| 1 | // -------------------------------------------------------------------------- |
| 2 | // |
| 3 | // Copyright |
| 4 | // Markus Wittmann, 2016-2017 |
| 5 | // RRZE, University of Erlangen-Nuremberg, Germany |
| 6 | // markus.wittmann -at- fau.de or hpc -at- rrze.fau.de |
| 7 | // |
| 8 | // Viktor Haag, 2016 |
| 9 | // LSS, University of Erlangen-Nuremberg, Germany |
| 10 | // |
| 11 | // This file is part of the Lattice Boltzmann Benchmark Kernels (LbmBenchKernels). |
| 12 | // |
| 13 | // LbmBenchKernels is free software: you can redistribute it and/or modify |
| 14 | // it under the terms of the GNU General Public License as published by |
| 15 | // the Free Software Foundation, either version 3 of the License, or |
| 16 | // (at your option) any later version. |
| 17 | // |
| 18 | // LbmBenchKernels is distributed in the hope that it will be useful, |
| 19 | // but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 20 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 21 | // GNU General Public License for more details. |
| 22 | // |
| 23 | // You should have received a copy of the GNU General Public License |
| 24 | // along with LbmBenchKernels. If not, see <http://www.gnu.org/licenses/>. |
| 25 | // |
| 26 | // -------------------------------------------------------------------------- |
| 27 | #include "BenchKernelD3Q19ListAaRiaCommon.h" |
| 28 | |
| 29 | #include "Memory.h" |
| 30 | #include "Vtk.h" |
| 31 | #include "LikwidIf.h" |
| 32 | |
| 33 | #include <inttypes.h> |
| 34 | #include <math.h> |
| 35 | |
| 36 | #ifdef _OPENMP |
| 37 | #include <omp.h> |
| 38 | #endif |
| 39 | |
| 40 | void FNAME(D3Q19ListAaRiaKernel)(LatticeDesc * ld, KernelData * kernelData, CaseData * cd) |
| 41 | { |
| 42 | Assert(ld != NULL); |
| 43 | Assert(kernelData != NULL); |
| 44 | Assert(cd != NULL); |
| 45 | |
| 46 | Assert(cd->Omega > 0.0); |
| 47 | Assert(cd->Omega < 2.0); |
| 48 | |
| 49 | KernelData * kd = (KernelData *)kernelData; |
| 50 | KernelDataList * kdl = KDL(kernelData); |
| 51 | KernelDataListRia * kdlr = KDLR(kernelData); |
| 52 | PdfT omega = cd->Omega; |
| 53 | PdfT omegaEven = omega; |
| 54 | // 1/ 4: best stability; |
| 55 | // 1/12: removes third-order advection error (best advection); |
| 56 | // 1/ 6: removes fourth-order diffusion error (best diffusion); |
| 57 | // 3/16: exact location of bounce back for poiseuille flow |
| 58 | PdfT magicParam = F(1.0) / F(12.0); |
| 59 | PdfT omegaOdd = F(1.0) / (F(0.5) + magicParam / (F(1.0) / omega - F(0.5))); |
| 60 | |
| 61 | PdfT evenPart = F(0.0); |
| 62 | PdfT oddPart = F(0.0); |
| 63 | PdfT dir_indep_trm = F(0.0); |
| 64 | |
| 65 | const PdfT w_0 = F(1.0) / F(3.0); |
| 66 | const PdfT w_1 = F(1.0) / F(18.0); |
