qt5base-lts/tests/auto/gui/math3d/qmatrixnxn/tst_qmatrixnxn.cpp

3350 lines
97 KiB
C++
Raw Normal View History

/****************************************************************************
**
** Copyright (C) 2016 The Qt Company Ltd.
** Contact: https://www.qt.io/licensing/
**
** This file is part of the test suite of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:GPL-EXCEPT$
** 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 General Public License Usage
** Alternatively, this file may be used under the terms of the GNU
** General Public License version 3 as published by the Free Software
** Foundation with exceptions as appearing in the file LICENSE.GPL3-EXCEPT
** 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-3.0.html.
**
** $QT_END_LICENSE$
**
****************************************************************************/
#include <QTest>
#include <QtCore/qmath.h>
#include <QtGui/qmatrix4x4.h>
class tst_QMatrixNxN : public QObject
{
Q_OBJECT
public:
tst_QMatrixNxN() {}
~tst_QMatrixNxN() {}
private slots:
void create2x2();
void create3x3();
void create4x4();
void create4x3();
void isIdentity2x2();
void isIdentity3x3();
void isIdentity4x4();
void isIdentity4x3();
void compare2x2();
void compare3x3();
void compare4x4();
void compare4x3();
void transposed2x2();
void transposed3x3();
void transposed4x4();
void transposed4x3();
void add2x2_data();
void add2x2();
void add3x3_data();
void add3x3();
void add4x4_data();
void add4x4();
void add4x3_data();
void add4x3();
void subtract2x2_data();
void subtract2x2();
void subtract3x3_data();
void subtract3x3();
void subtract4x4_data();
void subtract4x4();
void subtract4x3_data();
void subtract4x3();
void multiply2x2_data();
void multiply2x2();
void multiply3x3_data();
void multiply3x3();
void multiply4x4_data();
void multiply4x4();
void multiply4x3_data();
void multiply4x3();
void multiplyFactor2x2_data();
void multiplyFactor2x2();
void multiplyFactor3x3_data();
void multiplyFactor3x3();
void multiplyFactor4x4_data();
void multiplyFactor4x4();
void multiplyFactor4x3_data();
void multiplyFactor4x3();
void divideFactor2x2_data();
void divideFactor2x2();
void divideFactor3x3_data();
void divideFactor3x3();
void divideFactor4x4_data();
void divideFactor4x4();
void divideFactor4x3_data();
void divideFactor4x3();
void negate2x2_data();
void negate2x2();
void negate3x3_data();
void negate3x3();
void negate4x4_data();
void negate4x4();
void negate4x3_data();
void negate4x3();
void inverted4x4_data();
void inverted4x4();
void orthonormalInverse4x4();
void scale4x4_data();
void scale4x4();
void translate4x4_data();
void translate4x4();
void rotate4x4_data();
void rotate4x4();
void normalMatrix_data();
void normalMatrix();
void optimizedTransforms();
void ortho();
void frustum();
void perspective();
void viewport();
void flipCoordinates();
void convertGeneric();
void optimize_data();
void optimize();
void columnsAndRows();
void convertQTransform();
void fill();
void mapRect_data();
void mapRect();
void mapVector_data();
void mapVector();
void properties();
void metaTypes();
private:
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static void setMatrix(QMatrix2x2& m, const float *values);
static void setMatrixDirect(QMatrix2x2& m, const float *values);
static bool isSame(const QMatrix2x2& m, const float *values);
static bool isIdentity(const QMatrix2x2& m);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static void setMatrix(QMatrix3x3& m, const float *values);
static void setMatrixDirect(QMatrix3x3& m, const float *values);
static bool isSame(const QMatrix3x3& m, const float *values);
static bool isIdentity(const QMatrix3x3& m);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static void setMatrix(QMatrix4x4& m, const float *values);
static void setMatrixDirect(QMatrix4x4& m, const float *values);
static bool isSame(const QMatrix4x4& m, const float *values);
static bool isIdentity(const QMatrix4x4& m);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static void setMatrix(QMatrix4x3& m, const float *values);
static void setMatrixDirect(QMatrix4x3& m, const float *values);
static bool isSame(const QMatrix4x3& m, const float *values);
static bool isIdentity(const QMatrix4x3& m);
};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float nullValues2[] =
{0.0f, 0.0f,
0.0f, 0.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const identityValues2[16] =
{1.0f, 0.0f,
0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float doubleIdentity2[] =
{2.0f, 0.0f,
0.0f, 2.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const uniqueValues2[16] =
{1.0f, 2.0f,
5.0f, 6.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const transposedValues2[16] =
{1.0f, 5.0f,
2.0f, 6.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float nullValues3[] =
{0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const identityValues3[16] =
{1.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float doubleIdentity3[] =
{2.0f, 0.0f, 0.0f,
0.0f, 2.0f, 0.0f,
0.0f, 0.0f, 2.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const uniqueValues3[16] =
{1.0f, 2.0f, 3.0f,
5.0f, 6.0f, 7.0f,
9.0f, 10.0f, 11.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const transposedValues3[16] =
{1.0f, 5.0f, 9.0f,
2.0f, 6.0f, 10.0f,
3.0f, 7.0f, 11.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float nullValues4[] =
{0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const identityValues4[16] =
{1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float doubleIdentity4[] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 2.0f, 0.0f, 0.0f,
0.0f, 0.0f, 2.0f, 0.0f,
0.0f, 0.0f, 0.0f, 2.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const uniqueValues4[16] =
{1.0f, 2.0f, 3.0f, 4.0f,
5.0f, 6.0f, 7.0f, 8.0f,
9.0f, 10.0f, 11.0f, 12.0f,
13.0f, 14.0f, 15.0f, 16.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const transposedValues4[16] =
{1.0f, 5.0f, 9.0f, 13.0f,
2.0f, 6.0f, 10.0f, 14.0f,
3.0f, 7.0f, 11.0f, 15.0f,
4.0f, 8.0f, 12.0f, 16.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float nullValues4x3[] =
{0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const identityValues4x3[12] =
{1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const doubleIdentity4x3[12] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 2.0f, 0.0f, 0.0f,
0.0f, 0.0f, 2.0f, 0.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const uniqueValues4x3[12] =
{1.0f, 2.0f, 3.0f, 4.0f,
5.0f, 6.0f, 7.0f, 8.0f,
9.0f, 10.0f, 11.0f, 12.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const transposedValues3x4[12] =
{1.0f, 5.0f, 9.0f,
2.0f, 6.0f, 10.0f,
3.0f, 7.0f, 11.0f,
4.0f, 8.0f, 12.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
// We use a slightly better implementation of qFuzzyCompare here that
// handles the case where one of the values is exactly 0
static inline bool fuzzyCompare(float p1, float p2)
{
if (qFuzzyIsNull(p1))
return qFuzzyIsNull(p2);
else if (qFuzzyIsNull(p2))
return false;
else
return qFuzzyCompare(p1, p2);
}
// Set a matrix to a specified array of values, which are assumed
// to be in row-major order. This sets the values using floating-point.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrix(QMatrix2x2& m, const float *values)
{
for (int row = 0; row < 2; ++row)
for (int col = 0; col < 2; ++col)
m(row, col) = values[row * 2 + col];
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrix(QMatrix3x3& m, const float *values)
{
for (int row = 0; row < 3; ++row)
for (int col = 0; col < 3; ++col)
m(row, col) = values[row * 3 + col];
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrix(QMatrix4x4& m, const float *values)
{
for (int row = 0; row < 4; ++row)
for (int col = 0; col < 4; ++col)
m(row, col) = values[row * 4 + col];
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrix(QMatrix4x3& m, const float *values)
{
for (int row = 0; row < 3; ++row)
for (int col = 0; col < 4; ++col)
m(row, col) = values[row * 4 + col];
}
// Set a matrix to a specified array of values, which are assumed
// to be in row-major order. This sets the values directly into
// the internal data() array.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrixDirect(QMatrix2x2& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float *data = m.data();
for (int row = 0; row < 2; ++row) {
for (int col = 0; col < 2; ++col) {
data[row + col * 2] = values[row * 2 + col];
}
}
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrixDirect(QMatrix3x3& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float *data = m.data();
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 3; ++col) {
data[row + col * 3] = values[row * 3 + col];
}
}
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrixDirect(QMatrix4x4& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float *data = m.data();
for (int row = 0; row < 4; ++row) {
for (int col = 0; col < 4; ++col) {
data[row + col * 4] = values[row * 4 + col];
}
}
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
void tst_QMatrixNxN::setMatrixDirect(QMatrix4x3& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float *data = m.data();
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 4; ++col) {
data[row + col * 3] = values[row * 4 + col];
}
}
}
// Determine if a matrix is the same as a specified array of values.
// The values are assumed to be specified in row-major order.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
bool tst_QMatrixNxN::isSame(const QMatrix2x2& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *mv = m.constData();
for (int row = 0; row < 2; ++row) {
for (int col = 0; col < 2; ++col) {
// Check the values using the operator() function.
if (!fuzzyCompare(m(row, col), values[row * 2 + col])) {
qDebug() << "floating-point failure at" << row << col << "actual =" << m(row, col) << "expected =" << values[row * 2 + col];
return false;
}
// Check the values using direct access, which verifies that the values
// are stored internally in column-major order.
if (!fuzzyCompare(mv[col * 2 + row], values[row * 2 + col])) {
qDebug() << "column floating-point failure at" << row << col << "actual =" << mv[col * 2 + row] << "expected =" << values[row * 2 + col];
return false;
}
}
}
return true;
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
bool tst_QMatrixNxN::isSame(const QMatrix3x3& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *mv = m.constData();
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 3; ++col) {
// Check the values using the operator() access function.
if (!fuzzyCompare(m(row, col), values[row * 3 + col])) {
qDebug() << "floating-point failure at" << row << col << "actual =" << m(row, col) << "expected =" << values[row * 3 + col];
return false;
}
// Check the values using direct access, which verifies that the values
// are stored internally in column-major order.
if (!fuzzyCompare(mv[col * 3 + row], values[row * 3 + col])) {
qDebug() << "column floating-point failure at" << row << col << "actual =" << mv[col * 3 + row] << "expected =" << values[row * 3 + col];
return false;
}
}
}
return true;
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
bool tst_QMatrixNxN::isSame(const QMatrix4x4& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *mv = m.constData();
for (int row = 0; row < 4; ++row) {
for (int col = 0; col < 4; ++col) {
// Check the values using the operator() access function.
if (!fuzzyCompare(m(row, col), values[row * 4 + col])) {
qDebug() << "floating-point failure at" << row << col << "actual =" << m(row, col) << "expected =" << values[row * 4 + col];
return false;
}
// Check the values using direct access, which verifies that the values
// are stored internally in column-major order.
if (!fuzzyCompare(mv[col * 4 + row], values[row * 4 + col])) {
qDebug() << "column floating-point failure at" << row << col << "actual =" << mv[col * 4 + row] << "expected =" << values[row * 4 + col];
return false;
}
}
}
return true;
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
bool tst_QMatrixNxN::isSame(const QMatrix4x3& m, const float *values)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *mv = m.constData();
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 4; ++col) {
// Check the values using the operator() access function.
if (!fuzzyCompare(m(row, col), values[row * 4 + col])) {
qDebug() << "floating-point failure at" << row << col << "actual =" << m(row, col) << "expected =" << values[row * 4 + col];
return false;
}
// Check the values using direct access, which verifies that the values
// are stored internally in column-major order.
if (!fuzzyCompare(mv[col * 3 + row], values[row * 4 + col])) {
qDebug() << "column floating-point failure at" << row << col << "actual =" << mv[col * 3 + row] << "expected =" << values[row * 4 + col];
return false;
}
}
}
return true;
}
// Determine if a matrix is the identity.
bool tst_QMatrixNxN::isIdentity(const QMatrix2x2& m)
{
return isSame(m, identityValues2);
}
bool tst_QMatrixNxN::isIdentity(const QMatrix3x3& m)
{
return isSame(m, identityValues3);
}
bool tst_QMatrixNxN::isIdentity(const QMatrix4x4& m)
{
return isSame(m, identityValues4);
}
bool tst_QMatrixNxN::isIdentity(const QMatrix4x3& m)
{
return isSame(m, identityValues4x3);
}
// Test the creation of QMatrix2x2 objects in various ways:
// construct, copy, and modify.
void tst_QMatrixNxN::create2x2()
{
QMatrix2x2 m1;
QVERIFY(isIdentity(m1));
QVERIFY(m1.isIdentity());
QMatrix2x2 m2;
setMatrix(m2, uniqueValues2);
QVERIFY(isSame(m2, uniqueValues2));
QVERIFY(!m2.isIdentity());
QMatrix2x2 m3;
setMatrixDirect(m3, uniqueValues2);
QVERIFY(isSame(m3, uniqueValues2));
QMatrix2x2 m4(m3);
QVERIFY(isSame(m4, uniqueValues2));
QMatrix2x2 m5;
m5 = m3;
QVERIFY(isSame(m5, uniqueValues2));
m5.setToIdentity();
QVERIFY(isIdentity(m5));
QMatrix2x2 m6(uniqueValues2);
QVERIFY(isSame(m6, uniqueValues2));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float vals[4];
m6.copyDataTo(vals);
for (int index = 0; index < 4; ++index)
QCOMPARE(vals[index], uniqueValues2[index]);
}
// Test the creation of QMatrix3x3 objects in various ways:
// construct, copy, and modify.
