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 #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)  ( (GLubyte *) (A) + (uintptr_t) (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
 204 /**
 205  * Sometimes we treat GLfloats as GLints.  On x86 systems, moving a float
 206  * as a int (thereby using integer registers instead of FP registers) is
 207  * a performance win.  Typically, this can be done with ordinary casts.
 208  * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
 209  * these casts generate warnings.
 210  * The following union typedef is used to solve that.
 211  */
 212 typedef union { GLfloat f; GLint i; } fi_type;
 213
 214
 215
 216 /**********************************************************************
 217  * Math macros
 218  */
 219
 220 #define MAX_GLUSHORT    0xffff
 221 #define MAX_GLUINT      0xffffffff
 222
 223 #ifndef M_PI
 224 #define M_PI (3.1415926536)
 225 #endif
 226
 227 #ifndef M_E
 228 #define M_E (2.7182818284590452354)
 229 #endif
 230
 231
 232 /* XXX this is a bit of a hack needed for compilation within XFree86 */
 233 #ifndef FLT_MIN
 234 #define FLT_MIN (1.0e-37)
 235 #endif
 236
 237 /* Degrees to radians conversion: */
 238 #define DEG2RAD (M_PI/180.0)
 239
 240
 241 /***
 242  *** USE_IEEE: Determine if we're using IEEE floating point
 243  ***/
 244 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
 245     defined(__s390x__) || defined(__powerpc__) || \
 246     defined(__amd64__) || \
 247     defined(ia64) || defined(__ia64__) || \
 248     defined(__hppa__) || defined(hpux) || \
 249     defined(__mips) || defined(_MIPS_ARCH) || \
 250     defined(__arm__) || \
 251     (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
 252 #define USE_IEEE
 253 #define IEEE_ONE 0x3f800000
 254 #endif
 255
 256
 257 /***
 258  *** SQRTF: single-precision square root
 259  ***/
 260 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 261 #  define SQRTF(X)  _mesa_sqrtf(X)
 262 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 263 #  define SQRTF(X)  (float) xf86sqrt((float) (X))
 264 #else
 265 #  define SQRTF(X)  (float) sqrt((float) (X))
 266 #endif
 267
 268
 269 /***
 270  *** INV_SQRTF: single-precision inverse square root
 271  ***/
 272 #if 0
 273 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 274 #else
 275 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 276 #endif
 277
 278
 279 /***
 280  *** LOG2: Log base 2 of float
 281  ***/
 282 #ifdef USE_IEEE
 283 #if 0
 284 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 285  * Based on code from http://www.stereopsis.com/log2.html
 286  */
 287 static INLINE GLfloat LOG2(GLfloat x)
 288 {
 289    const GLfloat y = x * x * x * x;
 290    const GLuint ix = *((GLuint *) &y);
 291    const GLuint exp = (ix >> 23) & 0xFF;
 292    const GLint log2 = ((GLint) exp) - 127;
 293    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 294 }
 295 #endif
 296 /* Pretty fast, and accurate.
 297  * Based on code from http://www.flipcode.com/totd/
 298  */
 299 static INLINE GLfloat LOG2(GLfloat val)
 300 {
 301    fi_type num;
 302    GLint log_2;
 303    num.f = val;
 304    log_2 = ((num.i >> 23) & 255) - 128;
 305    num.i &= ~(255 << 23);
 306    num.i += 127 << 23;
 307    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 308    return num.f + log_2;
 309 }
 310 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 311 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 312 #else
 313 /*
 314  * NOTE: log_base_2(x) = log(x) / log(2)
 315  * NOTE: 1.442695 = 1/log(2).
 316  */
 317 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 318 #endif
 319
 320
 321 /***
 322  *** IS_INF_OR_NAN: test if float is infinite or NaN
 323  ***/
 324 #ifdef USE_IEEE
 325 static INLINE int IS_INF_OR_NAN( float x )
 326 {
 327    fi_type tmp;
 328    tmp.f = x;
 329    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 330 }
 331 #elif defined(isfinite)
 332 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 333 #elif defined(finite)
 334 #define IS_INF_OR_NAN(x)        (!finite(x))
 335 #elif defined(__VMS)
 336 #define IS_INF_OR_NAN(x)        (!finite(x))
 337 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 338 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 339 #else
 340 #define IS_INF_OR_NAN(x)        (!finite(x))
 341 #endif
 342
 343
 344 /***
 345  *** IS_NEGATIVE: test if float is negative
 346  ***/
 347 #if defined(USE_IEEE)
 348 static INLINE int GET_FLOAT_BITS( float x )
 349 {
 350    fi_type fi;
 351    fi.f = x;
 352    return fi.i;
 353 }
 354 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
 355 #else
 356 #define IS_NEGATIVE(x) (x < 0.0F)
 357 #endif
 358
 359
 360 /***
 361  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 362  ***/
 363 #if defined(USE_IEEE)
 364 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 365 #else
 366 /* Could just use (x*y<0) except for the flatshading requirements.
