Rename the various function types in t_context.h to include a tnl_ prefix.
[mesa.git] / src / mesa / drivers / dri / radeon / radeon_swtcl.c
1 /* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_swtcl.c,v 1.6 2003/05/06 23:52:08 daenzer Exp $ */
2 /**************************************************************************
3
4 Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
5 VA Linux Systems Inc., Fremont, California.
6
7 All Rights Reserved.
8
9 Permission is hereby granted, free of charge, to any person obtaining
10 a copy of this software and associated documentation files (the
11 "Software"), to deal in the Software without restriction, including
12 without limitation the rights to use, copy, modify, merge, publish,
13 distribute, sublicense, and/or sell copies of the Software, and to
14 permit persons to whom the Software is furnished to do so, subject to
15 the following conditions:
16
17 The above copyright notice and this permission notice (including the
18 next paragraph) shall be included in all copies or substantial
19 portions of the Software.
20
21 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
22 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
23 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
24 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
25 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
26 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
27 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28
29 **************************************************************************/
30
31 /*
32 * Authors:
33 * Keith Whitwell <keith@tungstengraphics.com>
34 */
35
36 #include "glheader.h"
37 #include "mtypes.h"
38 #include "colormac.h"
39 #include "enums.h"
40 #include "imports.h"
41 #include "macros.h"
42
43 #include "swrast_setup/swrast_setup.h"
44 #include "math/m_translate.h"
45 #include "tnl/tnl.h"
46 #include "tnl/t_context.h"
47 #include "tnl/t_pipeline.h"
48 #include "tnl/t_vtx_api.h" /* for _tnl_FlushVertices */
49
50 #include "radeon_context.h"
51 #include "radeon_ioctl.h"
52 #include "radeon_state.h"
53 #include "radeon_swtcl.h"
54 #include "radeon_tcl.h"
55
56 /***********************************************************************
57 * Build render functions from dd templates *
58 ***********************************************************************/
59
60
61 #define RADEON_XYZW_BIT 0x01
62 #define RADEON_RGBA_BIT 0x02
63 #define RADEON_SPEC_BIT 0x04
64 #define RADEON_TEX0_BIT 0x08
65 #define RADEON_TEX1_BIT 0x10
66 #define RADEON_PTEX_BIT 0x20
67 #define RADEON_MAX_SETUP 0x40
68
69 static void flush_last_swtcl_prim( radeonContextPtr rmesa );
70
71 static struct {
72 void (*emit)( GLcontext *, GLuint, GLuint, void *, GLuint );
73 tnl_interp_func interp;
74 tnl_copy_pv_func copy_pv;
75 GLboolean (*check_tex_sizes)( GLcontext *ctx );
76 GLuint vertex_size;
77 GLuint vertex_format;
78 } setup_tab[RADEON_MAX_SETUP];
79
80
81 #define TINY_VERTEX_FORMAT (RADEON_CP_VC_FRMT_XY | \
82 RADEON_CP_VC_FRMT_Z | \
83 RADEON_CP_VC_FRMT_PKCOLOR)
84
85 #define NOTEX_VERTEX_FORMAT (RADEON_CP_VC_FRMT_XY | \
86 RADEON_CP_VC_FRMT_Z | \
87 RADEON_CP_VC_FRMT_W0 | \
88 RADEON_CP_VC_FRMT_PKCOLOR | \
89 RADEON_CP_VC_FRMT_PKSPEC)
90
91 #define TEX0_VERTEX_FORMAT (RADEON_CP_VC_FRMT_XY | \
92 RADEON_CP_VC_FRMT_Z | \
93 RADEON_CP_VC_FRMT_W0 | \
94 RADEON_CP_VC_FRMT_PKCOLOR | \
95 RADEON_CP_VC_FRMT_PKSPEC | \
96 RADEON_CP_VC_FRMT_ST0)
97
98 #define TEX1_VERTEX_FORMAT (RADEON_CP_VC_FRMT_XY | \
99 RADEON_CP_VC_FRMT_Z | \
100 RADEON_CP_VC_FRMT_W0 | \
101 RADEON_CP_VC_FRMT_PKCOLOR | \
102 RADEON_CP_VC_FRMT_PKSPEC | \
103 RADEON_CP_VC_FRMT_ST0 | \
104 RADEON_CP_VC_FRMT_ST1)
105
106 #define PROJ_TEX1_VERTEX_FORMAT (RADEON_CP_VC_FRMT_XY | \
107 RADEON_CP_VC_FRMT_Z | \
108 RADEON_CP_VC_FRMT_W0 | \
109 RADEON_CP_VC_FRMT_PKCOLOR | \
110 RADEON_CP_VC_FRMT_PKSPEC | \
111 RADEON_CP_VC_FRMT_ST0 | \
112 RADEON_CP_VC_FRMT_Q0 | \
113 RADEON_CP_VC_FRMT_ST1 | \
114 RADEON_CP_VC_FRMT_Q1)
115
116 #define TEX2_VERTEX_FORMAT 0
117 #define TEX3_VERTEX_FORMAT 0
118 #define PROJ_TEX3_VERTEX_FORMAT 0
119
120 #define DO_XYZW (IND & RADEON_XYZW_BIT)
121 #define DO_RGBA (IND & RADEON_RGBA_BIT)
122 #define DO_SPEC (IND & RADEON_SPEC_BIT)
123 #define DO_FOG (IND & RADEON_SPEC_BIT)
124 #define DO_TEX0 (IND & RADEON_TEX0_BIT)
125 #define DO_TEX1 (IND & RADEON_TEX1_BIT)
126 #define DO_TEX2 0
127 #define DO_TEX3 0
128 #define DO_PTEX (IND & RADEON_PTEX_BIT)
129
130 #define VERTEX radeonVertex
131 #define VERTEX_COLOR radeon_color_t
132 #define GET_VIEWPORT_MAT() 0
133 #define GET_TEXSOURCE(n) n
134 #define GET_VERTEX_FORMAT() RADEON_CONTEXT(ctx)->swtcl.vertex_format
135 #define GET_VERTEX_STORE() RADEON_CONTEXT(ctx)->swtcl.verts
136 #define GET_VERTEX_SIZE() RADEON_CONTEXT(ctx)->swtcl.vertex_size * sizeof(GLuint)
137
138 #define HAVE_HW_VIEWPORT 1
139 /* Tiny vertices don't seem to work atm - haven't looked into why.
