src/mesa/main/imports.h

   1 /*
   2  * Mesa 3-D graphics library
   3  * Version:  6.2
   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 /*@}*/
  92
  93
  94 /*
  95  * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
  96  * as offsets into buffer stores.  Since the vertex array pointer and
  97  * buffer store pointer are both pointers and we need to add them, we use
  98  * this macro.
  99  * Both pointers/offsets are expressed in bytes.
 100  */
 101 #define ADD_POINTERS(A, B)  ( (GLubyte *) (A) + (uintptr_t) (B) )
 102
 103
 104 /**********************************************************************/
 105 /** \name [Pseudo] static array declaration.
 106  *
 107  * MACs and BeOS don't support static larger than 32kb, so ...
 108  */
 109 /*@{*/
 110
 111 /**
 112  * \def DEFARRAY
 113  * Define a [static] unidimensional array
 114  */
 115
 116 /**
 117  * \def DEFMARRAY
 118  * Define a [static] bi-dimensional array
 119  */
 120
 121 /**
 122  * \def DEFMNARRAY
 123  * Define a [static] tri-dimensional array
 124  */
 125
 126 /**
 127  * \def CHECKARRAY
 128  * Verifies a [static] array was properly allocated.
 129  */
 130
 131 /**
 132  * \def UNDEFARRAY
 133  * Undefine (free) a [static] array.
 134  */
 135
 136 #if defined(macintosh) && !defined(__MRC__)
 137 /*extern char *AGLAlloc(int size);*/
 138 /*extern void AGLFree(char* ptr);*/
 139 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE *NAME = (TYPE*)_mesa_alloc(sizeof(TYPE)*(SIZE))
 140 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
 141 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
 142 #  define CHECKARRAY(NAME,CMD)                          do {if (!(NAME)) {CMD;}} while (0)
 143 #  define UNDEFARRAY(NAME)                              do {if ((NAME)) {_mesa_free((char*)NAME);}  }while (0)
 144 #elif defined(__BEOS__)
 145 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE *NAME = (TYPE*)_mesa_malloc(sizeof(TYPE)*(SIZE))
 146 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
 147 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
 148 #  define CHECKARRAY(NAME,CMD)                          do {if (!(NAME)) {CMD;}} while (0)
 149 #  define UNDEFARRAY(NAME)                              do {if ((NAME)) {_mesa_free((char*)NAME);}  }while (0)
 150 #else
 151 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE NAME[SIZE]
 152 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE NAME[SIZE1][SIZE2]
 153 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE NAME[SIZE1][SIZE2][SIZE3]
 154 #  define CHECKARRAY(NAME,CMD)                          do {} while(0)
 155 #  define UNDEFARRAY(NAME)
 156 #endif
 157
 158 /*@}*/
 159
 160
 161 /**
 162  * Sometimes we treat GLfloats as GLints.  On x86 systems, moving a float
 163  * as a int (thereby using integer registers instead of FP registers) is
 164  * a performance win.  Typically, this can be done with ordinary casts.
 165  * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
 166  * these casts generate warnings.
 167  * The following union typedef is used to solve that.
