i915: Remove most of the code under gen >= 4 checks.
[mesa.git] / src / mesa / drivers / dri / i965 / 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 <string.h>
34 #include "main/mtypes.h"
35 #include "main/mm.h"
36
37 #ifdef __cplusplus
38 extern "C" {
39 /* Evil hack for using libdrm in a c++ compiler. */
40 #define virtual virt
41 #endif
42
43 #include "drm.h"
44 #include "intel_bufmgr.h"
45
46 #include "intel_screen.h"
47 #include "intel_tex_obj.h"
48 #include "i915_drm.h"
49
50 #ifdef __cplusplus
51 #undef virtual
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 struct intel_batchbuffer {
120 /** Current batchbuffer being queued up. */
121 drm_intel_bo *bo;
122 /** Last BO submitted to the hardware. Used for glFinish(). */
123 drm_intel_bo *last_bo;
124 /** BO for post-sync nonzero writes for gen6 workaround. */
125 drm_intel_bo *workaround_bo;
126 bool need_workaround_flush;
127
128 struct cached_batch_item *cached_items;
129
130 uint16_t emit, total;
131 uint16_t used, reserved_space;
132 uint32_t *map;
133 uint32_t *cpu_map;
134 #define BATCH_SZ (8192*sizeof(uint32_t))
135
136 uint32_t state_batch_offset;
137 bool is_blit;
138 bool needs_sol_reset;
139
140 struct {
141 uint16_t used;
142 int reloc_count;
143 } saved;
144 };
145
146 /**
147 * intel_context is derived from Mesa's context class: struct gl_context.
148 */
149 struct intel_context
150 {
151 struct gl_context ctx; /**< base class, must be first field */
152
153 struct
154 {
155 void (*destroy) (struct intel_context * intel);
156 void (*emit_state) (struct intel_context * intel);
157 void (*finish_batch) (struct intel_context * intel);
158 void (*new_batch) (struct intel_context * intel);
159 void (*emit_invarient_state) (struct intel_context * intel);
160 void (*update_texture_state) (struct intel_context * intel);
161
162 void (*render_start) (struct intel_context * intel);
163 void (*render_prevalidate) (struct intel_context * intel);
164 void (*set_draw_region) (struct intel_context * intel,
165 struct intel_region * draw_regions[],
166 struct intel_region * depth_region,
167 GLuint num_regions);
168 void (*update_draw_buffer)(struct intel_context *intel);
169
170 void (*reduced_primitive_state) (struct intel_context * intel,
171 GLenum rprim);
172
173 bool (*check_vertex_size) (struct intel_context * intel,
174 GLuint expected);
175 void (*invalidate_state) (struct intel_context *intel,
176 GLuint new_state);
177
178 void (*assert_not_dirty) (struct intel_context *intel);
179
180 void (*debug_batch)(struct intel_context *intel);
181 void (*annotate_aub)(struct intel_context *intel);
182 bool (*render_target_supported)(struct intel_context *intel,
183 struct gl_renderbuffer *rb);
184
185 /** Can HiZ be enabled on a depthbuffer of the given format? */
186 bool (*is_hiz_depth_format)(struct intel_context *intel,
187 gl_format format);
188
189 /**
190 * Surface state operations (i965+ only)
191 * \{
192 */
193 void (*update_texture_surface)(struct gl_context *ctx,
194 unsigned unit,
195 uint32_t *binding_table,
196 unsigned surf_index);
197 void (*update_renderbuffer_surface)(struct brw_context *brw,
198 struct gl_renderbuffer *rb,
199 bool layered,
200 unsigned unit);
201 void (*update_null_renderbuffer_surface)(struct brw_context *brw,
202 unsigned unit);
203 void (*create_constant_surface)(struct brw_context *brw,
204 drm_intel_bo *bo,
205 uint32_t offset,
206 uint32_t size,
207 uint32_t *out_offset,
208 bool dword_pitch);
209 /** \} */
210
211 /**
212 * Send the appropriate state packets to configure depth, stencil, and
213 * HiZ buffers (i965+ only)
214 */
215 void (*emit_depth_stencil_hiz)(struct brw_context *brw,
216 struct intel_mipmap_tree *depth_mt,
217 uint32_t depth_offset,
218 uint32_t depthbuffer_format,
219 uint32_t depth_surface_type,
220 struct intel_mipmap_tree *stencil_mt,
221 bool hiz, bool separate_stencil,
222 uint32_t width, uint32_t height,
223 uint32_t tile_x, uint32_t tile_y);
224
225 } vtbl;
226
227 GLbitfield Fallback; /**< mask of INTEL_FALLBACK_x bits */
228 GLuint NewGLState;
229
230 dri_bufmgr *bufmgr;
231 unsigned int maxBatchSize;
232
233 /**
234 * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965.
