i965: Add new vtable entries for surface state updating functions.
[mesa.git] / src / mesa / drivers / dri / intel / intel_context.h
1 /**************************************************************************
2 *
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 #ifndef INTELCONTEXT_INC
29 #define INTELCONTEXT_INC
30
31
32 #include <stdbool.h>
33 #include "main/mtypes.h"
34 #include "main/mm.h"
35
36 #ifdef __cplusplus
37 extern "C" {
38 /* Evil hack for using libdrm in a c++ compiler. */
39 #define virtual virt
40 #endif
41
42 #include "drm.h"
43 #include "intel_bufmgr.h"
44
45 #include "intel_screen.h"
46 #include "intel_tex_obj.h"
47 #include "i915_drm.h"
48
49 #ifdef __cplusplus
50 #undef virtual
51 }
52 #endif
53
54 #include "tnl/t_vertex.h"
55
56 #define TAG(x) intel##x
57 #include "tnl_dd/t_dd_vertex.h"
58 #undef TAG
59
60 #define DV_PF_555 (1<<8)
61 #define DV_PF_565 (2<<8)
62 #define DV_PF_8888 (3<<8)
63 #define DV_PF_4444 (8<<8)
64 #define DV_PF_1555 (9<<8)
65
66 struct intel_region;
67 struct intel_context;
68
69 typedef void (*intel_tri_func) (struct intel_context *, intelVertex *,
70 intelVertex *, intelVertex *);
71 typedef void (*intel_line_func) (struct intel_context *, intelVertex *,
72 intelVertex *);
73 typedef void (*intel_point_func) (struct intel_context *, intelVertex *);
74
75 /**
76 * Bits for intel->Fallback field
77 */
78 /*@{*/
79 #define INTEL_FALLBACK_DRAW_BUFFER 0x1
80 #define INTEL_FALLBACK_READ_BUFFER 0x2
81 #define INTEL_FALLBACK_DEPTH_BUFFER 0x4
82 #define INTEL_FALLBACK_STENCIL_BUFFER 0x8
83 #define INTEL_FALLBACK_USER 0x10
84 #define INTEL_FALLBACK_RENDERMODE 0x20
85 #define INTEL_FALLBACK_TEXTURE 0x40
86 #define INTEL_FALLBACK_DRIVER 0x1000 /**< first for drivers */
87 /*@}*/
88
89 extern void intelFallback(struct intel_context *intel, GLbitfield bit,
90 bool mode);
91 #define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode )
92
93
94 #define INTEL_WRITE_PART 0x1
95 #define INTEL_WRITE_FULL 0x2
96 #define INTEL_READ 0x4
97
98 #define INTEL_MAX_FIXUP 64
99
100 #ifndef likely
101 #ifdef __GNUC__
102 #define likely(expr) (__builtin_expect(expr, 1))
103 #define unlikely(expr) (__builtin_expect(expr, 0))
104 #else
105 #define likely(expr) (expr)
106 #define unlikely(expr) (expr)
107 #endif
108 #endif
109
110 struct intel_sync_object {
111 struct gl_sync_object Base;
112
113 /** Batch associated with this sync object */
114 drm_intel_bo *bo;
115 };
116
117 struct brw_context;
118
119 /**
120 * intel_context is derived from Mesa's context class: struct gl_context.