 168  */
 169 typedef union { GLfloat f; GLint i; } fi_type;
 170
 171
 172
 173 /**********************************************************************
 174  * Math macros
 175  */
 176
 177 #define MAX_GLUSHORT    0xffff
 178 #define MAX_GLUINT      0xffffffff
 179
 180 #ifndef M_PI
 181 #define M_PI (3.1415926536)
 182 #endif
 183
 184 #ifndef M_E
 185 #define M_E (2.7182818284590452354)
 186 #endif
 187
 188
 189 /* XXX this is a bit of a hack needed for compilation within XFree86 */
 190 #ifndef FLT_MIN
 191 #define FLT_MIN (1.0e-37)
 192 #endif
 193
 194 /* Degrees to radians conversion: */
 195 #define DEG2RAD (M_PI/180.0)
 196
 197
 198 /***
 199  *** USE_IEEE: Determine if we're using IEEE floating point
 200  ***/
 201 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
 202     defined(__s390x__) || defined(__powerpc__) || \
 203     defined(__amd64__) || \
 204     defined(ia64) || defined(__ia64__) || \
 205     defined(__hppa__) || defined(hpux) || \
 206     defined(__mips) || defined(_MIPS_ARCH) || \
 207     defined(__arm__) || \
 208     (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
 209 #define USE_IEEE
 210 #define IEEE_ONE 0x3f800000
 211 #endif
 212
 213
 214 /***
 215  *** SQRTF: single-precision square root
 216  ***/
 217 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 218 #  define SQRTF(X)  _mesa_sqrtf(X)
 219 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 220 #  define SQRTF(X)  (float) xf86sqrt((float) (X))
 221 #else
 222 #  define SQRTF(X)  (float) sqrt((float) (X))
 223 #endif
 224
 225
 226 /***
 227  *** INV_SQRTF: single-precision inverse square root
 228  ***/
 229 #if 0
 230 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 231 #else
 232 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 233 #endif
 234
 235
 236 /***
 237  *** LOG2: Log base 2 of float
 238  ***/
 239 #ifdef USE_IEEE
 240 #if 0
 241 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 242  * Based on code from http://www.stereopsis.com/log2.html
 243  */
 244 static INLINE GLfloat LOG2(GLfloat x)
 245 {
 246    const GLfloat y = x * x * x * x;
 247    const GLuint ix = *((GLuint *) &y);
 248    const GLuint exp = (ix >> 23) & 0xFF;
 249    const GLint log2 = ((GLint) exp) - 127;
 250    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 251 }
 252 #endif
 253 /* Pretty fast, and accurate.
 254  * Based on code from http://www.flipcode.com/totd/
 255  */
 256 static INLINE GLfloat LOG2(GLfloat val)
 257 {
 258    fi_type num;
 259    GLint log_2;
 260    num.f = val;
 261    log_2 = ((num.i >> 23) & 255) - 128;
 262    num.i &= ~(255 << 23);
 263    num.i += 127 << 23;
 264    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 265    return num.f + log_2;
 266 }
 267 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 268 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 269 #else
 270 /*
 271  * NOTE: log_base_2(x) = log(x) / log(2)
 272  * NOTE: 1.442695 = 1/log(2).
 273  */
 274 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 275 #endif
 276
 277
 278 /***
 279  *** IS_INF_OR_NAN: test if float is infinite or NaN
 280  ***/
 281 #ifdef USE_IEEE
 282 static INLINE int IS_INF_OR_NAN( float x )
 283 {
 284    fi_type tmp;
 285    tmp.f = x;
 286    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 287 }
 288 #elif defined(isfinite)
 289 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 290 #elif defined(finite)
 291 #define IS_INF_OR_NAN(x)        (!finite(x))
 292 #elif defined(__VMS)
 293 #define IS_INF_OR_NAN(x)        (!finite(x))
 294 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 295 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 296 #else
 297 #define IS_INF_OR_NAN(x)        (!finite(x))
 298 #endif
 299
 300
 301 /***
 302  *** IS_NEGATIVE: test if float is negative
 303  ***/
 304 #if defined(USE_IEEE)
 305 static INLINE int GET_FLOAT_BITS( float x )
 306 {
 307    fi_type fi;
 308    fi.f = x;
 309    return fi.i;
 310 }
 311 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
 312 #else
 313 #define IS_NEGATIVE(x) (x < 0.0F)
 314 #endif
 315
 316
 317 /***
 318  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 319  ***/
 320 #if defined(USE_IEEE)
 321 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 322 #else
 323 /* Could just use (x*y<0) except for the flatshading requirements.
 324  * Maybe there's a better way?
