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 <limits.h> 28#include <stdint.h> 29#include <math.h> 30 31#if defined(__SSE__) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 1)) || (defined(_M_X64) && !defined(_M_ARM64EC)) 32#include <xmmintrin.h> 33#include <emmintrin.h> 34#endif 35 36#ifdef __SSE4_1__ 37#include <smmintrin.h> 38#endif 39 40/* The C standard library has functions round()/rint()/nearbyint() that round 41 * their arguments according to the rounding mode set in the floating-point 42 * control register. While there are trunc()/ceil()/floor() functions that do 43 * a specific operation without modifying the rounding mode, there is no 44 * roundeven() in any version of C. 45 * 46 * Technical Specification 18661 (ISO/IEC TS 18661-1:2014) adds roundeven(), 47 * but it's unfortunately not implemented by glibc. 48 * 49 * This implementation differs in that it does not raise the inexact exception. 50 * 51 * We use rint() to implement these functions, with the assumption that the 52 * floating-point rounding mode has not been changed from the default Round 53 * to Nearest. 54 */ 55 56/** 57 * \brief Rounds \c x to the nearest integer, with ties to the even integer. 58 */ 59static inline float 60_mesa_roundevenf(float x) 61{ 62#ifdef __SSE4_1__ 63 float ret; 64 __m128 m = _mm_load_ss(&x); 65 m = _mm_round_ss(m, m, _MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC); 66 _mm_store_ss(&ret, m); 67 return ret; 68#else 69 return rintf(x); 70#endif 71} 72 73/** 74 * \brief Rounds \c x to the nearest integer, with ties to the even integer. 75 */ 76static inline double 77_mesa_roundeven(double x) 78{ 79#ifdef __SSE4_1__ 80 double ret; 81 __m128d m = _mm_load_sd(&x); 82 m = _mm_round_sd(m, m, _MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC); 83 _mm_store_sd(&ret, m); 84 return ret; 85#else 86 return rint(x); 87#endif 88} 89 90/** 91 * \brief Rounds \c x to the nearest integer, with ties to the even integer, 92 * and returns the value as a long int. 93 */ 94static inline long 95_mesa_lroundevenf(float x) 96{ 97#if defined(__SSE__) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 1)) || (defined(_M_X64) && !defined(_M_ARM64EC)) 98#if LONG_MAX == INT64_MAX 99 return _mm_cvtss_si64(_mm_load_ss(&x)); 100#elif LONG_MAX == INT32_MAX 101 return _mm_cvtss_si32(_mm_load_ss(&x)); 102#else 103#error "Unsupported long size" 104#endif 105#else 106 return lrintf(x); 107#endif 108} 109 110/** 111 * \brief Rounds \c x to the nearest integer, with ties to the even integer, 112 * and returns the value as a long int. 113 */ 114static inline long 115_mesa_lroundeven(double x) 116{ 117#if defined(__SSE2__) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || (defined(_M_X64) && !defined(_M_ARM64EC)) 118#if LONG_MAX == INT64_MAX 119 return _mm_cvtsd_si64(_mm_load_sd(&x)); 120#elif LONG_MAX == INT32_MAX 121 return _mm_cvtsd_si32(_mm_load_sd(&x)); 122#else 123#error "Unsupported long size" 124#endif 125#else 126 return lrint(x); 127#endif 128} 129 130/** 131 * \brief Rounds \c x to the nearest integer, with ties to the even integer, 132 * and returns the value as an int64_t. 133 */ 134static inline int64_t 135_mesa_i64roundevenf(float x) 136{ 137#if LONG_MAX == INT64_MAX 138 return _mesa_lroundevenf(x); 139#elif LONG_MAX == INT32_MAX 140 return llrintf(x); 141#else 142#error "Unsupported long size" 143#endif 144} 145 146#endif 147