projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge branch 'lp-offset-twoside'
[mesa.git] / src / mesa / drivers / dri / r200 / r200_swtcl.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/mtypes.h"
37 #include "main/colormac.h"
38 #include "main/enums.h"
39 #include "main/image.h"
40 #include "main/imports.h"
41 #include "main/macros.h"
42 #include "main/simple_list.h"
43
44 #include "swrast/s_context.h"
45 #include "swrast/s_fog.h"
46 #include "swrast_setup/swrast_setup.h"
47 #include "tnl/tnl.h"
48 #include "tnl/t_context.h"
49 #include "tnl/t_pipeline.h"
50
51 #include "r200_context.h"
52 #include "r200_ioctl.h"
53 #include "r200_state.h"
54 #include "r200_swtcl.h"
55 #include "r200_tcl.h"
56
57
58 /***********************************************************************
59 * Initialization
60 ***********************************************************************/
61
62 #define EMIT_ATTR( ATTR, STYLE, F0 ) \
63 do { \
64 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR); \
65 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE); \
66 rmesa->radeon.swtcl.vertex_attr_count++; \
67 fmt_0 |= F0; \
68 } while (0)
69
70 #define EMIT_PAD( N ) \
71 do { \
72 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0; \
73 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD; \
74 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N); \
75 rmesa->radeon.swtcl.vertex_attr_count++; \
76 } while (0)
77
78 static void r200SetVertexFormat( struct gl_context *ctx )
79 {
80 r200ContextPtr rmesa = R200_CONTEXT( ctx );
81 TNLcontext *tnl = TNL_CONTEXT(ctx);
82 struct vertex_buffer *VB = &tnl->vb;
83 DECLARE_RENDERINPUTS(index_bitset);
84 int fmt_0 = 0;
85 int fmt_1 = 0;
86 int offset = 0;
87
88 RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );
89
90 /* Important:
91 */
92 if ( VB->NdcPtr != NULL ) {
93 VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
94 }
95 else {
96 VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
97 }
98
99 assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
100 rmesa->radeon.swtcl.vertex_attr_count = 0;
101
102 /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
103 * build up a hardware vertex.
104 */
105 if ( !rmesa->swtcl.needproj ||
106 RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { /* need w coord for projected textures */
107 EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F, R200_VTX_XY | R200_VTX_Z0 | R200_VTX_W0 );
108 offset = 4;
109 }
110 else {
111 EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F, R200_VTX_XY | R200_VTX_Z0 );
112 offset = 3;
113 }
114
115 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) {
116 EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F, R200_VTX_POINT_SIZE );
117 offset += 1;
118 }
119
120 rmesa->swtcl.coloroffset = offset;
121 #if MESA_LITTLE_ENDIAN
122 EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA, (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT) );
123 #else
124 EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR, (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT) );
125 #endif
126 offset += 1;
127
128 rmesa->swtcl.specoffset = 0;
129 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) ||
130 RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
131
132 #if MESA_LITTLE_ENDIAN
133 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
134 rmesa->swtcl.specoffset = offset;
135 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_RGB, (R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT) );
136 }
137 else {
138 EMIT_PAD( 3 );
139 }
140
141 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
142 EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, (R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT) );
143 }
144 else {
145 EMIT_PAD( 1 );
146 }
147 #else
148 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
149 EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, (R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT) );
150 }
151 else {
152 EMIT_PAD( 1 );
153 }
154
155 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
156 rmesa->swtcl.specoffset = offset;
157 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, (R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT) );
158 }
159 else {
160 EMIT_PAD( 3 );
161 }
162 #endif
163 }
164
165 if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
166 int i;
167
168 for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
169 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
170 GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size;
171
