Merge branch 'vbo-0.2'
[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 * Authors:
32 * Keith Whitwell <keith@tungstengraphics.com>
33 * Nicolai Haehnle <prefect_@gmx.net>
34 */
35 #include "glheader.h"
36 #include "api_arrayelt.h"
37 #include "context.h"
38 #include "simple_list.h"
39 #include "imports.h"
40 #include "matrix.h"
41 #include "extensions.h"
42 #include "state.h"
43 #include "bufferobj.h"
44
45 #include "swrast/swrast.h"
46 #include "swrast_setup/swrast_setup.h"
47 #include "vbo/vbo.h"
48
49 #include "tnl/tnl.h"
50 #include "tnl/t_pipeline.h"
51 #include "tnl/t_vp_build.h"
52
53 #include "drivers/common/driverfuncs.h"
54
55 #include "radeon_ioctl.h"
56 #include "radeon_span.h"
57 #include "r300_context.h"
58 #include "r300_cmdbuf.h"
59 #include "r300_state.h"
60 #include "r300_ioctl.h"
61 #include "r300_tex.h"
62 #include "r300_maos.h"
63
64 #ifdef USER_BUFFERS
65 #include "radeon_mm.h"
66 #endif
67
68 #include "vblank.h"
69 #include "utils.h"
70 #include "xmlpool.h" /* for symbolic values of enum-type options */
71
72 /* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
73 int future_hw_tcl_on=1;
74 int hw_tcl_on=1;
75
76 #define need_GL_EXT_stencil_two_side
77 #define need_GL_ARB_multisample
78 #define need_GL_ARB_texture_compression
79 #define need_GL_ARB_vertex_buffer_object
80 #define need_GL_ARB_vertex_program
81 #define need_GL_EXT_blend_minmax
82 //#define need_GL_EXT_fog_coord
83 #define need_GL_EXT_secondary_color
84 #define need_GL_EXT_blend_equation_separate
85 #define need_GL_EXT_blend_func_separate
86 #define need_GL_EXT_gpu_program_parameters
87 #define need_GL_NV_vertex_program
88 #include "extension_helper.h"
89
90 const struct dri_extension card_extensions[] = {
91 {"GL_ARB_multisample", GL_ARB_multisample_functions},
92 {"GL_ARB_multitexture", NULL},
93 {"GL_ARB_texture_border_clamp", NULL},
94 {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
95 {"GL_ARB_texture_cube_map", NULL},
96 {"GL_ARB_texture_env_add", NULL},
97 {"GL_ARB_texture_env_combine", NULL},
98 {"GL_ARB_texture_env_crossbar", NULL},
99 {"GL_ARB_texture_env_dot3", NULL},
100 {"GL_ARB_texture_mirrored_repeat", NULL},
101 {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
102 {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
103 {"GL_ARB_fragment_program", NULL},
104 {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
105 {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
106 {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
107 {"GL_EXT_blend_subtract", NULL},
108 // {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
109 {"GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions},
110 {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
111 {"GL_EXT_stencil_wrap", NULL},
112 {"GL_EXT_texture_edge_clamp", NULL},
113 {"GL_EXT_texture_env_combine", NULL},
114 {"GL_EXT_texture_env_dot3", NULL},
115 {"GL_EXT_texture_filter_anisotropic", NULL},
116 {"GL_EXT_texture_lod_bias", NULL},
117 {"GL_EXT_texture_mirror_clamp", NULL},
118 {"GL_EXT_texture_rectangle", NULL},
119 {"GL_ATI_texture_env_combine3", NULL},
120 {"GL_ATI_texture_mirror_once", NULL},
121 {"GL_MESA_pack_invert", NULL},
122 {"GL_MESA_ycbcr_texture", NULL},
123 {"GL_MESAX_texture_float", NULL},
124 {"GL_NV_blend_square", NULL},
125 {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
126 {"GL_SGIS_generate_mipmap", NULL},
127 {NULL, NULL}
128 };
129
130 const struct dri_extension stencil_two_side[] = {
131 {"GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions},
132 };
133
134 extern struct tnl_pipeline_stage _r300_render_stage;
135 extern const struct tnl_pipeline_stage _r300_tcl_stage;
136 extern const struct tnl_pipeline_stage _r300_texrect_stage;
137
138 static const struct tnl_pipeline_stage *r300_pipeline[] = {
139
140 /* Try and go straight to t&l
141 */
142 &_r300_tcl_stage,
143
144 /* Catch any t&l fallbacks
145 */
146 &_tnl_vertex_transform_stage,
147 &_tnl_normal_transform_stage,
148 &_tnl_lighting_stage,
149 &_tnl_fog_coordinate_stage,
150 &_tnl_texgen_stage,
151 &_tnl_texture_transform_stage,
152 &_tnl_arb_vertex_program_stage,
153 &_tnl_vertex_program_stage,
154
155 /* Try again to go to tcl?
