78d16ed0f3b5ecd4114b49ab792a618149c6b033
[mesa.git] / src / mesa / drivers / dri / r200 / r200_tcl.c
1 /* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_tcl.c,v 1.2 2002/12/16 16:18:55 dawes Exp $ */
2 /*
3 Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved.
4
5 The Weather Channel (TM) funded Tungsten Graphics to develop the
6 initial release of the Radeon 8500 driver under the XFree86 license.
7 This notice must be preserved.
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 "imports.h"
38 #include "mtypes.h"
39 #include "enums.h"
40 #include "colormac.h"
41 #include "light.h"
42
43 #include "array_cache/acache.h"
44 #include "tnl/tnl.h"
45 #include "tnl/t_pipeline.h"
46
47 #include "r200_context.h"
48 #include "r200_state.h"
49 #include "r200_ioctl.h"
50 #include "r200_tex.h"
51 #include "r200_tcl.h"
52 #include "r200_swtcl.h"
53 #include "r200_maos.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 0 /* hw quad verts in wrong order??? */
66 #define HAVE_QUAD_STRIPS 1
67 #define HAVE_POLYGONS 1
68 #define HAVE_ELTS 1
69
70
71 #define HW_POINTS R200_VF_PRIM_POINTS
72 #define HW_LINES R200_VF_PRIM_LINES
73 #define HW_LINE_LOOP 0
74 #define HW_LINE_STRIP R200_VF_PRIM_LINE_STRIP
75 #define HW_TRIANGLES R200_VF_PRIM_TRIANGLES
76 #define HW_TRIANGLE_STRIP_0 R200_VF_PRIM_TRIANGLE_STRIP
77 #define HW_TRIANGLE_STRIP_1 0
78 #define HW_TRIANGLE_FAN R200_VF_PRIM_TRIANGLE_FAN
79 #define HW_QUADS R200_VF_PRIM_QUADS
80 #define HW_QUAD_STRIP R200_VF_PRIM_QUAD_STRIP
81 #define HW_POLYGON R200_VF_PRIM_POLYGON
82
83
84 static GLboolean discrete_prim[0x10] = {
85 0, /* 0 none */
86 1, /* 1 points */
87 1, /* 2 lines */
88 0, /* 3 line_strip */
89 1, /* 4 tri_list */
90 0, /* 5 tri_fan */
91 0, /* 6 tri_strip */
92 0, /* 7 tri_w_flags */
93 1, /* 8 rect list (unused) */
94 1, /* 9 3vert point */
95 1, /* a 3vert line */
96 0, /* b point sprite */
97 0, /* c line loop */
98 1, /* d quads */
99 0, /* e quad strip */
100 0, /* f polygon */
101 };
102
103
104 #define LOCAL_VARS r200ContextPtr rmesa = R200_CONTEXT(ctx); (void)rmesa
105 #define ELT_TYPE GLushort
106
107 #define ELT_INIT(prim, hw_prim) \
108 r200TclPrimitive( ctx, prim, hw_prim | R200_VF_PRIM_WALK_IND )
109
110 #define GET_MESA_ELTS() rmesa->tcl.Elts
111
112
113 /* Don't really know how many elts will fit in what's left of cmdbuf,
114 * as there is state to emit, etc:
115 */
116
117 /* Testing on isosurf shows a maximum around here. Don't know if it's
118 * the card or driver or kernel module that is causing the behaviour.
