-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
This constitutes a fairly complete submission of an entirely new floating point type which conforms to IEEE 754 as a 16-bit storage class. Conversion between qfloat16 and float is currently performed through a sequence of lookup tables. Global-level functions qRound(), qRound64(), qFuzzyCompare(), qFuzzyIsNull(), and qIsNull() each with a qfloat16 parameter have been included for completeness. [ChangeLog][QtCore] Added new qfloat16 class. Change-Id: Ia52eb27846965c14f8140c00faf5ba33c9443976 Reviewed-by: Thiago Macieira <[email protected]>
- Loading branch information
Glen Mabey
authored and
Glen Mabey
committed
Jan 31, 2017
1 parent
e5d303c
commit 3ab7016
Showing
13 changed files
with
932 additions
and
4 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,117 @@ | ||
/**************************************************************************** | ||
** | ||
** Copyright (C) 2016 by Southwest Research Institute (R) | ||
** Contact: http://www.qt-project.org/legal | ||
** | ||
** This file is part of the QtCore module of the Qt Toolkit. | ||
** | ||
** $QT_BEGIN_LICENSE:LGPL$ | ||
** Commercial License Usage | ||
** Licensees holding valid commercial Qt licenses may use this file in | ||
** accordance with the commercial license agreement provided with the | ||
** Software or, alternatively, in accordance with the terms contained in | ||
** a written agreement between you and The Qt Company. For licensing terms | ||
** and conditions see https://www.qt.io/terms-conditions. For further | ||
** information use the contact form at https://www.qt.io/contact-us. | ||
** | ||
** GNU Lesser General Public License Usage | ||
** Alternatively, this file may be used under the terms of the GNU Lesser | ||
** General Public License version 3 as published by the Free Software | ||
** Foundation and appearing in the file LICENSE.LGPL3 included in the | ||
** packaging of this file. Please review the following information to | ||
** ensure the GNU Lesser General Public License version 3 requirements | ||
** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. | ||
** | ||
** GNU General Public License Usage | ||
** Alternatively, this file may be used under the terms of the GNU | ||
** General Public License version 2.0 or (at your option) the GNU General | ||
** Public license version 3 or any later version approved by the KDE Free | ||
** Qt Foundation. The licenses are as published by the Free Software | ||
** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 | ||
** included in the packaging of this file. Please review the following | ||
** information to ensure the GNU General Public License requirements will | ||
** be met: https://www.gnu.org/licenses/gpl-2.0.html and | ||
** https://www.gnu.org/licenses/gpl-3.0.html. | ||
** | ||
** $QT_END_LICENSE$ | ||
** | ||
****************************************************************************/ | ||
|
||
#include "qfloat16_p.h" | ||
|
||
QT_BEGIN_NAMESPACE | ||
|
||
/*! \headerfile <QFloat16> | ||
This header file provides support for half-precision (16-bit) floating | ||
point data with the class \c qfloat16. It is fully compliant with IEEE | ||
754 as a storage type. This implies that any arithmetic operation on a | ||
\c qfloat16 instance results in the value first being converted to a | ||
\c float. This conversion to and from \c float is performed by hardware | ||
when possible, but on processors that do not natively support half-precision, | ||
the conversion is performed through a sequence of lookup table operations. | ||
\c qfloat16 should be treated as if it were a POD (plain old data) type. | ||
Consequently, none of the supported operations need any elaboration beyond | ||
stating that it supports all arithmetic operators incident to floating point | ||
types. | ||
