src/mesa/main/imports.h

   1 /*
   2  * Mesa 3-D graphics library
   3  * Version:  6.1
   4  *
   5  * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a
   8  * copy of this software and associated documentation files (the "Software"),
   9  * to deal in the Software without restriction, including without limitation
  10  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  11  * and/or sell copies of the Software, and to permit persons to whom the
  12  * Software is furnished to do so, subject to the following conditions:
  13  *
  14  * The above copyright notice and this permission notice shall be included
  15  * in all copies or substantial portions of the Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  18  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  20  * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
  21  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  22  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  23  */
  24
  25
  26 /**
  27  * \file imports.h
  28  * Standard C library function wrappers.
  29  *
  30  * This file provides wrappers for all the standard C library functions
  31  * like malloc(), free(), printf(), getenv(), etc.
  32  */
  33
  34
  35 #ifndef IMPORTS_H
  36 #define IMPORTS_H
  37
  38
  39 /* XXX some of the stuff in glheader.h should be moved into this file.
  40  */
  41 #include "glheader.h"
  42
  43
  44 #ifdef __cplusplus
  45 extern "C" {
  46 #endif
  47
  48
  49 /**********************************************************************/
  50 /** \name General macros */
  51 /*@{*/
  52
  53 #ifndef NULL
  54 #define NULL 0
  55 #endif
  56
  57 /*@}*/
  58
  59
  60 /**********************************************************************/
  61 /** Memory macros */
  62 /*@{*/
  63
  64 /** Allocate \p BYTES bytes */
  65 #define MALLOC(BYTES)      _mesa_malloc(BYTES)
  66 /** Allocate and zero \p BYTES bytes */
  67 #define CALLOC(BYTES)      _mesa_calloc(BYTES)
  68 /** Allocate a structure of type \p T */
  69 #define MALLOC_STRUCT(T)   (struct T *) _mesa_malloc(sizeof(struct T))
  70 /** Allocate and zero a structure of type \p T */
  71 #define CALLOC_STRUCT(T)   (struct T *) _mesa_calloc(sizeof(struct T))
  72 /** Free memory */
  73 #define FREE(PTR)          _mesa_free(PTR)
  74
  75 /** Allocate \p BYTES aligned at \p N bytes */
  76 #define ALIGN_MALLOC(BYTES, N)     _mesa_align_malloc(BYTES, N)
  77 /** Allocate and zero \p BYTES bytes aligned at \p N bytes */
  78 #define ALIGN_CALLOC(BYTES, N)     _mesa_align_calloc(BYTES, N)
  79 /** Allocate a structure of type \p T aligned at \p N bytes */
  80 #define ALIGN_MALLOC_STRUCT(T, N)  (struct T *) _mesa_align_malloc(sizeof(struct T), N)
  81 /** Allocate and zero a structure of type \p T aligned at \p N bytes */
  82 #define ALIGN_CALLOC_STRUCT(T, N)  (struct T *) _mesa_align_calloc(sizeof(struct T), N)
  83 /** Free aligned memory */
  84 #define ALIGN_FREE(PTR)            _mesa_align_free(PTR)
  85
  86 /** Copy \p BYTES bytes from \p SRC into \p DST */
  87 #define MEMCPY( DST, SRC, BYTES)   _mesa_memcpy(DST, SRC, BYTES)
  88 /** Set \p N bytes in \p DST to \p VAL */
  89 #define MEMSET( DST, VAL, N )      _mesa_memset(DST, VAL, N)
  90
  91 #define MEMSET16( DST, VAL, N )   _mesa_memset16( (DST), (VAL), (size_t) (N) )
  92
  93 /*@}*/
  94
  95
  96 /*
  97  * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
  98  * as offsets into buffer stores.  Since the vertex array pointer and
  99  * buffer store pointer are both pointers and we need to add them, we use
 100  * this macro.
 101  * Both pointers/offsets are expressed in bytes.
 102  */
 103 #define ADD_POINTERS(A, B)  ( (A) + (unsigned long) (B) )
 104
 105
 106 /**********************************************************************/
 107 /** \name [Pseudo] static array declaration.
 108  *
 109  * MACs and BeOS don't support static larger than 32kb, so ...