void tst_QMatrixNxN::create3x3()
{
QMatrix3x3 m1;
QVERIFY(isIdentity(m1));
QVERIFY(m1.isIdentity());
QMatrix3x3 m2;
setMatrix(m2, uniqueValues3);
QVERIFY(isSame(m2, uniqueValues3));
QVERIFY(!m2.isIdentity());
QMatrix3x3 m3;
setMatrixDirect(m3, uniqueValues3);
QVERIFY(isSame(m3, uniqueValues3));
QMatrix3x3 m4(m3);
QVERIFY(isSame(m4, uniqueValues3));
QMatrix3x3 m5;
m5 = m3;
QVERIFY(isSame(m5, uniqueValues3));
m5.setToIdentity();
QVERIFY(isIdentity(m5));
QMatrix3x3 m6(uniqueValues3);
QVERIFY(isSame(m6, uniqueValues3));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float vals[9];
m6.copyDataTo(vals);
for (int index = 0; index < 9; ++index)
QCOMPARE(vals[index], uniqueValues3[index]);
}
// Test the creation of QMatrix4x4 objects in various ways:
// construct, copy, and modify.
void tst_QMatrixNxN::create4x4()
{
QMatrix4x4 m1;
QVERIFY(isIdentity(m1));
QVERIFY(m1.isIdentity());
QMatrix4x4 m2;
setMatrix(m2, uniqueValues4);
QVERIFY(isSame(m2, uniqueValues4));
QVERIFY(!m2.isIdentity());
QMatrix4x4 m3;
setMatrixDirect(m3, uniqueValues4);
QVERIFY(isSame(m3, uniqueValues4));
QMatrix4x4 m4(m3);
QVERIFY(isSame(m4, uniqueValues4));
QMatrix4x4 m5;
m5 = m3;
QVERIFY(isSame(m5, uniqueValues4));
m5.setToIdentity();
QVERIFY(isIdentity(m5));
QMatrix4x4 m6(uniqueValues4);
QVERIFY(isSame(m6, uniqueValues4));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float vals[16];
m6.copyDataTo(vals);
for (int index = 0; index < 16; ++index)
QCOMPARE(vals[index], uniqueValues4[index]);
QMatrix4x4 m8
(uniqueValues4[0], uniqueValues4[1], uniqueValues4[2], uniqueValues4[3],
uniqueValues4[4], uniqueValues4[5], uniqueValues4[6], uniqueValues4[7],
uniqueValues4[8], uniqueValues4[9], uniqueValues4[10], uniqueValues4[11],
uniqueValues4[12], uniqueValues4[13], uniqueValues4[14], uniqueValues4[15]);
QVERIFY(isSame(m8, uniqueValues4));
}
// Test the creation of QMatrix4x3 objects in various ways:
// construct, copy, and modify.
void tst_QMatrixNxN::create4x3()
{
QMatrix4x3 m1;
QVERIFY(isIdentity(m1));
QVERIFY(m1.isIdentity());
QMatrix4x3 m2;
setMatrix(m2, uniqueValues4x3);
QVERIFY(isSame(m2, uniqueValues4x3));
QVERIFY(!m2.isIdentity());
QMatrix4x3 m3;
setMatrixDirect(m3, uniqueValues4x3);
QVERIFY(isSame(m3, uniqueValues4x3));
QMatrix4x3 m4(m3);
QVERIFY(isSame(m4, uniqueValues4x3));
QMatrix4x3 m5;
m5 = m3;
QVERIFY(isSame(m5, uniqueValues4x3));
m5.setToIdentity();
QVERIFY(isIdentity(m5));
QMatrix4x3 m6(uniqueValues4x3);
QVERIFY(isSame(m6, uniqueValues4x3));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float vals[12];
m6.copyDataTo(vals);
for (int index = 0; index < 12; ++index)
QCOMPARE(vals[index], uniqueValues4x3[index]);
}
// Test isIdentity() for 2x2 matrices.
void tst_QMatrixNxN::isIdentity2x2()
{
for (int i = 0; i < 2 * 2; ++i) {
QMatrix2x2 m;
QVERIFY(m.isIdentity());
m.data()[i] = 42.0f;
QVERIFY(!m.isIdentity());
}
}
// Test isIdentity() for 3x3 matrices.
void tst_QMatrixNxN::isIdentity3x3()
{
for (int i = 0; i < 3 * 3; ++i) {
QMatrix3x3 m;
QVERIFY(m.isIdentity());
m.data()[i] = 42.0f;
QVERIFY(!m.isIdentity());
}
}
// Test isIdentity() for 4x4 matrices.
void tst_QMatrixNxN::isIdentity4x4()
{
for (int i = 0; i < 4 * 4; ++i) {
QMatrix4x4 m;
QVERIFY(m.isIdentity());
m.data()[i] = 42.0f;
QVERIFY(!m.isIdentity());
}
// Force the "Identity" flag bit to be lost and check again.
QMatrix4x4 m2;
m2.data()[0] = 1.0f;
QVERIFY(m2.isIdentity());
}
// Test isIdentity() for 4x3 matrices.
void tst_QMatrixNxN::isIdentity4x3()
{
for (int i = 0; i < 4 * 3; ++i) {
QMatrix4x3 m;
QVERIFY(m.isIdentity());
m.data()[i] = 42.0f;
QVERIFY(!m.isIdentity());
}
}
// Test 2x2 matrix comparisons.
void tst_QMatrixNxN::compare2x2()
{
QMatrix2x2 m1(uniqueValues2);
QMatrix2x2 m2(uniqueValues2);
QMatrix2x2 m3(transposedValues2);
QCOMPARE(m1, m2);
QVERIFY(!(m1 != m2));
QVERIFY(m1 != m3);
QVERIFY(!(m1 == m3));
}
// Test 3x3 matrix comparisons.
void tst_QMatrixNxN::compare3x3()
{
QMatrix3x3 m1(uniqueValues3);
QMatrix3x3 m2(uniqueValues3);
QMatrix3x3 m3(transposedValues3);
QCOMPARE(m1, m2);
QVERIFY(!(m1 != m2));
QVERIFY(m1 != m3);
QVERIFY(!(m1 == m3));
}
// Test 4x4 matrix comparisons.
void tst_QMatrixNxN::compare4x4()
{
QMatrix4x4 m1(uniqueValues4);
QMatrix4x4 m2(uniqueValues4);
QMatrix4x4 m3(transposedValues4);
QCOMPARE(m1, m2);
QVERIFY(!(m1 != m2));
QVERIFY(m1 != m3);
QVERIFY(!(m1 == m3));
}
// Test 4x3 matrix comparisons.
void tst_QMatrixNxN::compare4x3()
{
QMatrix4x3 m1(uniqueValues4x3);
QMatrix4x3 m2(uniqueValues4x3);
QMatrix4x3 m3(transposedValues3x4);
QCOMPARE(m1, m2);
QVERIFY(!(m1 != m2));
QVERIFY(m1 != m3);
QVERIFY(!(m1 == m3));
}
// Test matrix 2x2 transpose operations.
void tst_QMatrixNxN::transposed2x2()
{
// Transposing the identity should result in the identity.
QMatrix2x2 m1;
QMatrix2x2 m2 = m1.transposed();
QVERIFY(isIdentity(m2));
// Transpose a more interesting matrix that allows us to track
// exactly where each source element ends up.
QMatrix2x2 m3(uniqueValues2);
QMatrix2x2 m4 = m3.transposed();
QVERIFY(isSame(m4, transposedValues2));
// Transpose in-place, just to check that the compiler is sane.
m3 = m3.transposed();
QVERIFY(isSame(m3, transposedValues2));
}
// Test matrix 3x3 transpose operations.
void tst_QMatrixNxN::transposed3x3()
{
// Transposing the identity should result in the identity.
QMatrix3x3 m1;
QMatrix3x3 m2 = m1.transposed();
QVERIFY(isIdentity(m2));
// Transpose a more interesting matrix that allows us to track
// exactly where each source element ends up.
QMatrix3x3 m3(uniqueValues3);
QMatrix3x3 m4 = m3.transposed();
QVERIFY(isSame(m4, transposedValues3));
// Transpose in-place, just to check that the compiler is sane.
m3 = m3.transposed();
QVERIFY(isSame(m3, transposedValues3));
}
// Test matrix 4x4 transpose operations.
void tst_QMatrixNxN::transposed4x4()
{
// Transposing the identity should result in the identity.
QMatrix4x4 m1;
QMatrix4x4 m2 = m1.transposed();
QVERIFY(isIdentity(m2));
// Transpose a more interesting matrix that allows us to track
// exactly where each source element ends up.
QMatrix4x4 m3(uniqueValues4);
QMatrix4x4 m4 = m3.transposed();
QVERIFY(isSame(m4, transposedValues4));
// Transpose in-place, just to check that the compiler is sane.
m3 = m3.transposed();
QVERIFY(isSame(m3, transposedValues4));
}
// Test matrix 4x3 transpose operations.
void tst_QMatrixNxN::transposed4x3()
{
QMatrix4x3 m3(uniqueValues4x3);
QMatrix3x4 m4 = m3.transposed();
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float values[12];
m4.copyDataTo(values);
for (int index = 0; index < 12; ++index)
QCOMPARE(values[index], transposedValues3x4[index]);
}
// Test matrix addition for 2x2 matrices.
void tst_QMatrixNxN::add2x2_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues2 << (void *)nullValues2 << (void *)nullValues2;
QTest::newRow("identity/null")
<< (void *)identityValues2 << (void *)nullValues2 << (void *)identityValues2;
QTest::newRow("identity/identity")
<< (void *)identityValues2 << (void *)identityValues2 << (void *)doubleIdentity2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const sumValues[16] =
{2.0f, 7.0f,
7.0f, 12.0f};
QTest::newRow("unique")
<< (void *)uniqueValues2 << (void *)transposedValues2 << (void *)sumValues;
}
void tst_QMatrixNxN::add2x2()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix2x2 m1((const float *)m1Values);
QMatrix2x2 m2((const float *)m2Values);
QMatrix2x2 m4(m1);
m4 += m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m3Values));
QMatrix2x2 m5;
m5 = m1 + m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
}
// Test matrix addition for 3x3 matrices.
void tst_QMatrixNxN::add3x3_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues3 << (void *)nullValues3 << (void *)nullValues3;
QTest::newRow("identity/null")
<< (void *)identityValues3 << (void *)nullValues3 << (void *)identityValues3;
QTest::newRow("identity/identity")
<< (void *)identityValues3 << (void *)identityValues3 << (void *)doubleIdentity3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const sumValues[16] =
{2.0f, 7.0f, 12.0f,
7.0f, 12.0f, 17.0f,
12.0f, 17.0f, 22.0f};
QTest::newRow("unique")
<< (void *)uniqueValues3 << (void *)transposedValues3 << (void *)sumValues;
}
void tst_QMatrixNxN::add3x3()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix3x3 m1((const float *)m1Values);
QMatrix3x3 m2((const float *)m2Values);
QMatrix3x3 m4(m1);
m4 += m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m3Values));
QMatrix3x3 m5;
m5 = m1 + m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
}
// Test matrix addition for 4x4 matrices.
void tst_QMatrixNxN::add4x4_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues4 << (void *)nullValues4 << (void *)nullValues4;
QTest::newRow("identity/null")
<< (void *)identityValues4 << (void *)nullValues4 << (void *)identityValues4;
QTest::newRow("identity/identity")
<< (void *)identityValues4 << (void *)identityValues4 << (void *)doubleIdentity4;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const sumValues[16] =
{2.0f, 7.0f, 12.0f, 17.0f,
7.0f, 12.0f, 17.0f, 22.0f,
12.0f, 17.0f, 22.0f, 27.0f,
17.0f, 22.0f, 27.0f, 32.0f};
QTest::newRow("unique")
<< (void *)uniqueValues4 << (void *)transposedValues4 << (void *)sumValues;
}
void tst_QMatrixNxN::add4x4()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m1((const float *)m1Values);
QMatrix4x4 m2((const float *)m2Values);
QMatrix4x4 m4(m1);
m4 += m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m3Values));
QMatrix4x4 m5;
m5 = m1 + m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
}
// Test matrix addition for 4x3 matrices.
void tst_QMatrixNxN::add4x3_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues4x3 << (void *)nullValues4x3 << (void *)nullValues4x3;
QTest::newRow("identity/null")
<< (void *)identityValues4x3 << (void *)nullValues4x3 << (void *)identityValues4x3;
QTest::newRow("identity/identity")
<< (void *)identityValues4x3 << (void *)identityValues4x3 << (void *)doubleIdentity4x3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const sumValues[16] =
{2.0f, 7.0f, 12.0f, 6.0f,
11.0f, 16.0f, 10.0f, 15.0f,
20.0f, 14.0f, 19.0f, 24.0f};
QTest::newRow("unique")
<< (void *)uniqueValues4x3 << (void *)transposedValues3x4 << (void *)sumValues;
}
void tst_QMatrixNxN::add4x3()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x3 m1((const float *)m1Values);
QMatrix4x3 m2((const float *)m2Values);
QMatrix4x3 m4(m1);
m4 += m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m3Values));
QMatrix4x3 m5;
m5 = m1 + m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
}
// Test matrix subtraction for 2x2 matrices.