 367  * Maybe there's a better way?
 368  */
 369 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 370 #endif
 371
 372
 373 /***
 374  *** CEILF: ceiling of float
 375  *** FLOORF: floor of float
 376  *** FABSF: absolute value of float
 377  ***/
 378 #if defined(XFree86LOADER) && defined(IN_MODULE)
 379 #define CEILF(x)   ((GLfloat) xf86ceil(x))
 380 #define FLOORF(x)  ((GLfloat) xf86floor(x))
 381 #define FABSF(x)   ((GLfloat) xf86fabs(x))
 382 #elif defined(__gnu_linux__)
 383 /* C99 functions */
 384 #define CEILF(x)   ceilf(x)
 385 #define FLOORF(x)  floorf(x)
 386 #define FABSF(x)   fabsf(x)
 387 #else
 388 #define CEILF(x)   ((GLfloat) ceil(x))
 389 #define FLOORF(x)  ((GLfloat) floor(x))
 390 #define FABSF(x)   ((GLfloat) fabs(x))
 391 #endif
 392
 393
 394 /***
 395  *** IROUND: return (as an integer) float rounded to nearest integer
 396  ***/
 397 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
 398 static INLINE int iround(float f)
 399 {
 400    int r;
 401    __asm__ ("fstoi %1, %0" : "=f" (r) : "f" (f));
 402    return r;
 403 }
 404 #define IROUND(x)  iround(x)
 405 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
 406                         (!defined(__BEOS__) || (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
 407 static INLINE int iround(float f)
 408 {
 409    int r;
 410    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 411    return r;
 412 }
 413 #define IROUND(x)  iround(x)
 414 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
 415 static INLINE int iround(float f)
 416 {
 417    int r;
 418    _asm {
 419          fld f
 420          fistp r
 421         }
 422    return r;
 423 }
 424 #define IROUND(x)  iround(x)
 425 #elif defined(__WATCOMC__) && defined(__386__)
 426 long iround(float f);
 427 #pragma aux iround =                    \
 428         "push   eax"                        \
 429         "fistp  dword ptr [esp]"            \
 430         "pop    eax"                        \
 431         parm [8087]                         \
 432         value [eax]                         \
 433         modify exact [eax];
 434 #define IROUND(x)  iround(x)
 435 #else
 436 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 437 #endif
 438
 439
 440 /***
 441  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 442  ***/
 443 #ifdef DEBUG
 444 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 445 #else
 446 #define IROUND_POS(f) (IROUND(f))
 447 #endif
 448
 449
 450 /***
 451  *** IFLOOR: return (as an integer) floor of float
 452  ***/
 453 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 454 /*
 455  * IEEE floor for computers that round to nearest or even.
 456  * 'f' must be between -4194304 and 4194303.
 457  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 458  * but uses some IEEE specific tricks for better speed.
 459  * Contributed by Josh Vanderhoof
 460  */
 461 static INLINE int ifloor(float f)
 462 {
 463    int ai, bi;
 464    double af, bf;
 465    af = (3 << 22) + 0.5 + (double)f;
 466    bf = (3 << 22) + 0.5 - (double)f;
 467    /* GCC generates an extra fstp/fld without this. */
 468    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 469    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 470    return (ai - bi) >> 1;
 471 }
 472 #define IFLOOR(x)  ifloor(x)
 473 #elif defined(USE_IEEE)
 474 static INLINE int ifloor(float f)
 475 {
 476    int ai, bi;
 477    double af, bf;
 478    fi_type u;
 479
 480    af = (3 << 22) + 0.5 + (double)f;
 481    bf = (3 << 22) + 0.5 - (double)f;
 482    u.f = (float) af;  ai = u.i;
 483    u.f = (float) bf;  bi = u.i;
 484    return (ai - bi) >> 1;
 485 }
 486 #define IFLOOR(x)  ifloor(x)
 487 #else
 488 static INLINE int ifloor(float f)
 489 {
 490    int i = IROUND(f);
 491    return (i > f) ? i - 1 : i;
 492 }
 493 #define IFLOOR(x)  ifloor(x)
 494 #endif
 495
 496
 497 /***
 498  *** ICEIL: return (as an integer) ceiling of float
 499  ***/
 500 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 501 /*
 502  * IEEE ceil for computers that round to nearest or even.
 503  * 'f' must be between -4194304 and 4194303.
