Home | History | Annotate | Download | only in util
      1 /**************************************************************************
      2  *
      3  * Copyright 2008 VMware, Inc.
      4  * All Rights Reserved.
      5  *
      6  * Permission is hereby granted, free of charge, to any person obtaining a
      7  * copy of this software and associated documentation files (the
      8  * "Software"), to deal in the Software without restriction, including
      9  * without limitation the rights to use, copy, modify, merge, publish,
     10  * distribute, sub license, and/or sell copies of the Software, and to
     11  * permit persons to whom the Software is furnished to do so, subject to
     12  * the following conditions:
     13  *
     14  * The above copyright notice and this permission notice (including the
     15  * next paragraph) shall be included in all copies or substantial portions
     16  * of the Software.
     17  *
     18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
     19  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
     20  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
     21  * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
     22  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
     23  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
     24  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
     25  *
     26  **************************************************************************/
     27 
     28 
     29 /**
     30  * Math utilities and approximations for common math functions.
     31  * Reduced precision is usually acceptable in shaders...
     32  *
     33  * "fast" is used in the names of functions which are low-precision,
     34  * or at least lower-precision than the normal C lib functions.
     35  */
     36 
     37 
     38 #ifndef U_MATH_H
     39 #define U_MATH_H
     40 
     41 
     42 #include "pipe/p_compiler.h"
     43 
     44 #include "c99_math.h"
     45 #include <assert.h>
     46 #include <float.h>
     47 #include <stdarg.h>
     48 
     49 #include "util/bitscan.h"
     50 
     51 #ifdef __cplusplus
     52 extern "C" {
     53 #endif
     54 
     55 
     56 #ifndef M_SQRT2
     57 #define M_SQRT2 1.41421356237309504880
     58 #endif
     59 
     60 #define POW2_TABLE_SIZE_LOG2 9
     61 #define POW2_TABLE_SIZE (1 << POW2_TABLE_SIZE_LOG2)
     62 #define POW2_TABLE_OFFSET (POW2_TABLE_SIZE/2)
     63 #define POW2_TABLE_SCALE ((float)(POW2_TABLE_SIZE/2))
     64 extern float pow2_table[POW2_TABLE_SIZE];
     65 
     66 
     67 /**
     68  * Initialize math module.  This should be called before using any
     69  * other functions in this module.
     70  */
     71 extern void
     72 util_init_math(void);
     73 
     74 
     75 union fi {
     76    float f;
     77    int32_t i;
     78    uint32_t ui;
     79 };
     80 
     81 
     82 union di {
     83    double d;
     84    int64_t i;
     85    uint64_t ui;
     86 };
     87 
     88 
     89 /**
     90  * Extract the IEEE float32 exponent.
     91  */
     92 static inline signed
     93 util_get_float32_exponent(float x)
     94 {
     95    union fi f;
     96 
     97    f.f = x;
     98 
     99    return ((f.ui >> 23) & 0xff) - 127;
    100 }
    101 
    102 
    103 /**
    104  * Fast version of 2^x
    105  * Identity: exp2(a + b) = exp2(a) * exp2(b)
    106  * Let ipart = int(x)
    107  * Let fpart = x - ipart;
    108  * So, exp2(x) = exp2(ipart) * exp2(fpart)
    109  * Compute exp2(ipart) with i << ipart
    110  * Compute exp2(fpart) with lookup table.
    111  */
    112 static inline float
    113 util_fast_exp2(float x)
    114 {
    115    int32_t ipart;
    116    float fpart, mpart;
    117    union fi epart;
    118 
    119    if(x > 129.00000f)
    120       return 3.402823466e+38f;
    121 
    122    if (x < -126.99999f)
    123       return 0.0f;
    124 
    125    ipart = (int32_t) x;
    126    fpart = x - (float) ipart;
    127 
    128    /* same as
    129     *   epart.f = (float) (1 << ipart)
    130     * but faster and without integer overflow for ipart > 31
    131     */
    132    epart.i = (ipart + 127 ) << 23;
    133 
    134    mpart = pow2_table[POW2_TABLE_OFFSET + (int)(fpart * POW2_TABLE_SCALE)];
    135 
    136    return epart.f * mpart;
    137 }
    138 
    139 
    140 /**
    141  * Fast approximation to exp(x).
