src/mesa/main/imports.h

   1 /**
   2  * \file imports.h
   3  * Standard C library function wrappers.
   4  *
   5  * This file provides wrappers for all the standard C library functions
   6  * like malloc(), free(), printf(), getenv(), etc.
   7  */
   8
   9 /*
  10  * Mesa 3-D graphics library
  11  * Version:  6.0
  12  *
  13  * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
  14  *
  15  * Permission is hereby granted, free of charge, to any person obtaining a
  16  * copy of this software and associated documentation files (the "Software"),
  17  * to deal in the Software without restriction, including without limitation
  18  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  19  * and/or sell copies of the Software, and to permit persons to whom the
  20  * Software is furnished to do so, subject to the following conditions:
  21  *
  22  * The above copyright notice and this permission notice shall be included
  23  * in all copies or substantial portions of the Software.
  24  *
  25  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  26  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  27  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  28  * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
  29  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  30  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  31  */
  32
  33
  34 #ifndef IMPORTS_H
  35 #define IMPORTS_H
  36
  37
  38 /* XXX some of the stuff in glheader.h should be moved into this file.
  39  */
  40 #include "glheader.h"
  41
  42
  43 #ifdef __cplusplus
  44 extern "C" {
  45 #endif
  46
  47
  48 /**********************************************************************/
  49 /** \name General macros */
  50 /*@{*/
  51
  52 #ifndef NULL
  53 #define NULL 0
  54 #endif
  55
  56 /*@}*/
  57
  58
  59 /**********************************************************************/
  60 /** Memory macros */
  61 /*@{*/
  62
  63 /** Allocate \p BYTES bytes */
  64 #define MALLOC(BYTES)      _mesa_malloc(BYTES)
  65 /** Allocate and zero \p BYTES bytes */
  66 #define CALLOC(BYTES)      _mesa_calloc(BYTES)
  67 /** Allocate a structure of type \p T */
  68 #define MALLOC_STRUCT(T)   (struct T *) _mesa_malloc(sizeof(struct T))
  69 /** Allocate and zero a structure of type \p T */
  70 #define CALLOC_STRUCT(T)   (struct T *) _mesa_calloc(sizeof(struct T))
  71 /** Free memory */
  72 #define FREE(PTR)          _mesa_free(PTR)
  73
  74 /** Allocate \p BYTES aligned at \p N bytes */
  75 #define ALIGN_MALLOC(BYTES, N)     _mesa_align_malloc(BYTES, N)
  76 /** Allocate and zero \p BYTES bytes aligned at \p N bytes */
  77 #define ALIGN_CALLOC(BYTES, N)     _mesa_align_calloc(BYTES, N)
  78 /** Allocate a structure of type \p T aligned at \p N bytes */
  79 #define ALIGN_MALLOC_STRUCT(T, N)  (struct T *) _mesa_align_malloc(sizeof(struct T), N)
  80 /** Allocate and zero a structure of type \p T aligned at \p N bytes */
  81 #define ALIGN_CALLOC_STRUCT(T, N)  (struct T *) _mesa_align_calloc(sizeof(struct T), N)
  82 /** Free aligned memory */
  83 #define ALIGN_FREE(PTR)            _mesa_align_free(PTR)
  84
  85 /** Copy \p BYTES bytes from \p SRC into \p DST */
  86 #define MEMCPY( DST, SRC, BYTES)   _mesa_memcpy(DST, SRC, BYTES)
  87 /** Set \p N bytes in \p DST to \p VAL */
  88 #define MEMSET( DST, VAL, N )      _mesa_memset(DST, VAL, N)
  89
  90 #define MEMSET16( DST, VAL, N )   _mesa_memset16( (DST), (VAL), (size_t) (N) )
  91
  92 /*@}*/
  93
  94
  95 /*
  96  * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
  97  * as offsets into buffer stores.  Since the vertex array pointer and
  98  * buffer store pointer are both pointers and we need to add them, we use
  99  * this macro.
 100  * Both pointers/offsets are expressed in bytes.
 101  */
 102 #define ADD_POINTERS(A, B)  ( (A) + (unsigned long) (B) )
 103
 104
 105 /**********************************************************************/
 106 /** \name [Pseudo] static array declaration.
 107  *
 108  * MACs and BeOS don't support static larger than 32kb, so ...
