projects / mesa.git / blob
? search:


f3f558b7def05c50a6e7133ab6945e426d621327
[mesa.git] / src / mesa / drivers / dri / r200 / r200_tcl.c
1 /*
2 Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved.
3
4 The Weather Channel (TM) funded Tungsten Graphics to develop the
5 initial release of the Radeon 8500 driver under the XFree86 license.
6 This notice must be preserved.
7
8 Permission is hereby granted, free of charge, to any person obtaining
9 a copy of this software and associated documentation files (the
10 "Software"), to deal in the Software without restriction, including
11 without limitation the rights to use, copy, modify, merge, publish,
12 distribute, sublicense, and/or sell copies of the Software, and to
13 permit persons to whom the Software is furnished to do so, subject to
14 the following conditions:
15
16 The above copyright notice and this permission notice (including the
17 next paragraph) shall be included in all copies or substantial
18 portions of the Software.
19
20 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
21 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
22 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
23 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
24 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
25 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
26 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
27
28 **************************************************************************/
29
30 /*
31 * Authors:
32 * Keith Whitwell <keith@tungstengraphics.com>
33 */
34
35 #include "main/glheader.h"
36 #include "main/imports.h"
37 #include "main/mtypes.h"
38 #include "main/enums.h"
39 #include "main/colormac.h"
40 #include "main/light.h"
41
42 #include "vbo/vbo.h"
43 #include "tnl/tnl.h"
44 #include "tnl/t_pipeline.h"
45
46 #include "r200_context.h"
47 #include "r200_state.h"
48 #include "r200_ioctl.h"
49 #include "r200_tcl.h"
50 #include "r200_swtcl.h"
51 #include "r200_maos.h"
52
53 #include "radeon_common_context.h"
54
55
56
57 #define HAVE_POINTS 1
58 #define HAVE_LINES 1
59 #define HAVE_LINE_LOOP 0
60 #define HAVE_LINE_STRIPS 1
61 #define HAVE_TRIANGLES 1
62 #define HAVE_TRI_STRIPS 1
63 #define HAVE_TRI_STRIP_1 0
64 #define HAVE_TRI_FANS 1
65 #define HAVE_QUADS 1
66 #define HAVE_QUAD_STRIPS 1
67 #define HAVE_POLYGONS 1
68 #define HAVE_ELTS 1
69
70
71 #define HW_POINTS ((ctx->Point.PointSprite || \
72 ((ctx->_TriangleCaps & (DD_POINT_SIZE | DD_POINT_ATTEN)) && \
73 !(ctx->_TriangleCaps & (DD_POINT_SMOOTH)))) ? \
74 R200_VF_PRIM_POINT_SPRITES : R200_VF_PRIM_POINTS)
75 #define HW_LINES R200_VF_PRIM_LINES
76 #define HW_LINE_LOOP 0
77 #define HW_LINE_STRIP R200_VF_PRIM_LINE_STRIP
78 #define HW_TRIANGLES R200_VF_PRIM_TRIANGLES
79 #define HW_TRIANGLE_STRIP_0 R200_VF_PRIM_TRIANGLE_STRIP
80 #define HW_TRIANGLE_STRIP_1 0
81 #define HW_TRIANGLE_FAN R200_VF_PRIM_TRIANGLE_FAN
82 #define HW_QUADS R200_VF_PRIM_QUADS
83 #define HW_QUAD_STRIP R200_VF_PRIM_QUAD_STRIP
84 #define HW_POLYGON R200_VF_PRIM_POLYGON
85
86
87 static GLboolean discrete_prim[0x10] = {
88 0, /* 0 none */
89 1, /* 1 points */
90 1, /* 2 lines */
91 0, /* 3 line_strip */
92 1, /* 4 tri_list */
93 0, /* 5 tri_fan */
94 0, /* 6 tri_strip */
95 0, /* 7 tri_w_flags */
96 1, /* 8 rect list (unused) */
97 1, /* 9 3vert point */
98 1, /* a 3vert line */
99 0, /* b point sprite */
100 0, /* c line loop */
101 1, /* d quads */
102 0, /* e quad strip */
103 0, /* f polygon */
104 };
105
106
107 #define LOCAL_VARS r200ContextPtr rmesa = R200_CONTEXT(ctx)
108 #define ELT_TYPE GLushort
109
110 #define ELT_INIT(prim, hw_prim) \
111 r200TclPrimitive( ctx, prim, hw_prim | R200_VF_PRIM_WALK_IND )
112
113 #define GET_MESA_ELTS() TNL_CONTEXT(ctx)->vb.Elts
114
115
116 /* Don't really know how many elts will fit in what's left of cmdbuf,
117 * as there is state to emit, etc:
118 */
119
120 /* Testing on isosurf shows a maximum around here. Don't know if it's
121 * the card or driver or kernel module that is causing the behaviour.