156 * - no good for asymmetric-twoside (do with multipass)
157 * - no good for asymmetric-unfilled (do with multipass)
158 * - good for material
159 * - good for texgen
160 * - need to manipulate a bit of state
161 *
162 * - worth it/not worth it?
163 */
164
165 /* Else do them here.
166 */
167 /* scale texture rectangle to 0..1. */
168 &_r300_texrect_stage,
169 &_r300_render_stage,
170 &_tnl_render_stage, /* FALLBACK */
171 0,
172 };
173
174
175 /* Create the device specific rendering context.
176 */
177 GLboolean r300CreateContext(const __GLcontextModes * glVisual,
178 __DRIcontextPrivate * driContextPriv,
179 void *sharedContextPrivate)
180 {
181 __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
182 radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
183 struct dd_function_table functions;
184 r300ContextPtr r300;
185 GLcontext *ctx;
186 int tcl_mode, i;
187
188 assert(glVisual);
189 assert(driContextPriv);
190 assert(screen);
191
192 /* Allocate the R300 context */
193 r300 = (r300ContextPtr)CALLOC(sizeof(*r300));
194 if (!r300)
195 return GL_FALSE;
196
197 /* Parse configuration files.
198 * Do this here so that initialMaxAnisotropy is set before we create
199 * the default textures.
200 */
201 driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
202 screen->driScreen->myNum, "r300");
203
204 //r300->texmicrotile = GL_TRUE;
205
206 /* Init default driver functions then plug in our R300-specific functions
207 * (the texture functions are especially important)
208 */
209 _mesa_init_driver_functions(&functions);
210 r300InitIoctlFuncs(&functions);
211 r300InitStateFuncs(&functions);
212 r300InitTextureFuncs(&functions);
213 r300InitShaderFuncs(&functions);
214
215 #ifdef USER_BUFFERS
216 radeon_mm_init(r300);
217 #endif
218 #ifdef HW_VBOS
219 if (hw_tcl_on) {
220 r300_init_vbo_funcs(&functions);
221 }
222 #endif
223 if (!radeonInitContext(&r300->radeon, &functions,
224 glVisual, driContextPriv, sharedContextPrivate)) {
225 FREE(r300);
226 return GL_FALSE;
227 }
228
229 /* Init r300 context data */
230 r300->dma.buf0_address = r300->radeon.radeonScreen->buffers->list[0].address;
231
232 (void)memset(r300->texture_heaps, 0, sizeof(r300->texture_heaps));
233 make_empty_list(&r300->swapped);
234
235 r300->nr_heaps = 1 /* screen->numTexHeaps */ ;
236 assert(r300->nr_heaps < RADEON_NR_TEX_HEAPS);
237 for (i = 0; i < r300->nr_heaps; i++) {
238 r300->texture_heaps[i] = driCreateTextureHeap(i, r300,
239 screen->
240 texSize[i], 12,
241 RADEON_NR_TEX_REGIONS,
242 (drmTextureRegionPtr)
243 r300->radeon.sarea->
244 tex_list[i],
245 &r300->radeon.sarea->
246 tex_age[i],
247 &r300->swapped,
248 sizeof
249 (r300TexObj),
250 (destroy_texture_object_t
251 *)
252 r300DestroyTexObj);
253 }
254 r300->texture_depth = driQueryOptioni(&r300->radeon.optionCache,
255 "texture_depth");
256 if (r300->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
257 r300->texture_depth = (screen->cpp == 4) ?
258 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
259
260 /* Set the maximum texture size small enough that we can guarentee that
261 * all texture units can bind a maximal texture and have them both in
262 * texturable memory at once.
263 */
264
265 ctx = r300->radeon.glCtx;
266
267 ctx->Const.MaxTextureImageUnits = driQueryOptioni(&r300->radeon.optionCache,
268 "texture_image_units");
269 ctx->Const.MaxTextureCoordUnits = driQueryOptioni(&r300->radeon.optionCache,
270 "texture_coord_units");
271 ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureImageUnits,
272 ctx->Const.MaxTextureCoordUnits);
273 ctx->Const.MaxTextureMaxAnisotropy = 16.0;
274
275 ctx->Const.MinPointSize = 1.0;
276 ctx->Const.MinPointSizeAA = 1.0;
277 ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
278 ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
279
280 ctx->Const.MinLineWidth = 1.0;
281 ctx->Const.MinLineWidthAA = 1.0;
282 ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
283 ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
284
285 #ifdef USER_BUFFERS
286 /* Needs further modifications */
287 #if 0
288 ctx->Const.MaxArrayLockSize = (/*512*/RADEON_BUFFER_SIZE*16*1024) / (4*4);
289 #endif
290 #endif
291
292 /* Initialize the software rasterizer and helper modules.