    142  */
    143 static inline float
    144 util_fast_exp(float x)
    145 {
    146    const float k = 1.44269f; /* = log2(e) */
    147    return util_fast_exp2(k * x);
    148 }
    149 
    150 
    151 #define LOG2_TABLE_SIZE_LOG2 16
    152 #define LOG2_TABLE_SCALE (1 << LOG2_TABLE_SIZE_LOG2)
    153 #define LOG2_TABLE_SIZE (LOG2_TABLE_SCALE + 1)
    154 extern float log2_table[LOG2_TABLE_SIZE];
    155 
    156 
    157 /**
    158  * Fast approximation to log2(x).
    159  */
    160 static inline float
    161 util_fast_log2(float x)
    162 {
    163    union fi num;
    164    float epart, mpart;
    165    num.f = x;
    166    epart = (float)(((num.i & 0x7f800000) >> 23) - 127);
    167    /* mpart = log2_table[mantissa*LOG2_TABLE_SCALE + 0.5] */
    168    mpart = log2_table[((num.i & 0x007fffff) + (1 << (22 - LOG2_TABLE_SIZE_LOG2))) >> (23 - LOG2_TABLE_SIZE_LOG2)];
    169    return epart + mpart;
    170 }
    171 
    172 
    173 /**
    174  * Fast approximation to x^y.
    175  */
    176 static inline float
    177 util_fast_pow(float x, float y)
    178 {
    179    return util_fast_exp2(util_fast_log2(x) * y);
    180 }
    181 
    182 /* Note that this counts zero as a power of two.
    183  */
    184 static inline boolean
    185 util_is_power_of_two( unsigned v )
    186 {
    187    return (v & (v-1)) == 0;
    188 }
    189 
    190 
    191 /**
    192  * Floor(x), returned as int.
    193  */
    194 static inline int
    195 util_ifloor(float f)
    196 {
    197    int ai, bi;
    198    double af, bf;
    199    union fi u;
    200    af = (3 << 22) + 0.5 + (double) f;
    201    bf = (3 << 22) + 0.5 - (double) f;
    202    u.f = (float) af;  ai = u.i;
    203    u.f = (float) bf;  bi = u.i;
    204    return (ai - bi) >> 1;
    205 }
    206 
    207 
    208 /**
    209  * Round float to nearest int.
    210  */
    211 static inline int
    212 util_iround(float f)
    213 {
    214 #if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86)
    215    int r;
    216    __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
    217    return r;
    218 #elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86)
    219    int r;
    220    _asm {
    221       fld f
    222       fistp r
    223    }
    224    return r;
    225 #else
    226    if (f >= 0.0f)
    227       return (int) (f + 0.5f);
    228    else
    229       return (int) (f - 0.5f);
    230 #endif
    231 }
    232 
    233 
    234 /**
    235  * Approximate floating point comparison
    236  */
    237 static inline boolean
    238 util_is_approx(float a, float b, float tol)
    239 {
    240    return fabsf(b - a) <= tol;
    241 }
    242 
    243 
    244 /**
    245  * util_is_X_inf_or_nan = test if x is NaN or +/- Inf
    246  * util_is_X_nan        = test if x is NaN
    247  * util_X_inf_sign      = return +1 for +Inf, -1 for -Inf, or 0 for not Inf
    248  *
    249  * NaN can be checked with x != x, however this fails with the fast math flag
    250  **/
    251 
    252 
    253 /**
    254  * Single-float
    255  */
    256 static inline boolean
    257 util_is_inf_or_nan(float x)
    258 {
    259    union fi tmp;
    260    tmp.f = x;
    261    return (tmp.ui & 0x7f800000) == 0x7f800000;
    262 }
    263 
    264 
    265 static inline boolean
    266 util_is_nan(float x)
    267 {
    268    union fi tmp;
    269    tmp.f = x;
    270    return (tmp.ui & 0x7fffffff) > 0x7f800000;
    271 }
    272 
    273 
    274 static inline int
    275 util_inf_sign(float x)
    276 {
    277    union fi tmp;
    278    tmp.f = x;
    279    if ((tmp.ui & 0x7fffffff) != 0x7f800000) {
    280       return 0;