void tst_QMatrixNxN::subtract2x2_data()
{
// Use the same test cases as the add test.
add2x2_data();
}
void tst_QMatrixNxN::subtract2x2()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix2x2 m1((const float *)m1Values);
QMatrix2x2 m2((const float *)m2Values);
QMatrix2x2 m3((const float *)m3Values);
QMatrix2x2 m4(m3);
m4 -= m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix2x2 m5;
m5 = m3 - m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
QMatrix2x2 m6(m3);
m6 -= m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m6, (const float *)m1Values));
QMatrix2x2 m7;
m7 = m3 - m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m7, (const float *)m1Values));
}
// Test matrix subtraction for 3x3 matrices.
void tst_QMatrixNxN::subtract3x3_data()
{
// Use the same test cases as the add test.
add3x3_data();
}
void tst_QMatrixNxN::subtract3x3()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix3x3 m1((const float *)m1Values);
QMatrix3x3 m2((const float *)m2Values);
QMatrix3x3 m3((const float *)m3Values);
QMatrix3x3 m4(m3);
m4 -= m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix3x3 m5;
m5 = m3 - m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
QMatrix3x3 m6(m3);
m6 -= m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m6, (const float *)m1Values));
QMatrix3x3 m7;
m7 = m3 - m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m7, (const float *)m1Values));
}
// Test matrix subtraction for 4x4 matrices.
void tst_QMatrixNxN::subtract4x4_data()
{
// Use the same test cases as the add test.
add4x4_data();
}
void tst_QMatrixNxN::subtract4x4()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m1((const float *)m1Values);
QMatrix4x4 m2((const float *)m2Values);
QMatrix4x4 m3((const float *)m3Values);
QMatrix4x4 m4(m3);
m4 -= m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix4x4 m5;
m5 = m3 - m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
QMatrix4x4 m6(m3);
m6 -= m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m6, (const float *)m1Values));
QMatrix4x4 m7;
m7 = m3 - m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m7, (const float *)m1Values));
}
// Test matrix subtraction for 4x3 matrices.
void tst_QMatrixNxN::subtract4x3_data()
{
// Use the same test cases as the add test.
add4x3_data();
}
void tst_QMatrixNxN::subtract4x3()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x3 m1((const float *)m1Values);
QMatrix4x3 m2((const float *)m2Values);
QMatrix4x3 m3((const float *)m3Values);
QMatrix4x3 m4(m3);
m4 -= m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix4x3 m5;
m5 = m3 - m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
QMatrix4x3 m6(m3);
m6 -= m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m6, (const float *)m1Values));
QMatrix4x3 m7;
m7 = m3 - m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m7, (const float *)m1Values));
}
// Test matrix multiplication for 2x2 matrices.
void tst_QMatrixNxN::multiply2x2_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues2 << (void *)nullValues2 << (void *)nullValues2;
QTest::newRow("null/unique")
<< (void *)nullValues2 << (void *)uniqueValues2 << (void *)nullValues2;
QTest::newRow("unique/null")
<< (void *)uniqueValues2 << (void *)nullValues2 << (void *)nullValues2;
QTest::newRow("unique/identity")
<< (void *)uniqueValues2 << (void *)identityValues2 << (void *)uniqueValues2;
QTest::newRow("identity/unique")
<< (void *)identityValues2 << (void *)uniqueValues2 << (void *)uniqueValues2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float uniqueResult[4];
for (int row = 0; row < 2; ++row) {
for (int col = 0; col < 2; ++col) {
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float sum = 0.0f;
for (int j = 0; j < 2; ++j)
sum += uniqueValues2[row * 2 + j] * transposedValues2[j * 2 + col];
uniqueResult[row * 2 + col] = sum;
}
}
QTest::newRow("unique/transposed")
<< (void *)uniqueValues2 << (void *)transposedValues2 << (void *)uniqueResult;
}
void tst_QMatrixNxN::multiply2x2()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix2x2 m1((const float *)m1Values);
QMatrix2x2 m2((const float *)m2Values);
QMatrix2x2 m5;
m5 = m1 * m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
}
// Test matrix multiplication for 3x3 matrices.
void tst_QMatrixNxN::multiply3x3_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues3 << (void *)nullValues3 << (void *)nullValues3;
QTest::newRow("null/unique")
<< (void *)nullValues3 << (void *)uniqueValues3 << (void *)nullValues3;
QTest::newRow("unique/null")
<< (void *)uniqueValues3 << (void *)nullValues3 << (void *)nullValues3;
QTest::newRow("unique/identity")
<< (void *)uniqueValues3 << (void *)identityValues3 << (void *)uniqueValues3;
QTest::newRow("identity/unique")
<< (void *)identityValues3 << (void *)uniqueValues3 << (void *)uniqueValues3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float uniqueResult[9];
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 3; ++col) {
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float sum = 0.0f;
for (int j = 0; j < 3; ++j)
sum += uniqueValues3[row * 3 + j] * transposedValues3[j * 3 + col];
uniqueResult[row * 3 + col] = sum;
}
}
QTest::newRow("unique/transposed")
<< (void *)uniqueValues3 << (void *)transposedValues3 << (void *)uniqueResult;
}
void tst_QMatrixNxN::multiply3x3()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix3x3 m1((const float *)m1Values);
QMatrix3x3 m2((const float *)m2Values);
QMatrix3x3 m5;
m5 = m1 * m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
}
// Test matrix multiplication for 4x4 matrices.
void tst_QMatrixNxN::multiply4x4_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues4 << (void *)nullValues4 << (void *)nullValues4;
QTest::newRow("null/unique")
<< (void *)nullValues4 << (void *)uniqueValues4 << (void *)nullValues4;
QTest::newRow("unique/null")
<< (void *)uniqueValues4 << (void *)nullValues4 << (void *)nullValues4;
QTest::newRow("unique/identity")
<< (void *)uniqueValues4 << (void *)identityValues4 << (void *)uniqueValues4;
QTest::newRow("identity/unique")
<< (void *)identityValues4 << (void *)uniqueValues4 << (void *)uniqueValues4;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float uniqueResult[16];
for (int row = 0; row < 4; ++row) {
for (int col = 0; col < 4; ++col) {
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float sum = 0.0f;
for (int j = 0; j < 4; ++j)
sum += uniqueValues4[row * 4 + j] * transposedValues4[j * 4 + col];
uniqueResult[row * 4 + col] = sum;
}
}
QTest::newRow("unique/transposed")
<< (void *)uniqueValues4 << (void *)transposedValues4 << (void *)uniqueResult;
}
void tst_QMatrixNxN::multiply4x4()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m1((const float *)m1Values);
QMatrix4x4 m2((const float *)m2Values);
QMatrix4x4 m4;
m4 = m1;
m4 *= m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m3Values));
QMatrix4x4 m5;
m5 = m1 * m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m3Values));
QMatrix4x4 m1xm1 = m1 * m1;
m1 *= m1;
QCOMPARE(m1, m1xm1);
}
// Test matrix multiplication for 4x3 matrices.
void tst_QMatrixNxN::multiply4x3_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<void *>("m3Values");
QTest::newRow("null")
<< (void *)nullValues4x3 << (void *)nullValues4x3 << (void *)nullValues3;
QTest::newRow("null/unique")
<< (void *)nullValues4x3 << (void *)uniqueValues4x3 << (void *)nullValues3;
QTest::newRow("unique/null")
<< (void *)uniqueValues4x3 << (void *)nullValues4x3 << (void *)nullValues3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float uniqueResult[9];
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 3; ++col) {
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float sum = 0.0f;
for (int j = 0; j < 4; ++j)
sum += uniqueValues4x3[row * 4 + j] * transposedValues3x4[j * 3 + col];
uniqueResult[row * 3 + col] = sum;
}
}
QTest::newRow("unique/transposed")
<< (void *)uniqueValues4x3 << (void *)transposedValues3x4 << (void *)uniqueResult;
}
void tst_QMatrixNxN::multiply4x3()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(void *, m3Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x3 m1((const float *)m1Values);
QMatrix3x4 m2((const float *)m2Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QGenericMatrix<3, 3, float> m4;
m4 = m1 * m2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float values[9];
m4.copyDataTo(values);
for (int index = 0; index < 9; ++index)
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(values[index], ((const float *)m3Values)[index]);
}
// Test matrix multiplication by a factor for 2x2 matrices.
void tst_QMatrixNxN::multiplyFactor2x2_data()
{
QTest::addColumn<void *>("m1Values");
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("factor");
QTest::addColumn<void *>("m2Values");
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)nullValues2 << (float)1.0f << (void *)nullValues2;
QTest::newRow("double identity")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)identityValues2 << (float)2.0f << (void *)doubleIdentity2;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const values[16] =
{1.0f, 2.0f,
5.0f, 6.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const doubleValues[16] =
{2.0f, 4.0f,
10.0f, 12.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const negDoubleValues[16] =
{-2.0f, -4.0f,
-10.0f, -12.0f};
QTest::newRow("unique")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)2.0f << (void *)doubleValues;
QTest::newRow("neg")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)-2.0f << (void *)negDoubleValues;
QTest::newRow("zero")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)0.0f << (void *)nullValues4;
}
void tst_QMatrixNxN::multiplyFactor2x2()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix2x2 m1((const float *)m1Values);
QMatrix2x2 m3;
m3 = m1;
m3 *= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m2Values));
QMatrix2x2 m4;
m4 = m1 * factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix2x2 m5;
m5 = factor * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
}
// Test matrix multiplication by a factor for 3x3 matrices.
void tst_QMatrixNxN::multiplyFactor3x3_data()
{
QTest::addColumn<void *>("m1Values");
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("factor");
QTest::addColumn<void *>("m2Values");
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)nullValues3 << (float)1.0f << (void *)nullValues3;
QTest::newRow("double identity")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)identityValues3 << (float)2.0f << (void *)doubleIdentity3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const values[16] =
{1.0f, 2.0f, 3.0f,
5.0f, 6.0f, 7.0f,
9.0f, 10.0f, 11.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const doubleValues[16] =
{2.0f, 4.0f, 6.0f,
10.0f, 12.0f, 14.0f,
18.0f, 20.0f, 22.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const negDoubleValues[16] =
{-2.0f, -4.0f, -6.0f,
-10.0f, -12.0f, -14.0f,
-18.0f, -20.0f, -22.0f};
QTest::newRow("unique")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)2.0f << (void *)doubleValues;
QTest::newRow("neg")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)-2.0f << (void *)negDoubleValues;
QTest::newRow("zero")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)0.0f << (void *)nullValues4;
}
void tst_QMatrixNxN::multiplyFactor3x3()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix3x3 m1((const float *)m1Values);
QMatrix3x3 m3;
m3 = m1;
m3 *= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m2Values));
QMatrix3x3 m4;
m4 = m1 * factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix3x3 m5;
m5 = factor * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
}
// Test matrix multiplication by a factor for 4x4 matrices.
void tst_QMatrixNxN::multiplyFactor4x4_data()
{
QTest::addColumn<void *>("m1Values");
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("factor");
QTest::addColumn<void *>("m2Values");
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)nullValues4 << (float)1.0f << (void *)nullValues4;
QTest::newRow("double identity")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)identityValues4 << (float)2.0f << (void *)doubleIdentity4;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const values[16] =
{1.0f, 2.0f, 3.0f, 4.0f,
5.0f, 6.0f, 7.0f, 8.0f,
9.0f, 10.0f, 11.0f, 12.0f,
13.0f, 14.0f, 15.0f, 16.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const doubleValues[16] =
{2.0f, 4.0f, 6.0f, 8.0f,
10.0f, 12.0f, 14.0f, 16.0f,
18.0f, 20.0f, 22.0f, 24.0f,
26.0f, 28.0f, 30.0f, 32.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const negDoubleValues[16] =
{-2.0f, -4.0f, -6.0f, -8.0f,
-10.0f, -12.0f, -14.0f, -16.0f,
-18.0f, -20.0f, -22.0f, -24.0f,
-26.0f, -28.0f, -30.0f, -32.0f};
QTest::newRow("unique")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)2.0f << (void *)doubleValues;
QTest::newRow("neg")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)-2.0f << (void *)negDoubleValues;
QTest::newRow("zero")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)0.0f << (void *)nullValues4;
}
void tst_QMatrixNxN::multiplyFactor4x4()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m1((const float *)m1Values);
QMatrix4x4 m3;
m3 = m1;
m3 *= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m2Values));
QMatrix4x4 m4;
m4 = m1 * factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix4x4 m5;
m5 = factor * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
}
// Test matrix multiplication by a factor for 4x3 matrices.
