From 0837872bed3d97d3433d7e0cbafffa7f271ebee1 Mon Sep 17 00:00:00 2001 From: "Emil A. Overbeck" Date: Sat, 18 Jul 2026 00:59:18 +0200 Subject: Deterministic physics doc --- research/tech-stack.org | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'research/tech-stack.org') diff --git a/research/tech-stack.org b/research/tech-stack.org index ce0a9c1..3e43e44 100644 --- a/research/tech-stack.org +++ b/research/tech-stack.org @@ -436,6 +436,12 @@ my_mod/ * Deterministic Simulation Design +NOTE: the authoritative plan is =research/deterministic-physics.org=. The +section below predates the custom solver and assumes Box2D owns the simulation. +It does not: the game is decided by the mass-spring solver in +=src/game/build_graph.cpp=, and Box2D is idle. The determinism work targets that +solver (strict IEEE-754 float, owned transcendentals), not Box2D. + For lockstep multiplayer, the simulation must be exactly identical on all machines. @@ -474,7 +480,9 @@ Projects to study during development: * Open Questions -- [ ] Fixed-point math fallback vs trusting Box2D v3 determinism +- [X] Fixed-point math fallback vs trusting Box2D v3 determinism — resolved: + strict IEEE-754 float on the custom solver, fixed-point held in reserve. See + =research/deterministic-physics.org=. - [ ] Exact rope simulation approach within Box2D (distance joints? soft constraints? custom spring?) - [ ] Fire propagation model (graph traversal along connected flammable -- cgit v1.3