140 */
141 #define HAVE_HW_DIVIDE (IND & ~(RADEON_XYZW_BIT|RADEON_RGBA_BIT))
142 #define HAVE_TINY_VERTICES 1
143 #define HAVE_RGBA_COLOR 1
144 #define HAVE_NOTEX_VERTICES 1
145 #define HAVE_TEX0_VERTICES 1
146 #define HAVE_TEX1_VERTICES 1
147 #define HAVE_TEX2_VERTICES 0
148 #define HAVE_TEX3_VERTICES 0
149 #define HAVE_PTEX_VERTICES 1
150
151 #define CHECK_HW_DIVIDE (!(ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE| \
152 DD_TRI_UNFILLED)))
153
154 #define INTERP_VERTEX setup_tab[RADEON_CONTEXT(ctx)->swtcl.SetupIndex].interp
155 #define COPY_PV_VERTEX setup_tab[RADEON_CONTEXT(ctx)->swtcl.SetupIndex].copy_pv
156
157
158 /***********************************************************************
159 * Generate pv-copying and translation functions *
160 ***********************************************************************/
161
162 #define TAG(x) radeon_##x
163 #define IND ~0
164 #include "tnl_dd/t_dd_vb.c"
165 #undef IND
166
167
168 /***********************************************************************
169 * Generate vertex emit and interp functions *
170 ***********************************************************************/
171
172 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT)
173 #define TAG(x) x##_wg
174 #include "tnl_dd/t_dd_vbtmp.h"
175
176 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_TEX0_BIT)
177 #define TAG(x) x##_wgt0
178 #include "tnl_dd/t_dd_vbtmp.h"
179
180 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_TEX0_BIT|RADEON_PTEX_BIT)
181 #define TAG(x) x##_wgpt0
182 #include "tnl_dd/t_dd_vbtmp.h"
183
184 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_TEX0_BIT|RADEON_TEX1_BIT)
185 #define TAG(x) x##_wgt0t1
186 #include "tnl_dd/t_dd_vbtmp.h"
187
188 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_TEX0_BIT|RADEON_TEX1_BIT|\
189 RADEON_PTEX_BIT)
190 #define TAG(x) x##_wgpt0t1
191 #include "tnl_dd/t_dd_vbtmp.h"
192
193 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_SPEC_BIT)
194 #define TAG(x) x##_wgfs
195 #include "tnl_dd/t_dd_vbtmp.h"
196
197 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_SPEC_BIT|\
198 RADEON_TEX0_BIT)
199 #define TAG(x) x##_wgfst0
200 #include "tnl_dd/t_dd_vbtmp.h"
201
202 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_SPEC_BIT|\
203 RADEON_TEX0_BIT|RADEON_PTEX_BIT)
204 #define TAG(x) x##_wgfspt0
205 #include "tnl_dd/t_dd_vbtmp.h"
206
207 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_SPEC_BIT|\
208 RADEON_TEX0_BIT|RADEON_TEX1_BIT)
209 #define TAG(x) x##_wgfst0t1
210 #include "tnl_dd/t_dd_vbtmp.h"
211
212 #define IND (RADEON_XYZW_BIT|RADEON_RGBA_BIT|RADEON_SPEC_BIT|\
213 RADEON_TEX0_BIT|RADEON_TEX1_BIT|RADEON_PTEX_BIT)
214 #define TAG(x) x##_wgfspt0t1
215 #include "tnl_dd/t_dd_vbtmp.h"
216
217
218 /***********************************************************************
219 * Initialization
220 ***********************************************************************/
221
222 static void init_setup_tab( void )
223 {
224 init_wg();
225 init_wgt0();
226 init_wgpt0();
227 init_wgt0t1();
228 init_wgpt0t1();
229 init_wgfs();
230 init_wgfst0();
231 init_wgfspt0();
232 init_wgfst0t1();
233 init_wgfspt0t1();
234 }
235
236
237
238 void radeonPrintSetupFlags(char *msg, GLuint flags )
239 {
240 fprintf(stderr, "%s(%x): %s%s%s%s%s%s\n",
241 msg,
242 (int)flags,
243 (flags & RADEON_XYZW_BIT) ? " xyzw," : "",
244 (flags & RADEON_RGBA_BIT) ? " rgba," : "",
245 (flags & RADEON_SPEC_BIT) ? " spec/fog," : "",
246 (flags & RADEON_TEX0_BIT) ? " tex-0," : "",
247 (flags & RADEON_TEX1_BIT) ? " tex-1," : "",
248 (flags & RADEON_PTEX_BIT) ? " proj-tex," : "");
249 }
250
251
252 static void radeonRenderStart( GLcontext *ctx )
253 {
254 TNLcontext *tnl = TNL_CONTEXT(ctx);
255 radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
256
257 if (!setup_tab[rmesa->swtcl.SetupIndex].check_tex_sizes(ctx)) {
258 GLuint ind = rmesa->swtcl.SetupIndex |= (RADEON_PTEX_BIT|RADEON_RGBA_BIT);
259
260 /* Projective textures are handled nicely; just have to change
261 * up to the new vertex format.