122 */
123 #define GET_MAX_HW_ELTS() 300
124
125 #define RESET_STIPPLE() do { \
126 R200_STATECHANGE( rmesa, lin ); \
127 radeonEmitState(&rmesa->radeon); \
128 } while (0)
129
130 #define AUTO_STIPPLE( mode ) do { \
131 R200_STATECHANGE( rmesa, lin ); \
132 if (mode) \
133 rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] |= \
134 R200_LINE_PATTERN_AUTO_RESET; \
135 else \
136 rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &= \
137 ~R200_LINE_PATTERN_AUTO_RESET; \
138 radeonEmitState(&rmesa->radeon); \
139 } while (0)
140
141
142 #define ALLOC_ELTS(nr) r200AllocElts( rmesa, nr )
143
144 static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr )
145 {
146 if (rmesa->radeon.dma.flush == r200FlushElts &&
147 rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) {
148
149 GLushort *dest = (GLushort *)(rmesa->radeon.tcl.elt_dma_bo->ptr +
150 rmesa->radeon.tcl.elt_dma_offset + rmesa->tcl.elt_used);
151
152 rmesa->tcl.elt_used += nr*2;
153
154 return dest;
155 }
156 else {
157 if (rmesa->radeon.dma.flush)
158 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
159
160 r200EmitAOS( rmesa,
161 rmesa->radeon.tcl.aos_count, 0 );
162
163 r200EmitMaxVtxIndex(rmesa, rmesa->radeon.tcl.aos[0].count);
164 return r200AllocEltsOpenEnded( rmesa, rmesa->tcl.hw_primitive, nr );
165 }
166 }
167
168
169 #define CLOSE_ELTS() \
170 do { \
171 if (0) R200_NEWPRIM( rmesa ); \
172 } \
173 while (0)
174
175
176 /* TODO: Try to extend existing primitive if both are identical,
177 * discrete and there are no intervening state changes. (Somewhat
178 * duplicates changes to DrawArrays code)
179 */
180 static void r200EmitPrim( GLcontext *ctx,
181 GLenum prim,
182 GLuint hwprim,
183 GLuint start,
184 GLuint count)
185 {
186 r200ContextPtr rmesa = R200_CONTEXT( ctx );
187 r200TclPrimitive( ctx, prim, hwprim );
188
189 // fprintf(stderr,"Emit prim %d\n", rmesa->radeon.tcl.aos_count);
190
191 r200EmitAOS( rmesa,
192 rmesa->radeon.tcl.aos_count,
193 start );
194
195 /* Why couldn't this packet have taken an offset param?