 504  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 505  * but uses some IEEE specific tricks for better speed.
 506  * Contributed by Josh Vanderhoof
 507  */
 508 static INLINE int iceil(float f)
 509 {
 510    int ai, bi;
 511    double af, bf;
 512    af = (3 << 22) + 0.5 + (double)f;
 513    bf = (3 << 22) + 0.5 - (double)f;
 514    /* GCC generates an extra fstp/fld without this. */
 515    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 516    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 517    return (ai - bi + 1) >> 1;
 518 }
 519 #define ICEIL(x)  iceil(x)
 520 #elif defined(USE_IEEE)
 521 static INLINE int iceil(float f)
 522 {
 523    int ai, bi;
 524    double af, bf;
 525    fi_type u;
 526    af = (3 << 22) + 0.5 + (double)f;
 527    bf = (3 << 22) + 0.5 - (double)f;
 528    u.f = (float) af; ai = u.i;
 529    u.f = (float) bf; bi = u.i;
 530    return (ai - bi + 1) >> 1;
 531 }
 532 #define ICEIL(x)  iceil(x)
 533 #else
 534 static INLINE int iceil(float f)
 535 {
 536    int i = IROUND(f);
 537    return (i < f) ? i + 1 : i;
 538 }
 539 #define ICEIL(x)  iceil(x)
 540 #endif
 541
 542
 543 /***
 544  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 545  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 546  ***/
 547 #if defined(USE_IEEE) && !defined(DEBUG)
 548 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 549 /* This function/macro is sensitive to precision.  Test very carefully
 550  * if you change it!
 551  */
 552 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 553         do {                                                            \
 554            fi_type __tmp;                                               \
 555            __tmp.f = (F);                                               \
 556            if (__tmp.i < 0)                                             \
 557               UB = (GLubyte) 0;                                         \
 558            else if (__tmp.i >= IEEE_0996)                               \
 559               UB = (GLubyte) 255;                                       \
 560            else {                                                       \
 561               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 562               UB = (GLubyte) __tmp.i;                                   \
 563            }                                                            \
 564         } while (0)
 565 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 566         do {                                                            \
 567            fi_type __tmp;                                               \
 568            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 569            UB = (GLubyte) __tmp.i;                                      \
 570         } while (0)
 571 #else
 572 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 573         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 574 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 575         ub = ((GLubyte) IROUND((f) * 255.0F))
 576 #endif
 577
 578
 579 /***
 580  *** COPY_FLOAT: copy a float from src to dest.
 581  ***/
 582 #define COPY_FLOAT( dst, src )          (dst) = (src)
 583
 584
 585 /***
 586  *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
 587  ***                  original mode to a temporary).
 588  *** END_FAST_MATH: Restore x86 FPU to original mode.
 589  ***/
 590 #if defined(__GNUC__) && defined(__i386__)
 591 /*
 592  * Set the x86 FPU control word to guarentee only 32 bits of precision
 593  * are stored in registers.  Allowing the FPU to store more introduces
 594  * differences between situations where numbers are pulled out of memory
 595  * vs. situations where the compiler is able to optimize register usage.
 596  *
 597  * In the worst case, we force the compiler to use a memory access to
 598  * truncate the float, by specifying the 'volatile' keyword.
 599  */
 600 /* Hardware default: All exceptions masked, extended double precision,
 601  * round to nearest (IEEE compliant):
 602  */
 603 #define DEFAULT_X86_FPU         0x037f
 604 /* All exceptions masked, single precision, round to nearest:
 605  */
 606 #define FAST_X86_FPU            0x003f
 607 /* The fldcw instruction will cause any pending FP exceptions to be
 608  * raised prior to entering the block, and we clear any pending
 609  * exceptions before exiting the block.  Hence, asm code has free
 610  * reign over the FPU while in the fast math block.
 611  */
 612 #if defined(NO_FAST_MATH)
 613 #define START_FAST_MATH(x)                                              \
 614 do {                                                                    \
 615    static GLuint mask = DEFAULT_X86_FPU;                                \
 616    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 617    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 618 } while (0)
 619 #else
 620 #define START_FAST_MATH(x)                                              \
 621 do {                                                                    \
 622    static GLuint mask = FAST_X86_FPU;                                   \
 623    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 624    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 625 } while (0)
 626 #endif
 627 /* Restore original FPU mode, and clear any exceptions that may have
 628  * occurred in the FAST_MATH block.