 109  */
 110 /*@{*/
 111
 112 /**
 113  * \def DEFARRAY
 114  * Define a [static] unidimensional array
 115  */
 116
 117 /**
 118  * \def DEFMARRAY
 119  * Define a [static] bi-dimensional array
 120  */
 121
 122 /**
 123  * \def DEFMNARRAY
 124  * Define a [static] tri-dimensional array
 125  */
 126
 127 /**
 128  * \def CHECKARRAY
 129  * Verifies a [static] array was properly allocated.
 130  */
 131
 132 /**
 133  * \def UNDEFARRAY
 134  * Undefine (free) a [static] array.
 135  */
 136
 137 #if defined(macintosh) && !defined(__MRC__)
 138 /*extern char *AGLAlloc(int size);*/
 139 /*extern void AGLFree(char* ptr);*/
 140 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE *NAME = (TYPE*)_mesa_alloc(sizeof(TYPE)*(SIZE))
 141 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
 142 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
 143 #  define CHECKARRAY(NAME,CMD)                          do {if (!(NAME)) {CMD;}} while (0)
 144 #  define UNDEFARRAY(NAME)                              do {if ((NAME)) {_mesa_free((char*)NAME);}  }while (0)
 145 #elif defined(__BEOS__)
 146 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE *NAME = (TYPE*)_mesa_malloc(sizeof(TYPE)*(SIZE))
 147 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
 148 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
 149 #  define CHECKARRAY(NAME,CMD)                          do {if (!(NAME)) {CMD;}} while (0)
 150 #  define UNDEFARRAY(NAME)                              do {if ((NAME)) {_mesa_free((char*)NAME);}  }while (0)
 151 #else
 152 #  define DEFARRAY(TYPE,NAME,SIZE)                      TYPE NAME[SIZE]
 153 #  define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2)              TYPE NAME[SIZE1][SIZE2]
 154 #  define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3)       TYPE NAME[SIZE1][SIZE2][SIZE3]
 155 #  define CHECKARRAY(NAME,CMD)                          do {} while(0)
 156 #  define UNDEFARRAY(NAME)
 157 #endif
 158
 159 /*@}*/
 160
 161
 162 /**********************************************************************/
 163 /** \name External pixel buffer allocation.
 164  *
 165  * If you want Mesa's depth/stencil/accum/etc buffers to be allocated with a
 166  * specialized allocator you can define MESA_EXTERNAL_BUFFERALLOC and implement
 167  * _ext_mesa_alloc_pixelbuffer() _ext_mesa_free_pixelbuffer() in your
 168  * application.
 169  *
 170  * \author
 171  * Contributed by Gerk Huisma (gerk@five-d.demon.nl).
 172  */
 173 /*@{*/
 174
 175 /**
 176  * \def MESA_PBUFFER_ALLOC
 177  * Allocate a pixel buffer.
 178  */
 179
 180 /**
 181  * \def MESA_PBUFFER_FREE
 182  * Free a pixel buffer.
 183  */
 184
 185 #ifdef MESA_EXTERNAL_BUFFERALLOC
 186 extern void *_ext_mesa_alloc_pixelbuffer( unsigned int size );
 187 extern void _ext_mesa_free_pixelbuffer( void *pb );
 188
 189 #define MESA_PBUFFER_ALLOC(BYTES)  (void *) _ext_mesa_alloc_pixelbuffer(BYTES)
 190 #define MESA_PBUFFER_FREE(PTR)     _ext_mesa_free_pixelbuffer(PTR)
 191 #else
 192 /* Default buffer allocation uses the aligned allocation routines: */
 193 #define MESA_PBUFFER_ALLOC(BYTES)  (void *) _mesa_align_malloc(BYTES, 512)
 194 #define MESA_PBUFFER_FREE(PTR)     _mesa_align_free(PTR)
 195 #endif
 196
 197 /*@}*/
 198
 199
 200
 201 /**********************************************************************
 202  * Math macros
 203  */
 204
 205 #define MAX_GLUSHORT    0xffff
 206 #define MAX_GLUINT      0xffffffff
 207
 208 #ifndef M_PI
 209 #define M_PI (3.1415926536)
 210 #endif
 211
 212 /* XXX this is a bit of a hack needed for compilation within XFree86 */
 213 #ifndef FLT_MIN