\since 5.9 | ||
*/ | ||
|
||
Q_STATIC_ASSERT_X(sizeof(float) == sizeof(quint32), | ||
"qfloat16 assumes that floats are 32 bits wide"); | ||
Q_STATIC_ASSERT_X(std::numeric_limits<float>::is_iec559, | ||
"Only works with IEEE 754 floating point"); | ||
|
||
/*! | ||
Returns true if the \c qfloat16 \a {f} is equivalent to infinity. | ||
\relates <QFloat16> | ||
\sa qIsInf | ||
*/ | ||
Q_REQUIRED_RESULT bool qIsInf(qfloat16 f) Q_DECL_NOTHROW { return qt_is_inf(f); } | ||
|
||
/*! | ||
Returns true if the \c qfloat16 \a {f} is not a number (NaN). | ||
\relates <QFloat16> | ||
\sa qIsNaN | ||
*/ | ||
Q_REQUIRED_RESULT bool qIsNaN(qfloat16 f) Q_DECL_NOTHROW { return qt_is_nan(f); } | ||
|
||
/*! | ||
Returns true if the \c qfloat16 \a {f} is a finite number. | ||
\relates <QFloat16> | ||
\sa qIsFinite | ||
*/ | ||
Q_REQUIRED_RESULT bool qIsFinite(qfloat16 f) Q_DECL_NOTHROW { return qt_is_finite(f); } | ||
|
||
/*! \fn int qRound(qfloat16 value) | ||
\relates <QFloat16> | ||
Rounds \a value to the nearest integer. | ||
\sa qRound | ||
*/ | ||
|
||
/*! \fn qint64 qRound64(qfloat16 value) | ||
\relates <QFloat16> | ||
Rounds \a value to the nearest 64-bit integer. | ||
\sa qRound64 | ||
*/ | ||
|
||
/*! \fn bool qFuzzyCompare(qfloat16 p1, qfloat16 p2) | ||
\relates <QFloat16> | ||
Compares the floating point value \a p1 and \a p2 and | ||
returns \c true if they are considered equal, otherwise \c false. | ||
The two numbers are compared in a relative way, where the | ||
exactness is stronger the smaller the numbers are. | ||
*/ | ||
|
||
QT_END_NAMESPACE |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,225 @@ | ||
/**************************************************************************** | ||
** | ||
** Copyright (C) 2016 by Southwest Research Institute (R) | ||
** Contact: http://www.qt-project.org/legal | ||
** | ||
** This file is part of the QtCore module of the Qt Toolkit. | ||
** | ||
** $QT_BEGIN_LICENSE:LGPL$ | ||
** Commercial License Usage | ||
** Licensees holding valid commercial Qt licenses may use this file in | ||
** accordance with the commercial license agreement provided with the | ||
** Software or, alternatively, in accordance with the terms contained in | ||
** a written agreement between you and The Qt Company. For licensing terms | ||
** and conditions see https://www.qt.io/terms-conditions. For further | ||
** information use the contact form at https://www.qt.io/contact-us. | ||
** | ||
** GNU Lesser General Public License Usage | ||
** Alternatively, this file may be used under the terms of the GNU Lesser | ||
** General Public License version 3 as published by the Free Software | ||
** Foundation and appearing in the file LICENSE.LGPL3 included in the | ||
** packaging of this file. Please review the following information to | ||
** ensure the GNU Lesser General Public License version 3 requirements | ||
** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. | ||
** | ||
** GNU General Public License Usage | ||
** Alternatively, this file may be used under the terms of the GNU | ||
** General Public License version 2.0 or (at your option) the GNU General | ||
** Public license version 3 or any later version approved by the KDE Free | ||
** Qt Foundation. The licenses are as published by the Free Software | ||
** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 | ||
** included in the packaging of this file. Please review the following | ||
** information to ensure the GNU General Public License requirements will | ||
** be met: https://www.gnu.org/licenses/gpl-2.0.html and | ||
** https://www.gnu.org/licenses/gpl-3.0.html. | ||
** | ||
** $QT_END_LICENSE$ | ||
** | ||
****************************************************************************/ | ||
|
||
#ifndef QFLOAT16_H | ||
#define QFLOAT16_H | ||
|
||
#include <QtCore/qglobal.h> | ||
