src/mesa/main/imports.h

   1 /*
   2  * Mesa 3-D graphics library
   3  * Version:  6.5.2
   4  *
   5  * Copyright (C) 1999-2006  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 #include <GL/internal/glcore.h>
  43
  44
  45 #ifdef __cplusplus
  46 extern "C" {
  47 #endif
  48
  49
  50 /**********************************************************************/
  51 /** \name General macros */
  52 /*@{*/
  53
  54 #ifndef NULL
  55 #define NULL 0
  56 #endif
  57
  58
  59 /** gcc -pedantic warns about long string literals, LONGSTRING silences that */
  60 #if !defined(__GNUC__) || (__GNUC__ < 2) || \
  61     ((__GNUC__ == 2) && (__GNUC_MINOR__ <= 7))
  62 # define LONGSTRING
  63 #else
  64 # define LONGSTRING __extension__
  65 #endif
  66
  67 /*@}*/
  68
  69
  70 /**********************************************************************/
  71 /** Memory macros */
  72 /*@{*/
  73
  74 /** Allocate \p BYTES bytes */
  75 #define MALLOC(BYTES)      _mesa_malloc(BYTES)
  76 /** Allocate and zero \p BYTES bytes */
  77 #define CALLOC(BYTES)      _mesa_calloc(BYTES)
  78 /** Allocate a structure of type \p T */
  79 #define MALLOC_STRUCT(T)   (struct T *) _mesa_malloc(sizeof(struct T))
  80 /** Allocate and zero a structure of type \p T */
  81 #define CALLOC_STRUCT(T)   (struct T *) _mesa_calloc(sizeof(struct T))
  82 /** Free memory */
  83 #define FREE(PTR)          _mesa_free(PTR)
  84
  85 /** Allocate \p BYTES aligned at \p N bytes */
  86 #define ALIGN_MALLOC(BYTES, N)     _mesa_align_malloc(BYTES, N)
  87 /** Allocate and zero \p BYTES bytes aligned at \p N bytes */
  88 #define ALIGN_CALLOC(BYTES, N)     _mesa_align_calloc(BYTES, N)
  89 /** Allocate a structure of type \p T aligned at \p N bytes */
  90 #define ALIGN_MALLOC_STRUCT(T, N)  (struct T *) _mesa_align_malloc(sizeof(struct T), N)
  91 /** Allocate and zero a structure of type \p T aligned at \p N bytes */
  92 #define ALIGN_CALLOC_STRUCT(T, N)  (struct T *) _mesa_align_calloc(sizeof(struct T), N)
  93 /** Free aligned memory */
  94 #define ALIGN_FREE(PTR)            _mesa_align_free(PTR)
  95
  96 /** Copy \p BYTES bytes from \p SRC into \p DST */
  97 #define MEMCPY( DST, SRC, BYTES)   _mesa_memcpy(DST, SRC, BYTES)
  98 /** Set \p N bytes in \p DST to \p VAL */
  99 #define MEMSET( DST, VAL, N )      _mesa_memset(DST, VAL, N)
 100
 101 /*@}*/
 102
 103
 104 /*
 105  * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
 106  * as offsets into buffer stores.  Since the vertex array pointer and
 107  * buffer store pointer are both pointers and we need to add them, we use
 108  * this macro.
 109  * Both pointers/offsets are expressed in bytes.
 110  */
 111 #define ADD_POINTERS(A, B)  ( (GLubyte *) (A) + (uintptr_t) (B) )
 112
 113
 114 /**
 115  * Sometimes we treat GLfloats as GLints.  On x86 systems, moving a float
 116  * as a int (thereby using integer registers instead of FP registers) is
 117  * a performance win.  Typically, this can be done with ordinary casts.
 118  * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
 119  * these casts generate warnings.
 120  * The following union typedef is used to solve that.
