tx258/lammps-gran-kokkos
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'USER-DPD_kokkos' into USER-DPD_GBhacks, bring in MPI deadlock fix

Browse Source
This commit is contained in:
Tim Mattox
2017-03-17 03:24:03 -04:00
parent ce9b3f4fcb 5ea096d08c
commit 4a7bc115cb
1 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
src/irregular.cpp
View File

@ -297,11 +297,9 @@ int Irregular::create_atom(int n, int *sizes, int *proclist, int sortflag)
// setup for collective comm
// work1 = 1 for procs I send a message to, not including self
// work2 = 1 for all procs, used for ReduceScatter
for (i = 0; i < nprocs; i++) {
work1[i] = 0;
work2[i] = 1;
}
for (i = 0; i < n; i++) work1[proclist[i]] = 1;
work1[me] = 0;
@ -318,7 +316,7 @@ int Irregular::create_atom(int n, int *sizes, int *proclist, int sortflag)
MPI_Allreduce(work1,work2,nprocs,MPI_INT,MPI_SUM,world);
nrecv_proc = work2[me];
#else
MPI_Reduce_scatter(work1,&nrecv_proc,work2,MPI_INT,MPI_SUM,world);
MPI_Reduce_scatter_block(work1,&nrecv_proc,1,MPI_INT,MPI_SUM,world);
#endif
#endif
@ -395,7 +393,9 @@ int Irregular::create_atom(int n, int *sizes, int *proclist, int sortflag)
sendmax_proc = 0;
for (i = 0; i < nsend_proc; i++) {
MPI_Send(&length_send[i],1,MPI_INT,proc_send[i],0,world);
MPI_Request tmpReq; // Use non-blocking send to avoid possible deadlock
MPI_Isend(&length_send[i],1,MPI_INT,proc_send[i],0,world,&tmpReq);
MPI_Request_free(&tmpReq); // the MPI_Barrier below marks completion
sendmax_proc = MAX(sendmax_proc,length_send[i]);
}
@ -543,11 +543,9 @@ int Irregular::create_data(int n, int *proclist, int sortflag)
// setup for collective comm
// work1 = 1 for procs I send a message to, not including self
// work2 = 1 for all procs, used for ReduceScatter
for (i = 0; i < nprocs; i++) {
work1[i] = 0;
work2[i] = 1;
}
for (i = 0; i < n; i++) work1[proclist[i]] = 1;
work1[me] = 0;
@ -564,7 +562,7 @@ int Irregular::create_data(int n, int *proclist, int sortflag)
MPI_Allreduce(work1,work2,nprocs,MPI_INT,MPI_SUM,world);
nrecv_proc = work2[me];
#else
MPI_Reduce_scatter(work1,&nrecv_proc,work2,MPI_INT,MPI_SUM,world);
MPI_Reduce_scatter_block(work1,&nrecv_proc,1,MPI_INT,MPI_SUM,world);
#endif
#endif
@ -641,7 +639,9 @@ int Irregular::create_data(int n, int *proclist, int sortflag)
sendmax_proc = 0;
for (i = 0; i < nsend_proc; i++) {
MPI_Send(&num_send[i],1,MPI_INT,proc_send[i],0,world);
MPI_Request tmpReq; // Use non-blocking send to avoid possible deadlock
MPI_Isend(&num_send[i],1,MPI_INT,proc_send[i],0,world,&tmpReq);
MPI_Request_free(&tmpReq); // the MPI_Barrier below marks completion
sendmax_proc = MAX(sendmax_proc,num_send[i]);
}