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)  ( (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__) || \
 227     defined(__AMD64__) || \
 228     defined(ia64) || defined(__ia64__) || \
 229     defined(__hppa__) || defined(hpux) || \
 230     defined(__mips) || defined(_MIPS_ARCH) || \
 231     (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
 232 #define USE_IEEE
 233 #define IEEE_ONE 0x3f800000
 234 #endif
 235
 236
 237 /***
 238  *** SQRTF: single-precision square root
 239  ***/
 240 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 241 #  define SQRTF(X)  _mesa_sqrtf(X)
 242 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 243 #  define SQRTF(X)  (float) xf86sqrt((float) (X))
 244 #else
 245 #  define SQRTF(X)  (float) sqrt((float) (X))
 246 #endif
 247
 248
 249 /***
 250  *** INV_SQRTF: single-precision inverse square root
 251  ***/
 252 #if 0
 253 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 254 #else
 255 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 256 #endif
 257
 258
 259 /***
 260  *** LOG2: Log base 2 of float
 261  ***/
 262 #ifdef USE_IEEE
 263 #if 0
 264 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 265  * Based on code from http://www.stereopsis.com/log2.html
 266  */
 267 static INLINE GLfloat LOG2(GLfloat x)
 268 {
 269    const GLfloat y = x * x * x * x;
 270    const GLuint ix = *((GLuint *) &y);
 271    const GLuint exp = (ix >> 23) & 0xFF;
 272    const GLint log2 = ((GLint) exp) - 127;
 273    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 274 }
 275 #endif
 276 /* Pretty fast, and accurate.
 277  * Based on code from http://www.flipcode.com/totd/
 278  */
 279 static INLINE GLfloat LOG2(GLfloat val)
 280 {
 281    fi_type num;
 282    GLint log_2;
 283    num.f = val;
 284    log_2 = ((num.i >> 23) & 255) - 128;
 285    num.i &= ~(255 << 23);
 286    num.i += 127 << 23;
 287    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 288    return num.f + log_2;
 289 }
 290 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 291 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 292 #else
 293 /*
 294  * NOTE: log_base_2(x) = log(x) / log(2)
 295  * NOTE: 1.442695 = 1/log(2).
 296  */
 297 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 298 #endif
 299
 300
 301 /***
 302  *** IS_INF_OR_NAN: test if float is infinite or NaN
 303  ***/
 304 #ifdef USE_IEEE
 305 static INLINE int IS_INF_OR_NAN( float x )
 306 {
 307    fi_type tmp;
 308    tmp.f = x;
 309    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 310 }
 311 #elif defined(isfinite)
 312 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 313 #elif defined(finite)
 314 #define IS_INF_OR_NAN(x)        (!finite(x))
 315 #elif defined(__VMS)
 316 #define IS_INF_OR_NAN(x)        (!finite(x))
 317 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 318 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 319 #else
 320 #define IS_INF_OR_NAN(x)        (!finite(x))
 321 #endif
 322
 323
 324 /***
 325  *** IS_NEGATIVE: test if float is negative
 326  ***/
 327 #if defined(USE_IEEE)
 328 #define GET_FLOAT_BITS(x) ((fi_type *) &(x))->i
 329 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) & (1<<31))
 330 #else
 331 #define IS_NEGATIVE(x) (x < 0.0F)
 332 #endif
 333
 334
 335 /***
 336  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 337  ***/
 338 #if defined(USE_IEEE)
 339 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 340 #else
 341 /* Could just use (x*y<0) except for the flatshading requirements.
 342  * Maybe there's a better way?