262 */
263 if (setup_tab[ind].vertex_format != rmesa->swtcl.vertex_format) {
264 RADEON_NEWPRIM(rmesa);
265 rmesa->swtcl.vertex_format = setup_tab[ind].vertex_format;
266 rmesa->swtcl.vertex_size = setup_tab[ind].vertex_size;
267 }
268
269 if (!(ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
270 tnl->Driver.Render.Interp = setup_tab[rmesa->swtcl.SetupIndex].interp;
271 tnl->Driver.Render.CopyPV = setup_tab[rmesa->swtcl.SetupIndex].copy_pv;
272 }
273 }
274
275 if (rmesa->dma.flush != 0 &&
276 rmesa->dma.flush != flush_last_swtcl_prim)
277 rmesa->dma.flush( rmesa );
278 }
279
280
281 void radeonBuildVertices( GLcontext *ctx, GLuint start, GLuint count,
282 GLuint newinputs )
283 {
284 radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
285 GLuint stride = rmesa->swtcl.vertex_size * sizeof(int);
286 GLubyte *v = ((GLubyte *)rmesa->swtcl.verts + (start * stride));
287
288 newinputs |= rmesa->swtcl.SetupNewInputs;
289 rmesa->swtcl.SetupNewInputs = 0;
290
291 if (!newinputs)
292 return;
293
294 setup_tab[rmesa->swtcl.SetupIndex].emit( ctx, start, count, v, stride );
295 }
296
297 void radeonChooseVertexState( GLcontext *ctx )
298 {
299 radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
300 TNLcontext *tnl = TNL_CONTEXT(ctx);
301 GLuint ind = (RADEON_XYZW_BIT | RADEON_RGBA_BIT);
302
303 if (!rmesa->TclFallback || rmesa->Fallback)
304 return;
305
306 if (ctx->Fog.Enabled || (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR))
307 ind |= RADEON_SPEC_BIT;
308
309 if (ctx->Texture._EnabledUnits & 0x2)
310 /* unit 1 enabled */
311 ind |= RADEON_TEX0_BIT|RADEON_TEX1_BIT;
312 else if (ctx->Texture._EnabledUnits & 0x1)
313 /* unit 0 enabled */
314 ind |= RADEON_TEX0_BIT;
315
316 rmesa->swtcl.SetupIndex = ind;
317
318 if (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED)) {
319 tnl->Driver.Render.Interp = radeon_interp_extras;
320 tnl->Driver.Render.CopyPV = radeon_copy_pv_extras;
321 }
322 else {
323 tnl->Driver.Render.Interp = setup_tab[ind].interp;
324 tnl->Driver.Render.CopyPV = setup_tab[ind].copy_pv;
325 }
326
327 if (setup_tab[ind].vertex_format != rmesa->swtcl.vertex_format) {
328 RADEON_NEWPRIM(rmesa);
329 rmesa->swtcl.vertex_format = setup_tab[ind].vertex_format;
330 rmesa->swtcl.vertex_size = setup_tab[ind].vertex_size;
331 }
332
333 {
334 GLuint se_coord_fmt, needproj;
335
336 /* HW perspective divide is a win, but tiny vertex formats are a
337 * bigger one.
338 */
339 if (setup_tab[ind].vertex_format == TINY_VERTEX_FORMAT ||
340 (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
341 needproj = GL_TRUE;
342 se_coord_fmt = (RADEON_VTX_XY_PRE_MULT_1_OVER_W0 |
343 RADEON_VTX_Z_PRE_MULT_1_OVER_W0 |
344 RADEON_TEX1_W_ROUTING_USE_Q1);
345 }
346 else {
347 needproj = GL_FALSE;
348 se_coord_fmt = (RADEON_VTX_W0_IS_NOT_1_OVER_W0 |
349 RADEON_TEX1_W_ROUTING_USE_Q1);
350 }
351
352 if ( se_coord_fmt != rmesa->hw.set.cmd[SET_SE_COORDFMT] ) {
353 RADEON_STATECHANGE( rmesa, set );
354 rmesa->hw.set.cmd[SET_SE_COORDFMT] = se_coord_fmt;
355 }
356 _tnl_need_projected_coords( ctx, needproj );
357 }
358 }
359
360
361 /* Flush vertices in the current dma region.
362 */
363 static void flush_last_swtcl_prim( radeonContextPtr rmesa )
364 {
365 if (RADEON_DEBUG & DEBUG_IOCTL)
366 fprintf(stderr, "%s\n", __FUNCTION__);
367
368 rmesa->dma.flush = 0;
369
370 if (rmesa->dma.current.buf) {
371 struct radeon_dma_region *current = &rmesa->dma.current;
372 GLuint current_offset = (rmesa->radeonScreen->gart_buffer_offset +
373 current->buf->buf->idx * RADEON_BUFFER_SIZE +
374 current->start);
375
376 assert (!(rmesa->swtcl.hw_primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
377
378 assert (current->start +
379 rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
380 current->ptr);
381
382 if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
383 radeonEmitVertexAOS( rmesa,
384 rmesa->swtcl.vertex_size,
385 current_offset);
386
387 radeonEmitVbufPrim( rmesa,
388 rmesa->swtcl.vertex_format,
389 rmesa->swtcl.hw_primitive,
390 rmesa->swtcl.numverts);
391 }
392
393 rmesa->swtcl.numverts = 0;
394 current->start = current->ptr;
395 }
396 }
397
398
399 /* Alloc space in the current dma region.
