Here's a more readable version:
sin(A) and sin(B) were being computed repeatedly in the most inner of inner loops, which I pulled out, because A and B don't change in the middle. sin(j) and cos(j) was being computed for every i, even though j wasn't changing. And the intensity was being computed even if the pixel was going to appear beyond the buffer (or wasn't going to appear at all because of a closer pixel in the z buffer).
I also refactored parts of it, and gave names to some constants which you can tweak for fun. DELTA_I can be increased without apparent detriment to image quality. I can't think of good names for the two constants in the middle of the code though: 2 and 5. Are those the minor and major radius of the donut?
IIRC, yes, 5 and 2 are the major and minor radii, meaning 5 is the radius of the ring at the center of the donut (a torus of thickness 0, as it were) and 2 is the thickness -- the radius of the circle extruded around the central ring.