2 * Mesa 3-D graphics library
5 * Copyright (C) 1999-2006 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.
42 #include <GL/internal/glcore.h>
50 /**********************************************************************/
51 /** \name General macros */
59 /** gcc -pedantic warns about long string literals, LONGSTRING silences that */
60 #if !defined(__GNUC__) || (__GNUC__ < 2) || \
61 ((__GNUC__ == 2) && (__GNUC_MINOR__ <= 7))
64 # define LONGSTRING __extension__
70 /**********************************************************************/
74 /** Allocate \p BYTES bytes */
75 #define MALLOC(BYTES) _mesa_malloc(BYTES)
76 /** Allocate and zero \p BYTES bytes */
77 #define CALLOC(BYTES) _mesa_calloc(BYTES)
78 /** Allocate a structure of type \p T */
79 #define MALLOC_STRUCT(T) (struct T *) _mesa_malloc(sizeof(struct T))
80 /** Allocate and zero a structure of type \p T */
81 #define CALLOC_STRUCT(T) (struct T *) _mesa_calloc(sizeof(struct T))
83 #define FREE(PTR) _mesa_free(PTR)
85 /** Allocate \p BYTES aligned at \p N bytes */
86 #define ALIGN_MALLOC(BYTES, N) _mesa_align_malloc(BYTES, N)
87 /** Allocate and zero \p BYTES bytes aligned at \p N bytes */
88 #define ALIGN_CALLOC(BYTES, N) _mesa_align_calloc(BYTES, N)
89 /** Allocate a structure of type \p T aligned at \p N bytes */
90 #define ALIGN_MALLOC_STRUCT(T, N) (struct T *) _mesa_align_malloc(sizeof(struct T), N)
91 /** Allocate and zero a structure of type \p T aligned at \p N bytes */
92 #define ALIGN_CALLOC_STRUCT(T, N) (struct T *) _mesa_align_calloc(sizeof(struct T), N)
93 /** Free aligned memory */
94 #define ALIGN_FREE(PTR) _mesa_align_free(PTR)
96 /** Copy \p BYTES bytes from \p SRC into \p DST */
97 #define MEMCPY( DST, SRC, BYTES) _mesa_memcpy(DST, SRC, BYTES)
98 /** Set \p N bytes in \p DST to \p VAL */
99 #define MEMSET( DST, VAL, N ) _mesa_memset(DST, VAL, N)
105 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
106 * as offsets into buffer stores. Since the vertex array pointer and
107 * buffer store pointer are both pointers and we need to add them, we use
109 * Both pointers/offsets are expressed in bytes.
111 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
115 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
116 * as a int (thereby using integer registers instead of FP registers) is
117 * a performance win. Typically, this can be done with ordinary casts.
118 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
119 * these casts generate warnings.
120 * The following union typedef is used to solve that.
122 typedef union { GLfloat f
; GLint i
; } fi_type
;
126 /**********************************************************************
130 #define MAX_GLUSHORT 0xffff
131 #define MAX_GLUINT 0xffffffff
134 #define M_PI (3.1415926536)
138 #define M_E (2.7182818284590452354)
142 #define FLT_MAX_EXP 128
145 /* XXX this is a bit of a hack needed for compilation within XFree86 */
147 #define FLT_MIN (1.0e-37)
150 /* Degrees to radians conversion: */
151 #define DEG2RAD (M_PI/180.0)
155 *** USE_IEEE: Determine if we're using IEEE floating point
157 #if defined(__i386__) || defined(__386__) || defined(__sparc__) || \
158 defined(__s390x__) || defined(__powerpc__) || \
159 defined(__amd64__) || \
160 defined(ia64) || defined(__ia64__) || \
161 defined(__hppa__) || defined(hpux) || \
162 defined(__mips) || defined(_MIPS_ARCH) || \
163 defined(__arm__) || \
164 defined(__sh__) || defined(__m32r__) || \
165 (defined(__alpha__) && (defined(__IEEE_FLOAT) || !defined(VMS)))
167 #define IEEE_ONE 0x3f800000
172 *** SQRTF: single-precision square root
174 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
175 # define SQRTF(X) _mesa_sqrtf(X)
176 #elif defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
177 # define SQRTF(X) (float) xf86sqrt((float) (X))
179 # define SQRTF(X) (float) sqrt((float) (X))
184 *** INV_SQRTF: single-precision inverse square root
187 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
189 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
194 *** LOG2: Log base 2 of float
198 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
199 * Based on code from http://www.stereopsis.com/log2.html
201 static INLINE GLfloat
LOG2(GLfloat x
)
203 const GLfloat y
= x
* x
* x
* x
;
204 const GLuint ix
= *((GLuint
*) &y
);
205 const GLuint exp
= (ix
>> 23) & 0xFF;
206 const GLint log2
= ((GLint
) exp
) - 127;
207 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
210 /* Pretty fast, and accurate.