void tst_QMatrixNxN::multiplyFactor4x3_data()
{
QTest::addColumn<void *>("m1Values");
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("factor");
QTest::addColumn<void *>("m2Values");
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)nullValues4x3 << (float)1.0f << (void *)nullValues4x3;
QTest::newRow("double identity")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)identityValues4x3 << (float)2.0f << (void *)doubleIdentity4x3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const values[12] =
{1.0f, 2.0f, 3.0f, 4.0f,
5.0f, 6.0f, 7.0f, 8.0f,
9.0f, 10.0f, 11.0f, 12.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const doubleValues[12] =
{2.0f, 4.0f, 6.0f, 8.0f,
10.0f, 12.0f, 14.0f, 16.0f,
18.0f, 20.0f, 22.0f, 24.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const negDoubleValues[12] =
{-2.0f, -4.0f, -6.0f, -8.0f,
-10.0f, -12.0f, -14.0f, -16.0f,
-18.0f, -20.0f, -22.0f, -24.0f};
QTest::newRow("unique")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)2.0f << (void *)doubleValues;
QTest::newRow("neg")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)-2.0f << (void *)negDoubleValues;
QTest::newRow("zero")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (void *)values << (float)0.0f << (void *)nullValues4x3;
}
void tst_QMatrixNxN::multiplyFactor4x3()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x3 m1((const float *)m1Values);
QMatrix4x3 m3;
m3 = m1;
m3 *= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m2Values));
QMatrix4x3 m4;
m4 = m1 * factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m2Values));
QMatrix4x3 m5;
m5 = factor * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)m2Values));
}
// Test matrix division by a factor for 2x2 matrices.
void tst_QMatrixNxN::divideFactor2x2_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor2x2_data();
}
void tst_QMatrixNxN::divideFactor2x2()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
if (factor == 0.0f)
return;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix2x2 m2((const float *)m2Values);
QMatrix2x2 m3;
m3 = m2;
m3 /= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m1Values));
QMatrix2x2 m4;
m4 = m2 / factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m1Values));
}
// Test matrix division by a factor for 3x3 matrices.
void tst_QMatrixNxN::divideFactor3x3_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor3x3_data();
}
void tst_QMatrixNxN::divideFactor3x3()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
if (factor == 0.0f)
return;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix3x3 m2((const float *)m2Values);
QMatrix3x3 m3;
m3 = m2;
m3 /= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m1Values));
QMatrix3x3 m4;
m4 = m2 / factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m1Values));
}
// Test matrix division by a factor for 4x4 matrices.
void tst_QMatrixNxN::divideFactor4x4_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor4x4_data();
}
void tst_QMatrixNxN::divideFactor4x4()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
if (factor == 0.0f)
return;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m2((const float *)m2Values);
QMatrix4x4 m3;
m3 = m2;
m3 /= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m1Values));
QMatrix4x4 m4;
m4 = m2 / factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m1Values));
}
// Test matrix division by a factor for 4x3 matrices.
void tst_QMatrixNxN::divideFactor4x3_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor4x3_data();
}
void tst_QMatrixNxN::divideFactor4x3()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, factor);
QFETCH(void *, m2Values);
if (factor == 0.0f)
return;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x3 m2((const float *)m2Values);
QMatrix4x3 m3;
m3 = m2;
m3 /= factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m3, (const float *)m1Values));
QMatrix4x3 m4;
m4 = m2 / factor;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m4, (const float *)m1Values));
}
// Test matrix negation for 2x2 matrices.
void tst_QMatrixNxN::negate2x2_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor2x2_data();
}
void tst_QMatrixNxN::negate2x2()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *values = (const float *)m1Values;
QMatrix2x2 m1(values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float negated[4];
for (int index = 0; index < 4; ++index)
negated[index] = -values[index];
QMatrix2x2 m2;
m2 = -m1;
QVERIFY(isSame(m2, negated));
}
// Test matrix negation for 3x3 matrices.
void tst_QMatrixNxN::negate3x3_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor3x3_data();
}
void tst_QMatrixNxN::negate3x3()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *values = (const float *)m1Values;
QMatrix3x3 m1(values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float negated[9];
for (int index = 0; index < 9; ++index)
negated[index] = -values[index];
QMatrix3x3 m2;
m2 = -m1;
QVERIFY(isSame(m2, negated));
}
// Test matrix negation for 4x4 matrices.
void tst_QMatrixNxN::negate4x4_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor4x4_data();
}
void tst_QMatrixNxN::negate4x4()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *values = (const float *)m1Values;
QMatrix4x4 m1(values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float negated[16];
for (int index = 0; index < 16; ++index)
negated[index] = -values[index];
QMatrix4x4 m2;
m2 = -m1;
QVERIFY(isSame(m2, negated));
}
// Test matrix negation for 4x3 matrices.
void tst_QMatrixNxN::negate4x3_data()
{
// Use the same test cases as the multiplyFactor test.
multiplyFactor4x3_data();
}
void tst_QMatrixNxN::negate4x3()
{
QFETCH(void *, m1Values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *values = (const float *)m1Values;
QMatrix4x3 m1(values);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float negated[12];
for (int index = 0; index < 12; ++index)
negated[index] = -values[index];
QMatrix4x3 m2;
m2 = -m1;
QVERIFY(isSame(m2, negated));
}
// Matrix inverted. This is a more straight-forward implementation
// of the algorithm at http://www.j3d.org/matrix_faq/matrfaq_latest.html#Q24
// than the optimized version in the QMatrix4x4 code. Hopefully it is
// easier to verify that this version is the same as the reference.
struct Matrix3
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float v[9];
};
struct Matrix4
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float v[16];
};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float m3Determinant(const Matrix3& m)
{
return m.v[0] * (m.v[4] * m.v[8] - m.v[7] * m.v[5]) -
m.v[1] * (m.v[3] * m.v[8] - m.v[6] * m.v[5]) +
m.v[2] * (m.v[3] * m.v[7] - m.v[6] * m.v[4]);
}
static bool m3Inverse(const Matrix3& min, Matrix3& mout)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float det = m3Determinant(min);
if (det == 0.0f)
return false;
mout.v[0] = (min.v[4] * min.v[8] - min.v[5] * min.v[7]) / det;
mout.v[1] = -(min.v[1] * min.v[8] - min.v[2] * min.v[7]) / det;
mout.v[2] = (min.v[1] * min.v[5] - min.v[4] * min.v[2]) / det;
mout.v[3] = -(min.v[3] * min.v[8] - min.v[5] * min.v[6]) / det;
mout.v[4] = (min.v[0] * min.v[8] - min.v[6] * min.v[2]) / det;
mout.v[5] = -(min.v[0] * min.v[5] - min.v[3] * min.v[2]) / det;
mout.v[6] = (min.v[3] * min.v[7] - min.v[6] * min.v[4]) / det;
mout.v[7] = -(min.v[0] * min.v[7] - min.v[6] * min.v[1]) / det;
mout.v[8] = (min.v[0] * min.v[4] - min.v[1] * min.v[3]) / det;
return true;
}
static void m3Transpose(Matrix3& m)
{
qSwap(m.v[1], m.v[3]);
qSwap(m.v[2], m.v[6]);
qSwap(m.v[5], m.v[7]);
}
static void m4Submatrix(const Matrix4& min, Matrix3& mout, int i, int j)
{
for (int di = 0; di < 3; ++di) {
for (int dj = 0; dj < 3; ++dj) {
int si = di + ((di >= i) ? 1 : 0);
int sj = dj + ((dj >= j) ? 1 : 0);
mout.v[di * 3 + dj] = min.v[si * 4 + sj];
}
}
}
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float m4Determinant(const Matrix4& m)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float det;
float result = 0.0f;
float i = 1.0f;
Matrix3 msub;
for (int n = 0; n < 4; ++n, i *= -1.0f) {
m4Submatrix(m, msub, 0, n);
det = m3Determinant(msub);
result += m.v[n] * det * i;
}
return result;
}
static void m4Inverse(const Matrix4& min, Matrix4& mout)
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float det = m4Determinant(min);
Matrix3 msub;
for (int i = 0; i < 4; ++i) {
for (int j = 0; j < 4; ++j) {
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float sign = 1.0f - ((i + j) % 2) * 2.0f;
m4Submatrix(min, msub, i, j);
mout.v[i + j * 4] = (m3Determinant(msub) * sign) / det;
}
}
}
// Test matrix inverted for 4x4 matrices.
void tst_QMatrixNxN::inverted4x4_data()
{
QTest::addColumn<void *>("m1Values");
QTest::addColumn<void *>("m2Values");
QTest::addColumn<bool>("invertible");
QTest::newRow("null")
<< (void *)nullValues4 << (void *)identityValues4 << false;
QTest::newRow("identity")
<< (void *)identityValues4 << (void *)identityValues4 << true;
QTest::newRow("unique")
<< (void *)uniqueValues4 << (void *)identityValues4 << false;
static Matrix4 const invertible = {
{5.0f, 0.0f, 0.0f, 2.0f,
0.0f, 6.0f, 0.0f, 3.0f,
0.0f, 0.0f, 7.0f, 4.0f,
0.0f, 0.0f, 0.0f, 1.0f}
};
static Matrix4 inverted;
m4Inverse(invertible, inverted);
QTest::newRow("invertible")
<< (void *)invertible.v << (void *)inverted.v << true;
static Matrix4 const invertible2 = {
{1.0f, 2.0f, 4.0f, 2.0f,
8.0f, 3.0f, 5.0f, 3.0f,
6.0f, 7.0f, 9.0f, 4.0f,
0.0f, 0.0f, 0.0f, 1.0f}
};
static Matrix4 inverted2;
m4Inverse(invertible2, inverted2);
QTest::newRow("invertible2")
<< (void *)invertible2.v << (void *)inverted2.v << true;
static Matrix4 const translate = {
{1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 1.0f, 0.0f, 3.0f,
0.0f, 0.0f, 1.0f, 4.0f,
0.0f, 0.0f, 0.0f, 1.0f}
};
static Matrix4 const inverseTranslate = {
{1.0f, 0.0f, 0.0f, -2.0f,
0.0f, 1.0f, 0.0f, -3.0f,
0.0f, 0.0f, 1.0f, -4.0f,
0.0f, 0.0f, 0.0f, 1.0f}
};
QTest::newRow("translate")
<< (void *)translate.v << (void *)inverseTranslate.v << true;
}
void tst_QMatrixNxN::inverted4x4()
{
QFETCH(void *, m1Values);
QFETCH(void *, m2Values);
QFETCH(bool, invertible);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m1((const float *)m1Values);
if (invertible)
QVERIFY(m1.determinant() != 0.0f);
else
QCOMPARE(m1.determinant(), 0.0f);
Matrix4 m1alt;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
memcpy(m1alt.v, (const float *)m1Values, sizeof(m1alt.v));
QCOMPARE(m1.determinant(), m4Determinant(m1alt));
QMatrix4x4 m2;
bool inv;
m2 = m1.inverted(&inv);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m2, (const float *)m2Values));
if (invertible) {
QVERIFY(inv);
Matrix4 m2alt;
m4Inverse(m1alt, m2alt);
QVERIFY(isSame(m2, m2alt.v));
QMatrix4x4 m3;
m3 = m1 * m2;
QVERIFY(isIdentity(m3));
QMatrix4x4 m4;
m4 = m2 * m1;
QVERIFY(isIdentity(m4));
} else {
QVERIFY(!inv);
}
// Test again, after inferring the special matrix type.
m1.optimize();
m2 = m1.inverted(&inv);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m2, (const float *)m2Values));
QCOMPARE(inv, invertible);
}
void tst_QMatrixNxN::orthonormalInverse4x4()
{
QMatrix4x4 m1;
QVERIFY(qFuzzyCompare(m1.inverted(), m1));
QMatrix4x4 m2;
m2.rotate(45.0, 1.0, 0.0, 0.0);
m2.translate(10.0, 0.0, 0.0);
// Use operator() to drop the internal flags that
// mark the matrix as orthonormal. This will force inverted()
// to compute m3.inverted() the long way. We can then compare
// the result to what the faster algorithm produces on m2.