 343  */
 344 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 345 #endif
 346
 347
 348 /***
 349  *** CEILF: ceiling of float
 350  *** FLOORF: floor of float
 351  *** FABSF: absolute value of float
 352  ***/
 353 #if defined(XFree86LOADER) && defined(IN_MODULE)
 354 #define CEILF(x)   ((GLfloat) xf86ceil(x))
 355 #define FLOORF(x)  ((GLfloat) xf86floor(x))
 356 #define FABSF(x)   ((GLfloat) xf86fabs(x))
 357 #elif defined(__gnu_linux__)
 358 /* C99 functions */
 359 #define CEILF(x)   ceilf(x)
 360 #define FLOORF(x)  floorf(x)
 361 #define FABSF(x)   fabsf(x)
 362 #else
 363 #define CEILF(x)   ((GLfloat) ceil(x))
 364 #define FLOORF(x)  ((GLfloat) floor(x))
 365 #define FABSF(x)   ((GLfloat) fabs(x))
 366 #endif
 367
 368
 369 /***
 370  *** IROUND: return (as an integer) float rounded to nearest integer
 371  ***/
 372 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
 373 static INLINE int iround(float f)
 374 {
 375    int r;
 376    __asm__ ("fstoi %1, %0" : "=f" (r) : "f" (f));
 377    return r;
 378 }
 379 #define IROUND(x)  iround(x)
 380 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
 381                         (!defined(__BEOS__) || (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
 382 static INLINE int iround(float f)
 383 {
 384    int r;
 385    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 386    return r;
 387 }
 388 #define IROUND(x)  iround(x)
 389 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
 390 static INLINE int iround(float f)
 391 {
 392    int r;
 393    _asm {
 394          fld f
 395          fistp r
 396         }
 397    return r;
 398 }
 399 #define IROUND(x)  iround(x)
 400 #elif defined(__WATCOMC__) && defined(__386__)
 401 long iround(float f);
 402 #pragma aux iround =                    \
 403         "push   eax"                        \
 404         "fistp  dword ptr [esp]"            \
 405         "pop    eax"                        \
 406         parm [8087]                         \
 407         value [eax]                         \
 408         modify exact [eax];
 409 #define IROUND(x)  iround(x)
 410 #else
 411 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 412 #endif
 413
 414
 415 /***
 416  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 417  ***/
 418 #ifdef DEBUG
 419 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 420 #else
 421 #define IROUND_POS(f) (IROUND(f))
 422 #endif
 423
 424
 425 /***
 426  *** IFLOOR: return (as an integer) floor of float
 427  ***/
 428 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 429 /*
 430  * IEEE floor for computers that round to nearest or even.
 431  * 'f' must be between -4194304 and 4194303.
 432  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 433  * but uses some IEEE specific tricks for better speed.
 434  * Contributed by Josh Vanderhoof
 435  */
 436 static INLINE int ifloor(float f)
 437 {
 438    int ai, bi;
 439    double af, bf;
 440    af = (3 << 22) + 0.5 + (double)f;
 441    bf = (3 << 22) + 0.5 - (double)f;
 442    /* GCC generates an extra fstp/fld without this. */
 443    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 444    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 445    return (ai - bi) >> 1;
 446 }
 447 #define IFLOOR(x)  ifloor(x)
 448 #elif defined(USE_IEEE)
 449 static INLINE int ifloor(float f)
 450 {
 451    int ai, bi;
 452    double af, bf;
 453    fi_type u;
 454
 455    af = (3 << 22) + 0.5 + (double)f;
 456    bf = (3 << 22) + 0.5 - (double)f;
 457    u.f = (float) af;  ai = u.i;
 458    u.f = (float) bf;  bi = u.i;
 459    return (ai - bi) >> 1;
 460 }
 461 #define IFLOOR(x)  ifloor(x)
 462 #else
 463 static INLINE int ifloor(float f)
 464 {
 465    int i = IROUND(f);
 466    return (i > f) ? i - 1 : i;
 467 }
 468 #define IFLOOR(x)  ifloor(x)
 469 #endif
 470
 471
 472 /***
 473  *** ICEIL: return (as an integer) ceiling of float
 474  ***/
 475 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 476 /*
 477  * IEEE ceil for computers that round to nearest or even.
 478  * 'f' must be between -4194304 and 4194303.
 479  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 480  * but uses some IEEE specific tricks for better speed.
 481  * Contributed by Josh Vanderhoof
 482  */
 483 static INLINE int iceil(float f)
 484 {
 485    int ai, bi;
 486    double af, bf;
 487    af = (3 << 22) + 0.5 + (double)f;
 488    bf = (3 << 22) + 0.5 - (double)f;
 489    /* GCC generates an extra fstp/fld without this. */
 490    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 491    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 492    return (ai - bi + 1) >> 1;
 493 }
 494 #define ICEIL(x)  iceil(x)
 495 #elif defined(USE_IEEE)
 496 static INLINE int iceil(float f)
 497 {
 498    int ai, bi;
 499    double af, bf;
 500    fi_type u;
 501    af = (3 << 22) + 0.5 + (double)f;
 502    bf = (3 << 22) + 0.5 - (double)f;
 503    u.f = (float) af; ai = u.i;
 504    u.f = (float) bf; bi = u.i;
 505    return (ai - bi + 1) >> 1;
 506 }
 507 #define ICEIL(x)  iceil(x)
 508 #else
 509 static INLINE int iceil(float f)
 510 {
 511    int i = IROUND(f);
 512    return (i < f) ? i + 1 : i;
 513 }
 514 #define ICEIL(x)  iceil(x)
 515 #endif
 516
 517
 518 /***
 519  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 520  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 521  ***/
 522 #if defined(USE_IEEE) && !defined(DEBUG)
 523 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 524 /* This function/macro is sensitive to precision.  Test very carefully
 525  * if you change it!
