[DrLegitamate]

Six Joba (level 11) functions matched from their retained near misses, plus measured notes for the ones that stay assembly.

Matched

Function	Bytes	What closed it
func_L11_00434580	1,144	Two control-flow fixes in the draft: a hit with a true predicate only removes the object (retail skips func_L00_003294F0), and a stopped projectile in frames 25-33 returns without expiring. Each arm writes out its own hit tail, which jump2 cross-jumps after allocation (object stays in $s2). func_002C2210's last two parameters are declared float-then-int (same $t2/$f14; other callers already use that interleaving). The distance call is func_00282E80, not func_00282E48.
func_L11_00472E10	896	The eight-child loop as a plain index loop over the state words swaps the sentinel and state into retail's $s4/$s3.
func_L11_00461308	1,236	The random bounds are one two-float block at D_L11_001B2388: the upper is read as D_L11_001B2388_block[1] (existing _block sda alias pattern), a two-instruction absolute load that stays out of the call's delay slot; the lower keeps its $gp view. Forming the linked moby's pointer before reading its state byte fixes the predicate's registers.
func_L11_003DCEB0	436	The state flag takes the switch variable only once the frame count reaches 0x79, giving retail's movn; the count is incremented in place and re-read (unmasked store, masked compare). The effect table uses the char D_L11_001A8F20[] __attribute__((sda)) declaration most callers already have, so its address is one macro load and the 0x810001 stack argument takes $a2 (diagnosed from the allocation dump). Prototypes added for three helpers it called undeclared.
func_L11_00437FD0	628	The direction flag and pass count carry next-state values set in each arm and copied back at the end of each pass, reproducing retail's $a1/$v1 values with stores after each definition and reloads before the shared copy and the pass test (the allocation dump showed a single pseudo crossing 14 calls could never be caller-saved). The lengths are indexed by the loop counter, length is block-local and next is formed after the floor clamp. -O2 -G0 -g in its own unit (-g supplies the line-label nop before div.s).
func_L11_00438880	1,108	Both flag scans as while (next < 13) { int j = next - 1; if (!flags[j]) break; next++; }. Written as flags[next-1], the front end folds the address to (flags+next)-1 and loop.c finds the address giv not worth reducing (0 vs 7 in the loop dump); the separate j keeps the next-1 giv, so the summed benefit makes loop.c strength-reduce it into retail's pointer cursor, including the pre-loop test through the computed address and the copy into the cursor.

Each lives in its own unit; the nonmatching/ copies are retired, and the two functions that had port-only C in nonmatching/functional.tsv lose those rows and files (as in 588f8f0).

Notes only (still assembly)

• func_L11_00452450: all 956 bytes match if func_002B60B0 is declared void in this caller. Not used, because the callee returns its pointer and COMPILER-NOTES forbids an invented void return; most existing callers in src/ already declare it void. A maintainer ruling would settle it.
• func_L11_00443D78: negative result appended to its notes (inline conversions do not serialize).
• func_L11_00443940: improved body and findings in its notes. With the MenuRecord/SDATA table shapes from the matched sibling and -fno-rerun-loop-opt, the selection loop matches retail; 20 bytes remain, mainly retail keeping each label pointer in its own register before the four draw calls.

Validation

• tools/check.py: each matched function 1 of 1 USA v1.01 place. PAL's catalogue has none of them; US2 not checked (no US2 dump on this machine).
• tools/build.py all --strict: 28 programs OK, zero C fallbacks (e.g. 11_joba: OK: 2363208 loaded bytes equal).
• python3 -m unittest discover -s tests: 339 tests OK.
• tests/check_repository.py: repository checks passed.
• Compiler rebuilt from this master before measuring. Progress report not regenerated.

🤖 Generated with Claude Code

https://claude.ai/code/session_017f7s8MnhdrdEynhYhxkLs6

[bytematchdev]

ByteMatch progress report

SCES_516.07

Code matched	54.15% · 3.09 MB of 5.70 MB
Fuzzy match	54.15%
Functions	6,996 / 8,672
Data matched	0.00%
Units complete	958 / 1,196

No function changes.

SCUS_972.68_v2

Code matched	81.01% · 4.52 MB of 5.58 MB
Fuzzy match	81.01%
Functions	7,595 / 8,634
Data matched	0.00%
Units complete	973 / 1,156

No function changes.

SCUS_972.68

Code matched	84.74% · 4.68 MB of 5.53 MB
Fuzzy match	84.74%
Functions	7,760 / 8,579
Data matched	0.00%
Units complete	924 / 1,018

No function changes.

_{Commit 910165a · Ratchet & Clank: Going Commando on ByteMatch}