119 */
120 #define GET_MAX_HW_ELTS() 300
121
122 #define RESET_STIPPLE() do { \
123 R200_STATECHANGE( rmesa, lin ); \
124 r200EmitState( rmesa ); \
125 } while (0)
126
127 #define AUTO_STIPPLE( mode ) do { \
128 R200_STATECHANGE( rmesa, lin ); \
129 if (mode) \
130 rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] |= \
131 R200_LINE_PATTERN_AUTO_RESET; \
132 else \
133 rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &= \
134 ~R200_LINE_PATTERN_AUTO_RESET; \
135 r200EmitState( rmesa ); \
136 } while (0)
137
138
139 #define ALLOC_ELTS(nr) r200AllocElts( rmesa, nr )
140
141 static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr )
142 {
143 if (rmesa->dma.flush == r200FlushElts &&
144 rmesa->store.cmd_used + nr*2 < R200_CMD_BUF_SZ) {
145
146 GLushort *dest = (GLushort *)(rmesa->store.cmd_buf +
147 rmesa->store.cmd_used);
148
149 rmesa->store.cmd_used += nr*2;
150
151 return dest;
152 }
153 else {
154 if (rmesa->dma.flush)
155 rmesa->dma.flush( rmesa );
156
157 r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
158 rmesa->hw.max_state_size + ELTS_BUFSZ(nr) );
159
160 r200EmitAOS( rmesa,
161 rmesa->tcl.aos_components,
162 rmesa->tcl.nr_aos_components, 0 );
163
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 EMIT_PRIM( 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 r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
190 rmesa->hw.max_state_size + VBUF_BUFSZ );
191
192 r200EmitAOS( rmesa,
193 rmesa->tcl.aos_components,
194 rmesa->tcl.nr_aos_components,
195 start );
196
197 /* Why couldn't this packet have taken an offset param?
198 */
199 r200EmitVbufPrim( rmesa,
200 rmesa->tcl.hw_primitive,
201 count - start );
202 }
203
204
205
206 /* Try & join small primitives
207 */
208 #if 0
209 #define PREFER_DISCRETE_ELT_PRIM( NR, PRIM ) 0
210 #else
211 #define PREFER_DISCRETE_ELT_PRIM( NR, PRIM ) \
212 ((NR) < 20 || \
213 ((NR) < 40 && \
214 rmesa->tcl.hw_primitive == (PRIM| \
215 R200_VF_TCL_OUTPUT_VTX_ENABLE| \
216 R200_VF_PRIM_WALK_IND)))
217 #endif
218
219 #ifdef MESA_BIG_ENDIAN
220 /* We could do without (most of) this ugliness if dest was always 32 bit word aligned... */
221 #define EMIT_ELT(dest, offset, x) do { \
222 int off = offset + ( ( (GLuint)dest & 0x2 ) >> 1 ); \
223 GLushort *des = (GLushort *)( (GLuint)dest & ~0x2 ); \
224 (des)[ off + 1 - 2 * ( off & 1 ) ] = (GLushort)(x); } while (0)
225 #else
226 #define EMIT_ELT(dest, offset, x) (dest)[offset] = (GLushort) (x)
227 #endif
228
229 #define EMIT_TWO_ELTS(dest, offset, x, y) *(GLuint *)((dest)+offset) = ((y)<<16)|(x);
230
231
232
233 #define TAG(x) tcl_##x
234 #include "tnl_dd/t_dd_dmatmp2.h"
235
236 /**********************************************************************/
237 /* External entrypoints */
238 /**********************************************************************/
239
240 void r200EmitPrimitive( GLcontext *ctx,
241 GLuint first,
242 GLuint last,
243 GLuint flags )
244 {
245 tcl_render_tab_verts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
246 }
247
248 void r200EmitEltPrimitive( GLcontext *ctx,
249 GLuint first,
250 GLuint last,
251 GLuint flags )
252 {
253 tcl_render_tab_elts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
254 }
255
256 void r200TclPrimitive( GLcontext *ctx,
257 GLenum prim,
258 int hw_prim )
259 {
260 r200ContextPtr rmesa = R200_CONTEXT(ctx);
261 GLuint newprim = hw_prim | R200_VF_TCL_OUTPUT_VTX_ENABLE;
262
263 if (newprim != rmesa->tcl.hw_primitive ||
264 !discrete_prim[hw_prim&0xf]) {
265 R200_NEWPRIM( rmesa );
266 rmesa->tcl.hw_primitive = newprim;
267 }
268 }
269
270
271 /**********************************************************************/
272 /* Render pipeline stage */
273 /**********************************************************************/
274
275
276 /* TCL render.
