ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Re-arranged memory allocation for cgrid_brick, some issues need to be fixed

Browse Source
This commit is contained in:
Trung Nguyen
2022-09-11 18:58:34 -05:00
parent 5e59c95be4
commit 7f4efa380a
4 changed files with 69 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
lib/gpu/lal_amoeba.cu
View File

@ -1646,7 +1646,7 @@ __kernel void k_fphi_uind(const __global numtyp4 *restrict x_,
//numtyp4 ix; fetch4(ix,ii,pos_tex); //x_[i];
acctyp fdip_buf[32];
int j,k,m;
int j,k;
int nlpts = (bsorder-1) / 2;
// extract the permanent multipole field at each site

11
lib/gpu/lal_amoeba_ext.cpp
View File

@ -164,13 +164,14 @@ void amoeba_gpu_compute_polar_real(int *host_amtype, int *host_amgroup, double *
void amoeba_gpu_fphi_uind(const int inum_full, const int bsorder,
double ***host_thetai1, double ***host_thetai2,
double ***host_thetai3, int** igrid,
double *host_grid_brick_start, double ****host_grid_brick,
double ***host_thetai3, int** igrid, double ****host_grid_brick,
void **host_fdip_phi1, void **host_fdip_phi2, void **host_fdip_sum_phi,
int nzlo_out, int nzhi_out, int nylo_out, int nyhi_out,
int nxlo_out, int nxhi_out, bool& first_iteration) {
const int nzlo_out, const int nzhi_out,
const int nylo_out, const int nyhi_out,
const int nxlo_out, const int nxhi_out,
bool& first_iteration) {
AMOEBAMF.compute_fphi_uind(inum_full, bsorder, host_thetai1, host_thetai2,
host_thetai3, igrid, host_grid_brick_start, host_grid_brick, host_fdip_phi1,
host_thetai3, igrid, host_grid_brick, host_fdip_phi1,
host_fdip_phi2, host_fdip_sum_phi, nzlo_out, nzhi_out,
nylo_out, nyhi_out, nxlo_out, nxhi_out, first_iteration);
}

83
lib/gpu/lal_base_amoeba.cpp
View File

@ -186,8 +186,10 @@ void BaseAmoebaT::clear_atomic() {
_igrid.clear();
_fdip_phi1.clear();
_fdip_phi2.clear();
_cgrid_brick.clear();
_fdip_sum_phi.clear();
_cgrid_brick.clear();
hview_cgrid.clear();
dev_nspecial15.clear();
dev_special15.clear();
dev_special15_t.clear();
@ -563,10 +565,9 @@ template <class numtyp, class acctyp>
void BaseAmoebaT::precompute_induce(const int inum_full, const int bsorder,
double ***host_thetai1, double ***host_thetai2,
double ***host_thetai3, int** host_igrid,
double* host_grid_brick_start, double**** host_grid_brick,
int nzlo_out, int nzhi_out,
int nylo_out, int nyhi_out,
int nxlo_out, int nxhi_out) {
const int nzlo_out, const int nzhi_out,
const int nylo_out, const int nyhi_out,
const int nxlo_out, const int nxhi_out) {
_bsorder = bsorder;
@ -643,32 +644,6 @@ void BaseAmoebaT::precompute_induce(const int inum_full, const int bsorder,
dview_int[idx+2] = host_igrid[i][2];
}
ucl_copy(_igrid, dview_int, false);
}
// ---------------------------------------------------------------------------
// fphi_uind = induced potential from grid
// fphi_uind extracts the induced dipole potential from the particle mesh Ewald grid
// ---------------------------------------------------------------------------
template <class numtyp, class acctyp>
void BaseAmoebaT::compute_fphi_uind(const int inum_full, const int bsorder,
double ***host_thetai1, double ***host_thetai2,
double ***host_thetai3, int** host_igrid,
double *host_grid_brick_start, double ****host_grid_brick,
void** host_fdip_phi1, void **host_fdip_phi2, void **host_fdip_sum_phi,
int nzlo_out, int nzhi_out, int nylo_out, int nyhi_out,
int nxlo_out, int nxhi_out, bool& first_iteration)
{
// allocation/resize and transfers before the first iteration
if (first_iteration) {
precompute_induce(inum_full, bsorder, host_thetai1, host_thetai2, host_thetai3,
host_igrid, host_grid_brick_start, host_grid_brick, nzlo_out, nzhi_out,
nylo_out, nyhi_out, nxlo_out, nxhi_out);
first_iteration = false;
}
// update the cgrid_brick with data host
_nzlo_out = nzlo_out;
_nzhi_out = nzhi_out;
@ -681,8 +656,47 @@ void BaseAmoebaT::compute_fphi_uind(const int inum_full, const int bsorder,
_ngridx = nxhi_out - nxlo_out + 1;
_num_grid_points = _ngridx * _ngridy * _ngridz;
UCL_H_Vec<double> hview_cgrid;
hview_cgrid.alloc(_num_grid_points*2, *(this->ucl_device), UCL_READ_WRITE);
int numel = _num_grid_points*2;
if (_cgrid_brick.cols() == 0) {
hview_cgrid.alloc(numel, *(this->ucl_device), UCL_READ_WRITE);
_cgrid_brick.alloc(numel, *(this->ucl_device), UCL_READ_ONLY);
} else if (numel > _cgrid_brick.cols()) {
hview_cgrid.resize(numel);
_cgrid_brick.resize(numel);
}
}
// ---------------------------------------------------------------------------
// fphi_uind = induced potential from grid
// fphi_uind extracts the induced dipole potential from the particle mesh Ewald grid
// ---------------------------------------------------------------------------
template <class numtyp, class acctyp>
void BaseAmoebaT::compute_fphi_uind(const int inum_full, const int bsorder,
double ***host_thetai1, double ***host_thetai2,
double ***host_thetai3, int** host_igrid,
double ****host_grid_brick,
void** host_fdip_phi1,
void **host_fdip_phi2,
void **host_fdip_sum_phi,
const int nzlo_out, const int nzhi_out,
const int nylo_out, const int nyhi_out,
const int nxlo_out, const int nxhi_out,
bool& first_iteration)
{
// TODO: find out why this alloc helps makes the cgrid_brick ucl_copy work
UCL_H_Vec<numtyp> hview;
hview.alloc(1, *(this->ucl_device), UCL_READ_ONLY);
// allocation/resize and transfers before the first iteration
if (first_iteration) {
precompute_induce(inum_full, bsorder, host_thetai1, host_thetai2,
host_thetai3, host_igrid, nzlo_out, nzhi_out,
nylo_out, nyhi_out, nxlo_out, nxhi_out);
first_iteration = false;
}
int n = 0;
for (int iz = nzlo_out; iz <= nzhi_out; iz++)
for (int iy = nylo_out; iy <= nyhi_out; iy++)
@ -691,11 +705,8 @@ void BaseAmoebaT::compute_fphi_uind(const int inum_full, const int bsorder,
hview_cgrid[n+1] = host_grid_brick[iz][iy][ix][1];
n += 2;
}
//hview_cgrid.view(host_grid_brick_start, _num_grid_points*2, *(this->ucl_device));
_cgrid_brick.alloc(_num_grid_points*2, *(this->ucl_device), UCL_READ_ONLY);
ucl_copy(_cgrid_brick, hview_cgrid, false);
const int red_blocks = fphi_uind();
_fdip_phi1.update_host(_max_thetai_size*10);