| 67 | const PdfT w_2 = F(1.0) / F(36.0); |
| 68 | |
| 69 | const PdfT w_1_x3 = w_1 * F(3.0); const PdfT w_1_nine_half = w_1 * F(9.0) / F(2.0); PdfT w_1_indep = F(0.0); |
| 70 | const PdfT w_2_x3 = w_2 * F(3.0); const PdfT w_2_nine_half = w_2 * F(9.0) / F(2.0); PdfT w_2_indep = F(0.0); |
| 71 | |
| 72 | PdfT ui; |
| 73 | |
| 74 | PdfT ux, uy, uz; |
| 75 | PdfT dens; |
| 76 | |
| 77 | // Declare pdf_N, pdf_E, pdf_S, pdf_W, ... |
| 78 | #define X(name, idx, idxinv, x, y, z) PdfT JOIN(pdf_,name); |
| 79 | D3Q19_LIST |
| 80 | #undef X |
| 81 | |
| 82 | // Declare pointers to pdfs ppdf_N, ppdf_E, ppdf_S, ppdf_W, ... |
| 83 | #define X(name, idx, idxinv, x, y, z) PdfT * JOIN(ppdf_,name) = NULL; |
| 84 | D3Q19_LIST |
| 85 | #undef X |
| 86 | |
| 87 | uint32_t nConsecNodes = kdlr->nConsecNodes; |
| 88 | uint32_t * consecNodes = kdlr->ConsecNodes; |
| 89 | uint32_t consecIndex = 0; |
| 90 | uint32_t consecValue = 0; |
| 91 | |
| 92 | #ifndef DEBUG |
| 93 | UNUSED(nConsecNodes); |
| 94 | #endif |
| 95 | |
| 96 | PdfT * src = kd->Pdfs[0]; |
| 97 | |
| 98 | int maxIterations = cd->MaxIterations; |
| 99 | int nFluid = kdl->nFluid; |
| 100 | int nCells = kdl->nCells; |
| 101 | |
| 102 | uint32_t adjListIndex; |
| 103 | uint32_t * adjList = kdl->AdjList; |
| 104 | |
| 105 | |
| 106 | |
| 107 | #ifdef VTK_OUTPUT |
| 108 | if (cd->VtkOutput) { |
| 109 | kd->PdfsActive = src; |
| 110 | VtkWrite(ld, kd, cd, -1); |
| 111 | } |
| 112 | #endif |
| 113 | |
| 114 | #ifdef STATISTICS |
| 115 | kd->PdfsActive = src; |
| 116 | KernelStatistics(kd, ld, cd, 0); |
| 117 | #endif |
| 118 | |
| 119 | X_KERNEL_START(kernelData); |
| 120 | |
| 121 | // TODO: outer openmp parallel |
| 122 | for(int iter = 0; iter < maxIterations; iter += 2) { |
| 123 | |
| 124 | X_LIKWID_START("list-aa-ria-even"); |
| 125 | |
| 126 | // -------------------------------------------------------------------- |
| 127 | // even time step |
| 128 | // -------------------------------------------------------------------- |
| 129 | #ifdef _OPENMP |
| 130 | #pragma omp parallel for default(none) \ |
| 131 | shared(stderr, nFluid, nCells, kd, kdl, adjList, omegaOdd, omegaEven, src) \ |
| 132 | private(ux, uy, uz, dens, adjListIndex, evenPart, oddPart, dir_indep_trm, w_1_indep, w_2_indep, ui,\ |
| 133 | pdf_C, \ |
| 134 | pdf_N, pdf_E, pdf_S, pdf_W, \ |
| 135 | pdf_NE, pdf_SE, pdf_SW, pdf_NW, \ |
| 136 | pdf_T, pdf_TN, pdf_TE, pdf_TS, pdf_TW, \ |
| 137 | pdf_B, pdf_BN, pdf_BE, pdf_BS, pdf_BW) |
| 138 | #endif |
| 139 | #ifdef INTEL_OPT_DIRECTIVES |
| 140 | #pragma ivdep |
| 141 | #pragma vector always |
| 142 | #pragma simd |
| 143 | #endif |
| 144 | for (int index = 0; index < nFluid; ++index) { // LOOP list-aa-ria-even |
| 145 | |
