The repository root carried seven markdown files that were not onboarding
material, plus two that had been superseded by the code they described. This
sorts them: reader-facing docs to docs/, working notes to docs/internal/, and
the two superseded ones merged into the docs that replaced them.
Moved to docs/: WIDESCREEN.md, ENHANCEMENTS.md, TCP_COMMANDS.md
Moved to docs/internal/: PRINCIPLES.md, ISSUES.md, DEBUG.md, GTE_LM_FIX.md,
and from docs/: bios-hardening.md, overlay-status.md,
overlay-plan.md, overlay-recompilation-design.md,
overlay-discovery.md, TOMBA2_EXC_CONSULT_RESPONSE.md,
CASE_A_AOT_GAP.md, audit_inventory.md
Moved to docs/internal/upstream/: UPSTREAM_2026-08-20.md
The root keeps only what belongs there: README.md, CONTRIBUTING.md, LICENSE,
CLAUDE.md, THIRD_PARTY_ATTRIBUTION.md, RELEASE_NOTES.md, the dotfiles and
psxrecomp_cli.py.
MERGES
SPLITGEN_MIGRATION.md folded into docs/BUILDING.md as "Split generated C". Its
migration instructions were stale in a way that mattered: it told every game
repo to hand-edit its CMakeLists and add a file(GLOB ...) over
<serial>_full*.c, but runtime/runtime.cmake now does that itself. GEN_FULL_GLOB
is a multi-value argument of psxrecomp_add_game_runtime(); the loop at
runtime.cmake:1985 makes each pattern absolute against CMAKE_CURRENT_SOURCE_DIR,
expands it with file(GLOB), and concatenates the hits into
GAME_GENERATED_FULL_C. So a title names the pattern and gets however many shards
a regeneration produced. The rewritten section describes that, the two-step
fallback when the globs match nothing (GEN_FULL_FALLBACK, else <serial>_full.c
derived from the dispatch marker by substituting _dispatch.c), and the one
caveat the code actually has: that file(GLOB) is plain, not CONFIGURE_DEPENDS,
so a regeneration which changes the shard COUNT needs a re-configure to be seen.
psxrecomp-toml-readme.md folded into docs/GAME_PROJECT_SETUP.md as "EXE-only
autofill (psxrecomp-toml)", placed after "Disc autofill (probe_disc.py)" and
before "Project Studio" so the two autofill paths sit together. Its documented
flags were checked against the usage text in recompiler/src/main_toml.cpp:28-42
and against the argument parser at :150-165 -- all seven options match with no
drift, so the flag list is reproduced as-is. Added a pointer that probe_disc.py
is preferable when a cue exists, since that tool does the same EXE-header work
plus disc identity and netplay gates.
FUNCTIONAL FIXES
tools/gen_tcp_commands.py wrote to REPO/"TCP_COMMANDS.md". After the move that
path does not exist, so the generator would have created a new stray root file
and --check would have failed against it forever. Retargeted to
REPO/"docs"/"TCP_COMMANDS.md", along with the seven docstring and message
strings that name the path.
Running it surfaced two further problems, both fixed here:
* The doc was genuinely stale -- pre-existing drift, not introduced by this
commit (the moved file is byte-identical to origin/master, and no
debug-server command table is touched here). Four commands registered since
the last regeneration were missing from the index -- cdrom_timing_dump,
savestate_status, ws_ui_groups, xprobe_watch -- and scanline had gained a
backend. Regenerated: 307 -> 311 commands, and the change is confined to the
AUTOGEN block (the hand-written head and tail are byte-identical to master).
tools/gen_tcp_commands.py --check now exits 0.
* DOC.write_text() has no newline argument, so Python translated every newline
to os.linesep. On Windows that rewrote all 585 lines of the file to CRLF and
turned a four-line index update into a whole-file diff -- measured: 589 CRLF
/ 0 LF after a run, against 0 CRLF / 585 LF in the repo. Now written through
an explicit newline="" handle, verified to leave the file at 0 CRLF.
runtime/include/debug_server.h:9 and runtime/src/debug_server.c:8 cited "TCP.md
and DEBUG.md". TCP.md has never existed in this repo under that name; both now
name docs/TCP_COMMANDS.md and docs/internal/DEBUG.md.
LINK AND REFERENCE FIXES
Every inbound reference to a moved file was located by grepping each basename
across the whole tracked tree, not just the markdown. Markdown hrefs in
README.md, CONTRIBUTING.md, docs/ARCHITECTURE.md, docs/COMPILING_OVERLAYS.md,
docs/AOT_OVERLAY_PLAN.md, docs/RELOCATION_MANIFEST_FORMAT.md,
docs/config_schema.md, docs/OVERLAY_XREF.md, docs/TCP_COMMANDS.md and several
docs/internal/ files now point at the new locations; where two files became
siblings under docs/internal/ the path was shortened rather than lengthened
(bios-hardening.md's reference to overlay-discovery.md, for instance).
docs/COMPILING_OVERLAYS.md's "in this folder" phrasing was wrong once two of the
three files it listed moved, so it now names internal/ explicitly.
