1 // Copyright (c) 2017-2024, Lawrence Livermore National Security, LLC and other CEED contributors. 2 // All Rights Reserved. See the top-level LICENSE and NOTICE files for details. 3 // 4 // SPDX-License-Identifier: BSD-2-Clause 5 // 6 // This file is part of CEED: http://github.com/ceed 7 8 /// @file 9 /// Internal header for CUDA backend macro and type definitions for JiT source 10 #include <ceed/types.h> 11 12 //------------------------------------------------------------------------------ 13 // Load matrices for basis actions 14 //------------------------------------------------------------------------------ 15 template <int P, int Q> 16 inline __device__ void LoadMatrix(SharedData_Cuda &data, const CeedScalar *__restrict__ d_B, CeedScalar *B) { 17 for (CeedInt i = data.t_id; i < P * Q; i += blockDim.x * blockDim.y * blockDim.z) B[i] = d_B[i]; 18 } 19 20 //------------------------------------------------------------------------------ 21 // AtPoints 22 //------------------------------------------------------------------------------ 23 24 //------------------------------------------------------------------------------ 25 // L-vector -> single point 26 //------------------------------------------------------------------------------ 27 template <int NUM_COMP, int COMP_STRIDE, int NUM_PTS> 28 inline __device__ void ReadPoint(SharedData_Cuda &data, const CeedInt elem, const CeedInt p, const CeedInt points_in_elem, 29 const CeedInt *__restrict__ indices, const CeedScalar *__restrict__ d_u, CeedScalar *r_u) { 30 const CeedInt ind = indices[p + elem * NUM_PTS]; 31 32 for (CeedInt comp = 0; comp < NUM_COMP; comp++) { 33 r_u[comp] = d_u[ind + comp * COMP_STRIDE]; 34 } 35 } 36 37 //------------------------------------------------------------------------------ 38 // Single point -> L-vector 39 //------------------------------------------------------------------------------ 40 template <int NUM_COMP, int COMP_STRIDE, int NUM_PTS> 41 inline __device__ void WritePoint(SharedData_Cuda &data, const CeedInt elem, const CeedInt p, const CeedInt points_in_elem, 42 const CeedInt *__restrict__ indices, const CeedScalar *__restrict__ r_u, CeedScalar *d_u) { 43 if (p < points_in_elem) { 44 const CeedInt ind = indices[p + elem * NUM_PTS]; 45 46 for (CeedInt comp = 0; comp < NUM_COMP; comp++) { 47 d_u[ind + comp * COMP_STRIDE] += r_u[comp]; 48 } 49 } 50 } 51 52 //------------------------------------------------------------------------------ 53 // 1D 54 //------------------------------------------------------------------------------ 55 56 //------------------------------------------------------------------------------ 57 // L-vector -> E-vector, offsets provided 58 //------------------------------------------------------------------------------ 59 template <int NUM_COMP, int COMP_STRIDE, int P_1d> 60 inline __device__ void ReadLVecStandard1d(SharedData_Cuda &data, const CeedInt num_nodes, const CeedInt elem, const CeedInt *__restrict__ indices, 61 const CeedScalar *__restrict__ d_u, CeedScalar *__restrict__ r_u) { 62 if (data.t_id_x < P_1d) { 63 const CeedInt node = data.t_id_x; 64 const CeedInt ind = indices[node + elem * P_1d]; 65 66 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[comp] = d_u[ind + COMP_STRIDE * comp]; 67 } 68 } 69 70 //------------------------------------------------------------------------------ 71 // L-vector -> E-vector, strided 72 //------------------------------------------------------------------------------ 73 template <int NUM_COMP, int P_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 74 inline __device__ void ReadLVecStrided1d(SharedData_Cuda &data, const CeedInt elem, const CeedScalar *__restrict__ d_u, 75 CeedScalar *__restrict__ r_u) { 76 if (data.t_id_x < P_1d) { 77 const CeedInt node = data.t_id_x; 78 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 79 80 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[comp] = d_u[ind + comp * STRIDES_COMP]; 81 } 82 } 83 84 //------------------------------------------------------------------------------ 85 // E-vector -> L-vector, offsets provided 86 //------------------------------------------------------------------------------ 87 template <int NUM_COMP, int COMP_STRIDE, int P_1d> 88 inline __device__ void WriteLVecStandard1d(SharedData_Cuda &data, const CeedInt num_nodes, const CeedInt elem, const CeedInt *__restrict__ indices, 89 const CeedScalar *__restrict__ r_v, CeedScalar *__restrict__ d_v) { 90 if (data.t_id_x < P_1d) { 91 const CeedInt node = data.t_id_x; 92 const CeedInt ind = indices[node + elem * P_1d]; 93 94 for (CeedInt comp = 0; comp < NUM_COMP; comp++) atomicAdd(&d_v[ind + COMP_STRIDE * comp], r_v[comp]); 95 } 96 } 97 98 //------------------------------------------------------------------------------ 99 // E-vector -> L-vector, strided 100 //------------------------------------------------------------------------------ 101 template <int NUM_COMP, int P_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 102 inline __device__ void WriteLVecStrided1d(SharedData_Cuda &data, const CeedInt elem, const CeedScalar *__restrict__ r_v, 103 CeedScalar *__restrict__ d_v) { 104 if (data.t_id_x < P_1d) { 105 const CeedInt node = data.t_id_x; 106 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 107 108 for (CeedInt comp = 0; comp < NUM_COMP; comp++) d_v[ind + comp * STRIDES_COMP] += r_v[comp]; 109 } 110 } 111 112 //------------------------------------------------------------------------------ 113 // 2D 114 //------------------------------------------------------------------------------ 115 116 //------------------------------------------------------------------------------ 117 // L-vector -> E-vector, offsets provided 118 //------------------------------------------------------------------------------ 119 template <int NUM_COMP, int COMP_STRIDE, int P_1d> 120 inline __device__ void ReadLVecStandard2d(SharedData_Cuda &data, const CeedInt num_nodes, const CeedInt elem, const CeedInt *__restrict__ indices, 121 const CeedScalar *__restrict__ d_u, CeedScalar *__restrict__ r_u) { 122 if (data.t_id_x < P_1d && data.t_id_y < P_1d) { 123 const CeedInt node = data.t_id_x + data.t_id_y * P_1d; 124 const CeedInt ind = indices[node + elem * P_1d * P_1d]; 125 126 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[comp] = d_u[ind + COMP_STRIDE * comp]; 127 } 128 } 129 130 //------------------------------------------------------------------------------ 131 // L-vector -> E-vector, strided 132 //------------------------------------------------------------------------------ 133 template <int NUM_COMP, int P_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 134 inline __device__ void ReadLVecStrided2d(SharedData_Cuda &data, const CeedInt elem, const CeedScalar *__restrict__ d_u, 135 CeedScalar *__restrict__ r_u) { 136 if (data.t_id_x < P_1d && data.t_id_y < P_1d) { 137 const CeedInt node = data.t_id_x + data.t_id_y * P_1d; 138 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 139 140 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[comp] = d_u[ind + comp * STRIDES_COMP]; 141 } 142 } 143 144 //------------------------------------------------------------------------------ 145 // E-vector -> L-vector, offsets provided 146 //------------------------------------------------------------------------------ 147 template <int NUM_COMP, int COMP_STRIDE, int P_1d> 148 inline __device__ void WriteLVecStandard2d(SharedData_Cuda &data, const CeedInt num_nodes, const CeedInt elem, const CeedInt *__restrict__ indices, 149 const CeedScalar *__restrict__ r_v, CeedScalar *__restrict__ d_v) { 150 if (data.t_id_x < P_1d && data.t_id_y < P_1d) { 151 const CeedInt node = data.t_id_x + data.t_id_y * P_1d; 152 const CeedInt ind = indices[node + elem * P_1d * P_1d]; 153 154 for (CeedInt comp = 0; comp < NUM_COMP; comp++) atomicAdd(&d_v[ind + COMP_STRIDE * comp], r_v[comp]); 155 } 156 } 157 158 //------------------------------------------------------------------------------ 159 // E-vector -> L-vector, strided 160 //------------------------------------------------------------------------------ 161 template <int NUM_COMP, int P_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 162 inline __device__ void WriteLVecStrided2d(SharedData_Cuda &data, const CeedInt elem, const CeedScalar *__restrict__ r_v, 163 CeedScalar *__restrict__ d_v) { 164 if (data.t_id_x < P_1d && data.t_id_y < P_1d) { 165 const CeedInt node = data.t_id_x + data.t_id_y * P_1d; 166 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 167 168 for (CeedInt comp = 0; comp < NUM_COMP; comp++) d_v[ind + comp * STRIDES_COMP] += r_v[comp]; 169 } 170 } 171 172 //------------------------------------------------------------------------------ 173 // 3D 174 //------------------------------------------------------------------------------ 175 176 //------------------------------------------------------------------------------ 177 // L-vector -> E-vector, offsets provided 178 //------------------------------------------------------------------------------ 179 template <int NUM_COMP, int COMP_STRIDE, int P_1d> 180 inline __device__ void ReadLVecStandard3d(SharedData_Cuda &data, const CeedInt num_nodes, const CeedInt elem, const CeedInt *__restrict__ indices, 181 const CeedScalar *__restrict__ d_u, CeedScalar *__restrict__ r_u) { 182 if (data.t_id_x < P_1d && data.t_id_y < P_1d) 183 for (CeedInt z = 0; z < P_1d; z++) { 184 const CeedInt node = data.t_id_x + data.t_id_y * P_1d + z * P_1d * P_1d; 185 const CeedInt ind = indices[node + elem * P_1d * P_1d * P_1d]; 186 187 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[z + comp * P_1d] = d_u[ind + COMP_STRIDE * comp]; 188 } 189 } 190 191 //------------------------------------------------------------------------------ 192 // L-vector -> E-vector, strided 193 //------------------------------------------------------------------------------ 194 template <int NUM_COMP, int P_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 195 inline __device__ void ReadLVecStrided3d(SharedData_Cuda &data, const CeedInt elem, const CeedScalar *__restrict__ d_u, 196 CeedScalar *__restrict__ r_u) { 197 if (data.t_id_x < P_1d && data.t_id_y < P_1d) 198 for (CeedInt z = 0; z < P_1d; z++) { 199 const CeedInt node = data.t_id_x + data.t_id_y * P_1d + z * P_1d * P_1d; 200 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 201 202 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[z + comp * P_1d] = d_u[ind + comp * STRIDES_COMP]; 203 } 204 } 205 206 //------------------------------------------------------------------------------ 207 // E-vector -> Q-vector, offests provided 208 //------------------------------------------------------------------------------ 209 template <int NUM_COMP, int COMP_STRIDE, int Q_1d> 210 inline __device__ void ReadEVecSliceStandard3d(SharedData_Cuda &data, const CeedInt nquads, const CeedInt elem, const CeedInt q, 211 const CeedInt *__restrict__ indices, const CeedScalar *__restrict__ d_u, 212 CeedScalar *__restrict__ r_u) { 213 if (data.t_id_x < Q_1d && data.t_id_y < Q_1d) { 214 const CeedInt node = data.t_id_x + data.t_id_y * Q_1d + q * Q_1d * Q_1d; 215 const CeedInt ind = indices[node + elem * Q_1d * Q_1d * Q_1d]; 216 217 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[comp] = d_u[ind + COMP_STRIDE * comp]; 218 } 219 } 220 221 //------------------------------------------------------------------------------ 222 // E-vector -> Q-vector, strided 223 //------------------------------------------------------------------------------ 224 template <int NUM_COMP, int Q_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 225 inline __device__ void ReadEVecSliceStrided3d(SharedData_Cuda &data, const CeedInt elem, const CeedInt q, const CeedScalar *__restrict__ d_u, 226 CeedScalar *__restrict__ r_u) { 227 if (data.t_id_x < Q_1d && data.t_id_y < Q_1d) { 228 const CeedInt node = data.t_id_x + data.t_id_y * Q_1d + q * Q_1d * Q_1d; 229 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 230 231 for (CeedInt comp = 0; comp < NUM_COMP; comp++) r_u[comp] = d_u[ind + comp * STRIDES_COMP]; 232 } 233 } 234 235 //------------------------------------------------------------------------------ 236 // E-vector -> L-vector, offsets provided 237 //------------------------------------------------------------------------------ 238 template <int NUM_COMP, int COMP_STRIDE, int P_1d> 239 