel, a, b, d ); \
566 intel->draw_tri( intel, b, c, d ); \
568 intel_draw_quad( intel, a, b, c, d ); \
571 #define LINE( v0, v1 ) \
574 intel->draw_line( intel, v0, v1 ); \
576 intel_draw_line( intel, v0, v1 ); \
579 #define POINT( v0 ) \
582 intel->draw_point( intel, v0 ); \
584 intel_draw_point( intel, v0 ); \
588 /***********************************************************************
589 * Build render functions from dd templates *
590 ***********************************************************************/
592 #define INTEL_OFFSET_BIT 0x01
593 #define INTEL_TWOSIDE_BIT 0x02
594 #define INTEL_UNFILLED_BIT 0x04
595 #define INTEL_FALLBACK_BIT 0x08
596 #define INTEL_MAX_TRIFUNC 0x10
601 tnl_points_func points
;
603 tnl_triangle_func triangle
;
605 } rast_tab
[INTEL_MAX_TRIFUNC
];
608 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
609 #define DO_OFFSET (IND & INTEL_OFFSET_BIT)
610 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
611 #define DO_TWOSIDE (IND & INTEL_TWOSIDE_BIT)
617 #define DO_FULL_QUAD 1
620 #define HAVE_BACK_COLORS 0
621 #define HAVE_HW_FLATSHADE 1
622 #define VERTEX intelVertex
625 /* Only used to pull back colors into vertices (ie, we know color is
628 #define INTEL_COLOR( dst, src ) \
630 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
631 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
632 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
633 UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]); \
636 #define INTEL_SPEC( dst, src ) \
638 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
639 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
640 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
644 #define DEPTH_SCALE intel->polygon_offset_scale
645 #define UNFILLED_TRI unfilled_tri
646 #define UNFILLED_QUAD unfilled_quad
647 #define VERT_X(_v) _v->v.x
648 #define VERT_Y(_v) _v->v.y
649 #define VERT_Z(_v) _v->v.z
650 #define AREA_IS_CCW( a ) (a > 0)
651 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
653 #define VERT_SET_RGBA( v, c ) if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
654 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
655 #define VERT_SAVE_RGBA( idx ) if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
656 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
658 #define VERT_SET_SPEC( v, c ) if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
659 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
660 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
661 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
663 #define LOCAL_VARS(n) \
664 struct intel_context *intel = intel_context(ctx); \
665 GLuint color[n] = { 0, }, spec[n] = { 0, }; \
666 GLuint coloroffset = intel->coloroffset; \
667 GLuint specoffset = intel->specoffset; \
668 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
671 /***********************************************************************
672 * Helpers for rendering unfilled primitives *
673 ***********************************************************************/
675 static const GLuint hw_prim
[GL_POLYGON
+ 1] = {
688 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
689 #define RENDER_PRIMITIVE intel->render_primitive
691 #define IND INTEL_FALLBACK_BIT
692 #include "tnl_dd/t_dd_unfilled.h"
695 /***********************************************************************
696 * Generate GL render functions *
697 ***********************************************************************/
701 #include "tnl_dd/t_dd_tritmp.h"
703 #define IND (INTEL_OFFSET_BIT)
704 #define TAG(x) x##_offset
705 #include "tnl_dd/t_dd_tritmp.h"
707 #define IND (INTEL_TWOSIDE_BIT)
708 #define TAG(x) x##_twoside
709 #include "tnl_dd/t_dd_tritmp.h"
711 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
712 #define TAG(x) x##_twoside_offset
713 #include "tnl_dd/t_dd_tritmp.h"
715 #define IND (INTEL_UNFILLED_BIT)
716 #define TAG(x) x##_unfilled
717 #include "tnl_dd/t_dd_tritmp.h"
719 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
720 #define TAG(x) x##_offset_unfilled
721 #include "tnl_dd/t_dd_tritmp.h"
723 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
724 #define TAG(x) x##_twoside_unfilled
725 #include "tnl_dd/t_dd_tritmp.h"
727 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
728 #define TAG(x) x##_twoside_offset_unfilled
729 #include "tnl_dd/t_dd_tritmp.h"
731 #define IND (INTEL_FALLBACK_BIT)
732 #define TAG(x) x##_fallback
733 #include "tnl_dd/t_dd_tritmp.h"