 526  */
 527 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 528         do {                                                            \
 529            fi_type __tmp;                                               \
 530            __tmp.f = (F);                                               \
 531            if (__tmp.i < 0)                                             \
 532               UB = (GLubyte) 0;                                         \
 533            else if (__tmp.i >= IEEE_0996)                               \
 534               UB = (GLubyte) 255;                                       \
 535            else {                                                       \
 536               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 537               UB = (GLubyte) __tmp.i;                                   \
 538            }                                                            \
 539         } while (0)
 540 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 541         do {                                                            \
 542            fi_type __tmp;                                               \
 543            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 544            UB = (GLubyte) __tmp.i;                                      \
 545         } while (0)
 546 #else
 547 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 548         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 549 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 550         ub = ((GLubyte) IROUND((f) * 255.0F))
 551 #endif
 552
 553
 554 /***
 555  *** COPY_FLOAT: copy a float from src to dest, avoid slow FP regs if possible
 556  ***/
 557 #if defined(USE_IEEE) && !defined(DEBUG)
 558 #define COPY_FLOAT( dst, src )                                  \
 559         ((fi_type *) &(dst))->i = ((fi_type *) (void *) &(src))->i
 560 #else
 561 #define COPY_FLOAT( dst, src )          (dst) = (src)
 562 #endif
 563
 564
 565 /***
 566  *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
 567  ***                  original mode to a temporary).
 568  *** END_FAST_MATH: Restore x86 FPU to original mode.
 569  ***/
 570 #if defined(__GNUC__) && defined(__i386__)
 571 /*
 572  * Set the x86 FPU control word to guarentee only 32 bits of precision
 573  * are stored in registers.  Allowing the FPU to store more introduces
 574  * differences between situations where numbers are pulled out of memory
 575  * vs. situations where the compiler is able to optimize register usage.
 576  *
 577  * In the worst case, we force the compiler to use a memory access to
 578  * truncate the float, by specifying the 'volatile' keyword.
 579  */
 580 /* Hardware default: All exceptions masked, extended double precision,
 581  * round to nearest (IEEE compliant):
 582  */
 583 #define DEFAULT_X86_FPU         0x037f
 584 /* All exceptions masked, single precision, round to nearest:
 585  */
 586 #define FAST_X86_FPU            0x003f
 587 /* The fldcw instruction will cause any pending FP exceptions to be
 588  * raised prior to entering the block, and we clear any pending
 589  * exceptions before exiting the block.  Hence, asm code has free
 590  * reign over the FPU while in the fast math block.
 591  */
 592 #if defined(NO_FAST_MATH)
 593 #define START_FAST_MATH(x)                                              \
 594 do {                                                                    \
 595    static GLuint mask = DEFAULT_X86_FPU;                                \
 596    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 597    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 598 } while (0)
 599 #else
 600 #define START_FAST_MATH(x)                                              \
 601 do {                                                                    \
 602    static GLuint mask = FAST_X86_FPU;                                   \
 603    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 604    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 605 } while (0)
 606 #endif
 607 /* Restore original FPU mode, and clear any exceptions that may have
 608  * occurred in the FAST_MATH block.