196 */
197 r200EmitVbufPrim( rmesa,
198 rmesa->tcl.hw_primitive,
199 count - start );
200 }
201
202 #define EMIT_PRIM(ctx, prim, hwprim, start, count) do { \
203 r200EmitPrim( ctx, prim, hwprim, start, count ); \
204 (void) rmesa; } while (0)
205
206 #define MAX_CONVERSION_SIZE 40
207 /* Try & join small primitives
208 */
209 #if 0
210 #define PREFER_DISCRETE_ELT_PRIM( NR, PRIM ) 0
211 #else
212 #define PREFER_DISCRETE_ELT_PRIM( NR, PRIM ) \
213 ((NR) < 20 || \
214 ((NR) < 40 && \
215 rmesa->tcl.hw_primitive == (PRIM| \
216 R200_VF_TCL_OUTPUT_VTX_ENABLE| \
217 R200_VF_PRIM_WALK_IND)))
218 #endif
219
220 #ifdef MESA_BIG_ENDIAN
221 /* We could do without (most of) this ugliness if dest was always 32 bit word aligned... */
222 #define EMIT_ELT(dest, offset, x) do { \
223 int off = offset + ( ( (GLuint)dest & 0x2 ) >> 1 ); \
224 GLushort *des = (GLushort *)( (GLuint)dest & ~0x2 ); \
225 (des)[ off + 1 - 2 * ( off & 1 ) ] = (GLushort)(x); \
226 (void)rmesa; } while (0)
227 #else
228 #define EMIT_ELT(dest, offset, x) do { \
229 (dest)[offset] = (GLushort) (x); \
230 (void)rmesa; } while (0)
231 #endif
232
233 #define EMIT_TWO_ELTS(dest, offset, x, y) *(GLuint *)((dest)+offset) = ((y)<<16)|(x);
234
235
236
237 #define TAG(x) tcl_##x
238 #include "tnl_dd/t_dd_dmatmp2.h"
239
240 /**********************************************************************/
241 /* External entrypoints */
242 /**********************************************************************/
243
244 void r200EmitPrimitive( GLcontext *ctx,
245 GLuint first,
246 GLuint last,
247 GLuint flags )
248 {
249 tcl_render_tab_verts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
250 }
251
252 void r200EmitEltPrimitive( GLcontext *ctx,
253 GLuint first,
254 GLuint last,
255 GLuint flags )
256 {
257 tcl_render_tab_elts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
258 }
259
260 void r200TclPrimitive( GLcontext *ctx,
261 GLenum prim,
262 int hw_prim )
263 {
264 r200ContextPtr rmesa = R200_CONTEXT(ctx);
265 GLuint newprim = hw_prim | R200_VF_TCL_OUTPUT_VTX_ENABLE;
266
267 if (newprim != rmesa->tcl.hw_primitive ||
268 !discrete_prim[hw_prim&0xf]) {
269 /* need to disable perspective-correct texturing for point sprites */
270 if ((prim & PRIM_MODE_MASK) == GL_POINTS && ctx->Point.PointSprite) {
271 if (rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE) {
272 R200_STATECHANGE( rmesa, set );
273 rmesa->hw.set.cmd[SET_RE_CNTL] &= ~R200_PERSPECTIVE_ENABLE;
274 }
275 }
276 else if (!(rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE)) {
277 R200_STATECHANGE( rmesa, set );
278 rmesa->hw.set.cmd[SET_RE_CNTL] |= R200_PERSPECTIVE_ENABLE;
279 }
280 R200_NEWPRIM( rmesa );
281 rmesa->tcl.hw_primitive = newprim;
282 }
283 }
284
285
286 /**********************************************************************/
287 /* Fog blend factor computation for hw tcl */
288 /* same calculation used as in t_vb_fog.c */
289 /**********************************************************************/
290
291 #define FOG_EXP_TABLE_SIZE 256
292 #define FOG_MAX (10.0)
293 #define EXP_FOG_MAX .0006595
294 #define FOG_INCR (FOG_MAX/FOG_EXP_TABLE_SIZE)
295 static GLfloat exp_table[FOG_EXP_TABLE_SIZE];
296
297 #if 1
298 #define NEG_EXP( result, narg ) \
299 do { \
300 GLfloat f = (GLfloat) (narg * (1.0/FOG_INCR)); \
301 GLint k = (GLint) f; \
302 if (k > FOG_EXP_TABLE_SIZE-2) \
303 result = (GLfloat) EXP_FOG_MAX; \
304 else \
305 result = exp_table[k] + (f-k)*(exp_table[k+1]-exp_table[k]); \
306 } while (0)
307 #else
308 #define NEG_EXP( result, narg ) \
309 do { \
310 result = exp(-narg); \
311 } while (0)
312 #endif
313
314
315 /**
316 * Initialize the exp_table[] lookup table for approximating exp().
