src/util/rounding.h

   1 /*
   2  * Copyright © 2015 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  */
  23
  24 #ifndef _ROUNDING_H
  25 #define _ROUNDING_H
  26
  27 #include <math.h>
  28 #include <limits.h>
  29
  30 #ifdef __x86_64__
  31 #include <xmmintrin.h>
  32 #include <emmintrin.h>
  33 #endif
  34
  35 #ifdef __SSE4_1__
  36 #include <smmintrin.h>
  37 #endif
  38
  39 /* The C standard library has functions round()/rint()/nearbyint() that round
  40  * their arguments according to the rounding mode set in the floating-point
  41  * control register. While there are trunc()/ceil()/floor() functions that do
  42  * a specific operation without modifying the rounding mode, there is no
  43  * roundeven() in any version of C.
  44  *
  45  * Technical Specification 18661 (ISO/IEC TS 18661-1:2014) adds roundeven(),
  46  * but it's unfortunately not implemented by glibc.
  47  *
  48  * This implementation differs in that it does not raise the inexact exception.
  49  *
  50  * We use rint() to implement these functions, with the assumption that the
  51  * floating-point rounding mode has not been changed from the default Round
  52  * to Nearest.
  53  */
  54
  55 /**
  56  * \brief Rounds \c x to the nearest integer, with ties to the even integer.
  57  */
  58 static inline float
  59 _mesa_roundevenf(float x)
  60 {
  61 #ifdef __SSE4_1__
  62    float ret;
  63    __m128 m = _mm_load_ss(&x);
  64    m = _mm_round_ss(m, m, _MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC);
  65    _mm_store_ss(&ret, m);
  66    return ret;
  67 #else
  68    return rintf(x);
  69 #endif
  70 }
  71
  72 /**
  73  * \brief Rounds \c x to the nearest integer, with ties to the even integer.
  74  */
  75 static inline double
  76 _mesa_roundeven(double x)
  77 {
  78 #ifdef __SSE4_1__
  79    double ret;
  80    __m128d m = _mm_load_sd(&x);
  81    m = _mm_round_sd(m, m, _MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC);
  82    _mm_store_sd(&ret, m);
  83    return ret;
  84 #else
  85    return rint(x);
  86 #endif
  87 }
  88
  89 /**
  90  * \brief Rounds \c x to the nearest integer, with ties to the even integer,
  91  * and returns the value as a long int.
  92  */
  93 static inline long
  94 _mesa_lroundevenf(float x)
  95 {
  96 #ifdef __x86_64__
  97 #if LONG_BIT == 64
  98    return _mm_cvtss_si64(_mm_load_ss(&x));
  99 #elif LONG_BIT == 32 || defined(_WIN32)
 100    return _mm_cvtss_si32(_mm_load_ss(&x));
 101 #else
 102 #error "Unsupported or undefined LONG_BIT"
 103 #endif
 104 #else
 105    return lrintf(x);
 106 #endif
 107 }
 108
 109 /**
 110  * \brief Rounds \c x to the nearest integer, with ties to the even integer,
 111  * and returns the value as a long int.
 112  */
 113 static inline long
 114 _mesa_lroundeven(double x)
 115 {
 116 #ifdef __x86_64__
 117 #if LONG_BIT == 64
 118    return _mm_cvtsd_si64(_mm_load_sd(&x));
 119 #elif LONG_BIT == 32 || defined(_WIN32)
 120    return _mm_cvtsd_si32(_mm_load_sd(&x));
 121 #else
 122 #error "Unsupported or undefined LONG_BIT"
 123 #endif
 124 #else
 125    return lrint(x);
 126 #endif
 127 }
 128
 129 #endif