 110  */
 111 /*@{*/
 112
 113 /**
 114  * \def DEFARRAY
 115  * Define a [static] unidimensional array
 116  */
 117
 118 /**
 119  * \def DEFMARRAY
 120  * Define a [static] bi-dimensional array
 121  */
 122
 123 /**
 124  * \def DEFMNARRAY
 125  * Define a [static] tri-dimensional array
 126  */
 127
 128 /**
 129  * \def CHECKARRAY
 130  * Verifies a [static] array was properly allocated.
 131  */
 132
 133 /**
 134  * \def UNDEFARRAY
 135  * Undefine (free) a [static] array.
 136  */
 137
 138 #if defined(macintosh) && !defined(__MRC__)
 139 /*extern char *AGLAlloc(int size);*/
 140 /*extern void AGLFree(char* ptr);*/
 141 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE *NAME = (TYPE*)_mesa_alloc(sizeof(TYPE)*(SIZE))
 142 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
 143 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
 144 #  define CHECKARRAY(NAME,CMD)                          do {if (!(NAME)) {CMD;}} while (0)
 145 #  define UNDEFARRAY(NAME)                              do {if ((NAME)) {_mesa_free((char*)NAME);}  }while (0)
 146 #elif defined(__BEOS__)
 147 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE *NAME = (TYPE*)_mesa_malloc(sizeof(TYPE)*(SIZE))
 148 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
 149 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
 150 #  define CHECKARRAY(NAME,CMD)                          do {if (!(NAME)) {CMD;}} while (0)
 151 #  define UNDEFARRAY(NAME)                              do {if ((NAME)) {_mesa_free((char*)NAME);}  }while (0)
 152 #else
 153 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE NAME[SIZE]
 154 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE NAME[SIZE1][SIZE2]
 155 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE NAME[SIZE1][SIZE2][SIZE3]
 156 #  define CHECKARRAY(NAME,CMD)                          do {} while(0)
 157 #  define UNDEFARRAY(NAME)
 158 #endif
 159
 160 /*@}*/
 161
 162
 163 /**********************************************************************/
 164 /** \name External pixel buffer allocation.
 165  *
 166  * If you want Mesa's depth/stencil/accum/etc buffers to be allocated with a
 167  * specialized allocator you can define MESA_EXTERNAL_BUFFERALLOC and implement
 168  * _ext_mesa_alloc_pixelbuffer() _ext_mesa_free_pixelbuffer() in your
 169  * application.
 170  *
 171  * \author
 172  * Contributed by Gerk Huisma (gerk@five-d.demon.nl).
 173  */
 174 /*@{*/
 175
 176 /**
 177  * \def MESA_PBUFFER_ALLOC
 178  * Allocate a pixel buffer.
 179  */
 180
 181 /**
 182  * \def MESA_PBUFFER_FREE
 183  * Free a pixel buffer.
 184  */
 185
 186 #ifdef MESA_EXTERNAL_BUFFERALLOC
 187 extern void *_ext_mesa_alloc_pixelbuffer( unsigned int size );
 188 extern void _ext_mesa_free_pixelbuffer( void *pb );
 189
 190 #define MESA_PBUFFER_ALLOC(BYTES)  (void *) _ext_mesa_alloc_pixelbuffer(BYTES)
 191 #define MESA_PBUFFER_FREE(PTR)     _ext_mesa_free_pixelbuffer(PTR)
 192 #else
 193 /* Default buffer allocation uses the aligned allocation routines: */
 194 #define MESA_PBUFFER_ALLOC(BYTES)  (void *) _mesa_align_malloc(BYTES, 512)
 195 #define MESA_PBUFFER_FREE(PTR)     _mesa_align_free(PTR)
 196 #endif
 197
 198 /*@}*/
 199
 200
 201
 202 /**********************************************************************
 203  * Math macros
 204  */
 205
 206 #define MAX_GLUSHORT    0xffff
 207 #define MAX_GLUINT      0xffffffff
 208
 209 #ifndef M_PI
 210 #define M_PI (3.1415926536)
 211 #endif
 212
 213 /* XXX this is a bit of a hack needed for compilation within XFree86 */
 214 #ifndef FLT_MIN
 215 #define FLT_MIN (1.0e-37)
 216 #endif
 217