735 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
736 #define TAG(x) x##_offset_fallback
737 #include "tnl_dd/t_dd_tritmp.h"
739 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
740 #define TAG(x) x##_twoside_fallback
741 #include "tnl_dd/t_dd_tritmp.h"
743 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
744 #define TAG(x) x##_twoside_offset_fallback
745 #include "tnl_dd/t_dd_tritmp.h"
747 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
748 #define TAG(x) x##_unfilled_fallback
749 #include "tnl_dd/t_dd_tritmp.h"
751 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
752 #define TAG(x) x##_offset_unfilled_fallback
753 #include "tnl_dd/t_dd_tritmp.h"
755 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
756 #define TAG(x) x##_twoside_unfilled_fallback
757 #include "tnl_dd/t_dd_tritmp.h"
759 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
761 #define TAG(x) x##_twoside_offset_unfilled_fallback
762 #include "tnl_dd/t_dd_tritmp.h"
771 init_twoside_offset();
773 init_offset_unfilled();
774 init_twoside_unfilled();
775 init_twoside_offset_unfilled();
777 init_offset_fallback();
778 init_twoside_fallback();
779 init_twoside_offset_fallback();
780 init_unfilled_fallback();
781 init_offset_unfilled_fallback();
782 init_twoside_unfilled_fallback();
783 init_twoside_offset_unfilled_fallback();
787 /***********************************************************************
788 * Rasterization fallback helpers *
789 ***********************************************************************/
792 /* This code is hit only when a mix of accelerated and unaccelerated
793 * primitives are being drawn, and only for the unaccelerated
797 intel_fallback_tri(struct intel_context
*intel
,
798 intelVertex
* v0
, intelVertex
* v1
, intelVertex
* v2
)
800 struct gl_context
*ctx
= &intel
->ctx
;
804 fprintf(stderr
, "\n%s\n", __FUNCTION__
);
806 INTEL_FIREVERTICES(intel
);
808 _swsetup_Translate(ctx
, v0
, &v
[0]);
809 _swsetup_Translate(ctx
, v1
, &v
[1]);
810 _swsetup_Translate(ctx
, v2
, &v
[2]);
811 intelSpanRenderStart(ctx
);
812 _swrast_Triangle(ctx
, &v
[0], &v
[1], &v
[2]);
813 intelSpanRenderFinish(ctx
);
818 intel_fallback_line(struct intel_context
*intel
,
819 intelVertex
* v0
, intelVertex
* v1
)
821 struct gl_context
*ctx
= &intel
->ctx
;
825 fprintf(stderr
, "\n%s\n", __FUNCTION__
);
827 INTEL_FIREVERTICES(intel
);
829 _swsetup_Translate(ctx
, v0
, &v
[0]);
830 _swsetup_Translate(ctx
, v1
, &v
[1]);
831 intelSpanRenderStart(ctx
);
832 _swrast_Line(ctx
, &v
[0], &v
[1]);
833 intelSpanRenderFinish(ctx
);
837 intel_fallback_point(struct intel_context
*intel
,
840 struct gl_context
*ctx
= &intel
->ctx
;
844 fprintf(stderr
, "\n%s\n", __FUNCTION__
);
846 INTEL_FIREVERTICES(intel
);
848 _swsetup_Translate(ctx
, v0
, &v
[0]);
849 intelSpanRenderStart(ctx
);
850 _swrast_Point(ctx
, &v
[0]);
851 intelSpanRenderFinish(ctx
);
855 /**********************************************************************/
856 /* Render unclipped begin/end objects */
857 /**********************************************************************/
860 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
861 #define RENDER_POINTS( start, count ) \
862 for ( ; start < count ; start++) POINT( V(ELT(start)) );
863 #define RENDER_LINE( v0, v1 ) LINE( V(v0), V(v1) )
864 #define RENDER_TRI( v0, v1, v2 ) TRI( V(v0), V(v1), V(v2) )
865 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
866 #define INIT(x) intelRenderPrimitive( ctx, x )
869 struct intel_context *intel = intel_context(ctx); \
870 GLubyte *vertptr = (GLubyte *)intel->verts; \
871 const GLuint vertsize = intel->vertex_size; \
872 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
874 #define RESET_STIPPLE
875 #define RESET_OCCLUSION
876 #define PRESERVE_VB_DEFS
878 #define TAG(x) intel_##x##_verts
879 #include "tnl/t_vb_rendertmp.h"
882 #define TAG(x) intel_##x##_elts
883 #define ELT(x) elt[x]
884 #include "tnl/t_vb_rendertmp.h"
886 /**********************************************************************/
887 /* Render clipped primitives */
888 /**********************************************************************/
893 intelRenderClippedPoly(struct gl_context
* ctx
, const GLuint
* elts
, GLuint n
)
895 struct intel_context
*intel
= intel_context(ctx
);
896 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
897 struct vertex_buffer
*VB
= &TNL_CONTEXT(ctx
)->vb
;
898 GLuint prim
= intel
->render_primitive
;
900 /* Render the new vertices as an unclipped polygon.
