i965: Drop dead i915 blend state code.
[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 #define INTEL_WRITE_PART 0x1
76 #define INTEL_WRITE_FULL 0x2
77 #define INTEL_READ 0x4
78
79 #define INTEL_MAX_FIXUP 64
80
81 #ifndef likely
82 #ifdef __GNUC__
83 #define likely(expr) (__builtin_expect(expr, 1))
84 #define unlikely(expr) (__builtin_expect(expr, 0))
85 #else
86 #define likely(expr) (expr)
87 #define unlikely(expr) (expr)
88 #endif
89 #endif
90
91 struct intel_sync_object {
92 struct gl_sync_object Base;
93
94 /** Batch associated with this sync object */
95 drm_intel_bo *bo;
96 };
97
98 struct brw_context;
99
100 struct intel_batchbuffer {
101 /** Current batchbuffer being queued up. */
102 drm_intel_bo *bo;
103 /** Last BO submitted to the hardware. Used for glFinish(). */
104 drm_intel_bo *last_bo;
105 /** BO for post-sync nonzero writes for gen6 workaround. */
106 drm_intel_bo *workaround_bo;
107 bool need_workaround_flush;
108
109 struct cached_batch_item *cached_items;
110
111 uint16_t emit, total;
112 uint16_t used, reserved_space;
113 uint32_t *map;
114 uint32_t *cpu_map;
115 #define BATCH_SZ (8192*sizeof(uint32_t))
116
117 uint32_t state_batch_offset;
118 bool is_blit;
119 bool needs_sol_reset;
120
121 struct {
122 uint16_t used;
123 int reloc_count;
124 } saved;
125 };
126
127 /**
128 * intel_context is derived from Mesa's context class: struct gl_context.
129 */
130 struct intel_context
131 {
132 struct gl_context ctx; /**< base class, must be first field */
133
134 struct
135 {
136 void (*destroy) (struct intel_context * intel);
137 void (*finish_batch) (struct intel_context * intel);
138 void (*new_batch) (struct intel_context * intel);
139
140 void (*update_draw_buffer)(struct intel_context *intel);
141
142 void (*invalidate_state) (struct intel_context *intel,
143 GLuint new_state);
144
145 void (*debug_batch)(struct intel_context *intel);
146 void (*annotate_aub)(struct intel_context *intel);
147 bool (*render_target_supported)(struct intel_context *intel,
148 struct gl_renderbuffer *rb);
149
150 /** Can HiZ be enabled on a depthbuffer of the given format? */
151 bool (*is_hiz_depth_format)(struct intel_context *intel,
152 gl_format format);
153
154 void (*update_texture_surface)(struct gl_context *ctx,
155 unsigned unit,
156 uint32_t *binding_table,
157 unsigned surf_index);
158 void (*update_renderbuffer_surface)(struct brw_context *brw,
159 struct gl_renderbuffer *rb,
160 bool layered,
161 unsigned unit);
162 void (*update_null_renderbuffer_surface)(struct brw_context *brw,
163 unsigned unit);
164 void (*create_constant_surface)(struct brw_context *brw,
165 drm_intel_bo *bo,
166 uint32_t offset,
167 uint32_t size,
168 uint32_t *out_offset,
169 bool dword_pitch);
170
171 /**
172 * Send the appropriate state packets to configure depth, stencil, and
173 * HiZ buffers (i965+ only)
174 */
175 void (*emit_depth_stencil_hiz)(struct brw_context *brw,
176 struct intel_mipmap_tree *depth_mt,
177 uint32_t depth_offset,
178 uint32_t depthbuffer_format,
179 uint32_t depth_surface_type,
180 struct intel_mipmap_tree *stencil_mt,
181 bool hiz, bool separate_stencil,
182 uint32_t width, uint32_t height,
183 uint32_t tile_x, uint32_t tile_y);
184
185 } vtbl;
186
187 GLuint NewGLState;
188
189 dri_bufmgr *bufmgr;
190 unsigned int maxBatchSize;
191
192 /**
193 * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965.
