2 * Mesa 3-D graphics library
5 * Copyright (C) 1999-2004 Brian Paul All Rights Reserved.
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:
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
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.
28 * Standard C library function wrappers.
30 * This file provides wrappers for all the standard C library functions
31 * like malloc(), free(), printf(), getenv(), etc.
39 /* XXX some of the stuff in glheader.h should be moved into this file.
49 /**********************************************************************/
50 /** \name General macros */
60 /**********************************************************************/
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))
73 #define FREE(PTR) _mesa_free(PTR)
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)
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)
95 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
96 * as offsets into buffer stores. Since the vertex array pointer and
97 * buffer store pointer are both pointers and we need to add them, we use
99 * Both pointers/offsets are expressed in bytes.
101 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
104 /**********************************************************************/
105 /** \name [Pseudo] static array declaration.
107 * MACs and BeOS don't support static larger than 32kb, so ...
113 * Define a [static] unidimensional array
118 * Define a [static] bi-dimensional array
123 * Define a [static] tri-dimensional array
128 * Verifies a [static] array was properly allocated.
133 * Undefine (free) a [static] array.
136 #if defined(macintosh) && !defined(__MRC__)
137 /*extern char *AGLAlloc(int size);*/
138 /*extern void AGLFree(char* ptr);*/
139 # define DEFARRAY(TYPE,NAME,SIZE) TYPE *NAME = (TYPE*)_mesa_alloc(sizeof(TYPE)*(SIZE))
140 # define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2) TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
141 # define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3) TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_alloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
142 # define CHECKARRAY(NAME,CMD) do {if (!(NAME)) {CMD;}} while (0)
143 # define UNDEFARRAY(NAME) do {if ((NAME)) {_mesa_free((char*)NAME);} }while (0)
144 #elif defined(__BEOS__)
145 # define DEFARRAY(TYPE,NAME,SIZE) TYPE *NAME = (TYPE*)_mesa_malloc(sizeof(TYPE)*(SIZE))
146 # define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2) TYPE (*NAME)[SIZE2] = (TYPE(*)[SIZE2])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2))
147 # define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3) TYPE (*NAME)[SIZE2][SIZE3] = (TYPE(*)[SIZE2][SIZE3])_mesa_malloc(sizeof(TYPE)*(SIZE1)*(SIZE2)*(SIZE3))
148 # define CHECKARRAY(NAME,CMD) do {if (!(NAME)) {CMD;}} while (0)
149 # define UNDEFARRAY(NAME) do {if ((NAME)) {_mesa_free((char*)NAME);} }while (0)
151 # define DEFARRAY(TYPE,NAME,SIZE) TYPE NAME[SIZE]
152 # define DEFMARRAY(TYPE,NAME,SIZE1,SIZE2) TYPE NAME[SIZE1][SIZE2]
153 # define DEFMNARRAY(TYPE,NAME,SIZE1,SIZE2,SIZE3) TYPE NAME[SIZE1][SIZE2][SIZE3]
154 # define CHECKARRAY(NAME,CMD) do {} while(0)
155 # define UNDEFARRAY(NAME)
162 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
163 * as a int (thereby using integer registers instead of FP registers) is
164 * a performance win. Typically, this can be done with ordinary casts.
165 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
166 * these casts generate warnings.
167 * The following union typedef is used to solve that.
