Merge remote branch 'main/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 r300ContextPtr r300 = (r300ContextPtr)radeon;
239 BATCH_LOCALS(radeon);
240
241 r300->vap_flush_needed = GL_TRUE;
242
243 cp_wait(radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
244 BEGIN_BATCH_NO_AUTOSTATE(2);
245 OUT_BATCH_REGVAL(R300_TX_INVALTAGS, R300_TX_FLUSH);
246 END_BATCH();
247 end_3d(radeon);
248 }
249
250 static void r300_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
251 {
252 r300ContextPtr r300 = R300_CONTEXT(ctx);
253 if (mode)
254 r300->radeon.Fallback |= bit;
255 else
256 r300->radeon.Fallback &= ~bit;
257 }
258
259 static void r300_init_vtbl(radeonContextPtr radeon)
260 {
261 radeon->vtbl.get_lock = r300_get_lock;
262 radeon->vtbl.update_viewport_offset = r300UpdateViewportOffset;
263 radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header;
264 radeon->vtbl.swtcl_flush = r300_swtcl_flush;
265 radeon->vtbl.pre_emit_atoms = r300_vtbl_pre_emit_atoms;
266 radeon->vtbl.fallback = r300_fallback;
267 }
268
269
270 /* Create the device specific rendering context.
271 */
272 GLboolean r300CreateContext(const __GLcontextModes * glVisual,
273 __DRIcontextPrivate * driContextPriv,
274 void *sharedContextPrivate)
275 {
276 __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
277 radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
278 struct dd_function_table functions;
279 r300ContextPtr r300;
280 GLcontext *ctx;
281 int tcl_mode;
282
283 assert(glVisual);
284 assert(driContextPriv);
285 assert(screen);
286
287 /* Allocate the R300 context */
288 r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
289 if (!r300)
290 return GL_FALSE;
291
292 if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
293 hw_tcl_on = future_hw_tcl_on = 0;
294
295 r300_init_vtbl(&r300->radeon);
296 /* Parse configuration files.
297 * Do this here so that initialMaxAnisotropy is set before we create
298 * the default textures.
299 */
300 driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
301 screen->driScreen->myNum, "r300");
302 r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
303 "def_max_anisotropy");
304
305 /* Init default driver functions then plug in our R300-specific functions
306 * (the texture functions are especially important)
307 */
308 _mesa_init_driver_functions(&functions);
309 r300InitIoctlFuncs(&functions);
310 r300InitStateFuncs(&functions);
311 r300InitTextureFuncs(&functions);
312 r300InitShaderFuncs(&functions);
313
314 if (!radeonInitContext(&r300->radeon, &functions,
315 glVisual, driContextPriv,
316 sharedContextPrivate)) {
317 FREE(r300);
318 return GL_FALSE;
319 }
320
321 /* Init r300 context data */
322 r300->radeon.texture_depth = driQueryOptioni(&r300->radeon.optionCache,
323 "texture_depth");
324 if (r300->radeon.texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
325 r300->radeon.texture_depth = (screen->cpp == 4) ?
326 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
327
328 /* Set the maximum texture size small enough that we can guarentee that
329 * all texture units can bind a maximal texture and have them both in
330 * texturable memory at once.
331 */
332
333 ctx = r300->radeon.glCtx;
334
335 ctx->Const.MaxTextureImageUnits =
336 driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
337 ctx->Const.MaxTextureCoordUnits =
338 driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
339 ctx->Const.MaxTextureUnits =
340 MIN2(ctx->Const.MaxTextureImageUnits,
341 ctx->Const.MaxTextureCoordUnits);
342 ctx->Const.MaxTextureMaxAnisotropy = 16.0;
343 ctx->Const.MaxTextureLodBias = 16.0;
344
345 if (screen->chip_family >= CHIP_FAMILY_RV515) {
346 ctx->Const.MaxTextureLevels = 13;
347 ctx->Const.MaxTextureRectSize = 4096;
348 }
349
350 ctx->Const.MinPointSize = 1.0;
351 ctx->Const.MinPointSizeAA = 1.0;
352 ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
353 ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
354
355 ctx->Const.MinLineWidth = 1.0;
356 ctx->Const.MinLineWidthAA = 1.0;
357 ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
358 ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
359
360 /* Needs further modifications */
361 #if 0
362 ctx->Const.MaxArrayLockSize =
363 ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
364 #endif
365
366 ctx->Const.MaxDrawBuffers = 1;
367
368 /* Initialize the software rasterizer and helper modules.