 121  */
 122 typedef union { GLfloat f; GLint i; } fi_type;
 123
 124
 125
 126 /**********************************************************************
 127  * Math macros
 128  */
 129
 130 #define MAX_GLUSHORT    0xffff
 131 #define MAX_GLUINT      0xffffffff
 132
 133 #ifndef M_PI
 134 #define M_PI (3.1415926536)
 135 #endif
 136
 137 #ifndef M_E
 138 #define M_E (2.7182818284590452354)
 139 #endif
 140
 141 #ifndef FLT_MAX_EXP
 142 #define FLT_MAX_EXP 128
 143 #endif
 144
 145 /* XXX this is a bit of a hack needed for compilation within XFree86 */
 146 #ifndef FLT_MIN
 147 #define FLT_MIN (1.0e-37)
 148 #endif
 149
 150 /* Degrees to radians conversion: */
 151 #define DEG2RAD (M_PI/180.0)
 152
 153
 154 /***
 155  *** USE_IEEE: Determine if we're using IEEE floating point
 156  ***/
 157 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
 158     defined(__s390x__) || defined(__powerpc__) || \
 159     defined(__amd64__) || \
 160     defined(ia64) || defined(__ia64__) || \
 161     defined(__hppa__) || defined(hpux) || \
 162     defined(__mips) || defined(_MIPS_ARCH) || \
 163     defined(__arm__) || \
 164     defined(__sh__) || defined(__m32r__) || \
 165     (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
 166 #define USE_IEEE
 167 #define IEEE_ONE 0x3f800000
 168 #endif
 169
 170
 171 /***
 172  *** SQRTF: single-precision square root
 173  ***/
 174 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 175 #  define SQRTF(X)  _mesa_sqrtf(X)
 176 #elif defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
 177 #  define SQRTF(X)  (float) xf86sqrt((float) (X))
 178 #else
 179 #  define SQRTF(X)  (float) sqrt((float) (X))
 180 #endif
 181
 182
 183 /***
 184  *** INV_SQRTF: single-precision inverse square root
 185  ***/
 186 #if 0
 187 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 188 #else
 189 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 190 #endif
 191
 192
 193 /***
 194  *** LOG2: Log base 2 of float
 195  ***/
 196 #ifdef USE_IEEE
 197 #if 0
 198 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 199  * Based on code from http://www.stereopsis.com/log2.html
 200  */
 201 static INLINE GLfloat LOG2(GLfloat x)
 202 {
 203    const GLfloat y = x * x * x * x;
 204    const GLuint ix = *((GLuint *) &y);
 205    const GLuint exp = (ix >> 23) & 0xFF;
 206    const GLint log2 = ((GLint) exp) - 127;
 207    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 208 }
 209 #endif
 210 /* Pretty fast, and accurate.
 211  * Based on code from http://www.flipcode.com/totd/
 212  */
 213 static INLINE GLfloat LOG2(GLfloat val)
 214 {
 215    fi_type num;
 216    GLint log_2;
 217    num.f = val;
 218    log_2 = ((num.i >> 23) & 255) - 128;
 219    num.i &= ~(255 << 23);
 220    num.i += 127 << 23;
 221    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 222    return num.f + log_2;
 223 }
 224 #elif defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
 225 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 226 #else
 227 /*
 228  * NOTE: log_base_2(x) = log(x) / log(2)
 229  * NOTE: 1.442695 = 1/log(2).
 230  */
 231 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 232 #endif
 233
 234
 235 /***
 236  *** IS_INF_OR_NAN: test if float is infinite or NaN
 237  ***/
 238 #ifdef USE_IEEE
 239 static INLINE int IS_INF_OR_NAN( float x )
 240 {
 241    fi_type tmp;
 242    tmp.f = x;
 243    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 244 }
 245 #elif defined(isfinite)
 246 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 247 #elif defined(finite)
 248 #define IS_INF_OR_NAN(x)        (!finite(x))
 249 #elif defined(__VMS)
 250 #define IS_INF_OR_NAN(x)        (!finite(x))
 251 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 252 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 253 #else
 254 #define IS_INF_OR_NAN(x)        (!finite(x))
 255 #endif
 256
 257
 258 /***
 259  *** IS_NEGATIVE: test if float is negative
 260  ***/
 261 #if defined(USE_IEEE)
 262 static INLINE int GET_FLOAT_BITS( float x )
 263 {
 264    fi_type fi;
 265    fi.f = x;
 266    return fi.i;
 267 }
 268 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
 269 #else
 270 #define IS_NEGATIVE(x) (x < 0.0F)
 271 #endif
 272
 273
 274 /***
 275  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 276  ***/
 277 #if defined(USE_IEEE)
 278 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 279 #else
 280 /* Could just use (x*y<0) except for the flatshading requirements.