169 typedef union { GLfloat f
; GLint i
; } fi_type
;
173 /**********************************************************************
177 #define MAX_GLUSHORT 0xffff
178 #define MAX_GLUINT 0xffffffff
181 #define M_PI (3.1415926536)
185 #define M_E (2.7182818284590452354)
189 /* XXX this is a bit of a hack needed for compilation within XFree86 */
191 #define FLT_MIN (1.0e-37)
194 /* Degrees to radians conversion: */
195 #define DEG2RAD (M_PI/180.0)
199 *** USE_IEEE: Determine if we're using IEEE floating point
201 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
202 defined(__s390x__) || defined(__powerpc__) || \
203 defined(__amd64__) || \
204 defined(ia64) || defined(__ia64__) || \
205 defined(__hppa__) || defined(hpux) || \
206 defined(__mips) || defined(_MIPS_ARCH) || \
207 defined(__arm__) || \
208 (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
210 #define IEEE_ONE 0x3f800000
215 *** SQRTF: single-precision square root
217 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
218 # define SQRTF(X) _mesa_sqrtf(X)
219 #elif defined(XFree86LOADER) && defined(IN_MODULE)
220 # define SQRTF(X) (float) xf86sqrt((float) (X))
222 # define SQRTF(X) (float) sqrt((float) (X))
227 *** INV_SQRTF: single-precision inverse square root
230 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
232 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
237 *** LOG2: Log base 2 of float
241 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
242 * Based on code from http://www.stereopsis.com/log2.html
244 static INLINE GLfloat
LOG2(GLfloat x
)
246 const GLfloat y
= x
* x
* x
* x
;
247 const GLuint ix
= *((GLuint
*) &y
);
248 const GLuint exp
= (ix
>> 23) & 0xFF;
249 const GLint log2
= ((GLint
) exp
) - 127;
250 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
253 /* Pretty fast, and accurate.
254 * Based on code from http://www.flipcode.com/totd/
256 static INLINE GLfloat
LOG2(GLfloat val
)
261 log_2
= ((num
.i
>> 23) & 255) - 128;
262 num
.i
&= ~(255 << 23);
264 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
265 return num
.f
+ log_2
;
267 #elif defined(XFree86LOADER) && defined(IN_MODULE)
268 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
271 * NOTE: log_base_2(x) = log(x) / log(2)
272 * NOTE: 1.442695 = 1/log(2).
274 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
279 *** IS_INF_OR_NAN: test if float is infinite or NaN
282 static INLINE
int IS_INF_OR_NAN( float x
)
286 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
288 #elif defined(isfinite)
289 #define IS_INF_OR_NAN(x) (!isfinite(x))
290 #elif defined(finite)
291 #define IS_INF_OR_NAN(x) (!finite(x))
293 #define IS_INF_OR_NAN(x) (!finite(x))
294 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
295 #define IS_INF_OR_NAN(x) (!isfinite(x))
297 #define IS_INF_OR_NAN(x) (!finite(x))
302 *** IS_NEGATIVE: test if float is negative
304 #if defined(USE_IEEE)
305 static INLINE
int GET_FLOAT_BITS( float x
)
311 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
313 #define IS_NEGATIVE(x) (x < 0.0F)
318 *** DIFFERENT_SIGNS: test if two floats have opposite signs
320 #if defined(USE_IEEE)
321 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
323 /* Could just use (x*y<0) except for the flatshading requirements.
324 * Maybe there's a better way?
326 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
331 *** CEILF: ceiling of float
332 *** FLOORF: floor of float
333 *** FABSF: absolute value of float
335 #if defined(XFree86LOADER) && defined(IN_MODULE)
336 #define CEILF(x) ((GLfloat) xf86ceil(x))
337 #define FLOORF(x) ((GLfloat) xf86floor(x))
338 #define FABSF(x) ((GLfloat) xf86fabs(x))
339 #elif defined(__gnu_linux__)
341 #define CEILF(x) ceilf(x)
342 #define FLOORF(x) floorf(x)
343 #define FABSF(x) fabsf(x)
345 #define CEILF(x) ((GLfloat) ceil(x))
346 #define FLOORF(x) ((GLfloat) floor(x))
347 #define FABSF(x) ((GLfloat) fabs(x))
352 *** IROUND: return (as an integer) float rounded to nearest integer
354 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
355 static INLINE
int iround(float f
)
358 __asm__ ("fstoi %1, %0" : "=f" (r
) : "f" (f
));
361 #define IROUND(x) iround(x)
362 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
363 (!defined(__BEOS__) || (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
364 static INLINE
int iround(float f
)
367 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
370 #define IROUND(x) iround(x)
371 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
372 static INLINE
int iround(float f
)
381 #define IROUND(x) iround(x)
382 #elif defined(__WATCOMC__) && defined(__386__)
383 long iround(float f
);
384 #pragma aux iround = \
386 "fistp dword ptr [esp]" \
391 #define IROUND(x) iround(x)
393 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
398 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
401 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
403 #define IROUND_POS(f) (IROUND(f))
408 *** IFLOOR: return (as an integer) floor of float
410 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
412 * IEEE floor for computers that round to nearest or even.
