r200/r300: swtcl fixups to use old dma buffers on top of BOs
[mesa.git] / src / mesa / drivers / dri / r300 / r300_context.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 * \file
32 *
33 * \author Keith Whitwell <keith@tungstengraphics.com>
34 *
35 * \author Nicolai Haehnle <prefect_@gmx.net>
36 */
37
38 #include "main/glheader.h"
39 #include "main/api_arrayelt.h"
40 #include "main/context.h"
41 #include "main/simple_list.h"
42 #include "main/imports.h"
43 #include "main/matrix.h"
44 #include "main/extensions.h"
45 #include "main/state.h"
46 #include "main/bufferobj.h"
47
48 #include "swrast/swrast.h"
49 #include "swrast_setup/swrast_setup.h"
50 #include "vbo/vbo.h"
51
52 #include "tnl/tnl.h"
53 #include "tnl/t_pipeline.h"
54 #include "tnl/t_vp_build.h"
55
56 #include "drivers/common/driverfuncs.h"
57
58 #include "radeon_ioctl.h"
59 #include "radeon_span.h"
60 #include "r300_context.h"
61 #include "r300_cmdbuf.h"
62 #include "r300_state.h"
63 #include "r300_ioctl.h"
64 #include "r300_tex.h"
65 #include "r300_emit.h"
66 #include "r300_swtcl.h"
67 #include "radeon_bo_legacy.h"
68
69
70 #include "vblank.h"
71 #include "utils.h"
72 #include "xmlpool.h" /* for symbolic values of enum-type options */
73
74 /* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
75 int future_hw_tcl_on = 1;
76 int hw_tcl_on = 1;
77
78 #define need_GL_EXT_stencil_two_side
79 #define need_GL_ARB_multisample
80 #define need_GL_ARB_point_parameters
81 #define need_GL_ARB_texture_compression
82 #define need_GL_ARB_vertex_buffer_object
83 #define need_GL_ARB_vertex_program
84 #define need_GL_EXT_blend_minmax
85 //#define need_GL_EXT_fog_coord
86 #define need_GL_EXT_multi_draw_arrays
87 #define need_GL_EXT_secondary_color
88 #define need_GL_EXT_blend_equation_separate
89 #define need_GL_EXT_blend_func_separate
90 #define need_GL_EXT_gpu_program_parameters
91 #define need_GL_NV_vertex_program
92 #include "extension_helper.h"
93
94 const struct dri_extension card_extensions[] = {
95 /* *INDENT-OFF* */
96 {"GL_ARB_depth_texture", NULL},
97 {"GL_ARB_fragment_program", NULL},
98 {"GL_ARB_multisample", GL_ARB_multisample_functions},
99 {"GL_ARB_multitexture", NULL},
100 {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
101 {"GL_ARB_shadow", NULL},
102 {"GL_ARB_shadow_ambient", NULL},
103 {"GL_ARB_texture_border_clamp", NULL},
104 {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
105 {"GL_ARB_texture_cube_map", NULL},
106 {"GL_ARB_texture_env_add", NULL},
107 {"GL_ARB_texture_env_combine", NULL},
108 {"GL_ARB_texture_env_crossbar", NULL},
109 {"GL_ARB_texture_env_dot3", NULL},
110 {"GL_ARB_texture_mirrored_repeat", NULL},
111 {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
112 {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
113 {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
114 {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
115 {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
116 {"GL_EXT_blend_subtract", NULL},
117 // {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
118 {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions},
119 {"GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions},
120 {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
121 {"GL_EXT_shadow_funcs", NULL},
122 {"GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions},
123 {"GL_EXT_stencil_wrap", NULL},
124 {"GL_EXT_texture_edge_clamp", NULL},
125 {"GL_EXT_texture_env_combine", NULL},
126 {"GL_EXT_texture_env_dot3", NULL},
127 {"GL_EXT_texture_filter_anisotropic", NULL},
128 {"GL_EXT_texture_lod_bias", NULL},
129 {"GL_EXT_texture_mirror_clamp", NULL},
130 {"GL_EXT_texture_rectangle", NULL},
131 {"GL_ATI_texture_env_combine3", NULL},
132 {"GL_ATI_texture_mirror_once", NULL},
133 {"GL_MESA_pack_invert", NULL},
134 {"GL_MESA_ycbcr_texture", NULL},
135 {"GL_MESAX_texture_float", NULL},
136 {"GL_NV_blend_square", NULL},
137 {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
138 {"GL_SGIS_generate_mipmap", NULL},
139 {NULL, NULL}
140 /* *INDENT-ON* */
141 };
142
143 extern struct tnl_pipeline_stage _r300_render_stage;
144 extern const struct tnl_pipeline_stage _r300_tcl_stage;
145
146 static const struct tnl_pipeline_stage *r300_pipeline[] = {
147
148 /* Try and go straight to t&l
149 */
150 &_r300_tcl_stage,
151
152 /* Catch any t&l fallbacks
153 */
154 &_tnl_vertex_transform_stage,
155 &_tnl_normal_transform_stage,
156 &_tnl_lighting_stage,
157 &_tnl_fog_coordinate_stage,
158 &_tnl_texgen_stage,
159 &_tnl_texture_transform_stage,
160 &_tnl_vertex_program_stage,
161
162 /* Try again to go to tcl?
