64-BitBrainstorm_2026/AMSS-NCKU
8
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Stabilize GPU AMR prolong/restrict paths

Browse Source
This commit is contained in:
CGH0S7
2026-04-10 21:57:58 +08:00
parent c578a15ecd
commit a089041c3b
3 changed files with 453 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
AMSS_NCKU_source/Parallel.C
View File

@@ -14,6 +14,11 @@ extern int bssn_cuda_prolong3_pack(int wei,
const double *llbf, const double *uubf, const int *extf, double *funf,
const double *llbp, const double *uubp,
const double *SoA, int symmetry) __attribute__((weak));
extern int bssn_cuda_restrict3_pack(int wei,
const double *llbc, const double *uubc, const int *extc, double *func,
const double *llbf, const double *uubf, const int *extf, const double *funf,
const double *llbr, const double *uubr,
const double *SoA, int symmetry) __attribute__((weak));
#endif
namespace {
@@ -4048,11 +4053,18 @@ int Parallel::data_packer(double *data, MyList<Parallel::gridseg> *src, MyList<P
src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
dst->data->llb, dst->data->uub);
break;
case 2:
f_restrict3(DIM, dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
dst->data->llb, dst->data->uub, varls->data->SoA, Symmetry);
break;
case 2:
if (!bssn_cuda_restrict3_pack ||
bssn_cuda_restrict3_pack(DIM,
dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
dst->data->llb, dst->data->uub, varls->data->SoA, Symmetry))
{
f_restrict3(DIM, dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
dst->data->llb, dst->data->uub, varls->data->SoA, Symmetry);
}
break;
case 3:
if (!bssn_cuda_prolong3_pack ||
bssn_cuda_prolong3_pack(DIM,