235 */
236 int gen;
237 int gt;
238 bool needs_ff_sync;
239 bool is_haswell;
240 bool is_baytrail;
241 bool is_g4x;
242 bool is_945;
243 bool has_separate_stencil;
244 bool must_use_separate_stencil;
245 bool has_hiz;
246 bool has_llc;
247 bool has_swizzling;
248
249 int urb_size;
250
251 drm_intel_context *hw_ctx;
252
253 struct intel_batchbuffer batch;
254
255 drm_intel_bo *first_post_swapbuffers_batch;
256 bool need_throttle;
257 bool no_batch_wrap;
258 bool tnl_pipeline_running; /**< Set while i915's _tnl_run_pipeline. */
259
260 /**
261 * Set if we're either a debug context or the INTEL_DEBUG=perf environment
262 * variable is set, this is the flag indicating to do expensive work that
263 * might lead to a perf_debug() call.
264 */
265 bool perf_debug;
266
267 struct
268 {
269 GLuint id;
270 uint32_t start_ptr; /**< for i8xx */
271 uint32_t primitive; /**< Current hardware primitive type */
272 void (*flush) (struct intel_context *);
273 drm_intel_bo *vb_bo;
274 uint8_t *vb;
275 unsigned int start_offset; /**< Byte offset of primitive sequence */
276 unsigned int current_offset; /**< Byte offset of next vertex */
277 unsigned int count; /**< Number of vertices in current primitive */
278 } prim;
279
280 struct {
281 drm_intel_bo *bo;
282 GLuint offset;
283 uint32_t buffer_len;
284 uint32_t buffer_offset;
285 char buffer[4096];
286 } upload;
287
288 uint32_t max_gtt_map_object_size;
289
290 GLuint stats_wm;
291
292 /* Offsets of fields within the current vertex:
293 */
294 GLuint coloroffset;
295 GLuint specoffset;
296 GLuint wpos_offset;
297
298 struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
299 GLuint vertex_attr_count;
300
301 bool hw_stencil;
302 bool hw_stipple;
303 bool no_rast;
304 bool always_flush_batch;
305 bool always_flush_cache;
306 bool disable_throttling;
307
308 /* State for intelvb.c and inteltris.c.
309 */
310 GLuint RenderIndex;
311 GLmatrix ViewportMatrix;
312 GLenum render_primitive;
313 GLenum reduced_primitive; /*< Only gen < 6 */
314 GLuint vertex_size;
315 GLubyte *verts; /* points to tnl->clipspace.vertex_buf */
316
317 /* Fallback rasterization functions
318 */
319 intel_point_func draw_point;
320 intel_line_func draw_line;
321 intel_tri_func draw_tri;
322
323 /**
324 * Set if rendering has occured to the drawable's front buffer.
325 *
326 * This is used in the DRI2 case to detect that glFlush should also copy
327 * the contents of the fake front buffer to the real front buffer.