903 GLuint
*tmp
= VB
->Elts
;
904 VB
->Elts
= (GLuint
*) elts
;
905 tnl
->Driver
.Render
.PrimTabElts
[GL_POLYGON
] (ctx
, 0, n
,
906 PRIM_BEGIN
| PRIM_END
);
910 /* Restore the render primitive
912 if (prim
!= GL_POLYGON
)
913 tnl
->Driver
.Render
.PrimitiveNotify(ctx
, prim
);
917 intelRenderClippedLine(struct gl_context
* ctx
, GLuint ii
, GLuint jj
)
919 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
921 tnl
->Driver
.Render
.Line(ctx
, ii
, jj
);
925 intelFastRenderClippedPoly(struct gl_context
* ctx
, const GLuint
* elts
, GLuint n
)
927 struct intel_context
*intel
= intel_context(ctx
);
928 const GLuint vertsize
= intel
->vertex_size
;
929 GLuint
*vb
= intel_get_prim_space(intel
, (n
- 2) * 3);
930 GLubyte
*vertptr
= (GLubyte
*) intel
->verts
;
931 const GLuint
*start
= (const GLuint
*) V(elts
[0]);
934 for (i
= 2; i
< n
; i
++) {
935 COPY_DWORDS(j
, vb
, vertsize
, V(elts
[i
- 1]));
936 COPY_DWORDS(j
, vb
, vertsize
, V(elts
[i
]));
937 COPY_DWORDS(j
, vb
, vertsize
, start
);
941 /**********************************************************************/
942 /* Choose render functions */
943 /**********************************************************************/
947 #define DD_LINE_STIPPLE (1 << 7)
948 #define DD_POINT_ATTEN (1 << 9)
950 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN)
951 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
954 intelChooseRenderState(struct gl_context
* ctx
)
956 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
957 struct intel_context
*intel
= intel_context(ctx
);
958 GLuint flags
= ctx
->_TriangleCaps
|
959 (ctx
->Line
.StippleFlag
? DD_LINE_STIPPLE
: 0) |
960 (ctx
->Point
._Attenuated
? DD_POINT_ATTEN
: 0);
961 const struct gl_fragment_program
*fprog
= ctx
->FragmentProgram
._Current
;
962 bool have_wpos
= (fprog
&& (fprog
->Base
.InputsRead
& VARYING_BIT_POS
));
965 if (INTEL_DEBUG
& DEBUG_STATE
)
966 fprintf(stderr
, "\n%s\n", __FUNCTION__
);
968 if ((flags
& (ANY_FALLBACK_FLAGS
| ANY_RASTER_FLAGS
)) || have_wpos
) {
970 if (flags
& ANY_RASTER_FLAGS
) {
971 if (flags
& DD_TRI_LIGHT_TWOSIDE
)
972 index
|= INTEL_TWOSIDE_BIT
;
973 if (flags
& DD_TRI_OFFSET
)
974 index
|= INTEL_OFFSET_BIT
;
975 if (flags
& DD_TRI_UNFILLED
)
976 index
|= INTEL_UNFILLED_BIT
;
980 intel
->draw_point
= intel_wpos_point
;
981 intel
->draw_line
= intel_wpos_line
;
982 intel
->draw_tri
= intel_wpos_triangle
;
984 /* Make sure these get called:
986 index
|= INTEL_FALLBACK_BIT
;
989 intel
->draw_point
= intel_draw_point
;
990 intel
->draw_line
= intel_draw_line
;
991 intel
->draw_tri
= intel_draw_triangle
;
994 /* Hook in fallbacks for specific primitives.
