An N × N grid of multiply-accumulate cells holds a weight tile still and pumps activations through it, one row per cycle, with no register file or instruction fetch in the inner loop. The cycle count, the fill and drain, and the formula that says why a 128-row tile wastes two thirds of the array.
Updated Sep 2026 · Grounded in real AI infrastructure interview loops and written to a senior-engineer editorial bar, with every number worked and every diagram hand-built.
An N × N grid of multiply-accumulate cells holds a weight tile still and pumps activations through it, one row per cycle, with no register file or instruction fetch in the inner loop. The cycle count, the fill and drain, and the formula that says why a 128-row tile wastes two thirds of the array.
Ranked by how closely each one overlaps this question's topic, so the first card is the thing to read if the answer above moved too fast.
Scored on a correct data-flow picture (stationary weights, activations shifting one cell per cycle, partial sums accumulating down the column), the cycle-count derivation, and the efficiency formula as a function of tile size.
No comments yet — be the first to share your approach.
