lemmaroutine
Multiplication with linear clock-gap storage
Multiplication keeps suspended storage despite long expansion
multiplication primitive for reciprocal and interface
Lemma·
lem:arith-multiplication·§10.3 Multiplication by minimum-index bandsStatement (paper)
For normalized exact input streams and , multiplication has all
the conclusions of Lemma . In particular,
its suspended storage at an operand place is
, despite the potentially long product expansion.
In plain English
For normalized exact input streams and , the product stream has all the properties of the elementary arithmetic operations. In particular, the storage suspended at an operand place is , even though the product expansion is long.
Why it's here
Multiplication is the one primitive whose naive expansion has far too many term pairs, so it needs separate treatment. The proof groups term pairs into bands by minimum index, truncates the tail with error under half a unit, and evaluates each band as a small-residue adder tree. The band path cost is then charged to the clock gap , so the depth-first evaluation lemma gives the linear storage bound that later results (reciprocal, the uniform interface) rely on.
Classified routine:
The proof is a technical adaptation of known techniques. It groups term pairs by minimum index and truncates the tail, which is standard bookkeeping for multiplying digit streams. It then evaluates the bands with the paper's depth-first, path-only circuit evaluation lemma and charges the cost to the clock gap.
Uses
Used by
Inside the proof
Sources
PaperOpen
- Section
- §10.3 Multiplication by minimum-index bands
- Page
- —
- Label
lem:arith-multiplication- Line
sections/arithmetic.tex:253
Leannot aligned
Phase 2 aligns this node to a Lean declaration.