glRotate/Translate/Scale vs. custom func's using the matrices

It seems to me that it would be faster to write the matrices for each of these functions myself, and then use glLoadMatrix to perform the desired actions.

Anyone done this and have some real feedback?

Thanks,

I calculate the matrix for rotation and use glLoadMatrix but then I use gltranslate and I never use scale. Certainly, I could easily add the translation into my matrix and avoid glTranslate. Probably save a few counts. But, I’ve been lazy of late.