src/mesa/main/imports.h

   1 /*
   2  * Mesa 3-D graphics library
   3  * Version:  7.5
   4  *
   5  * Copyright (C) 1999-2008  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 #include "compiler.h"
  40 #include "glheader.h"
  41
  42
  43 #ifdef __cplusplus
  44 extern "C" {
  45 #endif
  46
  47
  48 /**********************************************************************/
  49 /** Memory macros */
  50 /*@{*/
  51
  52 /** Allocate \p BYTES bytes */
  53 #define MALLOC(BYTES)      _mesa_malloc(BYTES)
  54 /** Allocate and zero \p BYTES bytes */
  55 #define CALLOC(BYTES)      _mesa_calloc(BYTES)
  56 /** Allocate a structure of type \p T */
  57 #define MALLOC_STRUCT(T)   (struct T *) _mesa_malloc(sizeof(struct T))
  58 /** Allocate and zero a structure of type \p T */
  59 #define CALLOC_STRUCT(T)   (struct T *) _mesa_calloc(sizeof(struct T))
  60 /** Free memory */
  61 #define FREE(PTR)          _mesa_free(PTR)
  62
  63 /** Allocate \p BYTES aligned at \p N bytes */
  64 #define ALIGN_MALLOC(BYTES, N)     _mesa_align_malloc(BYTES, N)
  65 /** Allocate and zero \p BYTES bytes aligned at \p N bytes */
  66 #define ALIGN_CALLOC(BYTES, N)     _mesa_align_calloc(BYTES, N)
  67 /** Allocate a structure of type \p T aligned at \p N bytes */
  68 #define ALIGN_MALLOC_STRUCT(T, N)  (struct T *) _mesa_align_malloc(sizeof(struct T), N)
  69 /** Allocate and zero a structure of type \p T aligned at \p N bytes */
  70 #define ALIGN_CALLOC_STRUCT(T, N)  (struct T *) _mesa_align_calloc(sizeof(struct T), N)
  71 /** Free aligned memory */
  72 #define ALIGN_FREE(PTR)            _mesa_align_free(PTR)
  73
  74 /** Set \p N bytes in \p DST to \p VAL */
  75 #define MEMSET( DST, VAL, N )      memset(DST, VAL, N)
  76
  77 /*@}*/
  78
  79
  80 /*
  81  * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
  82  * as offsets into buffer stores.  Since the vertex array pointer and
  83  * buffer store pointer are both pointers and we need to add them, we use
  84  * this macro.
  85  * Both pointers/offsets are expressed in bytes.
  86  */
  87 #define ADD_POINTERS(A, B)  ( (GLubyte *) (A) + (uintptr_t) (B) )
  88
  89
  90 /**
  91  * Sometimes we treat GLfloats as GLints.  On x86 systems, moving a float
  92  * as a int (thereby using integer registers instead of FP registers) is
  93  * a performance win.  Typically, this can be done with ordinary casts.
  94  * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
  95  * these casts generate warnings.
  96  * The following union typedef is used to solve that.
  97  */
  98 typedef union { GLfloat f; GLint i; } fi_type;
  99
 100
 101
 102 /**********************************************************************
 103  * Math macros
 104  */
 105
 106 #define MAX_GLUSHORT    0xffff
 107 #define MAX_GLUINT      0xffffffff
 108
 109 /* Degrees to radians conversion: */
 110 #define DEG2RAD (M_PI/180.0)
 111
 112
 113 /***
 114  *** SQRTF: single-precision square root
 115  ***/
 116 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 117 #  define SQRTF(X)  _mesa_sqrtf(X)
 118 #else
 119 #  define SQRTF(X)  (float) sqrt((float) (X))
 120 #endif
 121
 122
 123 /***
 124  *** INV_SQRTF: single-precision inverse square root
 125  ***/
 126 #if 0
 127 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
 128 #else
 129 #define INV_SQRTF(X) (1.0F / SQRTF(X))  /* this is faster on a P4 */
 130 #endif
 131
 132
 133 /***
 134  *** LOG2: Log base 2 of float
 135  ***/
 136 #ifdef USE_IEEE
 137 #if 0
 138 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
 139  * Based on code from http://www.stereopsis.com/log2.html
 140  */
 141 static INLINE GLfloat LOG2(GLfloat x)
 142 {
 143    const GLfloat y = x * x * x * x;
 144    const GLuint ix = *((GLuint *) &y);
 145    const GLuint exp = (ix >> 23) & 0xFF;
 146    const GLint log2 = ((GLint) exp) - 127;
 147    return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
 148 }
 149 #endif
 150 /* Pretty fast, and accurate.