172 fmt_1 |= sz << (3 * i);
173 EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1, 0 );
174 }
175 }
176 }
177
178 if ( (rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] & R200_FOG_USE_MASK)
179 != R200_FOG_USE_SPEC_ALPHA ) {
180 R200_STATECHANGE( rmesa, ctx );
181 rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~R200_FOG_USE_MASK;
182 rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= R200_FOG_USE_SPEC_ALPHA;
183 }
184
185 if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) ||
186 (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0) ||
187 (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
188 R200_NEWPRIM(rmesa);
189 R200_STATECHANGE( rmesa, vtx );
190 rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = fmt_0;
191 rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = fmt_1;
192
193 rmesa->radeon.swtcl.vertex_size =
194 _tnl_install_attrs( ctx,
195 rmesa->radeon.swtcl.vertex_attrs,
196 rmesa->radeon.swtcl.vertex_attr_count,
197 NULL, 0 );
198 rmesa->radeon.swtcl.vertex_size /= 4;
199 RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset );
200 }
201 }
202
203 static void r200_predict_emit_size( r200ContextPtr rmesa )
204 {
205 if (RADEON_DEBUG & RADEON_VERTS)
206 fprintf(stderr, "%s\n", __func__);
207 const int vertex_array_size = 7;
208 const int prim_size = 3;
209 if (!rmesa->radeon.swtcl.emit_prediction) {
210 const int state_size = radeonCountStateEmitSize(&rmesa->radeon);
211 if (rcommonEnsureCmdBufSpace(&rmesa->radeon,
212 state_size +
213 vertex_array_size + prim_size,
214 __FUNCTION__))
215 rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon);
216 else
217 rmesa->radeon.swtcl.emit_prediction = state_size;
218 rmesa->radeon.swtcl.emit_prediction += vertex_array_size + prim_size
219 + rmesa->radeon.cmdbuf.cs->cdw;
220 }
221 }
222
223
224 static void r200RenderStart( struct gl_context *ctx )
225 {
226 r200SetVertexFormat( ctx );
227 if (RADEON_DEBUG & RADEON_VERTS)
228 fprintf(stderr, "%s\n", __func__);
229 }
230
231
232 /**
233 * Set vertex state for SW TCL. The primary purpose of this function is to
234 * determine in advance whether or not the hardware can / should do the
235 * projection divide or Mesa should do it.
236 */
237 void r200ChooseVertexState( struct gl_context *ctx )
238 {
239 r200ContextPtr rmesa = R200_CONTEXT( ctx );
240 TNLcontext *tnl = TNL_CONTEXT(ctx);
241 GLuint vte;
242 GLuint vap;
243
244 /* We must ensure that we don't do _tnl_need_projected_coords while in a
245 * rasterization fallback. As this function will be called again when we
246 * leave a rasterization fallback, we can just skip it for now.
247 */
248 if (rmesa->radeon.Fallback != 0)
249 return;
250
251 vte = rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL];
252 vap = rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL];
253
254 /* HW perspective divide is a win, but tiny vertex formats are a
255 * bigger one.
256 */
257 if (!RENDERINPUTS_TEST_RANGE( tnl->render_inputs_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )
258 || (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
259 rmesa->swtcl.needproj = GL_TRUE;
260 vte |= R200_VTX_XY_FMT | R200_VTX_Z_FMT;
261 vte &= ~R200_VTX_W0_FMT;
262 if (RENDERINPUTS_TEST_RANGE( tnl->render_inputs_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
263 vap &= ~R200_VAP_FORCE_W_TO_ONE;
264 }
265 else {
266 vap |= R200_VAP_FORCE_W_TO_ONE;
267 }
268 }
269 else {
270 rmesa->swtcl.needproj = GL_FALSE;
271 vte &= ~(R200_VTX_XY_FMT | R200_VTX_Z_FMT);
272 vte |= R200_VTX_W0_FMT;
273 vap &= ~R200_VAP_FORCE_W_TO_ONE;
274 }
275
276 _tnl_need_projected_coords( ctx, rmesa->swtcl.needproj );
277
278 if (vte != rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL]) {
279 R200_STATECHANGE( rmesa, vte );
280 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] = vte;
281 }
282
283 if (vap != rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL]) {
284 R200_STATECHANGE( rmesa, vap );
285 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] = vap;
286 }
287 }
288
289 void r200_swtcl_flush(struct gl_context *ctx, uint32_t current_offset)
290 {
291 r200ContextPtr rmesa = R200_CONTEXT(ctx);
292 if (RADEON_DEBUG & RADEON_VERTS)
293 fprintf(stderr, "%s\n", __func__);
294
295
296 radeonEmitState(&rmesa->radeon);
297 r200EmitVertexAOS( rmesa,
298 rmesa->radeon.swtcl.vertex_size,
299 rmesa->radeon.swtcl.bo,
300 current_offset);
301
302
303 r200EmitVbufPrim( rmesa,
304 rmesa->radeon.swtcl.hw_primitive,
305 rmesa->radeon.swtcl.numverts);
306 if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw )
307 WARN_ONCE("Rendering was %d commands larger than predicted size."
308 " We might overflow command buffer.\n",
309 rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction );
310
311 rmesa->radeon.swtcl.emit_prediction = 0;
312
313 }
314
315 /**************************************************************************/
316
317
318 static INLINE GLuint reduced_hw_prim( struct gl_context *ctx, GLuint prim)
319 {
320 switch (prim) {
321 case GL_POINTS:
322 return (((R200_CONTEXT(ctx))->radeon.radeonScreen->drmSupportsPointSprites &&
323 !(ctx->_TriangleCaps & DD_POINT_SMOOTH)) ?
324 R200_VF_PRIM_POINT_SPRITES : R200_VF_PRIM_POINTS);
325 case GL_LINES:
326 /* fallthrough */
327 case GL_LINE_LOOP:
328 /* fallthrough */
329 case GL_LINE_STRIP:
330 return R200_VF_PRIM_LINES;
331 default:
332 /* all others reduced to triangles */
333 return R200_VF_PRIM_TRIANGLES;
334 }
335 }
336
337
338 static void r200RasterPrimitive( struct gl_context *ctx, GLuint hwprim );
339 static void r200RenderPrimitive( struct gl_context *ctx, GLenum prim );
340 static void r200ResetLineStipple( struct gl_context *ctx );
341
342 /***********************************************************************
343 * Emit primitives as inline vertices *
344 ***********************************************************************/
345
346 #define HAVE_POINTS 1
347 #define HAVE_LINES 1
348 #define HAVE_LINE_STRIPS 1
349 #define HAVE_TRIANGLES 1
350 #define HAVE_TRI_STRIPS 1
351 #define HAVE_TRI_STRIP_1 0
352 #define HAVE_TRI_FANS 1
353 #define HAVE_QUADS 0
354 #define HAVE_QUAD_STRIPS 0
355 #define HAVE_POLYGONS 1
356 #define HAVE_ELTS 0
357
358 static void* r200_alloc_verts( r200ContextPtr rmesa, GLuint n, GLuint size)
359 {
360 void *rv;
361 do {
362 r200_predict_emit_size( rmesa );
363 rv = rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 );
364 } while(!rv);
365 return rv;
366 }
367
368 #undef LOCAL_VARS
369 #undef ALLOC_VERTS
370 #define CTX_ARG r200ContextPtr rmesa
371 #define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
372 #define ALLOC_VERTS( n, size ) r200_alloc_verts(rmesa, n, size)
373 #define LOCAL_VARS \
374 r200ContextPtr rmesa = R200_CONTEXT(ctx); \
375 const char *r200verts = (char *)rmesa->radeon.swtcl.verts;
376 #define VERT(x) (radeonVertex *)(r200verts + ((x) * vertsize * sizeof(int)))
377 #define VERTEX radeonVertex
378 #define DO_DEBUG_VERTS (1 && (R200_DEBUG & RADEON_VERTS))
379
380 #undef TAG
381 #define TAG(x) r200_##x
382 #include "tnl_dd/t_dd_triemit.h"
383
384
385 /***********************************************************************
386 * Macros for t_dd_tritmp.h to draw basic primitives *
387 ***********************************************************************/