317 */
318 void
319 r200InitStaticFogData( void )
320 {
321 GLfloat f = 0.0F;
322 GLint i = 0;
323 for ( ; i < FOG_EXP_TABLE_SIZE ; i++, f += FOG_INCR) {
324 exp_table[i] = (GLfloat) exp(-f);
325 }
326 }
327
328
329 /**
330 * Compute per-vertex fog blend factors from fog coordinates by
331 * evaluating the GL_LINEAR, GL_EXP or GL_EXP2 fog function.
332 * Fog coordinates are distances from the eye (typically between the
333 * near and far clip plane distances).
334 * Note the fog (eye Z) coords may be negative so we use ABS(z) below.
335 * Fog blend factors are in the range [0,1].
336 */
337 float
338 r200ComputeFogBlendFactor( GLcontext *ctx, GLfloat fogcoord )
339 {
340 GLfloat end = ctx->Fog.End;
341 GLfloat d, temp;
342 const GLfloat z = FABSF(fogcoord);
343
344 switch (ctx->Fog.Mode) {
345 case GL_LINEAR:
346 if (ctx->Fog.Start == ctx->Fog.End)
347 d = 1.0F;
348 else
349 d = 1.0F / (ctx->Fog.End - ctx->Fog.Start);
350 temp = (end - z) * d;
351 return CLAMP(temp, 0.0F, 1.0F);
352 break;
353 case GL_EXP:
354 d = ctx->Fog.Density;
355 NEG_EXP( temp, d * z );
356 return temp;
357 break;
358 case GL_EXP2:
359 d = ctx->Fog.Density*ctx->Fog.Density;
360 NEG_EXP( temp, d * z * z );
361 return temp;
362 break;
363 default:
364 _mesa_problem(ctx, "Bad fog mode in make_fog_coord");
365 return 0;
366 }
367 }
368
369 /**
370 * Predict total emit size for next rendering operation so there is no flush in middle of rendering
371 * Prediction has to aim towards the best possible value that is worse than worst case scenario
372 */
373 static GLuint r200EnsureEmitSize( GLcontext * ctx , GLubyte* vimap_rev )
374 {
375 r200ContextPtr rmesa = R200_CONTEXT(ctx);
376 TNLcontext *tnl = TNL_CONTEXT(ctx);
377 struct vertex_buffer *VB = &tnl->vb;
378 GLuint space_required;
379 GLuint state_size;
380 GLuint nr_aos = 0;
381 int i;
382 /* predict number of aos to emit */
383 for (i = 0; i < 15; ++i)
384 {
385 if (vimap_rev[i] != 255)
386 {
387 ++nr_aos;
388 }
389 }
390
391 {
392 /* count the prediction for state size */
393 space_required = 0;
394 state_size = radeonCountStateEmitSize( &rmesa->radeon );
395 /* vtx may be changed in r200EmitArrays so account for it if not dirty */
396 if (!rmesa->hw.vtx.dirty)
397 state_size += rmesa->hw.vtx.check(rmesa->radeon.glCtx, &rmesa->hw.vtx);
398 /* predict size for elements */
399 for (i = 0; i < VB->PrimitiveCount; ++i)
400 {
401 if (!VB->Primitive[i].count)
402 continue;
403 /* If primitive.count is less than MAX_CONVERSION_SIZE
404 rendering code may decide convert to elts.
