half.h 17.6 KB
Newer Older
wester committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
///////////////////////////////////////////////////////////////////////////
//
// Copyright (c) 2002, Industrial Light & Magic, a division of Lucas
// Digital Ltd. LLC
//
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
// *       Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// *       Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// *       Neither the name of Industrial Light & Magic nor the names of
// its contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
///////////////////////////////////////////////////////////////////////////

// Primary authors:
//     Florian Kainz <kainz@ilm.com>
//     Rod Bogart <rgb@ilm.com>

//---------------------------------------------------------------------------
//
//	half -- a 16-bit floating point number class:
//
//	Type half can represent positive and negative numbers whose
//	magnitude is between roughly 6.1e-5 and 6.5e+4 with a relative
//	error of 9.8e-4; numbers smaller than 6.1e-5 can be represented
//	with an absolute error of 6.0e-8.  All integers from -2048 to
//	+2048 can be represented exactly.
//
//	Type half behaves (almost) like the built-in C++ floating point
//	types.  In arithmetic expressions, half, float and double can be
//	mixed freely.  Here are a few examples:
//
//	    half a (3.5);
//	    float b (a + sqrt (a));
//	    a += b;
//	    b += a;
//	    b = a + 7;
//
//	Conversions from half to float are lossless; all half numbers
//	are exactly representable as floats.
//
//	Conversions from float to half may not preserve a float's value
//	exactly.  If a float is not representable as a half, then the
//	float value is rounded to the nearest representable half.  If a
//	float value is exactly in the middle between the two closest
//	representable half values, then the float value is rounded to
//	the closest half whose least significant bit is zero.
//
//	Overflows during float-to-half conversions cause arithmetic
//	exceptions.  An overflow occurs when the float value to be
//	converted is too large to be represented as a half, or if the
//	float value is an infinity or a NAN.
//
//	The implementation of type half makes the following assumptions
//	about the implementation of the built-in C++ types:
//
//	    float is an IEEE 754 single-precision number
//	    sizeof (float) == 4
//	    sizeof (unsigned int) == sizeof (float)
//	    alignof (unsigned int) == alignof (float)
//	    sizeof (unsigned short) == 2
//
//---------------------------------------------------------------------------

#ifndef _HALF_H_
#define _HALF_H_

#include <iostream>

#if defined(OPENEXR_DLL)
    #if defined(HALF_EXPORTS)
    #define HALF_EXPORT __declspec(dllexport)
    #else
    #define HALF_EXPORT __declspec(dllimport)
    #endif
    #define HALF_EXPORT_CONST
#else
    #define HALF_EXPORT
    #define HALF_EXPORT_CONST const
#endif

class HALF_EXPORT half
{
  public:

    //-------------
    // Constructors
    //-------------

    half ();			// no initialization
    half (float f);


    //--------------------
    // Conversion to float
    //--------------------

    operator		float () const;


    //------------
    // Unary minus
    //------------

    half		operator - () const;


    //-----------
    // Assignment
    //-----------

    half &		operator = (half  h);
    half &		operator = (float f);

    half &		operator += (half  h);
    half &		operator += (float f);

    half &		operator -= (half  h);
    half &		operator -= (float f);

    half &		operator *= (half  h);
    half &		operator *= (float f);

    half &		operator /= (half  h);
    half &		operator /= (float f);


    //---------------------------------------------------------
    // Round to n-bit precision (n should be between 0 and 10).
    // After rounding, the significand's 10-n least significant
    // bits will be zero.
    //---------------------------------------------------------

    half		round (unsigned int n) const;


    //--------------------------------------------------------------------
    // Classification:
    //
    //	h.isFinite()		returns true if h is a normalized number,
    //				a denormalized number or zero
    //
    //	h.isNormalized()	returns true if h is a normalized number
    //
    //	h.isDenormalized()	returns true if h is a denormalized number
    //
    //	h.isZero()		returns true if h is zero
    //
    //	h.isNan()		returns true if h is a NAN
    //
    //	h.isInfinity()		returns true if h is a positive
    //				or a negative infinity
    //
    //	h.isNegative()		returns true if the sign bit of h
    //				is set (negative)
    //--------------------------------------------------------------------

    bool		isFinite () const;
    bool		isNormalized () const;
    bool		isDenormalized () const;
    bool		isZero () const;
    bool		isNan () const;
    bool		isInfinity () const;
    bool		isNegative () const;