388
389 #define QUAD( a, b, c, d ) r200_quad( rmesa, a, b, c, d )
390 #define TRI( a, b, c ) r200_triangle( rmesa, a, b, c )
391 #define LINE( a, b ) r200_line( rmesa, a, b )
392 #define POINT( a ) r200_point( rmesa, a )
393
394 /***********************************************************************
395 * Build render functions from dd templates *
396 ***********************************************************************/
397
398 #define R200_TWOSIDE_BIT 0x01
399 #define R200_UNFILLED_BIT 0x02
400 #define R200_MAX_TRIFUNC 0x04
401
402
403 static struct {
404 tnl_points_func points;
405 tnl_line_func line;
406 tnl_triangle_func triangle;
407 tnl_quad_func quad;
408 } rast_tab[R200_MAX_TRIFUNC];
409
410
411 #define DO_FALLBACK 0
412 #define DO_UNFILLED (IND & R200_UNFILLED_BIT)
413 #define DO_TWOSIDE (IND & R200_TWOSIDE_BIT)
414 #define DO_FLAT 0
415 #define DO_OFFSET 0
416 #define DO_TRI 1
417 #define DO_QUAD 1
418 #define DO_LINE 1
419 #define DO_POINTS 1
420 #define DO_FULL_QUAD 1
421
422 #define HAVE_SPEC 1
423 #define HAVE_BACK_COLORS 0
424 #define HAVE_HW_FLATSHADE 1
425 #define TAB rast_tab
426
427 #define DEPTH_SCALE 1.0
428 #define UNFILLED_TRI unfilled_tri
429 #define UNFILLED_QUAD unfilled_quad
430 #define VERT_X(_v) _v->v.x
431 #define VERT_Y(_v) _v->v.y
432 #define VERT_Z(_v) _v->v.z
433 #define AREA_IS_CCW( a ) (a < 0)
434 #define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
435
436 #define VERT_SET_RGBA( v, c ) \
437 do { \
438 radeon_color_t *color = (radeon_color_t *)&((v)->ui[coloroffset]); \
439 UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]); \
440 UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]); \
441 UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]); \
442 UNCLAMPED_FLOAT_TO_UBYTE(color->alpha, (c)[3]); \
443 } while (0)
444
445 #define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
446
447 #define VERT_SET_SPEC( v, c ) \
448 do { \
449 if (specoffset) { \
450 radeon_color_t *spec = (radeon_color_t *)&((v)->ui[specoffset]); \
451 UNCLAMPED_FLOAT_TO_UBYTE(spec->red, (c)[0]); \
452 UNCLAMPED_FLOAT_TO_UBYTE(spec->green, (c)[1]); \
453 UNCLAMPED_FLOAT_TO_UBYTE(spec->blue, (c)[2]); \
454 } \
455 } while (0)
456 #define VERT_COPY_SPEC( v0, v1 ) \
457 do { \
458 if (specoffset) { \
459 radeon_color_t *spec0 = (radeon_color_t *)&((v0)->ui[specoffset]); \
460 radeon_color_t *spec1 = (radeon_color_t *)&((v1)->ui[specoffset]); \
461 spec0->red = spec1->red; \
462 spec0->green = spec1->green; \
463 spec0->blue = spec1->blue; \
464 } \
465 } while (0)
466
467 /* These don't need LE32_TO_CPU() as they used to save and restore
468 * colors which are already in the correct format.