405 In that case we have to make pessimistic prediction.
406 and use larger of 2 paths. */
407 const GLuint elts = ELTS_BUFSZ(nr_aos);
408 const GLuint index = INDEX_BUFSZ;
409 const GLuint vbuf = VBUF_BUFSZ;
410 if ( (!VB->Elts && VB->Primitive[i].count >= MAX_CONVERSION_SIZE)
411 || vbuf > index + elts)
412 space_required += vbuf;
413 else
414 space_required += index + elts;
415 space_required += AOS_BUFSZ(nr_aos);
416 }
417 }
418
419 radeon_print(RADEON_RENDER,RADEON_VERBOSE,
420 "%s space %u, aos %d\n",
421 __func__, space_required, AOS_BUFSZ(nr_aos) );
422 /* flush the buffer in case we need more than is left. */
423 if (rcommonEnsureCmdBufSpace(&rmesa->radeon, space_required + state_size, __FUNCTION__))
424 return space_required + radeonCountStateEmitSize( &rmesa->radeon );
425 else
426 return space_required + state_size;
427 }
428
429
430 /**********************************************************************/
431 /* Render pipeline stage */
432 /**********************************************************************/
433
434
435 /* TCL render.
436 */
437 static GLboolean r200_run_tcl_render( GLcontext *ctx,
438 struct tnl_pipeline_stage *stage )
439 {
440 r200ContextPtr rmesa = R200_CONTEXT(ctx);
441 TNLcontext *tnl = TNL_CONTEXT(ctx);
442 struct vertex_buffer *VB = &tnl->vb;
443 GLuint i;
444 GLubyte *vimap_rev;
445 /* use hw fixed order for simplicity, pos 0, weight 1, normal 2, fog 3,
446 color0 - color3 4-7, texcoord0 - texcoord5 8-13, pos 1 14. Must not use
447 more than 12 of those at the same time. */
448 GLubyte map_rev_fixed[15] = {255, 255, 255, 255, 255, 255, 255, 255,
449 255, 255, 255, 255, 255, 255, 255};
450
451
452 /* TODO: separate this from the swtnl pipeline
453 */
454 if (rmesa->radeon.TclFallback)
455 return GL_TRUE; /* fallback to software t&l */
456
457 radeon_print(RADEON_RENDER, RADEON_NORMAL, "%s\n", __FUNCTION__);
458
459 if (VB->Count == 0)
460 return GL_FALSE;
461
462 /* Validate state:
463 */
464 if (rmesa->radeon.NewGLState)
465 if (!r200ValidateState( ctx ))
466 return GL_TRUE; /* fallback to sw t&l */
467
468 if (!ctx->VertexProgram._Enabled) {
469 /* NOTE: inputs != tnl->render_inputs - these are the untransformed
470 * inputs.
471 */
472 map_rev_fixed[0] = VERT_ATTRIB_POS;
473 /* technically there is no reason we always need VA_COLOR0. In theory
474 could disable it depending on lighting, color materials, texturing... */
475 map_rev_fixed[4] = VERT_ATTRIB_COLOR0;
476
477 if (ctx->Light.Enabled) {
478 map_rev_fixed[2] = VERT_ATTRIB_NORMAL;
479 }
480
481 /* this also enables VA_COLOR1 when using separate specular
482 lighting model, which is unnecessary.
