Question

glScalef(0.5f, 0.5f, 0);
// do some things
glScalef(2.0f, 2.0f, 0);

Or..

glPushMatrix();
glScalef(0.5f, 0.5f, 0);
// do some things
glPopMatrix();

Which is better, performance wise?

Answer 1

The second one. Because the first one computes and multiplies a 4×4 matrix, while the second only writes and reads it. Note that OpenGL can be intelligent about it and only create/multiply a diagonal matrix. Nevertheless, you are comparing 3 floating point multiplications with 16 variable stores. If you are concerned about the difference in performance, you need to write a test.

Precision-wise, the second one is definitely better because due to whatever rounding error 0.5 * x * 2 might give a different value than x. This is why, even if the other method is slightly faster, this method is preferred.