QMatrix4x4 m3 = m2;
m3(0, 0);
bool invertible;
QVERIFY(qFuzzyCompare(m2.inverted(&invertible), m3.inverted()));
QVERIFY(invertible);
QMatrix4x4 m4;
m4.rotate(45.0, 0.0, 1.0, 0.0);
QMatrix4x4 m5 = m4;
m5(0, 0);
QVERIFY(qFuzzyCompare(m4.inverted(), m5.inverted()));
QMatrix4x4 m6;
m1.rotate(88, 0.0, 0.0, 1.0);
m1.translate(-20.0, 20.0, 15.0);
m1.rotate(25, 1.0, 0.0, 0.0);
QMatrix4x4 m7 = m6;
m7(0, 0);
QVERIFY(qFuzzyCompare(m6.inverted(), m7.inverted()));
}
// Test the generation and use of 4x4 scale matrices.
void tst_QMatrixNxN::scale4x4_data()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("x");
QTest::addColumn<float>("y");
QTest::addColumn<float>("z");
QTest::addColumn<void *>("resultValues");
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float nullScale[] =
{0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)0.0f << (float)0.0f << (float)0.0f << (void *)nullScale;
QTest::newRow("identity")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)1.0f << (float)1.0f << (float)1.0f << (void *)identityValues4;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float doubleScale[] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 2.0f, 0.0f, 0.0f,
0.0f, 0.0f, 2.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("double")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)2.0f << (float)2.0f << (float)2.0f << (void *)doubleScale;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float complexScale[] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 11.0f, 0.0f, 0.0f,
0.0f, 0.0f, -6.5f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("complex")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)2.0f << (float)11.0f << (float)-6.5f << (void *)complexScale;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float complexScale2D[] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, -11.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("complex2D")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)2.0f << (float)-11.0f << (float)1.0f << (void *)complexScale2D;
}
void tst_QMatrixNxN::scale4x4()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, x);
QFETCH(float, y);
QFETCH(float, z);
QFETCH(void *, resultValues);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 result((const float *)resultValues);
QMatrix4x4 m1;
m1.scale(QVector3D(x, y, z));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m1, (const float *)resultValues));
QMatrix4x4 m2;
m2.scale(x, y, z);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m2, (const float *)resultValues));
if (z == 1.0f) {
QMatrix4x4 m2b;
m2b.scale(x, y);
QCOMPARE(m2b, m2);
}
QVector3D v1(2.0f, 3.0f, -4.0f);
QVector3D v2 = m1 * v1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v2.x(), (float)(2.0f * x));
QCOMPARE(v2.y(), (float)(3.0f * y));
QCOMPARE(v2.z(), (float)(-4.0f * z));
v2 = v1 * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v2.x(), (float)(2.0f * x));
QCOMPARE(v2.y(), (float)(3.0f * y));
QCOMPARE(v2.z(), (float)(-4.0f * z));
QVector4D v3(2.0f, 3.0f, -4.0f, 34.0f);
QVector4D v4 = m1 * v3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v4.x(), (float)(2.0f * x));
QCOMPARE(v4.y(), (float)(3.0f * y));
QCOMPARE(v4.z(), (float)(-4.0f * z));
QCOMPARE(v4.w(), (float)34.0f);
v4 = v3 * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v4.x(), (float)(2.0f * x));
QCOMPARE(v4.y(), (float)(3.0f * y));
QCOMPARE(v4.z(), (float)(-4.0f * z));
QCOMPARE(v4.w(), (float)34.0f);
QPoint p1(2, 3);
QPoint p2 = m1 * p1;
QCOMPARE(p2.x(), (int)(2.0f * x));
QCOMPARE(p2.y(), (int)(3.0f * y));
p2 = p1 * m1;
QCOMPARE(p2.x(), (int)(2.0f * x));
QCOMPARE(p2.y(), (int)(3.0f * y));
QPointF p3(2.0f, 3.0f);
QPointF p4 = m1 * p3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(p4.x(), (float)(2.0f * x));
QCOMPARE(p4.y(), (float)(3.0f * y));
p4 = p3 * m1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(p4.x(), (float)(2.0f * x));
QCOMPARE(p4.y(), (float)(3.0f * y));
QMatrix4x4 m3(uniqueValues4);
QMatrix4x4 m4(m3);
m4.scale(x, y, z);
QVERIFY(m4 == m3 * m1);
if (x == y && y == z) {
QMatrix4x4 m5;
m5.scale(x);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)resultValues));
}
if (z == 1.0f) {
QMatrix4x4 m4b(m3);
m4b.scale(x, y);
QCOMPARE(m4b, m4);
}
// Test coverage when the special matrix type is unknown.
QMatrix4x4 m6;
m6(0, 0) = 1.0f;
m6.scale(QVector3D(x, y, z));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m6, (const float *)resultValues));
QMatrix4x4 m7;
m7(0, 0) = 1.0f;
m7.scale(x, y, z);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m7, (const float *)resultValues));
if (x == y && y == z) {
QMatrix4x4 m8;
m8(0, 0) = 1.0f;
m8.scale(x);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m8, (const float *)resultValues));
m8.optimize();
m8.scale(1.0f);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m8, (const float *)resultValues));
QMatrix4x4 m9;
m9.translate(0.0f, 0.0f, 0.0f);
m9.scale(x);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m9, (const float *)resultValues));
}
}
// Test the generation and use of 4x4 translation matrices.
void tst_QMatrixNxN::translate4x4_data()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("x");
QTest::addColumn<float>("y");
QTest::addColumn<float>("z");
QTest::addColumn<void *>("resultValues");
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)0.0f << (float)0.0f << (float)0.0f << (void *)identityValues4;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float identityTranslate[] =
{1.0f, 0.0f, 0.0f, 1.0f,
0.0f, 1.0f, 0.0f, 1.0f,
0.0f, 0.0f, 1.0f, 1.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("identity")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)1.0f << (float)1.0f << (float)1.0f << (void *)identityTranslate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float complexTranslate[] =
{1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 1.0f, 0.0f, 11.0f,
0.0f, 0.0f, 1.0f, -6.5f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("complex")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)2.0f << (float)11.0f << (float)-6.5f << (void *)complexTranslate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float complexTranslate2D[] =
{1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 1.0f, 0.0f, -11.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("complex2D")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)2.0f << (float)-11.0f << (float)0.0f << (void *)complexTranslate2D;
}
void tst_QMatrixNxN::translate4x4()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, x);
QFETCH(float, y);
QFETCH(float, z);
QFETCH(void *, resultValues);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 result((const float *)resultValues);
QMatrix4x4 m1;
m1.translate(QVector3D(x, y, z));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m1, (const float *)resultValues));
QMatrix4x4 m2;
m2.translate(x, y, z);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m2, (const float *)resultValues));
if (z == 0.0f) {
QMatrix4x4 m2b;
m2b.translate(x, y);
QCOMPARE(m2b, m2);
}
QVector3D v1(2.0f, 3.0f, -4.0f);
QVector3D v2 = m1 * v1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v2.x(), (float)(2.0f + x));
QCOMPARE(v2.y(), (float)(3.0f + y));
QCOMPARE(v2.z(), (float)(-4.0f + z));
QVector4D v3(2.0f, 3.0f, -4.0f, 1.0f);
QVector4D v4 = m1 * v3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v4.x(), (float)(2.0f + x));
QCOMPARE(v4.y(), (float)(3.0f + y));
QCOMPARE(v4.z(), (float)(-4.0f + z));
QCOMPARE(v4.w(), (float)1.0f);
QVector4D v5(2.0f, 3.0f, -4.0f, 34.0f);
QVector4D v6 = m1 * v5;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(v6.x(), (float)(2.0f + x * 34.0f));
QCOMPARE(v6.y(), (float)(3.0f + y * 34.0f));
QCOMPARE(v6.z(), (float)(-4.0f + z * 34.0f));
QCOMPARE(v6.w(), (float)34.0f);
QPoint p1(2, 3);
QPoint p2 = m1 * p1;
QCOMPARE(p2.x(), (int)(2.0f + x));
QCOMPARE(p2.y(), (int)(3.0f + y));
QPointF p3(2.0f, 3.0f);
QPointF p4 = m1 * p3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QCOMPARE(p4.x(), (float)(2.0f + x));
QCOMPARE(p4.y(), (float)(3.0f + y));
QMatrix4x4 m3(uniqueValues4);
QMatrix4x4 m4(m3);
m4.translate(x, y, z);
QVERIFY(m4 == m3 * m1);
if (z == 0.0f) {
QMatrix4x4 m4b(m3);
m4b.translate(x, y);
QCOMPARE(m4b, m4);
}
}
// Test the generation and use of 4x4 rotation matrices.
void tst_QMatrixNxN::rotate4x4_data()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("angle");
QTest::addColumn<float>("x");
QTest::addColumn<float>("y");
QTest::addColumn<float>("z");
QTest::addColumn<void *>("resultValues");
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float nullRotate[] =
{0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)0.0f << (float)0.0f << (float)0.0f
<< (void *)nullRotate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float noRotate[] =
{1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("zerodegrees")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)0.0f
<< (float)2.0f << (float)3.0f << (float)-4.0f
<< (void *)noRotate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float xRotate[] =
{1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, -1.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("xrotate")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)1.0f << (float)0.0f << (float)0.0f
<< (void *)xRotate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float xRotateNeg[] =
{1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, -1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("-xrotate")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)-1.0f << (float)0.0f << (float)0.0f
<< (void *)xRotateNeg;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float yRotate[] =
{0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
-1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("yrotate")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)0.0f << (float)1.0f << (float)0.0f
<< (void *)yRotate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float yRotateNeg[] =
{0.0f, 0.0f, -1.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("-yrotate")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)0.0f << (float)-1.0f << (float)0.0f
<< (void *)yRotateNeg;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float zRotate[] =
{0.0f, -1.0f, 0.0f, 0.0f,
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("zrotate")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)0.0f << (float)0.0f << (float)1.0f
<< (void *)zRotate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float zRotateNeg[] =
{0.0f, 1.0f, 0.0f, 0.0f,
-1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("-zrotate")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)90.0f
<< (float)0.0f << (float)0.0f << (float)-1.0f
<< (void *)zRotateNeg;
// Algorithm from http://en.wikipedia.org/wiki/Rotation_matrix.
// Deliberately different from the one in the code for cross-checking.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float complexRotate[16];
float x = 1.0f;
float y = 2.0f;
float z = -6.0f;
float angle = -45.0f;
float c = std::cos(qDegreesToRadians(angle));
float s = std::sin(qDegreesToRadians(angle));
float len = std::sqrt(x * x + y * y + z * z);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float xu = x / len;
float yu = y / len;
float zu = z / len;
complexRotate[0] = (float)((1 - xu * xu) * c + xu * xu);
complexRotate[1] = (float)(-zu * s - xu * yu * c + xu * yu);
complexRotate[2] = (float)(yu * s - xu * zu * c + xu * zu);
complexRotate[3] = 0;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
complexRotate[4] = (float)(zu * s - xu * yu * c + xu * yu);
complexRotate[5] = (float)((1 - yu * yu) * c + yu * yu);
complexRotate[6] = (float)(-xu * s - yu * zu * c + yu * zu);
complexRotate[7] = 0;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
complexRotate[8] = (float)(-yu * s - xu * zu * c + xu * zu);
complexRotate[9] = (float)(xu * s - yu * zu * c + yu * zu);
complexRotate[10] = (float)((1 - zu * zu) * c + zu * zu);
complexRotate[11] = 0;
complexRotate[12] = 0;
complexRotate[13] = 0;
complexRotate[14] = 0;
complexRotate[15] = 1;
QTest::newRow("complex")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)angle
<< (float)x << (float)y << (float)z
<< (void *)complexRotate;
}
void tst_QMatrixNxN::rotate4x4()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, angle);
QFETCH(float, x);
QFETCH(float, y);
QFETCH(float, z);
QFETCH(void *, resultValues);
QMatrix4x4 m1;
m1.rotate(angle, QVector3D(x, y, z));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m1, (const float *)resultValues));
QMatrix4x4 m2;
m2.rotate(angle, x, y, z);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m2, (const float *)resultValues));
QMatrix4x4 m3(uniqueValues4);
QMatrix4x4 m4(m3);
m4.rotate(angle, x, y, z);
QVERIFY(qFuzzyCompare(m4, m3 * m1));
// Null vectors don't make sense for quaternion rotations.
