mirror of
https://github.com/Druthulu/BFM-decomp
synced 2026-10-05 00:47:54 -04:00
Drew T
72e6a096ca
CAUGHT BY VERIFYING THE CHECKPOINT INSTEAD OF ASSERTING IT. `.run/wave22/` was covered by the blanket `/.run/*` ignore, so the 13 UNBANKED drafts — 6 NEAR with precise residual diagnoses and 7 that reached match_one MATCH but did not bank whole-binary — existed only on disk. My own checkpoint's "START HERE" list names them as next-session fuel, and they cost ~2.59M subagent tokens to produce. One `git clean -fdx` would have destroyed them: exactly the exposure the .run/giants Fable5 cracks had before Phase 27 curated them (R20). Now allowlisted: .run/wave22/*.c + .run/wave22_targets.json (the per-target canonical callee/data declarations resolved from the real TU scope — the §17a-1 lever the wave was built on). Also commits docs/family-hseq.md, regenerated by this session's family_hseq runs.
52 lines
1.9 KiB
C
52 lines
1.9 KiB
C
/* canonical (sig_hints) — return type widened s32 so the `sw $v0` after the jal has a
|
|
source; the hint's `void` cannot express the store. */
|
|
extern s32 func_80163664(s32, u16, u16, s32, s32, s32, s32, s32, s32, s32, u16, s32, s32);
|
|
|
|
/* canonical (sig_hints) */
|
|
extern s32 D_80115100;
|
|
extern s32 D_80115200;
|
|
extern u16 D_80126B18[];
|
|
|
|
/* derived from the asm: separate lui/%lo per symbol => distinct externs */
|
|
extern s32 D_80115204;
|
|
extern s32 D_80115208;
|
|
extern u16 D_801270B0[];
|
|
extern u16 D_801270B2;
|
|
extern u16 D_801270B4;
|
|
/* LOAD-BEARING: D_80126B1A must be declared/stored as an ARRAY, not a scalar.
|
|
gcc-2.7.2 true_dependence() lets an in-struct MEM with an unstable (register)
|
|
address bypass a not-in-struct MEM with a stable (symbol) address. As a scalar,
|
|
the `sh %lo(D_80126B1A)` store does NOT conflict with `a5[2]`, so the block-6
|
|
`lhu $t0,4($t0)` gets hoisted into the D_80126B1A block's load-delay slot and the
|
|
whole filler queue shifts one slot early (75 ins, no `nop`). ARRAY_REF sets
|
|
MEM_IN_STRUCT_P on the store => the two MEMs conflict => the load stays put. */
|
|
extern u16 D_80126B1A[];
|
|
extern u16 D_80126B1C;
|
|
extern s32 D_80114EB0;
|
|
extern s32 D_80114EC8;
|
|
extern s32 D_8011DAF0;
|
|
extern s32 D_80115298;
|
|
extern s32 D_80126734;
|
|
|
|
void func_80163534(s32 a0, u16 a1, u16 a2, s32 a3, u16 a4, u16 *a5)
|
|
{
|
|
s32 *p = &D_80115200;
|
|
|
|
*p = 0;
|
|
D_80115204 = 0;
|
|
D_80115208 = 0;
|
|
|
|
D_801270B0[0] = *(u16 *)(a0 + 0x44) + a5[0];
|
|
D_801270B2 = *(u16 *)(a0 + 0x46) + a5[1];
|
|
D_801270B4 = *(u16 *)(a0 + 0x48) + a5[2];
|
|
|
|
D_80126B18[0] = *(u16 *)(a0 + 0x6) + a5[0];
|
|
D_80126B1A[0] = *(u16 *)(a0 + 0xA) + a5[1];
|
|
D_80126B1C = *(u16 *)(a0 + 0xE) + a5[2];
|
|
|
|
*p = func_80163664(a0, a1, a2, (s32)D_801270B0, (s32)D_80126B18,
|
|
(s32)&D_80114EB0, (s32)&D_80114EC8, (s32)&D_80115100,
|
|
(s32)&D_8011DAF0, a3, a4, (s32)&D_80115298,
|
|
(s32)&D_80126734);
|
|
}
|