| 146 | #define I(index, dir) P_INDEX_3((nCells), (index), (dir)) |
| 147 | |
| 148 | #define X(name, idx, idxinv, _x, _y, _z) JOIN(pdf_,name) = src[I(index, idx)]; |
| 149 | D3Q19_LIST |
| 150 | #undef X |
| 151 | |
| 152 | // #define LID_DRIVEN_CAVITY |
| 153 | |
| 154 | #ifdef LID_DRIVEN_CAVITY |
| 155 | int nX = kd->Dims[0]; |
| 156 | int nY = kd->Dims[1]; |
| 157 | int nZ = kd->Dims[2]; |
| 158 | |
| 159 | int x = kdl->Coords[C_INDEX_X(index)]; |
| 160 | int y = kdl->Coords[C_INDEX_Y(index)]; |
| 161 | int z = kdl->Coords[C_INDEX_Z(index)]; |
| 162 | |
| 163 | if (z == nZ - 4 && x > 3 && x < (nX - 4) && y > 3 && y < (nY - 4)) { |
| 164 | ux = F(0.1) * F(0.5)77; |
| 165 | uy = F(0.0); |
| 166 | uz = F(0.0); |
| 167 | } else { |
| 168 | #endif |
| 169 | ux = pdf_E + pdf_NE + pdf_SE + pdf_TE + pdf_BE - |
| 170 | pdf_W - pdf_NW - pdf_SW - pdf_TW - pdf_BW; |
| 171 | uy = pdf_N + pdf_NE + pdf_NW + pdf_TN + pdf_BN - |
| 172 | pdf_S - pdf_SE - pdf_SW - pdf_TS - pdf_BS; |
| 173 | uz = pdf_T + pdf_TE + pdf_TW + pdf_TN + pdf_TS - |
| 174 | pdf_B - pdf_BE - pdf_BW - pdf_BN - pdf_BS; |
| 175 | #ifdef LID_DRIVEN_CAVITY |
| 176 | } |
| 177 | #endif |
| 178 | |
| 179 | dens = pdf_C + |
| 180 | pdf_N + pdf_E + pdf_S + pdf_W + |
| 181 | pdf_NE + pdf_SE + pdf_SW + pdf_NW + |
| 182 | pdf_T + pdf_TN + pdf_TE + pdf_TS + pdf_TW + |
| 183 | pdf_B + pdf_BN + pdf_BE + pdf_BS + pdf_BW; |
| 184 | |
| 185 | dir_indep_trm = dens - (ux * ux + uy * uy + uz * uz)*F(3.0)/F(2.0); |
| 186 | |
| 187 | // direction: w_0 |
| 188 | src[I(index, D3Q19_C) ] = pdf_C - omegaEven*(pdf_C - w_0*dir_indep_trm); |
| 189 | |
| 190 | // direction: w_1 |
| 191 | w_1_indep = w_1*dir_indep_trm; |
| 192 | |
| 193 | ui = uy; |
| 194 | evenPart = omegaEven*( F(0.5)*(pdf_N + pdf_S) - ui*ui*w_1_nine_half - w_1_indep ); |
| 195 | oddPart = omegaOdd*(F(0.5)*(pdf_N - pdf_S) - ui*w_1_x3 ); |
| 196 | src[I(index, D3Q19_S)] = pdf_N - evenPart - oddPart; |
| 197 | src[I(index, D3Q19_N)] = pdf_S - evenPart + oddPart; |
| 198 | |
| 199 | ui = ux; |
| 200 | evenPart = omegaEven*( F(0.5)*(pdf_E + pdf_W) - ui*ui*w_1_nine_half - w_1_indep ); |
| 201 | oddPart = omegaOdd*(F(0.5)*(pdf_E - pdf_W) - ui*w_1_x3 ); |
| 202 | src[I(index, D3Q19_W)] = pdf_E - evenPart - oddPart; |
| 203 | src[I(index, D3Q19_E)] = pdf_W - evenPart + oddPart; |
| 204 | |
| 205 | ui = uz; |
| 206 | evenPart = omegaEven*( F(0.5)*(pdf_T + pdf_B) - ui*ui*w_1_nine_half - w_1_indep ); |
| 207 | oddPart = omegaOdd*(F(0.5)*(pdf_T - pdf_B) - ui*w_1_x3 ); |
| 208 | src[I(index, D3Q19_B)] = pdf_T - evenPart - oddPart; |
| 209 | src[I(index, D3Q19_T)] = pdf_B - evenPart + oddPart; |
| 210 | |
| 211 | // direction: w_2 |
| 212 | w_2_indep = w_2*dir_indep_trm; |
| 213 | |
| 214 | ui = -ux + uy; |
