- bone_sse: deduplicate findInterpIdx calls in bone loop — rotation's
search result reused for scale/translation when tracks share temporal
structure (canReuseInterp guard). Est. ~23% bone loop cycle reduction.
- silicon: SSE replacement for processLinkedListCollision (0x6ABC40,
1.57% CPU). V4 AABB overlap test replaces 6 x87 FCOMP/FNSTSW.
Benched at 3.2x speedup (378→115 cyc/call, 8 nodes).
- transform44: remove A/B toggle, always use SSE path (teardown guard
kept). A/B infrastructure remains for other hooks.
- bench: add processLinkedListCollision benchmark with fake linked list
test fixture and stubbed addGeometryToBuffer.
mulMat3x4 / mulMat3x4InPlace: translation row had A*B operands swapped.
Original computes A_translation * B_rotation + B_translation, but our
code was doing A_rotation * B_translation + A_translation. Verified by
tracing x87 disassembly: first element loads A[9]*B[col] pattern.
Fixed in both silicon_sse.zig and silicon.zig.
packParticleColor: x87 rounds 127.5 to 128 (round-to-nearest), but SSE
@intFromFloat truncates to 127. Added @round() before @intFromFloat.
42/42 benchmarks now pass correctness. 0 MISMATCHes.
Probe infrastructure using comptime probeDetour() that generates
per-function detours: atomic counter increment + callOriginal passthrough.
Hit counts reported on shutdown. Hooks installed at lateInit (engine init),
not DLL load time.
Key findings during decompilation:
- 0x40CF81 is GetFPUControlWord, not ftol (silicon mislabel)
- Real __ftol at 0x40A2B0 (51+ callers, hot path)
- 0x7B7A80/7B7B10 use normal stack floats, not FPU register params
- 0x686640/686820/6868E0 are bounding volume ops, not vector ops
- 0x7786A0 is a UI model constructor, not SetModelLighting
Complete catalog of libSiliconPatch240.dll hook targets with verified
calling conventions, parameter layouts, RET stack cleanup, and functional
descriptions from Ghidra analysis. Zero unknown stubs remain.
MarcelineVQ