211 * Based on code from http://www.flipcode.com/totd/
213 static INLINE GLfloat
LOG2(GLfloat val
)
218 log_2
= ((num
.i
>> 23) & 255) - 128;
219 num
.i
&= ~(255 << 23);
221 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
222 return num
.f
+ log_2
;
224 #elif defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
225 #define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
228 * NOTE: log_base_2(x) = log(x) / log(2)
229 * NOTE: 1.442695 = 1/log(2).
231 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
236 *** IS_INF_OR_NAN: test if float is infinite or NaN
239 static INLINE
int IS_INF_OR_NAN( float x
)
243 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
245 #elif defined(isfinite)
246 #define IS_INF_OR_NAN(x) (!isfinite(x))
247 #elif defined(finite)
248 #define IS_INF_OR_NAN(x) (!finite(x))
250 #define IS_INF_OR_NAN(x) (!finite(x))
251 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
252 #define IS_INF_OR_NAN(x) (!isfinite(x))
254 #define IS_INF_OR_NAN(x) (!finite(x))
259 *** IS_NEGATIVE: test if float is negative
261 #if defined(USE_IEEE)
262 static INLINE
int GET_FLOAT_BITS( float x
)
268 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
270 #define IS_NEGATIVE(x) (x < 0.0F)
275 *** DIFFERENT_SIGNS: test if two floats have opposite signs
277 #if defined(USE_IEEE)
278 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
280 /* Could just use (x*y<0) except for the flatshading requirements.
281 * Maybe there's a better way?
283 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
288 *** CEILF: ceiling of float
289 *** FLOORF: floor of float
290 *** FABSF: absolute value of float
291 *** LOGF: the natural logarithm (base e) of the value
292 *** EXPF: raise e to the value
293 *** LDEXPF: multiply value by an integral power of two
294 *** FREXPF: extract mantissa and exponent from value
296 #if defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
297 #define CEILF(x) ((GLfloat) xf86ceil(x))
298 #define FLOORF(x) ((GLfloat) xf86floor(x))
299 #define FABSF(x) ((GLfloat) xf86fabs(x))
300 #define LOGF(x) ((GLfloat) xf86log(x))
301 #define EXPF(x) ((GLfloat) xf86exp(x))
302 #define LDEXPF(x,y) ((GLfloat) xf86ldexp(x,y))
303 #define FREXPF(x,y) ((GLfloat) xf86frexp(x,y))
304 #elif defined(__gnu_linux__)
306 #define CEILF(x) ceilf(x)
307 #define FLOORF(x) floorf(x)
308 #define FABSF(x) fabsf(x)
309 #define LOGF(x) logf(x)
310 #define EXPF(x) expf(x)
311 #define LDEXPF(x,y) ldexpf(x,y)
312 #define FREXPF(x,y) frexpf(x,y)
314 #define CEILF(x) ((GLfloat) ceil(x))
315 #define FLOORF(x) ((GLfloat) floor(x))
316 #define FABSF(x) ((GLfloat) fabs(x))
317 #define LOGF(x) ((GLfloat) log(x))
318 #define EXPF(x) ((GLfloat) exp(x))
319 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
320 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
325 *** IROUND: return (as an integer) float rounded to nearest integer
327 #if defined(USE_SPARC_ASM) && defined(__GNUC__) && defined(__sparc__)
328 static INLINE
int iround(float f
)
331 __asm__ ("fstoi %1, %0" : "=f" (r
) : "f" (f
));
334 #define IROUND(x) iround(x)
335 #elif defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
336 (!defined(__BEOS__) || (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
337 static INLINE
int iround(float f
)
340 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
343 #define IROUND(x) iround(x)
344 #elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__)
345 static INLINE
int iround(float f
)
354 #define IROUND(x) iround(x)
355 #elif defined(__WATCOMC__) && defined(__386__)
356 long iround(float f
);
357 #pragma aux iround = \
359 "fistp dword ptr [esp]" \
364 #define IROUND(x) iround(x)
366 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
371 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
374 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
376 #define IROUND_POS(f) (IROUND(f))
381 *** IFLOOR: return (as an integer) floor of float
383 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
385 * IEEE floor for computers that round to nearest or even.
386 * 'f' must be between -4194304 and 4194303.
387 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
388 * but uses some IEEE specific tricks for better speed.