 214 #define FLT_MIN (1.0e-37)
 215 #endif
 216
 217 /* Degrees to radians conversion: */
 218 #define DEG2RAD (M_PI/180.0)
 219
 220
 221 /***
 222  *** USE_IEEE: Determine if we're using IEEE floating point
 223  ***/
 224 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
 225     defined(__s390x__) || defined(__powerpc__) || \
 226     ( defined(__alpha__) && ( defined(__IEEE_FLOAT) || !defined(VMS) ) )
 227 #define USE_IEEE
 228 #define IEEE_ONE 0x3f800000
 229 #endif
 230
 231
 232 /***
 233  *** SQRTF: single-precision square root
 234  ***/
 235 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 236 #  define SQRTF(X)  _mesa_sqrtf(X)
 237 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 238 #  define SQRTF(X)  (float) xf86sqrt((float) (X))
 239 #else
 240 #  define SQRTF(X)  (float) sqrt((float) (X))
 241 #endif
 242
 243
 244 /***
 245  *** INV_SQRTF: single-precision inverse square root
 246  ***/
 247 #if 0
 248 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 249 #else
 250 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 251 #endif
 252
 253
 254 /***
 255  *** LOG2: Log base 2 of float
 256  ***/
 257 #ifdef USE_IEEE
 258 #if 0
 259 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 260  * Based on code from http://www.stereopsis.com/log2.html
 261  */
 262 static INLINE GLfloat LOG2(GLfloat x)
 263 {
 264    const GLfloat y = x * x * x * x;
 265    const GLuint ix = *((GLuint *) &y);
 266    const GLuint exp = (ix >> 23) & 0xFF;
 267    const GLint log2 = ((GLint) exp) - 127;
 268    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 269 }
 270 #endif
 271 /* Pretty fast, and accurate.
 272  * Based on code from http://www.flipcode.com/totd/
 273  */
 274 static INLINE GLfloat LOG2(GLfloat val)
 275 {
 276    fi_type num;
 277    GLint log_2;
 278    num.f = val;
 279    log_2 = ((num.i >> 23) & 255) - 128;
 280    num.i &= ~(255 << 23);
 281    num.i += 127 << 23;
 282    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 283    return num.f + log_2;
 284 }
 285 #elif defined(XFree86LOADER) && defined(IN_MODULE)
 286 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 287 #else
 288 /*
 289  * NOTE: log_base_2(x) = log(x) / log(2)
 290  * NOTE: 1.442695 = 1/log(2).
 291  */
 292 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 293 #endif
 294
 295
 296 /***
 297  *** IS_INF_OR_NAN: test if float is infinite or NaN
 298  ***/
 299 #ifdef USE_IEEE
 300 static INLINE int IS_INF_OR_NAN( float x )
 301 {
 302    fi_type tmp;
 303    tmp.f = x;
 304    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 305 }
 306 #elif defined(isfinite)
 307 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 308 #elif defined(finite)
 309 #define IS_INF_OR_NAN(x)        (!finite(x))
 310 #elif __VMS
 311 #define IS_INF_OR_NAN(x)        (!finite(x))
 312 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 313 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 314 #else
 315 #define IS_INF_OR_NAN(x)        (!finite(x))
 316 #endif
 317
 318
 319 /***
 320  *** IS_NEGATIVE: test if float is negative
 321  ***/
 322 #if defined(USE_IEEE)
 323 #define GET_FLOAT_BITS(x) ((fi_type *) &(x))->i
 324 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) & (1<<31))
 325 #else
 326 #define IS_NEGATIVE(x) (x < 0.0F)
 327 #endif
 328
 329
 330 /***
 331  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 332  ***/
 333 #if defined(USE_IEEE)
 334 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 335 #else
 336 /* Could just use (x*y<0) except for the flatshading requirements.
 337  * Maybe there's a better way?