121 */
122 struct intel_context
123 {
124 struct gl_context ctx; /**< base class, must be first field */
125
126 struct
127 {
128 void (*destroy) (struct intel_context * intel);
129 void (*emit_state) (struct intel_context * intel);
130 void (*finish_batch) (struct intel_context * intel);
131 void (*new_batch) (struct intel_context * intel);
132 void (*emit_invarient_state) (struct intel_context * intel);
133 void (*update_texture_state) (struct intel_context * intel);
134
135 void (*render_start) (struct intel_context * intel);
136 void (*render_prevalidate) (struct intel_context * intel);
137 void (*set_draw_region) (struct intel_context * intel,
138 struct intel_region * draw_regions[],
139 struct intel_region * depth_region,
140 GLuint num_regions);
141 void (*update_draw_buffer)(struct intel_context *intel);
142
143 void (*reduced_primitive_state) (struct intel_context * intel,
144 GLenum rprim);
145
146 bool (*check_vertex_size) (struct intel_context * intel,
147 GLuint expected);
148 void (*invalidate_state) (struct intel_context *intel,
149 GLuint new_state);
150
151 void (*assert_not_dirty) (struct intel_context *intel);
152
153 void (*debug_batch)(struct intel_context *intel);
154 bool (*render_target_supported)(gl_format format);
155
156 /** Can HiZ be enabled on a depthbuffer of the given format? */
157 bool (*is_hiz_depth_format)(struct intel_context *intel,
158 gl_format format);
159
160 /**
161 * \name HiZ operations
162 *
163 * See the following sections of the Sandy Bridge PRM, Volume 1, Part2:
164 * - 7.5.3.1 Depth Buffer Clear
165 * - 7.5.3.2 Depth Buffer Resolve
166 * - 7.5.3.3 Hierarchical Depth Buffer Resolve
167 * \{
168 */
169 void (*hiz_resolve_depthbuffer)(struct intel_context *intel,
170 struct intel_region *depth_region);
171 void (*hiz_resolve_hizbuffer)(struct intel_context *intel,
172 struct intel_region *depth_region);
173 /** \} */
174
175 /**
176 * Surface state operations (i965+ only)
177 * \{
178 */
179 void (*update_texture_surface)(struct gl_context *ctx, unsigned unit);
180 void (*update_renderbuffer_surface)(struct brw_context *brw,
181 struct gl_renderbuffer *rb,
182 unsigned unit);
183 void (*update_null_renderbuffer_surface)(struct brw_context *brw,
184 unsigned unit);
185 void (*create_constant_surface)(struct brw_context *brw,
186 drm_intel_bo *bo,
187 int width,
188 uint32_t *out_offset);
189 /** \} */
190 } vtbl;
191
192 GLbitfield Fallback; /**< mask of INTEL_FALLBACK_x bits */
193 GLuint NewGLState;
194
195 dri_bufmgr *bufmgr;
196 unsigned int maxBatchSize;
197
198 /**
199 * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965.
200 */
201 int gen;
202 int gt;
203 bool needs_ff_sync;
204 bool is_g4x;
205 bool is_945;
206 bool has_separate_stencil;
207 bool must_use_separate_stencil;
208 bool has_hiz;
209
210 int urb_size;
211
212 struct intel_batchbuffer {
213 /** Current batchbuffer being queued up. */
214 drm_intel_bo *bo;
215 /** Last BO submitted to the hardware. Used for glFinish(). */
216 drm_intel_bo *last_bo;
217 /** BO for post-sync nonzero writes for gen6 workaround. */
218 drm_intel_bo *workaround_bo;
219 bool need_workaround_flush;
220
221 struct cached_batch_item *cached_items;
222
223 uint16_t emit, total;
224 uint16_t used, reserved_space;
225 uint32_t map[8192];
226 #define BATCH_SZ (8192*sizeof(uint32_t))
227
228 uint32_t state_batch_offset;
229 bool is_blit;
230
231 struct {
232 uint16_t used;
233 int reloc_count;
234 } saved;
235 } batch;
236
237 drm_intel_bo *first_post_swapbuffers_batch;
238 bool need_throttle;
239 bool no_batch_wrap;
240 bool tnl_pipeline_running; /**< Set while i915's _tnl_run_pipeline. */
241
242 struct
243 {
244 GLuint id;
245 uint32_t start_ptr; /**< for i8xx */
246 uint32_t primitive; /**< Current hardware primitive type */
247 void (*flush) (struct intel_context *);
248 drm_intel_bo *vb_bo;
249 uint8_t *vb;
250 unsigned int start_offset; /**< Byte offset of primitive sequence */
251 unsigned int current_offset; /**< Byte offset of next vertex */
252 unsigned int count; /**< Number of vertices in current primitive */
253 } prim;
254
255 struct {
256 drm_intel_bo *bo;
257 GLuint offset;
258 uint32_t buffer_len;
259 uint32_t buffer_offset;
260 char buffer[4096];
261 } upload;
262
263 GLuint stats_wm;
264
265 /* Offsets of fields within the current vertex:
266 */
267 GLuint coloroffset;
268 GLuint specoffset;
269 GLuint wpos_offset;
270
271 struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
272 GLuint vertex_attr_count;
273
274 GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */
275
276 bool hw_stencil;
277 bool hw_stipple;
278 bool depth_buffer_is_float;
279 bool no_rast;
280 bool always_flush_batch;
281 bool always_flush_cache;
282
283 /* 0 - nonconformant, best performance;
284 * 1 - fallback to sw for known conformance bugs
285 * 2 - always fallback to sw
286 */
287 GLuint conformance_mode;
288
289 /* State for intelvb.c and inteltris.c.
