Merge remote branch 'origin/master' into radeon-rewrite
[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 #include "main/texobj.h"
48
49 #include "swrast/swrast.h"
50 #include "swrast_setup/swrast_setup.h"
51 #include "vbo/vbo.h"
52
53 #include "tnl/tnl.h"
54 #include "tnl/t_pipeline.h"
55 #include "tnl/t_vp_build.h"
56
57 #include "drivers/common/driverfuncs.h"
58
59 #include "r300_context.h"
60 #include "radeon_context.h"
61 #include "radeon_span.h"
62 #include "r300_cmdbuf.h"
63 #include "r300_state.h"
64 #include "r300_ioctl.h"
65 #include "r300_tex.h"
66 #include "r300_emit.h"
67 #include "r300_swtcl.h"
68 #include "radeon_bocs_wrapper.h"
69
70
71 #include "vblank.h"
72 #include "utils.h"
73 #include "xmlpool.h" /* for symbolic values of enum-type options */
74
75 /* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
76 int future_hw_tcl_on = 1;
77 int hw_tcl_on = 1;
78
79 #define need_GL_VERSION_2_0
80 #define need_GL_ARB_point_parameters
81 #define need_GL_ARB_vertex_program
82 #define need_GL_EXT_blend_equation_separate
83 #define need_GL_EXT_blend_func_separate
84 #define need_GL_EXT_blend_minmax
85 //#define need_GL_EXT_fog_coord
86 #define need_GL_EXT_gpu_program_parameters
87 #define need_GL_EXT_secondary_color
88 #define need_GL_EXT_stencil_two_side
89 #define need_GL_ATI_separate_stencil
90 #define need_GL_NV_vertex_program
91 #include "extension_helper.h"
92
93 const struct dri_extension card_extensions[] = {
94 /* *INDENT-OFF* */
95 {"GL_ARB_depth_texture", NULL},
96 {"GL_ARB_fragment_program", NULL},
97 {"GL_ARB_multitexture", NULL},
98 {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
99 {"GL_ARB_shadow", NULL},
100 {"GL_ARB_shadow_ambient", NULL},
101 {"GL_ARB_texture_border_clamp", NULL},
102 {"GL_ARB_texture_cube_map", NULL},
103 {"GL_ARB_texture_env_add", NULL},
104 {"GL_ARB_texture_env_combine", NULL},
105 {"GL_ARB_texture_env_crossbar", NULL},
106 {"GL_ARB_texture_env_dot3", NULL},
107 {"GL_ARB_texture_mirrored_repeat", NULL},
108 {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
109 {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
110 {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
111 {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
112 {"GL_EXT_blend_subtract", NULL},
113 // {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
114 {"GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions},
115 {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
116 {"GL_EXT_shadow_funcs", NULL},
117 {"GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions},
118 {"GL_EXT_stencil_wrap", NULL},
119 {"GL_EXT_texture_edge_clamp", NULL},
120 {"GL_EXT_texture_env_combine", NULL},
121 {"GL_EXT_texture_env_dot3", NULL},
122 {"GL_EXT_texture_filter_anisotropic", NULL},
123 {"GL_EXT_texture_lod_bias", NULL},
124 {"GL_EXT_texture_mirror_clamp", NULL},
125 {"GL_EXT_texture_rectangle", NULL},
126 {"GL_ATI_separate_stencil", GL_ATI_separate_stencil_functions},
127 {"GL_ATI_texture_env_combine3", NULL},
128 {"GL_ATI_texture_mirror_once", NULL},
129 {"GL_MESA_pack_invert", NULL},
130 {"GL_MESA_ycbcr_texture", NULL},
131 {"GL_MESAX_texture_float", NULL},
132 {"GL_NV_blend_square", NULL},
133 {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
134 {"GL_SGIS_generate_mipmap", NULL},
135 {NULL, NULL}
136 /* *INDENT-ON* */
137 };
138
139
140 /**
141 * The GL 2.0 functions are needed to make display lists work with
142 * functions added by GL_ATI_separate_stencil.
