Merge commit 'origin/gallium-0.1' into gallium-0.2
[mesa.git] / src / mesa / drivers / dri / i915 / intel_tris.c
1 /**************************************************************************
2 *
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
4 * All Rights Reserved.
5 *
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:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
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.
25 *
26 **************************************************************************/
27
28 /** @file intel_tris.c
29 *
30 * This file contains functions for managing the vertex buffer and emitting
31 * primitives into it.
32 */
33
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"
40 #include "main/dd.h"
41
42 #include "swrast/swrast.h"
43 #include "swrast_setup/swrast_setup.h"
44 #include "tnl/t_context.h"
45 #include "tnl/t_pipeline.h"
46 #include "tnl/t_vertex.h"
47
48 #include "intel_screen.h"
49 #include "intel_context.h"
50 #include "intel_tris.h"
51 #include "intel_batchbuffer.h"
52 #include "intel_buffers.h"
53 #include "intel_reg.h"
54 #include "intel_span.h"
55 #include "intel_tex.h"
56 #include "intel_chipset.h"
57 #include "i830_context.h"
58 #include "i830_reg.h"
59
60 static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
61 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
62 GLuint hwprim);
63
64 /** Sets the primitive type for a primitive sequence, flushing as needed. */
65 void intel_set_prim(struct intel_context *intel, uint32_t prim)
66 {
67 if (prim != intel->prim.primitive) {
68 INTEL_FIREVERTICES(intel);
69 intel->prim.primitive = prim;
70 }
71 }
72
73 /** Returns mapped VB space for the given number of vertices */
74 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
75 {
76 uint32_t *addr;
77
78 /* Check for space in the existing VB */
79 if (intel->prim.vb_bo == NULL ||
80 (intel->prim.current_offset +
81 count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
82 (intel->prim.count + count) >= (1 << 16)) {
83 /* Flush existing prim if any */
84 INTEL_FIREVERTICES(intel);
85
86 intel_finish_vb(intel);
87
88 /* Start a new VB */
89 if (intel->prim.vb == NULL)
90 intel->prim.vb = malloc(INTEL_VB_SIZE);
91 intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb",
92 INTEL_VB_SIZE, 4);
93 intel->prim.start_offset = 0;
94 intel->prim.current_offset = 0;
95 }
96
97 intel->prim.flush = intel_flush_prim;
98
99 addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
100 intel->prim.current_offset += intel->vertex_size * 4 * count;
101 intel->prim.count += count;
102
103 return addr;
104 }
105
106 /** Dispatches the accumulated primitive to the batchbuffer. */
107 void intel_flush_prim(struct intel_context *intel)
108 {
109 BATCH_LOCALS;
110 dri_bo *aper_array[2];
111 dri_bo *vb_bo;
112 unsigned int offset, count;
113
114 /* Must be called after an intel_start_prim. */
115 assert(intel->prim.primitive != ~0);
116
117 if (intel->prim.count == 0)
118 return;
119
120 /* Clear the current prims out of the context state so that a batch flush
121 * flush triggered by wait_flips or emit_state doesn't loop back to
122 * flush_prim again.
123 */
124 vb_bo = intel->prim.vb_bo;
125 dri_bo_reference(vb_bo);
126 count = intel->prim.count;
127 intel->prim.count = 0;
128 offset = intel->prim.start_offset;
129 intel->prim.start_offset = intel->prim.current_offset;
130 if (!IS_9XX(intel->intelScreen->deviceID))
131 intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
132 intel->prim.flush = NULL;
133
134 intel_wait_flips(intel);
135
136 intel->vtbl.emit_state(intel);
137
138 aper_array[0] = intel->batch->buf;
139 aper_array[1] = vb_bo;
140 if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
141 intel_batchbuffer_flush(intel->batch);
142 intel->vtbl.emit_state(intel);
143 }
144
145 /* Ensure that we don't start a new batch for the following emit, which
146 * depends on the state just emitted. emit_state should be making sure we
147 * have the space for this.
148 */
149 intel->no_batch_wrap = GL_TRUE;
150
151 /* Check that we actually emitted the state into this batch, using the
152 * UPLOAD_CTX bit as the signal.