    281    }
    282 
    283    return (x < 0) ? -1 : 1;
    284 }
    285 
    286 
    287 /**
    288  * Double-float
    289  */
    290 static inline boolean
    291 util_is_double_inf_or_nan(double x)
    292 {
    293    union di tmp;
    294    tmp.d = x;
    295    return (tmp.ui & 0x7ff0000000000000ULL) == 0x7ff0000000000000ULL;
    296 }
    297 
    298 
    299 static inline boolean
    300 util_is_double_nan(double x)
    301 {
    302    union di tmp;
    303    tmp.d = x;
    304    return (tmp.ui & 0x7fffffffffffffffULL) > 0x7ff0000000000000ULL;
    305 }
    306 
    307 
    308 static inline int
    309 util_double_inf_sign(double x)
    310 {
    311    union di tmp;
    312    tmp.d = x;
    313    if ((tmp.ui & 0x7fffffffffffffffULL) != 0x7ff0000000000000ULL) {
    314       return 0;
    315    }
    316 
    317    return (x < 0) ? -1 : 1;
    318 }
    319 
    320 
    321 /**
    322  * Half-float
    323  */
    324 static inline boolean
    325 util_is_half_inf_or_nan(int16_t x)
    326 {
    327    return (x & 0x7c00) == 0x7c00;
    328 }
    329 
    330 
    331 static inline boolean
    332 util_is_half_nan(int16_t x)
    333 {
    334    return (x & 0x7fff) > 0x7c00;
    335 }
    336 
    337 
    338 static inline int
    339 util_half_inf_sign(int16_t x)
    340 {
    341    if ((x & 0x7fff) != 0x7c00) {
    342       return 0;
    343    }
    344 
    345    return (x < 0) ? -1 : 1;
    346 }
    347 
    348 
    349 /**
    350  * Return float bits.
    351  */
    352 static inline unsigned
    353 fui( float f )
    354 {
    355    union fi fi;
    356    fi.f = f;
    357    return fi.ui;
    358 }
    359 
    360 static inline float
    361 uif(uint32_t ui)
    362 {
    363    union fi fi;
    364    fi.ui = ui;
    365    return fi.f;
    366 }
    367 
    368 
    369 /**
    370  * Convert ubyte to float in [0, 1].
    371  * XXX a 256-entry lookup table would be slightly faster.
    372  */
    373 static inline float
    374 ubyte_to_float(ubyte ub)
    375 {
    376    return (float) ub * (1.0f / 255.0f);
    377 }
    378 
    379 
    380 /**
    381  * Convert float in [0,1] to ubyte in [0,255] with clamping.
    382  */
    383 static inline ubyte
    384 float_to_ubyte(float f)
    385 {
    386    union fi tmp;
    387 
    388    tmp.f = f;
    389    if (tmp.i < 0) {
    390       return (ubyte) 0;
    391    }
    392    else if (tmp.i >= 0x3f800000 /* 1.0f */) {
    393       return (ubyte) 255;
    394    }
    395    else {
    396       tmp.f = tmp.f * (255.0f/256.0f) + 32768.0f;
    397       return (ubyte) tmp.i;
    398    }
    399 }
    400 
    401 static inline float
    402 byte_to_float_tex(int8_t b)
    403 {
    404    return (b == -128) ? -1.0F : b * 1.0F / 127.0F;
    405 }
    406 
    407 static inline int8_t
    408 float_to_byte_tex(float f)
    409 {
    410    return (int8_t) (127.0F * f);
    411 }
    412 
    413 /**
    414  * Calc log base 2
    415  */
    416 static inline unsigned
    417 util_logbase2(unsigned n)
    418 {
    419 #if defined(HAVE___BUILTIN_CLZ)
    420    return ((sizeof(unsigned) * 8 - 1) - __builtin_clz(n | 1));
    421 #else
    422    unsigned pos = 0;
    423    if (n >= 1<<16) { n >>= 16; pos += 16; }
    424    if (n >= 1<< 8) { n >>=  8; pos +=  8; }
    425    if (n >= 1<< 4) { n >>=  4; pos +=  4; }
    426    if (n >= 1<< 2) { n >>=  2; pos +=  2; }
    427    if (n >= 1<< 1) {           pos +=  1; }
    428    return pos;
    429 #endif
    430 }
    431 
    432 /**
    433  * Returns the ceiling of log n base 2, and 0 when n == 0. Equivalently,
    434  * returns the smallest x such that n <= 2**x.