inline __device__ void WriteLVecStandard3d(SharedData_Cuda &data, const CeedInt num_nodes, const CeedInt elem, const CeedInt *__restrict__ indices, 240 const CeedScalar *__restrict__ r_v, CeedScalar *__restrict__ d_v) { 241 if (data.t_id_x < P_1d && data.t_id_y < P_1d) 242 for (CeedInt z = 0; z < P_1d; z++) { 243 const CeedInt node = data.t_id_x + data.t_id_y * P_1d + z * P_1d * P_1d; 244 const CeedInt ind = indices[node + elem * P_1d * P_1d * P_1d]; 245 246 for (CeedInt comp = 0; comp < NUM_COMP; comp++) atomicAdd(&d_v[ind + COMP_STRIDE * comp], r_v[z + comp * P_1d]); 247 } 248 } 249 250 //------------------------------------------------------------------------------ 251 // E-vector -> L-vector, strided 252 //------------------------------------------------------------------------------ 253 template <int NUM_COMP, int P_1d, int STRIDES_NODE, int STRIDES_COMP, int STRIDES_ELEM> 254 inline __device__ void WriteLVecStrided3d(SharedData_Cuda &data, const CeedInt elem, const CeedScalar *__restrict__ r_v, 255 CeedScalar *__restrict__ d_v) { 256 if (data.t_id_x < P_1d && data.t_id_y < P_1d) 257 for (CeedInt z = 0; z < P_1d; z++) { 258 const CeedInt node = data.t_id_x + data.t_id_y * P_1d + z * P_1d * P_1d; 259 const CeedInt ind = node * STRIDES_NODE + elem * STRIDES_ELEM; 260 261 for (CeedInt comp = 0; comp < NUM_COMP; comp++) d_v[ind + comp * STRIDES_COMP] += r_v[z + comp * P_1d]; 262 } 263 } 264 265 //------------------------------------------------------------------------------ 266 // 3D collocated derivatives computation 267 //------------------------------------------------------------------------------ 268 template <int NUM_COMP, int Q_1d> 269 inline __device__ void GradColloSlice3d(SharedData_Cuda &data, const CeedInt q, const CeedScalar *__restrict__ r_U, const CeedScalar *c_G, 270 CeedScalar *__restrict__ r_V) { 271 if (data.t_id_x < Q_1d && data.t_id_y < Q_1d) { 272 for (CeedInt comp = 0; comp < NUM_COMP; comp++) { 273 data.slice[data.t_id_x + data.t_id_y * T_1D] = r_U[q + comp * Q_1d]; 274 __syncthreads(); 275 // X derivative 276 r_V[comp + 0 * NUM_COMP] = 0.0; 277 for (CeedInt i = 0; i < Q_1d; i++) 278 r_V[comp + 0 * NUM_COMP] += c_G[i + data.t_id_x * Q_1d] * data.slice[i + data.t_id_y * T_1D]; // Contract x direction (X derivative) 279 // Y derivative 280 r_V[comp + 1 * NUM_COMP] = 0.0; 281 for (CeedInt i = 0; i < Q_1d; i++) 282 r_V[comp + 1 * NUM_COMP] += c_G[i + data.t_id_y * Q_1d] * data.slice[data.t_id_x + i * T_1D]; // Contract y direction (Y derivative) 283 // Z derivative 284 r_V[comp + 2 * NUM_COMP] = 0.0; 285 for (CeedInt i = 0; i < Q_1d; i++) r_V[comp + 2 * NUM_COMP] += c_G[i + q * Q_1d] * r_U[i + comp * Q_1d]; // Contract z direction (Z derivative) 286 __syncthreads(); 287 } 288 } 289 } 290 291 //------------------------------------------------------------------------------ 292 // 3D collocated derivatives transpose 293 //------------------------------------------------------------------------------ 294 template <int NUM_COMP, int Q_1d> 295 inline __device__ void GradColloSliceTranspose3d(SharedData_Cuda &data, const CeedInt q, const CeedScalar *__restrict__ r_U, const CeedScalar *c_G, 296 CeedScalar *__restrict__ r_V) { 297 if (data.t_id_x < Q_1d && data.t_id_y < Q_1d) { 298 for (CeedInt comp = 0; comp < NUM_COMP; comp++) { 299 // X derivative 300 data.slice[data.t_id_x + data.t_id_y * T_1D] = r_U[comp + 0 * NUM_COMP]; 301 __syncthreads(); 302 for (CeedInt i = 0; i < Q_1d; i++) 303 r_V[q + comp * Q_1d] += c_G[data.t_id_x + i * Q_1d] * data.slice[i + data.t_id_y * T_1D]; // Contract x direction (X derivative) 304 __syncthreads(); 305 // Y derivative 306 data.slice[data.t_id_x + data.t_id_y * T_1D] = r_U[comp + 1 * NUM_COMP]; 307 __syncthreads(); 308 for (CeedInt i = 0; i < Q_1d; i++) 309 r_V[q + comp * Q_1d] += c_G[data.t_id_y + i * Q_1d] * data.slice[data.t_id_x + i * T_1D]; // Contract y direction (Y derivative) 310 __syncthreads(); 311 // Z derivative 312 for (CeedInt i = 0; i < Q_1d; i++) 313 r_V[i + comp * Q_1d] += c_G[i + q * Q_1d] * r_U[comp + 2 * NUM_COMP]; // PARTIAL contract z direction (Z derivative) 314 } 315 } 316 } 317