413 * 'f' must be between -4194304 and 4194303.
414 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
415 * but uses some IEEE specific tricks for better speed.
416 * Contributed by Josh Vanderhoof
418 static INLINE
int ifloor(float f
)
422 af
= (3 << 22) + 0.5 + (double)f
;
423 bf
= (3 << 22) + 0.5 - (double)f
;
424 /* GCC generates an extra fstp/fld without this. */
425 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
426 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
427 return (ai
- bi
) >> 1;
429 #define IFLOOR(x) ifloor(x)
430 #elif defined(USE_IEEE)
431 static INLINE
int ifloor(float f
)
437 af
= (3 << 22) + 0.5 + (double)f
;
438 bf
= (3 << 22) + 0.5 - (double)f
;
439 u
.f
= (float) af
; ai
= u
.i
;
440 u
.f
= (float) bf
; bi
= u
.i
;
441 return (ai
- bi
) >> 1;
443 #define IFLOOR(x) ifloor(x)
445 static INLINE
int ifloor(float f
)
448 return (i
> f
) ? i
- 1 : i
;
450 #define IFLOOR(x) ifloor(x)
455 *** ICEIL: return (as an integer) ceiling of float
457 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
459 * IEEE ceil for computers that round to nearest or even.
460 * 'f' must be between -4194304 and 4194303.
461 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
462 * but uses some IEEE specific tricks for better speed.
463 * Contributed by Josh Vanderhoof
465 static INLINE
int iceil(float f
)
469 af
= (3 << 22) + 0.5 + (double)f
;
470 bf
= (3 << 22) + 0.5 - (double)f
;
471 /* GCC generates an extra fstp/fld without this. */
472 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
473 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
474 return (ai
- bi
+ 1) >> 1;
476 #define ICEIL(x) iceil(x)
477 #elif defined(USE_IEEE)
478 static INLINE
int iceil(float f
)
483 af
= (3 << 22) + 0.5 + (double)f
;
484 bf
= (3 << 22) + 0.5 - (double)f
;
485 u
.f
= (float) af
; ai
= u
.i
;
486 u
.f
= (float) bf
; bi
= u
.i
;
487 return (ai
- bi
+ 1) >> 1;
489 #define ICEIL(x) iceil(x)
491 static INLINE
int iceil(float f
)
494 return (i
< f
) ? i
+ 1 : i
;
496 #define ICEIL(x) iceil(x)
501 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
502 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
504 #if defined(USE_IEEE) && !defined(DEBUG)
505 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
506 /* This function/macro is sensitive to precision. Test very carefully
509 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
515 else if (__tmp.i >= IEEE_0996) \
516 UB = (GLubyte) 255; \
518 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
519 UB = (GLubyte) __tmp.i; \
522 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
525 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
526 UB = (GLubyte) __tmp.i; \
529 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
530 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
531 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
532 ub = ((GLubyte) IROUND((f) * 255.0F))
537 *** COPY_FLOAT: copy a float from src to dest.
539 #define COPY_FLOAT( dst, src ) (dst) = (src)
543 *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
544 *** original mode to a temporary).
545 *** END_FAST_MATH: Restore x86 FPU to original mode.
547 #if defined(__GNUC__) && defined(__i386__)
549 * Set the x86 FPU control word to guarentee only 32 bits of precision
550 * are stored in registers. Allowing the FPU to store more introduces
551 * differences between situations where numbers are pulled out of memory
552 * vs. situations where the compiler is able to optimize register usage.
554 * In the worst case, we force the compiler to use a memory access to
555 * truncate the float, by specifying the 'volatile' keyword.
557 /* Hardware default: All exceptions masked, extended double precision,
558 * round to nearest (IEEE compliant):
560 #define DEFAULT_X86_FPU 0x037f
561 /* All exceptions masked, single precision, round to nearest:
563 #define FAST_X86_FPU 0x003f
564 /* The fldcw instruction will cause any pending FP exceptions to be
565 * raised prior to entering the block, and we clear any pending
566 * exceptions before exiting the block. Hence, asm code has free
567 * reign over the FPU while in the fast math block.
