`nx_u384_mul.nx` source

↩ module page · 110 lines · 3122 B
1// nx_u384_mul.nx -- 384x384 -> 768-bit schoolbook multiplication +
2// wide-buffer helpers (alloc / cmp / fits_in_384 / copy_low).
3//
4// Mirror of nx_u256_mul.nx with 24-limb wide buffers in place of
5// 16-limb.  Foundation for nx_p384_field_mul.
6//
7// API:
8//   u384_wide_alloc()            -> *i64 24-limb buffer (zeroed)
9//   u384_mul_wide(out_24, a, b)  out_24 = a * b (no reduction)
10//   u384_wide_cmp(a_24, b_24)    -> -1 | 0 | +1
11//   u384_wide_fits_in_384(a_24)  -> 0 | 1
12//   u384_wide_copy_low(out_12, src_24)
13//
14// license_tier: INDEPENDENT_REDERIVE
15// genealogy_id: international-research-sources/nist/fips_186_5
16// lineage_id:   nishi_u384_mul_q10
17
18// nx_safety_envelope:
19//   intended_use:       AUTO_APPLIED -- primitive-specific tuning queued
20//   sil_target:         SIL1
21//   evidence:           [bulk_applied_2026-05-20, u384-wide-mul]
22//   verdict:            NOT_YET_EVALUATED
23
24import "nx_syscalls.nx"
25import "nx_u384.nx"
26
27const NX_U384_WIDE_LIMBS: i64 = 24
28const NX_U384_WIDE_BYTES: i64 = 96
29
30func u384_wide_alloc() -> *i64 {
31    let b: *u8 = sys_mmap(NX_U384_WIDE_LIMBS * 8)
32    let p: *i64 = b as *i64
33    var i: i64 = 0
34    while i < NX_U384_WIDE_LIMBS {
35        p[i] = 0
36        i = i + 1
37    }
38    return p
39}
40
41// Free a u384_wide_alloc buffer (the 768-bit product scratch).  Paired
42// with u384_free so p384_field_mul_fast leaves no per-call leak.
43func u384_wide_free(p: *i64) -> i64 {
44    if (p as i64) == 0 { return 0 }
45    sys_munmap(p as *u8, NX_U384_WIDE_LIMBS * 8)
46    return 0
47}
48
49// out_24 = a * b (384x384 -> 768).  No reduction.
50// Aliasing: out_24 MUST be a distinct buffer from a and b.
51func u384_mul_wide(out_24: *i64, a: *i64, b: *i64) -> i64 {
52    var k: i64 = 0
53    while k < NX_U384_WIDE_LIMBS {
54        out_24[k] = 0
55        k = k + 1
56    }
57    var i: i64 = 0
58    while i < NX_U384_LIMBS {
59        let ai: i64 = a[i] & NX_U384_LIMB_MASK
60        var carry: i64 = 0
61        var j: i64 = 0
62        while j < NX_U384_LIMBS {
63            let bj: i64 = b[j] & NX_U384_LIMB_MASK
64            let prev: i64 = out_24[i + j] & NX_U384_LIMB_MASK
65            let s: i64 = prev + ai * bj + carry
66            out_24[i + j] = s & NX_U384_LIMB_MASK
67            carry = (s >> NX_U384_LIMB_BITS) & NX_U384_LIMB_MASK
68            j = j + 1
69        }
70        out_24[i + 12] = carry
71        i = i + 1
72    }
73    return 0
74}
75
76func u384_wide_cmp(a_24: *i64, b_24: *i64) -> i64 {
77    var i: i64 = NX_U384_WIDE_LIMBS - 1
78    while i >= 0 {
79        let av: i64 = a_24[i] & NX_U384_LIMB_MASK
80        let bv: i64 = b_24[i] & NX_U384_LIMB_MASK
81        if av < bv { return 0 - 1 }
82        if av > bv { return 1 }
83        i = i - 1
84    }
85    return 0
86}
87
88func u384_wide_fits_in_384(a_24: *i64) -> i64 {
89    var i: i64 = 12
90    var acc: i64 = 0
91    while i < NX_U384_WIDE_LIMBS {
92        acc = acc | (a_24[i] & NX_U384_LIMB_MASK)
93        i = i + 1
94    }
95    if acc == 0 { return 1 }
96    return 0
97}
98
99func u384_wide_copy_low(out_12: *i64, src_24: *i64) -> i64 {
100    var i: i64 = 0
101    while i < NX_U384_LIMBS {
102        out_12[i] = src_24[i]
103        i = i + 1
104    }
105    return 0
106}
107
108func main() -> i64 {
109    return 0
110}
nx_u384_mul.nx source

`nx_u384_mul.nx` source