389 * Contributed by Josh Vanderhoof
391 static INLINE
int ifloor(float f
)
395 af
= (3 << 22) + 0.5 + (double)f
;
396 bf
= (3 << 22) + 0.5 - (double)f
;
397 /* GCC generates an extra fstp/fld without this. */
398 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
399 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
400 return (ai
- bi
) >> 1;
402 #define IFLOOR(x) ifloor(x)
403 #elif defined(USE_IEEE)
404 static INLINE
int ifloor(float f
)
410 af
= (3 << 22) + 0.5 + (double)f
;
411 bf
= (3 << 22) + 0.5 - (double)f
;
412 u
.f
= (float) af
; ai
= u
.i
;
413 u
.f
= (float) bf
; bi
= u
.i
;
414 return (ai
- bi
) >> 1;
416 #define IFLOOR(x) ifloor(x)
418 static INLINE
int ifloor(float f
)
421 return (i
> f
) ? i
- 1 : i
;
423 #define IFLOOR(x) ifloor(x)
428 *** ICEIL: return (as an integer) ceiling of float
430 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
432 * IEEE ceil for computers that round to nearest or even.
433 * 'f' must be between -4194304 and 4194303.
434 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
435 * but uses some IEEE specific tricks for better speed.
436 * Contributed by Josh Vanderhoof
438 static INLINE
int iceil(float f
)
442 af
= (3 << 22) + 0.5 + (double)f
;
443 bf
= (3 << 22) + 0.5 - (double)f
;
444 /* GCC generates an extra fstp/fld without this. */
445 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
446 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
447 return (ai
- bi
+ 1) >> 1;
449 #define ICEIL(x) iceil(x)
450 #elif defined(USE_IEEE)
451 static INLINE
int iceil(float f
)
456 af
= (3 << 22) + 0.5 + (double)f
;
457 bf
= (3 << 22) + 0.5 - (double)f
;
458 u
.f
= (float) af
; ai
= u
.i
;
459 u
.f
= (float) bf
; bi
= u
.i
;
460 return (ai
- bi
+ 1) >> 1;
462 #define ICEIL(x) iceil(x)
464 static INLINE
int iceil(float f
)
467 return (i
< f
) ? i
+ 1 : i
;
469 #define ICEIL(x) iceil(x)
474 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
475 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
477 #if defined(USE_IEEE) && !defined(DEBUG)
478 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
479 /* This function/macro is sensitive to precision. Test very carefully
482 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
488 else if (__tmp.i >= IEEE_0996) \
489 UB = (GLubyte) 255; \
491 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
492 UB = (GLubyte) __tmp.i; \
495 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
498 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
499 UB = (GLubyte) __tmp.i; \
502 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
503 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
504 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
505 ub = ((GLubyte) IROUND((f) * 255.0F))
510 *** START_FAST_MATH: Set x86 FPU to faster, 32-bit precision mode (and save
511 *** original mode to a temporary).
512 *** END_FAST_MATH: Restore x86 FPU to original mode.
514 #if defined(__GNUC__) && defined(__i386__)
516 * Set the x86 FPU control word to guarentee only 32 bits of precision
517 * are stored in registers. Allowing the FPU to store more introduces
518 * differences between situations where numbers are pulled out of memory
519 * vs. situations where the compiler is able to optimize register usage.
521 * In the worst case, we force the compiler to use a memory access to
522 * truncate the float, by specifying the 'volatile' keyword.
524 /* Hardware default: All exceptions masked, extended double precision,
525 * round to nearest (IEEE compliant):
527 #define DEFAULT_X86_FPU 0x037f
528 /* All exceptions masked, single precision, round to nearest:
530 #define FAST_X86_FPU 0x003f
531 /* The fldcw instruction will cause any pending FP exceptions to be
532 * raised prior to entering the block, and we clear any pending
533 * exceptions before exiting the block. Hence, asm code has free
534 * reign over the FPU while in the fast math block.
536 #if defined(NO_FAST_MATH)
537 #define START_FAST_MATH(x) \
539 static GLuint mask = DEFAULT_X86_FPU; \
540 __asm__ ( "fnstcw %0" : "=m" (*&(x)) ); \
541 __asm__ ( "fldcw %0" : : "m" (mask) ); \
544 #define START_FAST_MATH(x) \
546 static GLuint mask = FAST_X86_FPU; \
547 __asm__ ( "fnstcw %0" : "=m" (*&(x)) ); \
548 __asm__ ( "fldcw %0" : : "m" (mask) ); \
551 /* Restore original FPU mode, and clear any exceptions that may have
552 * occurred in the FAST_MATH block.