290 */
291 GLuint RenderIndex;
292 GLmatrix ViewportMatrix;
293 GLenum render_primitive;
294 GLenum reduced_primitive; /*< Only gen < 6 */
295 GLuint vertex_size;
296 GLubyte *verts; /* points to tnl->clipspace.vertex_buf */
297
298 /* Fallback rasterization functions
299 */
300 intel_point_func draw_point;
301 intel_line_func draw_line;
302 intel_tri_func draw_tri;
303
304 /**
305 * Set if rendering has occured to the drawable's front buffer.
306 *
307 * This is used in the DRI2 case to detect that glFlush should also copy
308 * the contents of the fake front buffer to the real front buffer.
309 */
310 bool front_buffer_dirty;
311
312 /**
313 * Track whether front-buffer rendering is currently enabled
314 *
315 * A separate flag is used to track this in order to support MRT more
316 * easily.
317 */
318 bool is_front_buffer_rendering;
319 /**
320 * Track whether front-buffer is the current read target.
321 *
322 * This is closely associated with is_front_buffer_rendering, but may
323 * be set separately. The DRI2 fake front buffer must be referenced
324 * either way.
325 */
326 bool is_front_buffer_reading;
327
328 /**
329 * Count of intel_regions that are mapped.
330 *
331 * This allows us to assert that no batch buffer is emitted if a
332 * region is mapped.
333 */
334 int num_mapped_regions;
335
336 bool use_texture_tiling;
337 bool use_early_z;
338
339 int driFd;
340
341 __DRIcontext *driContext;
342 struct intel_screen *intelScreen;
343 void (*saved_viewport)(struct gl_context * ctx,
344 GLint x, GLint y, GLsizei width, GLsizei height);
345
346 /**
347 * Configuration cache
348 */
349 driOptionCache optionCache;
350 };
351
352 extern char *__progname;
353
354
355 #define SUBPIXEL_X 0.125
356 #define SUBPIXEL_Y 0.125
357
358 #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
359
360 /**
361 * Align a value up to an alignment value
362 *
363 * If \c value is not already aligned to the requested alignment value, it
364 * will be rounded up.
365 *
366 * \param value Value to be rounded
367 * \param alignment Alignment value to be used. This must be a power of two.
368 *
369 * \sa ROUND_DOWN_TO()
370 */
371 #define ALIGN(value, alignment) (((value) + alignment - 1) & ~(alignment - 1))
372
373 /**
374 * Align a value down to an alignment value
375 *
376 * If \c value is not already aligned to the requested alignment value, it
377 * will be rounded down.
378 *
379 * \param value Value to be rounded
380 * \param alignment Alignment value to be used. This must be a power of two.
381 *
382 * \sa ALIGN()
383 */
384 #define ROUND_DOWN_TO(value, alignment) ((value) & ~(alignment - 1))
385
386 #define IS_POWER_OF_TWO(val) (((val) & (val - 1)) == 0)
387
388 static INLINE uint32_t
389 U_FIXED(float value, uint32_t frac_bits)
390 {
391 value *= (1 << frac_bits);
392 return value < 0 ? 0 : value;
393 }
394
395 static INLINE uint32_t
396 S_FIXED(float value, uint32_t frac_bits)
397 {
398 return value * (1 << frac_bits);
399 }
400
401 #define INTEL_FIREVERTICES(intel) \
402 do { \
403 if ((intel)->prim.flush) \
404 (intel)->prim.flush(intel); \
405 } while (0)
406
407 /* ================================================================
408 * From linux kernel i386 header files, copes with odd sizes better
409 * than COPY_DWORDS would:
410 * XXX Put this in src/mesa/main/imports.h ???