293 */
294 _swrast_CreateContext(ctx);
295 _vbo_CreateContext(ctx);
296 _tnl_CreateContext(ctx);
297 _swsetup_CreateContext(ctx);
298 _swsetup_Wakeup(ctx);
299 _ae_create_context(ctx);
300
301 /* Install the customized pipeline:
302 */
303 _tnl_destroy_pipeline(ctx);
304 _tnl_install_pipeline(ctx, r300_pipeline);
305
306 /* Try and keep materials and vertices separate:
307 */
308 /* _tnl_isolate_materials(ctx, GL_TRUE); */
309
310 /* Configure swrast and TNL to match hardware characteristics:
311 */
312 _swrast_allow_pixel_fog(ctx, GL_FALSE);
313 _swrast_allow_vertex_fog(ctx, GL_TRUE);
314 _tnl_allow_pixel_fog(ctx, GL_FALSE);
315 _tnl_allow_vertex_fog(ctx, GL_TRUE);
316
317 /* currently bogus data */
318 ctx->Const.VertexProgram.MaxInstructions=VSF_MAX_FRAGMENT_LENGTH/4;
319 ctx->Const.VertexProgram.MaxNativeInstructions=VSF_MAX_FRAGMENT_LENGTH/4;
320 ctx->Const.VertexProgram.MaxNativeAttribs=16; /* r420 */
321 ctx->Const.VertexProgram.MaxTemps=32;
322 ctx->Const.VertexProgram.MaxNativeTemps=/*VSF_MAX_FRAGMENT_TEMPS*/32;
323 ctx->Const.VertexProgram.MaxNativeParameters=256; /* r420 */
324 ctx->Const.VertexProgram.MaxNativeAddressRegs=1;
325
326 ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
327 ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* copy i915... */
328 ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
329 ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
330 ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
331 ctx->Const.FragmentProgram.MaxNativeInstructions = PFS_MAX_ALU_INST+PFS_MAX_TEX_INST;
332 ctx->Const.FragmentProgram.MaxNativeTexIndirections = PFS_MAX_TEX_INDIRECT;
333 ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* and these are?? */
334 _tnl_ProgramCacheInit(ctx);
335 ctx->_MaintainTexEnvProgram = GL_TRUE;
336
337 driInitExtensions(ctx, card_extensions, GL_TRUE);
338
339 if (driQueryOptionb(&r300->radeon.optionCache, "disable_stencil_two_side") == 0)
340 driInitSingleExtension(ctx, stencil_two_side);
341
342 if (r300->radeon.glCtx->Mesa_DXTn && !driQueryOptionb (&r300->radeon.optionCache, "disable_s3tc")) {
343 _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
344 _mesa_enable_extension( ctx, "GL_S3_s3tc" );
345 }
346 else if (driQueryOptionb (&r300->radeon.optionCache, "force_s3tc_enable")) {
347 _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
348 }
349
350 r300->disable_lowimpact_fallback = driQueryOptionb(&r300->radeon.optionCache, "disable_lowimpact_fallback");
351
352 radeonInitSpanFuncs(ctx);
353 r300InitCmdBuf(r300);
354 r300InitState(r300);
355
356 #ifdef RADEON_VTXFMT_A
357 radeon_init_vtxfmt_a(r300);
358 #endif
359
360 #if 0
361 /* plug in a few more device driver functions */
362 /* XXX these should really go right after _mesa_init_driver_functions() */
363 r300InitPixelFuncs(ctx);
364 r300InitSwtcl(ctx);
365 #endif
366 TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
367
368 tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
369 if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
370 fprintf(stderr, "disabling 3D acceleration\n");
371 #if R200_MERGED
372 FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
373 #endif
374 }
375 if (tcl_mode == DRI_CONF_TCL_SW ||
376 !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
377 if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
378 r300->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
379 fprintf(stderr, "Disabling HW TCL support\n");
380 }
381 TCL_FALLBACK(r300->radeon.glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
382 }
383
384 return GL_TRUE;
385 }
386
387 static void r300FreeGartAllocations(r300ContextPtr r300)
388 {
389 int i, ret, tries=0, done_age, in_use=0;
390 drm_radeon_mem_free_t memfree;
391
392 memfree.region = RADEON_MEM_REGION_GART;
393
394 #ifdef USER_BUFFERS
395 for (i = r300->rmm->u_last; i > 0; i--) {
396 if (r300->rmm->u_list[i].ptr == NULL) {