| 215 | evenPart = omegaEven*( F(0.5)*(pdf_NW + pdf_SE) - ui*ui*w_2_nine_half - w_2_indep ); |
| 216 | oddPart = omegaOdd*(F(0.5)*(pdf_NW - pdf_SE) - ui*w_2_x3 ); |
| 217 | src[I(index, D3Q19_SE)] = pdf_NW - evenPart - oddPart; |
| 218 | src[I(index, D3Q19_NW)] = pdf_SE - evenPart + oddPart; |
| 219 | |
| 220 | ui = ux + uy; |
| 221 | evenPart = omegaEven*( F(0.5)*(pdf_NE + pdf_SW) - ui*ui*w_2_nine_half - w_2_indep ); |
| 222 | oddPart = omegaOdd*(F(0.5)*(pdf_NE - pdf_SW) - ui*w_2_x3 ); |
| 223 | src[I(index, D3Q19_SW)] = pdf_NE - evenPart - oddPart; |
| 224 | src[I(index, D3Q19_NE)] = pdf_SW - evenPart + oddPart; |
| 225 | |
| 226 | ui = -ux + uz; |
| 227 | evenPart = omegaEven*( F(0.5)*(pdf_TW + pdf_BE) - ui*ui*w_2_nine_half - w_2_indep ); |
| 228 | oddPart = omegaOdd*(F(0.5)*(pdf_TW - pdf_BE) - ui*w_2_x3 ); |
| 229 | src[I(index, D3Q19_BE)] = pdf_TW - evenPart - oddPart; |
| 230 | src[I(index, D3Q19_TW)] = pdf_BE - evenPart + oddPart; |
| 231 | |
| 232 | ui = ux + uz; |
| 233 | evenPart = omegaEven*( F(0.5)*(pdf_TE + pdf_BW) - ui*ui*w_2_nine_half - w_2_indep ); |
| 234 | oddPart = omegaOdd*(F(0.5)*(pdf_TE - pdf_BW) - ui*w_2_x3 ); |
| 235 | src[I(index, D3Q19_BW)] = pdf_TE - evenPart - oddPart; |
| 236 | src[I(index, D3Q19_TE)] = pdf_BW - evenPart + oddPart; |
| 237 | |
| 238 | ui = -uy + uz; |
| 239 | evenPart = omegaEven*( F(0.5)*(pdf_TS + pdf_BN) - ui*ui*w_2_nine_half - w_2_indep ); |
| 240 | oddPart = omegaOdd*(F(0.5)*(pdf_TS - pdf_BN) - ui*w_2_x3 ); |
| 241 | src[I(index, D3Q19_BN)] = pdf_TS - evenPart - oddPart; |
| 242 | src[I(index, D3Q19_TS)] = pdf_BN - evenPart + oddPart; |
| 243 | |
| 244 | ui = uy + uz; |
| 245 | evenPart = omegaEven*( F(0.5)*(pdf_TN + pdf_BS) - ui*ui*w_2_nine_half - w_2_indep ); |
| 246 | oddPart = omegaOdd*(F(0.5)*(pdf_TN - pdf_BS) - ui*w_2_x3 ); |
| 247 | src[I(index, D3Q19_BS)] = pdf_TN - evenPart - oddPart; |
| 248 | src[I(index, D3Q19_TN)] = pdf_BS - evenPart + oddPart; |
| 249 | |
| 250 | } // (parallel) loop over fluid nodes |
| 251 | |
| 252 | X_LIKWID_STOP("list-aa-ria-even"); |
| 253 | |
| 254 | // save current iteration |
| 255 | kdl->Iteration = iter; |
| 256 | #ifdef VERIFICATION |
| 257 | kd->PdfsActive = src; |
| 258 | KernelAddBodyForce(kd, ld, cd); |
| 259 | #endif |
| 260 | |
| 261 | // -------------------------------------------------------------------- |
| 262 | // odd time step |
| 263 | // -------------------------------------------------------------------- |
| 264 | |
| 265 | X_LIKWID_START("list-aa-ria-odd"); |
| 266 | |
| 267 | #ifdef _OPENMP |
| 268 | #pragma omp parallel default(none) \ |
| 269 | shared(stderr, nFluid, nCells, kd, kdl, kdlr, adjList, omegaOdd, omegaEven, src, consecNodes, nConsecNodes) \ |
