src/mesa/main/imports.h

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