996 if (flags
& ANY_FALLBACK_FLAGS
) {
997 if (flags
& DD_LINE_STIPPLE
)
998 intel
->draw_line
= intel_fallback_line
;
1000 if ((flags
& DD_TRI_STIPPLE
) && !intel
->hw_stipple
)
1001 intel
->draw_tri
= intel_fallback_tri
;
1003 if (flags
& DD_POINT_ATTEN
) {
1005 intel
->draw_point
= intel_atten_point
;
1007 intel
->draw_point
= intel_fallback_point
;
1010 index
|= INTEL_FALLBACK_BIT
;
1014 if (intel
->RenderIndex
!= index
) {
1015 intel
->RenderIndex
= index
;
1017 tnl
->Driver
.Render
.Points
= rast_tab
[index
].points
;
1018 tnl
->Driver
.Render
.Line
= rast_tab
[index
].line
;
1019 tnl
->Driver
.Render
.Triangle
= rast_tab
[index
].triangle
;
1020 tnl
->Driver
.Render
.Quad
= rast_tab
[index
].quad
;
1023 tnl
->Driver
.Render
.PrimTabVerts
= intel_render_tab_verts
;
1024 tnl
->Driver
.Render
.PrimTabElts
= intel_render_tab_elts
;
1025 tnl
->Driver
.Render
.ClippedLine
= line
; /* from tritmp.h */
1026 tnl
->Driver
.Render
.ClippedPolygon
= intelFastRenderClippedPoly
;
1029 tnl
->Driver
.Render
.PrimTabVerts
= _tnl_render_tab_verts
;
1030 tnl
->Driver
.Render
.PrimTabElts
= _tnl_render_tab_elts
;
1031 tnl
->Driver
.Render
.ClippedLine
= intelRenderClippedLine
;
1032 tnl
->Driver
.Render
.ClippedPolygon
= intelRenderClippedPoly
;
1037 static const GLenum reduced_prim
[GL_POLYGON
+ 1] = {
1051 /**********************************************************************/
1052 /* High level hooks for t_vb_render.c */
1053 /**********************************************************************/
1059 intelRunPipeline(struct gl_context
* ctx
)
1061 struct intel_context
*intel
= intel_context(ctx
);
1063 _mesa_lock_context_textures(ctx
);
1066 _mesa_update_state_locked(ctx
);
1068 /* We need to get this done before we start the pipeline, or a
1069 * change in the INTEL_FALLBACK() of its intel_draw_buffers() call
1070 * while the pipeline is running will result in mismatched swrast
1071 * map/unmaps, and later assertion failures.
1073 intel_prepare_render(intel
);
1075 if (intel
->NewGLState
) {
1076 if (intel
->NewGLState
& _NEW_TEXTURE
) {
1077 intel
->vtbl
.update_texture_state(intel
);
1080 if (!intel
->Fallback
) {
1081 if (intel
->NewGLState
& _INTEL_NEW_RENDERSTATE
)
1082 intelChooseRenderState(ctx
);
1085 intel
->NewGLState
= 0;
1088 intel_map_vertex_shader_textures(ctx
);
1089 intel
->tnl_pipeline_running
= true;
1090 _tnl_run_pipeline(ctx
);
1091 intel
->tnl_pipeline_running
= false;
1092 intel_unmap_vertex_shader_textures(ctx
);
1094 _mesa_unlock_context_textures(ctx
);
1098 intelRenderStart(struct gl_context
* ctx
)
1100 struct intel_context
*intel
= intel_context(ctx
);
1102 intel_check_front_buffer_rendering(intel
);
1103 intel
->vtbl
.render_start(intel_context(ctx
));
1104 intel
->vtbl
.emit_state(intel
);
1108 intelRenderFinish(struct gl_context
* ctx
)
1110 struct intel_context
*intel
= intel_context(ctx
);
1112 if (intel
->RenderIndex
& INTEL_FALLBACK_BIT
)
1115 INTEL_FIREVERTICES(intel
);
1121 /* System to flush dma and emit state changes based on the rasterized
1125 intelRasterPrimitive(struct gl_context
* ctx
, GLenum rprim
, GLuint hwprim
)
1127 struct intel_context
*intel
= intel_context(ctx
);
1130 fprintf(stderr
, "%s %s %x\n", __FUNCTION__
,
1131 _mesa_lookup_enum_by_nr(rprim
), hwprim
);
1133 intel
->vtbl
.reduced_primitive_state(intel
, rprim
);
1135 /* Start a new primitive. Arrange to have it flushed later on.