| 270 | private(ux, uy, uz, dens, adjListIndex, evenPart, oddPart, dir_indep_trm, w_1_indep, w_2_indep, ui,\ |
| 271 | pdf_C, \ |
| 272 | pdf_N, pdf_E, pdf_S, pdf_W, \ |
| 273 | pdf_NE, pdf_SE, pdf_SW, pdf_NW, \ |
| 274 | pdf_T, pdf_TN, pdf_TE, pdf_TS, pdf_TW, \ |
| 275 | pdf_B, pdf_BN, pdf_BE, pdf_BS, pdf_BW, \ |
| 276 | ppdf_C, \ |
| 277 | ppdf_N, ppdf_E, ppdf_S, ppdf_W, \ |
| 278 | ppdf_NE, ppdf_SE, ppdf_SW, ppdf_NW, \ |
| 279 | ppdf_T, ppdf_TN, ppdf_TE, ppdf_TS, ppdf_TW, \ |
| 280 | ppdf_B, ppdf_BN, ppdf_BE, ppdf_BS, ppdf_BW, \ |
| 281 | consecValue, consecIndex) |
| 282 | #endif |
| 283 | { |
| 284 | int threadId = 0; |
| 285 | |
| 286 | #ifdef _OPENMP |
| 287 | threadId = omp_get_thread_num(); |
| 288 | #endif |
| 289 | |
| 290 | consecIndex = kdlr->ConsecThreadIndices[threadId]; |
| 291 | consecValue = 0; |
| 292 | |
| 293 | int * threadIndices = kdlr->FluidNodeThreadIndices; |
| 294 | |
| 295 | int nFluidThread = threadIndices[threadId + 1] - threadIndices[threadId]; |
| 296 | |
| 297 | int indexStart = threadIndices[threadId]; |
| 298 | int indexStop = threadIndices[threadId] + nFluidThread; |
| 299 | |
| 300 | // Because of runlength coding iterations are not independent. |
| 301 | for (int index = indexStart; index < indexStop; ++index) { // LOOP list-aa-ria-odd |
| 302 | |
| 303 | #define I(index, dir) P_INDEX_3((nCells), (index), (dir)) |
| 304 | |
| 305 | #if 1 |
| 306 | if (consecValue > 0) { |
| 307 | --consecValue; |
| 308 | // Increment all pdf pointers. |
| 309 | #define X(name, idx, idxinv, _x, _y, _z) ++JOIN(ppdf_,name); |
| 310 | D3Q19_LIST |
| 311 | #undef X |
| 312 | } |
| 313 | else { |
| 314 | Assert(consecIndex < nConsecNodes); |
| 315 | |
| 316 | consecValue = consecNodes[consecIndex] - 1; |
| 317 | // Load new pointers to PDFs of local cell: |
| 318 | |
| 319 | adjListIndex = index * N_D3Q19_IDX; |
| 320 | |
| 321 | #define X(name, idx, idxinv, _x, _y, _z) JOIN(ppdf_,name) = &(src[adjList[adjListIndex + idxinv]]); |
| 322 | D3Q19_LIST_WO_C |
| 323 | #undef X |
| 324 | |
| 325 | ppdf_C = &(src[P_INDEX_3(nCells, index, D3Q19_C)]); |
| 326 | ++consecIndex; |
| 327 | } |
| 328 | |
| 329 | #define X(name, idx, idxinv, _x, _y, _z) JOIN(pdf_,name) = *JOIN(ppdf_,name); |
| 330 | D3Q19_LIST |
| 331 | #undef X |
| 332 | #else |
| 333 | adjListIndex = index * N_D3Q19_IDX; |
| 334 | |
| 335 | // Load PDFs of local cell: pdf_N = src[adjList[adjListIndex + D3Q19_S]]; ... |
| 336 | pdf_C = src[P_INDEX_3(nCells, index, D3Q19_C)]; |
| 337 | |
| 338 | #define X(name, idx, idxinv, _x, _y, _z) JOIN(ppdf_,name) = &(src[adjList[adjListIndex + idxinv]]); |
| 339 | D3Q19_LIST_WO_C |
| 340 | #undef X |
| 341 | |
| 342 | #define X(name, idx, idxinv, _x, _y, _z) JOIN(pdf_,name) = src[adjList[adjListIndex + idxinv]]; |