 281  * Maybe there's a better way?
 282  */
 283 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 284 #endif
 285
 286
 287 /***
 288  *** CEILF: ceiling of float
 289  *** FLOORF: floor of float
 290  *** FABSF: absolute value of float
 291  *** LOGF: the natural logarithm (base e) of the value
 292  *** EXPF: raise e to the value
 293  *** LDEXPF: multiply value by an integral power of two
 294  *** FREXPF: extract mantissa and exponent from value
 295  ***/
 296 #if defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
 297 #define CEILF(x)   ((GLfloat) xf86ceil(x))
 298 #define FLOORF(x)  ((GLfloat) xf86floor(x))
 299 #define FABSF(x)   ((GLfloat) xf86fabs(x))
 300 #define LOGF(x)    ((GLfloat) xf86log(x))
 301 #define EXPF(x)    ((GLfloat) xf86exp(x))
 302 #define LDEXPF(x,y)   ((GLfloat) xf86ldexp(x,y))
 303 #define FREXPF(x,y)   ((GLfloat) xf86frexp(x,y))
 304 #elif defined(__gnu_linux__)
 305 /* C99 functions */
 306 #define CEILF(x)   ceilf(x)
 307 #define FLOORF(x)  floorf(x)
 308 #define FABSF(x)   fabsf(x)
 309 #define LOGF(x)    logf(x)
 310 #define EXPF(x)    expf(x)
 311 #define LDEXPF(x,y)  ldexpf(x,y)
 312 #define FREXPF(x,y)  frexpf(x,y)
 313 #else
 314 #define CEILF(x)   ((GLfloat) ceil(x))
 315 #define FLOORF(x)  ((GLfloat) floor(x))
 316 #define FABSF(x)   ((GLfloat) fabs(x))
 317 #define LOGF(x)    ((GLfloat) log(x))
 318 #define EXPF(x)    ((GLfloat) exp(x))
 319 #define LDEXPF(x,y)  ((GLfloat) ldexp(x,y))
 320 #define FREXPF(x,y)  ((GLfloat) frexp(x,y))
 321 #endif
 322
 323
 324 /***
 325  *** IROUND: return (as an integer) float rounded to nearest integer
 326  ***/
 327 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
 328 static INLINE int iround(float f)
 329 {
 330    int r;
 331    __asm__ ("fstoi %1, %0" : "=f" (r) : "f" (f));
 332    return r;
 333 }
 334 #define IROUND(x)  iround(x)
 335 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
 336                         (!defined(__BEOS__) || (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
 337 static INLINE int iround(float f)
 338 {
 339    int r;
 340    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 341    return r;
 342 }
 343 #define IROUND(x)  iround(x)
 344 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
 345 static INLINE int iround(float f)
 346 {
 347    int r;
 348    _asm {
 349          fld f
 350          fistp r
 351         }
 352    return r;
 353 }
 354 #define IROUND(x)  iround(x)
 355 #elif defined(__WATCOMC__) && defined(__386__)
 356 long iround(float f);
 357 #pragma aux iround =                    \
 358         "push   eax"                        \
 359         "fistp  dword ptr [esp]"            \
 360         "pop    eax"                        \
 361         parm [8087]                         \
 362         value [eax]                         \
 363         modify exact [eax];
 364 #define IROUND(x)  iround(x)
 365 #else
 366 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 367 #endif
 368
 369
 370 /***
 371  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 372  ***/
 373 #ifdef DEBUG
 374 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 375 #else
 376 #define IROUND_POS(f) (IROUND(f))
 377 #endif
 378
 379
 380 /***
 381  *** IFLOOR: return (as an integer) floor of float
 382  ***/
 383 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 384 /*
 385  * IEEE floor for computers that round to nearest or even.