400 */
401 static __inline void *radeonAllocDmaLowVerts( radeonContextPtr rmesa,
402 int nverts, int vsize )
403 {
404 GLuint bytes = vsize * nverts;
405
406 if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end )
407 radeonRefillCurrentDmaRegion( rmesa );
408
409 if (!rmesa->dma.flush) {
410 rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
411 rmesa->dma.flush = flush_last_swtcl_prim;
412 }
413
414 assert( vsize == rmesa->swtcl.vertex_size * 4 );
415 assert( rmesa->dma.flush == flush_last_swtcl_prim );
416 assert (rmesa->dma.current.start +
417 rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
418 rmesa->dma.current.ptr);
419
420
421 {
422 GLubyte *head = (GLubyte *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
423 rmesa->dma.current.ptr += bytes;
424 rmesa->swtcl.numverts += nverts;
425 return head;
426 }
427
428 }
429
430
431
432
433 static void *radeon_emit_contiguous_verts( GLcontext *ctx,
434 GLuint start,
435 GLuint count,
436 void *dest)
437 {
438 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
439 GLuint stride = rmesa->swtcl.vertex_size * 4;
440 setup_tab[rmesa->swtcl.SetupIndex].emit( ctx, start, count, dest, stride );
441 return (void *)((char *)dest + stride * (count - start));
442 }
443
444
445
446 void radeon_emit_indexed_verts( GLcontext *ctx, GLuint start, GLuint count )
447 {
448 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
449
450 radeonAllocDmaRegionVerts( rmesa,
451 &rmesa->swtcl.indexed_verts,
452 count - start,
453 rmesa->swtcl.vertex_size * 4,
454 64);
455
456 setup_tab[rmesa->swtcl.SetupIndex].emit(
457 ctx, start, count,
458 rmesa->swtcl.indexed_verts.address + rmesa->swtcl.indexed_verts.start,
459 rmesa->swtcl.vertex_size * 4 );
460 }
461
462
463 /*
464 * Render unclipped vertex buffers by emitting vertices directly to
465 * dma buffers. Use strip/fan hardware primitives where possible.
466 * Try to simulate missing primitives with indexed vertices.
467 */
468 #define HAVE_POINTS 1
469 #define HAVE_LINES 1
470 #define HAVE_LINE_STRIPS 1
471 #define HAVE_TRIANGLES 1
472 #define HAVE_TRI_STRIPS 1
473 #define HAVE_TRI_STRIP_1 0
474 #define HAVE_TRI_FANS 1
475 #define HAVE_QUADS 0
476 #define HAVE_QUAD_STRIPS 0
477 #define HAVE_POLYGONS 0
478 #define HAVE_ELTS 1
479
480 static const GLuint hw_prim[GL_POLYGON+1] = {
481 RADEON_CP_VC_CNTL_PRIM_TYPE_POINT,
482 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE,
483 0,
484 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE_STRIP,
485 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
486 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_STRIP,
487 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_FAN,
488 0,
489 0,
490 0
491 };
492
493 static __inline void radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim )
494 {
495 RADEON_NEWPRIM( rmesa );
496 rmesa->swtcl.hw_primitive = hw_prim[prim];
497 assert(rmesa->dma.current.ptr == rmesa->dma.current.start);
498 }
499
500 static __inline void radeonEltPrimitive( radeonContextPtr rmesa, GLenum prim )
501 {
502 RADEON_NEWPRIM( rmesa );
503 rmesa->swtcl.hw_primitive = hw_prim[prim] | RADEON_CP_VC_CNTL_PRIM_WALK_IND;
504 }
505
506
507
508
509 #define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx)
510 #define ELTS_VARS( buf ) GLushort *dest = buf
511 #define INIT( prim ) radeonDmaPrimitive( rmesa, prim )
512 #define ELT_INIT(prim) radeonEltPrimitive( rmesa, prim )
513 #define FLUSH() RADEON_NEWPRIM( rmesa )
514 #define GET_CURRENT_VB_MAX_VERTS() \
515 (((int)rmesa->dma.current.end - (int)rmesa->dma.current.ptr) / (rmesa->swtcl.vertex_size*4))
516 #define GET_SUBSEQUENT_VB_MAX_VERTS() \
517 ((RADEON_BUFFER_SIZE) / (rmesa->swtcl.vertex_size*4))
518
519 #if RADEON_OLD_PACKETS
520 # define GET_CURRENT_VB_MAX_ELTS() \
521 ((RADEON_CMD_BUF_SZ - (rmesa->store.cmd_used + 24)) / 2)
522 #else
523 # define GET_CURRENT_VB_MAX_ELTS() \
524 ((RADEON_CMD_BUF_SZ - (rmesa->store.cmd_used + 16)) / 2)
525 #endif
526 #define GET_SUBSEQUENT_VB_MAX_ELTS() \
527 ((RADEON_CMD_BUF_SZ - 1024) / 2)
528
529
530 static void *radeon_alloc_elts( radeonContextPtr rmesa, int nr )
531 {
532 if (rmesa->dma.flush == radeonFlushElts &&
533 rmesa->store.cmd_used + nr*2 < RADEON_CMD_BUF_SZ) {
534
535 rmesa->store.cmd_used += nr*2;
536
537 return (void *)(rmesa->store.cmd_buf + rmesa->store.cmd_used);
538 }
539 else {
540 if (rmesa->dma.flush) {
541 rmesa->dma.flush( rmesa );
542 }
543
544 radeonEmitVertexAOS( rmesa,
545 rmesa->swtcl.vertex_size,
546 (rmesa->radeonScreen->gart_buffer_offset +
547 rmesa->swtcl.indexed_verts.buf->buf->idx *
548 RADEON_BUFFER_SIZE +
549 rmesa->swtcl.indexed_verts.start));
550
551 return (void *) radeonAllocEltsOpenEnded( rmesa,
552 rmesa->swtcl.vertex_format,
553 rmesa->swtcl.hw_primitive,
554 nr );
555 }
556 }
557
558 #define ALLOC_ELTS(nr) radeon_alloc_elts(rmesa, nr)
559
560 #ifdef MESA_BIG_ENDIAN
561 /* We could do without (most of) this ugliness if dest was always 32 bit word aligned... */
562 #define EMIT_ELT(offset, x) do { \
563 int off = offset + ( ( (GLuint)dest & 0x2 ) >> 1 ); \