 218 /* Degrees to radians conversion: */
 219 #define DEG2RAD (M_PI/180.0)
 220
 221
 222 /***
 223  *** USE_IEEE: Determine if we're using IEEE floating point
 224  ***/
 225 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
 226     defined(__s390x__) || defined(__powerpc__) || \
 227     defined(__AMD64__) || \
 228     defined(ia64) || defined(__ia64__) || \
 229     (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
 230 #define USE_IEEE
 231 #define IEEE_ONE 0x3f800000
 232 #endif
 233
 234
 235 /***
 236  *** SQRTF: single-precision square root
 237  ***/
 238 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 239 #  define SQRTF(X)  _mesa_sqrtf(X)
 240 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 241 #  define SQRTF(X)  (float) xf86sqrt((float) (X))
 242 #else
 243 #  define SQRTF(X)  (float) sqrt((float) (X))
 244 #endif
 245
 246
 247 /***
 248  *** INV_SQRTF: single-precision inverse square root
 249  ***/
 250 #if 0
 251 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 252 #else
 253 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 254 #endif
 255
 256
 257 /***
 258  *** LOG2: Log base 2 of float
 259  ***/
 260 #ifdef USE_IEEE
 261 #if 0
 262 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 263  * Based on code from http://www.stereopsis.com/log2.html
 264  */
 265 static INLINE GLfloat LOG2(GLfloat x)
 266 {
 267    const GLfloat y = x * x * x * x;
 268    const GLuint ix = *((GLuint *) &y);
 269    const GLuint exp = (ix >> 23) & 0xFF;
 270    const GLint log2 = ((GLint) exp) - 127;
 271    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 272 }
 273 #endif
 274 /* Pretty fast, and accurate.
 275  * Based on code from http://www.flipcode.com/totd/
 276  */
 277 static INLINE GLfloat LOG2(GLfloat val)
 278 {
 279    fi_type num;
 280    GLint log_2;
 281    num.f = val;
 282    log_2 = ((num.i >> 23) & 255) - 128;
 283    num.i &= ~(255 << 23);
 284    num.i += 127 << 23;
 285    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 286    return num.f + log_2;
 287 }
 288 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 289 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 290 #else
 291 /*
 292  * NOTE: log_base_2(x) = log(x) / log(2)
 293  * NOTE: 1.442695 = 1/log(2).
 294  */
 295 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 296 #endif
 297
 298
 299 /***
 300  *** IS_INF_OR_NAN: test if float is infinite or NaN
 301  ***/
 302 #ifdef USE_IEEE
 303 static INLINE int IS_INF_OR_NAN( float x )
 304 {
 305    fi_type tmp;
 306    tmp.f = x;
 307    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 308 }
 309 #elif defined(isfinite)
 310 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 311 #elif defined(finite)
 312 #define IS_INF_OR_NAN(x)        (!finite(x))
 313 #elif defined(__VMS)
 314 #define IS_INF_OR_NAN(x)        (!finite(x))
 315 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 316 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 317 #else
 318 #define IS_INF_OR_NAN(x)        (!finite(x))
 319 #endif
 320
 321
 322 /***
 323  *** IS_NEGATIVE: test if float is negative
 324  ***/
 325 #if defined(USE_IEEE)
 326 #define GET_FLOAT_BITS(x) ((fi_type *) &(x))->i
 327 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) & (1<<31))
 328 #else
 329 #define IS_NEGATIVE(x) (x < 0.0F)
 330 #endif
 331
 332
 333 /***
 334  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 335  ***/
 336 #if defined(USE_IEEE)
 337 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 338 #else
 339 /* Could just use (x*y<0) except for the flatshading requirements.
 340  * Maybe there's a better way?