163 * - no good for asymmetric-twoside (do with multipass)
164 * - no good for asymmetric-unfilled (do with multipass)
165 * - good for material
166 * - good for texgen
167 * - need to manipulate a bit of state
168 *
169 * - worth it/not worth it?
170 */
171
172 /* Else do them here.
173 */
174 &_r300_render_stage,
175 &_tnl_render_stage, /* FALLBACK */
176 0,
177 };
178
179 static void r300RunPipeline(GLcontext * ctx)
180 {
181 _mesa_lock_context_textures(ctx);
182
183 if (ctx->NewState)
184 _mesa_update_state_locked(ctx);
185
186 _tnl_run_pipeline(ctx);
187 _mesa_unlock_context_textures(ctx);
188 }
189
190 static void r300_get_lock(radeonContextPtr rmesa)
191 {
192 drm_radeon_sarea_t *sarea = rmesa->sarea;
193
194 if (sarea->ctx_owner != rmesa->dri.hwContext) {
195 sarea->ctx_owner = rmesa->dri.hwContext;
196 if (!rmesa->radeonScreen->kernel_mm)
197 radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
198 }
199 }
200
201 static void r300_vtbl_flush(GLcontext *ctx)
202 {
203 r300Flush(ctx);
204 }
205
206 static void r300_vtbl_set_all_dirty(GLcontext *ctx)
207 {
208 r300ContextPtr rmesa = R300_CONTEXT(ctx);
209 rmesa->hw.all_dirty = GL_TRUE;
210 }
211
212 static void r300_vtbl_emit_state(radeonContextPtr rmesa)
213 {
214 r300EmitState((r300ContextPtr)rmesa);
215 }
216
217 extern int cs_write_dword(struct radeon_cs *cs, uint32_t dword);
218
219 static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
220 {
221 /* please flush pipe do all pending work */
222 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
223 R300_SC_SCREENDOOR, 1));
224 cs_write_dword(cs, 0x0);
225 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
226 R300_SC_SCREENDOOR, 1));
227 cs_write_dword(cs, 0x00FFFFFF);
228 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
229 R300_SC_HYPERZ, 1));
230 cs_write_dword(cs, 0x0);
231 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
232 R300_US_CONFIG, 1));
233 cs_write_dword(cs, 0x0);
234 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
235 R300_ZB_CNTL, 1));
236 cs_write_dword(cs, 0x0);
237 cs_write_dword(cs, cmdwait(rmesa->radeonScreen, R300_WAIT_3D));
238 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
239 R300_RB3D_DSTCACHE_CTLSTAT, 1));
240 cs_write_dword(cs, R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
241 cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
242 R300_ZB_ZCACHE_CTLSTAT, 1));
243 cs_write_dword(cs, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE);
244 cs_write_dword(cs, cmdwait(rmesa->radeonScreen,
245 R300_WAIT_3D | R300_WAIT_3D_CLEAN));
246 }
247
248 static void r300_vtbl_flush_vertices(radeonContextPtr rmesa)
249 {
250 R300_FIREVERTICES(((r300ContextPtr)rmesa));
251 }
252
253 static void r300_init_vtbl(radeonContextPtr radeon)
254 {
255 radeon->vtbl.get_lock = r300_get_lock;
256 radeon->vtbl.update_viewport_offset = r300UpdateViewportOffset;
257 radeon->vtbl.flush = r300_vtbl_flush;
258 radeon->vtbl.set_all_dirty = r300_vtbl_set_all_dirty;
259 radeon->vtbl.update_draw_buffer = r300UpdateDrawBuffer;
260 radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header;
261 radeon->vtbl.emit_state = r300_vtbl_emit_state;
262 radeon->vtbl.flush_vertices = r300_vtbl_flush_vertices;
263 radeon->vtbl.swtcl_flush = r300_swtcl_flush;
264 }
265
266
267 /* Create the device specific rendering context.
