Native bit complexity of unweighted matching #
The certificates use the complete framed encGraph input and encMatching output. Retained
vertex/name bits, explicit one-hot table rows, scope positions, every intermediate solver
state and the output edge list enter ordinary native multitape polynomial composition.
No graph validity, degree bound, finite relation language or runtime provider is an input.
For framed input length L, there are at most L distinct vertices and normalized edges.
Writing these counts as n and m, each choice table has at most m + 1 rows of width
m + 1, and each pin has one unary row. Thus choice-row storage is bounded by
n * (m + 1) ^ 2 Boolean entries, before the explicitly charged framing and binary scope
names. Tagging a retained vertex name by 2 * v + 1 adds at most one binary digit;
original edge names are tagged list positions. Solver storage growth is charged by the
all-raw explicit-table optimizer certificate used in the composition.
The nonempty choice scope is built in polynomial time, retaining binary names.
The complete generated explicit-table instance is native polynomial-time on every graph.
Reading mutual selection scans only actual occurrence records and their bit buffer.
The selected original position list is native polynomial-time.
Converting positions to actual original endpoint pairs is native polynomial-time.
The graph-only maximum matching endpoint has an ordinary native polynomial certificate.
Checking full vertex coverage scans the actual lists of vertices and selected edges.
Constructive perfect-matching search is native polynomial-time on complete raw graphs.
Full perfect-matching decision is native polynomial-time on complete raw graphs.