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 "c99_math.h" 28 29#include <limits.h> 30#include <stdint.h> 31 32#if defined(__SSE__) || defined(_MSC_VER) 33/* MSVC always has SSE nowadays */ 34#include <xmmintrin.h> 35#include <emmintrin.h> 36#endif 37 38#ifdef __SSE4_1__ 39#include <smmintrin.h> 40#endif 41 42/* The C standard library has functions round()/rint()/nearbyint() that round 43 * their arguments according to the rounding mode set in the floating-point 44 * control register. While there are trunc()/ceil()/floor() functions that do 45 * a specific operation without modifying the rounding mode, there is no 46 * roundeven() in any version of C. 47 * 48 * Technical Specification 18661 (ISO/IEC TS 18661-1:2014) adds roundeven(), 49 * but it's unfortunately not implemented by glibc. 50 * 51 * This implementation differs in that it does not raise the inexact exception. 52 * 53 * We use rint() to implement these functions, with the assumption that the 54 * floating-point rounding mode has not been changed from the default Round 55 * to Nearest. 56 */ 57 58/** 59 * \brief Rounds \c x to the nearest integer, with ties to the even integer. 60 */ 61static inline float 62_mesa_roundevenf(float x) 63{ 64#ifdef __SSE4_1__ 65 float ret; 66 __m128 m = _mm_load_ss(&x); 67 m = _mm_round_ss(m, m, _MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC); 68 _mm_store_ss(&ret, m); 69 return ret; 70#else 71 return rintf(x); 72#endif 73} 74 75/** 76 * \brief Rounds \c x to the nearest integer, with ties to the even integer. 77 */ 78static inline double 79_mesa_roundeven(double x) 80{ 81#ifdef __SSE4_1__ 82 double ret; 83 __m128d m = _mm_load_sd(&x); 84 m = _mm_round_sd(m, m, _MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC); 85 _mm_store_sd(&ret, m); 86 return ret; 87#else 88 return rint(x); 89#endif 90} 91 92/** 93 * \brief Rounds \c x to the nearest integer, with ties to the even integer, 94 * and returns the value as a long int. 95 */ 96static inline long 97_mesa_lroundevenf(float x) 98{ 99#if defined(__SSE__) || defined(_MSC_VER) 100#if LONG_MAX == INT64_MAX 101 return _mm_cvtss_si64(_mm_load_ss(&x)); 102#elif LONG_MAX == INT32_MAX 103 return _mm_cvtss_si32(_mm_load_ss(&x)); 104#else 105#error "Unsupported long size" 106#endif 107#else 108 return lrintf(x); 109#endif 110} 111 112/** 113 * \brief Rounds \c x to the nearest integer, with ties to the even integer, 114 * and returns the value as a long int. 115 */ 116static inline long 117_mesa_lroundeven(double x) 118{ 119#if defined(__SSE2__) || defined(_MSC_VER) 120#if LONG_MAX == INT64_MAX 121 return _mm_cvtsd_si64(_mm_load_sd(&x)); 122#elif LONG_MAX == INT32_MAX 123 return _mm_cvtsd_si32(_mm_load_sd(&x)); 124#else 125#error "Unsupported long size" 126#endif 127#else 128 return lrint(x); 129#endif 130} 131 132/** 133 * \brief Rounds \c x to the nearest integer, with ties to the even integer, 134 * and returns the value as an int64_t. 135 */ 136static inline int64_t 137_mesa_i64roundevenf(float x) 138{ 139#if LONG_MAX == INT64_MAX 140 return _mesa_lroundevenf(x); 141#elif LONG_MAX == INT32_MAX 142 return llrintf(x); 143#else 144#error "Unsupported long size" 145#endif 146} 147 148#endif 149