Commit Graph

13 Commits

Author SHA1 Message Date
Opok 4412bc7a7d widescreen: add per-primitive X-reject bound-load cull sites
Add opt-in [widescreen.cull] xclip_load_sites: a configured `lw rt,off(rs)`
loading a per-primitive X-reject bound (sltu-compared against ANDI-masked
u16 screen X) routes through psx_ws_xclip_bound() - INT32_MAX while the
margins are revealed (reject disabled; the wide-surface scissor clips the
overflow, wrapped off-left u16 coords pass), vanilla value at 4:3. Matched
native codegen emit and interpreter hook; overlay ABI v18 forwarder (NULL
preserves the vanilla loaded value); codegen ver 8. Parser, codegen and
overlay-mismatch regression coverage.

Split from #49: replaces the xclip_globals RAM poke with a recompile-time
instruction transform so native shards and the interpreter behave identically.
2026-07-21 15:02:20 -07:00
Opok a5fdf1ad9a widescreen: add side frustum-plane normal-X cull sites
Add opt-in [widescreen.cull] plane_nx_sites: a configured `lw rt,off(rs)`
loading the X component of a side frustum-plane normal (sign-tested per
corner, dot = nx*px + nz*pz) routes through psx_ws_plane_nx() — scaled by
the inverse aspect factor (4*den)/(3*num) while the margins are revealed,
widening the plane cone by exactly atan((3*num)/(4*den)*tan(theta));
identity at 4:3. Matched native codegen emit and interpreter hook; overlay
ABI v17 forwarder (NULL preserves the vanilla normal); codegen ver 7.
Parser, codegen and overlay-mismatch regression coverage.

Split from #49: replaces the plane sign-test RAM pokes with a
recompile-time instruction transform so native shards and the interpreter
behave identically.
2026-07-21 14:38:55 -07:00
Matthew Stanley e38daa3415 fix: mirror widescreen range sites in interpreter
Register configured range_sites with the dirty-RAM interpreter so overlay-resident classifiers use the same two-margin widening already emitted by native codegen.

Co-authored-by: Opok <novelasevilliousesp@gmail.com>
2026-07-20 19:39:36 -07:00
Matthew Stanley 8f78ddbd89 integrate interpreter hotpath and AOT sharding 2026-07-20 15:58:43 -07:00
Matthew Stanley fa626bdb4c feat: add aspect-scaled depth cull sites
Add opt-in slti/sltiu far-bound scaling with matched native and interpreter semantics, correct immediate sign extension, and an overlay ABI forwarder for live aspect state.

Co-authored-by: Opok <novelasevilliousesp@gmail.com>
2026-07-20 12:14:18 -07:00
Matthew Stanley 288926111f feat: add masked-u16 widescreen cull sites
Add a shared native/interpreter helper for sltiu classifiers operating on ANDI-masked 16-bit screen X values, preserving MIPS immediate sign extension and overlay linkage.

Co-authored-by: Opok <novelasevilliousesp@gmail.com>
2026-07-20 12:11:58 -07:00
Matthew Stanley 5ded80e61e Merge master (AOT overlay generation) into perf/interpreter-hotpath
Integration branch: the interpreter-perf series (interpreter fallback +
additive shard autocompile + always-on production observability + the
fail-closed alias-body emitter fix) rebased-by-merge onto master's AOT
overlay generation (play-free extraction, bounded indirect discovery,
AOT recall, shadow-diff redesign).

Semantic-union highlights:
- gte.cpp: HEAD's direct-field marshaling fast paths kept; master's
  s_gte_replay_sandbox guards grafted onto every affected site.
- psx_icache.c: HEAD's global tag array + fetch_miss split kept;
  master's transactional shadow record/replay retargeted onto it.
- overlay_capture.c: HEAD's commit/retry/wait-pending pipeline kept;
  master's durable history + autocapture APIs wired into it
  (persist_history_snapshot fed from the immutable .d contribution).
- overlay_loader.c: master's interp-record/native-replay shadow-diff
  redesign kept as base with HEAD's scheduler-bail hardening grafted
  (new SHDIFF_SCHED divergence kind); strict manifest v2 parsing keeps
  master's P pair-id lines; always-on native-call ring preserved.
- main.cpp shutdown calls BOTH overlay_autocapture_shutdown() and
  overlay_capture_wait_pending() (the latter drains the preserve-write
  FIFO the former does not).
- recompiler: master's generate_alias_group refactor retains the
  fail-closed alias-body re-entry defaults from c85f1d1.

Validated on this tree: recompiler + runtime + tests build; all six
runtime unit tests PASS; interpreter perf guards PASS; recompiler
patch/structural tests PASS (l1 decoder oracle noise = 376 pre-existing
on both parents).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-20 00:43:41 -07:00
Matthew Stanley bdeb106e5c feat: add negsub widescreen cull sites
Add an opt-in horizontal low-edge transform for subu rd,zero,rt classifiers, with matching native and dirty-RAM interpreter behavior and guarded overlay variants.

Co-authored-by: Opok <novelasevilliousesp@gmail.com>
2026-07-19 23:25:03 -07:00
Matthew Stanley 504db26445 feat: add bounded indirect overlay discovery 2026-07-19 17:15:17 -07:00
Matthew Stanley 05f6ef3e02 Optimize GTE fallback and isolate interpreter perf runs 2026-07-19 14:23:36 -07:00
Matthew Stanley a544809edf aot: persist overlay capture history atomically 2026-07-17 23:56:58 -07:00
Matthew Stanley 2f73f73f92 fix: keep patch config header C++17-compatible
Avoid exposing std::span through config_loader.h, which is consumed by C++17 game runtimes. Add a dedicated C++17 header compile test alongside the guarded-patch suite.
2026-07-13 22:33:02 -07:00
Matthew Stanley e7b61c807c recompiler: add guarded config MIPS patches
Add alias-aware, opcode-verified game-owned instruction replacements before discovery and code generation. Validate duplicate IDs and physical sites, preserve guarded overlay variants, and cover parser and generated-code behavior with CTest.

Co-authored-by: douglasjv <vanderveerdouglas@gmail.com>
2026-07-13 21:59:18 -07:00