24 source-comment mentions of ENHANCEMENTS.md sections G1.x across 19 files
(recompiler, runtime, cmake) were retargeted to docs/ENHANCEMENTS.md with the
section numbers untouched. docs/SHADOW_ENHANCEMENTS.md is a DIFFERENT file and
was deliberately excluded.
20 rationale strings citing docs/CASE_A_AOT_GAP.md across the four
recompiler/seeds/*.json files were updated. Done as a byte-level substring
replacement rather than a JSON round-trip: two of those files are stored with
CRLF endings and two with LF, and re-serializing them reformatted 5,700 lines.
The byte-level version changes exactly 20 lines and each file still parses.
References that look similar but point elsewhere were deliberately left alone,
and are worth naming so a future pass does not "fix" them:
* "recomp-template PRINCIPLES.md" / "NES recomp template PRINCIPLES.md" in
CLAUDE.md:43, docs/SHADOW_ENHANCEMENTS.md:28,
docs/internal/HLE_SCHEDULER_CARVEOUT_PLAN.md:5,
runtime/src/debug_server.c:4934 and tools/parity_diff.py:8 are a different
repository's file. Only the five that mean this repo's PRINCIPLES.md were
updated.
* Every remaining bare ISSUES.md mention is a GAME repo's tracker
(MegaManX6Recomp 2/7/8/9, TombaRecomp 6), not this repo's.
* docs/disc-speed.md:4,54,56 and docs/internal/bios-hardening.md:5 name
overlay-discovery as a BRANCH, not a file, so no path applies.
INDEXES
README.md's "New here?" table gains rows for docs/WIDESCREEN.md,
docs/ENHANCEMENTS.md and docs/TCP_COMMANDS.md. docs/internal/README.md gains
sections indexing all fourteen documents that moved in, plus an "Upstream
contributor reviews" section describing upstream/ and the new sweep file, plus a
"Removed" bullet for CYCLE_TIMING_ARCH.md (deleted in the previous commit)
pointing at FAITHFUL_TIMING_PLAN.md, CYCLE_MODEL_BEETLE.md and
PRECISE_IRQ_SLICE.md -- following the SLJIT_PERSIST_CACHE.md precedent already
in that file.
docs/ENHANCEMENTS.md:7 claimed framework bugs live in "game repos' ISSUES.md
until a framework tracker exists". A framework tracker does exist; it now says
so and names the Beads epic.
VERIFICATION
A link checker was written to scan every tracked *.md for relative links,
resolving each against the tree being scanned. Baseline at origin/master: 129
files, 137 relative links, 3 broken. After both commits: 127 files, 159 relative
links, 3 broken -- the SAME three, all pre-existing and untouched here:
docs/DATA_SHARDS.md -> INSTANT_LOADS_PARKED.md (never existed)
docs/DATA_SHARDS.md -> INSTANT_LOADS_SPIKE2.md (never existed)
docs/NETPLAY.md -> ../lib/recomp-net/README.md (inside a submodule)
Zero new broken links. The checker's own coverage was checked against a raw grep
for markdown link openers: the 3-link difference is C++ lambdas inside code
blocks, correctly not treated as links.
Also verified: py_compile over all 113 tracked tools/*.py plus psxrecomp_cli.py;
all four seed JSONs still parse; the test-registration guard still reports 123
test files all registered; recompiler/tests/test_coverage_vault.py still passes;
and no file in this commit had its line endings flipped (checked mechanically
across all 50 modified files, after two files were caught and redone).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
7.0 KiB
Overlay Discovery
Updated: 2026-05-30 Phase gate: Phase 5+ (game EXE loaded, dirty-RAM hits observed in game address space)
Problem
PSX games load code from disc at runtime into RAM. The main EXE on disc does not contain these overlays — they arrive via CD-ROM DMA during gameplay. The static recompiler sees only the main EXE at build time; it cannot know "at address 0x80090000, after area 2 loads, there will be function X" without actually running the game.
The dirty-RAM interpreter handles these hits today, but it is slow and has no coverage guarantee. This doc describes a self-improving two-layer system that eventually eliminates all runtime interpretation of overlay code.
Distinction from BIOS dirty-RAM
BIOS dirty-RAM (Rule 18 in CLAUDE.md) is code the BIOS assembles in-place at boot — e.g. the 4-instruction SIO handler written to RAM 0xCF0. That code never existed on disc.
Game overlays are different: they are compiled MIPS modules loaded to fixed RAM addresses. Their bytes are known after they land in RAM; the recompiler just hasn't seen them yet. This system fixes that.
Missed heuristics (static-EXE functions the recompiler failed to identify) are orthogonal: they exist in the main EXE at build time. Fix missed heuristics by improving heuristics; fix overlays by this system.