469 */
470 #define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[coloroffset]
471 #define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
472 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
473 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
474
475 #undef LOCAL_VARS
476 #undef TAG
477 #undef INIT
478
479 #define LOCAL_VARS(n) \
480 r200ContextPtr rmesa = R200_CONTEXT(ctx); \
481 GLuint color[n] = {0}, spec[n] = {0}; \
482 GLuint coloroffset = rmesa->swtcl.coloroffset; \
483 GLuint specoffset = rmesa->swtcl.specoffset; \
484 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
485
486 /***********************************************************************
487 * Helpers for rendering unfilled primitives *
488 ***********************************************************************/
489
490 #define RASTERIZE(x) r200RasterPrimitive( ctx, reduced_hw_prim(ctx, x) )
491 #define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
492 #undef TAG
493 #define TAG(x) x
494 #include "tnl_dd/t_dd_unfilled.h"
495 #undef IND
496
497
498 /***********************************************************************
499 * Generate GL render functions *
500 ***********************************************************************/
501
502
503 #define IND (0)
504 #define TAG(x) x
505 #include "tnl_dd/t_dd_tritmp.h"
506
507 #define IND (R200_TWOSIDE_BIT)
508 #define TAG(x) x##_twoside
509 #include "tnl_dd/t_dd_tritmp.h"
510
511 #define IND (R200_UNFILLED_BIT)
512 #define TAG(x) x##_unfilled
513 #include "tnl_dd/t_dd_tritmp.h"
514
515 #define IND (R200_TWOSIDE_BIT|R200_UNFILLED_BIT)
516 #define TAG(x) x##_twoside_unfilled
517 #include "tnl_dd/t_dd_tritmp.h"
518
519
520 static void init_rast_tab( void )
521 {
522 init();
523 init_twoside();
524 init_unfilled();
525 init_twoside_unfilled();
526 }
527
528 /**********************************************************************/
529 /* Render unclipped begin/end objects */
530 /**********************************************************************/
531
532 #define RENDER_POINTS( start, count ) \
533 for ( ; start < count ; start++) \
534 r200_point( rmesa, VERT(start) )
535 #define RENDER_LINE( v0, v1 ) \
536 r200_line( rmesa, VERT(v0), VERT(v1) )
537 #define RENDER_TRI( v0, v1, v2 ) \
538 r200_triangle( rmesa, VERT(v0), VERT(v1), VERT(v2) )
539 #define RENDER_QUAD( v0, v1, v2, v3 ) \
540 r200_quad( rmesa, VERT(v0), VERT(v1), VERT(v2), VERT(v3) )
541 #define INIT(x) do { \
542 r200RenderPrimitive( ctx, x ); \
543 } while (0)
544 #undef LOCAL_VARS
545 #define LOCAL_VARS \
546 r200ContextPtr rmesa = R200_CONTEXT(ctx); \
547 const GLuint vertsize = rmesa->radeon.swtcl.vertex_size; \
548 const char *r200verts = (char *)rmesa->radeon.swtcl.verts; \
549 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
550 const GLboolean stipple = ctx->Line.StippleFlag; \
551 (void) elt; (void) stipple;
552 #define RESET_STIPPLE if ( stipple ) r200ResetLineStipple( ctx );
553 #define RESET_OCCLUSION
554 #define PRESERVE_VB_DEFS
555 #define ELT(x) (x)
556 #define TAG(x) r200_##x##_verts
557 #include "tnl/t_vb_rendertmp.h"
558 #undef ELT
559 #undef TAG
560 #define TAG(x) r200_##x##_elts
561 #define ELT(x) elt[x]
562 #include "tnl/t_vb_rendertmp.h"
563
564
565
566 /**********************************************************************/
567 /* Choose render functions */
568 /**********************************************************************/
569
570 void r200ChooseRenderState( struct gl_context *ctx )
571 {
572 TNLcontext *tnl = TNL_CONTEXT(ctx);
573 r200ContextPtr rmesa = R200_CONTEXT(ctx);
574 GLuint index = 0;
575 GLuint flags = ctx->_TriangleCaps;
576
577 if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback)
578 return;
579
580 if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R200_TWOSIDE_BIT;
581 if (flags & DD_TRI_UNFILLED) index |= R200_UNFILLED_BIT;
582
583 if (index != rmesa->radeon.swtcl.RenderIndex) {
584 tnl->Driver.Render.Points = rast_tab[index].points;
585 tnl->Driver.Render.Line = rast_tab[index].line;
586 tnl->Driver.Render.ClippedLine = rast_tab[index].line;
587 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
588 tnl->Driver.Render.Quad = rast_tab[index].quad;