564 GLushort *des = (GLushort *)( (GLuint)dest & ~0x2 ); \
565 (des)[ off + 1 - 2 * ( off & 1 ) ] = (GLushort)(x); } while (0)
566 #else
567 #define EMIT_ELT(offset, x) (dest)[offset] = (GLushort) (x)
568 #endif
569 #define EMIT_TWO_ELTS(offset, x, y) *(GLuint *)(dest+offset) = ((y)<<16)|(x);
570 #define INCR_ELTS( nr ) dest += nr
571 #define ELTPTR dest
572 #define RELEASE_ELT_VERTS() \
573 radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, __FUNCTION__ )
574 #define EMIT_INDEXED_VERTS( ctx, start, count ) \
575 radeon_emit_indexed_verts( ctx, start, count )
576
577
578 #define ALLOC_VERTS( nr ) \
579 radeonAllocDmaLowVerts( rmesa, nr, rmesa->swtcl.vertex_size * 4 )
580 #define EMIT_VERTS( ctx, j, nr, buf ) \
581 radeon_emit_contiguous_verts(ctx, j, (j)+(nr), buf)
582
583 #define TAG(x) radeon_dma_##x
584 #include "tnl_dd/t_dd_dmatmp.h"
585
586
587 /**********************************************************************/
588 /* Render pipeline stage */
589 /**********************************************************************/
590
591
592 static GLboolean radeon_run_render( GLcontext *ctx,
593 struct tnl_pipeline_stage *stage )
594 {
595 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
596 TNLcontext *tnl = TNL_CONTEXT(ctx);
597 struct vertex_buffer *VB = &tnl->vb;
598 tnl_render_func *tab = TAG(render_tab_verts);
599 GLuint i;
600
601 if (rmesa->swtcl.indexed_verts.buf && (!VB->Elts || stage->changed_inputs))
602 RELEASE_ELT_VERTS();
603
604 if (rmesa->swtcl.RenderIndex != 0 ||
605 !radeon_dma_validate_render( ctx, VB ))
606 return GL_TRUE;
607
608 tnl->Driver.Render.Start( ctx );
609
610 if (VB->Elts) {
611 tab = TAG(render_tab_elts);
612 if (!rmesa->swtcl.indexed_verts.buf) {
613 if (VB->Count > GET_SUBSEQUENT_VB_MAX_VERTS())
614 return GL_TRUE;
615 EMIT_INDEXED_VERTS(ctx, 0, VB->Count);
616 }
617 }
618
619 for (i = 0 ; i < VB->PrimitiveCount ; i++)
620 {
621 GLuint prim = VB->Primitive[i].mode;
622 GLuint start = VB->Primitive[i].start;
623 GLuint length = VB->Primitive[i].count;
624
625 if (!length)
626 continue;
627
628 if (RADEON_DEBUG & DEBUG_PRIMS)
629 fprintf(stderr, "radeon_render.c: prim %s %d..%d\n",
630 _mesa_lookup_enum_by_nr(prim & PRIM_MODE_MASK),
631 start, start+length);
632
633 if (length)
634 tab[prim & PRIM_MODE_MASK]( ctx, start, start + length, prim );
635 }
636
637 tnl->Driver.Render.Finish( ctx );
638
639 return GL_FALSE; /* finished the pipe */
640 }
641
642
643
644 static void radeon_check_render( GLcontext *ctx,
645 struct tnl_pipeline_stage *stage )
646 {
647 stage->inputs = TNL_CONTEXT(ctx)->render_inputs;
648 }
649
650
651 static void dtr( struct tnl_pipeline_stage *stage )
652 {
653 (void)stage;
654 }
655
656
657 const struct tnl_pipeline_stage _radeon_render_stage =
658 {
659 "radeon render",
660 (_DD_NEW_SEPARATE_SPECULAR |
661 _NEW_TEXTURE|
662 _NEW_FOG|
663 _NEW_RENDERMODE), /* re-check (new inputs) */
664 0, /* re-run (always runs) */
665 GL_TRUE, /* active */
666 0, 0, /* inputs (set in check_render), outputs */
667 0, 0, /* changed_inputs, private */
668 dtr, /* destructor */
669 radeon_check_render, /* check - initially set to alloc data */
670 radeon_run_render /* run */
671 };
672
673
674 /**************************************************************************/
675
676 /* Radeon texture rectangle expects coords in 0..1 range, not 0..dimension
677 * as in the extension spec. Need to translate here.
678 *
679 * Note that swrast expects 0..dimension, so if a fallback is active,
680 * don't do anything. (Maybe need to configure swrast to match hw)
681 */
682 struct texrect_stage_data {
683 GLvector4f texcoord[MAX_TEXTURE_UNITS];
684 };
685
686 #define TEXRECT_STAGE_DATA(stage) ((struct texrect_stage_data *)stage->privatePtr)
687
688
689 static GLboolean run_texrect_stage( GLcontext *ctx,
690 struct tnl_pipeline_stage *stage )
691 {
692 struct texrect_stage_data *store = TEXRECT_STAGE_DATA(stage);
693 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
694 TNLcontext *tnl = TNL_CONTEXT(ctx);
695 struct vertex_buffer *VB = &tnl->vb;
696 GLuint i;
697
698 if (rmesa->Fallback)
699 return GL_TRUE;
700
701 for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++) {
702 if (!(ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_RECT_BIT))
703 continue;
704
705 if (stage->changed_inputs & VERT_BIT_TEX(i)) {
706 struct gl_texture_object *texObj = ctx->Texture.Unit[i].CurrentRect;
707 struct gl_texture_image *texImage = texObj->Image[0][texObj->BaseLevel];
708 const GLfloat iw = 1.0/texImage->Width;
709 const GLfloat ih = 1.0/texImage->Height;
710 GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data;
711 GLint instride = VB->TexCoordPtr[i]->stride;
712 GLfloat (*out)[4] = store->texcoord[i].data;
713 GLint j;
714
715 for (j = 0 ; j < VB->Count ; j++) {
716 out[j][0] = in[0] * iw;
717 out[j][1] = in[1] * ih;
718 in = (GLfloat *)((GLubyte *)in + instride);
719 }
720 }
721
722 VB->TexCoordPtr[i] = &store->texcoord[i];
723 }
724
725 return GL_TRUE;
726 }
727
728
729 /* Called the first time stage->run() is invoked.