    435  */
    436 static inline unsigned
    437 util_logbase2_ceil(unsigned n)
    438 {
    439    if (n <= 1)
    440       return 0;
    441 
    442    return 1 + util_logbase2(n - 1);
    443 }
    444 
    445 /**
    446  * Returns the smallest power of two >= x
    447  */
    448 static inline unsigned
    449 util_next_power_of_two(unsigned x)
    450 {
    451 #if defined(HAVE___BUILTIN_CLZ)
    452    if (x <= 1)
    453        return 1;
    454 
    455    return (1 << ((sizeof(unsigned) * 8) - __builtin_clz(x - 1)));
    456 #else
    457    unsigned val = x;
    458 
    459    if (x <= 1)
    460       return 1;
    461 
    462    if (util_is_power_of_two(x))
    463       return x;
    464 
    465    val--;
    466    val = (val >> 1) | val;
    467    val = (val >> 2) | val;
    468    val = (val >> 4) | val;
    469    val = (val >> 8) | val;
    470    val = (val >> 16) | val;
    471    val++;
    472    return val;
    473 #endif
    474 }
    475 
    476 
    477 /**
    478  * Return number of bits set in n.
    479  */
    480 static inline unsigned
    481 util_bitcount(unsigned n)
    482 {
    483 #if defined(HAVE___BUILTIN_POPCOUNT)
    484    return __builtin_popcount(n);
    485 #else
    486    /* K&R classic bitcount.
    487     *
    488     * For each iteration, clear the LSB from the bitfield.
    489     * Requires only one iteration per set bit, instead of
    490     * one iteration per bit less than highest set bit.
    491     */
    492    unsigned bits;
    493    for (bits = 0; n; bits++) {
    494       n &= n - 1;
    495    }
    496    return bits;
    497 #endif
    498 }
    499 
    500 
    501 static inline unsigned
    502 util_bitcount64(uint64_t n)
    503 {
    504 #ifdef HAVE___BUILTIN_POPCOUNTLL
    505    return __builtin_popcountll(n);
    506 #else
    507    return util_bitcount(n) + util_bitcount(n >> 32);
    508 #endif
    509 }
    510 
    511 
    512 /**
    513  * Reverse bits in n
    514  * Algorithm taken from:
    515  * http://stackoverflow.com/questions/9144800/c-reverse-bits-in-unsigned-integer
    516  */
    517 static inline unsigned
    518 util_bitreverse(unsigned n)
    519 {
    520     n = ((n >> 1) & 0x55555555u) | ((n & 0x55555555u) << 1);
    521     n = ((n >> 2) & 0x33333333u) | ((n & 0x33333333u) << 2);
    522     n = ((n >> 4) & 0x0f0f0f0fu) | ((n & 0x0f0f0f0fu) << 4);
    523     n = ((n >> 8) & 0x00ff00ffu) | ((n & 0x00ff00ffu) << 8);
    524     n = ((n >> 16) & 0xffffu) | ((n & 0xffffu) << 16);
    525     return n;
    526 }
    527 
    528 /**
    529  * Convert from little endian to CPU byte order.
    530  */
    531 
    532 #ifdef PIPE_ARCH_BIG_ENDIAN
    533 #define util_le64_to_cpu(x) util_bswap64(x)
    534 #define util_le32_to_cpu(x) util_bswap32(x)
    535 #define util_le16_to_cpu(x) util_bswap16(x)
    536 #else
    537 #define util_le64_to_cpu(x) (x)
    538 #define util_le32_to_cpu(x) (x)
    539 #define util_le16_to_cpu(x) (x)
    540 #endif
    541 
    542 #define util_cpu_to_le64(x) util_le64_to_cpu(x)
    543 #define util_cpu_to_le32(x) util_le32_to_cpu(x)
    544 #define util_cpu_to_le16(x) util_le16_to_cpu(x)
    545 
    546 /**
    547  * Reverse byte order of a 32 bit word.
    548  */
    549 static inline uint32_t
    550 util_bswap32(uint32_t n)
    551 {
    552 #if defined(HAVE___BUILTIN_BSWAP32)
    553    return __builtin_bswap32(n);
    554 #else
    555    return (n >> 24) |
    556           ((n >> 8) & 0x0000ff00) |
    557           ((n << 8) & 0x00ff0000) |
    558           (n << 24);
    559 #endif
    560 }
    561 
    562 /**
    563  * Reverse byte order of a 64bit word.
    564  */
    565 static inline uint64_t
    566 util_bswap64(uint64_t n)
    567 {
    568 #if defined(HAVE___BUILTIN_BSWAP64)
    569    return __builtin_bswap64(n);
    570 #else
    571    return ((uint64_t)util_bswap32((uint32_t)n) << 32) |
    572           util_bswap32((n >> 32));
    573 #endif
    574 }
    575 
    576 
    577 /**
    578  * Reverse byte order of a 16 bit word.