#include <QtCore/qmetatype.h> | ||
#include <string.h> | ||
|
||
QT_BEGIN_NAMESPACE | ||
|
||
#if 0 | ||
#pragma qt_class(QFloat16) | ||
#endif | ||
|
||
class qfloat16 | ||
{ | ||
public: | ||
#ifndef Q_QDOC | ||
Q_DECL_CONSTEXPR inline qfloat16() Q_DECL_NOTHROW : b16(0) { } | ||
inline qfloat16(float f) Q_DECL_NOTHROW; | ||
inline operator float() const Q_DECL_NOTHROW; | ||
inline operator double() const Q_DECL_NOTHROW; | ||
inline operator long double() const Q_DECL_NOTHROW; | ||
#endif | ||
|
||
private: | ||
quint16 b16; | ||
|
||
Q_CORE_EXPORT static const quint32 mantissatable[]; | ||
Q_CORE_EXPORT static const quint32 exponenttable[]; | ||
Q_CORE_EXPORT static const quint32 offsettable[]; | ||
Q_CORE_EXPORT static const quint32 basetable[]; | ||
Q_CORE_EXPORT static const quint32 shifttable[]; | ||
|
||
friend bool qIsNull(qfloat16 f) Q_DECL_NOTHROW; | ||
friend qfloat16 operator-(qfloat16 a) Q_DECL_NOTHROW; | ||
}; | ||
|
||
Q_DECLARE_TYPEINFO(qfloat16, Q_PRIMITIVE_TYPE); | ||
|
||
Q_CORE_EXPORT Q_REQUIRED_RESULT bool qIsInf(qfloat16 f) Q_DECL_NOTHROW; // complements qnumeric.h | ||
Q_CORE_EXPORT Q_REQUIRED_RESULT bool qIsNaN(qfloat16 f) Q_DECL_NOTHROW; // complements qnumeric.h | ||
Q_CORE_EXPORT Q_REQUIRED_RESULT bool qIsFinite(qfloat16 f) Q_DECL_NOTHROW; // complements qnumeric.h | ||
|
||
// The remainder of these utility functions complement qglobal.h | ||
inline Q_REQUIRED_RESULT int qRound(qfloat16 d) Q_DECL_NOTHROW | ||
{ return qRound(static_cast<float>(d)); } | ||
|
||
inline Q_REQUIRED_RESULT qint64 qRound64(qfloat16 d) Q_DECL_NOTHROW | ||
{ return qRound64(static_cast<float>(d)); } | ||
|
||
inline Q_REQUIRED_RESULT bool qFuzzyCompare(qfloat16 p1, qfloat16 p2) Q_DECL_NOTHROW | ||
{ | ||
float f1 = static_cast<float>(p1); | ||
float f2 = static_cast<float>(p2); | ||
// The significand precision for IEEE754 half precision is | ||
// 11 bits (10 explicitly stored), or approximately 3 decimal | ||
// digits. In selecting the fuzzy comparison factor of 102.5f | ||
// (that is, (2^10+1)/10) below, we effectively select a | ||
// window of about 1 (least significant) decimal digit about | ||
// which the two operands can vary and still return true. | ||
return (qAbs(f1 - f2) * 102.5f <= qMin(qAbs(f1), qAbs(f2))); | ||
} | ||
|
||
inline Q_REQUIRED_RESULT bool qIsNull(qfloat16 f) Q_DECL_NOTHROW | ||
{ | ||
return (f.b16 & static_cast<quint16>(0x7fff)) == 0; | ||
} | ||
|
||
inline int qIntCast(qfloat16 f) Q_DECL_NOTHROW | ||
{ return int(static_cast<float>(f)); } | ||
|
||
inline qfloat16::qfloat16(float f) Q_DECL_NOTHROW | ||
{ | ||
quint32 u; | ||
memcpy(&u, &f, sizeof(quint32)); | ||
b16 = basetable[(u >> 23) & 0x1ff] | ||
+ ((u & 0x007fffff) >> shifttable[(u >> 23) & 0x1ff]); | ||
} | ||
|
||
inline qfloat16::operator float() const Q_DECL_NOTHROW | ||
{ | ||
quint32 u = mantissatable[offsettable[b16 >> 10] + (b16 & 0x3ff)] | ||
+ exponenttable[b16 >> 10]; | ||
float f; | ||
memcpy(&f, &u, sizeof(quint32)); | ||
return f; | ||
} | ||
|
||
inline qfloat16::operator double() const Q_DECL_NOTHROW | ||
{ | ||
return static_cast<double>(float(*this)); | ||
} | ||
|
||
inline qfloat16::operator long double() const Q_DECL_NOTHROW | ||
{ | ||
return static_cast<long double>(float(*this)); | ||
} | ||
|
||
inline qfloat16 operator-(qfloat16 a) Q_DECL_NOTHROW | ||
{ | ||
qfloat16 f; | ||
f.b16 = a.b16 ^ quint16(0x8000); | ||
return f; | ||
} | ||
|
||
inline qfloat16 operator+(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return qfloat16(static_cast<float>(a) + static_cast<float>(b)); } | ||
inline qfloat16 operator-(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return qfloat16(static_cast<float>(a) - static_cast<float>(b)); } | ||
inline qfloat16 operator*(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return qfloat16(static_cast<float>(a) * static_cast<float>(b)); } | ||
