Frank14f/CelerisLab
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge K4 helper withdrawal into K5

Browse Source 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Claude
2026-10-22 01:32:42 +00:00
co-authored by Claude Opus 5.5
parent 0ddc6c3d62 78c0f1811b
commit af406b44bf
3 changed files with 9 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/CelerisLab/lbm/kernels/blocks.cu
+8 -7
View File
@@ -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/core/layout.cuh
-15
View File
@@ -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)
// ---------------------------------------------------------------------------