 325  */
 326 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 327 #endif
 328
 329
 330 /***
 331  *** CEILF: ceiling of float
 332  *** FLOORF: floor of float
 333  *** FABSF: absolute value of float
 334  ***/
 335 #if defined(XFree86LOADER) && defined(IN_MODULE)
 336 #define CEILF(x)   ((GLfloat) xf86ceil(x))
 337 #define FLOORF(x)  ((GLfloat) xf86floor(x))
 338 #define FABSF(x)   ((GLfloat) xf86fabs(x))
 339 #elif defined(__gnu_linux__)
 340 /* C99 functions */
 341 #define CEILF(x)   ceilf(x)
 342 #define FLOORF(x)  floorf(x)
 343 #define FABSF(x)   fabsf(x)
 344 #else
 345 #define CEILF(x)   ((GLfloat) ceil(x))
 346 #define FLOORF(x)  ((GLfloat) floor(x))
 347 #define FABSF(x)   ((GLfloat) fabs(x))
 348 #endif
 349
 350
 351 /***
 352  *** IROUND: return (as an integer) float rounded to nearest integer
 353  ***/
 354 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
 355 static INLINE int iround(float f)
 356 {
 357    int r;
 358    __asm__ ("fstoi %1, %0" : "=f" (r) : "f" (f));
 359    return r;
 360 }
 361 #define IROUND(x)  iround(x)
 362 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
 363                         (!defined(__BEOS__) || (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
 364 static INLINE int iround(float f)
 365 {
 366    int r;
 367    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 368    return r;
 369 }
 370 #define IROUND(x)  iround(x)
 371 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
 372 static INLINE int iround(float f)
 373 {
 374    int r;
 375    _asm {
 376          fld f
 377          fistp r
 378         }
 379    return r;
 380 }
 381 #define IROUND(x)  iround(x)
 382 #elif defined(__WATCOMC__) && defined(__386__)
 383 long iround(float f);
 384 #pragma aux iround =                    \
 385         "push   eax"                        \
 386         "fistp  dword ptr [esp]"            \
 387         "pop    eax"                        \
 388         parm [8087]                         \
 389         value [eax]                         \
 390         modify exact [eax];
 391 #define IROUND(x)  iround(x)
 392 #else
 393 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 394 #endif
 395
 396
 397 /***
 398  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 399  ***/
 400 #ifdef DEBUG
 401 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 402 #else
 403 #define IROUND_POS(f) (IROUND(f))
 404 #endif
 405
 406
 407 /***
 408  *** IFLOOR: return (as an integer) floor of float
 409  ***/
 410 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 411 /*
 412  * IEEE floor for computers that round to nearest or even.
 413  * 'f' must be between -4194304 and 4194303.
 414  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 415  * but uses some IEEE specific tricks for better speed.
 416  * Contributed by Josh Vanderhoof
 417  */
 418 static INLINE int ifloor(float f)
 419 {
 420    int ai, bi;
 421    double af, bf;
 422    af = (3 << 22) + 0.5 + (double)f;
 423    bf = (3 << 22) + 0.5 - (double)f;
 424    /* GCC generates an extra fstp/fld without this. */
 425    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 426    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 427    return (ai - bi) >> 1;
 428 }
 429 #define IFLOOR(x)  ifloor(x)
 430 #elif defined(USE_IEEE)
 431 static INLINE int ifloor(float f)
 432 {
 433    int ai, bi;
 434    double af, bf;
 435    fi_type u;
 436
 437    af = (3 << 22) + 0.5 + (double)f;
 438    bf = (3 << 22) + 0.5 - (double)f;
 439    u.f = (float) af;  ai = u.i;
 440    u.f = (float) bf;  bi = u.i;
 441    return (ai - bi) >> 1;
 442 }
 443 #define IFLOOR(x)  ifloor(x)
 444 #else
 445 static INLINE int ifloor(float f)
 446 {
 447    int i = IROUND(f);
 448    return (i > f) ? i - 1 : i;
 449 }
 450 #define IFLOOR(x)  ifloor(x)
 451 #endif
 452
 453
 454 /***
 455  *** ICEIL: return (as an integer) ceiling of float
 456  ***/
 457 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 458 /*
 459  * IEEE ceil for computers that round to nearest or even.
 460  * 'f' must be between -4194304 and 4194303.
 461  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 462  * but uses some IEEE specific tricks for better speed.