    579  */
    580 static inline uint16_t
    581 util_bswap16(uint16_t n)
    582 {
    583    return (n >> 8) |
    584           (n << 8);
    585 }
    586 
    587 static inline void*
    588 util_memcpy_cpu_to_le32(void * restrict dest, const void * restrict src, size_t n)
    589 {
    590 #ifdef PIPE_ARCH_BIG_ENDIAN
    591    size_t i, e;
    592    assert(n % 4 == 0);
    593 
    594    for (i = 0, e = n / 4; i < e; i++) {
    595       uint32_t * restrict d = (uint32_t* restrict)dest;
    596       const uint32_t * restrict s = (const uint32_t* restrict)src;
    597       d[i] = util_bswap32(s[i]);
    598    }
    599    return dest;
    600 #else
    601    return memcpy(dest, src, n);
    602 #endif
    603 }
    604 
    605 /**
    606  * Clamp X to [MIN, MAX].
    607  * This is a macro to allow float, int, uint, etc. types.
    608  */
    609 #define CLAMP( X, MIN, MAX )  ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) )
    610 
    611 #define MIN2( A, B )   ( (A)<(B) ? (A) : (B) )
    612 #define MAX2( A, B )   ( (A)>(B) ? (A) : (B) )
    613 
    614 #define MIN3( A, B, C ) ((A) < (B) ? MIN2(A, C) : MIN2(B, C))
    615 #define MAX3( A, B, C ) ((A) > (B) ? MAX2(A, C) : MAX2(B, C))
    616 
    617 #define MIN4( A, B, C, D ) ((A) < (B) ? MIN3(A, C, D) : MIN3(B, C, D))
    618 #define MAX4( A, B, C, D ) ((A) > (B) ? MAX3(A, C, D) : MAX3(B, C, D))
    619 
    620 
    621 /**
    622  * Align a value, only works pot alignemnts.
    623  */
    624 static inline int
    625 align(int value, int alignment)
    626 {
    627    return (value + alignment - 1) & ~(alignment - 1);
    628 }
    629 
    630 static inline uint64_t
    631 align64(uint64_t value, unsigned alignment)
    632 {
    633    return (value + alignment - 1) & ~((uint64_t)alignment - 1);
    634 }
    635 
    636 /**
    637  * Works like align but on npot alignments.
    638  */
    639 static inline size_t
    640 util_align_npot(size_t value, size_t alignment)
    641 {
    642    if (value % alignment)
    643       return value + (alignment - (value % alignment));
    644    return value;
    645 }
    646 
    647 static inline unsigned
    648 u_minify(unsigned value, unsigned levels)
    649 {
    650     return MAX2(1, value >> levels);
    651 }
    652 
    653 #ifndef COPY_4V
    654 #define COPY_4V( DST, SRC )         \
    655 do {                                \
    656    (DST)[0] = (SRC)[0];             \
    657    (DST)[1] = (SRC)[1];             \
    658    (DST)[2] = (SRC)[2];             \
    659    (DST)[3] = (SRC)[3];             \
    660 } while (0)
    661 #endif
    662 
    663 
    664 #ifndef COPY_4FV
    665 #define COPY_4FV( DST, SRC )  COPY_4V(DST, SRC)
    666 #endif
    667 
    668 
    669 #ifndef ASSIGN_4V
    670 #define ASSIGN_4V( DST, V0, V1, V2, V3 ) \
    671 do {                                     \
    672    (DST)[0] = (V0);                      \
    673    (DST)[1] = (V1);                      \
    674    (DST)[2] = (V2);                      \
    675    (DST)[3] = (V3);                      \
    676 } while (0)
    677 #endif
    678 
    679 
    680 static inline uint32_t
    681 util_unsigned_fixed(float value, unsigned frac_bits)
    682 {
    683    return value < 0 ? 0 : (uint32_t)(value * (1<<frac_bits));
    684 }
    685 
    686 static inline int32_t
    687 util_signed_fixed(float value, unsigned frac_bits)
    688 {
    689    return (int32_t)(value * (1<<frac_bits));
    690 }
    691 
    692 unsigned
    693 util_fpstate_get(void);
    694 unsigned
    695 util_fpstate_set_denorms_to_zero(unsigned current_fpstate);
    696 void
    697 util_fpstate_set(unsigned fpstate);
    698 
    699 
    700 
    701 #ifdef __cplusplus
    702 }
    703 #endif
    704 
    705 #endif /* U_MATH_H */
    706