589
590 if (index == 0) {
591 tnl->Driver.Render.PrimTabVerts = r200_render_tab_verts;
592 tnl->Driver.Render.PrimTabElts = r200_render_tab_elts;
593 tnl->Driver.Render.ClippedPolygon = r200_fast_clipped_poly;
594 } else {
595 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
596 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
597 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
598 }
599
600 rmesa->radeon.swtcl.RenderIndex = index;
601 }
602 }
603
604
605 /**********************************************************************/
606 /* High level hooks for t_vb_render.c */
607 /**********************************************************************/
608
609
610 static void r200RasterPrimitive( struct gl_context *ctx, GLuint hwprim )
611 {
612 r200ContextPtr rmesa = R200_CONTEXT(ctx);
613
614 radeon_prepare_render(&rmesa->radeon);
615 if (rmesa->radeon.NewGLState)
616 r200ValidateState( ctx );
617
618
619 if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
620 /* need to disable perspective-correct texturing for point sprites */
621 if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) {
622 if (rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE) {
623 R200_STATECHANGE( rmesa, set );
624 rmesa->hw.set.cmd[SET_RE_CNTL] &= ~R200_PERSPECTIVE_ENABLE;
625 }
626 }
627 else if (!(rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE)) {
628 R200_STATECHANGE( rmesa, set );
629 rmesa->hw.set.cmd[SET_RE_CNTL] |= R200_PERSPECTIVE_ENABLE;
630 }
631 R200_NEWPRIM( rmesa );
632 rmesa->radeon.swtcl.hw_primitive = hwprim;
633 }
634 }
635
636 static void r200RenderPrimitive( struct gl_context *ctx, GLenum prim )
637 {
638 r200ContextPtr rmesa = R200_CONTEXT(ctx);
639 rmesa->radeon.swtcl.render_primitive = prim;
640 if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED))
641 r200RasterPrimitive( ctx, reduced_hw_prim(ctx, prim) );
642 }
643
644 static void r200RenderFinish( struct gl_context *ctx )
645 {
646 }
647
648 static void r200ResetLineStipple( struct gl_context *ctx )
649 {
650 r200ContextPtr rmesa = R200_CONTEXT(ctx);
651 R200_STATECHANGE( rmesa, lin );
652 }
653
654
655 /**********************************************************************/
656 /* Transition to/from hardware rasterization. */
657 /**********************************************************************/
658
659 static const char * const fallbackStrings[] = {
660 "Texture mode",
661 "glDrawBuffer(GL_FRONT_AND_BACK)",
662 "glEnable(GL_STENCIL) without hw stencil buffer",
663 "glRenderMode(selection or feedback)",
664 "R200_NO_RAST",
665 "Mixing GL_CLAMP_TO_BORDER and GL_CLAMP (or GL_MIRROR_CLAMP_ATI)"
666 };
667
668
669 static const char *getFallbackString(GLuint bit)
670 {
671 int i = 0;
672 while (bit > 1) {
673 i++;
674 bit >>= 1;
675 }
676 return fallbackStrings[i];
677 }
678
679
680 void r200Fallback( struct gl_context *ctx, GLuint bit, GLboolean mode )
681 {
682 r200ContextPtr rmesa = R200_CONTEXT(ctx);
683 TNLcontext *tnl = TNL_CONTEXT(ctx);
684 GLuint oldfallback = rmesa->radeon.Fallback;
685
686 if (mode) {
687 rmesa->radeon.Fallback |= bit;
688 if (oldfallback == 0) {
689 radeon_firevertices(&rmesa->radeon);
690 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_TRUE );
691 _swsetup_Wakeup( ctx );
692 rmesa->radeon.swtcl.RenderIndex = ~0;
693 if (R200_DEBUG & RADEON_FALLBACKS) {
694 fprintf(stderr, "R200 begin rasterization fallback: 0x%x %s\n",
695 bit, getFallbackString(bit));
696 }
697 }
698 }
699 else {
700 rmesa->radeon.Fallback &= ~bit;
701 if (oldfallback == bit) {
702
703 _swrast_flush( ctx );
704 tnl->Driver.Render.Start = r200RenderStart;
705 tnl->Driver.Render.PrimitiveNotify = r200RenderPrimitive;
706 tnl->Driver.Render.Finish = r200RenderFinish;
707
708 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
709 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
710 tnl->Driver.Render.Interp = _tnl_interp;
711
712 tnl->Driver.Render.ResetLineStipple = r200ResetLineStipple;
713 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_FALSE );
714 if (rmesa->radeon.TclFallback) {
715 /* These are already done if rmesa->radeon.TclFallback goes to
716 * zero above. But not if it doesn't (R200_NO_TCL for
717 * example?)