 151  * Based on code from http://www.flipcode.com/totd/
 152  */
 153 static INLINE GLfloat LOG2(GLfloat val)
 154 {
 155    fi_type num;
 156    GLint log_2;
 157    num.f = val;
 158    log_2 = ((num.i >> 23) & 255) - 128;
 159    num.i &= ~(255 << 23);
 160    num.i += 127 << 23;
 161    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
 162    return num.f + log_2;
 163 }
 164 #else
 165 /*
 166  * NOTE: log_base_2(x) = log(x) / log(2)
 167  * NOTE: 1.442695 = 1/log(2).
 168  */
 169 #define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
 170 #endif
 171
 172
 173 /***
 174  *** IS_INF_OR_NAN: test if float is infinite or NaN
 175  ***/
 176 #ifdef USE_IEEE
 177 static INLINE int IS_INF_OR_NAN( float x )
 178 {
 179    fi_type tmp;
 180    tmp.f = x;
 181    return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
 182 }
 183 #elif defined(isfinite)
 184 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 185 #elif defined(finite)
 186 #define IS_INF_OR_NAN(x)        (!finite(x))
 187 #elif defined(__VMS)
 188 #define IS_INF_OR_NAN(x)        (!finite(x))
 189 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
 190 #define IS_INF_OR_NAN(x)        (!isfinite(x))
 191 #else
 192 #define IS_INF_OR_NAN(x)        (!finite(x))
 193 #endif
 194
 195
 196 /***
 197  *** IS_NEGATIVE: test if float is negative
 198  ***/
 199 #if defined(USE_IEEE)
 200 static INLINE int GET_FLOAT_BITS( float x )
 201 {
 202    fi_type fi;
 203    fi.f = x;
 204    return fi.i;
 205 }
 206 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
 207 #else
 208 #define IS_NEGATIVE(x) (x < 0.0F)
 209 #endif
 210
 211
 212 /***
 213  *** DIFFERENT_SIGNS: test if two floats have opposite signs
 214  ***/
 215 #if defined(USE_IEEE)
 216 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
 217 #else
 218 /* Could just use (x*y<0) except for the flatshading requirements.
 219  * Maybe there's a better way?