730 */
731 static GLboolean alloc_texrect_data( GLcontext *ctx,
732 struct tnl_pipeline_stage *stage )
733 {
734 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
735 struct texrect_stage_data *store;
736 GLuint i;
737
738 stage->privatePtr = CALLOC(sizeof(*store));
739 store = TEXRECT_STAGE_DATA(stage);
740 if (!store)
741 return GL_FALSE;
742
743 for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++)
744 _mesa_vector4f_alloc( &store->texcoord[i], 0, VB->Size, 32 );
745
746 /* Now run the stage.
747 */
748 stage->run = run_texrect_stage;
749 return stage->run( ctx, stage );
750 }
751
752
753 static void check_texrect( GLcontext *ctx,
754 struct tnl_pipeline_stage *stage )
755 {
756 GLuint flags = 0;
757
758 if (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT)
759 flags |= VERT_BIT_TEX0;
760
761 if (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT)
762 flags |= VERT_BIT_TEX1;
763
764 stage->inputs = flags;
765 stage->outputs = flags;
766 stage->active = (flags != 0);
767 }
768
769
770 static void free_texrect_data( struct tnl_pipeline_stage *stage )
771 {
772 struct texrect_stage_data *store = TEXRECT_STAGE_DATA(stage);
773 GLuint i;
774
775 if (store) {
776 for (i = 0 ; i < MAX_TEXTURE_UNITS ; i++)
777 if (store->texcoord[i].data)
778 _mesa_vector4f_free( &store->texcoord[i] );
779 FREE( store );
780 stage->privatePtr = 0;
781 }
782 }
783
784
785 const struct tnl_pipeline_stage _radeon_texrect_stage =
786 {
787 "radeon texrect stage", /* name */
788 _NEW_TEXTURE, /* check_state */
789 _NEW_TEXTURE, /* run_state */
790 GL_TRUE, /* active? */
791 0, /* inputs */
792 0, /* outputs */
793 0, /* changed_inputs */
794 NULL, /* private data */
795 free_texrect_data, /* destructor */
796 check_texrect, /* check */
797 alloc_texrect_data, /* run -- initially set to init */
798 };
799
800
801 /**************************************************************************/
802
803
804 static const GLuint reduced_hw_prim[GL_POLYGON+1] = {
805 RADEON_CP_VC_CNTL_PRIM_TYPE_POINT,
806 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE,
807 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE,
808 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE,
809 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
810 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
811 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
812 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
813 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
814 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
815 };
816
817 static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim );
818 static void radeonRenderPrimitive( GLcontext *ctx, GLenum prim );
819 static void radeonResetLineStipple( GLcontext *ctx );
820
821
822 /***********************************************************************
823 * Emit primitives as inline vertices *
824 ***********************************************************************/
825
826 #undef LOCAL_VARS
827 #undef ALLOC_VERTS
828 #define CTX_ARG radeonContextPtr rmesa
829 #define CTX_ARG2 rmesa
830 #define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
831 #define ALLOC_VERTS( n, size ) radeonAllocDmaLowVerts( rmesa, n, size * 4 )
832 #undef LOCAL_VARS
833 #define LOCAL_VARS \
834 radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \
835 const char *radeonverts = (char *)rmesa->swtcl.verts;
836 #define VERT(x) (radeonVertex *)(radeonverts + (x * vertsize * sizeof(int)))
837 #define VERTEX radeonVertex
838 #undef TAG
839 #define TAG(x) radeon_##x
840 #include "tnl_dd/t_dd_triemit.h"
841
842
843 /***********************************************************************
844 * Macros for t_dd_tritmp.h to draw basic primitives *
845 ***********************************************************************/
846
847 #define QUAD( a, b, c, d ) radeon_quad( rmesa, a, b, c, d )
848 #define TRI( a, b, c ) radeon_triangle( rmesa, a, b, c )
849 #define LINE( a, b ) radeon_line( rmesa, a, b )
850 #define POINT( a ) radeon_point( rmesa, a )
851
852 /***********************************************************************
853 * Build render functions from dd templates *
854 ***********************************************************************/
855
856 #define RADEON_TWOSIDE_BIT 0x01
857 #define RADEON_UNFILLED_BIT 0x02
858 #define RADEON_MAX_TRIFUNC 0x08
859
860
861 static struct {
862 tnl_points_func points;
863 tnl_line_func line;
864 tnl_triangle_func triangle;
865 tnl_quad_func quad;
866 } rast_tab[RADEON_MAX_TRIFUNC];
867
868
869 #define DO_FALLBACK 0
870 #define DO_OFFSET 0
871 #define DO_UNFILLED (IND & RADEON_UNFILLED_BIT)
872 #define DO_TWOSIDE (IND & RADEON_TWOSIDE_BIT)
873 #define DO_FLAT 0
874 #define DO_TRI 1
875 #define DO_QUAD 1
876 #define DO_LINE 1
877 #define DO_POINTS 1
878 #define DO_FULL_QUAD 1
879
880 #define HAVE_RGBA 1
881 #define HAVE_SPEC 1
882 #define HAVE_INDEX 0
883 #define HAVE_BACK_COLORS 0
884 #define HAVE_HW_FLATSHADE 1
885 #define TAB rast_tab
886
887 #define DEPTH_SCALE 1.0
888 #define UNFILLED_TRI unfilled_tri
889 #define UNFILLED_QUAD unfilled_quad
890 #define VERT_X(_v) _v->v.x
891 #define VERT_Y(_v) _v->v.y
892 #define VERT_Z(_v) _v->v.z
893 #define AREA_IS_CCW( a ) (a < 0)
894 #define GET_VERTEX(e) (rmesa->swtcl.verts + (e * rmesa->swtcl.vertex_size * sizeof(int)))
895
896 #define VERT_SET_RGBA( v, c ) \
897 do { \
898 radeon_color_t *color = (radeon_color_t *)&((v)->ui[coloroffset]); \
899 UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]); \
900 UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]); \
901 UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]); \
902 UNCLAMPED_FLOAT_TO_UBYTE(color->alpha, (c)[3]); \
903 } while (0)
904
905 #define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
906
907 #define VERT_SET_SPEC( v0, c ) \
908 do { \
909 if (havespec) { \
910 UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.red, (c)[0]); \
911 UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.green, (c)[1]); \
912 UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.blue, (c)[2]); \
913 } \
914 } while (0)
915 #define VERT_COPY_SPEC( v0, v1 ) \
916 do { \
917 if (havespec) { \
918 v0->v.specular.red = v1->v.specular.red; \
919 v0->v.specular.green = v1->v.specular.green; \
920 v0->v.specular.blue = v1->v.specular.blue; \
921 } \
922 } while (0)
923
924 /* These don't need LE32_TO_CPU() as they used to save and restore
925 * colors which are already in the correct format.