 609  */
 610 #define END_FAST_MATH(x)                                                \
 611 do {                                                                    \
 612    __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) );                     \
 613 } while (0)
 614
 615 #elif defined(__WATCOMC__) && defined(__386__)
 616 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 617 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 618 void _watcom_start_fast_math(unsigned short *x,unsigned short *mask);
 619 #pragma aux _watcom_start_fast_math =                                   \
 620    "fnstcw  word ptr [eax]"                                             \
 621    "fldcw   word ptr [ecx]"                                             \
 622    parm [eax] [ecx]                                                     \
 623    modify exact [];
 624 void _watcom_end_fast_math(unsigned short *x);
 625 #pragma aux _watcom_end_fast_math =                                     \
 626    "fnclex"                                                             \
 627    "fldcw   word ptr [eax]"                                             \
 628    parm [eax]                                                           \
 629    modify exact [];
 630 #if defined(NO_FAST_MATH)
 631 #define START_FAST_MATH(x)                                              \
 632 do {                                                                    \
 633    static GLushort mask = DEFAULT_X86_FPU;                                  \
 634    _watcom_start_fast_math(&x,&mask);                                   \
 635 } while (0)
 636 #else
 637 #define START_FAST_MATH(x)                                              \
 638 do {                                                                    \
 639    static GLushort mask = FAST_X86_FPU;                                 \
 640    _watcom_start_fast_math(&x,&mask);                                   \
 641 } while (0)
 642 #endif
 643 #define END_FAST_MATH(x)  _watcom_end_fast_math(&x)
 644 #else
 645 #define START_FAST_MATH(x)  x = 0
 646 #define END_FAST_MATH(x)  (void)(x)
 647 #endif
 648
 649
 650
 651 /**********************************************************************
 652  * Functions
 653  */
 654
 655 extern void *
 656 _mesa_malloc( size_t bytes );
 657
 658 extern void *
 659 _mesa_calloc( size_t bytes );
 660
 661 extern void
 662 _mesa_free( void *ptr );
 663
 664 extern void *
 665 _mesa_align_malloc( size_t bytes, unsigned long alignment );
 666
 667 extern void *
 668 _mesa_align_calloc( size_t bytes, unsigned long alignment );
 669
 670 extern void
 671 _mesa_align_free( void *ptr );
 672
 673 extern void *
 674 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 675
 676 extern void *
 677 _mesa_memcpy( void *dest, const void *src, size_t n );
 678
 679 extern void
 680 _mesa_memset( void *dst, int val, size_t n );
 681
 682 extern void
 683 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 684
 685 extern void
 686 _mesa_bzero( void *dst, size_t n );
 687
 688
 689 extern double
 690 _mesa_sin(double a);
 691
 692 extern double
 693 _mesa_cos(double a);
 694
 695 extern double
 696 _mesa_sqrtd(double x);
 697
 698 extern float
 699 _mesa_sqrtf(float x);
 700
 701 extern float
 702 _mesa_inv_sqrtf(float x);
 703
 704 extern double
 705 _mesa_pow(double x, double y);
 706
 707 extern float
 708 _mesa_log2(float x);
 709
 710 extern unsigned int
 711 _mesa_bitcount(unsigned int n);
 712
 713 extern GLhalfARB
 714 _mesa_float_to_half(float f);
 715
 716 extern float
 717 _mesa_half_to_float(GLhalfARB h);
 718
 719
 720 extern char *
 721 _mesa_getenv( const char *var );
 722
 723 extern char *
 724 _mesa_strstr( const char *haystack, const char *needle );
 725
 726 extern char *
 727 _mesa_strncat( char *dest, const char *src, size_t n );
 728
 729 extern char *
 730 _mesa_strcpy( char *dest, const char *src );
 731
 732 extern char *
 733 _mesa_strncpy( char *dest, const char *src, size_t n );
 734
 735 extern size_t
 736 _mesa_strlen( const char *s );
 737
 738 extern int
 739 _mesa_strcmp( const char *s1, const char *s2 );
 740
 741 extern int
 742 _mesa_strncmp( const char *s1, const char *s2, size_t n );
 743
 744 extern char *
 745 _mesa_strdup( const char *s );
 746
 747 extern int
 748 _mesa_atoi( const char *s );
 749
 750 extern double
 751 _mesa_strtod( const char *s, char **end );
 752
 753 extern int
 754 _mesa_sprintf( char *str, const char *fmt, ... );
 755
 756 extern void
 757 _mesa_printf( const char *fmtString, ... );
 758
 759
 760 extern void
 761 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 762
 763 extern void
 764 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 765
 766 extern void
 767 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 768
 769 extern void
 770 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 771
 772
 773 extern void
 774 _mesa_init_default_imports( __GLimports *imports, void *driverCtx );
 775
 776
 777 #ifdef __cplusplus
 778 }
 779 #endif
 780
 781
 782 #endif /* IMPORTS_H */