153 */
154 assert((intel->batch->dirty_state & (1<<1)) == 0);
155
156 #if 0
157 printf("emitting %d..%d=%d vertices size %d\n", offset,
158 intel->prim.current_offset, intel->prim.count,
159 intel->vertex_size * 4);
160 #endif
161
162 if (IS_9XX(intel->intelScreen->deviceID)) {
163 BEGIN_BATCH(5, LOOP_CLIPRECTS);
164 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
165 I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
166 assert((offset & !S0_VB_OFFSET_MASK) == 0);
167 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
168 OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
169 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
170
171 OUT_BATCH(_3DPRIMITIVE |
172 PRIM_INDIRECT |
173 PRIM_INDIRECT_SEQUENTIAL |
174 intel->prim.primitive |
175 count);
176 OUT_BATCH(0); /* Beginning vertex index */
177 ADVANCE_BATCH();
178 } else {
179 struct i830_context *i830 = i830_context(&intel->ctx);
180
181 BEGIN_BATCH(5, LOOP_CLIPRECTS);
182 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
183 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
184 /* S0 */
185 assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
186 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
187 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
188 S0_VB_ENABLE_830);
189 /* S1
190 * This is somewhat unfortunate -- VB width is tied up with
191 * vertex format data that we've already uploaded through
192 * _3DSTATE_VFT[01]_CMD. We may want to replace emits of VFT state with
193 * STATE_IMMEDIATE_1 like this to avoid duplication.
194 */
195 OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
196 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
197 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
198 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
199
200 OUT_BATCH(_3DPRIMITIVE |
201 PRIM_INDIRECT |
202 PRIM_INDIRECT_SEQUENTIAL |
203 intel->prim.primitive |
204 count);
205 OUT_BATCH(0); /* Beginning vertex index */
206 ADVANCE_BATCH();
207 }
208
209 intel->no_batch_wrap = GL_FALSE;
210
211 dri_bo_unreference(vb_bo);
212 }
213
214 /**
215 * Uploads the locally-accumulated VB into the buffer object.
216 *
217 * This avoids us thrashing the cachelines in and out as the buffer gets
218 * filled, dispatched, then reused as the hardware completes rendering from it,
219 * and also lets us clflush less if we dispatch with a partially-filled VB.
220 *
221 * This is called normally from get_space when we're finishing a BO, but also
222 * at batch flush time so that we don't try accessing the contents of a
223 * just-dispatched buffer.
224 */
225 void intel_finish_vb(struct intel_context *intel)
226 {
227 if (intel->prim.vb_bo == NULL)
228 return;
229
230 dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
231 intel->prim.vb);
232 dri_bo_unreference(intel->prim.vb_bo);
233 intel->prim.vb_bo = NULL;
234 }
235
236 /***********************************************************************
237 * Emit primitives as inline vertices *
238 ***********************************************************************/
239
240 #ifdef __i386__
241 #define COPY_DWORDS( j, vb, vertsize, v ) \
242 do { \
243 int __tmp; \
244 __asm__ __volatile__( "rep ; movsl" \
245 : "=%c" (j), "=D" (vb), "=S" (__tmp) \
246 : "0" (vertsize), \
247 "D" ((long)vb), \
248 "S" ((long)v) ); \
249 } while (0)
250 #else
251 #define COPY_DWORDS( j, vb, vertsize, v ) \
252 do { \
253 for ( j = 0 ; j < vertsize ; j++ ) { \
254 vb[j] = ((GLuint *)v)[j]; \
255 } \
256 vb += vertsize; \
257 } while (0)
258 #endif
259
260 static void
261 intel_draw_quad(struct intel_context *intel,
262 intelVertexPtr v0,
263 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
264 {
265 GLuint vertsize = intel->vertex_size;
266 GLuint *vb = intel_get_prim_space(intel, 6);
267 int j;
268
269 COPY_DWORDS(j, vb, vertsize, v0);
270 COPY_DWORDS(j, vb, vertsize, v1);
271
272 /* If smooth shading, draw like a trifan which gives better
273 * rasterization. Otherwise draw as two triangles with provoking
274 * vertex in third position as required for flat shading.
275 */
276 if (intel->ctx.Light.ShadeModel == GL_FLAT) {
277 COPY_DWORDS(j, vb, vertsize, v3);
278 COPY_DWORDS(j, vb, vertsize, v1);
279 }
280 else {
281 COPY_DWORDS(j, vb, vertsize, v2);
282 COPY_DWORDS(j, vb, vertsize, v0);
283 }
284
285 COPY_DWORDS(j, vb, vertsize, v2);
286 COPY_DWORDS(j, vb, vertsize, v3);
287 }
288
289 static void
290 intel_draw_triangle(struct intel_context *intel,
291 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
292 {
293 GLuint vertsize = intel->vertex_size;
294 GLuint *vb = intel_get_prim_space(intel, 3);
295 int j;
296
297 COPY_DWORDS(j, vb, vertsize, v0);
298 COPY_DWORDS(j, vb, vertsize, v1);
299 COPY_DWORDS(j, vb, vertsize, v2);
300 }
301
302
303 static void
304 intel_draw_line(struct intel_context *intel,
305 intelVertexPtr v0, intelVertexPtr v1)
306 {
307 GLuint vertsize = intel->vertex_size;
308 GLuint *vb = intel_get_prim_space(intel, 2);
309 int j;
310
311 COPY_DWORDS(j, vb, vertsize, v0);
312 COPY_DWORDS(j, vb, vertsize, v1);
313 }
314
315
316 static void
317 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
318 {
319 GLuint vertsize = intel->vertex_size;
320 GLuint *vb = intel_get_prim_space(intel, 1);
321 int j;
322
323 /* Adjust for sub pixel position -- still required for conform. */
324 *(float *) &vb[0] = v0->v.x;
325 *(float *) &vb[1] = v0->v.y;
326 for (j = 2; j < vertsize; j++)
327 vb[j] = v0->ui[j];
328 }
329
330
331
332 /***********************************************************************
333 * Fixup for ARB_point_parameters *
334 ***********************************************************************/
335
336 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
337 * represented in the fragment program InputsRead field.