    //--------------------------------------------
    // Special values
    //
    //	posInf()	returns +infinity
    //
    //	negInf()	returns -infinity
    //
    //	qNan()		returns a NAN with the bit
    //			pattern 0111111111111111
    //
    //	sNan()		returns a NAN with the bit
    //			pattern 0111110111111111
    //--------------------------------------------

    static half		posInf ();
    static half		negInf ();
    static half		qNan ();
    static half		sNan ();


    //--------------------------------------
    // Access to the internal representation
    //--------------------------------------

    unsigned short	bits () const;
    void		setBits (unsigned short bits);


  public:

    union uif
    {
    unsigned int	i;
    float		f;
    };

  private:

    static short	convert (int i);
    static float	overflow ();

    unsigned short	_h;

    static HALF_EXPORT_CONST uif		_toFloat[1 << 16];
    static HALF_EXPORT_CONST unsigned short _eLut[1 << 9];
};

//-----------
// Stream I/O
//-----------

HALF_EXPORT std::ostream &		operator << (std::ostream &os, half  h);
HALF_EXPORT std::istream &		operator >> (std::istream &is, half &h);


//----------
// Debugging
//----------

HALF_EXPORT void			printBits   (std::ostream &os, half  h);
HALF_EXPORT void			printBits   (std::ostream &os, float f);
HALF_EXPORT void			printBits   (char  c[19], half  h);
HALF_EXPORT void			printBits   (char  c[35], float f);


//-------------------------------------------------------------------------
// Limits
//
// Visual C++ will complain if HALF_MIN, HALF_NRM_MIN etc. are not float
// constants, but at least one other compiler (gcc 2.96) produces incorrect
// results if they are.
//-------------------------------------------------------------------------

#if (defined _WIN32 || defined _WIN64) && defined _MSC_VER

  #define HALF_MIN	5.96046448e-08f	// Smallest positive half

  #define HALF_NRM_MIN	6.10351562e-05f	// Smallest positive normalized half

  #define HALF_MAX	65504.0f	// Largest positive half

  #define HALF_EPSILON	0.00097656f	// Smallest positive e for which
                    // half (1.0 + e) != half (1.0)
#else

  #define HALF_MIN	5.96046448e-08	// Smallest positive half

  #define HALF_NRM_MIN	6.10351562e-05	// Smallest positive normalized half

  #define HALF_MAX	65504.0		// Largest positive half

  #define HALF_EPSILON	0.00097656	// Smallest positive e for which
                    // half (1.0 + e) != half (1.0)
#endif


#define HALF_MANT_DIG	11		// Number of digits in mantissa
                    // (significand + hidden leading 1)

#define HALF_DIG	2		// Number of base 10 digits that
                    // can be represented without change

#define HALF_RADIX	2		// Base of the exponent

#define HALF_MIN_EXP	-13		// Minimum negative integer such that
                    // HALF_RADIX raised to the power of
                    // one less than that integer is a
                    // normalized half

#define HALF_MAX_EXP	16		// Maximum positive integer such that
                    // HALF_RADIX raised to the power of
                    // one less than that integer is a
                    // normalized half

#define HALF_MIN_10_EXP	-4		// Minimum positive integer such
                    // that 10 raised to that power is
                    // a normalized half

#define HALF_MAX_10_EXP	4		// Maximum positive integer such
                    // that 10 raised to that power is
                    // a normalized half