1137 if (hwprim
!= intel
->prim
.primitive
) {
1138 INTEL_FIREVERTICES(intel
);
1140 intel_set_prim(intel
, hwprim
);
1148 intelRenderPrimitive(struct gl_context
* ctx
, GLenum prim
)
1150 struct intel_context
*intel
= intel_context(ctx
);
1153 fprintf(stderr
, "%s %s\n", __FUNCTION__
, _mesa_lookup_enum_by_nr(prim
));
1155 /* Let some clipping routines know which primitive they're dealing
1158 intel
->render_primitive
= prim
;
1160 /* Shortcircuit this when called for unfilled triangles. The rasterized
1161 * primitive will always be reset by lower level functions in that case,
1162 * potentially pingponging the state:
1164 if (reduced_prim
[prim
] == GL_TRIANGLES
&&
1165 (ctx
->_TriangleCaps
& DD_TRI_UNFILLED
))
1168 /* Set some primitive-dependent state and Start? a new primitive.
1170 intelRasterPrimitive(ctx
, reduced_prim
[prim
], hw_prim
[prim
]);
1174 /**********************************************************************/
1175 /* Transition to/from hardware rasterization. */
1176 /**********************************************************************/
1178 static char *fallbackStrings
[] = {
1179 [0] = "Draw buffer",
1180 [1] = "Read buffer",
1181 [2] = "Depth buffer",
1182 [3] = "Stencil buffer",
1183 [4] = "User disable",
1184 [5] = "Render mode",
1187 [13] = "Color mask",
1192 [18] = "Smooth polygon",
1193 [19] = "Smooth point",
1194 [20] = "point sprite coord origin",
1195 [21] = "depth/color drawing offset",
1196 [22] = "coord replace(SPRITE POINT ENABLE)",
1201 getFallbackString(GLuint bit
)
1208 return fallbackStrings
[i
];
1214 * Enable/disable a fallback flag.
1215 * \param bit one of INTEL_FALLBACK_x flags.
1218 intelFallback(struct intel_context
*intel
, GLbitfield bit
, bool mode
)
1220 struct gl_context
*ctx
= &intel
->ctx
;
1221 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
1222 const GLbitfield oldfallback
= intel
->Fallback
;
1225 intel
->Fallback
|= bit
;
1226 if (oldfallback
== 0) {
1227 assert(!intel
->tnl_pipeline_running
);
1230 if (INTEL_DEBUG
& DEBUG_PERF
)
1231 fprintf(stderr
, "ENTER FALLBACK %x: %s\n",
1232 bit
, getFallbackString(bit
));
1233 _swsetup_Wakeup(ctx
);
1234 intel
->RenderIndex
= ~0;
1238 intel
->Fallback
&= ~bit
;
1239 if (oldfallback
== bit
) {
1240 assert(!intel
->tnl_pipeline_running
);
1243 if (INTEL_DEBUG
& DEBUG_PERF
)
1244 fprintf(stderr
, "LEAVE FALLBACK %s\n", getFallbackString(bit
));
1245 tnl
->Driver
.Render
.Start
= intelRenderStart
;
1246 tnl
->Driver
.Render
.PrimitiveNotify
= intelRenderPrimitive
;
1247 tnl
->Driver
.Render
.Finish
= intelRenderFinish
;
1248 tnl
->Driver
.Render
.BuildVertices
= _tnl_build_vertices
;
1249 tnl
->Driver
.Render
.CopyPV
= _tnl_copy_pv
;
1250 tnl
->Driver
.Render
.Interp
= _tnl_interp
;
1252 _tnl_invalidate_vertex_state(ctx
, ~0);
1253 _tnl_invalidate_vertices(ctx
, ~0);
1254 _tnl_install_attrs(ctx
,
1255 intel
->vertex_attrs
,
1256 intel
->vertex_attr_count
,
1257 intel
->ViewportMatrix
.m
, 0);
1259 intel
->NewGLState
|= _INTEL_NEW_RENDERSTATE
;
1270 /**********************************************************************/
1271 /* Initialization. */
1272 /**********************************************************************/
1276 intelInitTriFuncs(struct gl_context
* ctx
)
1278 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
1279 static int firsttime
= 1;
1286 tnl
->Driver
.RunPipeline
= intelRunPipeline
;
1287 tnl
->Driver
.Render
.Start
= intelRenderStart
;
1288 tnl
->Driver
.Render
.Finish
= intelRenderFinish
;
1289 tnl
->Driver
.Render
.PrimitiveNotify
= intelRenderPrimitive
;
1290 tnl
->Driver
.Render
.ResetLineStipple
= _swrast_ResetLineStipple
;
1291 tnl
->Driver
.Render
.BuildVertices
= _tnl_build_vertices
;
1292 tnl
->Driver
.Render
.CopyPV
= _tnl_copy_pv
;
1293 tnl
->Driver
.Render
.Interp
= _tnl_interp
;