 220  */
 221 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
 222 #endif
 223
 224
 225 /***
 226  *** CEILF: ceiling of float
 227  *** FLOORF: floor of float
 228  *** FABSF: absolute value of float
 229  *** LOGF: the natural logarithm (base e) of the value
 230  *** EXPF: raise e to the value
 231  *** LDEXPF: multiply value by an integral power of two
 232  *** FREXPF: extract mantissa and exponent from value
 233  ***/
 234 #if defined(__gnu_linux__)
 235 /* C99 functions */
 236 #define CEILF(x)   ceilf(x)
 237 #define FLOORF(x)  floorf(x)
 238 #define FABSF(x)   fabsf(x)
 239 #define LOGF(x)    logf(x)
 240 #define EXPF(x)    expf(x)
 241 #define LDEXPF(x,y)  ldexpf(x,y)
 242 #define FREXPF(x,y)  frexpf(x,y)
 243 #else
 244 #define CEILF(x)   ((GLfloat) ceil(x))
 245 #define FLOORF(x)  ((GLfloat) floor(x))
 246 #define FABSF(x)   ((GLfloat) fabs(x))
 247 #define LOGF(x)    ((GLfloat) log(x))
 248 #define EXPF(x)    ((GLfloat) exp(x))
 249 #define LDEXPF(x,y)  ((GLfloat) ldexp(x,y))
 250 #define FREXPF(x,y)  ((GLfloat) frexp(x,y))
 251 #endif
 252
 253
 254 /***
 255  *** IROUND: return (as an integer) float rounded to nearest integer
 256  ***/
 257 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
 258                         (!(defined(__BEOS__) || defined(__HAIKU__))  || \
 259                         (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
 260 static INLINE int iround(float f)
 261 {
 262    int r;
 263    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
 264    return r;
 265 }
 266 #define IROUND(x)  iround(x)
 267 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
 268 static INLINE int iround(float f)
 269 {
 270    int r;
 271    _asm {
 272          fld f
 273          fistp r
 274         }
 275    return r;
 276 }
 277 #define IROUND(x)  iround(x)
 278 #elif defined(__WATCOMC__) && defined(__386__)
 279 long iround(float f);
 280 #pragma aux iround =                    \
 281         "push   eax"                        \
 282         "fistp  dword ptr [esp]"            \
 283         "pop    eax"                        \
 284         parm [8087]                         \
 285         value [eax]                         \
 286         modify exact [eax];
 287 #define IROUND(x)  iround(x)
 288 #else
 289 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 290 #endif
 291
 292 #define IROUND64(f)  ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 293
 294 /***
 295  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
 296  ***/
 297 #ifdef DEBUG
 298 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
 299 #else
 300 #define IROUND_POS(f) (IROUND(f))
 301 #endif
 302
 303
 304 /***
 305  *** IFLOOR: return (as an integer) floor of float
 306  ***/
 307 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 308 /*
 309  * IEEE floor for computers that round to nearest or even.
 310  * 'f' must be between -4194304 and 4194303.
 311  * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
 312  * but uses some IEEE specific tricks for better speed.
 313  * Contributed by Josh Vanderhoof
 314  */
 315 static INLINE int ifloor(float f)
 316 {
 317    int ai, bi;
 318    double af, bf;
 319    af = (3 << 22) + 0.5 + (double)f;
 320    bf = (3 << 22) + 0.5 - (double)f;
 321    /* GCC generates an extra fstp/fld without this. */
 322    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 323    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 324    return (ai - bi) >> 1;
 325 }
 326 #define IFLOOR(x)  ifloor(x)
 327 #elif defined(USE_IEEE)
 328 static INLINE int ifloor(float f)
 329 {
 330    int ai, bi;
 331    double af, bf;
 332    fi_type u;
 333
 334    af = (3 << 22) + 0.5 + (double)f;
 335    bf = (3 << 22) + 0.5 - (double)f;
 336    u.f = (float) af;  ai = u.i;
 337    u.f = (float) bf;  bi = u.i;
 338    return (ai - bi) >> 1;
 339 }
 340 #define IFLOOR(x)  ifloor(x)
 341 #else
 342 static INLINE int ifloor(float f)
 343 {
 344    int i = IROUND(f);
 345    return (i > f) ? i - 1 : i;
 346 }
 347 #define IFLOOR(x)  ifloor(x)
 348 #endif
 349
 350
 351 /***
 352  *** ICEIL: return (as an integer) ceiling of float
 353  ***/
 354 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
 355 /*
 356  * IEEE ceil for computers that round to nearest or even.
 357  * 'f' must be between -4194304 and 4194303.
 358  * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
 359  * but uses some IEEE specific tricks for better speed.