483 FIXME: OTOH, we're missing the case where a ATI_fragment_shader accesses
484 the secondary color (if lighting is disabled). The chip seems
485 misconfigured for that though elsewhere (tcl output, might lock up) */
486 if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
487 map_rev_fixed[5] = VERT_ATTRIB_COLOR1;
488 }
489
490 if ( (ctx->Fog.FogCoordinateSource == GL_FOG_COORD) && ctx->Fog.Enabled ) {
491 map_rev_fixed[3] = VERT_ATTRIB_FOG;
492 }
493
494 for (i = 0 ; i < ctx->Const.MaxTextureUnits; i++) {
495 if (ctx->Texture.Unit[i]._ReallyEnabled) {
496 if (rmesa->TexGenNeedNormals[i]) {
497 map_rev_fixed[2] = VERT_ATTRIB_NORMAL;
498 }
499 map_rev_fixed[8 + i] = VERT_ATTRIB_TEX0 + i;
500 }
501 }
502 vimap_rev = &map_rev_fixed[0];
503 }
504 else {
505 /* vtx_tcl_output_vtxfmt_0/1 need to match configuration of "fragment
506 part", since using some vertex interpolator later which is not in
507 out_vtxfmt0/1 will lock up. It seems to be ok to write in vertex
508 prog to a not enabled output however, so just don't mess with it.
509 We only need to change compsel. */
510 GLuint out_compsel = 0;
511 const GLbitfield64 vp_out =
512 rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten;
513
514 vimap_rev = &rmesa->curr_vp_hw->inputmap_rev[0];
515 assert(vp_out & BITFIELD64_BIT(VERT_RESULT_HPOS));
516 out_compsel = R200_OUTPUT_XYZW;
517 if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL0)) {
518 out_compsel |= R200_OUTPUT_COLOR_0;
519 }
520 if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL1)) {
521 out_compsel |= R200_OUTPUT_COLOR_1;
522 }
523 if (vp_out & BITFIELD64_BIT(VERT_RESULT_FOGC)) {
524 out_compsel |= R200_OUTPUT_DISCRETE_FOG;
525 }
526 if (vp_out & BITFIELD64_BIT(VERT_RESULT_PSIZ)) {
527 out_compsel |= R200_OUTPUT_PT_SIZE;
528 }
529 for (i = VERT_RESULT_TEX0; i < VERT_RESULT_TEX6; i++) {
530 if (vp_out & BITFIELD64_BIT(i)) {
531 out_compsel |= R200_OUTPUT_TEX_0 << (i - VERT_RESULT_TEX0);
532 }
533 }
534 if (rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] != out_compsel) {
535 R200_STATECHANGE( rmesa, vtx );
536 rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] = out_compsel;
537 }
538 }
539
540 /* Do the actual work:
541 */
542 radeonReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
543 GLuint emit_end = r200EnsureEmitSize( ctx, vimap_rev )
544 + rmesa->radeon.cmdbuf.cs->cdw;
545 r200EmitArrays( ctx, vimap_rev );
546
547 for (i = 0 ; i < VB->PrimitiveCount ; i++)
548 {
549 GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
550 GLuint start = VB->Primitive[i].start;
551 GLuint length = VB->Primitive[i].count;
552
553 if (!length)
554 continue;
555
556 if (VB->Elts)
557 r200EmitEltPrimitive( ctx, start, start+length, prim );
558 else
559 r200EmitPrimitive( ctx, start, start+length, prim );
560 }
561 if ( emit_end < rmesa->radeon.cmdbuf.cs->cdw )
562 WARN_ONCE("Rendering was %d commands larger than predicted size."
563 " We might overflow command buffer.\n", rmesa->radeon.cmdbuf.cs->cdw - emit_end);
564
565 return GL_FALSE; /* finished the pipe */
566 }
567
568
569
570 /* Initial state for tcl stage.