194 */
195 int gen;
196 int gt;
197 bool needs_ff_sync;
198 bool is_haswell;
199 bool is_baytrail;
200 bool is_g4x;
201 bool is_945;
202 bool has_separate_stencil;
203 bool must_use_separate_stencil;
204 bool has_hiz;
205 bool has_llc;
206 bool has_swizzling;
207
208 int urb_size;
209
210 drm_intel_context *hw_ctx;
211
212 struct intel_batchbuffer batch;
213
214 drm_intel_bo *first_post_swapbuffers_batch;
215 bool need_throttle;
216 bool no_batch_wrap;
217
218 /**
219 * Set if we're either a debug context or the INTEL_DEBUG=perf environment
220 * variable is set, this is the flag indicating to do expensive work that
221 * might lead to a perf_debug() call.
222 */
223 bool perf_debug;
224
225 struct {
226 drm_intel_bo *bo;
227 GLuint offset;
228 uint32_t buffer_len;
229 uint32_t buffer_offset;
230 char buffer[4096];
231 } upload;
232
233 uint32_t max_gtt_map_object_size;
234
235 GLuint stats_wm;
236
237 bool hw_stencil;
238 bool hw_stipple;
239 bool no_rast;
240 bool always_flush_batch;
241 bool always_flush_cache;
242 bool disable_throttling;
243
244 GLenum reduced_primitive;
245
246 /**
247 * Set if rendering has occured to the drawable's front buffer.
248 *
249 * This is used in the DRI2 case to detect that glFlush should also copy
250 * the contents of the fake front buffer to the real front buffer.
251 */
252 bool front_buffer_dirty;
253
254 /**
255 * Track whether front-buffer rendering is currently enabled
256 *
257 * A separate flag is used to track this in order to support MRT more
258 * easily.
259 */
260 bool is_front_buffer_rendering;
261 /**
262 * Track whether front-buffer is the current read target.
263 *
264 * This is closely associated with is_front_buffer_rendering, but may
265 * be set separately. The DRI2 fake front buffer must be referenced
266 * either way.
267 */
268 bool is_front_buffer_reading;
269
270 bool use_early_z;
271
272 int driFd;
273
274 __DRIcontext *driContext;
275 struct intel_screen *intelScreen;
276 void (*saved_viewport)(struct gl_context * ctx,
277 GLint x, GLint y, GLsizei width, GLsizei height);
278
279 /**
280 * Configuration cache
281 */
282 driOptionCache optionCache;
283 };
284
285 #define SUBPIXEL_X 0.125
286 #define SUBPIXEL_Y 0.125
287
288 /**
289 * Align a value down to an alignment value
290 *
291 * If \c value is not already aligned to the requested alignment value, it
292 * will be rounded down.
293 *
294 * \param value Value to be rounded
295 * \param alignment Alignment value to be used. This must be a power of two.
296 *
297 * \sa ALIGN()
298 */
299 #define ROUND_DOWN_TO(value, alignment) ((value) & ~(alignment - 1))
300
301 static INLINE uint32_t
302 U_FIXED(float value, uint32_t frac_bits)
303 {
304 value *= (1 << frac_bits);
305 return value < 0 ? 0 : value;
306 }
307
308 static INLINE uint32_t
309 S_FIXED(float value, uint32_t frac_bits)
310 {
311 return value * (1 << frac_bits);
312 }
313
314 /* ================================================================
315 * From linux kernel i386 header files, copes with odd sizes better
316 * than COPY_DWORDS would:
317 * XXX Put this in src/mesa/main/imports.h ???