143 */
144 const struct dri_extension gl_20_extension[] = {
145 {"GL_VERSION_2_0", GL_VERSION_2_0_functions },
146 };
147
148
149 extern struct tnl_pipeline_stage _r300_render_stage;
150 extern const struct tnl_pipeline_stage _r300_tcl_stage;
151
152 static const struct tnl_pipeline_stage *r300_pipeline[] = {
153
154 /* Try and go straight to t&l
155 */
156 &_r300_tcl_stage,
157
158 /* Catch any t&l fallbacks
159 */
160 &_tnl_vertex_transform_stage,
161 &_tnl_normal_transform_stage,
162 &_tnl_lighting_stage,
163 &_tnl_fog_coordinate_stage,
164 &_tnl_texgen_stage,
165 &_tnl_texture_transform_stage,
166 &_tnl_vertex_program_stage,
167
168 /* Try again to go to tcl?
169 * - no good for asymmetric-twoside (do with multipass)
170 * - no good for asymmetric-unfilled (do with multipass)
171 * - good for material
172 * - good for texgen
173 * - need to manipulate a bit of state
174 *
175 * - worth it/not worth it?
176 */
177
178 /* Else do them here.
179 */
180 &_r300_render_stage,
181 &_tnl_render_stage, /* FALLBACK */
182 0,
183 };
184
185 static void r300RunPipeline(GLcontext * ctx)
186 {
187 _mesa_lock_context_textures(ctx);
188
189 if (ctx->NewState)
190 _mesa_update_state_locked(ctx);
191
192 _tnl_run_pipeline(ctx);
193 _mesa_unlock_context_textures(ctx);
194 }
195
196 static void r300_get_lock(radeonContextPtr rmesa)
197 {
198 drm_radeon_sarea_t *sarea = rmesa->sarea;
199
200 if (sarea->ctx_owner != rmesa->dri.hwContext) {
201 sarea->ctx_owner = rmesa->dri.hwContext;
202 if (!rmesa->radeonScreen->kernel_mm)
203 radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
204 }
205 }
206
207 static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
208 {
209 /* please flush pipe do all pending work */
210 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
211 R300_SC_SCREENDOOR, 1));
212 radeon_cs_write_dword(cs, 0x0);
213 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
214 R300_SC_SCREENDOOR, 1));
215 radeon_cs_write_dword(cs, 0x00FFFFFF);
216 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
217 R300_SC_HYPERZ, 1));
218 radeon_cs_write_dword(cs, 0x0);
219 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
220 R300_US_CONFIG, 1));
221 radeon_cs_write_dword(cs, 0x0);
222 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
223 R300_ZB_CNTL, 1));
224 radeon_cs_write_dword(cs, 0x0);
225 radeon_cs_write_dword(cs, cmdwait(rmesa->radeonScreen, R300_WAIT_3D));
226 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
227 R300_RB3D_DSTCACHE_CTLSTAT, 1));
228 radeon_cs_write_dword(cs, R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
229 radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
230 R300_ZB_ZCACHE_CTLSTAT, 1));
231 radeon_cs_write_dword(cs, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE);
232 radeon_cs_write_dword(cs, cmdwait(rmesa->radeonScreen,
233 R300_WAIT_3D | R300_WAIT_3D_CLEAN));
234 }
235
236 static void r300_vtbl_pre_emit_atoms(radeonContextPtr radeon)
237 {
238 BATCH_LOCALS(radeon);
239 cp_wait(radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
240 BEGIN_BATCH_NO_AUTOSTATE(2);
241 OUT_BATCH(cmdpacket0(radeon->radeonScreen, R300_TX_INVALTAGS, 1));
242 OUT_BATCH(R300_TX_FLUSH);
243 END_BATCH();
244 end_3d(radeon);
245 }
246
247 static void r300_init_vtbl(radeonContextPtr radeon)
248 {
249 radeon->vtbl.get_lock = r300_get_lock;
250 radeon->vtbl.update_viewport_offset = r300UpdateViewportOffset;
251 radeon->vtbl.update_draw_buffer = r300UpdateDrawBuffer;
252 radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header;
253 radeon->vtbl.swtcl_flush = r300_swtcl_flush;
254 radeon->vtbl.pre_emit_atoms = r300_vtbl_pre_emit_atoms;
255 }
256
257
258 /* Create the device specific rendering context.
