From 4ec94c97879aafef15f7663135745e4ba61e62cf Mon Sep 17 00:00:00 2001
From: Adrian Kummerlaender
Date: Mon, 17 May 2021 00:15:33 +0200
Subject: Extract first public LiterateLB version

---
 tangle/benchmark-ldc.cu | 95 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 95 insertions(+)
 create mode 100644 tangle/benchmark-ldc.cu

(limited to 'tangle/benchmark-ldc.cu')

diff --git a/tangle/benchmark-ldc.cu b/tangle/benchmark-ldc.cu
new file mode 100644
index 0000000..2443afe
--- /dev/null
+++ b/tangle/benchmark-ldc.cu
@@ -0,0 +1,95 @@
+#include <LLBM/base.h>
+
+#include <LLBM/kernel/collide.h>
+#include <LLBM/kernel/bounce_back.h>
+#include <LLBM/kernel/bounce_back_moving_wall.h>
+
+#include "util/timer.h"
+
+#include <iostream>
+
+using DESCRIPTOR = descriptor::D3Q19;
+
+template <typename T>
+void simulate(descriptor::Cuboid<DESCRIPTOR> cuboid, std::size_t nStep) {
+  cudaSetDevice(0);
+
+  Lattice<DESCRIPTOR,T> lattice(cuboid);
+  
+  CellMaterials<DESCRIPTOR> materials(cuboid, [&cuboid](uint3 p) -> int {
+    if (p.x == 0 || p.x == cuboid.nX-1 || p.y == 0 || p.y == cuboid.nY-1 || p.z == 0) {
+      return 2; // boundary cell
+    } else if (p.z == cuboid.nZ-1) {
+      return 3; // lid cell
+    } else {
+      return 1; // bulk
+    }
+  });
+  
+  auto bulk_mask = materials.mask_of_material(1);
+  auto box_mask  = materials.mask_of_material(2);
+  auto lid_mask  = materials.mask_of_material(3);
+  
+  auto bulk_cells = materials.list_of_material(1);
+  auto box_cells  = materials.list_of_material(2);
+  auto lid_cells  = materials.list_of_material(3);
+  
+  lattice.template apply<InitializeO>(bulk_cells);
+  lattice.template apply<InitializeO>(box_cells);
+  lattice.template apply<InitializeO>(lid_cells);
+  
+  cudaDeviceSynchronize();
+
+  for (std::size_t iStep=0; iStep < 100; ++iStep) {
+    lattice.apply(Operator(BgkCollideO(), bulk_mask, 0.56),
+                  Operator(BounceBackO(), box_mask),
+                  Operator(BounceBackMovingWallO(), lid_mask, 0.05f, 0.f, 0.f));
+    lattice.stream();
+  }
+
+  cudaDeviceSynchronize();
+
+  auto start = timer::now();
+
+  for (std::size_t iStep=0; iStep < nStep; ++iStep) {
+    lattice.apply(Operator(BgkCollideO(), bulk_mask, 0.56),
+                  Operator(BounceBackO(), box_mask),
+                  Operator(BounceBackMovingWallO(), lid_mask, 0.05f, 0.f, 0.f));
+    lattice.stream();
+  }
+
+  cudaDeviceSynchronize();
+
+  auto mlups = timer::mlups(cuboid.volume, nStep, start);
+
+  std::cout << sizeof(T) << ", " << cuboid.nX << ", " << nStep << ", " << mlups << std::endl;
+}
+
+int main(int argc, char* argv[]) {
+  if (argc < 3 || argc > 4) {
+    std::cerr << "Invalid parameter count" << std::endl;
+    return -1;
+  }
+
+  const std::size_t n     = atoi(argv[1]);
+  const std::size_t steps = atoi(argv[2]);
+
+  unsigned precision = 4;
+  if (argc == 4) {
+    precision = atoi(argv[3]);
+  }
+
+  switch (precision) {
+  case 4:
+    simulate<float>({ n, n, n}, steps);
+    break;
+  case 8:
+    simulate<double>({ n, n, n}, steps);
+    break;
+  default:
+    std::cerr << "Invalid precision" << std::endl;
+    return -1;
+  }
+
+  return 0;
+}
-- 
cgit v1.2.3