| 343 | D3Q19_LIST_WO_C |
| 344 | #undef X |
| 345 | #endif |
| 346 | |
| 347 | #ifdef LID_DRIVEN_CAVITY |
| 348 | int nX = kd->Dims[0]; |
| 349 | int nY = kd->Dims[1]; |
| 350 | int nZ = kd->Dims[2]; |
| 351 | |
| 352 | int x = kdl->Coords[C_INDEX_X(index)]; |
| 353 | int y = kdl->Coords[C_INDEX_Y(index)]; |
| 354 | int z = kdl->Coords[C_INDEX_Z(index)]; |
| 355 | |
| 356 | if (z == nZ - 4 && x > 3 && x < (nX - 4) && y > 3 && y < (nY - 4)) { |
| 357 | ux = F(0.1) * F(0.5)77; |
| 358 | uy = F(0.0); |
| 359 | uz = F(0.0); |
| 360 | } else { |
| 361 | #endif |
| 362 | ux = pdf_E + pdf_NE + pdf_SE + pdf_TE + pdf_BE - |
| 363 | pdf_W - pdf_NW - pdf_SW - pdf_TW - pdf_BW; |
| 364 | uy = pdf_N + pdf_NE + pdf_NW + pdf_TN + pdf_BN - |
| 365 | pdf_S - pdf_SE - pdf_SW - pdf_TS - pdf_BS; |
| 366 | uz = pdf_T + pdf_TE + pdf_TW + pdf_TN + pdf_TS - |
| 367 | pdf_B - pdf_BE - pdf_BW - pdf_BN - pdf_BS; |
| 368 | #ifdef LID_DRIVEN_CAVITY |
| 369 | } |
| 370 | #endif |
| 371 | |
| 372 | dens = pdf_C + |
| 373 | pdf_N + pdf_E + pdf_S + pdf_W + |
| 374 | pdf_NE + pdf_SE + pdf_SW + pdf_NW + |
| 375 | pdf_T + pdf_TN + pdf_TE + pdf_TS + pdf_TW + |
| 376 | pdf_B + pdf_BN + pdf_BE + pdf_BS + pdf_BW; |
| 377 | |
| 378 | dir_indep_trm = dens - (ux * ux + uy * uy + uz * uz)*F(3.0)/F(2.0); |
| 379 | |
| 380 | adjListIndex = index * N_D3Q19_IDX; |
| 381 | |
| 382 | // direction: w_0 |
| 383 | src[I(index, D3Q19_C) ] = pdf_C - omegaEven*(pdf_C - w_0*dir_indep_trm); |
| 384 | |
| 385 | // direction: w_1 |
| 386 | w_1_indep = w_1*dir_indep_trm; |
| 387 | |
| 388 | ui = uy; |
| 389 | evenPart = omegaEven*( F(0.5)*(pdf_N + pdf_S) - ui*ui*w_1_nine_half - w_1_indep ); |
| 390 | oddPart = omegaOdd*(F(0.5)*(pdf_N - pdf_S) - ui*w_1_x3 ); |
| 391 | *ppdf_S = pdf_N - evenPart - oddPart; |
| 392 | *ppdf_N = pdf_S - evenPart + oddPart; |
| 393 | |
| 394 | ui = ux; |
| 395 | evenPart = omegaEven*( F(0.5)*(pdf_E + pdf_W) - ui*ui*w_1_nine_half - w_1_indep ); |
| 396 | oddPart = omegaOdd*(F(0.5)*(pdf_E - pdf_W) - ui*w_1_x3 ); |
| 397 | *ppdf_W = pdf_E - evenPart - oddPart; |
| 398 | *ppdf_E = pdf_W - evenPart + oddPart; |
| 399 | |
| 400 | ui = uz; |
| 401 | evenPart = omegaEven*( F(0.5)*(pdf_T + pdf_B) - ui*ui*w_1_nine_half - w_1_indep ); |
| 402 | oddPart = omegaOdd*(F(0.5)*(pdf_T - pdf_B) - ui*w_1_x3 ); |
| 403 | *ppdf_B = pdf_T - evenPart - oddPart; |
| 404 | *ppdf_T = pdf_B - evenPart + oddPart; |
| 405 | |
| 406 | // direction: w_2 |
| 407 | w_2_indep = w_2*dir_indep_trm; |
| 408 | |
| 409 | ui = -ux + uy; |
| 410 | evenPart = omegaEven*( F(0.5)*(pdf_NW + pdf_SE) - ui*ui*w_2_nine_half - w_2_indep ); |
| 411 | oddPart = omegaOdd*(F(0.5)*(pdf_NW - pdf_SE) - ui*w_2_x3 ); |
| 412 | *ppdf_SE = pdf_NW - evenPart - oddPart; |