718 */
719 _tnl_invalidate_vertex_state( ctx, ~0 );
720 _tnl_invalidate_vertices( ctx, ~0 );
721 RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset );
722 r200ChooseVertexState( ctx );
723 r200ChooseRenderState( ctx );
724 }
725 if (R200_DEBUG & RADEON_FALLBACKS) {
726 fprintf(stderr, "R200 end rasterization fallback: 0x%x %s\n",
727 bit, getFallbackString(bit));
728 }
729 }
730 }
731 }
732
733
734
735
736 /**
737 * Cope with depth operations by drawing individual pixels as points.
738 *
739 * \todo
740 * The way the vertex state is set in this routine is hokey. It seems to
741 * work, but it's very hackish. This whole routine is pretty hackish. If
742 * the bitmap is small enough, it seems like it would be faster to copy it
743 * to AGP memory and use it as a non-power-of-two texture (i.e.,
744 * NV_texture_rectangle).
745 */
746 void
747 r200PointsBitmap( struct gl_context *ctx, GLint px, GLint py,
748 GLsizei width, GLsizei height,
749 const struct gl_pixelstore_attrib *unpack,
750 const GLubyte *bitmap )
751 {
752 r200ContextPtr rmesa = R200_CONTEXT(ctx);
753 const GLfloat *rc = ctx->Current.RasterColor;
754 GLint row, col;
755 radeonVertex vert;
756 GLuint orig_vte;
757 GLuint h;
758
759
760 /* Turn off tcl.
761 */
762 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_BITMAP, 1 );
763
764 /* Choose tiny vertex format
765 */
766 {
767 const GLuint fmt_0 = R200_VTX_XY | R200_VTX_Z0 | R200_VTX_W0
768 | (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT);
769 const GLuint fmt_1 = 0;
770 GLuint vte = rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL];
771 GLuint vap = rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL];
772
773 vte &= ~(R200_VTX_XY_FMT | R200_VTX_Z_FMT);
774 vte |= R200_VTX_W0_FMT;
775 vap &= ~R200_VAP_FORCE_W_TO_ONE;
776
777 rmesa->radeon.swtcl.vertex_size = 5;
778
779 if ( (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0)
780 || (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
781 R200_NEWPRIM(rmesa);
782 R200_STATECHANGE( rmesa, vtx );
783 rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = fmt_0;
784 rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = fmt_1;
785 }
786
787 if (vte != rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL]) {
788 R200_STATECHANGE( rmesa, vte );
789 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] = vte;
790 }
791
792 if (vap != rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL]) {
793 R200_STATECHANGE( rmesa, vap );
794 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] = vap;
795 }
796 }
797
798 /* Ready for point primitives:
799 */
800 r200RenderPrimitive( ctx, GL_POINTS );
801
802 /* Turn off the hw viewport transformation:
803 */
804 R200_STATECHANGE( rmesa, vte );
805 orig_vte = rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL];
806 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] &= ~(R200_VPORT_X_SCALE_ENA |
807 R200_VPORT_Y_SCALE_ENA |
808 R200_VPORT_Z_SCALE_ENA |
809 R200_VPORT_X_OFFSET_ENA |
810 R200_VPORT_Y_OFFSET_ENA |
811 R200_VPORT_Z_OFFSET_ENA);
812
813 /* Turn off other stuff: Stipple?, texture?, blending?, etc.