 360  * Contributed by Josh Vanderhoof
 361  */
 362 static INLINE int iceil(float f)
 363 {
 364    int ai, bi;
 365    double af, bf;
 366    af = (3 << 22) + 0.5 + (double)f;
 367    bf = (3 << 22) + 0.5 - (double)f;
 368    /* GCC generates an extra fstp/fld without this. */
 369    __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
 370    __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
 371    return (ai - bi + 1) >> 1;
 372 }
 373 #define ICEIL(x)  iceil(x)
 374 #elif defined(USE_IEEE)
 375 static INLINE int iceil(float f)
 376 {
 377    int ai, bi;
 378    double af, bf;
 379    fi_type u;
 380    af = (3 << 22) + 0.5 + (double)f;
 381    bf = (3 << 22) + 0.5 - (double)f;
 382    u.f = (float) af; ai = u.i;
 383    u.f = (float) bf; bi = u.i;
 384    return (ai - bi + 1) >> 1;
 385 }
 386 #define ICEIL(x)  iceil(x)
 387 #else
 388 static INLINE int iceil(float f)
 389 {
 390    int i = IROUND(f);
 391    return (i < f) ? i + 1 : i;
 392 }
 393 #define ICEIL(x)  iceil(x)
 394 #endif
 395
 396
 397 /**
 398  * Is x a power of two?
 399  */
 400 static INLINE int
 401 _mesa_is_pow_two(int x)
 402 {
 403    return !(x & (x - 1));
 404 }
 405
 406 /**
 407  * Round given integer to next higer power of two
 408  * If X is zero result is undefined.
 409  *
 410  * Source for the fallback implementation is
 411  * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
 412  * http://graphics.stanford.edu/~seander/bithacks.html
 413  *
 414  * When using builtin function have to do some work
 415  * for case when passed values 1 to prevent hiting
 416  * undefined result from __builtin_clz. Undefined
 417  * results would be different depending on optimization
 418  * level used for build.
 419  */
 420 static INLINE int32_t
 421 _mesa_next_pow_two_32(uint32_t x)
 422 {
 423 #ifdef __GNUC__
 424         uint32_t y = (x != 1);
 425         return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
 426 #else
 427         x--;
 428         x |= x >> 1;
 429         x |= x >> 2;
 430         x |= x >> 4;
 431         x |= x >> 8;
 432         x |= x >> 16;
 433         x++;
 434         return x;
 435 #endif
 436 }
 437
 438 static INLINE int64_t
 439 _mesa_next_pow_two_64(uint64_t x)
 440 {
 441 #ifdef __GNUC__
 442         uint64_t y = (x != 1);
 443         if (sizeof(x) == sizeof(long))
 444                 return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
 445         else
 446                 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
 447 #else
 448         x--;
 449         x |= x >> 1;
 450         x |= x >> 2;
 451         x |= x >> 4;
 452         x |= x >> 8;
 453         x |= x >> 16;
 454         x |= x >> 32;
 455         x++;
 456         return x;
 457 #endif
 458 }
 459
 460
 461 /***
 462  *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
 463  *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
 464  ***/
 465 #if defined(USE_IEEE) && !defined(DEBUG)
 466 #define IEEE_0996 0x3f7f0000    /* 0.996 or so */
 467 /* This function/macro is sensitive to precision.  Test very carefully
 468  * if you change it!
 469  */
 470 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F)                                 \
 471         do {                                                            \
 472            fi_type __tmp;                                               \
 473            __tmp.f = (F);                                               \
 474            if (__tmp.i < 0)                                             \
 475               UB = (GLubyte) 0;                                         \
 476            else if (__tmp.i >= IEEE_0996)                               \
 477               UB = (GLubyte) 255;                                       \
 478            else {                                                       \
 479               __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F;           \
 480               UB = (GLubyte) __tmp.i;                                   \
 481            }                                                            \
 482         } while (0)
 483 #define CLAMPED_FLOAT_TO_UBYTE(UB, F)                                   \
 484         do {                                                            \
 485            fi_type __tmp;                                               \
 486            __tmp.f = (F) * (255.0F/256.0F) + 32768.0F;                  \
 487            UB = (GLubyte) __tmp.i;                                      \
 488         } while (0)
 489 #else
 490 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
 491         ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
 492 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
 493         ub = ((GLubyte) IROUND((f) * 255.0F))
 494 #endif
 495
 496
 497 /**
 498  * Return 1 if this is a little endian machine, 0 if big endian.