 338  */
 339 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 340 #endif
 341
 342
 343 /***
 344  *** CEILF: ceiling of float
 345  *** FLOORF: floor of float
 346  *** FABSF: absolute value of float
 347  ***/
 348 #if defined(XFree86LOADER) && defined(IN_MODULE)
 349 #define CEILF(x)   ((GLfloat) xf86ceil(x))
 350 #define FLOORF(x)  ((GLfloat) xf86floor(x))
 351 #define FABSF(x)   ((GLfloat) xf86fabs(x))
 352 #elif defined(__gnu_linux__)
 353 /* C99 functions */
 354 #define CEILF(x)   ceilf(x)
 355 #define FLOORF(x)  floorf(x)
 356 #define FABSF(x)   fabsf(x)
 357 #else
 358 #define CEILF(x)   ((GLfloat) ceil(x))
 359 #define FLOORF(x)  ((GLfloat) floor(x))
 360 #define FABSF(x)   ((GLfloat) fabs(x))
 361 #endif
 362
 363
 364 /***
 365  *** IROUND: return (as an integer) float rounded to nearest integer
 366  ***/
 367 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
 368 static INLINE int iround(float f)
 369 {
 370    int r;
 371    __asm__ ("fstoi %1, %0" : "=f" (r) : "f" (f));
 372    return r;
 373 }
 374 #define IROUND(x)  iround(x)
 375 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 376 static INLINE int iround(float f)
 377 {
 378    int r;
 379    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 380    return r;
 381 }
 382 #define IROUND(x)  iround(x)
 383 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
 384 static INLINE int iround(float f)
 385 {
 386    int r;
 387    _asm {
 388          fld f
 389          fistp r
 390         }
 391    return r;
 392 }
 393 #define IROUND(x)  iround(x)
 394 #elif defined(__WATCOMC__) && defined(__386__)
 395 long iround(float f);
 396 #pragma aux iround =                    \
 397         "push   eax"                        \
 398         "fistp  dword ptr [esp]"            \
 399         "pop    eax"                        \
 400         parm [8087]                         \
 401         value [eax]                         \
 402         modify exact [eax];
 403 #define IROUND(x)  iround(x)
 404 #else
 405 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 406 #endif
 407
 408
 409 /***
 410  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 411  ***/
 412 #ifdef DEBUG
 413 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 414 #else
 415 #define IROUND_POS(f) (IROUND(f))
 416 #endif
 417
 418
 419 /***
 420  *** IFLOOR: return (as an integer) floor of float
 421  ***/
 422 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 423 /*
 424  * IEEE floor for computers that round to nearest or even.
 425  * 'f' must be between -4194304 and 4194303.
 426  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 427  * but uses some IEEE specific tricks for better speed.
 428  * Contributed by Josh Vanderhoof
 429  */
 430 static INLINE int ifloor(float f)
 431 {
 432    int ai, bi;
 433    double af, bf;
 434    af = (3 << 22) + 0.5 + (double)f;
 435    bf = (3 << 22) + 0.5 - (double)f;
 436    /* GCC generates an extra fstp/fld without this. */
 437    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 438    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 439    return (ai - bi) >> 1;
 440 }
 441 #define IFLOOR(x)  ifloor(x)
 442 #elif defined(USE_IEEE)
 443 static INLINE int ifloor(float f)
 444 {
 445    int ai, bi;
 446    double af, bf;
 447    fi_type u;
 448
 449    af = (3 << 22) + 0.5 + (double)f;
 450    bf = (3 << 22) + 0.5 - (double)f;
 451    u.f = (float) af;  ai = u.i;
 452    u.f = (float) bf;  bi = u.i;
 453    return (ai - bi) >> 1;
 454 }
 455 #define IFLOOR(x)  ifloor(x)
 456 #else
 457 static INLINE int ifloor(float f)
 458 {
 459    int i = IROUND(f);
 460    return (i > f) ? i - 1 : i;
 461 }
 462 #define IFLOOR(x)  ifloor(x)
 463 #endif
 464
 465
 466 /***
 467  *** ICEIL: return (as an integer) ceiling of float
 468  ***/
 469 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 470 /*
 471  * IEEE ceil for computers that round to nearest or even.
 472  * 'f' must be between -4194304 and 4194303.
 473  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 474  * but uses some IEEE specific tricks for better speed.
 475  * Contributed by Josh Vanderhoof
 476  */
 477 static INLINE int iceil(float f)
 478 {
 479    int ai, bi;
 480    double af, bf;
 481    af = (3 << 22) + 0.5 + (double)f;
 482    bf = (3 << 22) + 0.5 - (double)f;
 483    /* GCC generates an extra fstp/fld without this. */
 484    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 485    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 486    return (ai - bi + 1) >> 1;
 487 }
 488 #define ICEIL(x)  iceil(x)
 489 #elif defined(USE_IEEE)
 490 static INLINE int iceil(float f)
 491 {
 492    int ai, bi;
 493    double af, bf;
 494    fi_type u;
 495    af = (3 << 22) + 0.5 + (double)f;
 496    bf = (3 << 22) + 0.5 - (double)f;
 497    u.f = (float) af; ai = u.i;
 498    u.f = (float) bf; bi = u.i;
 499    return (ai - bi + 1) >> 1;
 500 }
 501 #define ICEIL(x)  iceil(x)
 502 #else
 503 static INLINE int iceil(float f)
 504 {
 505    int i = IROUND(f);
 506    return (i < f) ? i + 1 : i;
 507 }
 508 #define ICEIL(x)  iceil(x)
 509 #endif
 510
 511
 512 /***
 513  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 514  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 515  ***/
 516 #if defined(USE_IEEE) && !defined(DEBUG)
 517 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 518 /* This function/macro is sensitive to precision.  Test very carefully
 519  * if you change it!