554 #define END_FAST_MATH(x) \
556 __asm__ ( "fnclex ; fldcw %0" : : "m" (*&(x)) ); \
559 #elif defined(__WATCOMC__) && defined(__386__)
560 #define DEFAULT_X86_FPU 0x037f /* See GCC comments above */
561 #define FAST_X86_FPU 0x003f /* See GCC comments above */
562 void _watcom_start_fast_math(unsigned short *x
,unsigned short *mask
);
563 #pragma aux _watcom_start_fast_math = \
564 "fnstcw word ptr [eax]" \
565 "fldcw word ptr [ecx]" \
568 void _watcom_end_fast_math(unsigned short *x
);
569 #pragma aux _watcom_end_fast_math = \
571 "fldcw word ptr [eax]" \
574 #if defined(NO_FAST_MATH)
575 #define START_FAST_MATH(x) \
577 static GLushort mask = DEFAULT_X86_FPU; \
578 _watcom_start_fast_math(&x,&mask); \
581 #define START_FAST_MATH(x) \
583 static GLushort mask = FAST_X86_FPU; \
584 _watcom_start_fast_math(&x,&mask); \
587 #define END_FAST_MATH(x) _watcom_end_fast_math(&x)
589 #elif defined(_MSC_VER) && defined(_M_IX86)
590 #define DEFAULT_X86_FPU 0x037f /* See GCC comments above */
591 #define FAST_X86_FPU 0x003f /* See GCC comments above */
592 #if defined(NO_FAST_MATH)
593 #define START_FAST_MATH(x) do {\
594 static GLuint mask = DEFAULT_X86_FPU;\
595 __asm fnstcw word ptr [x]\
596 __asm fldcw word ptr [mask]\
599 #define START_FAST_MATH(x) do {\
600 static GLuint mask = FAST_X86_FPU;\
601 __asm fnstcw word ptr [x]\
602 __asm fldcw word ptr [mask]\
605 #define END_FAST_MATH(x) do {\
607 __asm fldcw word ptr [x]\
611 #define START_FAST_MATH(x) x = 0
612 #define END_FAST_MATH(x) (void)(x)
617 * Return 1 if this is a little endian machine, 0 if big endian.
619 static INLINE GLboolean
620 _mesa_little_endian(void)
622 const GLuint ui
= 1; /* intentionally not static */
623 return *((const GLubyte
*) &ui
);
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_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
652 unsigned long alignment
);
655 _mesa_exec_malloc( GLuint size
);
658 _mesa_exec_free( void *addr
);
661 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
664 _mesa_memcpy( void *dest
, const void *src
, size_t n
);
667 _mesa_memset( void *dst
, int val
, size_t n
);
670 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
673 _mesa_bzero( void *dst
, size_t n
);
676 _mesa_memcmp( const void *s1
, const void *s2
, size_t n
);
688 _mesa_asinf(float x
);
691 _mesa_atanf(float x
);
694 _mesa_sqrtd(double x
);
697 _mesa_sqrtf(float x
);
700 _mesa_inv_sqrtf(float x
);
703 _mesa_pow(double x
, double y
);
709 _mesa_bitcount(unsigned int n
);
712 _mesa_float_to_half(float f
);
715 _mesa_half_to_float(GLhalfARB h
);
719 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
720 int (*compar
)(const void *, const void *) );
723 _mesa_getenv( const char *var
);
726 _mesa_strstr( const char *haystack
, const char *needle
);
729 _mesa_strncat( char *dest
, const char *src
, size_t n
);
732 _mesa_strcpy( char *dest
, const char *src
);
735 _mesa_strncpy( char *dest
, const char *src
, size_t n
);
738 _mesa_strlen( const char *s
);
741 _mesa_strcmp( const char *s1
, const char *s2
);
744 _mesa_strncmp( const char *s1
, const char *s2
, size_t n
);
747 _mesa_strdup( const char *s
);
750 _mesa_atoi( const char *s
);
753 _mesa_strtod( const char *s
, char **end
);
756 _mesa_sprintf( char *str
, const char *fmt
, ... );
759 _mesa_printf( const char *fmtString
, ... );
762 _mesa_vsprintf( char *str
, const char *fmt
, va_list args
);
766 _mesa_warning( __GLcontext
*gc
, const char *fmtString
, ... );
769 _mesa_problem( const __GLcontext
*ctx
, const char *fmtString
, ... );
772 _mesa_error( __GLcontext
*ctx
, GLenum error
, const char *fmtString
, ... );
775 _mesa_debug( const __GLcontext
*ctx
, const char *fmtString
, ... );
778 _mesa_exit( int status
);
782 _mesa_init_default_imports( __GLimports
*imports
, void *driverCtx
);
790 #endif /* IMPORTS_H */