mirror of
https://github.com/Druthulu/BFM-decomp
synced 2026-09-30 07:31:59 -04:00
ee957d11c6
- M4's 8 fns byte-correct in ISOLATION but drift 8-69 in-TU (codegen scheduling/volatile-loss, not data-type casts; 4 have no data conflict) -> reconcile_decls banks 0/8. R14 correction: the T6 'mechanical M4' projection was a 3rd object-probe over-count (after jumptable + linkwall). The 8 re-tier to F-band (permuter/§31). cookbook §41b-addendum + decision-log R31 entry. - T7 truly-mechanical tier EXHAUSTED: M1 37 + M3-clean 2 = 39 exemplars swept ×134 = ~4,694 fleet fns, fleet 72.29 -> 73.66%, R22 136/136 (committed commit:0495/commit:0496/commit:0497). CHECKPOINT for a fresh session; phase OPEN; NEXT = the F-band frontier (~39 fns) + jumptable/linkwall specialist workflows