338 */
339 static void
340 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
341 {
342 GLcontext *ctx = &intel->ctx;
343 GLfloat psz[4], col[4], restore_psz, restore_alpha;
344
345 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
346 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
347
348 restore_psz = psz[0];
349 restore_alpha = col[3];
350
351 if (psz[0] >= ctx->Point.Threshold) {
352 psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
353 }
354 else {
355 GLfloat dsize = psz[0] / ctx->Point.Threshold;
356 psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
357 col[3] *= dsize * dsize;
358 }
359
360 if (psz[0] < 1.0)
361 psz[0] = 1.0;
362
363 if (restore_psz != psz[0] || restore_alpha != col[3]) {
364 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
365 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
366
367 intel_draw_point(intel, v0);
368
369 psz[0] = restore_psz;
370 col[3] = restore_alpha;
371
372 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
373 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
374 }
375 else
376 intel_draw_point(intel, v0);
377 }
378
379
380
381
382
383 /***********************************************************************
384 * Fixup for I915 WPOS texture coordinate *
385 ***********************************************************************/
386
387
388
389 static void
390 intel_wpos_triangle(struct intel_context *intel,
391 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
392 {
393 GLuint offset = intel->wpos_offset;
394 GLuint size = intel->wpos_size;
395 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
396 GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
397 GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
398
399 __memcpy(v0_wpos, v0, size);
400 __memcpy(v1_wpos, v1, size);
401 __memcpy(v2_wpos, v2, size);
402
403 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
404 v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
405 v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
406
407
408 intel_draw_triangle(intel, v0, v1, v2);
409 }
410
411
412 static void
413 intel_wpos_line(struct intel_context *intel,
414 intelVertexPtr v0, intelVertexPtr v1)
415 {
416 GLuint offset = intel->wpos_offset;
417 GLuint size = intel->wpos_size;
418 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
419 GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
420
421 __memcpy(v0_wpos, v0, size);
422 __memcpy(v1_wpos, v1, size);
423
424 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
425 v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
426
427 intel_draw_line(intel, v0, v1);
428 }
429
430
431 static void
432 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
433 {
434 GLuint offset = intel->wpos_offset;
435 GLuint size = intel->wpos_size;
436 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
437
438 __memcpy(v0_wpos, v0, size);
439 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
440
441 intel_draw_point(intel, v0);
442 }
443
444
445
446
447
448
449 /***********************************************************************
450 * Macros for t_dd_tritmp.h to draw basic primitives *
451 ***********************************************************************/
452
453 #define TRI( a, b, c ) \
454 do { \
455 if (DO_FALLBACK) \
456 intel->draw_tri( intel, a, b, c ); \
457 else \
458 intel_draw_triangle( intel, a, b, c ); \
459 } while (0)
460
461 #define QUAD( a, b, c, d ) \
462 do { \
463 if (DO_FALLBACK) { \
464 intel->draw_tri( intel, a, b, d ); \
465 intel->draw_tri( intel, b, c, d ); \
466 } else \
467 intel_draw_quad( intel, a, b, c, d ); \
468 } while (0)
469
470 #define LINE( v0, v1 ) \
471 do { \
472 if (DO_FALLBACK) \
473 intel->draw_line( intel, v0, v1 ); \
474 else \
475 intel_draw_line( intel, v0, v1 ); \
476 } while (0)
477
478 #define POINT( v0 ) \
479 do { \
480 if (DO_FALLBACK) \
481 intel->draw_point( intel, v0 ); \
482 else \
483 intel_draw_point( intel, v0 ); \
484 } while (0)
485
486
487 /***********************************************************************
488 * Build render functions from dd templates *
489 ***********************************************************************/
490
491 #define INTEL_OFFSET_BIT 0x01
492 #define INTEL_TWOSIDE_BIT 0x02
493 #define INTEL_UNFILLED_BIT 0x04
494 #define INTEL_FALLBACK_BIT 0x08
495 #define INTEL_MAX_TRIFUNC 0x10
496
497
498 static struct
499 {
500 tnl_points_func points;
501 tnl_line_func line;
502 tnl_triangle_func triangle;
503 tnl_quad_func quad;
504 } rast_tab[INTEL_MAX_TRIFUNC];
505
506
507 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
508 #define DO_OFFSET (IND & INTEL_OFFSET_BIT)
509 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
510 #define DO_TWOSIDE (IND & INTEL_TWOSIDE_BIT)
511 #define DO_FLAT 0
512 #define DO_TRI 1
513 #define DO_QUAD 1
514 #define DO_LINE 1
515 #define DO_POINTS 1
516 #define DO_FULL_QUAD 1
517
518 #define HAVE_RGBA 1
519 #define HAVE_SPEC 1
520 #define HAVE_BACK_COLORS 0
521 #define HAVE_HW_FLATSHADE 1
522 #define VERTEX intelVertex
523 #define TAB rast_tab
524
525 /* Only used to pull back colors into vertices (ie, we know color is
526 * floating point).