328 */
329 bool front_buffer_dirty;
330
331 /**
332 * Track whether front-buffer rendering is currently enabled
333 *
334 * A separate flag is used to track this in order to support MRT more
335 * easily.
336 */
337 bool is_front_buffer_rendering;
338 /**
339 * Track whether front-buffer is the current read target.
340 *
341 * This is closely associated with is_front_buffer_rendering, but may
342 * be set separately. The DRI2 fake front buffer must be referenced
343 * either way.
344 */
345 bool is_front_buffer_reading;
346
347 bool use_early_z;
348
349 int driFd;
350
351 __DRIcontext *driContext;
352 struct intel_screen *intelScreen;
353 void (*saved_viewport)(struct gl_context * ctx,
354 GLint x, GLint y, GLsizei width, GLsizei height);
355
356 /**
357 * Configuration cache
358 */
359 driOptionCache optionCache;
360 };
361
362 #define SUBPIXEL_X 0.125
363 #define SUBPIXEL_Y 0.125
364
365 /**
366 * Align a value down to an alignment value
367 *
368 * If \c value is not already aligned to the requested alignment value, it
369 * will be rounded down.
370 *
371 * \param value Value to be rounded
372 * \param alignment Alignment value to be used. This must be a power of two.
373 *
374 * \sa ALIGN()
375 */
376 #define ROUND_DOWN_TO(value, alignment) ((value) & ~(alignment - 1))
377
378 static INLINE uint32_t
379 U_FIXED(float value, uint32_t frac_bits)
380 {
381 value *= (1 << frac_bits);
382 return value < 0 ? 0 : value;
383 }
384
385 static INLINE uint32_t
386 S_FIXED(float value, uint32_t frac_bits)
387 {
388 return value * (1 << frac_bits);
389 }
390
391 #define INTEL_FIREVERTICES(intel) \
392 do { \
393 if ((intel)->prim.flush) \
394 (intel)->prim.flush(intel); \
395 } while (0)
396
397 /* ================================================================
398 * From linux kernel i386 header files, copes with odd sizes better
399 * than COPY_DWORDS would:
400 * XXX Put this in src/mesa/main/imports.h ???
401 */
402 #if defined(i386) || defined(__i386__)
403 static INLINE void * __memcpy(void * to, const void * from, size_t n)
404 {
405 int d0, d1, d2;
406 __asm__ __volatile__(
407 "rep ; movsl\n\t"
408 "testb $2,%b4\n\t"
409 "je 1f\n\t"
410 "movsw\n"
411 "1:\ttestb $1,%b4\n\t"
412 "je 2f\n\t"
413 "movsb\n"
414 "2:"
415 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
416 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
417 : "memory");
418 return (to);
419 }
420 #else
421 #define __memcpy(a,b,c) memcpy(a,b,c)
422 #endif
423
424
425 /* ================================================================
426 * Debugging:
427 */
428 extern int INTEL_DEBUG;
429
430 #define DEBUG_TEXTURE 0x1
431 #define DEBUG_STATE 0x2
432 #define DEBUG_IOCTL 0x4
433 #define DEBUG_BLIT 0x8
434 #define DEBUG_MIPTREE 0x10
435 #define DEBUG_PERF 0x20
436 #define DEBUG_BATCH 0x80
437 #define DEBUG_PIXEL 0x100
438 #define DEBUG_BUFMGR 0x200
439 #define DEBUG_REGION 0x400
440 #define DEBUG_FBO 0x800
441 #define DEBUG_GS 0x1000