 341  */
 342 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 343 #endif
 344
 345
 346 /***
 347  *** CEILF: ceiling of float
 348  *** FLOORF: floor of float
 349  *** FABSF: absolute value of float
 350  ***/
 351 #if defined(XFree86LOADER) && defined(IN_MODULE)
 352 #define CEILF(x)   ((GLfloat) xf86ceil(x))
 353 #define FLOORF(x)  ((GLfloat) xf86floor(x))
 354 #define FABSF(x)   ((GLfloat) xf86fabs(x))
 355 #elif defined(__gnu_linux__)
 356 /* C99 functions */
 357 #define CEILF(x)   ceilf(x)
 358 #define FLOORF(x)  floorf(x)
 359 #define FABSF(x)   fabsf(x)
 360 #else
 361 #define CEILF(x)   ((GLfloat) ceil(x))
 362 #define FLOORF(x)  ((GLfloat) floor(x))
 363 #define FABSF(x)   ((GLfloat) fabs(x))
 364 #endif
 365
 366
 367 /***
 368  *** IROUND: return (as an integer) float rounded to nearest integer
 369  ***/
 370 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
 371 static INLINE int iround(float f)
 372 {
 373    int r;
 374    __asm__ ("fstoi %1, %0" : "=f" (r) : "f" (f));
 375    return r;
 376 }
 377 #define IROUND(x)  iround(x)
 378 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 379 static INLINE int iround(float f)
 380 {
 381    int r;
 382    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 383    return r;
 384 }
 385 #define IROUND(x)  iround(x)
 386 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
 387 static INLINE int iround(float f)
 388 {
 389    int r;
 390    _asm {
 391          fld f
 392          fistp r
 393         }
 394    return r;
 395 }
 396 #define IROUND(x)  iround(x)
 397 #elif defined(__WATCOMC__) && defined(__386__)
 398 long iround(float f);
 399 #pragma aux iround =                    \
 400         "push   eax"                        \
 401         "fistp  dword ptr [esp]"            \
 402         "pop    eax"                        \
 403         parm [8087]                         \
 404         value [eax]                         \
 405         modify exact [eax];
 406 #define IROUND(x)  iround(x)
 407 #else
 408 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 409 #endif
 410
 411
 412 /***
 413  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 414  ***/
 415 #ifdef DEBUG
 416 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 417 #else
 418 #define IROUND_POS(f) (IROUND(f))
 419 #endif
 420
 421
 422 /***
 423  *** IFLOOR: return (as an integer) floor of float
 424  ***/
 425 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 426 /*
 427  * IEEE floor for computers that round to nearest or even.
 428  * 'f' must be between -4194304 and 4194303.
 429  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 430  * but uses some IEEE specific tricks for better speed.
 431  * Contributed by Josh Vanderhoof
 432  */
 433 static INLINE int ifloor(float f)
 434 {
 435    int ai, bi;
 436    double af, bf;
 437    af = (3 << 22) + 0.5 + (double)f;
 438    bf = (3 << 22) + 0.5 - (double)f;
 439    /* GCC generates an extra fstp/fld without this. */
 440    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 441    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 442    return (ai - bi) >> 1;
 443 }
 444 #define IFLOOR(x)  ifloor(x)
 445 #elif defined(USE_IEEE)
 446 static INLINE int ifloor(float f)
 447 {
 448    int ai, bi;
 449    double af, bf;
 450    fi_type u;
 451
 452    af = (3 << 22) + 0.5 + (double)f;
 453    bf = (3 << 22) + 0.5 - (double)f;
 454    u.f = (float) af;  ai = u.i;
 455    u.f = (float) bf;  bi = u.i;
 456    return (ai - bi) >> 1;
 457 }
 458 #define IFLOOR(x)  ifloor(x)
 459 #else
 460 static INLINE int ifloor(float f)
 461 {
 462    int i = IROUND(f);
 463    return (i > f) ? i - 1 : i;
 464 }
 465 #define IFLOOR(x)  ifloor(x)
 466 #endif
 467
 468
 469 /***
 470  *** ICEIL: return (as an integer) ceiling of float
 471  ***/
 472 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 473 /*
 474  * IEEE ceil for computers that round to nearest or even.
 475  * 'f' must be between -4194304 and 4194303.
 476  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 477  * but uses some IEEE specific tricks for better speed.
 478  * Contributed by Josh Vanderhoof
 479  */
 480 static INLINE int iceil(float f)
 481 {
 482    int ai, bi;
 483    double af, bf;
 484    af = (3 << 22) + 0.5 + (double)f;
 485    bf = (3 << 22) + 0.5 - (double)f;
 486    /* GCC generates an extra fstp/fld without this. */
 487    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 488    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 489    return (ai - bi + 1) >> 1;
 490 }
 491 #define ICEIL(x)  iceil(x)
 492 #elif defined(USE_IEEE)
 493 static INLINE int iceil(float f)
 494 {
 495    int ai, bi;
 496    double af, bf;
 497    fi_type u;
 498    af = (3 << 22) + 0.5 + (double)f;
 499    bf = (3 << 22) + 0.5 - (double)f;
 500    u.f = (float) af; ai = u.i;
 501    u.f = (float) bf; bi = u.i;
 502    return (ai - bi + 1) >> 1;
 503 }
 504 #define ICEIL(x)  iceil(x)
 505 #else
 506 static INLINE int iceil(float f)
 507 {
 508    int i = IROUND(f);
 509    return (i < f) ? i + 1 : i;
 510 }
 511 #define ICEIL(x)  iceil(x)
 512 #endif
 513
 514
 515 /***
 516  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 517  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 518  ***/
 519 #if defined(USE_IEEE) && !defined(DEBUG)
 520 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 521 /* This function/macro is sensitive to precision.  Test very carefully
 522  * if you change it!
