You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
malvarezcastillo
changed the title
[WIP] Misc matches
Decompile hsd_8039930C particle bytecode interpreter (93.95%)
Feb 11, 2026
malvarezcastillo
changed the title
Decompile hsd_8039930C particle bytecode interpreter (93.95%)
Decompile all particle.c functions: 61/119 at 100%, 0 stubs remaining
Feb 12, 2026
malvarezcastillo
changed the title
Decompile all particle.c functions: 61/119 at 100%, 0 stubs remaining
particle.c: decompile 35 stubs, push 5 to 100%, improve 52 functions
Feb 12, 2026
The fn_80392CD8 code is byte-identical to the target — all 85 instructions match. The 99.02% comes from .data section offset differences in the .o file.
Root cause: On master, fn_80392CD8's string literals (the MCC error messages) were the first data in the .data section (offset 0x000). The compiler loaded the base via lis/addi and used mr r5, r31 (offset 0) for the first string. This PR decompiles functions that appear before fn_80392CD8 in the source (fn_80392A3C, DrawASCII, hsd_803921B8, etc.) — their string literals and jump tables now precede fn_80392CD8's data in .data, pushing the error strings to offset 0x84.
The compiler then generates addi r5, r31, 0x84 instead of mr r5, r31 for the first switch case — one different opcode. All other switch case offsets shift by the same 0x84 delta.
Why it's unfixable: mwcc places .data items in source order. The only fix would be either (a) reverting the new function implementations or (b) decompiling the entire particle.c TU so the full .data layout matches the target. Neither is practical. The regression will resolve naturally as more of the TU gets decompiled.
@ribbanya I rebased this branch and your commit (619eb5c "Remove unused label and format code") got dropped in the process because it conflicted extensively with the match improvement commits that were being developed concurrently. To make up for it, I've re-applied the formatting changes in a new commit (886871c4c):
Ran git clang-format on particle.c
Reordered includes alphabetically (dolphin/pad.h before dolphin/vi.h, MSL/trigf.h moved after MetroTRK/ppc_reg.h)
Applied the same header line-wrapping fixes to particle.h
The loop_end: label you removed was already absent in our version. The sphere_random: and sphere_common: labels are still referenced by goto statements so they were kept.
Sorry for the force push — the formatting commit touched 1666 lines which made rebasing impractical.
Update: particle.c analysis complete — all remaining non-100% functions are at their ceiling
I systematically analyzed all 57 non-100% functions in particle.c by comparing target vs compiled assembly instruction-by-instruction. Every single one has byte-identical instructions — the match percentage gaps are entirely due to relocation metadata differences that cannot be fixed in C source.
The two relocation patterns causing the gaps:
.bss.0 symbol naming: mwcc uses section-relative BSS relocations (.bss.0) for same-TU definitions, while the target uses named symbols (e.g., hsd_804CF810). This affects most functions that reference BSS globals.
.data string literal ordering: String literals appear in a different order in the compiled .data section vs target, causing relocation offset mismatches for string references. This is a file-level ordering issue, not per-function.
Bottom line: No further C source improvements are possible for particle.c. The 91.45% .text match represents 100% instruction-level accuracy. All remaining percentage gaps are tooling artifacts from objdiff's relocation scoring.
Final stats vs master:
7 NEW 100% matches
51 improved functions
.text section: 22.57% → 91.45%
+1788 bytes matched code
Also pushed review fixes: normalized TODO comment format, updated all stale match percentages, and added explanatory comments.
Decompile all remaining stubbed functions in particle.c including the
large hsd_8039930C particle bytecode interpreter. Fix BSS variable
sizes (psstructs.h array bounds 64->65), improve match percentages
across 52 functions, and push 7 functions to 100% match.
All remaining non-100% functions have byte-identical instructions;
match percentage gaps are purely relocation metadata differences
(.bss.0 symbol naming and .data string ordering).
.text section: 22.57% -> 91.45% (+1788 bytes matched)
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
malvarezcastillo
changed the title
No description provided.