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

reformat for increased readability

Browse Source
This commit is contained in:
Axel Kohlmeyer
2021-09-16 07:25:04 -04:00
parent dc50db0675
commit 353b3a2bb3
5 changed files with 25 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
lib/gpu/lal_answer.h
View File

@ -127,9 +127,8 @@ class Answer {
/// Add forces and torques from the GPU into a LAMMPS pointer /// Add forces and torques from the GPU into a LAMMPS pointer
void get_answers(double **f, double **tor); void get_answers(double **f, double **tor);
inline double get_answers(double **f, double **tor, double *eatom, inline double get_answers(double **f, double **tor, double *eatom, double **vatom,
double **vatom, double *virial, double &ecoul, double *virial, double &ecoul, int &error_flag_in) {
int &error_flag_in) {
double ta=MPI_Wtime(); double ta=MPI_Wtime();
time_answer.sync_stop(); time_answer.sync_stop();
_time_cpu_idle+=MPI_Wtime()-ta; _time_cpu_idle+=MPI_Wtime()-ta;

11
lib/gpu/lal_device.cpp
View File

@ -992,10 +992,8 @@ int DeviceT::compile_kernels() {
static_cast<size_t>(_block_cell_2d) > gpu->group_size_dim(0) || static_cast<size_t>(_block_cell_2d) > gpu->group_size_dim(0) ||
static_cast<size_t>(_block_cell_2d) > gpu->group_size_dim(1) || static_cast<size_t>(_block_cell_2d) > gpu->group_size_dim(1) ||
static_cast<size_t>(_block_cell_id) > gpu->group_size_dim(0) || static_cast<size_t>(_block_cell_id) > gpu->group_size_dim(0) ||
static_cast<size_t>(_max_shared_types*_max_shared_types* static_cast<size_t>(_max_shared_types*_max_shared_types*sizeof(numtyp)*17 > gpu->slm_size()) ||
sizeof(numtyp)*17 > gpu->slm_size()) || static_cast<size_t>(_max_bio_shared_types*2*sizeof(numtyp) > gpu->slm_size()))
static_cast<size_t>(_max_bio_shared_types*2*sizeof(numtyp) >
gpu->slm_size()))
return -13; return -13;
if (_block_pair % _simd_size != 0 || _block_bio_pair % _simd_size != 0 || if (_block_pair % _simd_size != 0 || _block_bio_pair % _simd_size != 0 ||
@ -1071,9 +1069,8 @@ void lmp_clear_device() {
global_device.clear_device(); global_device.clear_device();
} }
double lmp_gpu_forces(double **f, double **tor, double *eatom, double lmp_gpu_forces(double **f, double **tor, double *eatom, double **vatom,
double **vatom, double *virial, double &ecoul, double *virial, double &ecoul, int &error_flag) {
int &error_flag) {
return global_device.fix_gpu(f,tor,eatom,vatom,virial,ecoul,error_flag); return global_device.fix_gpu(f,tor,eatom,vatom,virial,ecoul,error_flag);
} }

8
lib/gpu/lal_device.h
View File

@ -163,17 +163,15 @@ class Device {
{ ans_queue.push(ans); } { ans_queue.push(ans); }
/// Add "answers" (force,energies,etc.) into LAMMPS structures /// Add "answers" (force,energies,etc.) into LAMMPS structures
inline double fix_gpu(double **f, double **tor, double *eatom, inline double fix_gpu(double **f, double **tor, double *eatom, double **vatom,
double **vatom, double *virial, double &ecoul, double *virial, double &ecoul, int &error_flag) {
int &error_flag) {
error_flag=0; error_flag=0;
atom.data_unavail(); atom.data_unavail();
if (ans_queue.empty()==false) { if (ans_queue.empty()==false) {
stop_host_timer(); stop_host_timer();
double evdw=0.0; double evdw=0.0;
while (ans_queue.empty()==false) { while (ans_queue.empty()==false) {
evdw+=ans_queue.front()->get_answers(f,tor,eatom,vatom,virial,ecoul, evdw += ans_queue.front()->get_answers(f,tor,eatom,vatom,virial,ecoul,error_flag);
error_flag);
ans_queue.pop(); ans_queue.pop();
} }
return evdw; return evdw;

18
src/GPU/fix_gpu.cpp
View File

@ -50,12 +50,10 @@ extern int lmp_init_device(MPI_Comm world, MPI_Comm replica, const int ngpu,
const int ocl_platform, char *device_type_flags, const int ocl_platform, char *device_type_flags,
const int block_pair); const int block_pair);
extern void lmp_clear_device(); extern void lmp_clear_device();
extern double lmp_gpu_forces(double **f, double **tor, double *eatom, extern double lmp_gpu_forces(double **f, double **tor, double *eatom, double **vatom,
double **vatom, double *virial, double &ecoul, double *virial, double &ecoul, int &err_flag);
int &err_flag); extern double lmp_gpu_update_bin_size(const double subx, const double suby, const double subz,
extern double lmp_gpu_update_bin_size(const double subx, const double suby, const int nlocal, const double cut);
const double subz, const int nlocal,
const double cut);
static const char cite_gpu_package[] = static const char cite_gpu_package[] =
"GPU package (short-range, long-range and three-body potentials):\n\n" "GPU package (short-range, long-range and three-body potentials):\n\n"
@ -327,13 +325,11 @@ void FixGPU::post_force(int /* vflag */)
double lvirial[6]; double lvirial[6];
for (int i = 0; i < 6; i++) lvirial[i] = 0.0; for (int i = 0; i < 6; i++) lvirial[i] = 0.0;
int err_flag; int err_flag;
double my_eng = lmp_gpu_forces(atom->f, atom->torque, force->pair->eatom, double my_eng = lmp_gpu_forces(atom->f, atom->torque, force->pair->eatom, force->pair->vatom,
force->pair->vatom, lvirial, lvirial, force->pair->eng_coul, err_flag);
force->pair->eng_coul, err_flag);
if (err_flag) { if (err_flag) {
if (err_flag==1) if (err_flag==1)
error->one(FLERR, error->one(FLERR,"Too many neighbors on GPU. Use neigh_modify one to increase limit.");
"Too many neighbors on GPU. Use neigh_modify one to increase limit.");
} }
force->pair->eng_vdwl += my_eng; force->pair->eng_vdwl += my_eng;