318 */
319 #if defined(i386) || defined(__i386__)
320 static INLINE void * __memcpy(void * to, const void * from, size_t n)
321 {
322 int d0, d1, d2;
323 __asm__ __volatile__(
324 "rep ; movsl\n\t"
325 "testb $2,%b4\n\t"
326 "je 1f\n\t"
327 "movsw\n"
328 "1:\ttestb $1,%b4\n\t"
329 "je 2f\n\t"
330 "movsb\n"
331 "2:"
332 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
333 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
334 : "memory");
335 return (to);
336 }
337 #else
338 #define __memcpy(a,b,c) memcpy(a,b,c)
339 #endif
340
341
342 /* ================================================================
343 * Debugging:
344 */
345 extern int INTEL_DEBUG;
346
347 #define DEBUG_TEXTURE 0x1
348 #define DEBUG_STATE 0x2
349 #define DEBUG_IOCTL 0x4
350 #define DEBUG_BLIT 0x8
351 #define DEBUG_MIPTREE 0x10
352 #define DEBUG_PERF 0x20
353 #define DEBUG_BATCH 0x80
354 #define DEBUG_PIXEL 0x100
355 #define DEBUG_BUFMGR 0x200
356 #define DEBUG_REGION 0x400
357 #define DEBUG_FBO 0x800
358 #define DEBUG_GS 0x1000
359 #define DEBUG_SYNC 0x2000
360 #define DEBUG_PRIMS 0x4000
361 #define DEBUG_VERTS 0x8000
362 #define DEBUG_DRI 0x10000
363 #define DEBUG_SF 0x20000
364 #define DEBUG_STATS 0x100000
365 #define DEBUG_WM 0x400000
366 #define DEBUG_URB 0x800000
367 #define DEBUG_VS 0x1000000
368 #define DEBUG_CLIP 0x2000000
369 #define DEBUG_AUB 0x4000000
370 #define DEBUG_SHADER_TIME 0x8000000
371 #define DEBUG_BLORP 0x10000000
372 #define DEBUG_NO16 0x20000000
373
374 #ifdef HAVE_ANDROID_PLATFORM
375 #define LOG_TAG "INTEL-MESA"
376 #include <cutils/log.h>
377 #ifndef ALOGW
378 #define ALOGW LOGW
379 #endif
380 #define dbg_printf(...) ALOGW(__VA_ARGS__)
381 #else
382 #define dbg_printf(...) printf(__VA_ARGS__)
383 #endif /* HAVE_ANDROID_PLATFORM */
384
385 #define DBG(...) do { \
386 if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG)) \
387 dbg_printf(__VA_ARGS__); \
388 } while(0)
389
390 #define perf_debug(...) do { \
391 static GLuint msg_id = 0; \
392 if (unlikely(INTEL_DEBUG & DEBUG_PERF)) \
393 dbg_printf(__VA_ARGS__); \
394 if (intel->perf_debug) \
395 _mesa_gl_debug(&intel->ctx, &msg_id, \
396 MESA_DEBUG_TYPE_PERFORMANCE, \
397 MESA_DEBUG_SEVERITY_MEDIUM, \
398 __VA_ARGS__); \
399 } while(0)
400
401 #define WARN_ONCE(cond, fmt...) do { \
402 if (unlikely(cond)) { \
403 static bool _warned = false; \
404 static GLuint msg_id = 0; \
405 if (!_warned) { \
406 fprintf(stderr, "WARNING: "); \
407 fprintf(stderr, fmt); \
408 _warned = true; \
409 \
410 _mesa_gl_debug(ctx, &msg_id, \
411 MESA_DEBUG_TYPE_OTHER, \
412 MESA_DEBUG_SEVERITY_HIGH, fmt); \
413 } \
414 } \
415 } while (0)
416
417 #define PCI_CHIP_845_G 0x2562
418 #define PCI_CHIP_I830_M 0x3577
419 #define PCI_CHIP_I855_GM 0x3582
420 #define PCI_CHIP_I865_G 0x2572
421 #define PCI_CHIP_I915_G 0x2582
422 #define PCI_CHIP_I915_GM 0x2592
423 #define PCI_CHIP_I945_G 0x2772
424 #define PCI_CHIP_I945_GM 0x27A2
425 #define PCI_CHIP_I945_GME 0x27AE
426 #define PCI_CHIP_G33_G 0x29C2
427 #define PCI_CHIP_Q35_G 0x29B2
428 #define PCI_CHIP_Q33_G 0x29D2
429
430
431 /* ================================================================
432 * intel_context.c:
433 */
434
435 extern bool intelInitContext(struct intel_context *intel,
436 int api,
437 unsigned major_version,
438 unsigned minor_version,
439 const struct gl_config * mesaVis,
440 __DRIcontext * driContextPriv,
441 void *sharedContextPrivate,
