int WalkerControlBase::copyWalkers(MCWalkerConfiguration& W) {
//clear the WalkerList to populate them with the good walkers
W.clear();
W.insert(W.begin(), good_w.begin(), good_w.end());
int cur_walker = good_w.size();
for(int i=0; i<good_w.size(); i++) { //,ie+=ncols) {
for(int j=0; j<ncopy_w[i]; j++, cur_walker++)
{
Walker_t* awalker=new Walker_t(*(good_w[i]));
awalker->ID=(++NumWalkersCreated)*NumContexts+MyContext;
awalker->ParentID=good_w[i]->ParentID;
W.push_back(awalker);
}
}
//clear good_w and ncopy_w for the next branch
good_w.clear();
ncopy_w.clear();
return W.getActiveWalkers();
}
/** swap walkers using (low,high) ordered pairs
*
* The communication occur only between the (low,high) pairs.
* This does not guarantee a perfect balance swapWalkersBlocked and swapWalkersAsync
* try to achieve. However, the number of messages and their sizes are less than
* other methods.
*/
void GlobalWalkerControl::swapWalkersMap(MCWalkerConfiguration& W) {
NumSwaps++;
multimap<int,int> nw_map;
for(int i=0; i<NumContexts; i++) {
nw_map.insert(pair<int,int>(NumPerNode[i],i));
}
// multimap key is sorted with ascending order
multimap<int,int>::iterator it(nw_map.begin());
multimap<int,int>::reverse_iterator it_b(nw_map.end());
bool notpaired=true;
int target_context=-1;
int half=NumContexts/2;
int item=0;
bool minorcontext;
while(notpaired &&item<half) {
int i=(*it).second;
int j=(*it_b).second;
if(i == MyContext) {
target_context=j;
notpaired=false;
minorcontext=true;
} else if(j == MyContext) {
target_context= i;
notpaired=false;
minorcontext=false;
}
++it; ++it_b; ++item;
}
int nw_tot=NumPerNode[MyContext]+NumPerNode[target_context];
int nw_L=nw_tot/2;
int nw_R=nw_tot-nw_L;
int dnw(0);
if(minorcontext) {
dnw=nw_L-NumPerNode[MyContext];//how many to get
} else {
dnw=NumPerNode[MyContext]-nw_R;//how many to send
}
if(dnw) {//something to swap
if(minorcontext) {//open recv buffer
Walker_t& wRef(*W[0]);
OOMPI_Packed recvBuffer(dnw*wRef.byteSize(),OOMPI_COMM_WORLD);
OOMPI_COMM_WORLD[target_context].Recv(recvBuffer);
//create walkers
int last = W.getActiveWalkers();
while(dnw) {
Walker_t *awalker= new Walker_t(wRef);
awalker->getMessage(recvBuffer);
W.push_back(awalker);
--dnw; ++last;
}
} else {
Walker_t& wRef(*W[0]);
OOMPI_Packed sendBuffer(dnw*wRef.byteSize(),OOMPI_COMM_WORLD);
int last=W.getActiveWalkers()-1;
while(dnw) {
W[last]->putMessage(sendBuffer);
--dnw; --last;
}
OOMPI_COMM_WORLD[target_context].Send(sendBuffer);
//last=WalkerList.size()-1;
//while(dnw_save) {
// delete WalkerList[last];
// WalkerList.pop_back();
// --dnw_save; --last;
//}
//destroyWalkers(WalkerList.begin()+nsub[MyContext], WalkerList.end());
W.destroyWalkers(W.begin()+nw_R, W.end());
}
}
/* not used yet
struct lessNode {
inline bool operator()(const pair<int,int>& a,
const pair<int,int>& b) const {
return a.second < b.second;
}
};
//typedef pair<int,int> mytype;
//vector<mytype> id2n(NumContexts);
//for(int i=0; i<NumContexts; i++) {
// id2n=pair<int,int>(i,NumPerNode[i]);
//}
//
//std::sort(id2n.begin(),id2n.end(),lessNode);
*/
}
/** swap Walkers with Irecv/Send
*
* The algorithm ensures that the load per node can differ only by one walker.
* The communication is one-dimensional.
*/
void GlobalWalkerControl::swapWalkersAsync(MCWalkerConfiguration& W) {
NumSwaps++;
OffSet[0]=0;
for(int i=0; i<NumContexts; i++) OffSet[i+1]=OffSet[i]+NumPerNode[i];
FairPartition(Cur_pop,NumContexts,FairOffSet);
int toLeft=FairOffSet[MyContext]-OffSet[MyContext];
int toRight=FairOffSet[MyContext+1]-OffSet[MyContext+1];
Walker_t& wRef(*W[0]);
int num_deleted=0;
int last=NumPerNode[MyContext]-1;
OOMPI_Packed* recvLeftBuffer=0;
OOMPI_Packed* recvRightBuffer=0;
OOMPI_Request recvLeftRequest;
OOMPI_Request recvRightRequest;
//initiate Irecv
if(toLeft<0) { // recv from node-1
int dn=-toLeft;
recvLeftBuffer = new OOMPI_Packed(dn*wRef.byteSize(),OOMPI_COMM_WORLD);
recvLeftRequest = OOMPI_COMM_WORLD[MyContext-1].Irecv(*recvLeftBuffer,MPI_ANY_TAG);
}
if(toRight>0) { //recv from node+1
recvRightBuffer = new OOMPI_Packed(toRight*wRef.byteSize(),OOMPI_COMM_WORLD);
recvRightRequest = OOMPI_COMM_WORLD[MyContext+1].Irecv(*recvRightBuffer,MPI_ANY_TAG);
}
//send the data
if(toLeft>0) { //send to node-1
int dn=toLeft;
num_deleted+=dn;
OOMPI_Packed sendBuffer(dn*wRef.byteSize(),OOMPI_COMM_WORLD);
while(dn) {
W[last]->putMessage(sendBuffer); --dn; --last;
}
OOMPI_COMM_WORLD[MyContext-1].Send(sendBuffer);
}
if(toRight<0) { // send to node+1
int dn=-toRight;
num_deleted+=dn;
OOMPI_Packed sendBuffer(dn*wRef.byteSize(),OOMPI_COMM_WORLD);
while(dn) {
W[last]->putMessage(sendBuffer); --dn; --last;
}
OOMPI_COMM_WORLD[MyContext+1].Send(sendBuffer);
}
while(num_deleted>0) {
W.pop_back(); --num_deleted;
}
if(recvLeftBuffer) {
recvLeftRequest.Wait();
int dn=-toLeft;
while(dn) {
Walker_t *awalker= new Walker_t(wRef);
awalker->getMessage(*recvLeftBuffer);
W.push_back(awalker);
--dn;
}
delete recvLeftBuffer;
recvLeftBuffer=0;
}
if(recvRightBuffer) {
recvRightRequest.Wait();
int dn=toRight;
while(dn) {
Walker_t *awalker= new Walker_t(wRef);
awalker->getMessage(*recvRightBuffer);
W.push_back(awalker);
--dn;
}
delete recvRightBuffer;
recvRightBuffer=0;
}
}