 499  */
 500 static INLINE GLboolean
 501 _mesa_little_endian(void)
 502 {
 503    const GLuint ui = 1; /* intentionally not static */
 504    return *((const GLubyte *) &ui);
 505 }
 506
 507
 508
 509 /**********************************************************************
 510  * Functions
 511  */
 512
 513 extern void *
 514 _mesa_malloc( size_t bytes );
 515
 516 extern void *
 517 _mesa_calloc( size_t bytes );
 518
 519 extern void
 520 _mesa_free( void *ptr );
 521
 522 extern void *
 523 _mesa_align_malloc( size_t bytes, unsigned long alignment );
 524
 525 extern void *
 526 _mesa_align_calloc( size_t bytes, unsigned long alignment );
 527
 528 extern void
 529 _mesa_align_free( void *ptr );
 530
 531 extern void *
 532 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
 533                     unsigned long alignment);
 534
 535 extern void *
 536 _mesa_exec_malloc( GLuint size );
 537
 538 extern void
 539 _mesa_exec_free( void *addr );
 540
 541 extern void *
 542 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
 543
 544 extern void
 545 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
 546
 547 extern void
 548 _mesa_bzero( void *dst, size_t n );
 549
 550 extern double
 551 _mesa_sin(double a);
 552
 553 extern float
 554 _mesa_sinf(float a);
 555
 556 extern double
 557 _mesa_cos(double a);
 558
 559 extern float
 560 _mesa_asinf(float x);
 561
 562 extern float
 563 _mesa_atanf(float x);
 564
 565 extern double
 566 _mesa_sqrtd(double x);
 567
 568 extern float
 569 _mesa_sqrtf(float x);
 570
 571 extern float
 572 _mesa_inv_sqrtf(float x);
 573
 574 extern void
 575 _mesa_init_sqrt_table(void);
 576
 577 extern double
 578 _mesa_pow(double x, double y);
 579
 580 extern int
 581 _mesa_ffs(int32_t i);
 582
 583 extern int
 584 _mesa_ffsll(int64_t i);
 585
 586 extern unsigned int
 587 _mesa_bitcount(unsigned int n);
 588
 589 extern GLhalfARB
 590 _mesa_float_to_half(float f);
 591
 592 extern float
 593 _mesa_half_to_float(GLhalfARB h);
 594
 595
 596 extern void *
 597 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
 598                int (*compar)(const void *, const void *) );
 599
 600 extern char *
 601 _mesa_getenv( const char *var );
 602
 603 extern char *
 604 _mesa_strdup( const char *s );
 605
 606 extern double
 607 _mesa_strtod( const char *s, char **end );
 608
 609 extern unsigned int
 610 _mesa_str_checksum(const char *str);
 611
 612 extern int
 613 _mesa_sprintf( char *str, const char *fmt, ... );
 614
 615 extern int
 616 _mesa_snprintf( char *str, size_t size, const char *fmt, ... );
 617
 618 extern void
 619 _mesa_printf( const char *fmtString, ... );
 620
 621 extern void
 622 _mesa_fprintf( FILE *f, const char *fmtString, ... );
 623
 624 extern int
 625 _mesa_vsprintf( char *str, const char *fmt, va_list args );
 626
 627
 628 extern void
 629 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
 630
 631 extern void
 632 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
 633
 634 extern void
 635 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
 636
 637 extern void
 638 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
 639
 640 #ifdef __cplusplus
 641 }
 642 #endif
 643
 644
 645 #endif /* IMPORTS_H */