1 // Copyright (c) 2017-2026, 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 HIP offset element restriction kernels
10 #include <ceed/types.h>
11
12 //------------------------------------------------------------------------------
13 // E-vector -> L-vector, standard (with offsets)
14 //------------------------------------------------------------------------------
15 #if !USE_DETERMINISTIC
AtPointsTranspose(const CeedInt * __restrict__ indices,const CeedInt * __restrict__ points_per_elem,const CeedScalar * __restrict__ u,CeedScalar * __restrict__ v)16 extern "C" __global__ void AtPointsTranspose(const CeedInt *__restrict__ indices, const CeedInt *__restrict__ points_per_elem,
17 const CeedScalar *__restrict__ u, CeedScalar *__restrict__ v) {
18 for (CeedInt node = blockIdx.x * blockDim.x + threadIdx.x; node < RSTR_NUM_ELEM * RSTR_ELEM_SIZE; node += blockDim.x * gridDim.x) {
19 const CeedInt ind = indices[node];
20 const CeedInt loc_node = node % RSTR_ELEM_SIZE;
21 const CeedInt elem = node / RSTR_ELEM_SIZE;
22
23 if (loc_node >= points_per_elem[elem]) continue;
24 for (CeedInt comp = 0; comp < RSTR_NUM_COMP; comp++) {
25 atomicAdd(&v[ind + comp * RSTR_COMP_STRIDE], u[loc_node + comp * RSTR_ELEM_SIZE * RSTR_NUM_ELEM + elem * RSTR_ELEM_SIZE]);
26 }
27 }
28 }
29 #else
AtPointsTranspose(const CeedInt * __restrict__ l_vec_indices,const CeedInt * __restrict__ t_indices,const CeedInt * __restrict__ points_per_elem,const CeedInt * __restrict__ t_offsets,const CeedScalar * __restrict__ u,CeedScalar * __restrict__ v)30 extern "C" __global__ void AtPointsTranspose(const CeedInt *__restrict__ l_vec_indices, const CeedInt *__restrict__ t_indices,
31 const CeedInt *__restrict__ points_per_elem, const CeedInt *__restrict__ t_offsets,
32 const CeedScalar *__restrict__ u, CeedScalar *__restrict__ v) {
33 CeedScalar value[RSTR_NUM_COMP];
34
35 for (CeedInt i = blockIdx.x * blockDim.x + threadIdx.x; i < RSTR_NUM_NODES; i += blockDim.x * gridDim.x) {
36 const CeedInt ind = l_vec_indices[i];
37 const CeedInt range_1 = t_offsets[i];
38 const CeedInt range_N = t_offsets[i + 1];
39
40 for (CeedInt comp = 0; comp < RSTR_NUM_COMP; comp++) value[comp] = 0.0;
41
42 for (CeedInt j = range_1; j < range_N; j++) {
43 const CeedInt t_ind = t_indices[j];
44 const CeedInt loc_node = t_ind % RSTR_ELEM_SIZE;
45 const CeedInt elem = t_ind / RSTR_ELEM_SIZE;
46
47 if (loc_node >= points_per_elem[elem]) continue;
48 for (CeedInt comp = 0; comp < RSTR_NUM_COMP; comp++) {
49 value[comp] += u[loc_node + comp * RSTR_ELEM_SIZE * RSTR_NUM_ELEM + elem * RSTR_ELEM_SIZE];
50 }
51 }
52
53 for (CeedInt comp = 0; comp < RSTR_NUM_COMP; comp++) v[ind + comp * RSTR_COMP_STRIDE] += value[comp];
54 }
55 }
56 #endif
57