 386  * 'f' must be between -4194304 and 4194303.
 387  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 388  * but uses some IEEE specific tricks for better speed.
 389  * Contributed by Josh Vanderhoof
 390  */
 391 static INLINE int ifloor(float f)
 392 {
 393    int ai, bi;
 394    double af, bf;
 395    af = (3 << 22) + 0.5 + (double)f;
 396    bf = (3 << 22) + 0.5 - (double)f;
 397    /* GCC generates an extra fstp/fld without this. */
 398    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 399    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 400    return (ai - bi) >> 1;
 401 }
 402 #define IFLOOR(x)  ifloor(x)
 403 #elif defined(USE_IEEE)
 404 static INLINE int ifloor(float f)
 405 {
 406    int ai, bi;
 407    double af, bf;
 408    fi_type u;
 409
 410    af = (3 << 22) + 0.5 + (double)f;
 411    bf = (3 << 22) + 0.5 - (double)f;
 412    u.f = (float) af;  ai = u.i;
 413    u.f = (float) bf;  bi = u.i;
 414    return (ai - bi) >> 1;
 415 }
 416 #define IFLOOR(x)  ifloor(x)
 417 #else
 418 static INLINE int ifloor(float f)
 419 {
 420    int i = IROUND(f);
 421    return (i > f) ? i - 1 : i;
 422 }
 423 #define IFLOOR(x)  ifloor(x)
 424 #endif
 425
 426
 427 /***
 428  *** ICEIL: return (as an integer) ceiling of float
 429  ***/
 430 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 431 /*
 432  * IEEE ceil for computers that round to nearest or even.
 433  * 'f' must be between -4194304 and 4194303.
 434  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 435  * but uses some IEEE specific tricks for better speed.
 436  * Contributed by Josh Vanderhoof
 437  */
 438 static INLINE int iceil(float f)
 439 {
 440    int ai, bi;
 441    double af, bf;
 442    af = (3 << 22) + 0.5 + (double)f;
 443    bf = (3 << 22) + 0.5 - (double)f;
 444    /* GCC generates an extra fstp/fld without this. */
 445    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 446    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 447    return (ai - bi + 1) >> 1;
 448 }
 449 #define ICEIL(x)  iceil(x)
 450 #elif defined(USE_IEEE)
 451 static INLINE int iceil(float f)
 452 {
 453    int ai, bi;
 454    double af, bf;
 455    fi_type u;
 456    af = (3 << 22) + 0.5 + (double)f;
 457    bf = (3 << 22) + 0.5 - (double)f;
 458    u.f = (float) af; ai = u.i;
 459    u.f = (float) bf; bi = u.i;
 460    return (ai - bi + 1) >> 1;
 461 }
 462 #define ICEIL(x)  iceil(x)
 463 #else
 464 static INLINE int iceil(float f)
 465 {
 466    int i = IROUND(f);
 467    return (i < f) ? i + 1 : i;
 468 }
 469 #define ICEIL(x)  iceil(x)
 470 #endif
 471
 472
 473 /***
 474  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 475  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 476  ***/
 477 #if defined(USE_IEEE) && !defined(DEBUG)
 478 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 479 /* This function/macro is sensitive to precision.  Test very carefully
 480  * if you change it!
 481  */
 482 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 483         do {                                                            \
 484            fi_type __tmp;                                               \
 485            __tmp.f = (F);                                               \
 486            if (__tmp.i < 0)                                             \
 487               UB = (GLubyte) 0;                                         \
 488            else if (__tmp.i >= IEEE_0996)                               \
 489               UB = (GLubyte) 255;                                       \
 490            else {                                                       \
 491               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 492               UB = (GLubyte) __tmp.i;                                   \
 493            }                                                            \
 494         } while (0)
 495 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 496         do {                                                            \
 497            fi_type __tmp;                                               \
 498            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 499            UB = (GLubyte) __tmp.i;                                      \
 500         } while (0)
 501 #else
 502 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 503         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 504 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 505         ub = ((GLubyte) IROUND((f) * 255.0F))
 506 #endif
 507
 508
 509 /***
 510  *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
 511  ***                  original mode to a temporary).