268 */
269 GLboolean r300CreateContext(const __GLcontextModes * glVisual,
270 __DRIcontextPrivate * driContextPriv,
271 void *sharedContextPrivate)
272 {
273 __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
274 radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
275 struct dd_function_table functions;
276 r300ContextPtr r300;
277 GLcontext *ctx;
278 int tcl_mode;
279
280 assert(glVisual);
281 assert(driContextPriv);
282 assert(screen);
283
284 /* Allocate the R300 context */
285 r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
286 if (!r300)
287 return GL_FALSE;
288
289 if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
290 hw_tcl_on = future_hw_tcl_on = 0;
291
292 r300_init_vtbl(&r300->radeon);
293 /* Parse configuration files.
294 * Do this here so that initialMaxAnisotropy is set before we create
295 * the default textures.
296 */
297 driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
298 screen->driScreen->myNum, "r300");
299 r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
300 "def_max_anisotropy");
301
302 /* Init default driver functions then plug in our R300-specific functions
303 * (the texture functions are especially important)
304 */
305 _mesa_init_driver_functions(&functions);
306 r300InitIoctlFuncs(&functions);
307 r300InitStateFuncs(&functions);
308 r300InitTextureFuncs(&functions);
309 r300InitShaderFuncs(&functions);
310
311 if (!radeonInitContext(&r300->radeon, &functions,
312 glVisual, driContextPriv,
313 sharedContextPrivate)) {
314 FREE(r300);
315 return GL_FALSE;
316 }
317
318 /* Init r300 context data */
319 r300->radeon.texture_depth = driQueryOptioni(&r300->radeon.optionCache,
320 "texture_depth");
321 if (r300->radeon.texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
322 r300->radeon.texture_depth = (screen->cpp == 4) ?
323 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
324
325 /* Set the maximum texture size small enough that we can guarentee that
326 * all texture units can bind a maximal texture and have them both in
327 * texturable memory at once.
328 */
329
330 ctx = r300->radeon.glCtx;
331
332 ctx->Const.MaxTextureImageUnits =
333 driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
334 ctx->Const.MaxTextureCoordUnits =
335 driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
336 ctx->Const.MaxTextureUnits =
337 MIN2(ctx->Const.MaxTextureImageUnits,
338 ctx->Const.MaxTextureCoordUnits);
339 ctx->Const.MaxTextureMaxAnisotropy = 16.0;
340 ctx->Const.MaxTextureLodBias = 16.0;
341
342 if (screen->chip_family >= CHIP_FAMILY_RV515) {
343 ctx->Const.MaxTextureLevels = 13;
344 ctx->Const.MaxTextureRectSize = 4096;
345 }
346
347 ctx->Const.MinPointSize = 1.0;
348 ctx->Const.MinPointSizeAA = 1.0;
349 ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
350 ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
351
352 ctx->Const.MinLineWidth = 1.0;
353 ctx->Const.MinLineWidthAA = 1.0;
354 ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
355 ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
356
357 /* Needs further modifications */
358 #if 0
359 ctx->Const.MaxArrayLockSize =
360 ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
361 #endif
362
363 /* Initialize the software rasterizer and helper modules.