411 */
412 #if defined(i386) || defined(__i386__)
413 static INLINE void * __memcpy(void * to, const void * from, size_t n)
414 {
415 int d0, d1, d2;
416 __asm__ __volatile__(
417 "rep ; movsl\n\t"
418 "testb $2,%b4\n\t"
419 "je 1f\n\t"
420 "movsw\n"
421 "1:\ttestb $1,%b4\n\t"
422 "je 2f\n\t"
423 "movsb\n"
424 "2:"
425 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
426 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
427 : "memory");
428 return (to);
429 }
430 #else
431 #define __memcpy(a,b,c) memcpy(a,b,c)
432 #endif
433
434
435 /* ================================================================
436 * Debugging:
437 */
438 extern int INTEL_DEBUG;
439
440 #define DEBUG_TEXTURE 0x1
441 #define DEBUG_STATE 0x2
442 #define DEBUG_IOCTL 0x4
443 #define DEBUG_BLIT 0x8
444 #define DEBUG_MIPTREE 0x10
445 #define DEBUG_FALLBACKS 0x20
446 #define DEBUG_VERBOSE 0x40
447 #define DEBUG_BATCH 0x80
448 #define DEBUG_PIXEL 0x100
449 #define DEBUG_BUFMGR 0x200
450 #define DEBUG_REGION 0x400
451 #define DEBUG_FBO 0x800
452 #define DEBUG_GS 0x1000
453 #define DEBUG_SYNC 0x2000
454 #define DEBUG_PRIMS 0x4000
455 #define DEBUG_VERTS 0x8000
456 #define DEBUG_DRI 0x10000
457 #define DEBUG_SF 0x20000
458 #define DEBUG_SANITY 0x40000
459 #define DEBUG_SLEEP 0x80000
460 #define DEBUG_STATS 0x100000
461 #define DEBUG_TILE 0x200000
462 #define DEBUG_WM 0x400000
463 #define DEBUG_URB 0x800000
464 #define DEBUG_VS 0x1000000
465 #define DEBUG_CLIP 0x2000000
466
467 #define DBG(...) do { \
468 if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG)) \
469 printf(__VA_ARGS__); \
470 } while(0)
471
472 #define fallback_debug(...) do { \
473 if (unlikely(INTEL_DEBUG & DEBUG_FALLBACKS)) \
474 printf(__VA_ARGS__); \
475 } while(0)
476
477 #define PCI_CHIP_845_G 0x2562
478 #define PCI_CHIP_I830_M 0x3577
479 #define PCI_CHIP_I855_GM 0x3582
480 #define PCI_CHIP_I865_G 0x2572
481 #define PCI_CHIP_I915_G 0x2582
482 #define PCI_CHIP_I915_GM 0x2592
483 #define PCI_CHIP_I945_G 0x2772
484 #define PCI_CHIP_I945_GM 0x27A2
485 #define PCI_CHIP_I945_GME 0x27AE
486 #define PCI_CHIP_G33_G 0x29C2
487 #define PCI_CHIP_Q35_G 0x29B2
488 #define PCI_CHIP_Q33_G 0x29D2
489
490
491 /* ================================================================
492 * intel_context.c:
493 */
494
495 extern bool intelInitContext(struct intel_context *intel,
496 int api,
497 const struct gl_config * mesaVis,
498 __DRIcontext * driContextPriv,
499 void *sharedContextPrivate,
500 struct dd_function_table *functions);
501
502 extern void intelFinish(struct gl_context * ctx);
503 extern void intel_flush_rendering_to_batch(struct gl_context *ctx);
504 extern void intel_flush(struct gl_context * ctx);
505
506 extern void intelInitDriverFunctions(struct dd_function_table *functions);
507
508 void intel_init_syncobj_functions(struct dd_function_table *functions);
509
510