 512  *** END_FAST_MATH: Restore x86 FPU to original mode.
 513  ***/
 514 #if defined(__GNUC__) && defined(__i386__)
 515 /*
 516  * Set the x86 FPU control word to guarentee only 32 bits of precision
 517  * are stored in registers.  Allowing the FPU to store more introduces
 518  * differences between situations where numbers are pulled out of memory
 519  * vs. situations where the compiler is able to optimize register usage.
 520  *
 521  * In the worst case, we force the compiler to use a memory access to
 522  * truncate the float, by specifying the 'volatile' keyword.
 523  */
 524 /* Hardware default: All exceptions masked, extended double precision,
 525  * round to nearest (IEEE compliant):
 526  */
 527 #define DEFAULT_X86_FPU         0x037f
 528 /* All exceptions masked, single precision, round to nearest:
 529  */
 530 #define FAST_X86_FPU            0x003f
 531 /* The fldcw instruction will cause any pending FP exceptions to be
 532  * raised prior to entering the block, and we clear any pending
 533  * exceptions before exiting the block.  Hence, asm code has free
 534  * reign over the FPU while in the fast math block.
 535  */
 536 #if defined(NO_FAST_MATH)
 537 #define START_FAST_MATH(x)                                              \
 538 do {                                                                    \
 539    static GLuint mask = DEFAULT_X86_FPU;                                \
 540    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 541    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 542 } while (0)
 543 #else
 544 #define START_FAST_MATH(x)                                              \
 545 do {                                                                    \
 546    static GLuint mask = FAST_X86_FPU;                                   \
 547    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 548    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 549 } while (0)
 550 #endif
 551 /* Restore original FPU mode, and clear any exceptions that may have
 552  * occurred in the FAST_MATH block.
 553  */
 554 #define END_FAST_MATH(x)                                                \
 555 do {                                                                    \
 556    __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) );                     \
 557 } while (0)
 558
 559 #elif defined(__WATCOMC__) && defined(__386__)
 560 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 561 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 562 void _watcom_start_fast_math(unsigned short *x,unsigned short *mask);
 563 #pragma aux _watcom_start_fast_math =                                   \
 564    "fnstcw  word ptr [eax]"                                             \
 565    "fldcw   word ptr [ecx]"                                             \
 566    parm [eax] [ecx]                                                     \
 567    modify exact [];
 568 void _watcom_end_fast_math(unsigned short *x);
 569 #pragma aux _watcom_end_fast_math =                                     \
 570    "fnclex"                                                             \
 571    "fldcw   word ptr [eax]"                                             \
 572    parm [eax]                                                           \
 573    modify exact [];
 574 #if defined(NO_FAST_MATH)
 575 #define START_FAST_MATH(x)                                              \
 576 do {                                                                    \
 577    static GLushort mask = DEFAULT_X86_FPU;                                  \
 578    _watcom_start_fast_math(&x,&mask);                                   \
 579 } while (0)
 580 #else
 581 #define START_FAST_MATH(x)                                              \
 582 do {                                                                    \
 583    static GLushort mask = FAST_X86_FPU;                                 \
 584    _watcom_start_fast_math(&x,&mask);                                   \
 585 } while (0)
 586 #endif
 587 #define END_FAST_MATH(x)  _watcom_end_fast_math(&x)
 588
 589 #elif defined(_MSC_VER) && defined(_M_IX86)
 590 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 591 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 592 #if defined(NO_FAST_MATH)
 593 #define START_FAST_MATH(x) do {\
 594         static GLuint mask = DEFAULT_X86_FPU;\
 595         __asm fnstcw word ptr [x]\
 596         __asm fldcw word ptr [mask]\
 597 } while(0)
 598 #else
 599 #define START_FAST_MATH(x) do {\
 600         static GLuint mask = FAST_X86_FPU;\
 601         __asm fnstcw word ptr [x]\
 602         __asm fldcw word ptr [mask]\
 603 } while(0)
 604 #endif
 605 #define END_FAST_MATH(x) do {\
 606         __asm fnclex\
 607         __asm fldcw word ptr [x]\
 608 } while(0)
 609
 610 #else
 611 #define START_FAST_MATH(x)  x = 0
 612 #define END_FAST_MATH(x)  (void)(x)
 613 #endif
 614
 615
 616 /**
 617  * Return 1 if this is a little endian machine, 0 if big endian.