| 413 | *ppdf_NW = pdf_SE - evenPart + oddPart; |
| 414 | |
| 415 | ui = ux + uy; |
| 416 | evenPart = omegaEven*( F(0.5)*(pdf_NE + pdf_SW) - ui*ui*w_2_nine_half - w_2_indep ); |
| 417 | oddPart = omegaOdd*(F(0.5)*(pdf_NE - pdf_SW) - ui*w_2_x3 ); |
| 418 | *ppdf_SW = pdf_NE - evenPart - oddPart; |
| 419 | *ppdf_NE = pdf_SW - evenPart + oddPart; |
| 420 | |
| 421 | ui = -ux + uz; |
| 422 | evenPart = omegaEven*( F(0.5)*(pdf_TW + pdf_BE) - ui*ui*w_2_nine_half - w_2_indep ); |
| 423 | oddPart = omegaOdd*(F(0.5)*(pdf_TW - pdf_BE) - ui*w_2_x3 ); |
| 424 | *ppdf_BE = pdf_TW - evenPart - oddPart; |
| 425 | *ppdf_TW = pdf_BE - evenPart + oddPart; |
| 426 | |
| 427 | ui = ux + uz; |
| 428 | evenPart = omegaEven*( F(0.5)*(pdf_TE + pdf_BW) - ui*ui*w_2_nine_half - w_2_indep ); |
| 429 | oddPart = omegaOdd*(F(0.5)*(pdf_TE - pdf_BW) - ui*w_2_x3 ); |
| 430 | *ppdf_BW = pdf_TE - evenPart - oddPart; |
| 431 | *ppdf_TE = pdf_BW - evenPart + oddPart; |
| 432 | |
| 433 | ui = -uy + uz; |
| 434 | evenPart = omegaEven*( F(0.5)*(pdf_TS + pdf_BN) - ui*ui*w_2_nine_half - w_2_indep ); |
| 435 | oddPart = omegaOdd*(F(0.5)*(pdf_TS - pdf_BN) - ui*w_2_x3 ); |
| 436 | *ppdf_BN = pdf_TS - evenPart - oddPart; |
| 437 | *ppdf_TS = pdf_BN - evenPart + oddPart; |
| 438 | |
| 439 | ui = uy + uz; |
| 440 | evenPart = omegaEven*( F(0.5)*(pdf_TN + pdf_BS) - ui*ui*w_2_nine_half - w_2_indep ); |
| 441 | oddPart = omegaOdd*(F(0.5)*(pdf_TN - pdf_BS) - ui*w_2_x3 ); |
| 442 | *ppdf_BS = pdf_TN - evenPart - oddPart; |
| 443 | *ppdf_TN = pdf_BS - evenPart + oddPart; |
| 444 | |
| 445 | #undef I |
| 446 | } // loop over fluid nodes |
| 447 | } // end pragma omp parallel |
| 448 | |
| 449 | X_LIKWID_STOP("list-aa-ria-odd"); |
| 450 | |
| 451 | // save current iteration |
| 452 | kdl->Iteration = iter + 1; |
| 453 | |
| 454 | #ifdef VERIFICATION |
| 455 | kd->PdfsActive = src; |
| 456 | KernelAddBodyForce(kd, ld, cd); |
| 457 | #endif |
| 458 | |
| 459 | #ifdef VTK_OUTPUT |
| 460 | if (cd->VtkOutput && (iter % cd->VtkModulus) == 0) { |
| 461 | kd->PdfsActive = src; |
| 462 | VtkWrite(ld, kd, cd, iter); |
| 463 | } |
| 464 | #endif |
| 465 | |
| 466 | #ifdef STATISTICS |
| 467 | kd->PdfsActive = src; |
| 468 | KernelStatistics(kd, ld, cd, iter); |
| 469 | #endif |
| 470 | |
| 471 | |
| 472 | } // for (int iter = 0; ... |
| 473 | |
| 474 | X_KERNEL_END(kernelData); |
| 475 | |
| 476 | #ifdef VTK_OUTPUT |
| 477 | if (cd->VtkOutput) { |
| 478 | kd->PdfsActive = src; |
| 479 | VtkWrite(ld, kd, cd, maxIterations); |
| 480 | } |
| 481 | #endif |
| 482 | |
| 483 | #ifdef STATISTICS |
| 484 | kd->PdfsActive = src; |
| 485 | KernelStatistics(kd, ld, cd, maxIterations); |
| 486 | #endif |
| 487 | |
| 488 | return; |
| 489 | } |