 520  */
 521 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 522         do {                                                            \
 523            fi_type __tmp;                                               \
 524            __tmp.f = (F);                                               \
 525            if (__tmp.i < 0)                                             \
 526               UB = (GLubyte) 0;                                         \
 527            else if (__tmp.i >= IEEE_0996)                               \
 528               UB = (GLubyte) 255;                                       \
 529            else {                                                       \
 530               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 531               UB = (GLubyte) __tmp.i;                                   \
 532            }                                                            \
 533         } while (0)
 534 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 535         do {                                                            \
 536            fi_type __tmp;                                               \
 537            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 538            UB = (GLubyte) __tmp.i;                                      \
 539         } while (0)
 540 #else
 541 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 542         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 543 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 544         ub = ((GLubyte) IROUND((f) * 255.0F))
 545 #endif
 546
 547
 548 /***
 549  *** COPY_FLOAT: copy a float from src to dest, avoid slow FP regs if possible
 550  ***/
 551 #if defined(USE_IEEE) && !defined(DEBUG)
 552 #define COPY_FLOAT( dst, src )                                  \
 553         ((fi_type *) &(dst))->i = ((fi_type *) &(src))->i
 554 #else
 555 #define COPY_FLOAT( dst, src )          (dst) = (src)
 556 #endif
 557
 558
 559 /***
 560  *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
 561  ***                  original mode to a temporary).
 562  *** END_FAST_MATH: Restore x86 FPU to original mode.
 563  ***/
 564 #if defined(__GNUC__) && defined(__i386__)
 565 /*
 566  * Set the x86 FPU control word to guarentee only 32 bits of precision
 567  * are stored in registers.  Allowing the FPU to store more introduces
 568  * differences between situations where numbers are pulled out of memory
 569  * vs. situations where the compiler is able to optimize register usage.
 570  *
 571  * In the worst case, we force the compiler to use a memory access to
 572  * truncate the float, by specifying the 'volatile' keyword.
 573  */
 574 /* Hardware default: All exceptions masked, extended double precision,
 575  * round to nearest (IEEE compliant):
 576  */
 577 #define DEFAULT_X86_FPU         0x037f
 578 /* All exceptions masked, single precision, round to nearest:
 579  */
 580 #define FAST_X86_FPU            0x003f
 581 /* The fldcw instruction will cause any pending FP exceptions to be
 582  * raised prior to entering the block, and we clear any pending
 583  * exceptions before exiting the block.  Hence, asm code has free
 584  * reign over the FPU while in the fast math block.
 585  */
 586 #if defined(NO_FAST_MATH)
 587 #define START_FAST_MATH(x)                                              \
 588 do {                                                                    \
 589    static GLuint mask = DEFAULT_X86_FPU;                                \
 590    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 591    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 592 } while (0)
 593 #else
 594 #define START_FAST_MATH(x)                                              \
 595 do {                                                                    \
 596    static GLuint mask = FAST_X86_FPU;                                   \
 597    __asm__ ( "fnstcw %0" : "=m" (*&(x)) );                              \
 598    __asm__ ( "fldcw %0" : : "m" (mask) );                               \
 599 } while (0)
 600 #endif
 601 /* Restore original FPU mode, and clear any exceptions that may have
 602  * occurred in the FAST_MATH block.
