kernels (K4): withdraw the address helpers; keep only fill_invalid

Horsea's gates showed block_ddf/block_slot change real PTX in
BlockHaloCopy, BlockExplode and BlockExplodeProjected (6/10 modules;
BlockExplodeProjected 38 -> 40 registers) although the results stayed
exact. The stack's bar is byte-identical PTX for naming-only steps, so
those three kernels return to their K3 expressions, the helpers leave
core/layout.cuh, and test_projected_explosion goes back to its K3 form.
fill_invalid stays (its kernels were byte-identical). K4 vs K3 is now
only the NaN-fill helper.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/CelerisLab/lbm/kernels/blocks.cu

@@ -45,8 +45,8 @@ extern "C" __global__ void BlockHaloCopy(
     unsigned int row = blockIdx.x * blockDim.x + threadIdx.x;
     if (row >= count) return;
     const uint64_t d = dst[row], s = src[row];
-    fpxx* to = block_ddf(f, d);
-    const fpxx* from = block_ddf(f, s);
+    fpxx* to = f + (d / TOTAL_CELLS) * NQ * TOTAL_CELLS;
+    const fpxx* from = f + (s / TOTAL_CELLS) * NQ * TOTAL_CELLS;
     for (unsigned int q = 0; q < NQ; ++q)
         to[index_f(d % TOTAL_CELLS, q)] = from[index_f(s % TOTAL_CELLS, q)];
 }
@@ -61,9 +61,9 @@ extern "C" __global__ void BlockExplode(
     unsigned int row = blockIdx.x * blockDim.x + threadIdx.x;
     if (row >= count) return;
     const uint64_t t = target[row];
-    fpxx* to = block_ddf(fine, t);
+    fpxx* to = fine + (t / TOTAL_CELLS) * NQ * TOTAL_CELLS;
     const uint64_t d = donor[row];
-    const fpxx* from = block_ddf(coarse, d, coarse_cells);
+    const fpxx* from = coarse + (d / coarse_cells) * NQ * coarse_cells;
     for (unsigned int q = 0; q < NQ; ++q)
         to[index_f(t % TOTAL_CELLS, q)] = from[q * coarse_cells + d % coarse_cells];
 }
@@ -80,12 +80,13 @@ extern "C" __global__ void BlockExplodeProjected(
     unsigned int row = blockIdx.x * blockDim.x + threadIdx.x;
     if (row >= count) return;
     const uint64_t t = target[row], d = donor[row];
-    fpxx* to = block_ddf(fine, t);
+    fpxx* to = fine + (t / TOTAL_CELLS) * NQ * TOTAL_CELLS;
     for (unsigned int q = 0; q < NQ; ++q) {
         float v[4];
         for (unsigned int a = 0; a < 4; ++a) {
             const uint64_t n = neighbors[4ul * row + a];
-            v[a] = coarse[block_slot(n, coarse_cells, q)];
+            v[a] = coarse[(n / coarse_cells) * NQ * coarse_cells
+                          + q * coarse_cells + n % coarse_cells];
         }
         float gx = .5f * (v[1] - v[0]), gy = .5f * (v[3] - v[2]);
         const float ex = d_cx[q], ey = d_cy[q];
@@ -95,7 +96,7 @@ extern "C" __global__ void BlockExplodeProjected(
             gy -= ey * along;
         }
         to[index_f(t % TOTAL_CELLS, q)] =
-            coarse[block_slot(d, coarse_cells, q)]
+            coarse[(d / coarse_cells) * NQ * coarse_cells + q * coarse_cells + d % coarse_cells]
             + offsets[2ul * row] * gx + offsets[2ul * row + 1] * gy;
     }
 }

src/CelerisLab/lbm/kernels/boundary/bounce_back.cuh

@@ -15,13 +15,10 @@
 // D2Q9 y-wall half-way bounce-back for pull double-buffer streaming.
 // ---------------------------------------------------------------------------
 #if LATTICE_MODEL == LATTICE_D2Q9
-// Rows 1 and top_row (= physical ny - 2) next to walls at rows 0 and ny - 1.
-// y is the physical row; k addresses the node in fi_in (a block window may
-// have NX/NY strides different from the physical extent).
-__device__ inline void apply_wall_bb_d2q9_rows(unsigned int y, unsigned int top_row,
-                                               float* __restrict__ f,
-                                               const fpxx* __restrict__ fi_in,
-                                               unsigned long k)
+__device__ inline void apply_wall_bb_d2q9_y_pull(unsigned int y,
+                                                  float* __restrict__ f,
+                                                  const fpxx* __restrict__ fi_in,
+                                                  unsigned long k)
 {
     if (y == 1) {
         // Directions sourced from y=0 wall in pull step: +y, +x+y, -x+y.
@@ -29,22 +26,14 @@ __device__ inline void apply_wall_bb_d2q9_rows(unsigned int y, unsigned int top_
         f[5] = load_ddf(fi_in, index_f(k, 6u));
         f[8] = load_ddf(fi_in, index_f(k, 7u));
     }
-    else if (y == top_row) {
-        // Directions sourced from the y = ny - 1 wall in pull step: -y, -x-y, +x-y.
+    else if (y == (unsigned int)(NY - 2)) {
+        // Directions sourced from y=NY-1 wall in pull step: -y, -x-y, +x-y.
         f[4] = load_ddf(fi_in, index_f(k, 3u));
         f[6] = load_ddf(fi_in, index_f(k, 5u));
         f[7] = load_ddf(fi_in, index_f(k, 8u));
     }
 }
 
