Review week 16: branches and tolerances

This week was about jumps without labels and comparisons with leeway. For review:

• Relative branches (breq … brge, jr) jump by a distance: 1 is the next line, 2 skips one. Every line counts, comments too. A label as a distance is almost always wrong.
• Against zero: breqz … brgez and brnan. Build the deviation first, then the sign decides. On NaN only brnan jumps.
• With devices: brdse and brdns for short fallback lines. Over whole sections a label is better.
• Approximately equal: sap, sna, bap, bna, brap, brna with relative tolerance c × max(|a|, |b|). On NaN sap and sna both return 0 in the game.
• Approximately zero: the z forms practically only take effect at exactly 0. For “almost zero” an absolute limit with abs and slt. epsilon is not the tolerance of the ap instructions.

You pass from 60 %, the explanations appear after handing in.