277 */
278 static GLboolean r200_run_tcl_render( GLcontext *ctx,
279 struct tnl_pipeline_stage *stage )
280 {
281 r200ContextPtr rmesa = R200_CONTEXT(ctx);
282 TNLcontext *tnl = TNL_CONTEXT(ctx);
283 struct vertex_buffer *VB = &tnl->vb;
284 GLuint i;
285
286 /* TODO: separate this from the swtnl pipeline
287 */
288 if (rmesa->TclFallback)
289 return GL_TRUE; /* fallback to software t&l */
290
291 if (R200_DEBUG & DEBUG_PRIMS)
292 fprintf(stderr, "%s\n", __FUNCTION__);
293
294 if (VB->Count == 0)
295 return GL_FALSE;
296
297 r200ReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
298 r200EmitArrays( ctx, stage->inputs );
299
300 rmesa->tcl.Elts = VB->Elts;
301
302 for (i = 0 ; i < VB->PrimitiveCount ; i++)
303 {
304 GLuint prim = VB->Primitive[i].mode;
305 GLuint start = VB->Primitive[i].start;
306 GLuint length = VB->Primitive[i].count;
307
308 if (!length)
309 continue;
310
311 if (rmesa->tcl.Elts)
312 r200EmitEltPrimitive( ctx, start, start+length, prim );
313 else
314 r200EmitPrimitive( ctx, start, start+length, prim );
315 }
316
317 return GL_FALSE; /* finished the pipe */
318 }
319
320
321
322 static void r200_check_tcl_render( GLcontext *ctx,
323 struct tnl_pipeline_stage *stage )
324 {
325 r200ContextPtr rmesa = R200_CONTEXT(ctx);
326 GLuint inputs = VERT_BIT_POS;
327 GLuint unit;
328
329 /* Validate state:
330 */
331 if (rmesa->NewGLState)
332 r200ValidateState( ctx );
333
334 if (ctx->RenderMode == GL_RENDER) {
335 /* Make all this event-driven:
336 */
337 if (ctx->Light.Enabled) {
338 inputs |= VERT_BIT_NORMAL;
339
340 if (1 || ctx->Light.ColorMaterialEnabled) {
341 inputs |= VERT_BIT_COLOR0;
342 }
343 }
344 else {
345 inputs |= VERT_BIT_COLOR0;
346
347 if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
348 inputs |= VERT_BIT_COLOR1;
349 }
350 }
351
352 for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
353 if (ctx->Texture.Unit[unit]._ReallyEnabled) {
354 if (rmesa->TexGenNeedNormals[unit]) {
355 inputs |= VERT_BIT_NORMAL;
356 }
357 inputs |= VERT_BIT_TEX(unit);
358 }
359 }
360
361 stage->inputs = inputs;
362 stage->active = 1;
363 }
364 else
365 stage->active = 0;
366 }
367
368 static void r200_init_tcl_render( GLcontext *ctx,
369 struct tnl_pipeline_stage *stage )
370 {
371 stage->check = r200_check_tcl_render;
372 stage->check( ctx, stage );
373 }
374
375 static void dtr( struct tnl_pipeline_stage *stage )
376 {
377 (void)stage;
378 }
379
380
381 /* Initial state for tcl stage.