259 */
260 GLboolean r300CreateContext(const __GLcontextModes * glVisual,
261 __DRIcontextPrivate * driContextPriv,
262 void *sharedContextPrivate)
263 {
264 __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
265 radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
266 struct dd_function_table functions;
267 r300ContextPtr r300;
268 GLcontext *ctx;
269 int tcl_mode;
270
271 assert(glVisual);
272 assert(driContextPriv);
273 assert(screen);
274
275 /* Allocate the R300 context */
276 r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
277 if (!r300)
278 return GL_FALSE;
279
280 if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
281 hw_tcl_on = future_hw_tcl_on = 0;
282
283 r300_init_vtbl(&r300->radeon);
284 /* Parse configuration files.
285 * Do this here so that initialMaxAnisotropy is set before we create
286 * the default textures.
287 */
288 driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
289 screen->driScreen->myNum, "r300");
290 r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
291 "def_max_anisotropy");
292
293 /* Init default driver functions then plug in our R300-specific functions
294 * (the texture functions are especially important)
295 */
296 _mesa_init_driver_functions(&functions);
297 r300InitIoctlFuncs(&functions);
298 r300InitStateFuncs(&functions);
299 r300InitTextureFuncs(&functions);
300 r300InitShaderFuncs(&functions);
301
302 if (!radeonInitContext(&r300->radeon, &functions,
303 glVisual, driContextPriv,
304 sharedContextPrivate)) {
305 FREE(r300);
306 return GL_FALSE;
307 }
308
309 /* Init r300 context data */
310 r300->radeon.texture_depth = driQueryOptioni(&r300->radeon.optionCache,
311 "texture_depth");
312 if (r300->radeon.texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
313 r300->radeon.texture_depth = (screen->cpp == 4) ?
314 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
315
316 /* Set the maximum texture size small enough that we can guarentee that
317 * all texture units can bind a maximal texture and have them both in
318 * texturable memory at once.
319 */
320
321 ctx = r300->radeon.glCtx;
322
323 ctx->Const.MaxTextureImageUnits =
324 driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
325 ctx->Const.MaxTextureCoordUnits =
326 driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
327 ctx->Const.MaxTextureUnits =
328 MIN2(ctx->Const.MaxTextureImageUnits,
329 ctx->Const.MaxTextureCoordUnits);
330 ctx->Const.MaxTextureMaxAnisotropy = 16.0;
331 ctx->Const.MaxTextureLodBias = 16.0;
332
333 if (screen->chip_family >= CHIP_FAMILY_RV515) {
334 ctx->Const.MaxTextureLevels = 13;
335 ctx->Const.MaxTextureRectSize = 4096;
336 }
337
338 ctx->Const.MinPointSize = 1.0;
339 ctx->Const.MinPointSizeAA = 1.0;
340 ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
341 ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
342
343 ctx->Const.MinLineWidth = 1.0;
344 ctx->Const.MinLineWidthAA = 1.0;
345 ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
346 ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
347
348 /* Needs further modifications */
349 #if 0
350 ctx->Const.MaxArrayLockSize =
351 ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
352 #endif
353
354 ctx->Const.MaxDrawBuffers = 1;
355
356 /* Initialize the software rasterizer and helper modules.
