rtx-cfd/rtx-fsi: overset A-P0 GATED + M1 precision probe — curvilinear collocated PISO: relative-reduction pressure stop (the absolute stop floored |du/dt| at 2e-4 on 64²), line-implicit-n sign fix, adjustPhi; gates: Cartesian reduction 1.37–1.40x the staggered error at orders 0.83/0.90; skewed stretched periodic annulus Stokes orders 2.30/2.06 (explicit and line-implicit), upwind 1.08/0.80; Poiseuille exact to 1e-9 on Cartesian and affine-sheared periodic channels (both diffusion variants), varying-skew channel order 2.02 (v 1.9), cell mass 1e-14; divergence ≤ 1e-11 relative every step; snapshot/restore bit-identical. M1: poisson.rs multigrid hierarchy generic over MgScalar (f32/f64), f64 CG keeps its own fine level; MgPrecision on MultigridParameters/EmbeddedParameters/PisoParameters, set_poisson_precision, harness RTX_FSI2_POISSON_F32 (march + noise probe, printed marker); f64 arm bit-identical in vivo (FSI2 default line-for-line with 08-31), f32 arm holds the noise floor and stall pins and the FSI2 band; poisson_equivalence f32 arm

Co-Authored-By: Claude Fable 5.1 <[email protected]>
Claude-Session: https://claude.ai/code/session_01X2GmJXeQ2njUecEKiJZ1G2

crates/specialized/rtx-cfd/src/solvers/incompressible/curvilinear/mod.rs

@@ -95,7 +95,12 @@ pub struct CurvilinearParameters {
     /// Projections per step (the first removes the divergence; the rest
     /// mop up inner-solver truncation).
     pub corrector_steps: usize,
-    /// Pressure-correction stop, relative to the step's flux scale.
+    /// Pressure-correction stop: each projection reduces the L1 cell mass
+    /// imbalance by this factor (plus a rounding floor of 1e-15 × Σ|F|).
+    /// Relative to the incoming divergence, not to the flux scale, so the
+    /// stop tightens as the flow settles — an absolute stop leaves a
+    /// velocity-noise floor that grows with the grid (measured: |du/dt|
+    /// floored at 2e-4 on the 64² Cartesian MMS with `1e-10 × Σ|F|`).
     pub tolerance: f64,
     /// Convection scheme.
     pub convection: PatchConvection,
@@ -111,7 +116,7 @@ impl Default for CurvilinearParameters {
     fn default() -> Self {
         Self {
             corrector_steps: 2,
-            tolerance: 1e-8,
+            tolerance: 1e-4,
             convection: PatchConvection::Upwind,
             normal_diffusion: NormalDiffusion::Explicit,
             boundaries: PatchBoundaries::default(),
@@ -304,22 +309,24 @@ impl CurvilinearPisoSolver {
         }
         let (_, matrix, anchor) = self.matrix.as_ref().expect("assembled");
         let flux_scale: f64 = field.flux.iter().map(|f| f.abs()).sum::<f64>().max(1e-300);
-        let tolerance = self.params.tolerance * flux_scale;
+        let floor = 1e-15 * flux_scale;
 
         let mut iterations = 0;
         let mut converged = true;
         let mut performed = 0;
         let mut max_div = self.max_divergence(&field.flux);
         for _ in 0..self.params.corrector_steps {
+            let incoming = self.divergence_l1(&field.flux);
+            if incoming <= floor {
+                break;
+            }
+            let tolerance = self.params.tolerance * incoming + floor;
             let (pc, out) = self.solve_pressure_correction(matrix, *anchor, &field.flux, tolerance);
             iterations += out.iterations;
             converged &= out.converged;
             self.apply_correction(field, &pc, dt);
             performed += 1;
             max_div = self.max_divergence(&field.flux);
-            if max_div <= tolerance {
-                break;
-            }
         }
         self.time = t_new;
         Ok(CurvilinearResult {