 463  * Contributed by Josh Vanderhoof
 464  */
 465 static INLINE int iceil(float f)
 466 {
 467    int ai, bi;
 468    double af, bf;
 469    af = (3 << 22) + 0.5 + (double)f;
 470    bf = (3 << 22) + 0.5 - (double)f;
 471    /* GCC generates an extra fstp/fld without this. */
 472    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 473    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 474    return (ai - bi + 1) >> 1;
 475 }
 476 #define ICEIL(x)  iceil(x)
 477 #elif defined(USE_IEEE)
 478 static INLINE int iceil(float f)
 479 {
 480    int ai, bi;
 481    double af, bf;
 482    fi_type u;
 483    af = (3 << 22) + 0.5 + (double)f;
 484    bf = (3 << 22) + 0.5 - (double)f;
 485    u.f = (float) af; ai = u.i;
 486    u.f = (float) bf; bi = u.i;
 487    return (ai - bi + 1) >> 1;
 488 }
 489 #define ICEIL(x)  iceil(x)
 490 #else
 491 static INLINE int iceil(float f)
 492 {
 493    int i = IROUND(f);
 494    return (i < f) ? i + 1 : i;
 495 }
 496 #define ICEIL(x)  iceil(x)
 497 #endif
 498
 499
 500 /***
 501  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 502  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 503  ***/
 504 #if defined(USE_IEEE) && !defined(DEBUG)
 505 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 506 /* This function/macro is sensitive to precision.  Test very carefully
 507  * if you change it!
 508  */
 509 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 510         do {                                                            \
 511            fi_type __tmp;                                               \
 512            __tmp.f = (F);                                               \
 513            if (__tmp.i < 0)                                             \
 514               UB = (GLubyte) 0;                                         \
 515            else if (__tmp.i >= IEEE_0996)                               \
 516               UB = (GLubyte) 255;                                       \
 517            else {                                                       \
 518               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 519               UB = (GLubyte) __tmp.i;                                   \
 520            }                                                            \
 521         } while (0)
 522 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 523         do {                                                            \
 524            fi_type __tmp;                                               \
 525            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 526            UB = (GLubyte) __tmp.i;                                      \
 527         } while (0)
 528 #else
 529 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 530         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 531 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 532         ub = ((GLubyte) IROUND((f) * 255.0F))
 533 #endif
 534
 535
 536 /***
 537  *** COPY_FLOAT: copy a float from src to dest.
 538  ***/
 539 #define COPY_FLOAT( dst, src )          (dst) = (src)
 540
 541
 542 /***
 543  *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
 544  ***                  original mode to a temporary).
 545  *** END_FAST_MATH: Restore x86 FPU to original mode.
 546  ***/
 547 #if defined(__GNUC__) && defined(__i386__)
 548 /*
 549  * Set the x86 FPU control word to guarentee only 32 bits of precision
 550  * are stored in registers.  Allowing the FPU to store more introduces
 551  * differences between situations where numbers are pulled out of memory
 552  * vs. situations where the compiler is able to optimize register usage.
 553  *
 554  * In the worst case, we force the compiler to use a memory access to
 555  * truncate the float, by specifying the 'volatile' keyword.
 556  */
 557 /* Hardware default: All exceptions masked, extended double precision,
 558  * round to nearest (IEEE compliant):
 559  */
 560 #define DEFAULT_X86_FPU         0x037f
 561 /* All exceptions masked, single precision, round to nearest:
 562  */
 563 #define FAST_X86_FPU            0x003f
 564 /* The fldcw instruction will cause any pending FP exceptions to be
 565  * raised prior to entering the block, and we clear any pending
 566  * exceptions before exiting the block.  Hence, asm code has free
 567  * reign over the FPU while in the fast math block.
 568  */
 569 #if defined(NO_FAST_MATH)
 570 #define START_FAST_MATH(x)                                              \
 571 do {                                                                    \
 572    static GLuint mask = DEFAULT_X86_FPU;                                \
 573    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 574    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 575 } while (0)
 576 #else
 577 #define START_FAST_MATH(x)                                              \
 578 do {                                                                    \
 579    static GLuint mask = FAST_X86_FPU;                                   \
 580    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 581    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 582 } while (0)
 583 #endif
 584 /* Restore original FPU mode, and clear any exceptions that may have
 585  * occurred in the FAST_MATH block.