357 */
358 _swrast_CreateContext(ctx);
359 _vbo_CreateContext(ctx);
360 _tnl_CreateContext(ctx);
361 _swsetup_CreateContext(ctx);
362 _swsetup_Wakeup(ctx);
363 _ae_create_context(ctx);
364
365 /* Install the customized pipeline:
366 */
367 _tnl_destroy_pipeline(ctx);
368 _tnl_install_pipeline(ctx, r300_pipeline);
369
370 /* Try and keep materials and vertices separate:
371 */
372 /* _tnl_isolate_materials(ctx, GL_TRUE); */
373
374 /* Configure swrast and TNL to match hardware characteristics:
375 */
376 _swrast_allow_pixel_fog(ctx, GL_FALSE);
377 _swrast_allow_vertex_fog(ctx, GL_TRUE);
378 _tnl_allow_pixel_fog(ctx, GL_FALSE);
379 _tnl_allow_vertex_fog(ctx, GL_TRUE);
380
381 /* currently bogus data */
382 if (screen->chip_flags & RADEON_CHIPSET_TCL) {
383 ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
384 ctx->Const.VertexProgram.MaxNativeInstructions =
385 VSF_MAX_FRAGMENT_LENGTH / 4;
386 ctx->Const.VertexProgram.MaxNativeAttribs = 16; /* r420 */
387 ctx->Const.VertexProgram.MaxTemps = 32;
388 ctx->Const.VertexProgram.MaxNativeTemps =
389 /*VSF_MAX_FRAGMENT_TEMPS */ 32;
390 ctx->Const.VertexProgram.MaxNativeParameters = 256; /* r420 */
391 ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
392 }
393
394 ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
395 ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* copy i915... */
396 ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
397 ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
398 ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
399 ctx->Const.FragmentProgram.MaxNativeInstructions =
400 PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
401 ctx->Const.FragmentProgram.MaxNativeTexIndirections =
402 PFS_MAX_TEX_INDIRECT;
403 ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* and these are?? */
404 ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
405 ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
406
407 driInitExtensions(ctx, card_extensions, GL_TRUE);
408
409 if (driQueryOptionb
410 (&r300->radeon.optionCache, "disable_stencil_two_side"))
411 _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
412
413 if (r300->radeon.glCtx->Mesa_DXTn
414 && !driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc")) {
415 _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
416 _mesa_enable_extension(ctx, "GL_S3_s3tc");
417 } else
418 if (driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable"))
419 {
420 _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
421 }
422
423 r300->disable_lowimpact_fallback =
424 driQueryOptionb(&r300->radeon.optionCache,
425 "disable_lowimpact_fallback");
426
427 radeonInitSpanFuncs( ctx );
428 r300InitCmdBuf(r300);
429 r300InitState(r300);
430 if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
431 r300InitSwtcl(ctx);
432
433 TNL_CONTEXT(ctx)->Driver.RunPipeline = r300RunPipeline;
434
435 tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
436 if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
437 fprintf(stderr, "disabling 3D acceleration\n");
438 #if R200_MERGED
439 FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
440 #endif
441 }
442 if (tcl_mode == DRI_CONF_TCL_SW ||
443 !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
444 if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
445 r300->radeon.radeonScreen->chip_flags &=
446 ~RADEON_CHIPSET_TCL;
447 fprintf(stderr, "Disabling HW TCL support\n");
448 }
449 TCL_FALLBACK(r300->radeon.glCtx,
450 RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
451 }
452
453 return GL_TRUE;
454 }
455
456 /* Destroy the device specific context.
457 */
458 void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
459 {
460 GET_CURRENT_CONTEXT(ctx);
461 r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate;
462 radeonContextPtr radeon = (radeonContextPtr) r300;
463 radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
464
465 if (RADEON_DEBUG & DEBUG_DRI) {
466 fprintf(stderr, "Destroying context !\n");
467 }
468
469 /* check if we're deleting the currently bound context */
470 if (&r300->radeon == current) {
471 radeonFlush(r300->radeon.glCtx);
472 _mesa_make_current(NULL, NULL, NULL);
473 }
474
475 /* Free r300 context resources */
476 assert(r300); /* should never be null */
477
478 if (r300) {
479 _swsetup_DestroyContext(r300->radeon.glCtx);
480 _tnl_DestroyContext(r300->radeon.glCtx);
481 _vbo_DestroyContext(r300->radeon.glCtx);
482 _swrast_DestroyContext(r300->radeon.glCtx);
483
484 rcommonFlushCmdBuf(&r300->radeon, __FUNCTION__);
485
486 if (radeon->state.scissor.pClipRects) {
487 FREE(radeon->state.scissor.pClipRects);
488 radeon->state.scissor.pClipRects = NULL;
489 }
490
491 r300DestroyCmdBuf(r300);
492
493 radeonCleanupContext(&r300->radeon);
494
495
496 /* the memory manager might be accessed when Mesa frees the shared
497 * state, so don't destroy it earlier
498 */
499
500
501 FREE(r300);
502 }
503 }