-__device__ inline void apply_wall_bb_d2q9_y_pull(unsigned int y,
-                                                  float* __restrict__ f,
-                                                  const fpxx* __restrict__ fi_in,
-                                                  unsigned long k)
-{
-    apply_wall_bb_d2q9_rows(y, (unsigned int)(NY - 2), f, fi_in, k);
-}
-
 // ---------------------------------------------------------------------------
 // D2Q9 y-wall free-slip (specular reflection) for pull double-buffer streaming.
 //

src/CelerisLab/lbm/kernels/core/layout.cuh

@@ -65,21 +65,6 @@ __device__ __forceinline__ void block_global_xy(unsigned int x, unsigned int y,
     gy = (blockIdx.z / nbx) * by + y - halo;
 }
 
-// Block-contiguous storage: block b holds NQ*cells values (Q-major inside) and
-// block cell id b*cells + c names cell c of block b. cells is TOTAL_CELLS for a
-// block window, or the whole domain for a dense (one-window) parent.
-template <typename T>
-__device__ __forceinline__ T* block_ddf(T* f, uint64_t block_cell, unsigned long cells = TOTAL_CELLS)
-{
-    return f + (block_cell / cells) * NQ * cells;
-}
-
-// Offset of population q of block cell block_cell from the start of storage.
-__device__ __forceinline__ uint64_t block_slot(uint64_t block_cell, unsigned long cells, unsigned int q)
-{
-    return (block_cell / cells) * NQ * cells + q * cells + block_cell % cells;
-}
-
 // ---------------------------------------------------------------------------
 // Neighbor list (periodic BCs via modular arithmetic)
 // ---------------------------------------------------------------------------

src/CelerisLab/lbm/kernels/operators/bounce_back_swap.cuh

@@ -1,14 +0,0 @@
-// CelerisLab – operators/bounce_back_swap.cuh
-// Full-way bounce-back at a wall node: swap every opposite pair (paired
-// order q, q+1). Needs only NQ; shared by the uniform and blocked steps.
-#ifndef CELERIS_OPERATORS_BOUNCE_BACK_SWAP_CUH
-#define CELERIS_OPERATORS_BOUNCE_BACK_SWAP_CUH
-
-__device__ __forceinline__ void bounce_back_swap(float* f) {
-    #pragma unroll
-    for (int i = 1; i < NQ; i += 2) {
-        float t = f[i]; f[i] = f[i+1]; f[i+1] = t;
-    }
-}
-
-#endif

src/CelerisLab/lbm/kernels/operators/helpers.cuh

@@ -5,7 +5,12 @@
 #ifndef CELERIS_OPERATORS_HELPERS_CUH
 #define CELERIS_OPERATORS_HELPERS_CUH
 
-#include "bounce_back_swap.cuh"
+__device__ __forceinline__ void bounce_back_swap(float* f) {
+    #pragma unroll
+    for (int i = 1; i < NQ; i += 2) {
+        float t = f[i]; f[i] = f[i+1]; f[i+1] = t;
+    }
+}
 