if (x != 0 || y != 0 || z != 0) {
QMatrix4x4 m5;
m5.rotate(QQuaternion::fromAxisAndAngle(QVector3D(x, y, z), angle));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(isSame(m5, (const float *)resultValues));
}
#define ROTATE4(xin,yin,zin,win,xout,yout,zout,wout) \
do { \
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
xout = ((const float *)resultValues)[0] * xin + \
((const float *)resultValues)[1] * yin + \
((const float *)resultValues)[2] * zin + \
((const float *)resultValues)[3] * win; \
yout = ((const float *)resultValues)[4] * xin + \
((const float *)resultValues)[5] * yin + \
((const float *)resultValues)[6] * zin + \
((const float *)resultValues)[7] * win; \
zout = ((const float *)resultValues)[8] * xin + \
((const float *)resultValues)[9] * yin + \
((const float *)resultValues)[10] * zin + \
((const float *)resultValues)[11] * win; \
wout = ((const float *)resultValues)[12] * xin + \
((const float *)resultValues)[13] * yin + \
((const float *)resultValues)[14] * zin + \
((const float *)resultValues)[15] * win; \
} while (0)
// Rotate various test vectors using the straight-forward approach.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float v1x, v1y, v1z, v1w;
ROTATE4(2.0f, 3.0f, -4.0f, 1.0f, v1x, v1y, v1z, v1w);
v1x /= v1w;
v1y /= v1w;
v1z /= v1w;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float v3x, v3y, v3z, v3w;
ROTATE4(2.0f, 3.0f, -4.0f, 1.0f, v3x, v3y, v3z, v3w);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float v5x, v5y, v5z, v5w;
ROTATE4(2.0f, 3.0f, -4.0f, 34.0f, v5x, v5y, v5z, v5w);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float p1x, p1y, p1z, p1w;
ROTATE4(2.0f, 3.0f, 0.0f, 1.0f, p1x, p1y, p1z, p1w);
p1x /= p1w;
p1y /= p1w;
p1z /= p1w;
QVector3D v1(2.0f, 3.0f, -4.0f);
QVector3D v2 = m1 * v1;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(v2.x(), v1x));
QVERIFY(qFuzzyCompare(v2.y(), v1y));
QVERIFY(qFuzzyCompare(v2.z(), v1z));
QVector4D v3(2.0f, 3.0f, -4.0f, 1.0f);
QVector4D v4 = m1 * v3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(v4.x(), v3x));
QVERIFY(qFuzzyCompare(v4.y(), v3y));
QVERIFY(qFuzzyCompare(v4.z(), v3z));
QVERIFY(qFuzzyCompare(v4.w(), v3w));
QVector4D v5(2.0f, 3.0f, -4.0f, 34.0f);
QVector4D v6 = m1 * v5;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(v6.x(), v5x));
QVERIFY(qFuzzyCompare(v6.y(), v5y));
QVERIFY(qFuzzyCompare(v6.z(), v5z));
QVERIFY(qFuzzyCompare(v6.w(), v5w));
QPoint p1(2, 3);
QPoint p2 = m1 * p1;
QCOMPARE(p2.x(), qRound(p1x));
QCOMPARE(p2.y(), qRound(p1y));
QPointF p3(2.0f, 3.0f);
QPointF p4 = m1 * p3;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(p4.x()), p1x));
QVERIFY(qFuzzyCompare(float(p4.y()), p1y));
if (x != 0 || y != 0 || z != 0) {
QQuaternion q = QQuaternion::fromAxisAndAngle(QVector3D(x, y, z), angle);
QVector3D vq = q.rotatedVector(v1);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(vq.x(), v1x));
QVERIFY(qFuzzyCompare(vq.y(), v1y));
QVERIFY(qFuzzyCompare(vq.z(), v1z));
}
}
static bool isSame(const QMatrix3x3& m1, const Matrix3& m2)
{
for (int row = 0; row < 3; ++row) {
for (int col = 0; col < 3; ++col) {
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
if (!qFuzzyCompare(m1(row, col), m2.v[row * 3 + col]))
return false;
}
}
return true;
}
// Test the computation of normal matrices from 4x4 transformation matrices.
void tst_QMatrixNxN::normalMatrix_data()
{
QTest::addColumn<void *>("mValues");
QTest::newRow("identity")
<< (void *)identityValues4;
QTest::newRow("unique")
<< (void *)uniqueValues4; // Not invertible because determinant == 0.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const translateValues[16] =
{1.0f, 0.0f, 0.0f, 4.0f,
0.0f, 1.0f, 0.0f, 5.0f,
0.0f, 0.0f, 1.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const scaleValues[16] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 7.0f, 0.0f, 0.0f,
0.0f, 0.0f, 9.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const bothValues[16] =
{2.0f, 0.0f, 0.0f, 4.0f,
0.0f, 7.0f, 0.0f, 5.0f,
0.0f, 0.0f, 9.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const rotateValues[16] =
{0.0f, 0.0f, 1.0f, 0.0f,
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const nullScaleValues1[16] =
{0.0f, 0.0f, 0.0f, 4.0f,
0.0f, 7.0f, 0.0f, 5.0f,
0.0f, 0.0f, 9.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const nullScaleValues2[16] =
{2.0f, 0.0f, 0.0f, 4.0f,
0.0f, 0.0f, 0.0f, 5.0f,
0.0f, 0.0f, 9.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const nullScaleValues3[16] =
{2.0f, 0.0f, 0.0f, 4.0f,
0.0f, 7.0f, 0.0f, 5.0f,
0.0f, 0.0f, 0.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("translate") << (void *)translateValues;
QTest::newRow("scale") << (void *)scaleValues;
QTest::newRow("both") << (void *)bothValues;
QTest::newRow("rotate") << (void *)rotateValues;
QTest::newRow("null scale 1") << (void *)nullScaleValues1;
QTest::newRow("null scale 2") << (void *)nullScaleValues2;
QTest::newRow("null scale 3") << (void *)nullScaleValues3;
}
void tst_QMatrixNxN::normalMatrix()
{
QFETCH(void *, mValues);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
const float *values = (const float *)mValues;
// Compute the expected answer the long way.
Matrix3 min;
Matrix3 answer;
min.v[0] = values[0];
min.v[1] = values[1];
min.v[2] = values[2];
min.v[3] = values[4];
min.v[4] = values[5];
min.v[5] = values[6];
min.v[6] = values[8];
min.v[7] = values[9];
min.v[8] = values[10];
bool invertible = m3Inverse(min, answer);
m3Transpose(answer);
// Perform the test.
QMatrix4x4 m1(values);
QMatrix3x3 n1 = m1.normalMatrix();
if (invertible)
QVERIFY(::isSame(n1, answer));
else
QVERIFY(isIdentity(n1));
// Perform the test again, after inferring special matrix types.
// This tests the optimized paths in the normalMatrix() function.
m1.optimize();
n1 = m1.normalMatrix();
if (invertible)
QVERIFY(::isSame(n1, answer));
else
QVERIFY(isIdentity(n1));
}
// Test optimized transformations on 4x4 matrices.
void tst_QMatrixNxN::optimizedTransforms()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const translateValues[16] =
{1.0f, 0.0f, 0.0f, 4.0f,
0.0f, 1.0f, 0.0f, 5.0f,
0.0f, 0.0f, 1.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const translateDoubleValues[16] =
{1.0f, 0.0f, 0.0f, 8.0f,
0.0f, 1.0f, 0.0f, 10.0f,
0.0f, 0.0f, 1.0f, -6.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const scaleValues[16] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 7.0f, 0.0f, 0.0f,
0.0f, 0.0f, 9.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const scaleDoubleValues[16] =
{4.0f, 0.0f, 0.0f, 0.0f,
0.0f, 49.0f, 0.0f, 0.0f,
0.0f, 0.0f, 81.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const bothValues[16] =
{2.0f, 0.0f, 0.0f, 4.0f,
0.0f, 7.0f, 0.0f, 5.0f,
0.0f, 0.0f, 9.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const bothReverseValues[16] =
{2.0f, 0.0f, 0.0f, 4.0f * 2.0f,
0.0f, 7.0f, 0.0f, 5.0f * 7.0f,
0.0f, 0.0f, 9.0f, -3.0f * 9.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const bothThenTranslateValues[16] =
{2.0f, 0.0f, 0.0f, 4.0f + 2.0f * 4.0f,
0.0f, 7.0f, 0.0f, 5.0f + 7.0f * 5.0f,
0.0f, 0.0f, 9.0f, -3.0f + 9.0f * -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const bothThenScaleValues[16] =
{4.0f, 0.0f, 0.0f, 4.0f,
0.0f, 49.0f, 0.0f, 5.0f,
0.0f, 0.0f, 81.0f, -3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QMatrix4x4 translate(translateValues);
QMatrix4x4 scale(scaleValues);
QMatrix4x4 both(bothValues);
QMatrix4x4 m1;
m1.translate(4.0f, 5.0f, -3.0f);
QVERIFY(isSame(m1, translateValues));
m1.translate(4.0f, 5.0f, -3.0f);
QVERIFY(isSame(m1, translateDoubleValues));
QMatrix4x4 m2;
m2.translate(QVector3D(4.0f, 5.0f, -3.0f));
QVERIFY(isSame(m2, translateValues));
m2.translate(QVector3D(4.0f, 5.0f, -3.0f));
QVERIFY(isSame(m2, translateDoubleValues));
QMatrix4x4 m3;
m3.scale(2.0f, 7.0f, 9.0f);
QVERIFY(isSame(m3, scaleValues));
m3.scale(2.0f, 7.0f, 9.0f);
QVERIFY(isSame(m3, scaleDoubleValues));
QMatrix4x4 m4;
m4.scale(QVector3D(2.0f, 7.0f, 9.0f));
QVERIFY(isSame(m4, scaleValues));
m4.scale(QVector3D(2.0f, 7.0f, 9.0f));
QVERIFY(isSame(m4, scaleDoubleValues));
QMatrix4x4 m5;
m5.translate(4.0f, 5.0f, -3.0f);
m5.scale(2.0f, 7.0f, 9.0f);
QVERIFY(isSame(m5, bothValues));
m5.translate(4.0f, 5.0f, -3.0f);
QVERIFY(isSame(m5, bothThenTranslateValues));
QMatrix4x4 m6;
m6.translate(QVector3D(4.0f, 5.0f, -3.0f));
m6.scale(QVector3D(2.0f, 7.0f, 9.0f));
QVERIFY(isSame(m6, bothValues));
m6.translate(QVector3D(4.0f, 5.0f, -3.0f));
QVERIFY(isSame(m6, bothThenTranslateValues));
QMatrix4x4 m7;
m7.scale(2.0f, 7.0f, 9.0f);
m7.translate(4.0f, 5.0f, -3.0f);
QVERIFY(isSame(m7, bothReverseValues));
QMatrix4x4 m8;
m8.scale(QVector3D(2.0f, 7.0f, 9.0f));
m8.translate(QVector3D(4.0f, 5.0f, -3.0f));
QVERIFY(isSame(m8, bothReverseValues));
QMatrix4x4 m9;
m9.translate(4.0f, 5.0f, -3.0f);
m9.scale(2.0f, 7.0f, 9.0f);
QVERIFY(isSame(m9, bothValues));
m9.scale(2.0f, 7.0f, 9.0f);
QVERIFY(isSame(m9, bothThenScaleValues));
QMatrix4x4 m10;
m10.translate(QVector3D(4.0f, 5.0f, -3.0f));
m10.scale(QVector3D(2.0f, 7.0f, 9.0f));
QVERIFY(isSame(m10, bothValues));
m10.scale(QVector3D(2.0f, 7.0f, 9.0f));
QVERIFY(isSame(m10, bothThenScaleValues));
}
// Test orthographic projections.
void tst_QMatrixNxN::ortho()
{
QMatrix4x4 m1;
m1.ortho(QRect(0, 0, 300, 150));
QPointF p1 = m1 * QPointF(0, 0);
QPointF p2 = m1 * QPointF(300, 0);
QPointF p3 = m1 * QPointF(0, 150);
QPointF p4 = m1 * QPointF(300, 150);
QVector3D p5 = m1 * QVector3D(300, 150, 1);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(p1.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p1.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p3.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p3.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p4.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p4.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p5.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.z()), -1.0f));
QMatrix4x4 m2;
m2.ortho(QRectF(0, 0, 300, 150));
p1 = m2 * QPointF(0, 0);
p2 = m2 * QPointF(300, 0);
p3 = m2 * QPointF(0, 150);
p4 = m2 * QPointF(300, 150);
p5 = m2 * QVector3D(300, 150, 1);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(p1.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p1.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p3.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p3.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p4.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p4.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p5.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.z()), -1.0f));
QMatrix4x4 m3;
m3.ortho(0, 300, 150, 0, -1, 1);
p1 = m3 * QPointF(0, 0);
p2 = m3 * QPointF(300, 0);
p3 = m3 * QPointF(0, 150);
p4 = m3 * QPointF(300, 150);
p5 = m3 * QVector3D(300, 150, 1);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(p1.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p1.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p3.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p3.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p4.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p4.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p5.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.z()), -1.0f));
QMatrix4x4 m4;
m4.ortho(0, 300, 150, 0, -2, 3);
p1 = m4 * QPointF(0, 0);
p2 = m4 * QPointF(300, 0);
p3 = m4 * QPointF(0, 150);
p4 = m4 * QPointF(300, 150);
p5 = m4 * QVector3D(300, 150, 1);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(p1.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p1.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p2.y()), 1.0f));
QVERIFY(qFuzzyCompare(float(p3.x()), -1.0f));
QVERIFY(qFuzzyCompare(float(p3.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p4.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p4.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.x()), 1.0f));
QVERIFY(qFuzzyCompare(float(p5.y()), -1.0f));
QVERIFY(qFuzzyCompare(float(p5.z()), -0.6f));
// An empty view volume should leave the matrix alone.