19
lib/gpu/lal_base_amoeba.h
View File

@ -153,10 +153,9 @@ class BaseAmoeba {
virtual void precompute_induce(const int inum_full, const int bsorder,
double ***host_thetai1, double ***host_thetai2,
double ***host_thetai3, int** igrid,
double *host_grid_brick_start, double ****host_grid_brick,
int nzlo_out, int nzhi_out,
int nylo_out, int nyhi_out,
int nxlo_out, int nxhi_out);
const int nzlo_out, const int nzhi_out,
const int nylo_out, const int nyhi_out,
const int nxlo_out, const int nxhi_out);
/// Compute multipole real-space with device neighboring
virtual int** compute_multipole_real(const int ago, const int inum_full, const int nall,
@ -183,10 +182,12 @@ class BaseAmoeba {
virtual void compute_fphi_uind(const int inum_full, const int bsorder,
double ***host_thetai1, double ***host_thetai2,
double ***host_thetai3, int** igrid,
double *host_grid_brick_start, double ****host_grid_brick,
double ****host_grid_brick,
void **host_fdip_phi1, void **host_fdip_phi2, void **host_fdip_sum_phi,
int nzlo_out, int nzhi_out, int nylo_out, int nyhi_out,
int nxlo_out, int nxhi_out, bool& first_iteration);
const int nzlo_out, const int nzhi_out,
const int nylo_out, const int nyhi_out,
const int nxlo_out, const int nxhi_out,
bool& first_iteration);
/// Compute polar real-space with device neighboring
virtual void compute_polar_real(int *host_amtype, int *host_amgroup, double **host_rpole,
@ -255,7 +256,9 @@ class BaseAmoeba {
int _nmax, _max_tep_size, _max_fieldp_size;
int _bsorder;
UCL_D_Vec<numtyp> _thetai1, _thetai2, _thetai3, _cgrid_brick;
UCL_D_Vec<numtyp> _thetai1, _thetai2, _thetai3;
UCL_H_Vec<numtyp> hview_cgrid;
UCL_D_Vec<numtyp> _cgrid_brick;
UCL_D_Vec<int> _igrid;
UCL_Vector<numtyp,numtyp> _fdip_phi1, _fdip_phi2, _fdip_sum_phi;
int _max_thetai_size;