571 */
572 const struct tnl_pipeline_stage _r200_tcl_stage =
573 {
574 "r200 render",
575 NULL, /* private */
576 NULL,
577 NULL,
578 NULL,
579 r200_run_tcl_render /* run */
580 };
581
582
583
584 /**********************************************************************/
585 /* Validate state at pipeline start */
586 /**********************************************************************/
587
588
589 /*-----------------------------------------------------------------------
590 * Manage TCL fallbacks
591 */
592
593
594 static void transition_to_swtnl( GLcontext *ctx )
595 {
596 r200ContextPtr rmesa = R200_CONTEXT(ctx);
597 TNLcontext *tnl = TNL_CONTEXT(ctx);
598
599 R200_NEWPRIM( rmesa );
600
601 r200ChooseVertexState( ctx );
602 r200ChooseRenderState( ctx );
603
604 _mesa_validate_all_lighting_tables( ctx );
605
606 tnl->Driver.NotifyMaterialChange =
607 _mesa_validate_all_lighting_tables;
608
609 radeonReleaseArrays( ctx, ~0 );
610
611 /* Still using the D3D based hardware-rasterizer from the radeon;
612 * need to put the card into D3D mode to make it work:
613 */
614 R200_STATECHANGE( rmesa, vap );
615 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~(R200_VAP_TCL_ENABLE|R200_VAP_PROG_VTX_SHADER_ENABLE);
616 }
617
618 static void transition_to_hwtnl( GLcontext *ctx )
619 {
620 r200ContextPtr rmesa = R200_CONTEXT(ctx);
621 TNLcontext *tnl = TNL_CONTEXT(ctx);
622
623 _tnl_need_projected_coords( ctx, GL_FALSE );
624
625 r200UpdateMaterial( ctx );
626
627 tnl->Driver.NotifyMaterialChange = r200UpdateMaterial;
628
629 if ( rmesa->radeon.dma.flush )
630 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
631
632 rmesa->radeon.dma.flush = NULL;
633
634 R200_STATECHANGE( rmesa, vap );
635 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_TCL_ENABLE;
636 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_FORCE_W_TO_ONE;
637
638 if (ctx->VertexProgram._Enabled) {
639 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_PROG_VTX_SHADER_ENABLE;
640 }
641
642 if ( ((rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] & R200_FOG_USE_MASK)
643 == R200_FOG_USE_SPEC_ALPHA) &&
644 (ctx->Fog.FogCoordinateSource == GL_FOG_COORD )) {
645 R200_STATECHANGE( rmesa, ctx );
646 rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~R200_FOG_USE_MASK;
647 rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= R200_FOG_USE_VTX_FOG;
648 }
649
650 R200_STATECHANGE( rmesa, vte );
651 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] &= ~(R200_VTX_XY_FMT|R200_VTX_Z_FMT);
652 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] |= R200_VTX_W0_FMT;
653
654 if (R200_DEBUG & RADEON_FALLBACKS)
655 fprintf(stderr, "R200 end tcl fallback\n");
656 }
657
658
659 static char *fallbackStrings[] = {
660 "Rasterization fallback",
661 "Unfilled triangles",
662 "Twosided lighting, differing materials",
663 "Materials in VB (maybe between begin/end)",
664 "Texgen unit 0",
665 "Texgen unit 1",
666 "Texgen unit 2",
667 "Texgen unit 3",
668 "Texgen unit 4",
669 "Texgen unit 5",
670 "User disable",
671 "Bitmap as points",
672 "Vertex program"
673 };
674
675
676 static char *getFallbackString(GLuint bit)
677 {
678 int i = 0;
679 while (bit > 1) {
680 i++;
681 bit >>= 1;
682 }
683 return fallbackStrings[i];
684 }
685
686
687
688 void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
689 {
690 r200ContextPtr rmesa = R200_CONTEXT(ctx);
691 GLuint oldfallback = rmesa->radeon.TclFallback;
692
693 if (mode) {
694 rmesa->radeon.TclFallback |= bit;
695 if (oldfallback == 0) {
696 if (R200_DEBUG & RADEON_FALLBACKS)
697 fprintf(stderr, "R200 begin tcl fallback %s\n",
698 getFallbackString( bit ));
699 transition_to_swtnl( ctx );
700 }
701 }
702 else {
703 rmesa->radeon.TclFallback &= ~bit;
704 if (oldfallback == bit) {
705 if (R200_DEBUG & RADEON_FALLBACKS)
706 fprintf(stderr, "R200 end tcl fallback %s\n",
707 getFallbackString( bit ));
708 transition_to_hwtnl( ctx );
709 }
710 }
711 }