 586  */
 587 #define END_FAST_MATH(x)                                                \
 588 do {                                                                    \
 589    __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) );                     \
 590 } while (0)
 591
 592 #elif defined(__WATCOMC__) && defined(__386__)
 593 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 594 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 595 void _watcom_start_fast_math(unsigned short *x,unsigned short *mask);
 596 #pragma aux _watcom_start_fast_math =                                   \
 597    "fnstcw  word ptr [eax]"                                             \
 598    "fldcw   word ptr [ecx]"                                             \
 599    parm [eax] [ecx]                                                     \
 600    modify exact [];
 601 void _watcom_end_fast_math(unsigned short *x);
 602 #pragma aux _watcom_end_fast_math =                                     \
 603    "fnclex"                                                             \
 604    "fldcw   word ptr [eax]"                                             \
 605    parm [eax]                                                           \
 606    modify exact [];
 607 #if defined(NO_FAST_MATH)
 608 #define START_FAST_MATH(x)                                              \
 609 do {                                                                    \
 610    static GLushort mask = DEFAULT_X86_FPU;                                  \
 611    _watcom_start_fast_math(&x,&mask);                                   \
 612 } while (0)
 613 #else
 614 #define START_FAST_MATH(x)                                              \
 615 do {                                                                    \
 616    static GLushort mask = FAST_X86_FPU;                                 \
 617    _watcom_start_fast_math(&x,&mask);                                   \
 618 } while (0)
 619 #endif
 620 #define END_FAST_MATH(x)  _watcom_end_fast_math(&x)
 621 #else
 622 #define START_FAST_MATH(x)  x = 0
 623 #define END_FAST_MATH(x)  (void)(x)
 624 #endif
 625
 626
 627
 628 /**********************************************************************
 629  * Functions
 630  */
 631
 632 extern void *
 633 _mesa_malloc( size_t bytes );
 634
 635 extern void *
 636 _mesa_calloc( size_t bytes );
 637
 638 extern void
 639 _mesa_free( void *ptr );
 640
 641 extern void *
 642 _mesa_align_malloc( size_t bytes, unsigned long alignment );
 643
 644 extern void *
 645 _mesa_align_calloc( size_t bytes, unsigned long alignment );
 646
 647 extern void
 648 _mesa_align_free( void *ptr );
 649
 650 extern void *
 651 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 652
 653 extern void *
 654 _mesa_memcpy( void *dest, const void *src, size_t n );
 655
 656 extern void
 657 _mesa_memset( void *dst, int val, size_t n );
 658
 659 extern void
 660 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 661
 662 extern void
 663 _mesa_bzero( void *dst, size_t n );
 664
 665
 666 extern double
 667 _mesa_sin(double a);
 668
 669 extern double
 670 _mesa_cos(double a);
 671
 672 extern double
 673 _mesa_sqrtd(double x);
 674
 675 extern float
 676 _mesa_sqrtf(float x);
 677
 678 extern float
 679 _mesa_inv_sqrtf(float x);
 680
 681 extern double
 682 _mesa_pow(double x, double y);
 683
 684 extern float
 685 _mesa_log2(float x);
 686
 687 extern unsigned int
 688 _mesa_bitcount(unsigned int n);
 689
 690 extern GLhalfARB
 691 _mesa_float_to_half(float f);
 692
 693 extern float
 694 _mesa_half_to_float(GLhalfARB h);
 695
 696
 697 extern char *
 698 _mesa_getenv( const char *var );
 699
 700 extern char *
 701 _mesa_strstr( const char *haystack, const char *needle );
 702
 703 extern char *
 704 _mesa_strncat( char *dest, const char *src, size_t n );
 705
 706 extern char *
 707 _mesa_strcpy( char *dest, const char *src );
 708
 709 extern char *
 710 _mesa_strncpy( char *dest, const char *src, size_t n );
 711
 712 extern size_t
 713 _mesa_strlen( const char *s );
 714
 715 extern int
 716 _mesa_strcmp( const char *s1, const char *s2 );
 717
 718 extern int
 719 _mesa_strncmp( const char *s1, const char *s2, size_t n );
 720
 721 extern char *
 722 _mesa_strdup( const char *s );
 723
 724 extern int
 725 _mesa_atoi( const char *s );
 726
 727 extern double
 728 _mesa_strtod( const char *s, char **end );
 729
 730 extern int
 731 _mesa_sprintf( char *str, const char *fmt, ... );
 732
 733 extern void
 734 _mesa_printf( const char *fmtString, ... );
 735
 736
 737 extern void
 738 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 739
 740 extern void
 741 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 742
 743 extern void
 744 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 745
 746 extern void
 747 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 748
 749
 750 extern void
 751 _mesa_init_default_imports( __GLimports *imports, void *driverCtx );
 752
 753
 754 #ifdef __cplusplus
 755 }
 756 #endif
 757
 758
 759 #endif /* IMPORTS_H */