Why disc extraction is not used
The naive approach — scan the disc image for overlay bytes by CRC — does not work for Tomba and
many other PSX titles. Investigation confirmed that Tomba stores overlays in a scatter-load
format: each disc sector is a (12-byte block descriptor, 2048-byte data block) pair, where
the descriptor tells the loader which RAM address to write the data block to. The resulting RAM
content is not present anywhere on disc as a contiguous byte range, so no CRC or brute-force scan
can find it.
The universal correct primitive is to capture bytes at DMA time — intercepting them as they land in RAM, after all scatter-reassembly and decompression. This works for:
- Scatter-load formats (Tomba, others)
- Game-specific LZ or RLE compression (bytes arrive decompressed via DMA)
- Direct file loads (trivially captured)
Architecture
Two layers, composing. As Layer B grows, Layer A fires less. Eventually Layer A goes silent.
Layer B is a manifest of Layer A's output: the same (load_addr, bytes, seeds) data, committed
to the repo and baked into the binary ahead of time.
Layer B — Build-time pre-compilation
Runtime side (capture): on every CD DMA completion into game-code RAM, the runtime appends
(load_addr, size, bytes) to a ring buffer. On process exit it writes
overlay_captures.json automatically — no user action required. The user discovers the file
exists and can attach it to a GitHub issue.
Build-time side: psxrecomp-game reads game.toml [[overlays]] entries, each pointing to
a captured .bin file and a list of seed entry points. It disassembles the MIPS bytes at
load_addr, emits C functions, and adds them to the static dispatch table. The shipped binary
has those overlays pre-compiled at zero runtime cost.
Coverage grows with every release as developer playthroughs and user contributions feed new captures into the build.
Layer A — Runtime background compilation with persistent cache
On CD DMA completion into game-code RAM:
- Hash the written bytes.
- Check
cache/<SCUS_ID>/<hash>.dll.- Cache hit →
LoadLibrarythe DLL, register its functions in the live dispatch table for[load_addr, load_addr + size). Interpreter never fires for this overlay. - Cache miss → fall through to dirty-RAM interpretation. Enqueue a background thread to
compile this overlay: emit C for the bytes at
load_addr, invoke the C compiler, write the DLL to cache. On the next load of this overlay the cache hit fires.
- Cache hit →
Result: within a single session, each new overlay is interpreted once and compiled once. Every subsequent encounter — same session or future sessions — dispatches to native code.
A user who plays through the entire game accumulates a complete cache. From that point forward the game runs entirely as compiled native code, never touching the interpreter.
How they compose
Dispatch (branch to overlay address):
1. Static dispatch table [Layer B] → precompiled function zero runtime cost
2. Compiled cache [Layer A] → LoadLibrary'd function one-time per session after first
3. Dirty-RAM interpreter → fallback first-ever encounter only
Layer B coverage grows with every rebuild. Layer A covers the gap within a session before the rebuild happens. Dirty-RAM fires exactly once per undiscovered overlay.
Game identification
Cache and capture files are namespaced by SCUS/SLUS/SCES product code, read from SYSTEM.CNF:
cache/
SCUS-94236/ ← Tomba (NTSC-U)
<hash>.dll
logs/
SCUS-94236/
overlay_map.jsonl
overlay_captures.json ← written at exit, next to the executable
Cache invalidation
A cached DLL is keyed by the hash of the overlay bytes. The hash IS the identity. If a disc dump differs between users or regions, the hash differs and the old cache entry is simply never matched. No explicit invalidation needed.
When Layer B statically compiles an overlay, Layer A's cached DLL for that hash becomes unreachable (static dispatch wins at step 1). It can be pruned post-build but is benign if left.
Implementation phases
| Phase | Work | Status |
|---|---|---|
| — | Dirty-RAM interpreter (Layer A last-resort fallback) | Done |
| B-1 | CD DMA byte capture ring (always-on, retail-safe, dest < 0x1C0000 filter) |
To build |
| B-1 | Automatic overlay_captures.json written at process exit |
To build |
| B-2 | game.toml [[overlays]] + psxrecomp-game overlay compilation |
To build |
| B-3 | Contribution ingestion script: JSON → .bin + game.toml entries |
To build |
| A-1 | On DMA completion, check cache DLL; if present, LoadLibrary + register |
To build |
| A-2 | Background thread: emit C, compile to DLL, write to cache | To build |
| A-3 | On next overlay encounter, A-1 fires instead of dirty-RAM | Follows from A-1/A-2 |
B-1 through B-3 ship before A-1. The system is useful with only Layer B (requires a rebuild after each new playthrough contribution). Layer A eliminates the "rebuild required" friction and makes the game self-healing for any user.
Scatter-load discovery (Tomba)
Tomba's CD DMA log (captured 2026-05-30) confirmed the scatter-load format:
- Every sector pair:
12-byte block descriptor(always to a header buffer at0x973C0) +2048-byte data block(scattered to the address the descriptor specifies). - A single ISO file load produces DMA entries with non-contiguous destination addresses.
- The 12-byte descriptor encodes the block's destination; the loader dispatches accordingly.
This format is game-specific but the DMA capture approach handles it transparently: the captured bytes are the post-scatter RAM contents, which is exactly the execution form the recompiler needs.