369 */
370 _swrast_CreateContext(ctx);
371 _vbo_CreateContext(ctx);
372 _tnl_CreateContext(ctx);
373 _swsetup_CreateContext(ctx);
374 _swsetup_Wakeup(ctx);
375 _ae_create_context(ctx);
376
377 /* Install the customized pipeline:
378 */
379 _tnl_destroy_pipeline(ctx);
380 _tnl_install_pipeline(ctx, r300_pipeline);
381
382 /* Try and keep materials and vertices separate:
383 */
384 /* _tnl_isolate_materials(ctx, GL_TRUE); */
385
386 /* Configure swrast and TNL to match hardware characteristics:
387 */
388 _swrast_allow_pixel_fog(ctx, GL_FALSE);
389 _swrast_allow_vertex_fog(ctx, GL_TRUE);
390 _tnl_allow_pixel_fog(ctx, GL_FALSE);
391 _tnl_allow_vertex_fog(ctx, GL_TRUE);
392
393 /* currently bogus data */
394 if (screen->chip_flags & RADEON_CHIPSET_TCL) {
395 ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
396 ctx->Const.VertexProgram.MaxNativeInstructions =
397 VSF_MAX_FRAGMENT_LENGTH / 4;
398 ctx->Const.VertexProgram.MaxNativeAttribs = 16; /* r420 */
399 ctx->Const.VertexProgram.MaxTemps = 32;
400 ctx->Const.VertexProgram.MaxNativeTemps =
401 /*VSF_MAX_FRAGMENT_TEMPS */ 32;
402 ctx->Const.VertexProgram.MaxNativeParameters = 256; /* r420 */
403 ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
404 }
405
406 ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
407 ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* copy i915... */
408 ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
409 ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
410 ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
411 ctx->Const.FragmentProgram.MaxNativeInstructions =
412 PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
413 ctx->Const.FragmentProgram.MaxNativeTexIndirections =
414 PFS_MAX_TEX_INDIRECT;
415 ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* and these are?? */
416 ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
417 ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
418
419 driInitExtensions(ctx, card_extensions, GL_TRUE);
420
421 if (driQueryOptionb
422 (&r300->radeon.optionCache, "disable_stencil_two_side"))
423 _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
424
425 if (r300->radeon.glCtx->Mesa_DXTn
426 && !driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc")) {
427 _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
428 _mesa_enable_extension(ctx, "GL_S3_s3tc");
429 } else
430 if (driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable"))
431 {
432 _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
433 }
434
435 r300->disable_lowimpact_fallback =
436 driQueryOptionb(&r300->radeon.optionCache,
437 "disable_lowimpact_fallback");
438
439 radeonInitSpanFuncs( ctx );
440 r300InitCmdBuf(r300);
441 r300InitState(r300);
442 if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
443 r300InitSwtcl(ctx);
444
445 TNL_CONTEXT(ctx)->Driver.RunPipeline = r300RunPipeline;
446
447 tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
448 if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
449 fprintf(stderr, "disabling 3D acceleration\n");
450 #if R200_MERGED
451 FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
452 #endif
453 }
454 if (tcl_mode == DRI_CONF_TCL_SW ||
455 !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
456 if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
457 r300->radeon.radeonScreen->chip_flags &=
458 ~RADEON_CHIPSET_TCL;
459 fprintf(stderr, "Disabling HW TCL support\n");
460 }
461 TCL_FALLBACK(r300->radeon.glCtx,
462 RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
463 }
464
465 return GL_TRUE;
466 }
467
468 /* Destroy the device specific context.
469 */
470 void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
471 {
472 GET_CURRENT_CONTEXT(ctx);
473 r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate;
474 radeonContextPtr radeon = (radeonContextPtr) r300;
475 radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
476
477 if (RADEON_DEBUG & DEBUG_DRI) {
478 fprintf(stderr, "Destroying context !\n");
479 }
480
481 /* check if we're deleting the currently bound context */
482 if (&r300->radeon == current) {
483 radeonFlush(r300->radeon.glCtx);
484 _mesa_make_current(NULL, NULL, NULL);
485 }
486
487 /* Free r300 context resources */
488 assert(r300); /* should never be null */
489
490 if (r300) {
491 _swsetup_DestroyContext(r300->radeon.glCtx);
492 _tnl_DestroyContext(r300->radeon.glCtx);
493 _vbo_DestroyContext(r300->radeon.glCtx);
494 _swrast_DestroyContext(r300->radeon.glCtx);
495
496 rcommonFlushCmdBuf(&r300->radeon, __FUNCTION__);
497
498 if (radeon->state.scissor.pClipRects) {
499 FREE(radeon->state.scissor.pClipRects);
500 radeon->state.scissor.pClipRects = NULL;
501 }
502
503 r300DestroyCmdBuf(r300);
504
505 radeonCleanupContext(&r300->radeon);
506
507
508 /* the memory manager might be accessed when Mesa frees the shared
509 * state, so don't destroy it earlier
510 */
511
512
513 FREE(r300);
514 }
515 }