926 */
927 #define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[coloroffset]
928 #define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
929 #define VERT_SAVE_SPEC( idx ) if (havespec) spec[idx] = v[idx]->ui[5]
930 #define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[5] = spec[idx]
931
932 #undef LOCAL_VARS
933 #undef TAG
934 #undef INIT
935
936 #define LOCAL_VARS(n) \
937 radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \
938 GLuint color[n], spec[n]; \
939 GLuint coloroffset = (rmesa->swtcl.vertex_size == 4 ? 3 : 4); \
940 GLboolean havespec = (rmesa->swtcl.vertex_size > 4); \
941 (void) color; (void) spec; (void) coloroffset; (void) havespec;
942
943 /***********************************************************************
944 * Helpers for rendering unfilled primitives *
945 ***********************************************************************/
946
947 #define RASTERIZE(x) radeonRasterPrimitive( ctx, reduced_hw_prim[x] )
948 #define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
949 #undef TAG
950 #define TAG(x) x
951 #include "tnl_dd/t_dd_unfilled.h"
952 #undef IND
953
954
955 /***********************************************************************
956 * Generate GL render functions *
957 ***********************************************************************/
958
959
960 #define IND (0)
961 #define TAG(x) x
962 #include "tnl_dd/t_dd_tritmp.h"
963
964 #define IND (RADEON_TWOSIDE_BIT)
965 #define TAG(x) x##_twoside
966 #include "tnl_dd/t_dd_tritmp.h"
967
968 #define IND (RADEON_UNFILLED_BIT)
969 #define TAG(x) x##_unfilled
970 #include "tnl_dd/t_dd_tritmp.h"
971
972 #define IND (RADEON_TWOSIDE_BIT|RADEON_UNFILLED_BIT)
973 #define TAG(x) x##_twoside_unfilled
974 #include "tnl_dd/t_dd_tritmp.h"
975
976
977 static void init_rast_tab( void )
978 {
979 init();
980 init_twoside();
981 init_unfilled();
982 init_twoside_unfilled();
983 }
984
985 /**********************************************************************/
986 /* Render unclipped begin/end objects */
987 /**********************************************************************/
988
989 #define VERT(x) (radeonVertex *)(radeonverts + (x * vertsize * sizeof(int)))
990 #define RENDER_POINTS( start, count ) \
991 for ( ; start < count ; start++) \
992 radeon_point( rmesa, VERT(start) )
993 #define RENDER_LINE( v0, v1 ) \
994 radeon_line( rmesa, VERT(v0), VERT(v1) )
995 #define RENDER_TRI( v0, v1, v2 ) \
996 radeon_triangle( rmesa, VERT(v0), VERT(v1), VERT(v2) )
997 #define RENDER_QUAD( v0, v1, v2, v3 ) \
998 radeon_quad( rmesa, VERT(v0), VERT(v1), VERT(v2), VERT(v3) )
999 #undef INIT
1000 #define INIT(x) do { \
1001 radeonRenderPrimitive( ctx, x ); \
1002 } while (0)
1003 #undef LOCAL_VARS
1004 #define LOCAL_VARS \
1005 radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \
1006 const GLuint vertsize = rmesa->swtcl.vertex_size; \
1007 const char *radeonverts = (char *)rmesa->swtcl.verts; \
1008 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
1009 const GLboolean stipple = ctx->Line.StippleFlag; \
1010 (void) elt; (void) stipple;
1011 #define RESET_STIPPLE if ( stipple ) radeonResetLineStipple( ctx );
1012 #define RESET_OCCLUSION
1013 #define PRESERVE_VB_DEFS
1014 #define ELT(x) (x)
1015 #define TAG(x) radeon_##x##_verts
1016 #include "tnl/t_vb_rendertmp.h"
1017 #undef ELT
1018 #undef TAG
1019 #define TAG(x) radeon_##x##_elts
1020 #define ELT(x) elt[x]
1021 #include "tnl/t_vb_rendertmp.h"
1022
1023
1024
1025 /**********************************************************************/
1026 /* Choose render functions */
1027 /**********************************************************************/
1028
1029 void radeonChooseRenderState( GLcontext *ctx )
1030 {
1031 TNLcontext *tnl = TNL_CONTEXT(ctx);
1032 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1033 GLuint index = 0;
1034 GLuint flags = ctx->_TriangleCaps;
1035
1036 if (!rmesa->TclFallback || rmesa->Fallback)
1037 return;
1038
1039 if (flags & DD_TRI_LIGHT_TWOSIDE) index |= RADEON_TWOSIDE_BIT;
1040 if (flags & DD_TRI_UNFILLED) index |= RADEON_UNFILLED_BIT;
1041
1042 if (index != rmesa->swtcl.RenderIndex) {
1043 tnl->Driver.Render.Points = rast_tab[index].points;
1044 tnl->Driver.Render.Line = rast_tab[index].line;
1045 tnl->Driver.Render.ClippedLine = rast_tab[index].line;
1046 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1047 tnl->Driver.Render.Quad = rast_tab[index].quad;
1048
1049 if (index == 0) {
1050 tnl->Driver.Render.PrimTabVerts = radeon_render_tab_verts;
1051 tnl->Driver.Render.PrimTabElts = radeon_render_tab_elts;
1052 tnl->Driver.Render.ClippedPolygon = radeon_fast_clipped_poly;
1053 } else {
1054 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1055 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1056 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
1057 }
1058
1059 rmesa->swtcl.RenderIndex = index;
1060 }
1061 }
1062
1063
1064 /**********************************************************************/