//---------------------------------------------------------------------------
//
// Implementation --
//
// Representation of a float:
//
//	We assume that a float, f, is an IEEE 754 single-precision
//	floating point number, whose bits are arranged as follows:
//
//	    31 (msb)
//	    |
//	    | 30     23
//	    | |      |
//	    | |      | 22                    0 (lsb)
//	    | |      | |                     |
//	    X XXXXXXXX XXXXXXXXXXXXXXXXXXXXXXX
//
//	    s e        m
//
//	S is the sign-bit, e is the exponent and m is the significand.
//
//	If e is between 1 and 254, f is a normalized number:
//
//	            s    e-127
//	    f = (-1)  * 2      * 1.m
//
//	If e is 0, and m is not zero, f is a denormalized number:
//
//	            s    -126
//	    f = (-1)  * 2      * 0.m
//
//	If e and m are both zero, f is zero:
//
//	    f = 0.0
//
//	If e is 255, f is an "infinity" or "not a number" (NAN),
//	depending on whether m is zero or not.
//
//	Examples:
//
//	    0 00000000 00000000000000000000000 = 0.0
//	    0 01111110 00000000000000000000000 = 0.5
//	    0 01111111 00000000000000000000000 = 1.0
//	    0 10000000 00000000000000000000000 = 2.0
//	    0 10000000 10000000000000000000000 = 3.0
//	    1 10000101 11110000010000000000000 = -124.0625
//	    0 11111111 00000000000000000000000 = +infinity
//	    1 11111111 00000000000000000000000 = -infinity
//	    0 11111111 10000000000000000000000 = NAN
//	    1 11111111 11111111111111111111111 = NAN
//
// Representation of a half:
//
//	Here is the bit-layout for a half number, h:
//
//	    15 (msb)
//	    |
//	    | 14  10
//	    | |   |
//	    | |   | 9        0 (lsb)
//	    | |   | |        |
//	    X XXXXX XXXXXXXXXX
//
//	    s e     m
//
//	S is the sign-bit, e is the exponent and m is the significand.
//
//	If e is between 1 and 30, h is a normalized number:
//
//	            s    e-15
//	    h = (-1)  * 2     * 1.m
//
//	If e is 0, and m is not zero, h is a denormalized number:
//
//	            S    -14
//	    h = (-1)  * 2     * 0.m
//
//	If e and m are both zero, h is zero:
//
//	    h = 0.0
//
//	If e is 31, h is an "infinity" or "not a number" (NAN),
//	depending on whether m is zero or not.
//
//	Examples:
//
//	    0 00000 0000000000 = 0.0
//	    0 01110 0000000000 = 0.5
//	    0 01111 0000000000 = 1.0
//	    0 10000 0000000000 = 2.0
//	    0 10000 1000000000 = 3.0
//	    1 10101 1111000001 = -124.0625
//	    0 11111 0000000000 = +infinity
//	    1 11111 0000000000 = -infinity
//	    0 11111 1000000000 = NAN
//	    1 11111 1111111111 = NAN
//
// Conversion:
//
//	Converting from a float to a half requires some non-trivial bit
//	manipulations.  In some cases, this makes conversion relatively
//	slow, but the most common case is accelerated via table lookups.
//
//	Converting back from a half to a float is easier because we don't
//	have to do any rounding.  In addition, there are only 65536
//	different half numbers; we can convert each of those numbers once
//	and store the results in a table.  Later, all conversions can be
//	done using only simple table lookups.
//
//---------------------------------------------------------------------------


//--------------------
// Simple constructors
//--------------------

inline
half::half ()
{
    // no initialization
}


//----------------------------
// Half-from-float constructor
//----------------------------

inline
half::half (float f)
{
    uif x;

    x.f = f;

    if (f == 0)
    {
    //
    // Common special case - zero.
    // Preserve the zero's sign bit.
    //

    _h = (x.i >> 16);
    }
    else
    {
    //
    // We extract the combined sign and exponent, e, from our
    // floating-point number, f.  Then we convert e to the sign
    // and exponent of the half number via a table lookup.
    //
    // For the most common case, where a normalized half is produced,
    // the table lookup returns a non-zero value; in this case, all
    // we have to do is round f's significand to 10 bits and combine
    // the result with e.
    //
    // For all other cases (overflow, zeroes, denormalized numbers
    // resulting from underflow, infinities and NANs), the table
    // lookup returns zero, and we call a longer, non-inline function
    // to do the float-to-half conversion.
    //

    register int e = (x.i >> 23) & 0x000001ff;

    e = _eLut[e];

    if (e)
    {
        //
        // Simple case - round the significand, m, to 10
        // bits and combine it with the sign and exponent.
        //

        register int m = x.i & 0x007fffff;
        _h = e + ((m + 0x00000fff + ((m >> 13) & 1)) >> 13);
    }
    else
    {
        //
        // Difficult case - call a function.
        //