442 struct dd_function_table *functions,
443 unsigned *dri_ctx_error);
444
445 extern void intelFinish(struct gl_context * ctx);
446 extern void _intel_flush(struct gl_context * ctx, const char *file, int line);
447
448 #define intel_flush(ctx) _intel_flush(ctx, __FILE__, __LINE__)
449
450 extern void intelInitDriverFunctions(struct dd_function_table *functions);
451
452 void intel_init_syncobj_functions(struct dd_function_table *functions);
453
454
455 /* ================================================================
456 * intel_state.c:
457 */
458
459 #define COMPAREFUNC_ALWAYS 0
460 #define COMPAREFUNC_NEVER 0x1
461 #define COMPAREFUNC_LESS 0x2
462 #define COMPAREFUNC_EQUAL 0x3
463 #define COMPAREFUNC_LEQUAL 0x4
464 #define COMPAREFUNC_GREATER 0x5
465 #define COMPAREFUNC_NOTEQUAL 0x6
466 #define COMPAREFUNC_GEQUAL 0x7
467
468 #define STENCILOP_KEEP 0
469 #define STENCILOP_ZERO 0x1
470 #define STENCILOP_REPLACE 0x2
471 #define STENCILOP_INCRSAT 0x3
472 #define STENCILOP_DECRSAT 0x4
473 #define STENCILOP_INCR 0x5
474 #define STENCILOP_DECR 0x6
475 #define STENCILOP_INVERT 0x7
476
477 #define LOGICOP_CLEAR 0
478 #define LOGICOP_NOR 0x1
479 #define LOGICOP_AND_INV 0x2
480 #define LOGICOP_COPY_INV 0x3
481 #define LOGICOP_AND_RVRSE 0x4
482 #define LOGICOP_INV 0x5
483 #define LOGICOP_XOR 0x6
484 #define LOGICOP_NAND 0x7
485 #define LOGICOP_AND 0x8
486 #define LOGICOP_EQUIV 0x9
487 #define LOGICOP_NOOP 0xa
488 #define LOGICOP_OR_INV 0xb
489 #define LOGICOP_COPY 0xc
490 #define LOGICOP_OR_RVRSE 0xd
491 #define LOGICOP_OR 0xe
492 #define LOGICOP_SET 0xf
493
494 #define BLENDFACT_ZERO 0x01
495 #define BLENDFACT_ONE 0x02
496 #define BLENDFACT_SRC_COLR 0x03
497 #define BLENDFACT_INV_SRC_COLR 0x04
498 #define BLENDFACT_SRC_ALPHA 0x05
499 #define BLENDFACT_INV_SRC_ALPHA 0x06
500 #define BLENDFACT_DST_ALPHA 0x07
501 #define BLENDFACT_INV_DST_ALPHA 0x08
502 #define BLENDFACT_DST_COLR 0x09
503 #define BLENDFACT_INV_DST_COLR 0x0a
504 #define BLENDFACT_SRC_ALPHA_SATURATE 0x0b
505 #define BLENDFACT_CONST_COLOR 0x0c
506 #define BLENDFACT_INV_CONST_COLOR 0x0d
507 #define BLENDFACT_CONST_ALPHA 0x0e
508 #define BLENDFACT_INV_CONST_ALPHA 0x0f
509 #define BLENDFACT_MASK 0x0f
510
511 enum {
512 DRI_CONF_BO_REUSE_DISABLED,
513 DRI_CONF_BO_REUSE_ALL
514 };
515
516 extern int intel_translate_shadow_compare_func(GLenum func);
517 extern int intel_translate_compare_func(GLenum func);
518 extern int intel_translate_stencil_op(GLenum op);
519 extern int intel_translate_logic_op(GLenum opcode);
520
521 void intel_update_renderbuffers(__DRIcontext *context,
522 __DRIdrawable *drawable);
523 void intel_prepare_render(struct intel_context *intel);
524
525 void
526 intel_resolve_for_dri2_flush(struct intel_context *intel,
527 __DRIdrawable *drawable);
528
529 void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
530 uint32_t buffer_id);
531 void intel_init_texture_formats(struct gl_context *ctx);
532
533 /*======================================================================
534 * Inline conversion functions.
535 * These are better-typed than the macros used previously:
536 */
537 static INLINE struct intel_context *
538 intel_context(struct gl_context * ctx)
539 {
540 return (struct intel_context *) ctx;
541 }
542
543 static INLINE bool
544 is_power_of_two(uint32_t value)
545 {
546 return (value & (value - 1)) == 0;
547 }
548
549 #ifdef __cplusplus
550 }
551 #endif
552
553 #endif