442 #define DEBUG_SYNC 0x2000
443 #define DEBUG_PRIMS 0x4000
444 #define DEBUG_VERTS 0x8000
445 #define DEBUG_DRI 0x10000
446 #define DEBUG_SF 0x20000
447 #define DEBUG_STATS 0x100000
448 #define DEBUG_WM 0x400000
449 #define DEBUG_URB 0x800000
450 #define DEBUG_VS 0x1000000
451 #define DEBUG_CLIP 0x2000000
452 #define DEBUG_AUB 0x4000000
453 #define DEBUG_SHADER_TIME 0x8000000
454 #define DEBUG_BLORP 0x10000000
455 #define DEBUG_NO16 0x20000000
456
457 #ifdef HAVE_ANDROID_PLATFORM
458 #define LOG_TAG "INTEL-MESA"
459 #include <cutils/log.h>
460 #ifndef ALOGW
461 #define ALOGW LOGW
462 #endif
463 #define dbg_printf(...) ALOGW(__VA_ARGS__)
464 #else
465 #define dbg_printf(...) printf(__VA_ARGS__)
466 #endif /* HAVE_ANDROID_PLATFORM */
467
468 #define DBG(...) do { \
469 if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG)) \
470 dbg_printf(__VA_ARGS__); \
471 } while(0)
472
473 #define perf_debug(...) do { \
474 static GLuint msg_id = 0; \
475 if (unlikely(INTEL_DEBUG & DEBUG_PERF)) \
476 dbg_printf(__VA_ARGS__); \
477 if (intel->perf_debug) \
478 _mesa_gl_debug(&intel->ctx, &msg_id, \
479 MESA_DEBUG_TYPE_PERFORMANCE, \
480 MESA_DEBUG_SEVERITY_MEDIUM, \
481 __VA_ARGS__); \
482 } while(0)
483
484 #define WARN_ONCE(cond, fmt...) do { \
485 if (unlikely(cond)) { \
486 static bool _warned = false; \
487 static GLuint msg_id = 0; \
488 if (!_warned) { \
489 fprintf(stderr, "WARNING: "); \
490 fprintf(stderr, fmt); \
491 _warned = true; \
492 \
493 _mesa_gl_debug(ctx, &msg_id, \
494 MESA_DEBUG_TYPE_OTHER, \
495 MESA_DEBUG_SEVERITY_HIGH, fmt); \
496 } \
497 } \
498 } while (0)
499
500 #define PCI_CHIP_845_G 0x2562
501 #define PCI_CHIP_I830_M 0x3577
502 #define PCI_CHIP_I855_GM 0x3582
503 #define PCI_CHIP_I865_G 0x2572
504 #define PCI_CHIP_I915_G 0x2582
505 #define PCI_CHIP_I915_GM 0x2592
506 #define PCI_CHIP_I945_G 0x2772
507 #define PCI_CHIP_I945_GM 0x27A2
508 #define PCI_CHIP_I945_GME 0x27AE
509 #define PCI_CHIP_G33_G 0x29C2
510 #define PCI_CHIP_Q35_G 0x29B2
511 #define PCI_CHIP_Q33_G 0x29D2
512
513
514 /* ================================================================
515 * intel_context.c:
516 */
517
518 extern bool intelInitContext(struct intel_context *intel,
519 int api,
520 unsigned major_version,
521 unsigned minor_version,
522 const struct gl_config * mesaVis,
523 __DRIcontext * driContextPriv,
524 void *sharedContextPrivate,
525 struct dd_function_table *functions,
526 unsigned *dri_ctx_error);
527
528 extern void intelFinish(struct gl_context * ctx);
529 extern void intel_flush_rendering_to_batch(struct gl_context *ctx);
530 extern void _intel_flush(struct gl_context * ctx, const char *file, int line);
531
532 #define intel_flush(ctx) _intel_flush(ctx, __FILE__, __LINE__)
533
534 extern void intelInitDriverFunctions(struct dd_function_table *functions);
535
536 void intel_init_syncobj_functions(struct dd_function_table *functions);
537
538
539 /* ================================================================
