embedded3 PERF-3 P1-5 (c) part 1: scatter kernels (e3_scat_{f64,f32,u32,i32}) in the CG module; the refresh's seven f64 coefficient arrays and z links, and the V-cycle fine level's f32 arrays and maps, scattered on the changed rows into persistent buffers (DeviceVcycle::refresh_fine_rows) — slab CSV byte-identical, band check passed, device moving/cg green; ny 124 rebuild block 1,895 → 1,748 ms per step. The first form redirected the V-cycle's own e3_cg_scatter_f64 launch (a replace-all on the launch name) and failed every gate — the symbol table is the first thing to read when adding to a module

Co-Authored-By: Claude Fable 5.1 <[email protected]>

crates/specialized/rtx-cfd/src/kernels/cuda/e3_cg.cu

@@ -198,3 +198,27 @@ extern "C" __global__ void e3_cg_scatter_f64(
     int g = cells[t];
     z[g] = (double) x0[g];
 }
+
+
+
+/* P1-5 (c): scatter compact (index, value) lists into persistent buffers. */
+extern "C" __global__ void e3_scat_f64(int n, const unsigned int* __restrict__ idx, const double* __restrict__ val, double* __restrict__ dst)
+{
+    int t = blockIdx.x * blockDim.x + threadIdx.x;
+    if (t < n) dst[idx[t]] = val[t];
+}
+extern "C" __global__ void e3_scat_f32(int n, const unsigned int* __restrict__ idx, const float* __restrict__ val, float* __restrict__ dst)
+{
+    int t = blockIdx.x * blockDim.x + threadIdx.x;
+    if (t < n) dst[idx[t]] = val[t];
+}
+extern "C" __global__ void e3_scat_u32(int n, const unsigned int* __restrict__ idx, const unsigned int* __restrict__ val, unsigned int* __restrict__ dst)
+{
+    int t = blockIdx.x * blockDim.x + threadIdx.x;
+    if (t < n) dst[idx[t]] = val[t];
+}
+extern "C" __global__ void e3_scat_i32(int n, const unsigned int* __restrict__ idx, const int* __restrict__ val, int* __restrict__ dst)
+{
+    int t = blockIdx.x * blockDim.x + threadIdx.x;
+    if (t < n) dst[idx[t]] = val[t];
+}