`nx_dr_ocm.nx` source

↩ module page · 101 lines · 3723 B
1// nx_dr_ocm.nx -- SOVEREIGN Opportunity-Cost Model ranker for research vectors (DR-6).
2// The operator's "integrate opportunity cost analysis" made a live capability: rank
3// candidate research rungs by Priority = (Value * Momentum) / Cost, report the
4// OPPORTUNITY COST of each pick (the forgone best alternative), and do budget-aware
5// greedy selection (FutureWeaver arXiv:2512.11213: score-cheap-before-spending
6// g = geomean(self-consistency, prior); only spend within the compute budget = h).
7// Composes the frontier ranking law (impact x maturity-gap x momentum) with an
8// explicit COST denominator. Integer, deterministic, bit-exact reproducible; imports
9// ONLY nx_syscalls = drift-immune. No hardware writes (Rule 26).
10//
11// module: nishi-core.research.dr_ocm
12// depends: nx_syscalls.nx
13// wired_status: LIBRARY  (conductor dispatch = DR-4; cost inputs = nx_cost_decompose)
14// genealogy_id: opportunity_cost_econ + futureweaver_2026_budget_planner
15import "nx_syscalls.nx"
16
17// integer sqrt (Newton) -- for the FutureWeaver cheap pre-score g = geomean.
18func ocm_isqrt(n: i64) -> i64 {
19    if n < 2 { return n }
20    var x: i64 = n
21    var y: i64 = (x + 1) / 2
22    while y < x { x = y; y = (x + n / x) / 2 }
23    return x
24}
25
26// Priority = (value * momentum * 1000) / cost. Guard cost<1 -> 1 (a free rung is
27// maximally preferred). Higher V, higher M, LOWER C -> higher priority.
28func ocm_priority(value: i64, momentum: i64, cost: i64) -> i64 {
29    var c: i64 = cost
30    if c < 1 { c = 1 }
31    return (value * momentum * 1000) / c
32}
33
34// FutureWeaver cheap pre-score g(alpha) = geomean(self-consistency, prior).
35func ocm_gscore(selfcons: i64, prior: i64) -> i64 {
36    return ocm_isqrt(selfcons * prior)
37}
38
39// pr[i] = ocm_priority(v[i],m[i],c[i]).
40func ocm_fill_priorities(v: *i64, m: *i64, c: *i64, n: i64, pr: *i64) -> i64 {
41    var i: i64 = 0
42    while i < n { pr[i] = ocm_priority(v[i], m[i], c[i]); i = i + 1 }
43    return 0
44}
45
46// Rank ids by priority DESC (ties -> lower id). out_order[n].
47func ocm_rank(pr: *i64, n: i64, out_order: *i64) -> i64 {
48    let used: *u8 = sys_mmap(n)
49    var i: i64 = 0
50    while i < n { used[i] = 0 as u8; i = i + 1 }
51    var r: i64 = 0
52    while r < n {
53        var best: i64 = 0 - 1
54        var j: i64 = 0
55        while j < n {
56            if used[j] == (0 as u8) {
57                if best < 0 { best = j }
58                else { if pr[j] > pr[best] { best = j } }
59            }
60            j = j + 1
61        }
62        out_order[r] = best
63        used[best] = 1 as u8
64        r = r + 1
65    }
66    return 0
67}
68
69// Opportunity cost per rank position: oc[r] = priority of the NEXT-best deferred
70// vector (the value forgone by taking order[r] now). Last position -> 0.
71func ocm_opp_cost(pr: *i64, order: *i64, n: i64, oc: *i64) -> i64 {
72    var r: i64 = 0
73    while r < n {
74        if r + 1 < n { oc[r] = pr[order[r + 1]] } else { oc[r] = 0 }
75        r = r + 1
76    }
77    return 0
78}
79
80// Budget-aware greedy selection in priority order (FutureWeaver budget-feasibility h):
81// pick a vector iff cumulative cost + its cost <= budget. sel[id]=1 if picked.
82// Returns count selected; out_spent[0] = total cost of the selected set. Greedy by
83// priority (NOT optimal knapsack -- declared, not silently optimal).
84func ocm_select(pr: *i64, c: *i64, order: *i64, n: i64, budget: i64, sel: *i64, out_spent: *i64) -> i64 {
85    var i: i64 = 0
86    while i < n { sel[i] = 0; i = i + 1 }
87    var spent: i64 = 0
88    var cnt: i64 = 0
89    var r: i64 = 0
90    while r < n {
91        let id: i64 = order[r]
92        if (spent + c[id]) <= budget {
93            sel[id] = 1
94            spent = spent + c[id]
95            cnt = cnt + 1
96        }
97        r = r + 1
98    }
99    out_spent[0] = spent
100    return cnt
101}
nx_dr_ocm.nx source

`nx_dr_ocm.nx` source