        _h = convert (x.i);
    }
    }
}


//------------------------------------------
// Half-to-float conversion via table lookup
//------------------------------------------

inline
half::operator float () const
{
    return _toFloat[_h].f;
}


//-------------------------
// Round to n-bit precision
//-------------------------

inline half
half::round (unsigned int n) const
{
    //
    // Parameter check.
    //

    if (n >= 10)
    return *this;

    //
    // Disassemble h into the sign, s,
    // and the combined exponent and significand, e.
    //

    unsigned short s = _h & 0x8000;
    unsigned short e = _h & 0x7fff;

    //
    // Round the exponent and significand to the nearest value
    // where ones occur only in the (10-n) most significant bits.
    // Note that the exponent adjusts automatically if rounding
    // up causes the significand to overflow.
    //

    e >>= 9 - n;
    e  += e & 1;
    e <<= 9 - n;

    //
    // Check for exponent overflow.
    //

    if (e >= 0x7c00)
    {
    //
    // Overflow occurred -- truncate instead of rounding.
    //

    e = _h;
    e >>= 10 - n;
    e <<= 10 - n;
    }

    //
    // Put the original sign bit back.
    //

    half h;
    h._h = s | e;

    return h;
}


//-----------------------
// Other inline functions
//-----------------------

inline half
half::operator - () const
{
    half h;
    h._h = _h ^ 0x8000;
    return h;
}


inline half &
half::operator = (half h)
{
    _h = h._h;
    return *this;
}


inline half &
half::operator = (float f)
{
    *this = half (f);
    return *this;
}


inline half &
half::operator += (half h)
{
    *this = half (float (*this) + float (h));
    return *this;
}


inline half &
half::operator += (float f)
{
    *this = half (float (*this) + f);
    return *this;
}


inline half &
half::operator -= (half h)
{
    *this = half (float (*this) - float (h));
    return *this;
}


inline half &
half::operator -= (float f)
{
    *this = half (float (*this) - f);
    return *this;
}


inline half &
half::operator *= (half h)
{
    *this = half (float (*this) * float (h));
    return *this;
}


inline half &
half::operator *= (float f)
{
    *this = half (float (*this) * f);
    return *this;
}


inline half &
half::operator /= (half h)
{
    *this = half (float (*this) / float (h));
    return *this;
}


inline half &
half::operator /= (float f)
{
    *this = half (float (*this) / f);
    return *this;
}


inline bool
half::isFinite () const
{
    unsigned short e = (_h >> 10) & 0x001f;
    return e < 31;
}


inline bool
half::isNormalized () const
{
    unsigned short e = (_h >> 10) & 0x001f;
    return e > 0 && e < 31;
}


inline bool
half::isDenormalized () const
{
    unsigned short e = (_h >> 10) & 0x001f;
    unsigned short m =  _h & 0x3ff;
    return e == 0 && m != 0;
}


inline bool
half::isZero () const
{
    return (_h & 0x7fff) == 0;
}


inline bool
half::isNan () const
{
    unsigned short e = (_h >> 10) & 0x001f;
    unsigned short m =  _h & 0x3ff;
    return e == 31 && m != 0;
}


inline bool
half::isInfinity () const
{
    unsigned short e = (_h >> 10) & 0x001f;
    unsigned short m =  _h & 0x3ff;
    return e == 31 && m == 0;
}


inline bool
half::isNegative () const
{
    return (_h & 0x8000) != 0;
}


inline half
half::posInf ()
{
    half h;
    h._h = 0x7c00;
    return h;
}


inline half
half::negInf ()
{
    half h;
    h._h = 0xfc00;
    return h;
}


inline half
half::qNan ()
{
    half h;
    h._h = 0x7fff;
    return h;
}


inline half
half::sNan ()
{
    half h;
    h._h = 0x7dff;
    return h;
}


inline unsigned short
half::bits () const
{
    return _h;
}


inline void
half::setBits (unsigned short bits)
{
    _h = bits;
}

#endif