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