527 */
528 #define INTEL_COLOR( dst, src ) \
529 do { \
530 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
531 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
532 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
533 UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]); \
534 } while (0)
535
536 #define INTEL_SPEC( dst, src ) \
537 do { \
538 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
539 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
540 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
541 } while (0)
542
543
544 #define DEPTH_SCALE intel->polygon_offset_scale
545 #define UNFILLED_TRI unfilled_tri
546 #define UNFILLED_QUAD unfilled_quad
547 #define VERT_X(_v) _v->v.x
548 #define VERT_Y(_v) _v->v.y
549 #define VERT_Z(_v) _v->v.z
550 #define AREA_IS_CCW( a ) (a > 0)
551 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
552
553 #define VERT_SET_RGBA( v, c ) if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
554 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
555 #define VERT_SAVE_RGBA( idx ) if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
556 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
557
558 #define VERT_SET_SPEC( v, c ) if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
559 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
560 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
561 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
562
563 #define LOCAL_VARS(n) \
564 struct intel_context *intel = intel_context(ctx); \
565 GLuint color[n] = { 0, }, spec[n] = { 0, }; \
566 GLuint coloroffset = intel->coloroffset; \
567 GLboolean specoffset = intel->specoffset; \
568 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
569
570
571 /***********************************************************************
572 * Helpers for rendering unfilled primitives *
573 ***********************************************************************/
574
575 static const GLuint hw_prim[GL_POLYGON + 1] = {
576 PRIM3D_POINTLIST,
577 PRIM3D_LINELIST,
578 PRIM3D_LINELIST,
579 PRIM3D_LINELIST,
580 PRIM3D_TRILIST,
581 PRIM3D_TRILIST,
582 PRIM3D_TRILIST,
583 PRIM3D_TRILIST,
584 PRIM3D_TRILIST,
585 PRIM3D_TRILIST
586 };
587
588 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
589 #define RENDER_PRIMITIVE intel->render_primitive
590 #define TAG(x) x
591 #define IND INTEL_FALLBACK_BIT
592 #include "tnl_dd/t_dd_unfilled.h"
593 #undef IND
594
595 /***********************************************************************
596 * Generate GL render functions *
597 ***********************************************************************/
598
599 #define IND (0)
600 #define TAG(x) x
601 #include "tnl_dd/t_dd_tritmp.h"
602
603 #define IND (INTEL_OFFSET_BIT)
604 #define TAG(x) x##_offset
605 #include "tnl_dd/t_dd_tritmp.h"
606
607 #define IND (INTEL_TWOSIDE_BIT)
608 #define TAG(x) x##_twoside
609 #include "tnl_dd/t_dd_tritmp.h"
610
611 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
612 #define TAG(x) x##_twoside_offset
613 #include "tnl_dd/t_dd_tritmp.h"
614
615 #define IND (INTEL_UNFILLED_BIT)
616 #define TAG(x) x##_unfilled
617 #include "tnl_dd/t_dd_tritmp.h"
618
619 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
620 #define TAG(x) x##_offset_unfilled
621 #include "tnl_dd/t_dd_tritmp.h"
622
623 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
624 #define TAG(x) x##_twoside_unfilled
625 #include "tnl_dd/t_dd_tritmp.h"
626
627 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
628 #define TAG(x) x##_twoside_offset_unfilled
629 #include "tnl_dd/t_dd_tritmp.h"
630
631 #define IND (INTEL_FALLBACK_BIT)
632 #define TAG(x) x##_fallback
633 #include "tnl_dd/t_dd_tritmp.h"
634
635 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
636 #define TAG(x) x##_offset_fallback
637 #include "tnl_dd/t_dd_tritmp.h"
638
639 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
640 #define TAG(x) x##_twoside_fallback
641 #include "tnl_dd/t_dd_tritmp.h"
642
643 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
644 #define TAG(x) x##_twoside_offset_fallback
645 #include "tnl_dd/t_dd_tritmp.h"
646
647 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
648 #define TAG(x) x##_unfilled_fallback
649 #include "tnl_dd/t_dd_tritmp.h"
650
651 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
652 #define TAG(x) x##_offset_unfilled_fallback