 603  */
 604 #define END_FAST_MATH(x)                                                \
 605 do {                                                                    \
 606    __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) );                     \
 607 } while (0)
 608
 609 #elif defined(__WATCOMC__) && defined(__386__)
 610 #define DEFAULT_X86_FPU         0x037f /* See GCC comments above */
 611 #define FAST_X86_FPU            0x003f /* See GCC comments above */
 612 void _watcom_start_fast_math(unsigned short *x,unsigned short *mask);
 613 #pragma aux _watcom_start_fast_math =                                   \
 614    "fnstcw  word ptr [eax]"                                             \
 615    "fldcw   word ptr [ecx]"                                             \
 616    parm [eax] [ecx]                                                     \
 617    modify exact [];
 618 void _watcom_end_fast_math(unsigned short *x);
 619 #pragma aux _watcom_end_fast_math =                                     \
 620    "fnclex"                                                             \
 621    "fldcw   word ptr [eax]"                                             \
 622    parm [eax]                                                           \
 623    modify exact [];
 624 #if defined(NO_FAST_MATH)
 625 #define START_FAST_MATH(x)                                              \
 626 do {                                                                    \
 627    static GLushort mask = DEFAULT_X86_FPU;                                  \
 628    _watcom_start_fast_math(&x,&mask);                                   \
 629 } while (0)
 630 #else
 631 #define START_FAST_MATH(x)                                              \
 632 do {                                                                    \
 633    static GLushort mask = FAST_X86_FPU;                                 \
 634    _watcom_start_fast_math(&x,&mask);                                   \
 635 } while (0)
 636 #endif
 637 #define END_FAST_MATH(x)  _watcom_end_fast_math(&x)
 638 #else
 639 #define START_FAST_MATH(x)  x = 0
 640 #define END_FAST_MATH(x)  (void)(x)
 641 #endif
 642
 643
 644
 645 /**********************************************************************
 646  * Functions
 647  */
 648
 649 extern void *
 650 _mesa_malloc( size_t bytes );
 651
 652 extern void *
 653 _mesa_calloc( size_t bytes );
 654
 655 extern void
 656 _mesa_free( void *ptr );
 657
 658 extern void *
 659 _mesa_align_malloc( size_t bytes, unsigned long alignment );
 660
 661 extern void *
 662 _mesa_align_calloc( size_t bytes, unsigned long alignment );
 663
 664 extern void
 665 _mesa_align_free( void *ptr );
 666
 667 extern void *
 668 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 669
 670 extern void *
 671 _mesa_memcpy( void *dest, const void *src, size_t n );
 672
 673 extern void
 674 _mesa_memset( void *dst, int val, size_t n );
 675
 676 extern void
 677 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 678
 679 extern void
 680 _mesa_bzero( void *dst, size_t n );
 681
 682
 683 extern double
 684 _mesa_sin(double a);
 685
 686 extern double
 687 _mesa_cos(double a);
 688
 689 extern double
 690 _mesa_sqrtd(double x);
 691
 692 extern float
 693 _mesa_sqrtf(float x);
 694
 695 extern float
 696 _mesa_inv_sqrtf(float x);
 697
 698 extern double
 699 _mesa_pow(double x, double y);
 700
 701 extern float
 702 _mesa_log2(float x);
 703
 704 extern unsigned int
 705 _mesa_bitcount(unsigned int n);
 706
 707 extern GLhalfNV
 708 _mesa_float_to_half(float f);
 709
 710 extern float
 711 _mesa_half_to_float(GLhalfNV h);
 712
 713
 714 extern char *
 715 _mesa_getenv( const char *var );
 716
 717 extern char *
 718 _mesa_strstr( const char *haystack, const char *needle );
 719
 720 extern char *
 721 _mesa_strncat( char *dest, const char *src, size_t n );
 722
 723 extern char *
 724 _mesa_strcpy( char *dest, const char *src );
 725
 726 extern char *
 727 _mesa_strncpy( char *dest, const char *src, size_t n );
 728
 729 extern size_t
 730 _mesa_strlen( const char *s );
 731
 732 extern int
 733 _mesa_strcmp( const char *s1, const char *s2 );
 734
 735 extern int
 736 _mesa_strncmp( const char *s1, const char *s2, size_t n );
 737
 738 extern char *
 739 _mesa_strdup( const char *s );
 740
 741 extern int
 742 _mesa_atoi( const char *s );
 743
 744 extern double
 745 _mesa_strtod( const char *s, char **end );
 746
 747 extern int
 748 _mesa_sprintf( char *str, const char *fmt, ... );
 749
 750 extern void
 751 _mesa_printf( const char *fmtString, ... );
 752
 753
 754 extern void
 755 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 756
 757 extern void
 758 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 759
 760 extern void
 761 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 762
 763 extern void
 764 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 765
 766
 767 extern void
 768 _mesa_init_default_imports( __GLimports *imports, void *driverCtx );
 769
 770
 771 #ifdef __cplusplus
 772 }
 773 #endif
 774
 775
 776 #endif /* IMPORTS_H */