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 struct intel_region;
  57 struct intel_context;
  58
  59 #define INTEL_WRITE_PART  0x1
  60 #define INTEL_WRITE_FULL  0x2
  61 #define INTEL_READ        0x4
  62
  63 #ifndef likely
  64 #ifdef __GNUC__
  65 #define likely(expr) (__builtin_expect(expr, 1))
  66 #define unlikely(expr) (__builtin_expect(expr, 0))
  67 #else
  68 #define likely(expr) (expr)
  69 #define unlikely(expr) (expr)
  70 #endif
  71 #endif
  72
  73 struct intel_sync_object {
  74    struct gl_sync_object Base;
  75
  76    /** Batch associated with this sync object */
  77    drm_intel_bo *bo;
  78 };
  79
  80 struct brw_context;
  81
  82 struct intel_batchbuffer {
  83    /** Current batchbuffer being queued up. */
  84    drm_intel_bo *bo;
  85    /** Last BO submitted to the hardware.  Used for glFinish(). */
  86    drm_intel_bo *last_bo;
  87    /** BO for post-sync nonzero writes for gen6 workaround. */
  88    drm_intel_bo *workaround_bo;
  89    bool need_workaround_flush;
  90
  91    struct cached_batch_item *cached_items;
  92
  93    uint16_t emit, total;
  94    uint16_t used, reserved_space;
  95    uint32_t *map;
  96    uint32_t *cpu_map;
  97 #define BATCH_SZ (8192*sizeof(uint32_t))
  98
  99    uint32_t state_batch_offset;
 100    bool is_blit;
 101    bool needs_sol_reset;
 102
 103    struct {
 104       uint16_t used;
 105       int reloc_count;
 106    } saved;
 107 };
 108
 109 /**
 110  * intel_context is derived from Mesa's context class: struct gl_context.
 111  */
 112 struct intel_context
 113 {
 114    struct gl_context ctx;  /**< base class, must be first field */
 115
 116    struct
 117    {
 118       void (*destroy) (struct intel_context * intel);
 119       void (*finish_batch) (struct intel_context * intel);
 120       void (*new_batch) (struct intel_context * intel);
 121
 122       void (*update_texture_surface)(struct gl_context *ctx,
 123                                      unsigned unit,
 124                                      uint32_t *binding_table,
 125                                      unsigned surf_index);
 126       void (*update_renderbuffer_surface)(struct brw_context *brw,
 127                                           struct gl_renderbuffer *rb,
 128                                           bool layered,
 129                                           unsigned unit);
 130       void (*update_null_renderbuffer_surface)(struct brw_context *brw,
 131                                                unsigned unit);
 132       void (*create_constant_surface)(struct brw_context *brw,
 133                                       drm_intel_bo *bo,
 134                                       uint32_t offset,
 135                                       uint32_t size,
 136                                       uint32_t *out_offset,
 137                                       bool dword_pitch);
 138
 139       /**
 140        * Send the appropriate state packets to configure depth, stencil, and
 141        * HiZ buffers (i965+ only)
 142        */
 143       void (*emit_depth_stencil_hiz)(struct brw_context *brw,
 144                                      struct intel_mipmap_tree *depth_mt,
 145                                      uint32_t depth_offset,
 146                                      uint32_t depthbuffer_format,
 147                                      uint32_t depth_surface_type,
 148                                      struct intel_mipmap_tree *stencil_mt,
 149                                      bool hiz, bool separate_stencil,
 150                                      uint32_t width, uint32_t height,
 151                                      uint32_t tile_x, uint32_t tile_y);
 152
 153    } vtbl;
 154
 155    GLuint NewGLState;
 156
 157    dri_bufmgr *bufmgr;
 158
 159    /**
 160     * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965.
 161     */
 162    int gen;
 163    int gt;
 164    bool needs_ff_sync;
 165    bool is_haswell;
 166    bool is_baytrail;
 167    bool is_g4x;
 168    bool has_separate_stencil;
 169    bool must_use_separate_stencil;
 170    bool has_hiz;
 171    bool has_llc;
 172    bool has_swizzling;
 173
 174    drm_intel_context *hw_ctx;
 175
 176    struct intel_batchbuffer batch;
 177
 178    drm_intel_bo *first_post_swapbuffers_batch;
 179    bool need_throttle;
 180    bool no_batch_wrap;
 181
 182    /**
 183     * Set if we're either a debug context or the INTEL_DEBUG=perf environment
 184     * variable is set, this is the flag indicating to do expensive work that
 185     * might lead to a perf_debug() call.