653 #include "tnl_dd/t_dd_tritmp.h"
654
655 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
656 #define TAG(x) x##_twoside_unfilled_fallback
657 #include "tnl_dd/t_dd_tritmp.h"
658
659 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
660 INTEL_FALLBACK_BIT)
661 #define TAG(x) x##_twoside_offset_unfilled_fallback
662 #include "tnl_dd/t_dd_tritmp.h"
663
664
665 static void
666 init_rast_tab(void)
667 {
668 init();
669 init_offset();
670 init_twoside();
671 init_twoside_offset();
672 init_unfilled();
673 init_offset_unfilled();
674 init_twoside_unfilled();
675 init_twoside_offset_unfilled();
676 init_fallback();
677 init_offset_fallback();
678 init_twoside_fallback();
679 init_twoside_offset_fallback();
680 init_unfilled_fallback();
681 init_offset_unfilled_fallback();
682 init_twoside_unfilled_fallback();
683 init_twoside_offset_unfilled_fallback();
684 }
685
686
687 /***********************************************************************
688 * Rasterization fallback helpers *
689 ***********************************************************************/
690
691
692 /* This code is hit only when a mix of accelerated and unaccelerated
693 * primitives are being drawn, and only for the unaccelerated
694 * primitives.
695 */
696 static void
697 intel_fallback_tri(struct intel_context *intel,
698 intelVertex * v0, intelVertex * v1, intelVertex * v2)
699 {
700 GLcontext *ctx = &intel->ctx;
701 SWvertex v[3];
702
703 if (0)
704 fprintf(stderr, "\n%s\n", __FUNCTION__);
705
706 INTEL_FIREVERTICES(intel);
707
708 _swsetup_Translate(ctx, v0, &v[0]);
709 _swsetup_Translate(ctx, v1, &v[1]);
710 _swsetup_Translate(ctx, v2, &v[2]);
711 intelSpanRenderStart(ctx);
712 _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
713 intelSpanRenderFinish(ctx);
714 }
715
716
717 static void
718 intel_fallback_line(struct intel_context *intel,
719 intelVertex * v0, intelVertex * v1)
720 {
721 GLcontext *ctx = &intel->ctx;
722 SWvertex v[2];
723
724 if (0)
725 fprintf(stderr, "\n%s\n", __FUNCTION__);
726
727 INTEL_FIREVERTICES(intel);
728
729 _swsetup_Translate(ctx, v0, &v[0]);
730 _swsetup_Translate(ctx, v1, &v[1]);
731 intelSpanRenderStart(ctx);
732 _swrast_Line(ctx, &v[0], &v[1]);
733 intelSpanRenderFinish(ctx);
734 }
735
736 static void
737 intel_fallback_point(struct intel_context *intel,
738 intelVertex * v0)
739 {
740 GLcontext *ctx = &intel->ctx;
741 SWvertex v[1];
742
743 if (0)
744 fprintf(stderr, "\n%s\n", __FUNCTION__);
745
746 INTEL_FIREVERTICES(intel);
747
748 _swsetup_Translate(ctx, v0, &v[0]);
749 intelSpanRenderStart(ctx);
750 _swrast_Point(ctx, &v[0]);
751 intelSpanRenderFinish(ctx);
752 }
753
754
755 /**********************************************************************/
756 /* Render unclipped begin/end objects */
757 /**********************************************************************/
758
759 #define IND 0
760 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
761 #define RENDER_POINTS( start, count ) \
762 for ( ; start < count ; start++) POINT( V(ELT(start)) );
763 #define RENDER_LINE( v0, v1 ) LINE( V(v0), V(v1) )
764 #define RENDER_TRI( v0, v1, v2 ) TRI( V(v0), V(v1), V(v2) )
765 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
766 #define INIT(x) intelRenderPrimitive( ctx, x )
767 #undef LOCAL_VARS
768 #define LOCAL_VARS \
769 struct intel_context *intel = intel_context(ctx); \
770 GLubyte *vertptr = (GLubyte *)intel->verts; \
771 const GLuint vertsize = intel->vertex_size; \
772 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
773 (void) elt;
774 #define RESET_STIPPLE
775 #define RESET_OCCLUSION
776 #define PRESERVE_VB_DEFS
777 #define ELT(x) x
778 #define TAG(x) intel_##x##_verts
779 #include "tnl/t_vb_rendertmp.h"
780 #undef ELT
781 #undef TAG
782 #define TAG(x) intel_##x##_elts
783 #define ELT(x) elt[x]
784 #include "tnl/t_vb_rendertmp.h"
785
786 /**********************************************************************/
787 /* Render clipped primitives */
788 /**********************************************************************/
789
790
791
792 static void
793 intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
794 {
795 struct intel_context *intel = intel_context(ctx);
796 TNLcontext *tnl = TNL_CONTEXT(ctx);
797 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
798 GLuint prim = intel->render_primitive;
799
800 /* Render the new vertices as an unclipped polygon.