QMatrix4x4 m5;
m5.ortho(0, 0, 150, 0, -2, 3);
QVERIFY(m5.isIdentity());
m5.ortho(0, 300, 150, 150, -2, 3);
QVERIFY(m5.isIdentity());
m5.ortho(0, 300, 150, 0, 2, 2);
QVERIFY(m5.isIdentity());
}
// Test perspective frustum projections.
void tst_QMatrixNxN::frustum()
{
QMatrix4x4 m1;
m1.frustum(-1.0f, 1.0f, -1.0f, 1.0f, -1.0f, 1.0f);
QVector3D p1 = m1 * QVector3D(-1.0f, -1.0f, 1.0f);
QVector3D p2 = m1 * QVector3D(1.0f, -1.0f, 1.0f);
QVector3D p3 = m1 * QVector3D(-1.0f, 1.0f, 1.0f);
QVector3D p4 = m1 * QVector3D(1.0f, 1.0f, 1.0f);
QVector3D p5 = m1 * QVector3D(0.0f, 0.0f, 2.0f);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(p1.x(), -1.0f));
QVERIFY(qFuzzyCompare(p1.y(), -1.0f));
QVERIFY(qFuzzyCompare(p1.z(), -1.0f));
QVERIFY(qFuzzyCompare(p2.x(), 1.0f));
QVERIFY(qFuzzyCompare(p2.y(), -1.0f));
QVERIFY(qFuzzyCompare(p2.z(), -1.0f));
QVERIFY(qFuzzyCompare(p3.x(), -1.0f));
QVERIFY(qFuzzyCompare(p3.y(), 1.0f));
QVERIFY(qFuzzyCompare(p3.z(), -1.0f));
QVERIFY(qFuzzyCompare(p4.x(), 1.0f));
QVERIFY(qFuzzyCompare(p4.y(), 1.0f));
QVERIFY(qFuzzyCompare(p4.z(), -1.0f));
QVERIFY(qFuzzyCompare(p5.x(), 0.0f));
QVERIFY(qFuzzyCompare(p5.y(), 0.0f));
QVERIFY(qFuzzyCompare(p5.z(), -0.5f));
// An empty view volume should leave the matrix alone.
QMatrix4x4 m5;
m5.frustum(0, 0, 150, 0, -2, 3);
QVERIFY(m5.isIdentity());
m5.frustum(0, 300, 150, 150, -2, 3);
QVERIFY(m5.isIdentity());
m5.frustum(0, 300, 150, 0, 2, 2);
QVERIFY(m5.isIdentity());
}
// Test perspective field-of-view projections.
void tst_QMatrixNxN::perspective()
{
QMatrix4x4 m1;
m1.perspective(45.0f, 1.0f, -1.0f, 1.0f);
QVector3D p1 = m1 * QVector3D(-1.0f, -1.0f, 1.0f);
QVector3D p2 = m1 * QVector3D(1.0f, -1.0f, 1.0f);
QVector3D p3 = m1 * QVector3D(-1.0f, 1.0f, 1.0f);
QVector3D p4 = m1 * QVector3D(1.0f, 1.0f, 1.0f);
QVector3D p5 = m1 * QVector3D(0.0f, 0.0f, 2.0f);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(p1.x(), 2.41421f));
QVERIFY(qFuzzyCompare(p1.y(), 2.41421f));
QVERIFY(qFuzzyCompare(p1.z(), -1.0f));
QVERIFY(qFuzzyCompare(p2.x(), -2.41421f));
QVERIFY(qFuzzyCompare(p2.y(), 2.41421f));
QVERIFY(qFuzzyCompare(p2.z(), -1.0f));
QVERIFY(qFuzzyCompare(p3.x(), 2.41421f));
QVERIFY(qFuzzyCompare(p3.y(), -2.41421f));
QVERIFY(qFuzzyCompare(p3.z(), -1.0f));
QVERIFY(qFuzzyCompare(p4.x(), -2.41421f));
QVERIFY(qFuzzyCompare(p4.y(), -2.41421f));
QVERIFY(qFuzzyCompare(p4.z(), -1.0f));
QVERIFY(qFuzzyCompare(p5.x(), 0.0f));
QVERIFY(qFuzzyCompare(p5.y(), 0.0f));
QVERIFY(qFuzzyCompare(p5.z(), -0.5f));
// An empty view volume should leave the matrix alone.
QMatrix4x4 m5;
m5.perspective(45.0f, 1.0f, 0.0f, 0.0f);
QVERIFY(m5.isIdentity());
m5.perspective(45.0f, 0.0f, -1.0f, 1.0f);
QVERIFY(m5.isIdentity());
m5.perspective(0.0f, 1.0f, -1.0f, 1.0f);
QVERIFY(m5.isIdentity());
}
// Test viewport transformations
void tst_QMatrixNxN::viewport()
{
// Uses default depth range of 0->1
QMatrix4x4 m1;
m1.viewport(0.0f, 0.0f, 1024.0f, 768.0f);
// Lower left
QVector4D p1 = m1 * QVector4D(-1.0f, -1.0f, 0.0f, 1.0f);
QVERIFY(qFuzzyIsNull(p1.x()));
QVERIFY(qFuzzyIsNull(p1.y()));
QVERIFY(qFuzzyCompare(p1.z(), 0.5f));
// Lower right
QVector4D p2 = m1 * QVector4D(1.0f, -1.0f, 0.0f, 1.0f);
QVERIFY(qFuzzyCompare(p2.x(), 1024.0f));
QVERIFY(qFuzzyIsNull(p2.y()));
// Upper right
QVector4D p3 = m1 * QVector4D(1.0f, 1.0f, 0.0f, 1.0f);
QVERIFY(qFuzzyCompare(p3.x(), 1024.0f));
QVERIFY(qFuzzyCompare(p3.y(), 768.0f));
// Upper left
QVector4D p4 = m1 * QVector4D(-1.0f, 1.0f, 0.0f, 1.0f);
QVERIFY(qFuzzyIsNull(p4.x()));
QVERIFY(qFuzzyCompare(p4.y(), 768.0f));
// Center
QVector4D p5 = m1 * QVector4D(0.0f, 0.0f, 0.0f, 1.0f);
QVERIFY(qFuzzyCompare(p5.x(), 1024.0f / 2.0f));
QVERIFY(qFuzzyCompare(p5.y(), 768.0f / 2.0f));
}
// Test left-handed vs right-handed coordinate flipping.
void tst_QMatrixNxN::flipCoordinates()
{
QMatrix4x4 m1;
m1.flipCoordinates();
QVector3D p1 = m1 * QVector3D(2, 3, 4);
QVERIFY(p1 == QVector3D(2, -3, -4));
QMatrix4x4 m2;
m2.scale(2.0f, 3.0f, 1.0f);
m2.flipCoordinates();
QVector3D p2 = m2 * QVector3D(2, 3, 4);
QVERIFY(p2 == QVector3D(4, -9, -4));
QMatrix4x4 m3;
m3.translate(2.0f, 3.0f, 1.0f);
m3.flipCoordinates();
QVector3D p3 = m3 * QVector3D(2, 3, 4);
QVERIFY(p3 == QVector3D(4, 0, -3));
QMatrix4x4 m4;
m4.rotate(90.0f, 0.0f, 0.0f, 1.0f);
m4.flipCoordinates();
QVector3D p4 = m4 * QVector3D(2, 3, 4);
QVERIFY(p4 == QVector3D(3, 2, -4));
}
// Test conversion of generic matrices to and from the non-generic types.
void tst_QMatrixNxN::convertGeneric()
{
QMatrix4x3 m1(uniqueValues4x3);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const unique4x4[16] = {
1.0f, 2.0f, 3.0f, 4.0f,
5.0f, 6.0f, 7.0f, 8.0f,
9.0f, 10.0f, 11.0f, 12.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QMatrix4x4 m4(m1);
QVERIFY(isSame(m4, unique4x4));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float const conv4x4[12] = {
1.0f, 2.0f, 3.0f, 4.0f,
5.0f, 6.0f, 7.0f, 8.0f,
9.0f, 10.0f, 11.0f, 12.0f
};
QMatrix4x4 m9(uniqueValues4);
QMatrix4x3 m10 = m9.toGenericMatrix<4, 3>();
QVERIFY(isSame(m10, conv4x4));
}
// Copy of "flagBits" in qmatrix4x4.h.
enum {
Identity = 0x0000, // Identity matrix
Translation = 0x0001, // Contains a translation
Scale = 0x0002, // Contains a scale
Rotation2D = 0x0004, // Contains a rotation about the Z axis
Rotation = 0x0008, // Contains an arbitrary rotation
Perspective = 0x0010, // Last row is different from (0, 0, 0, 1)
General = 0x001f // General matrix, unknown contents
};
// Structure that allows direct access to "flagBits" for testing.
struct Matrix4x4
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float m[4][4];
int flagBits;
};
// Test the inferring of special matrix types.
void tst_QMatrixNxN::optimize_data()
{
QTest::addColumn<void *>("mValues");
QTest::addColumn<int>("flagBits");
QTest::newRow("null")
<< (void *)nullValues4 << (int)General;
QTest::newRow("identity")
<< (void *)identityValues4 << (int)Identity;
QTest::newRow("unique")
<< (void *)uniqueValues4 << (int)General;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float scaleValues[16] = {
2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 3.0f, 0.0f, 0.0f,
0.0f, 0.0f, 4.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("scale")
<< (void *)scaleValues << (int)Scale;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float translateValues[16] = {
1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 1.0f, 0.0f, 3.0f,
0.0f, 0.0f, 1.0f, 4.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("translate")
<< (void *)translateValues << (int)Translation;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float scaleTranslateValues[16] = {
1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 2.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 4.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("scaleTranslate")
<< (void *)scaleTranslateValues << (int)(Scale | Translation);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float rotateValues[16] = {
0.0f, 1.0f, 0.0f, 0.0f,
-1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("rotate")
<< (void *)rotateValues << (int)Rotation2D;
// Left-handed system, not a simple rotation.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float scaleRotateValues[16] = {
0.0f, 1.0f, 0.0f, 0.0f,
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("scaleRotate")
<< (void *)scaleRotateValues << (int)(Scale | Rotation2D);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float matrix2x2Values[16] = {
1.0f, 2.0f, 0.0f, 0.0f,
8.0f, 3.0f, 0.0f, 0.0f,
0.0f, 0.0f, 9.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("matrix2x2")
<< (void *)matrix2x2Values << (int)(Scale | Rotation2D);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float matrix3x3Values[16] = {
1.0f, 2.0f, 4.0f, 0.0f,
8.0f, 3.0f, 5.0f, 0.0f,
6.0f, 7.0f, 9.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("matrix3x3")
<< (void *)matrix3x3Values << (int)(Scale | Rotation2D | Rotation);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float rotateTranslateValues[16] = {
0.0f, 1.0f, 0.0f, 1.0f,
-1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 0.0f, 1.0f, 3.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("rotateTranslate")
<< (void *)rotateTranslateValues << (int)(Translation | Rotation2D);
// Left-handed system, not a simple rotation.
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float scaleRotateTranslateValues[16] = {
0.0f, 1.0f, 0.0f, 1.0f,
1.0f, 0.0f, 0.0f, 2.0f,
0.0f, 0.0f, 1.0f, 3.0f,
0.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("scaleRotateTranslate")
<< (void *)scaleRotateTranslateValues << (int)(Translation | Scale | Rotation2D);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static float belowValues[16] = {
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
4.0f, 0.0f, 0.0f, 1.0f
};
QTest::newRow("below")
<< (void *)belowValues << (int)General;
}
void tst_QMatrixNxN::optimize()
{
QFETCH(void *, mValues);
QFETCH(int, flagBits);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m((const float *)mValues);
m.optimize();
QCOMPARE(reinterpret_cast<Matrix4x4 *>(&m)->flagBits, flagBits);
}
void tst_QMatrixNxN::columnsAndRows()
{
QMatrix4x4 m1(uniqueValues4);
QVERIFY(m1.column(0) == QVector4D(1, 5, 9, 13));
QVERIFY(m1.column(1) == QVector4D(2, 6, 10, 14));
QVERIFY(m1.column(2) == QVector4D(3, 7, 11, 15));
QVERIFY(m1.column(3) == QVector4D(4, 8, 12, 16));
QVERIFY(m1.row(0) == QVector4D(1, 2, 3, 4));
QVERIFY(m1.row(1) == QVector4D(5, 6, 7, 8));
QVERIFY(m1.row(2) == QVector4D(9, 10, 11, 12));
QVERIFY(m1.row(3) == QVector4D(13, 14, 15, 16));
m1.setColumn(0, QVector4D(-1, -5, -9, -13));
m1.setColumn(1, QVector4D(-2, -6, -10, -14));
m1.setColumn(2, QVector4D(-3, -7, -11, -15));
m1.setColumn(3, QVector4D(-4, -8, -12, -16));
QVERIFY(m1.column(0) == QVector4D(-1, -5, -9, -13));
QVERIFY(m1.column(1) == QVector4D(-2, -6, -10, -14));
QVERIFY(m1.column(2) == QVector4D(-3, -7, -11, -15));
QVERIFY(m1.column(3) == QVector4D(-4, -8, -12, -16));
QVERIFY(m1.row(0) == QVector4D(-1, -2, -3, -4));
QVERIFY(m1.row(1) == QVector4D(-5, -6, -7, -8));
QVERIFY(m1.row(2) == QVector4D(-9, -10, -11, -12));
QVERIFY(m1.row(3) == QVector4D(-13, -14, -15, -16));
m1.setRow(0, QVector4D(1, 5, 9, 13));
m1.setRow(1, QVector4D(2, 6, 10, 14));
m1.setRow(2, QVector4D(3, 7, 11, 15));
m1.setRow(3, QVector4D(4, 8, 12, 16));
QVERIFY(m1.column(0) == QVector4D(1, 2, 3, 4));
QVERIFY(m1.column(1) == QVector4D(5, 6, 7, 8));
QVERIFY(m1.column(2) == QVector4D(9, 10, 11, 12));
QVERIFY(m1.column(3) == QVector4D(13, 14, 15, 16));
QVERIFY(m1.row(0) == QVector4D(1, 5, 9, 13));
QVERIFY(m1.row(1) == QVector4D(2, 6, 10, 14));
QVERIFY(m1.row(2) == QVector4D(3, 7, 11, 15));
QVERIFY(m1.row(3) == QVector4D(4, 8, 12, 16));
}
// Test converting QTransform objects into QMatrix4x4 and then
// checking that transformations in the original perform the
// equivalent transformations in the new matrix.