1065 /* High level hooks for t_vb_render.c */
1066 /**********************************************************************/
1067
1068
1069 static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim )
1070 {
1071 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1072
1073 if (rmesa->swtcl.hw_primitive != hwprim) {
1074 RADEON_NEWPRIM( rmesa );
1075 rmesa->swtcl.hw_primitive = hwprim;
1076 }
1077 }
1078
1079 static void radeonRenderPrimitive( GLcontext *ctx, GLenum prim )
1080 {
1081 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1082 rmesa->swtcl.render_primitive = prim;
1083 if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED))
1084 radeonRasterPrimitive( ctx, reduced_hw_prim[prim] );
1085 }
1086
1087 static void radeonRenderFinish( GLcontext *ctx )
1088 {
1089 }
1090
1091 static void radeonResetLineStipple( GLcontext *ctx )
1092 {
1093 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1094 RADEON_STATECHANGE( rmesa, lin );
1095 }
1096
1097
1098 /**********************************************************************/
1099 /* Transition to/from hardware rasterization. */
1100 /**********************************************************************/
1101
1102 static const char * const fallbackStrings[] = {
1103 "Texture mode",
1104 "glDrawBuffer(GL_FRONT_AND_BACK)",
1105 "glEnable(GL_STENCIL) without hw stencil buffer",
1106 "glRenderMode(selection or feedback)",
1107 "glBlendEquation",
1108 "glBlendFunc",
1109 "RADEON_NO_RAST",
1110 "Mixing GL_CLAMP_TO_BORDER and GL_CLAMP (or GL_MIRROR_CLAMP_ATI)"
1111 };
1112
1113
1114 static const char *getFallbackString(GLuint bit)
1115 {
1116 int i = 0;
1117 while (bit > 1) {
1118 i++;
1119 bit >>= 1;
1120 }
1121 return fallbackStrings[i];
1122 }
1123
1124
1125 void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
1126 {
1127 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1128 TNLcontext *tnl = TNL_CONTEXT(ctx);
1129 GLuint oldfallback = rmesa->Fallback;
1130
1131 if (mode) {
1132 rmesa->Fallback |= bit;
1133 if (oldfallback == 0) {
1134 RADEON_FIREVERTICES( rmesa );
1135 TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE );
1136 _swsetup_Wakeup( ctx );
1137 _tnl_need_projected_coords( ctx, GL_TRUE );
1138 rmesa->swtcl.RenderIndex = ~0;
1139 if (RADEON_DEBUG & DEBUG_FALLBACKS) {
1140 fprintf(stderr, "Radeon begin rasterization fallback: 0x%x %s\n",
1141 bit, getFallbackString(bit));
1142 }
1143 }
1144 }
1145 else {
1146 rmesa->Fallback &= ~bit;
1147 if (oldfallback == bit) {
1148 _swrast_flush( ctx );
1149 tnl->Driver.Render.Start = radeonRenderStart;
1150 tnl->Driver.Render.PrimitiveNotify = radeonRenderPrimitive;
1151 tnl->Driver.Render.Finish = radeonRenderFinish;
1152 tnl->Driver.Render.BuildVertices = radeonBuildVertices;
1153 tnl->Driver.Render.ResetLineStipple = radeonResetLineStipple;
1154 TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_FALSE );
1155 if (rmesa->TclFallback) {
1156 /* These are already done if rmesa->TclFallback goes to
1157 * zero above. But not if it doesn't (RADEON_NO_TCL for
1158 * example?)
1159 */
1160 radeonChooseVertexState( ctx );
1161 radeonChooseRenderState( ctx );
1162 }
1163 if (RADEON_DEBUG & DEBUG_FALLBACKS) {
1164 fprintf(stderr, "Radeon end rasterization fallback: 0x%x %s\n",
1165 bit, getFallbackString(bit));
1166 }
1167 }
1168 }
1169 }
1170
1171
1172 void radeonFlushVertices( GLcontext *ctx, GLuint flags )
1173 {
1174 _tnl_FlushVertices( ctx, flags );
1175
1176 if (flags & FLUSH_STORED_VERTICES)
1177 RADEON_NEWPRIM( RADEON_CONTEXT( ctx ) );
1178 }
1179
1180 /**********************************************************************/
1181 /* Initialization. */
1182 /**********************************************************************/
1183
1184 void radeonInitSwtcl( GLcontext *ctx )
1185 {
1186 TNLcontext *tnl = TNL_CONTEXT(ctx);
1187 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1188 GLuint size = TNL_CONTEXT(ctx)->vb.Size;
1189 static int firsttime = 1;
1190
1191 if (firsttime) {
1192 init_rast_tab();
1193 init_setup_tab();
1194 firsttime = 0;
1195 }
1196
1197 tnl->Driver.Render.Start = radeonRenderStart;
1198 tnl->Driver.Render.Finish = radeonRenderFinish;
1199 tnl->Driver.Render.PrimitiveNotify = radeonRenderPrimitive;
1200 tnl->Driver.Render.ResetLineStipple = radeonResetLineStipple;
1201 tnl->Driver.Render.BuildVertices = radeonBuildVertices;
1202
1203 rmesa->swtcl.verts = (GLubyte *)ALIGN_MALLOC( size * 16 * 4, 32 );
1204 rmesa->swtcl.RenderIndex = ~0;
1205 rmesa->swtcl.render_primitive = GL_TRIANGLES;
1206 rmesa->swtcl.hw_primitive = 0;
1207 }
1208
1209
1210 void radeonDestroySwtcl( GLcontext *ctx )
1211 {
1212 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
1213
1214 if (rmesa->swtcl.indexed_verts.buf)
1215 radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts,
1216 __FUNCTION__ );
1217
1218 if (rmesa->swtcl.verts) {
1219 ALIGN_FREE(rmesa->swtcl.verts);
1220 rmesa->swtcl.verts = 0;
1221 }
1222
1223 }