17d8d0e25Snbeams // Copyright (c) 2017-2018, Lawrence Livermore National Security, LLC. 27d8d0e25Snbeams // Produced at the Lawrence Livermore National Laboratory. LLNL-CODE-734707. 37d8d0e25Snbeams // All Rights reserved. See files LICENSE and NOTICE for details. 47d8d0e25Snbeams // 57d8d0e25Snbeams // This file is part of CEED, a collection of benchmarks, miniapps, software 67d8d0e25Snbeams // libraries and APIs for efficient high-order finite element and spectral 77d8d0e25Snbeams // element discretizations for exascale applications. For more information and 87d8d0e25Snbeams // source code availability see http://github.com/ceed. 97d8d0e25Snbeams // 107d8d0e25Snbeams // The CEED research is supported by the Exascale Computing Project 17-SC-20-SC, 117d8d0e25Snbeams // a collaborative effort of two U.S. Department of Energy organizations (Office 127d8d0e25Snbeams // of Science and the National Nuclear Security Administration) responsible for 137d8d0e25Snbeams // the planning and preparation of a capable exascale ecosystem, including 147d8d0e25Snbeams // software, applications, hardware, advanced system engineering and early 157d8d0e25Snbeams // testbed platforms, in support of the nation's exascale computing imperative. 167d8d0e25Snbeams #include "../hip/ceed-hip.h" 177d8d0e25Snbeams 187d8d0e25Snbeams typedef struct { 197d8d0e25Snbeams hipModule_t module; 207d8d0e25Snbeams hipFunction_t interp; 217d8d0e25Snbeams hipFunction_t grad; 227d8d0e25Snbeams hipFunction_t weight; 23*9e31c45bSnbeams CeedInt blksizes[3]; // interp, grad, weight thread block sizes 247d8d0e25Snbeams CeedScalar *d_interp1d; 257d8d0e25Snbeams CeedScalar *d_grad1d; 267d8d0e25Snbeams CeedScalar *d_collograd1d; 277d8d0e25Snbeams CeedScalar *d_qweight1d; 287d8d0e25Snbeams CeedScalar *c_B; 297d8d0e25Snbeams CeedScalar *c_G; 307d8d0e25Snbeams } CeedBasis_Hip_shared; 317d8d0e25Snbeams 327d8d0e25Snbeams typedef struct { 337d8d0e25Snbeams Ceed_Hip base; 347d8d0e25Snbeams } Ceed_Hip_shared; 357d8d0e25Snbeams 367d8d0e25Snbeams CEED_INTERN int CeedBasisCreateTensorH1_Hip_shared(CeedInt dim, CeedInt P1d, 377d8d0e25Snbeams CeedInt Q1d, const CeedScalar *interp1d, const CeedScalar *grad1d, 387d8d0e25Snbeams const CeedScalar *qref1d, const CeedScalar *qweight1d, CeedBasis basis); 39