Watching Optimizations Happen

Optimizations have signatures

• Constant folding: a function computing 21 * 2 with locals becomes a load of an immediate (or return 42 inline) — no multiply instruction remains.
• Dead-code elimination: code whose result is never used disappears entirely.
• Loop-invariant code motion: a computation that does not depend on the loop variable is hoisted before the loop — one evaluation instead of n.
• Strength reduction: multiplications by powers of two become shifts.

Each has a visible footprint: instruction counts, removed loads, hoisted lines. That is why this course grounds optimization talk in comparing dumps, not folklore.

The honest measuring loop

Count something concrete: instructions in the dump, branches taken in a benchmark harness, bytes of the .text section (module 11's size). 'Faster' claims without a counter are vibes. The challenges below give you the counters.