364 */
365 _swrast_CreateContext(ctx);
366 _vbo_CreateContext(ctx);
367 _tnl_CreateContext(ctx);
368 _swsetup_CreateContext(ctx);
369 _swsetup_Wakeup(ctx);
370 _ae_create_context(ctx);
371
372 /* Install the customized pipeline:
373 */
374 _tnl_destroy_pipeline(ctx);
375 _tnl_install_pipeline(ctx, r300_pipeline);
376
377 /* Try and keep materials and vertices separate:
378 */
379 /* _tnl_isolate_materials(ctx, GL_TRUE); */
380
381 /* Configure swrast and TNL to match hardware characteristics:
382 */
383 _swrast_allow_pixel_fog(ctx, GL_FALSE);
384 _swrast_allow_vertex_fog(ctx, GL_TRUE);
385 _tnl_allow_pixel_fog(ctx, GL_FALSE);
386 _tnl_allow_vertex_fog(ctx, GL_TRUE);
387
388 /* currently bogus data */
389 if (screen->chip_flags & RADEON_CHIPSET_TCL) {
390 ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
391 ctx->Const.VertexProgram.MaxNativeInstructions =
392 VSF_MAX_FRAGMENT_LENGTH / 4;
393 ctx->Const.VertexProgram.MaxNativeAttribs = 16; /* r420 */
394 ctx->Const.VertexProgram.MaxTemps = 32;
395 ctx->Const.VertexProgram.MaxNativeTemps =
396 /*VSF_MAX_FRAGMENT_TEMPS */ 32;
397 ctx->Const.VertexProgram.MaxNativeParameters = 256; /* r420 */
398 ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
399 }
400
401 ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
402 ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* copy i915... */
403 ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
404 ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
405 ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
406 ctx->Const.FragmentProgram.MaxNativeInstructions =
407 PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
408 ctx->Const.FragmentProgram.MaxNativeTexIndirections =
409 PFS_MAX_TEX_INDIRECT;
410 ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* and these are?? */
411 ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
412 ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
413
414 driInitExtensions(ctx, card_extensions, GL_TRUE);
415
416 if (driQueryOptionb
417 (&r300->radeon.optionCache, "disable_stencil_two_side"))
418 _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
419
420 if (r300->radeon.glCtx->Mesa_DXTn
421 && !driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc")) {
422 _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
423 _mesa_enable_extension(ctx, "GL_S3_s3tc");
424 } else
425 if (driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable"))
426 {
427 _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
428 }
429
430 r300->disable_lowimpact_fallback =
431 driQueryOptionb(&r300->radeon.optionCache,
432 "disable_lowimpact_fallback");
433
434 radeonInitSpanFuncs(ctx);
435 r300InitCmdBuf(r300);
436 r300InitState(r300);
437 if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
438 r300InitSwtcl(ctx);
439
440 TNL_CONTEXT(ctx)->Driver.RunPipeline = r300RunPipeline;
441
442 tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
443 if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
444 fprintf(stderr, "disabling 3D acceleration\n");
445 #if R200_MERGED
446 FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
447 #endif
448 }
449 if (tcl_mode == DRI_CONF_TCL_SW ||
450 !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
451 if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
452 r300->radeon.radeonScreen->chip_flags &=
453 ~RADEON_CHIPSET_TCL;
454 fprintf(stderr, "Disabling HW TCL support\n");
455 }
456 TCL_FALLBACK(r300->radeon.glCtx,
457 RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
458 }
459
460 return GL_TRUE;
461 }
462
463 /* Destroy the device specific context.
464 */
465 void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
466 {
467 GET_CURRENT_CONTEXT(ctx);
468 r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate;
469 radeonContextPtr radeon = (radeonContextPtr) r300;
470 radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
471
472 if (RADEON_DEBUG & DEBUG_DRI) {
473 fprintf(stderr, "Destroying context !\n");
474 }
475
476 /* check if we're deleting the currently bound context */
477 if (&r300->radeon == current) {
478 radeonFlush(r300->radeon.glCtx);
479 _mesa_make_current(NULL, NULL, NULL);
480 }
481
482 /* Free r300 context resources */
483 assert(r300); /* should never be null */
484
485 if (r300) {
486 _swsetup_DestroyContext(r300->radeon.glCtx);
487 _tnl_DestroyContext(r300->radeon.glCtx);
488 _vbo_DestroyContext(r300->radeon.glCtx);
489 _swrast_DestroyContext(r300->radeon.glCtx);
490
491 rcommonFlushCmdBuf(&r300->radeon, __FUNCTION__);
492 r300DestroyCmdBuf(r300);
493
494 if (radeon->state.scissor.pClipRects) {
495 FREE(radeon->state.scissor.pClipRects);
496 radeon->state.scissor.pClipRects = NULL;
497 }
498
499 radeonCleanupContext(&r300->radeon);
500
501 /* the memory manager might be accessed when Mesa frees the shared
502 * state, so don't destroy it earlier
503 */
504
505
506 FREE(r300);
507 }
508 }