portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-17 07:17:06 +01:00
Code Releases Activity

Update to use shared memory to contain the stencil comms buffers

Browse Source 
Tested on 2.1.1.1 1.2.1.1 4.1.1.1 1.4.1.1 2.2.1.1 subnode decompositions
This commit is contained in:
azusayamaguchi
2016-10-24 17:30:43 +01:00
parent ea25a4d9ac
commit b6a65059a2
13 changed files with 706 additions and 458 deletions
Download Patch File Download Diff File Expand all files Collapse all files

73
lib/Stencil.h
View File

@ -290,10 +290,11 @@ PARALLEL_FOR_LOOP
// Unified Comms buffers for all directions
///////////////////////////////////////////////////////////
// Vectors that live on the symmetric heap in case of SHMEM
std::vector<commVector<scalar_object> > u_simd_send_buf_hide;
std::vector<commVector<scalar_object> > u_simd_recv_buf_hide;
commVector<cobj> u_send_buf;
commVector<cobj> u_recv_buf_hide;
// std::vector<commVector<scalar_object> > u_simd_send_buf_hide;
// std::vector<commVector<scalar_object> > u_simd_recv_buf_hide;
// commVector<cobj> u_send_buf_hide;
// commVector<cobj> u_recv_buf_hide;
// These are used; either SHM objects or refs to the above symmetric heap vectors
// depending on comms target
cobj* u_recv_buf_p;
@ -439,36 +440,19 @@ PARALLEL_FOR_LOOP
/////////////////////////////////////////////////////////////////////////////////
const int Nsimd = grid->Nsimd();
uint8_t *shm_ptr = (uint8_t *)_grid->ShmBufferSelf();
_grid->ShmBufferFreeAll();
u_simd_send_buf.resize(Nsimd);
u_simd_recv_buf.resize(Nsimd);
u_send_buf.resize(_unified_buffer_size);
if( ShmDirectCopy && shm_ptr != NULL ) {
u_recv_buf_p=(cobj *)shm_ptr; shm_ptr+= _unified_buffer_size*sizeof(cobj);
for(int l=0;l<Nsimd;l++){
u_simd_send_buf[l] = (scalar_object *)shm_ptr; shm_ptr += _unified_buffer_size*sizeof(scalar_object);
u_simd_recv_buf[l] = (scalar_object *)shm_ptr; shm_ptr += _unified_buffer_size*sizeof(scalar_object);
}
} else {
u_recv_buf_hide.resize(_unified_buffer_size);
u_simd_send_buf_hide.resize(Nsimd,commVector<scalar_object>(_unified_buffer_size));
u_simd_recv_buf_hide.resize(Nsimd,commVector<scalar_object>(_unified_buffer_size));
u_recv_buf_p=&u_recv_buf_hide[0];
for(int l=0;l<Nsimd;l++){
u_simd_send_buf[l] = & u_simd_send_buf_hide[l][0];
u_simd_recv_buf[l] = & u_simd_recv_buf_hide[l][0];
}
u_send_buf_p=(cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
u_recv_buf_p=(cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
for(int l=0;l<Nsimd;l++){
u_simd_recv_buf[l] = (scalar_object *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(scalar_object));
u_simd_send_buf[l] = (scalar_object *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(scalar_object));
}
PrecomputeByteOffsets();
}
void Local (int point, int dimension,int shiftpm,int cbmask)
@ -698,6 +682,7 @@ PARALLEL_FOR_LOOP
calls++;
Mergers.resize(0);
Packets.resize(0);
_grid->StencilBarrier();
HaloGather(source,compress);
this->CommunicateBegin(reqs);
this->CommunicateComplete(reqs);
@ -836,19 +821,17 @@ PARALLEL_FOR_LOOP
// try the direct copy if possible
/////////////////////////////////////////////////////////
cobj *u_send_buf_p = &u_send_buf[0];
if (ShmDirectCopy) {
cobj *shm = (cobj *) _grid->ShmBuffer(xmit_to_rank);
if ( shm!=NULL) {
u_send_buf_p = shm;
}
cobj *send_buf = (cobj *)_grid->ShmBufferTranslate(xmit_to_rank,u_recv_buf_p);
if ( (ShmDirectCopy==0)||send_buf==NULL ) {
cobj *send_buf = u_send_buf_p;
}
t_data-=usecond();
Gather_plane_simple_table (face_table[face_idx],rhs,u_send_buf_p,compress,u_comm_offset,so); face_idx++;
Gather_plane_simple_table (face_table[face_idx],rhs,send_buf,compress,u_comm_offset,so); face_idx++;
t_data+=usecond();
AddPacket((void *)&u_send_buf_p[u_comm_offset],
AddPacket((void *)&send_buf[u_comm_offset],
(void *)&u_recv_buf_p[u_comm_offset],
xmit_to_rank,
recv_from_rank,
@ -947,18 +930,16 @@ PARALLEL_FOR_LOOP
_grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank);
AddPacket((void *)sp,(void *)rp,xmit_to_rank,recv_from_rank,bytes);
auto shm_or_rp = rp;
if (ShmDirectCopy) {
scalar_object *shm = (scalar_object *) _grid->ShmBufferTranslate(xmit_to_rank,sp);
if ( shm!=NULL) {
shm_or_rp = shm;
}
}
scalar_object *shm = (scalar_object *) _grid->ShmBufferTranslate(recv_from_rank,sp);
if ((ShmDirectCopy==0)||(shm==NULL)) {
shm = rp;
}
rpointers[i] = shm_or_rp;
// if Direct, StencilSendToRecvFrom will suppress copy to a peer on node
// assuming above pointer flip
AddPacket((void *)sp,(void *)rp,xmit_to_rank,recv_from_rank,bytes);
rpointers[i] = shm;
} else {