void tst_QMatrixNxN::convertQTransform()
{
QTransform m1;
m1.translate(-3.5, 2.0);
QPointF p1 = m1.map(QPointF(100.0, 150.0));
QCOMPARE(p1.x(), 100.0 - 3.5);
QCOMPARE(p1.y(), 150.0 + 2.0);
QMatrix4x4 m2(m1);
QPointF p2 = m2 * QPointF(100.0, 150.0);
QCOMPARE((double)p2.x(), 100.0 - 3.5);
QCOMPARE((double)p2.y(), 150.0 + 2.0);
QCOMPARE(m1, m2.toTransform());
QTransform m3;
m3.scale(1.5, -2.0);
QPointF p3 = m3.map(QPointF(100.0, 150.0));
QCOMPARE(p3.x(), 1.5 * 100.0);
QCOMPARE(p3.y(), -2.0 * 150.0);
QMatrix4x4 m4(m3);
QPointF p4 = m4 * QPointF(100.0, 150.0);
QCOMPARE((double)p4.x(), 1.5 * 100.0);
QCOMPARE((double)p4.y(), -2.0 * 150.0);
QCOMPARE(m3, m4.toTransform());
QTransform m5;
m5.rotate(45.0);
QPointF p5 = m5.map(QPointF(100.0, 150.0));
QMatrix4x4 m6(m5);
QPointF p6 = m6 * QPointF(100.0, 150.0);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(p5.x()), float(p6.x())));
QVERIFY(qFuzzyCompare(float(p5.y()), float(p6.y())));
QTransform m7 = m6.toTransform();
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(m5.m11()), float(m7.m11())));
QVERIFY(qFuzzyCompare(float(m5.m12()), float(m7.m12())));
QVERIFY(qFuzzyCompare(float(m5.m21()), float(m7.m21())));
QVERIFY(qFuzzyCompare(float(m5.m22()), float(m7.m22())));
QVERIFY(qFuzzyCompare(float(m5.dx()), float(m7.dx())));
QVERIFY(qFuzzyCompare(float(m5.dy()), float(m7.dy())));
QVERIFY(qFuzzyCompare(float(m5.m13()), float(m7.m13())));
QVERIFY(qFuzzyCompare(float(m5.m23()), float(m7.m23())));
QVERIFY(qFuzzyCompare(float(m5.m33()), float(m7.m33())));
}
// Test filling matrices with specific values.
void tst_QMatrixNxN::fill()
{
QMatrix4x4 m1;
m1.fill(0.0f);
QVERIFY(isSame(m1, nullValues4));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float fillValues4[] =
{2.5f, 2.5f, 2.5f, 2.5f,
2.5f, 2.5f, 2.5f, 2.5f,
2.5f, 2.5f, 2.5f, 2.5f,
2.5f, 2.5f, 2.5f, 2.5f};
m1.fill(2.5f);
QVERIFY(isSame(m1, fillValues4));
QMatrix4x3 m2;
m2.fill(0.0f);
QVERIFY(isSame(m2, nullValues4x3));
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float fillValues4x3[] =
{2.5f, 2.5f, 2.5f, 2.5f,
2.5f, 2.5f, 2.5f, 2.5f,
2.5f, 2.5f, 2.5f, 2.5f};
m2.fill(2.5f);
QVERIFY(isSame(m2, fillValues4x3));
}
// Test the mapRect() function for QRect and QRectF.
void tst_QMatrixNxN::mapRect_data()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QTest::addColumn<float>("x");
QTest::addColumn<float>("y");
QTest::addColumn<float>("width");
QTest::addColumn<float>("height");
QTest::newRow("null")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)0.0f << (float)0.0f << (float)0.0f << (float)0.0f;
QTest::newRow("rect")
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
<< (float)1.0f << (float)-20.5f << (float)100.0f << (float)63.75f;
}
void tst_QMatrixNxN::mapRect()
{
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QFETCH(float, x);
QFETCH(float, y);
QFETCH(float, width);
QFETCH(float, height);
QRectF rect(x, y, width, height);
QRect recti(qRound(x), qRound(y), qRound(width), qRound(height));
QMatrix4x4 m1;
QCOMPARE(m1.mapRect(rect), rect);
QCOMPARE(m1.mapRect(recti), recti);
QMatrix4x4 m2;
m2.translate(-100.5f, 64.0f);
QRectF translated = rect.translated(-100.5f, 64.0f);
QRect translatedi = QRect(qRound(recti.x() - 100.5f), recti.y() + 64,
recti.width(), recti.height());
QCOMPARE(m2.mapRect(rect), translated);
QCOMPARE(m2.mapRect(recti), translatedi);
QMatrix4x4 m3;
m3.scale(-100.5f, 64.0f);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float scalex = x * -100.5f;
float scaley = y * 64.0f;
float scalewid = width * -100.5f;
float scaleht = height * 64.0f;
if (scalewid < 0.0f) {
scalewid = -scalewid;
scalex -= scalewid;
}
if (scaleht < 0.0f) {
scaleht = -scaleht;
scaley -= scaleht;
}
QRectF scaled(scalex, scaley, scalewid, scaleht);
QCOMPARE(m3.mapRect(rect), scaled);
scalex = recti.x() * -100.5f;
scaley = recti.y() * 64.0f;
scalewid = recti.width() * -100.5f;
scaleht = recti.height() * 64.0f;
if (scalewid < 0.0f) {
scalewid = -scalewid;
scalex -= scalewid;
}
if (scaleht < 0.0f) {
scaleht = -scaleht;
scaley -= scaleht;
}
QRect scaledi(qRound(scalex), qRound(scaley),
qRound(scalewid), qRound(scaleht));
QCOMPARE(m3.mapRect(recti), scaledi);
QMatrix4x4 m4;
m4.translate(-100.5f, 64.0f);
m4.scale(-2.5f, 4.0f);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
float transx1 = x * -2.5f - 100.5f;
float transy1 = y * 4.0f + 64.0f;
float transx2 = (x + width) * -2.5f - 100.5f;
float transy2 = (y + height) * 4.0f + 64.0f;
if (transx1 > transx2)
qSwap(transx1, transx2);
if (transy1 > transy2)
qSwap(transy1, transy2);
QRectF trans(transx1, transy1, transx2 - transx1, transy2 - transy1);
QCOMPARE(m4.mapRect(rect), trans);
transx1 = recti.x() * -2.5f - 100.5f;
transy1 = recti.y() * 4.0f + 64.0f;
transx2 = (recti.x() + recti.width()) * -2.5f - 100.5f;
transy2 = (recti.y() + recti.height()) * 4.0f + 64.0f;
if (transx1 > transx2)
qSwap(transx1, transx2);
if (transy1 > transy2)
qSwap(transy1, transy2);
QRect transi(qRound(transx1), qRound(transy1),
qRound(transx2) - qRound(transx1),
qRound(transy2) - qRound(transy1));
QCOMPARE(m4.mapRect(recti), transi);
m4.rotate(45.0f, 0.0f, 0.0f, 1.0f);
QTransform t4;
t4.translate(-100.5f, 64.0f);
t4.scale(-2.5f, 4.0f);
t4.rotate(45.0f);
QRectF mr = m4.mapRect(rect);
QRectF tr = t4.mapRect(rect);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(float(mr.x()), float(tr.x())));
QVERIFY(qFuzzyCompare(float(mr.y()), float(tr.y())));
QVERIFY(qFuzzyCompare(float(mr.width()), float(tr.width())));
QVERIFY(qFuzzyCompare(float(mr.height()), float(tr.height())));
QRect mri = m4.mapRect(recti);
QRect tri = t4.mapRect(recti);
QCOMPARE(mri, tri);
}
void tst_QMatrixNxN::mapVector_data()
{
QTest::addColumn<void *>("mValues");
QTest::newRow("null")
<< (void *)nullValues4;
QTest::newRow("identity")
<< (void *)identityValues4;
QTest::newRow("unique")
<< (void *)uniqueValues4;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float scale[] =
{2.0f, 0.0f, 0.0f, 0.0f,
0.0f, 11.0f, 0.0f, 0.0f,
0.0f, 0.0f, -6.5f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("scale")
<< (void *)scale;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float scaleTranslate[] =
{2.0f, 0.0f, 0.0f, 1.0f,
0.0f, 11.0f, 0.0f, 2.0f,
0.0f, 0.0f, -6.5f, 3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("scaleTranslate")
<< (void *)scaleTranslate;
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
static const float translate[] =
{1.0f, 0.0f, 0.0f, 1.0f,
0.0f, 1.0f, 0.0f, 2.0f,
0.0f, 0.0f, 1.0f, 3.0f,
0.0f, 0.0f, 0.0f, 1.0f};
QTest::newRow("translate")
<< (void *)translate;
}
void tst_QMatrixNxN::mapVector()
{
QFETCH(void *, mValues);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QMatrix4x4 m1((const float *)mValues);
QVector3D v(3.5f, -1.0f, 2.5f);
QVector3D expected
(v.x() * m1(0, 0) + v.y() * m1(0, 1) + v.z() * m1(0, 2),
v.x() * m1(1, 0) + v.y() * m1(1, 1) + v.z() * m1(1, 2),
v.x() * m1(2, 0) + v.y() * m1(2, 1) + v.z() * m1(2, 2));
QVector3D actual = m1.mapVector(v);
m1.optimize();
QVector3D actual2 = m1.mapVector(v);
Make gui/math3d classes use float rather than qreal This corrects the mismatch between using floats for internal storage and qreal in the API of QVector*D which leads to lots of implicit casts between double and float. This change also stops users from being surprised by the loss of precision when using these classes on desktop platforms and removes the need for the private constructors taking a dummy int as the final argument. The QMatrix4x4 and QQuaternion classes have been changed to use float for their internal storage since these are meant to be used in conjunction with the QVector*D classes. This is to prevent unexpected loss of precision and to improve performance. The on-disk format has also been changed from double to float thereby reducing the storage required when streaming vectors and matrices. This is potentially a large saving when working with complex 3D meshes etc. This also has a significant performance improvement when passing matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no longer have to iterate and convert the data to floats. This is an operation that could easily be needed many times per frame. This change also opens the door for further optimisations of these classes to be implemented by using SIMD intrinsics. This needs to be applied in conjunction with https://codereview.qt-project.org/#change,33548 Task-number: QTBUG-21035 Task-number: QTBUG-20661 Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-08-20 19:55:40 +00:00
QVERIFY(qFuzzyCompare(actual.x(), expected.x()));
QVERIFY(qFuzzyCompare(actual.y(), expected.y()));
QVERIFY(qFuzzyCompare(actual.z(), expected.z()));
QVERIFY(qFuzzyCompare(actual2.x(), expected.x()));
QVERIFY(qFuzzyCompare(actual2.y(), expected.y()));
QVERIFY(qFuzzyCompare(actual2.z(), expected.z()));
}
class tst_QMatrixNxN4x4Properties : public QObject
{
Q_OBJECT
Q_PROPERTY(QMatrix4x4 matrix READ matrix WRITE setMatrix)
public:
tst_QMatrixNxN4x4Properties(QObject *parent = nullptr) : QObject(parent) {}
QMatrix4x4 matrix() const { return m; }
void setMatrix(const QMatrix4x4& value) { m = value; }
private:
QMatrix4x4 m;
};
// Test getting and setting matrix properties via the metaobject system.
void tst_QMatrixNxN::properties()
{
tst_QMatrixNxN4x4Properties obj;
QMatrix4x4 m1(uniqueValues4);
obj.setMatrix(m1);
QMatrix4x4 m2 = qvariant_cast<QMatrix4x4>(obj.property("matrix"));
QVERIFY(isSame(m2, uniqueValues4));
QMatrix4x4 m3(transposedValues4);
obj.setProperty("matrix", QVariant::fromValue(m3));
m2 = qvariant_cast<QMatrix4x4>(obj.property("matrix"));
QVERIFY(isSame(m2, transposedValues4));
}
void tst_QMatrixNxN::metaTypes()
{
QCOMPARE(QMetaType::type("QMatrix4x4"), int(QMetaType::QMatrix4x4));
QCOMPARE(QByteArray(QMetaType::typeName(QMetaType::QMatrix4x4)),
QByteArray("QMatrix4x4"));
QVERIFY(QMetaType::isRegistered(QMetaType::QMatrix4x4));
QCOMPARE(qMetaTypeId<QMatrix4x4>(), int(QMetaType::QMatrix4x4));
}
QTEST_APPLESS_MAIN(tst_QMatrixNxN)
#include "tst_qmatrixnxn.moc"