 523  */
 524 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 525         do {                                                            \
 526            fi_type __tmp;                                               \
 527            __tmp.f = (F);                                               \
 528            if (__tmp.i < 0)                                             \
 529               UB = (GLubyte) 0;                                         \
 530            else if (__tmp.i >= IEEE_0996)                               \
 531               UB = (GLubyte) 255;                                       \
 532            else {                                                       \
 533               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 534               UB = (GLubyte) __tmp.i;                                   \
 535            }                                                            \
 536         } while (0)
 537 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 538         do {                                                            \
 539            fi_type __tmp;                                               \
 540            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 541            UB = (GLubyte) __tmp.i;                                      \
 542         } while (0)
 543 #else
 544 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 545         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 546 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 547         ub = ((GLubyte) IROUND((f) * 255.0F))
 548 #endif
 549
 550
 551 /***
 552  *** COPY_FLOAT: copy a float from src to dest, avoid slow FP regs if possible
 553  ***/
 554 #if defined(USE_IEEE) && !defined(DEBUG)
 555 #define COPY_FLOAT( dst, src )                                  \
 556         ((fi_type *) &(dst))->i = ((fi_type *) &(src))->i
 557 #else
 558 #define COPY_FLOAT( dst, src )          (dst) = (src)
 559 #endif
 560
 561
 562 /***
 563  *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
 564  ***                  original mode to a temporary).
 565  *** END_FAST_MATH: Restore x86 FPU to original mode.
 566  ***/
 567 #if defined(__GNUC__) && defined(__i386__)
 568 /*
 569  * Set the x86 FPU control word to guarentee only 32 bits of precision
 570  * are stored in registers.  Allowing the FPU to store more introduces
 571  * differences between situations where numbers are pulled out of memory
 572  * vs. situations where the compiler is able to optimize register usage.
 573  *
 574  * In the worst case, we force the compiler to use a memory access to
 575  * truncate the float, by specifying the 'volatile' keyword.
 576  */
 577 /* Hardware default: All exceptions masked, extended double precision,
 578  * round to nearest (IEEE compliant):
 579  */
 580 #define DEFAULT_X86_FPU         0x037f
 581 /* All exceptions masked, single precision, round to nearest:
 582  */
 583 #define FAST_X86_FPU            0x003f
 584 /* The fldcw instruction will cause any pending FP exceptions to be
 585  * raised prior to entering the block, and we clear any pending
 586  * exceptions before exiting the block.  Hence, asm code has free
 587  * reign over the FPU while in the fast math block.
 588  */
 589 #if defined(NO_FAST_MATH)
 590 #define START_FAST_MATH(x)                                              \
 591 do {                                                                    \
 592    static GLuint mask = DEFAULT_X86_FPU;                                \
 593    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 594    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 595 } while (0)
 596 #else
 597 #define START_FAST_MATH(x)                                              \
 598 do {                                                                    \
 599    static GLuint mask = FAST_X86_FPU;                                   \
 600    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 601    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 602 } while (0)
 603 #endif
 604 /* Restore original FPU mode, and clear any exceptions that may have
 605  * occurred in the FAST_MATH block.