 629  */
 630 #define END_FAST_MATH(x)                                                \
 631 do {                                                                    \
 632    __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) );                     \
 633 } while (0)
 634
 635 #elif defined(__WATCOMC__) && defined(__386__)
 636 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 637 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 638 void _watcom_start_fast_math(unsigned short *x,unsigned short *mask);
 639 #pragma aux _watcom_start_fast_math =                                   \
 640    "fnstcw  word ptr [eax]"                                             \
 641    "fldcw   word ptr [ecx]"                                             \
 642    parm [eax] [ecx]                                                     \
 643    modify exact [];
 644 void _watcom_end_fast_math(unsigned short *x);
 645 #pragma aux _watcom_end_fast_math =                                     \
 646    "fnclex"                                                             \
 647    "fldcw   word ptr [eax]"                                             \
 648    parm [eax]                                                           \
 649    modify exact [];
 650 #if defined(NO_FAST_MATH)
 651 #define START_FAST_MATH(x)                                              \
 652 do {                                                                    \
 653    static GLushort mask = DEFAULT_X86_FPU;                                  \
 654    _watcom_start_fast_math(&x,&mask);                                   \
 655 } while (0)
 656 #else
 657 #define START_FAST_MATH(x)                                              \
 658 do {                                                                    \
 659    static GLushort mask = FAST_X86_FPU;                                 \
 660    _watcom_start_fast_math(&x,&mask);                                   \
 661 } while (0)
 662 #endif
 663 #define END_FAST_MATH(x)  _watcom_end_fast_math(&x)
 664 #else
 665 #define START_FAST_MATH(x)  x = 0
 666 #define END_FAST_MATH(x)  (void)(x)
 667 #endif
 668
 669
 670
 671 /**********************************************************************
 672  * Functions
 673  */
 674
 675 extern void *
 676 _mesa_malloc( size_t bytes );
 677
 678 extern void *
 679 _mesa_calloc( size_t bytes );
 680
 681 extern void
 682 _mesa_free( void *ptr );
 683
 684 extern void *
 685 _mesa_align_malloc( size_t bytes, unsigned long alignment );
 686
 687 extern void *
 688 _mesa_align_calloc( size_t bytes, unsigned long alignment );
 689
 690 extern void
 691 _mesa_align_free( void *ptr );
 692
 693 extern void *
 694 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 695
 696 extern void *
 697 _mesa_memcpy( void *dest, const void *src, size_t n );
 698
 699 extern void
 700 _mesa_memset( void *dst, int val, size_t n );
 701
 702 extern void
 703 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 704
 705 extern void
 706 _mesa_bzero( void *dst, size_t n );
 707
 708
 709 extern double
 710 _mesa_sin(double a);
 711
 712 extern double
 713 _mesa_cos(double a);
 714
 715 extern double
 716 _mesa_sqrtd(double x);
 717
 718 extern float
 719 _mesa_sqrtf(float x);
 720
 721 extern float
 722 _mesa_inv_sqrtf(float x);
 723
 724 extern double
 725 _mesa_pow(double x, double y);
 726
 727 extern float
 728 _mesa_log2(float x);
 729
 730 extern unsigned int
 731 _mesa_bitcount(unsigned int n);
 732
 733 extern GLhalfARB
 734 _mesa_float_to_half(float f);
 735
 736 extern float
 737 _mesa_half_to_float(GLhalfARB h);
 738
 739
 740 extern char *
 741 _mesa_getenv( const char *var );
 742
 743 extern char *
 744 _mesa_strstr( const char *haystack, const char *needle );
 745
 746 extern char *
 747 _mesa_strncat( char *dest, const char *src, size_t n );
 748
 749 extern char *
 750 _mesa_strcpy( char *dest, const char *src );
 751
 752 extern char *
 753 _mesa_strncpy( char *dest, const char *src, size_t n );
 754
 755 extern size_t
 756 _mesa_strlen( const char *s );
 757
 758 extern int
 759 _mesa_strcmp( const char *s1, const char *s2 );
 760
 761 extern int
 762 _mesa_strncmp( const char *s1, const char *s2, size_t n );
 763
 764 extern char *
 765 _mesa_strdup( const char *s );
 766
 767 extern int
 768 _mesa_atoi( const char *s );
 769
 770 extern double
 771 _mesa_strtod( const char *s, char **end );
 772
 773 extern int
 774 _mesa_sprintf( char *str, const char *fmt, ... );
 775
 776 extern void
 777 _mesa_printf( const char *fmtString, ... );
 778
 779
 780 extern void
 781 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 782
 783 extern void
 784 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 785
 786 extern void
 787 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 788
 789 extern void
 790 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 791
 792
 793 extern void
 794 _mesa_init_default_imports( __GLimports *imports, void *driverCtx );
 795
 796
 797 #ifdef __cplusplus
 798 }
 799 #endif
 800
 801
 802 #endif /* IMPORTS_H */