397 continue;
398 }
399
400 /* check whether this buffer is still in use */
401 if (r300->rmm->u_list[i].pending) {
402 in_use++;
403 }
404 }
405 /* Cannot flush/lock if no context exists. */
406 if (in_use)
407 r300FlushCmdBuf(r300, __FUNCTION__);
408
409 done_age = radeonGetAge((radeonContextPtr)r300);
410
411 for (i = r300->rmm->u_last; i > 0; i--) {
412 if (r300->rmm->u_list[i].ptr == NULL) {
413 continue;
414 }
415
416 /* check whether this buffer is still in use */
417 if (!r300->rmm->u_list[i].pending) {
418 continue;
419 }
420
421 assert(r300->rmm->u_list[i].h_pending == 0);
422
423 tries = 0;
424 while(r300->rmm->u_list[i].age > done_age && tries++ < 1000) {
425 usleep(10);
426 done_age = radeonGetAge((radeonContextPtr)r300);
427 }
428 if (tries >= 1000) {
429 WARN_ONCE("Failed to idle region!");
430 }
431
432 memfree.region_offset = (char *)r300->rmm->u_list[i].ptr -
433 (char *)r300->radeon.radeonScreen->gartTextures.map;
434
435 ret = drmCommandWrite(r300->radeon.radeonScreen->driScreen->fd,
436 DRM_RADEON_FREE, &memfree, sizeof(memfree));
437 if (ret) {
438 fprintf(stderr, "Failed to free at %p\nret = %s\n",
439 r300->rmm->u_list[i].ptr, strerror(-ret));
440 } else {
441 if (i == r300->rmm->u_last)
442 r300->rmm->u_last--;
443
444 r300->rmm->u_list[i].pending = 0;
445 r300->rmm->u_list[i].ptr = NULL;
446 if (r300->rmm->u_list[i].fb) {
447 LOCK_HARDWARE(&(r300->radeon));
448 ret = mmFreeMem(r300->rmm->u_list[i].fb);
449 UNLOCK_HARDWARE(&(r300->radeon));
450 if (ret) fprintf(stderr, "failed to free!\n");
451 r300->rmm->u_list[i].fb = NULL;
452 }
453 r300->rmm->u_list[i].ref_count = 0;
454 }
455 }
456 r300->rmm->u_head = i;
457 #endif /* USER_BUFFERS */
458 }
459
460 /* Destroy the device specific context.
461 */
462 void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
463 {
464 GET_CURRENT_CONTEXT(ctx);
465 r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate;
466 radeonContextPtr radeon = (radeonContextPtr) r300;
467 radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
468
469 if (RADEON_DEBUG & DEBUG_DRI) {
470 fprintf(stderr, "Destroying context !\n");
471 }
472
473 /* check if we're deleting the currently bound context */
474 if (&r300->radeon == current) {
475 radeonFlush(r300->radeon.glCtx);
476 _mesa_make_current(NULL, NULL, NULL);
477 }
478
479 /* Free r300 context resources */
480 assert(r300); /* should never be null */
481
482 if (r300) {
483 GLboolean release_texture_heaps;
484
485 release_texture_heaps = (r300->radeon.glCtx->Shared->RefCount == 1);
486 _swsetup_DestroyContext(r300->radeon.glCtx);
487 _tnl_ProgramCacheDestroy(r300->radeon.glCtx);
488 _tnl_DestroyContext(r300->radeon.glCtx);
489 _vbo_DestroyContext(r300->radeon.glCtx);
490 _swrast_DestroyContext(r300->radeon.glCtx);
491
492 if (r300->dma.current.buf) {
493 r300ReleaseDmaRegion(r300, &r300->dma.current, __FUNCTION__ );
494 #ifndef USER_BUFFERS
495 r300FlushCmdBuf(r300, __FUNCTION__);
496 #endif
497 }
498 r300FreeGartAllocations(r300);
499 r300DestroyCmdBuf(r300);
500
501 if (radeon->state.scissor.pClipRects) {
502 FREE(radeon->state.scissor.pClipRects);
503 radeon->state.scissor.pClipRects = NULL;
504 }
505
506 if (release_texture_heaps) {
507 /* This share group is about to go away, free our private
508 * texture object data.
509 */
510 int i;
511
512 for (i = 0; i < r300->nr_heaps; i++) {
513 driDestroyTextureHeap(r300->texture_heaps[i]);
514 r300->texture_heaps[i] = NULL;
515 }
516
517 assert(is_empty_list(&r300->swapped));
518 }
519
520 radeonCleanupContext(&r300->radeon);
521
522 #ifdef USER_BUFFERS
523 /* the memory manager might be accessed when Mesa frees the shared
524 * state, so don't destroy it earlier
525 */
526 radeon_mm_destroy(r300);
527 #endif
528
529 /* free the option cache */
530 driDestroyOptionCache(&r300->radeon.optionCache);
531
532 FREE(r300);
533 }
534 }