 618  */
 619 static INLINE GLboolean
 620 _mesa_little_endian(void)
 621 {
 622    const GLuint ui = 1; /* intentionally not static */
 623    return *((const GLubyte *) &ui);
 624 }
 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_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
 652                     unsigned long alignment);
 653
 654 extern void *
 655 _mesa_exec_malloc( GLuint size );
 656
 657 extern void
 658 _mesa_exec_free( void *addr );
 659
 660 extern void *
 661 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 662
 663 extern void *
 664 _mesa_memcpy( void *dest, const void *src, size_t n );
 665
 666 extern void
 667 _mesa_memset( void *dst, int val, size_t n );
 668
 669 extern void
 670 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 671
 672 extern void
 673 _mesa_bzero( void *dst, size_t n );
 674
 675 extern int
 676 _mesa_memcmp( const void *s1, const void *s2, size_t n );
 677
 678 extern double
 679 _mesa_sin(double a);
 680
 681 extern float
 682 _mesa_sinf(float a);
 683
 684 extern double
 685 _mesa_cos(double a);
 686
 687 extern float
 688 _mesa_asinf(float x);
 689
 690 extern float
 691 _mesa_atanf(float x);
 692
 693 extern double
 694 _mesa_sqrtd(double x);
 695
 696 extern float
 697 _mesa_sqrtf(float x);
 698
 699 extern float
 700 _mesa_inv_sqrtf(float x);
 701
 702 extern double
 703 _mesa_pow(double x, double y);
 704
 705 extern int
 706 _mesa_ffs(int i);
 707
 708 extern unsigned int
 709 _mesa_bitcount(unsigned int n);
 710
 711 extern GLhalfARB
 712 _mesa_float_to_half(float f);
 713
 714 extern float
 715 _mesa_half_to_float(GLhalfARB h);
 716
 717
 718 extern void *
 719 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
 720                int (*compar)(const void *, const void *) );
 721
 722 extern char *
 723 _mesa_getenv( const char *var );
 724
 725 extern char *
 726 _mesa_strstr( const char *haystack, const char *needle );
 727
 728 extern char *
 729 _mesa_strncat( char *dest, const char *src, size_t n );
 730
 731 extern char *
 732 _mesa_strcpy( char *dest, const char *src );
 733
 734 extern char *
 735 _mesa_strncpy( char *dest, const char *src, size_t n );
 736
 737 extern size_t
 738 _mesa_strlen( const char *s );
 739
 740 extern int
 741 _mesa_strcmp( const char *s1, const char *s2 );
 742
 743 extern int
 744 _mesa_strncmp( const char *s1, const char *s2, size_t n );
 745
 746 extern char *
 747 _mesa_strdup( const char *s );
 748
 749 extern int
 750 _mesa_atoi( const char *s );
 751
 752 extern double
 753 _mesa_strtod( const char *s, char **end );
 754
 755 extern int
 756 _mesa_sprintf( char *str, const char *fmt, ... );
 757
 758 extern void
 759 _mesa_printf( const char *fmtString, ... );
 760
 761 extern int
 762 _mesa_vsprintf( char *str, const char *fmt, va_list args );
 763
 764
 765 extern void
 766 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 767
 768 extern void
 769 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 770
 771 extern void
 772 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 773
 774 extern void
 775 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 776
 777 extern void
 778 _mesa_exit( int status );
 779
 780
 781 extern void
 782 _mesa_init_default_imports( __GLimports *imports, void *driverCtx );
 783
 784
 785 #ifdef __cplusplus
 786 }
 787 #endif
 788
 789
 790 #endif /* IMPORTS_H */