511 /* ================================================================
512 * intel_state.c:
513 */
514 extern void intelInitStateFuncs(struct dd_function_table *functions);
515
516 #define COMPAREFUNC_ALWAYS 0
517 #define COMPAREFUNC_NEVER 0x1
518 #define COMPAREFUNC_LESS 0x2
519 #define COMPAREFUNC_EQUAL 0x3
520 #define COMPAREFUNC_LEQUAL 0x4
521 #define COMPAREFUNC_GREATER 0x5
522 #define COMPAREFUNC_NOTEQUAL 0x6
523 #define COMPAREFUNC_GEQUAL 0x7
524
525 #define STENCILOP_KEEP 0
526 #define STENCILOP_ZERO 0x1
527 #define STENCILOP_REPLACE 0x2
528 #define STENCILOP_INCRSAT 0x3
529 #define STENCILOP_DECRSAT 0x4
530 #define STENCILOP_INCR 0x5
531 #define STENCILOP_DECR 0x6
532 #define STENCILOP_INVERT 0x7
533
534 #define LOGICOP_CLEAR 0
535 #define LOGICOP_NOR 0x1
536 #define LOGICOP_AND_INV 0x2
537 #define LOGICOP_COPY_INV 0x3
538 #define LOGICOP_AND_RVRSE 0x4
539 #define LOGICOP_INV 0x5
540 #define LOGICOP_XOR 0x6
541 #define LOGICOP_NAND 0x7
542 #define LOGICOP_AND 0x8
543 #define LOGICOP_EQUIV 0x9
544 #define LOGICOP_NOOP 0xa
545 #define LOGICOP_OR_INV 0xb
546 #define LOGICOP_COPY 0xc
547 #define LOGICOP_OR_RVRSE 0xd
548 #define LOGICOP_OR 0xe
549 #define LOGICOP_SET 0xf
550
551 #define BLENDFACT_ZERO 0x01
552 #define BLENDFACT_ONE 0x02
553 #define BLENDFACT_SRC_COLR 0x03
554 #define BLENDFACT_INV_SRC_COLR 0x04
555 #define BLENDFACT_SRC_ALPHA 0x05
556 #define BLENDFACT_INV_SRC_ALPHA 0x06
557 #define BLENDFACT_DST_ALPHA 0x07
558 #define BLENDFACT_INV_DST_ALPHA 0x08
559 #define BLENDFACT_DST_COLR 0x09
560 #define BLENDFACT_INV_DST_COLR 0x0a
561 #define BLENDFACT_SRC_ALPHA_SATURATE 0x0b
562 #define BLENDFACT_CONST_COLOR 0x0c
563 #define BLENDFACT_INV_CONST_COLOR 0x0d
564 #define BLENDFACT_CONST_ALPHA 0x0e
565 #define BLENDFACT_INV_CONST_ALPHA 0x0f
566 #define BLENDFACT_MASK 0x0f
567
568 enum {
569 DRI_CONF_BO_REUSE_DISABLED,
570 DRI_CONF_BO_REUSE_ALL
571 };
572
573 extern int intel_translate_shadow_compare_func(GLenum func);
574 extern int intel_translate_compare_func(GLenum func);
575 extern int intel_translate_stencil_op(GLenum op);
576 extern int intel_translate_blend_factor(GLenum factor);
577 extern int intel_translate_logic_op(GLenum opcode);
578
579 void intel_update_renderbuffers(__DRIcontext *context,
580 __DRIdrawable *drawable);
581 void intel_prepare_render(struct intel_context *intel);
582
583 void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
584 uint32_t buffer_id);
585
586 /*======================================================================
587 * Inline conversion functions.
588 * These are better-typed than the macros used previously:
589 */
590 static INLINE struct intel_context *
591 intel_context(struct gl_context * ctx)
592 {
593 return (struct intel_context *) ctx;
594 }
595
596 static INLINE bool
597 is_power_of_two(uint32_t value)
598 {
599 return (value & (value - 1)) == 0;
600 }
601
602 #endif