inline qfloat16 operator/(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return qfloat16(static_cast<float>(a) / static_cast<float>(b)); } | ||
|
||
#define QF16_MAKE_ARITH_OP_FP(FP, OP) \ | ||
inline FP operator OP(qfloat16 lhs, FP rhs) Q_DECL_NOTHROW { return static_cast<FP>(lhs) OP rhs; } \ | ||
inline FP operator OP(FP lhs, qfloat16 rhs) Q_DECL_NOTHROW { return lhs OP static_cast<FP>(rhs); } | ||
#define QF16_MAKE_ARITH_OP_EQ_FP(FP, OP_EQ, OP) \ | ||
inline qfloat16& operator OP_EQ(qfloat16& lhs, FP rhs) Q_DECL_NOTHROW { lhs = qfloat16(static_cast<FP>(lhs) OP rhs); return lhs; } | ||
#define QF16_MAKE_ARITH_OP(FP) \ | ||
QF16_MAKE_ARITH_OP_FP(FP, +) \ | ||
QF16_MAKE_ARITH_OP_FP(FP, -) \ | ||
QF16_MAKE_ARITH_OP_FP(FP, *) \ | ||
QF16_MAKE_ARITH_OP_FP(FP, /) \ | ||
QF16_MAKE_ARITH_OP_EQ_FP(FP, +=, +) \ | ||
QF16_MAKE_ARITH_OP_EQ_FP(FP, -=, -) \ | ||
QF16_MAKE_ARITH_OP_EQ_FP(FP, *=, *) \ | ||
QF16_MAKE_ARITH_OP_EQ_FP(FP, /=, /) | ||
QF16_MAKE_ARITH_OP(long double) | ||
QF16_MAKE_ARITH_OP(double) | ||
QF16_MAKE_ARITH_OP(float) | ||
#undef QF16_MAKE_ARITH_OP | ||
#undef QF16_MAKE_ARITH_OP_FP | ||
|
||
#define QF16_MAKE_ARITH_OP_INT(OP) \ | ||
inline double operator OP(qfloat16 lhs, int rhs) Q_DECL_NOTHROW { return static_cast<double>(lhs) OP rhs; } \ | ||
inline double operator OP(int lhs, qfloat16 rhs) Q_DECL_NOTHROW { return lhs OP static_cast<double>(rhs); } | ||
QF16_MAKE_ARITH_OP_INT(+) | ||
QF16_MAKE_ARITH_OP_INT(-) | ||
QF16_MAKE_ARITH_OP_INT(*) | ||
QF16_MAKE_ARITH_OP_INT(/) | ||
#undef QF16_MAKE_ARITH_OP_INT | ||
|
||
inline bool operator>(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return static_cast<float>(a) > static_cast<float>(b); } | ||
inline bool operator<(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return static_cast<float>(a) < static_cast<float>(b); } | ||
inline bool operator>=(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return static_cast<float>(a) >= static_cast<float>(b); } | ||
inline bool operator<=(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return static_cast<float>(a) <= static_cast<float>(b); } | ||
inline bool operator==(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return static_cast<float>(a) == static_cast<float>(b); } | ||
inline bool operator!=(qfloat16 a, qfloat16 b) Q_DECL_NOTHROW { return static_cast<float>(a) != static_cast<float>(b); } | ||
|
||
#define QF16_MAKE_BOOL_OP_FP(FP, OP) \ | ||
inline bool operator OP(qfloat16 lhs, FP rhs) Q_DECL_NOTHROW { return static_cast<FP>(lhs) OP rhs; } \ | ||
inline bool operator OP(FP lhs, qfloat16 rhs) Q_DECL_NOTHROW { return lhs OP static_cast<FP>(rhs); } | ||
#define QF16_MAKE_BOOL_OP(FP) \ | ||
QF16_MAKE_BOOL_OP_FP(FP, <) \ | ||
QF16_MAKE_BOOL_OP_FP(FP, >) \ | ||
QF16_MAKE_BOOL_OP_FP(FP, >=) \ | ||
QF16_MAKE_BOOL_OP_FP(FP, <=) \ | ||
QF16_MAKE_BOOL_OP_FP(FP, ==) \ | ||
QF16_MAKE_BOOL_OP_FP(FP, !=) | ||
QF16_MAKE_BOOL_OP(long double) | ||
QF16_MAKE_BOOL_OP(double) | ||
QF16_MAKE_BOOL_OP(float) | ||
#undef QF16_MAKE_BOOL_OP | ||
#undef QF16_MAKE_BOOL_OP_FP | ||
|
||
#define QF16_MAKE_BOOL_OP_INT(OP) \ | ||
inline bool operator OP(qfloat16 a, int b) Q_DECL_NOTHROW { return static_cast<float>(a) OP b; } \ | ||
inline bool operator OP(int a, qfloat16 b) Q_DECL_NOTHROW { return a OP static_cast<float>(b); } | ||
QF16_MAKE_BOOL_OP_INT(>) | ||
QF16_MAKE_BOOL_OP_INT(<) | ||
QF16_MAKE_BOOL_OP_INT(>=) | ||
QF16_MAKE_BOOL_OP_INT(<=) | ||
QF16_MAKE_BOOL_OP_INT(==) | ||
QF16_MAKE_BOOL_OP_INT(!=) | ||
#undef QF16_MAKE_BOOL_OP_INT | ||
|
||
/*! | ||
\internal | ||
*/ | ||
inline Q_REQUIRED_RESULT bool qFuzzyIsNull(qfloat16 f) Q_DECL_NOTHROW | ||
{ | ||
return qAbs(static_cast<float>(f)) <= 0.001f; | ||
} | ||
|
||
QT_END_NAMESPACE | ||
|
||
Q_DECLARE_METATYPE(qfloat16) | ||
|
||
#endif // QFLOAT16_H |
Oops, something went wrong.