 186     */
 187    bool perf_debug;
 188
 189    struct {
 190       drm_intel_bo *bo;
 191       GLuint offset;
 192       uint32_t buffer_len;
 193       uint32_t buffer_offset;
 194       char buffer[4096];
 195    } upload;
 196
 197    uint32_t max_gtt_map_object_size;
 198
 199    GLuint stats_wm;
 200
 201    bool no_rast;
 202    bool always_flush_batch;
 203    bool always_flush_cache;
 204    bool disable_throttling;
 205
 206    GLenum reduced_primitive;
 207
 208    /**
 209     * Set if rendering has occured to the drawable's front buffer.
 210     *
 211     * This is used in the DRI2 case to detect that glFlush should also copy
 212     * the contents of the fake front buffer to the real front buffer.
 213     */
 214    bool front_buffer_dirty;
 215
 216    /**
 217     * Track whether front-buffer rendering is currently enabled
 218     *
 219     * A separate flag is used to track this in order to support MRT more
 220     * easily.
 221     */
 222    bool is_front_buffer_rendering;
 223    /**
 224     * Track whether front-buffer is the current read target.
 225     *
 226     * This is closely associated with is_front_buffer_rendering, but may
 227     * be set separately.  The DRI2 fake front buffer must be referenced
 228     * either way.
 229     */
 230    bool is_front_buffer_reading;
 231
 232    int driFd;
 233
 234    __DRIcontext *driContext;
 235    struct intel_screen *intelScreen;
 236
 237    /**
 238     * Configuration cache
 239     */
 240    driOptionCache optionCache;
 241 };
 242
 243 /**
 244  * Align a value down to an alignment value
 245  *
 246  * If \c value is not already aligned to the requested alignment value, it
 247  * will be rounded down.
 248  *
 249  * \param value  Value to be rounded
 250  * \param alignment  Alignment value to be used.  This must be a power of two.
 251  *
 252  * \sa ALIGN()
 253  */
 254 #define ROUND_DOWN_TO(value, alignment) ((value) & ~(alignment - 1))
 255
 256 static INLINE uint32_t
 257 U_FIXED(float value, uint32_t frac_bits)
 258 {
 259    value *= (1 << frac_bits);
 260    return value < 0 ? 0 : value;
 261 }
 262
 263 static INLINE uint32_t
 264 S_FIXED(float value, uint32_t frac_bits)
 265 {
 266    return value * (1 << frac_bits);
 267 }
 268
 269 /* ================================================================
 270  * From linux kernel i386 header files, copes with odd sizes better
 271  * than COPY_DWORDS would:
 272  * XXX Put this in src/mesa/main/imports.h ???
 273  */
 274 #if defined(i386) || defined(__i386__)
 275 static INLINE void * __memcpy(void * to, const void * from, size_t n)
 276 {
 277    int d0, d1, d2;
 278    __asm__ __volatile__(
 279       "rep ; movsl\n\t"
 280       "testb $2,%b4\n\t"
 281       "je 1f\n\t"
 282       "movsw\n"
 283       "1:\ttestb $1,%b4\n\t"
 284       "je 2f\n\t"
 285       "movsb\n"
 286       "2:"
 287       : "=&c" (d0), "=&D" (d1), "=&S" (d2)
 288       :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
 289       : "memory");
 290    return (to);
 291 }
 292 #else
 293 #define __memcpy(a,b,c) memcpy(a,b,c)
 294 #endif
 295
 296
 297 /* ================================================================
 298  * Debugging:
 299  */
 300 extern int INTEL_DEBUG;
 301
 302 #define DEBUG_TEXTURE   0x1
 303 #define DEBUG_STATE     0x2
 304 #define DEBUG_IOCTL     0x4
 305 #define DEBUG_BLIT      0x8
 306 #define DEBUG_MIPTREE   0x10
 307 #define DEBUG_PERF      0x20
 308 #define DEBUG_BATCH     0x80
 309 #define DEBUG_PIXEL     0x100
 310 #define DEBUG_BUFMGR    0x200
 311 #define DEBUG_REGION    0x400
 312 #define DEBUG_FBO       0x800
 313 #define DEBUG_GS        0x1000
 314 #define DEBUG_SYNC      0x2000
 315 #define DEBUG_PRIMS     0x4000
 316 #define DEBUG_VERTS     0x8000
 317 #define DEBUG_DRI       0x10000
 318 #define DEBUG_SF        0x20000
 319 #define DEBUG_STATS     0x100000
 320 #define DEBUG_WM        0x400000
 321 #define DEBUG_URB       0x800000
 322 #define DEBUG_VS        0x1000000
 323 #define DEBUG_CLIP      0x2000000
 324 #define DEBUG_AUB       0x4000000
 325 #define DEBUG_SHADER_TIME 0x8000000
 326 #define DEBUG_BLORP     0x10000000
 327 #define DEBUG_NO16      0x20000000
 328
 329 #ifdef HAVE_ANDROID_PLATFORM
 330 #define LOG_TAG "INTEL-MESA"
 331 #include <cutils/log.h>
 332 #ifndef ALOGW
 333 #define ALOGW LOGW
 334 #endif
 335 #define dbg_printf(...) ALOGW(__VA_ARGS__)