801 */
802 {
803 GLuint *tmp = VB->Elts;
804 VB->Elts = (GLuint *) elts;
805 tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
806 PRIM_BEGIN | PRIM_END);
807 VB->Elts = tmp;
808 }
809
810 /* Restore the render primitive
811 */
812 if (prim != GL_POLYGON)
813 tnl->Driver.Render.PrimitiveNotify(ctx, prim);
814 }
815
816 static void
817 intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj)
818 {
819 TNLcontext *tnl = TNL_CONTEXT(ctx);
820
821 tnl->Driver.Render.Line(ctx, ii, jj);
822 }
823
824 static void
825 intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
826 {
827 struct intel_context *intel = intel_context(ctx);
828 const GLuint vertsize = intel->vertex_size;
829 GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
830 GLubyte *vertptr = (GLubyte *) intel->verts;
831 const GLuint *start = (const GLuint *) V(elts[0]);
832 int i, j;
833
834 for (i = 2; i < n; i++) {
835 COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
836 COPY_DWORDS(j, vb, vertsize, V(elts[i]));
837 COPY_DWORDS(j, vb, vertsize, start);
838 }
839 }
840
841 /**********************************************************************/
842 /* Choose render functions */
843 /**********************************************************************/
844
845
846
847
848 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
849 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
850
851 void
852 intelChooseRenderState(GLcontext * ctx)
853 {
854 TNLcontext *tnl = TNL_CONTEXT(ctx);
855 struct intel_context *intel = intel_context(ctx);
856 GLuint flags = ctx->_TriangleCaps;
857 const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
858 GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
859 GLuint index = 0;
860
861 if (INTEL_DEBUG & DEBUG_STATE)
862 fprintf(stderr, "\n%s\n", __FUNCTION__);
863
864 if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
865
866 if (flags & ANY_RASTER_FLAGS) {
867 if (flags & DD_TRI_LIGHT_TWOSIDE)
868 index |= INTEL_TWOSIDE_BIT;
869 if (flags & DD_TRI_OFFSET)
870 index |= INTEL_OFFSET_BIT;
871 if (flags & DD_TRI_UNFILLED)
872 index |= INTEL_UNFILLED_BIT;
873 }
874
875 if (have_wpos) {
876 intel->draw_point = intel_wpos_point;
877 intel->draw_line = intel_wpos_line;
878 intel->draw_tri = intel_wpos_triangle;
879
880 /* Make sure these get called:
881 */
882 index |= INTEL_FALLBACK_BIT;
883 }
884 else {
885 intel->draw_point = intel_draw_point;
886 intel->draw_line = intel_draw_line;
887 intel->draw_tri = intel_draw_triangle;
888 }
889
890 /* Hook in fallbacks for specific primitives.