569 #if defined(NO_FAST_MATH)
570 #define START_FAST_MATH(x) \
572 static GLuint mask = DEFAULT_X86_FPU; \
573 __asm__ ( "fnstcw %0" : "=m" (*&(x)) ); \
574 __asm__ ( "fldcw %0" : : "m" (mask) ); \
577 #define START_FAST_MATH(x) \
579 static GLuint mask = FAST_X86_FPU; \
580 __asm__ ( "fnstcw %0" : "=m" (*&(x)) ); \
581 __asm__ ( "fldcw %0" : : "m" (mask) ); \
584 /* Restore original FPU mode, and clear any exceptions that may have
585 * occurred in the FAST_MATH block.
587 #define END_FAST_MATH(x) \
589 __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) ); \
592 #elif defined(__WATCOMC__) && defined(__386__)
593 #define DEFAULT_X86_FPU 0x037f /* See GCC comments above */
594 #define FAST_X86_FPU 0x003f /* See GCC comments above */
595 void _watcom_start_fast_math(unsigned short *x
,unsigned short *mask
);
596 #pragma aux _watcom_start_fast_math = \
597 "fnstcw word ptr [eax]" \
598 "fldcw word ptr [ecx]" \
601 void _watcom_end_fast_math(unsigned short *x
);
602 #pragma aux _watcom_end_fast_math = \
604 "fldcw word ptr [eax]" \
607 #if defined(NO_FAST_MATH)
608 #define START_FAST_MATH(x) \
610 static GLushort mask = DEFAULT_X86_FPU; \
611 _watcom_start_fast_math(&x,&mask); \
614 #define START_FAST_MATH(x) \
616 static GLushort mask = FAST_X86_FPU; \
617 _watcom_start_fast_math(&x,&mask); \
620 #define END_FAST_MATH(x) _watcom_end_fast_math(&x)
622 #define START_FAST_MATH(x) x = 0
623 #define END_FAST_MATH(x) (void)(x)
628 /**********************************************************************
633 _mesa_malloc( size_t bytes
);
636 _mesa_calloc( size_t bytes
);
639 _mesa_free( void *ptr
);
642 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
645 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
648 _mesa_align_free( void *ptr
);
651 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
654 _mesa_memcpy( void *dest
, const void *src
, size_t n
);
657 _mesa_memset( void *dst
, int val
, size_t n
);
660 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
663 _mesa_bzero( void *dst
, size_t n
);
673 _mesa_sqrtd(double x
);
676 _mesa_sqrtf(float x
);
679 _mesa_inv_sqrtf(float x
);
682 _mesa_pow(double x
, double y
);
688 _mesa_bitcount(unsigned int n
);
691 _mesa_float_to_half(float f
);
694 _mesa_half_to_float(GLhalfARB h
);
698 _mesa_getenv( const char *var
);
701 _mesa_strstr( const char *haystack
, const char *needle
);
704 _mesa_strncat( char *dest
, const char *src
, size_t n
);
707 _mesa_strcpy( char *dest
, const char *src
);
710 _mesa_strncpy( char *dest
, const char *src
, size_t n
);
713 _mesa_strlen( const char *s
);
716 _mesa_strcmp( const char *s1
, const char *s2
);
719 _mesa_strncmp( const char *s1
, const char *s2
, size_t n
);
722 _mesa_strdup( const char *s
);
725 _mesa_atoi( const char *s
);
728 _mesa_strtod( const char *s
, char **end
);
731 _mesa_sprintf( char *str
, const char *fmt
, ... );
734 _mesa_printf( const char *fmtString
, ... );
738 _mesa_warning( __GLcontext
*gc
, const char *fmtString
, ... );
741 _mesa_problem( const __GLcontext
*ctx
, const char *fmtString
, ... );
744 _mesa_error( __GLcontext
*ctx
, GLenum error
, const char *fmtString
, ... );
747 _mesa_debug( const __GLcontext
*ctx
, const char *fmtString
, ... );
751 _mesa_init_default_imports( __GLimports
*imports
, void *driverCtx
);
759 #endif /* IMPORTS_H */