Modify QMatrix4x4 and QQuaternion to use qreal internally

Some concerns were expressed about the float precision of QMatrix4x4,
which this change addresses by using qreal instead.

The QVector2D/3D/4D classes still use float internally, so that they
can be used directly in large arrays of vertex values to be uploaded
to an OpenGL server.

QQuaternion is a client-side class, and it should produce rotations
that are consistent with QMatrix4x4.  So its precision was changed too.

A consequence of this change is that the following no longer works
in a portable fashion:

    QMatrix4x4 mat;
    ...
    glLoadMatrixf(mat.constData());

The caller must now repack the argument to convert from qreal to GLfloat.

Reviewed-by: Michael Goddard
Reviewed-by: Andreas

1 files changed, 6 insertions, 6 deletions

diff --git a/tests/benchmarks/qmatrix4x4/tst_qmatrix4x4.cpp b/tests/benchmarks/qmatrix4x4/tst_qmatrix4x4.cpp
index b4c8202..c72bc27 100644
--- a/tests/benchmarks/qmatrix4x4/tst_qmatrix4x4.cpp
+++ b/tests/benchmarks/qmatrix4x4/tst_qmatrix4x4.cpp
@@ -171,9 +171,9 @@ void tst_QMatrix4x4::multiplyDirect()
 
     QMatrix4x4 m3;
 
-    const float *m1data = m1.constData();
-    const float *m2data = m2.constData();
-    float *m3data = m3.data();
+    const qreal *m1data = m1.constData();
+    const qreal *m2data = m2.constData();
+    qreal *m3data = m3.data();
 
     QBENCHMARK {
         for (int row = 0; row < 4; ++row) {
@@ -266,9 +266,9 @@ void tst_QMatrix4x4::mapVectorDirect()
 {
     QFETCH(QMatrix4x4, m1);
 
-    const float *m1data = m1.constData();
-    float v[4] = {10.5f, -2.0f, 3.0f, 1.0f};
-    float result[4];
+    const qreal *m1data = m1.constData();
+    qreal v[4] = {10.5f, -2.0f, 3.0f, 1.0f};
+    qreal result[4];
 
     QBENCHMARK {
         for (int row = 0; row < 4; ++row) {