891 */
892 if (flags & ANY_FALLBACK_FLAGS) {
893 if (flags & DD_LINE_STIPPLE)
894 intel->draw_line = intel_fallback_line;
895
896 if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
897 intel->draw_tri = intel_fallback_tri;
898
899 if (flags & DD_TRI_SMOOTH) {
900 if (intel->strict_conformance)
901 intel->draw_tri = intel_fallback_tri;
902 }
903
904 if (flags & DD_POINT_ATTEN) {
905 if (0)
906 intel->draw_point = intel_atten_point;
907 else
908 intel->draw_point = intel_fallback_point;
909 }
910
911 if (flags & DD_POINT_SMOOTH) {
912 if (intel->strict_conformance)
913 intel->draw_point = intel_fallback_point;
914 }
915
916 index |= INTEL_FALLBACK_BIT;
917 }
918 }
919
920 if (intel->RenderIndex != index) {
921 intel->RenderIndex = index;
922
923 tnl->Driver.Render.Points = rast_tab[index].points;
924 tnl->Driver.Render.Line = rast_tab[index].line;
925 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
926 tnl->Driver.Render.Quad = rast_tab[index].quad;
927
928 if (index == 0) {
929 tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
930 tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
931 tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
932 tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
933 }
934 else {
935 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
936 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
937 tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
938 tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
939 }
940 }
941 }
942
943 static const GLenum reduced_prim[GL_POLYGON + 1] = {
944 GL_POINTS,
945 GL_LINES,
946 GL_LINES,
947 GL_LINES,
948 GL_TRIANGLES,
949 GL_TRIANGLES,
950 GL_TRIANGLES,
951 GL_TRIANGLES,
952 GL_TRIANGLES,
953 GL_TRIANGLES
954 };
955
956
957 /**********************************************************************/
958 /* High level hooks for t_vb_render.c */
959 /**********************************************************************/
960
961
962
963
964 static void
965 intelRunPipeline(GLcontext * ctx)
966 {
967 struct intel_context *intel = intel_context(ctx);
968
969 _mesa_lock_context_textures(ctx);
970
971 if (ctx->NewState)
972 _mesa_update_state_locked(ctx);
973
974 if (intel->NewGLState) {
975 if (intel->NewGLState & _NEW_TEXTURE) {
976 intel->vtbl.update_texture_state(intel);
977 }
978
979 if (!intel->Fallback) {
980 if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
981 intelChooseRenderState(ctx);
982 }
983
984 intel->NewGLState = 0;
985 }
986
987 _tnl_run_pipeline(ctx);
988
989 _mesa_unlock_context_textures(ctx);
990 }
991
992 static void
993 intelRenderStart(GLcontext * ctx)
994 {
995 struct intel_context *intel = intel_context(ctx);
996
997 intel->vtbl.render_start(intel_context(ctx));
998 intel->vtbl.emit_state(intel);
999 }
1000
1001 static void
1002 intelRenderFinish(GLcontext * ctx)
1003 {
1004 struct intel_context *intel = intel_context(ctx);
1005
1006 if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1007 _swrast_flush(ctx);
1008
1009 INTEL_FIREVERTICES(intel);
1010 }
1011
1012
1013
1014
1015 /* System to flush dma and emit state changes based on the rasterized
1016 * primitive.
1017 */
1018 static void
1019 intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim)
1020 {
1021 struct intel_context *intel = intel_context(ctx);
1022
1023 if (0)
1024 fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1025 _mesa_lookup_enum_by_nr(rprim), hwprim);
1026
1027 intel->vtbl.reduced_primitive_state(intel, rprim);
1028
1029 /* Start a new primitive. Arrange to have it flushed later on.
1030 */
1031 if (hwprim != intel->prim.primitive) {
1032 INTEL_FIREVERTICES(intel);
1033
1034 intel_set_prim(intel, hwprim);
1035 }
1036 }
1037
1038
1039 /*
1040 */
1041 static void
1042 intelRenderPrimitive(GLcontext * ctx, GLenum prim)
1043 {
1044 struct intel_context *intel = intel_context(ctx);
1045
1046 if (0)
1047 fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1048
1049 /* Let some clipping routines know which primitive they're dealing
1050 * with.
1051 */
1052 intel->render_primitive = prim;
1053
1054 /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1055 * triangles. The rasterized primitive will always be reset by
1056 * lower level functions in that case, potentially pingponging the
1057 * state:
1058 */
1059 if (reduced_prim[prim] == GL_TRIANGLES &&
1060 (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1061 return;
1062
1063 /* Set some primitive-dependent state and Start? a new primitive.