 336 #else
 337 #define dbg_printf(...) printf(__VA_ARGS__)
 338 #endif /* HAVE_ANDROID_PLATFORM */
 339
 340 #define DBG(...) do {                                           \
 341         if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG))            \
 342                 dbg_printf(__VA_ARGS__);                        \
 343 } while(0)
 344
 345 #define perf_debug(...) do {                                    \
 346    static GLuint msg_id = 0;                                    \
 347    if (unlikely(INTEL_DEBUG & DEBUG_PERF))                      \
 348       dbg_printf(__VA_ARGS__);                                  \
 349    if (intel->perf_debug)                                       \
 350       _mesa_gl_debug(&intel->ctx, &msg_id,                      \
 351                      MESA_DEBUG_TYPE_PERFORMANCE,               \
 352                      MESA_DEBUG_SEVERITY_MEDIUM,                \
 353                      __VA_ARGS__);                              \
 354 } while(0)
 355
 356 #define WARN_ONCE(cond, fmt...) do {                            \
 357    if (unlikely(cond)) {                                        \
 358       static bool _warned = false;                              \
 359       static GLuint msg_id = 0;                                 \
 360       if (!_warned) {                                           \
 361          fprintf(stderr, "WARNING: ");                          \
 362          fprintf(stderr, fmt);                                  \
 363          _warned = true;                                        \
 364                                                                 \
 365          _mesa_gl_debug(ctx, &msg_id,                           \
 366                         MESA_DEBUG_TYPE_OTHER,                  \
 367                         MESA_DEBUG_SEVERITY_HIGH, fmt);         \
 368       }                                                         \
 369    }                                                            \
 370 } while (0)
 371
 372 /* ================================================================
 373  * intel_context.c:
 374  */
 375
 376 extern bool intelInitContext(struct intel_context *intel,
 377                              int api,
 378                              unsigned major_version,
 379                              unsigned minor_version,
 380                              const struct gl_config * mesaVis,
 381                              __DRIcontext * driContextPriv,
 382                              void *sharedContextPrivate,
 383                              struct dd_function_table *functions,
 384                              unsigned *dri_ctx_error);
 385
 386 extern void intelFinish(struct gl_context * ctx);
 387 extern void _intel_flush(struct gl_context * ctx, const char *file, int line);
 388
 389 #define intel_flush(ctx) _intel_flush(ctx, __FILE__, __LINE__)
 390
 391 extern void intelInitDriverFunctions(struct dd_function_table *functions);
 392
 393 void intel_init_syncobj_functions(struct dd_function_table *functions);
 394
 395 enum {
 396    DRI_CONF_BO_REUSE_DISABLED,
 397    DRI_CONF_BO_REUSE_ALL
 398 };
 399
 400 extern int intel_translate_shadow_compare_func(GLenum func);
 401 extern int intel_translate_compare_func(GLenum func);
 402 extern int intel_translate_stencil_op(GLenum op);
 403 extern int intel_translate_logic_op(GLenum opcode);
 404
 405 void intel_update_renderbuffers(__DRIcontext *context,
 406                                 __DRIdrawable *drawable);
 407 void intel_prepare_render(struct intel_context *intel);
 408
 409 void
 410 intel_resolve_for_dri2_flush(struct intel_context *intel,
 411                              __DRIdrawable *drawable);
 412
 413 extern void
 414 intelInitExtensions(struct gl_context *ctx);
 415 extern void
 416 intelInitClearFuncs(struct dd_function_table *functions);
 417
 418 /*======================================================================
 419  * Inline conversion functions.
 420  * These are better-typed than the macros used previously:
 421  */
 422 static INLINE struct intel_context *
 423 intel_context(struct gl_context * ctx)
 424 {
 425    return (struct intel_context *) ctx;
 426 }
 427
 428 static INLINE bool
 429 is_power_of_two(uint32_t value)
 430 {
 431    return (value & (value - 1)) == 0;
 432 }
 433
 434 #ifdef __cplusplus
 435 }
 436 #endif
 437
 438 #endif