 __device__ __forceinline__ void collide_dispatch(
     float* f, float rho_n,

src/CelerisLab/lbm/kernels/step/block_channel.cu

@@ -5,8 +5,6 @@
 // Root-only blocked counterpart of OneStep's D2Q9/SRT channel closure.
 // NX/NY are LOCAL strides; all physical boundary decisions use global coordinates.
 #include "../core/cell_kind.cuh"
-#include "../operators/bounce_back_swap.cuh"
-#include "../boundary/bounce_back.cuh"
 #include "../boundary/inlet/common.cuh"
 #include "../boundary/inlet/zou_he_local.cuh"
 #include "../boundary/outlet/pressure_neq.cuh"
@@ -36,17 +34,27 @@ extern "C" __global__ void BlockChannelStep(
         const bool wall = gy == 0 || gy == d_block_channel.ny - 1;
         const bool outlet = gx == d_block_channel.nx - 1;
         if (wall) {
-            bounce_back_swap(f);  // Wall owns all four corners; ghosts never collide.
+            // Wall owns all four corners. Ghosts pull and swap, never collide.
+            for (int q = 1; q < NQ; q += 2) {
+                const float t = f[q]; f[q] = f[q + 1]; f[q + 1] = t;
+            }
         } else if (gx == 0) {
             apply_zou_he_left_velocity_inlet_d2q9(
                 f, inlet_target_u((float)gy, d_block_channel.ny, d_block_channel.velocity), 0.0f);
         } else if (outlet) {
             // Immutable input donor, receiver-local (also valid across a block seam).
             float donor[NQ];
-            load_node(in + base, k - 1, donor);
+            for (int q = 0; q < NQ; ++q)
+                donor[q] = load_ddf(in + base, index_f(k - 1, q));
             apply_pressure_outlet_d2q9(f, donor);
-        } else {
-            apply_wall_bb_d2q9_rows(gy, d_block_channel.ny - 2, f, in + base, k);
+        } else if (gy == 1) {
+            f[3] = load_ddf(in + base, index_f(k, 4));
+            f[5] = load_ddf(in + base, index_f(k, 6));
+            f[8] = load_ddf(in + base, index_f(k, 7));
+        } else if (gy == d_block_channel.ny - 2) {
+            f[4] = load_ddf(in + base, index_f(k, 3));
+            f[6] = load_ddf(in + base, index_f(k, 5));
+            f[7] = load_ddf(in + base, index_f(k, 8));
         }
         if (!wall && !outlet) {
 #if COLLISION_MODEL == 2

src/CelerisLab/lbm/kernels/streaming/pull_double_buffer.cuh

@@ -21,14 +21,6 @@
 //   i.e. the population traveling in direction i at node n
 //   arrived from the node at n - c_i  = neighbor in direction opp(i)
 // ---------------------------------------------------------------------------
-// load_node: all NQ populations of node n as stored (no streaming).
-__device__ __forceinline__ void load_node(const fpxx* __restrict__ fi, unsigned long n,
-                                          float* __restrict__ f)
-{
-    for (int q = 0; q < NQ; ++q)
-        f[q] = load_ddf(fi, index_f(n, q));
-}
-
 __device__ __forceinline__ void stream_pull_load(unsigned long n,
                                                  float* __restrict__ f,
                                                  const fpxx* __restrict__ fi,

tests/unit/test_projected_explosion.py

@@ -163,12 +163,6 @@ class ProjectedKernelTests(unittest.TestCase):
         donors = parent.owner.ravel()[p.ghost_coarse]
         source = (Path(__file__).resolve().parents[2] /
                   "src/CelerisLab/lbm/kernels/blocks.cu").read_text()
-        # The kernel addresses storage through the real core/layout.cuh helpers.
-        layout = (Path(__file__).resolve().parents[2] /
-                  "src/CelerisLab/lbm/kernels/core/layout.cuh").read_text()
-        start = layout.index("template <typename T>\n__device__ __forceinline__ T* block_ddf(")
-        end = layout.index("}\n", layout.index("__device__ __forceinline__ uint64_t block_slot(")) + 2
-        helpers = layout[start:end].replace("__device__ __forceinline__ ", "inline ")
         kernel = source.split('extern "C" __global__ void BlockExplodeProjected(',1)[1]
         kernel = 'void BlockExplodeProjected(' + kernel.split('// Targets are receiver-local SOURCE slots',1)[0]
         def array(name, values, kind):
@@ -189,7 +183,7 @@ class ProjectedKernelTests(unittest.TestCase):
             code += f'constexpr unsigned NQ=9, TOTAL_CELLS={int(np.prod(child.local_shape))};\n'
             code += 'struct Id {unsigned x;}; Id blockIdx{0},blockDim{1},threadIdx{0};\n'
             code += 'unsigned index_f(unsigned k,unsigned q){return q*TOTAL_CELLS+k;}\n'
-            code += 'int d_cx[]={0,1,-1,0,0,1,-1,1,-1},d_cy[]={0,0,0,1,-1,1,-1,-1,1};\n'+helpers+kernel
+            code += 'int d_cx[]={0,1,-1,0,0,1,-1,1,-1},d_cy[]={0,0,0,1,-1,1,-1,-1,1};\n'+kernel
             code += data+array('target',targets,'uint64_t')+array('donor',donors,'uint64_t')
             code += array('neighbor',ids,'uint64_t')+array('offset',offsets,'float')
             code += f'float fine[{child.nblocks*9*int(np.prod(child.local_shape))}]={{}};\n'