1064 */
1065 intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1066 }
1067
1068
1069 /**********************************************************************/
1070 /* Transition to/from hardware rasterization. */
1071 /**********************************************************************/
1072
1073 static char *fallbackStrings[] = {
1074 [0] = "Draw buffer",
1075 [1] = "Read buffer",
1076 [2] = "Depth buffer",
1077 [3] = "Stencil buffer",
1078 [4] = "User disable",
1079 [5] = "Render mode",
1080
1081 [12] = "Texture",
1082 [13] = "Color mask",
1083 [14] = "Stencil",
1084 [15] = "Stipple",
1085 [16] = "Program",
1086 [17] = "Logic op",
1087 [18] = "Smooth polygon",
1088 [19] = "Smooth point",
1089 };
1090
1091
1092 static char *
1093 getFallbackString(GLuint bit)
1094 {
1095 int i = 0;
1096 while (bit > 1) {
1097 i++;
1098 bit >>= 1;
1099 }
1100 return fallbackStrings[i];
1101 }
1102
1103
1104
1105 void
1106 intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode)
1107 {
1108 GLcontext *ctx = &intel->ctx;
1109 TNLcontext *tnl = TNL_CONTEXT(ctx);
1110 GLuint oldfallback = intel->Fallback;
1111
1112 if (mode) {
1113 intel->Fallback |= bit;
1114 if (oldfallback == 0) {
1115 intelFlush(ctx);
1116 if (INTEL_DEBUG & DEBUG_FALLBACKS)
1117 fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1118 bit, getFallbackString(bit));
1119 _swsetup_Wakeup(ctx);
1120 intel->RenderIndex = ~0;
1121 }
1122 }
1123 else {
1124 intel->Fallback &= ~bit;
1125 if (oldfallback == bit) {
1126 _swrast_flush(ctx);
1127 if (INTEL_DEBUG & DEBUG_FALLBACKS)
1128 fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1129 tnl->Driver.Render.Start = intelRenderStart;
1130 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1131 tnl->Driver.Render.Finish = intelRenderFinish;
1132 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1133 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1134 tnl->Driver.Render.Interp = _tnl_interp;
1135
1136 _tnl_invalidate_vertex_state(ctx, ~0);
1137 _tnl_invalidate_vertices(ctx, ~0);
1138 _tnl_install_attrs(ctx,
1139 intel->vertex_attrs,
1140 intel->vertex_attr_count,
1141 intel->ViewportMatrix.m, 0);
1142
1143 intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1144 }
1145 }
1146 }
1147
1148 union fi
1149 {
1150 GLfloat f;
1151 GLint i;
1152 };
1153
1154
1155 /**********************************************************************/
1156 /* Used only with the metaops callbacks. */
1157 /**********************************************************************/
1158 static void
1159 intel_meta_draw_poly(struct intel_context *intel,
1160 GLuint n,
1161 GLfloat xy[][2],
1162 GLfloat z, GLuint color, GLfloat tex[][2])
1163 {
1164 union fi *vb;
1165 GLint i;
1166 GLboolean was_locked = intel->locked;
1167 unsigned int saved_vertex_size = intel->vertex_size;
1168
1169 if (!was_locked)
1170 LOCK_HARDWARE(intel);
1171
1172 intel->vertex_size = 6;
1173
1174 /* All 3d primitives should be emitted with LOOP_CLIPRECTS,
1175 * otherwise the drawing origin (DR4) might not be set correctly.
1176 */
1177 intel_set_prim(intel, PRIM3D_TRIFAN);
1178 vb = (union fi *) intel_get_prim_space(intel, n);
1179
1180 for (i = 0; i < n; i++) {
1181 vb[0].f = xy[i][0];
1182 vb[1].f = xy[i][1];
1183 vb[2].f = z;
1184 vb[3].i = color;
1185 vb[4].f = tex[i][0];
1186 vb[5].f = tex[i][1];
1187 vb += 6;
1188 }
1189
1190 INTEL_FIREVERTICES(intel);
1191
1192 intel->vertex_size = saved_vertex_size;
1193
1194 if (!was_locked)
1195 UNLOCK_HARDWARE(intel);
1196 }
1197
1198 static void
1199 intel_meta_draw_quad(struct intel_context *intel,
1200 GLfloat x0, GLfloat x1,
1201 GLfloat y0, GLfloat y1,
1202 GLfloat z,
1203 GLuint color,
1204 GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1)
1205 {
1206 GLfloat xy[4][2];
1207 GLfloat tex[4][2];
1208
1209 xy[0][0] = x0;
1210 xy[0][1] = y0;
1211 xy[1][0] = x1;
1212 xy[1][1] = y0;
1213 xy[2][0] = x1;
1214 xy[2][1] = y1;
1215 xy[3][0] = x0;
1216 xy[3][1] = y1;
1217
1218 tex[0][0] = s0;
1219 tex[0][1] = t0;
1220 tex[1][0] = s1;
1221 tex[1][1] = t0;
1222 tex[2][0] = s1;
1223 tex[2][1] = t1;
1224 tex[3][0] = s0;
1225 tex[3][1] = t1;
1226
1227 intel_meta_draw_poly(intel, 4, xy, z, color, tex);
1228 }
1229
1230
1231
1232 /**********************************************************************/
1233 /* Initialization. */
1234 /**********************************************************************/
1235
1236
1237 void
1238 intelInitTriFuncs(GLcontext * ctx)
1239 {
1240 struct intel_context *intel = intel_context(ctx);
1241 TNLcontext *tnl = TNL_CONTEXT(ctx);
1242 static int firsttime = 1;
1243
1244 if (firsttime) {
1245 init_rast_tab();
1246 firsttime = 0;
1247 }
1248
1249 tnl->Driver.RunPipeline = intelRunPipeline;
1250 tnl->Driver.Render.Start = intelRenderStart;
1251 tnl->Driver.Render.Finish = intelRenderFinish;
1252 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1253 tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1254 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1255 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1256 tnl->Driver.Render.Interp = _tnl_interp;
1257
1258 intel->vtbl.meta_draw_quad = intel_meta_draw_quad;
1259 }