814 */
815
816
817 /* Populate the vertex
818 *
819 * Incorporate FOG into RGBA
820 */
821 if (ctx->Fog.Enabled) {
822 const GLfloat *fc = ctx->Fog.Color;
823 GLfloat color[4];
824 GLfloat f;
825
826 if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT)
827 f = _swrast_z_to_fogfactor(ctx, ctx->Current.Attrib[VERT_ATTRIB_FOG][0]);
828 else
829 f = _swrast_z_to_fogfactor(ctx, ctx->Current.RasterDistance);
830
831 color[0] = f * rc[0] + (1.F - f) * fc[0];
832 color[1] = f * rc[1] + (1.F - f) * fc[1];
833 color[2] = f * rc[2] + (1.F - f) * fc[2];
834 color[3] = rc[3];
835
836 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.red, color[0]);
837 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.green, color[1]);
838 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.blue, color[2]);
839 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.alpha, color[3]);
840 }
841 else {
842 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.red, rc[0]);
843 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.green, rc[1]);
844 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.blue, rc[2]);
845 UNCLAMPED_FLOAT_TO_CHAN(vert.tv.color.alpha, rc[3]);
846 }
847
848
849 vert.tv.z = ctx->Current.RasterPos[2];
850
851
852 /* Update window height
853 */
854 LOCK_HARDWARE( &rmesa->radeon );
855 UNLOCK_HARDWARE( &rmesa->radeon );
856 h = radeon_get_drawable(&rmesa->radeon)->h + radeon_get_drawable(&rmesa->radeon)->y;
857 px += radeon_get_drawable(&rmesa->radeon)->x;
858
859 /* Clipping handled by existing mechansims in r200_ioctl.c?
860 */
861 for (row=0; row<height; row++) {
862 const GLubyte *src = (const GLubyte *)
863 _mesa_image_address2d(unpack, bitmap, width, height,
864 GL_COLOR_INDEX, GL_BITMAP, row, 0 );
865
866 if (unpack->LsbFirst) {
867 /* Lsb first */
868 GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
869 for (col=0; col<width; col++) {
870 if (*src & mask) {
871 vert.tv.x = px+col;
872 vert.tv.y = h - (py+row) - 1;
873 r200_point( rmesa, &vert );
874 }
875 src += (mask >> 7);
876 mask = ((mask << 1) & 0xff) | (mask >> 7);
877 }
878
879 /* get ready for next row */
880 if (mask != 1)
881 src++;
882 }
883 else {
884 /* Msb first */
885 GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
886 for (col=0; col<width; col++) {
887 if (*src & mask) {
888 vert.tv.x = px+col;
889 vert.tv.y = h - (py+row) - 1;
890 r200_point( rmesa, &vert );
891 }
892 src += mask & 1;
893 mask = ((mask << 7) & 0xff) | (mask >> 1);
894 }
895 /* get ready for next row */
896 if (mask != 128)
897 src++;
898 }
899 }
900
901 /* Fire outstanding vertices, restore state
902 */
903 R200_STATECHANGE( rmesa, vte );
904 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] = orig_vte;
905
906 /* Unfallback
907 */
908 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_BITMAP, 0 );
909
910 /* Need to restore vertexformat?
911 */
912 if (rmesa->radeon.TclFallback)
913 r200ChooseVertexState( ctx );
914 }
915
916
917
918 /**********************************************************************/
919 /* Initialization. */
920 /**********************************************************************/
921
922 void r200InitSwtcl( struct gl_context *ctx )
923 {
924 TNLcontext *tnl = TNL_CONTEXT(ctx);
925 r200ContextPtr rmesa = R200_CONTEXT(ctx);
926 static int firsttime = 1;
927
928 if (firsttime) {
929 init_rast_tab();
930 firsttime = 0;
931 }
932 rmesa->radeon.swtcl.emit_prediction = 0;
933
934 tnl->Driver.Render.Start = r200RenderStart;
935 tnl->Driver.Render.Finish = r200RenderFinish;
936 tnl->Driver.Render.PrimitiveNotify = r200RenderPrimitive;
937 tnl->Driver.Render.ResetLineStipple = r200ResetLineStipple;
938 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
939 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
940 tnl->Driver.Render.Interp = _tnl_interp;
941
942 /* FIXME: what are these numbers? */
943 _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12,
944 36 * sizeof(GLfloat) );
945
946 rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
947 rmesa->radeon.swtcl.RenderIndex = ~0;
948 rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
949 rmesa->radeon.swtcl.hw_primitive = 0;
950 }
951