 606  */
 607 #define END_FAST_MATH(x)                                                \
 608 do {                                                                    \
 609    __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) );                     \
 610 } while (0)
 611
 612 #elif defined(__WATCOMC__) && defined(__386__)
 613 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 614 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 615 void _watcom_start_fast_math(unsigned short *x,unsigned short *mask);
 616 #pragma aux _watcom_start_fast_math =                                   \
 617    "fnstcw  word ptr [eax]"                                             \
 618    "fldcw   word ptr [ecx]"                                             \
 619    parm [eax] [ecx]                                                     \
 620    modify exact [];
 621 void _watcom_end_fast_math(unsigned short *x);
 622 #pragma aux _watcom_end_fast_math =                                     \
 623    "fnclex"                                                             \
 624    "fldcw   word ptr [eax]"                                             \
 625    parm [eax]                                                           \
 626    modify exact [];
 627 #if defined(NO_FAST_MATH)
 628 #define START_FAST_MATH(x)                                              \
 629 do {                                                                    \
 630    static GLushort mask = DEFAULT_X86_FPU;                                  \
 631    _watcom_start_fast_math(&x,&mask);                                   \
 632 } while (0)
 633 #else
 634 #define START_FAST_MATH(x)                                              \
 635 do {                                                                    \
 636    static GLushort mask = FAST_X86_FPU;                                 \
 637    _watcom_start_fast_math(&x,&mask);                                   \
 638 } while (0)
 639 #endif
 640 #define END_FAST_MATH(x)  _watcom_end_fast_math(&x)
 641 #else
 642 #define START_FAST_MATH(x)  x = 0
 643 #define END_FAST_MATH(x)  (void)(x)
 644 #endif
 645
 646
 647
 648 /**********************************************************************
 649  * Functions
 650  */
 651
 652 extern void *
 653 _mesa_malloc( size_t bytes );
 654
 655 extern void *
 656 _mesa_calloc( size_t bytes );
 657
 658 extern void
 659 _mesa_free( void *ptr );
 660
 661 extern void *
 662 _mesa_align_malloc( size_t bytes, unsigned long alignment );
 663
 664 extern void *
 665 _mesa_align_calloc( size_t bytes, unsigned long alignment );
 666
 667 extern void
 668 _mesa_align_free( void *ptr );
 669
 670 extern void *
 671 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 672
 673 extern void *
 674 _mesa_memcpy( void *dest, const void *src, size_t n );
 675
 676 extern void
 677 _mesa_memset( void *dst, int val, size_t n );
 678
 679 extern void
 680 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 681
 682 extern void
 683 _mesa_bzero( void *dst, size_t n );
 684
 685
 686 extern double
 687 _mesa_sin(double a);
 688
 689 extern double
 690 _mesa_cos(double a);
 691
 692 extern double
 693 _mesa_sqrtd(double x);
 694
 695 extern float
 696 _mesa_sqrtf(float x);
 697
 698 extern float
 699 _mesa_inv_sqrtf(float x);
 700
 701 extern double
 702 _mesa_pow(double x, double y);
 703
 704 extern float
 705 _mesa_log2(float x);
 706
 707 extern unsigned int
 708 _mesa_bitcount(unsigned int n);
 709
 710 extern GLhalfARB
 711 _mesa_float_to_half(float f);
 712
 713 extern float
 714 _mesa_half_to_float(GLhalfARB h);
 715
 716
 717 extern char *
 718 _mesa_getenv( const char *var );
 719
 720 extern char *
 721 _mesa_strstr( const char *haystack, const char *needle );
 722
 723 extern char *
 724 _mesa_strncat( char *dest, const char *src, size_t n );
 725
 726 extern char *
 727 _mesa_strcpy( char *dest, const char *src );
 728
 729 extern char *
 730 _mesa_strncpy( char *dest, const char *src, size_t n );
 731
 732 extern size_t
 733 _mesa_strlen( const char *s );
 734
 735 extern int
 736 _mesa_strcmp( const char *s1, const char *s2 );
 737
 738 extern int
 739 _mesa_strncmp( const char *s1, const char *s2, size_t n );
 740
 741 extern char *
 742 _mesa_strdup( const char *s );
 743
 744 extern int
 745 _mesa_atoi( const char *s );
 746
 747 extern double
 748 _mesa_strtod( const char *s, char **end );
 749
 750 extern int
 751 _mesa_sprintf( char *str, const char *fmt, ... );
 752
 753 extern void
 754 _mesa_printf( const char *fmtString, ... );
 755
 756
 757 extern void
 758 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 759
 760 extern void
 761 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 762
 763 extern void
 764 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 765
 766 extern void
 767 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 768
 769
 770 extern void
 771 _mesa_init_default_imports( __GLimports *imports, void *driverCtx );
 772
 773
 774 #ifdef __cplusplus
 775 }
 776 #endif
 777
 778
 779 #endif /* IMPORTS_H */