% A program to calculate pi in parallel fashion
% Modelled after the MPI "pi.f" program

% clear;
% "hosts" is defined at startup in .octaverc - don't clear!

PI25DT = 3.141592653589793238462643
NMAX=10000000;
% NMAX is the number of terms in the expansion for pi

% First, single processor code:
t0 = clock();

h=1/NMAX;
a=1:NMAX;
b=h*(a-0.5);
PI=h*sum(4./(1+b.*b))
ERROR=PI25DT-PI

elapsed_time=etime(clock(),t0)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Now, multiprocessor code:

sockets = connect(hosts);
% open a connection to the variable hosts; hosts=["n50"; "n51"; ... "n55"];

if 1==2,
% verify all connections - myid runs from 1-5
% myid = 0 is the master node
  for k=2:6,
    reval( "send(myid,sockets(1,:))", sockets(k,:));
    myid_k = recv(sockets(k,:))
  end
end


t0 = clock();

nproc=6;
h=1/NMAX;

% first send over a few variables:
for l=2:6,
  send(nproc,sockets(l,:));
  reval( "nproc=recv(sockets(1,:));",sockets(l,:));
  send(NMAX,sockets(l,:));
  reval( "NMAX=recv(sockets(1,:));",sockets(l,:));
  send(h,sockets(l,:));
  reval( "h=recv(sockets(1,:));",sockets(l,:));
end

% now send over the command(s) to execute:
for l=2:6,
  reval(["a=(myid+1):nproc:NMAX"; "b=h*(a-0.5)"; "c=h*sum(4./(1+b.*b))"],sockets(l,:));
end

% now calculate for l=1, the master node, while the others grind away as well...
h=1/NMAX;
a=1:nproc:NMAX;
b=h*(a-0.5);
c=h*sum(4.0./(1 + b.*b));

% send the answers, "c", from the slave nodes to the master node.  Store them in psum:
psum=zeros(1,6);
for l=2:6,
  reval("send(c,sockets(1,:))",sockets(l,:));
  psum(l)=recv(sockets(l,:));
% blocking receives; code won't continue until something is received
end

PI=c + sum(psum)
ERROR=PI25DT-PI

elapsed_time=etime(clock(),t0)

closeall(sockets);