code wiki / _hdl_build / nx_autoopt_library_test.nx
nx_autoopt_library_test.nx
buildroot/runtime/_hdl_build/nx_autoopt_library_test.nx
about
nx_autoopt_library_test.nx -- the team builds its OWN optimization library, autonomously.
No hand-authored answers: it sweeps a family of kernels it chose (x*C, C=2..13),
and for EACH the GENERATOR (superopt) authors the shortest program, the ENGINEER
(execution) verifies it 1:1 on held-out inputs, and the verified-optimal machine code is
BANKED into a persistent library. The team also KNOWS its own coverage: a kernel it can't
reach within its search depth is recorded as a GAP (self-knowledge, not a failure), which
is exactly the signal for self-directed escalation. Who wrote the answers? The team did.
Gate: every kernel it CLAIMS to have solved is verified by execution -> exit 0.
dependencies 3 imports · 0 importers
imports: nx_superopt_emit.nxnx_mulchain.nxnx_engineer_crash.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| none |
functions
| 14 | func al_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 } |
| 15 | func al_num(v: i64) -> i64 |
| 24 | func tally(r: *i64, n: i64) -> i64 { var ec: i64 = 0; var i: i64 = 0; while i < n { if r[i] != 1 { if ec == 0 { ec = i + 1 } } i = i + 1 } return ec } called by 1: main |
| 25 | func al_write(path: *u8, buf: *u8, len: i64) -> i64 { let fd: i64 = sys_openat_wr(path, 0x1a4); if fd < 0 { return 0 - 1 } sys_write(fd, buf, len); sys_close(fd); return 0 } |
| 28 | func opt_one(c: i64, res: *i64) -> i64 |
| 61 | func autorun(lo: i64, hi: i64, mat: *i64, r: *i64) -> i64 |
| 82 | func report(mat: *i64, nk: i64) -> i64 |
| 109 | func main() -> i64 |