At -O2, CorrelatedValuePropagation and SimplifyCFG fold a translated
shader's repeated tests of one value into switches whose cases jump past
their own join. Apple's GPU backend then runs out of control-flow IDs
("Max Unstruct ID 758 ... exceeds supported limit", a note: metal-tt
still exits 0) and lowers the function a far slower way. Odyssey's
alRenderMaterialUber took 24.6 minutes and 17.3 MB of applegpu_g13s code;
at -O1 it takes 31 s and 1.3 MB.
Sampled archive bases from seven games compile to code within a few
percent of -O2. Odyssey's full g13s archive builds in 15.6 minutes
(exit 0), its Cap Kingdom frames match -O2 within 3/255 on scattered
pixels, and GPU time per frame is within run-to-run noise.
Document the mechanism, and correct the Odyssey note: the 758 line never
failed the build; the reporter's exit 1 came from the BAHS link errors
fixed in ff44c59.
Refs #276
Subsystem references
How the host meets each part of a Switch game, and every failure met so far, one entry per mechanism: symptom, cause, fix (file and function), the games it was found on (linked to their notes in dm/games/<id>/README.md). Start from the symptom index when you have a log line or a visible result; the workflows that lead here are in docs/agents.
| File | Covers |
|---|---|
| packing.md | NSP script, compressed inputs and build artifacts |
| rewriter.md | The NSO rewriter patch_nso.py: jump tables, exclusive windows, patch area reach, cuts and observation gates, functions without unwind entries, stack and system registers, relocations |
| memory.md | Guest layout, heap and application pool, virtual address memory and the alias region (the SDK's node area), engine allocators, mapping SVCs, running out of memory |
| threads.md | Thread creation and cores, address arbitration, yields and spin waits, events |
| services.md | SDK calls and system services: sm connections, names that changed between SDKs, file system and saves, add-on content, input, keyboard and applets, display and operation mode, accounts and network, settings and fonts, C library and time, modules |
| movies.md | The SDK's movie library: decoder lifecycle, output slots, native-texture output, demuxers and codecs, game movie players |
| audio.md | The audio renderer and its command lists by SDK, AudioOut and devices, codecs, timing |
| frame-rate.md | The host's frame step, engine frame limiters, present intervals and display rates, CPU cost |
| nvn-objects.md | NVN object sizes, memory pools, command memory, queues and syncs, CPU-written textures, NVN entry points |
| rendering.md | The replay on Metal: textures, passes and clears, blend / depth / stencil, vertex input, uniforms and ordering, conditional rendering; the Vulkan and OpenGL layers |
| shaders.md | Shader containers by engine, control sections and code, translation, bindings and handles, shaders compiled at run time, pipelines and caches |
| game-state.md | Saves, scripted state, and game behaviour that looks like a host bug but is not |
A new general finding goes into the file and section it belongs to, in the same form, with a line in the symptom index; a fact about one game goes into that game's note.