540 * intel_state.c:
541 */
542
543 #define COMPAREFUNC_ALWAYS 0
544 #define COMPAREFUNC_NEVER 0x1
545 #define COMPAREFUNC_LESS 0x2
546 #define COMPAREFUNC_EQUAL 0x3
547 #define COMPAREFUNC_LEQUAL 0x4
548 #define COMPAREFUNC_GREATER 0x5
549 #define COMPAREFUNC_NOTEQUAL 0x6
550 #define COMPAREFUNC_GEQUAL 0x7
551
552 #define STENCILOP_KEEP 0
553 #define STENCILOP_ZERO 0x1
554 #define STENCILOP_REPLACE 0x2
555 #define STENCILOP_INCRSAT 0x3
556 #define STENCILOP_DECRSAT 0x4
557 #define STENCILOP_INCR 0x5
558 #define STENCILOP_DECR 0x6
559 #define STENCILOP_INVERT 0x7
560
561 #define LOGICOP_CLEAR 0
562 #define LOGICOP_NOR 0x1
563 #define LOGICOP_AND_INV 0x2
564 #define LOGICOP_COPY_INV 0x3
565 #define LOGICOP_AND_RVRSE 0x4
566 #define LOGICOP_INV 0x5
567 #define LOGICOP_XOR 0x6
568 #define LOGICOP_NAND 0x7
569 #define LOGICOP_AND 0x8
570 #define LOGICOP_EQUIV 0x9
571 #define LOGICOP_NOOP 0xa
572 #define LOGICOP_OR_INV 0xb
573 #define LOGICOP_COPY 0xc
574 #define LOGICOP_OR_RVRSE 0xd
575 #define LOGICOP_OR 0xe
576 #define LOGICOP_SET 0xf
577
578 #define BLENDFACT_ZERO 0x01
579 #define BLENDFACT_ONE 0x02
580 #define BLENDFACT_SRC_COLR 0x03
581 #define BLENDFACT_INV_SRC_COLR 0x04
582 #define BLENDFACT_SRC_ALPHA 0x05
583 #define BLENDFACT_INV_SRC_ALPHA 0x06
584 #define BLENDFACT_DST_ALPHA 0x07
585 #define BLENDFACT_INV_DST_ALPHA 0x08
586 #define BLENDFACT_DST_COLR 0x09
587 #define BLENDFACT_INV_DST_COLR 0x0a
588 #define BLENDFACT_SRC_ALPHA_SATURATE 0x0b
589 #define BLENDFACT_CONST_COLOR 0x0c
590 #define BLENDFACT_INV_CONST_COLOR 0x0d
591 #define BLENDFACT_CONST_ALPHA 0x0e
592 #define BLENDFACT_INV_CONST_ALPHA 0x0f
593 #define BLENDFACT_MASK 0x0f
594
595 enum {
596 DRI_CONF_BO_REUSE_DISABLED,
597 DRI_CONF_BO_REUSE_ALL
598 };
599
600 extern int intel_translate_shadow_compare_func(GLenum func);
601 extern int intel_translate_compare_func(GLenum func);
602 extern int intel_translate_stencil_op(GLenum op);
603 extern int intel_translate_blend_factor(GLenum factor);
604 extern int intel_translate_logic_op(GLenum opcode);
605
606 void intel_update_renderbuffers(__DRIcontext *context,
607 __DRIdrawable *drawable);
608 void intel_prepare_render(struct intel_context *intel);
609
610 void
611 intel_resolve_for_dri2_flush(struct intel_context *intel,
612 __DRIdrawable *drawable);
613
614 void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
615 uint32_t buffer_id);
616 void intel_init_texture_formats(struct gl_context *ctx);
617
618 /*======================================================================
619 * Inline conversion functions.
620 * These are better-typed than the macros used previously:
621 */
622 static INLINE struct intel_context *
623 intel_context(struct gl_context * ctx)
624 {
625 return (struct intel_context *) ctx;
626 }
627
628 static INLINE bool
629 is_power_of_two(uint32_t value)
630 {
631 return (value & (value - 1)) == 0;
632 }
633
634 #ifdef __cplusplus
635 }
636 #endif
637
638 #endif