382 */
383 const struct tnl_pipeline_stage _r200_tcl_stage =
384 {
385 "r200 render",
386 (_DD_NEW_SEPARATE_SPECULAR |
387 _NEW_LIGHT|
388 _NEW_TEXTURE|
389 _NEW_FOG|
390 _NEW_RENDERMODE), /* re-check (new inputs) */
391 0, /* re-run (always runs) */
392 GL_TRUE, /* active */
393 0, 0, /* inputs (set in check_render), outputs */
394 0, 0, /* changed_inputs, private */
395 dtr, /* destructor */
396 r200_init_tcl_render, /* check - initially set to alloc data */
397 r200_run_tcl_render /* run */
398 };
399
400
401
402 /**********************************************************************/
403 /* Validate state at pipeline start */
404 /**********************************************************************/
405
406
407 /*-----------------------------------------------------------------------
408 * Manage TCL fallbacks
409 */
410
411
412 static void transition_to_swtnl( GLcontext *ctx )
413 {
414 r200ContextPtr rmesa = R200_CONTEXT(ctx);
415 TNLcontext *tnl = TNL_CONTEXT(ctx);
416
417 R200_NEWPRIM( rmesa );
418
419 r200ChooseVertexState( ctx );
420 r200ChooseRenderState( ctx );
421
422 _mesa_validate_all_lighting_tables( ctx );
423
424 tnl->Driver.NotifyMaterialChange =
425 _mesa_validate_all_lighting_tables;
426
427 r200ReleaseArrays( ctx, ~0 );
428
429 /* Still using the D3D based hardware-rasterizer from the radeon;
430 * need to put the card into D3D mode to make it work:
431 */
432 R200_STATECHANGE( rmesa, vap );
433 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_TCL_ENABLE;
434 }
435
436 static void transition_to_hwtnl( GLcontext *ctx )
437 {
438 r200ContextPtr rmesa = R200_CONTEXT(ctx);
439 TNLcontext *tnl = TNL_CONTEXT(ctx);
440
441 _tnl_need_projected_coords( ctx, GL_FALSE );
442
443 r200UpdateMaterial( ctx );
444
445 tnl->Driver.NotifyMaterialChange = r200UpdateMaterial;
446
447 if ( rmesa->dma.flush )
448 rmesa->dma.flush( rmesa );
449
450 rmesa->dma.flush = 0;
451
452 if (rmesa->swtcl.indexed_verts.buf)
453 r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts,
454 __FUNCTION__ );
455
456 R200_STATECHANGE( rmesa, vap );
457 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_TCL_ENABLE;
458 rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_FORCE_W_TO_ONE;
459
460 R200_STATECHANGE( rmesa, vte );
461 rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] &= ~(R200_VTX_XY_FMT|R200_VTX_Z_FMT);
462
463 if (R200_DEBUG & DEBUG_FALLBACKS)
464 fprintf(stderr, "R200 end tcl fallback\n");
465 }
466
467
468 static char *fallbackStrings[] = {
469 "Rasterization fallback",
470 "Unfilled triangles",
471 "Twosided lighting, differing materials",
472 "Materials in VB (maybe between begin/end)",
473 "Texgen unit 0",
474 "Texgen unit 1",
475 "Texgen unit 2",
476 "Texgen unit 3",
477 "Texgen unit 4",
478 "Texgen unit 5",
479 "User disable",
480 "Bitmap as points"
481 };
482
483
484 static char *getFallbackString(GLuint bit)
485 {
486 int i = 0;
487 while (bit > 1) {
488 i++;
489 bit >>= 1;
490 }
491 return fallbackStrings[i];
492 }
493
494
495
496 void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
497 {
498 r200ContextPtr rmesa = R200_CONTEXT(ctx);
499 GLuint oldfallback = rmesa->TclFallback;
500
501 if (mode) {
502 rmesa->TclFallback |= bit;
503 if (oldfallback == 0) {
504 if (R200_DEBUG & DEBUG_FALLBACKS)
505 fprintf(stderr, "R200 begin tcl fallback %s\n",
506 getFallbackString( bit ));
507 transition_to_swtnl( ctx );
508 }
509 }
510 else {
511 rmesa->TclFallback &= ~bit;
512 if (oldfallback == bit) {
513 if (R200_DEBUG & DEBUG_FALLBACKS)
514 fprintf(stderr, "R200 end tcl fallback %s\n",
515 getFallbackString( bit ));
516 transition_to_hwtnl( ctx );
517 }
518 }
519 }