I am executing a ParallelTable but the code fails to finish. This is the setup. f is a function.

uVec[t_] := Table[Subscript[u, i][t], {i, 0, n}](*Functions unknown*) 
f = Simplify[(Total[bVec[x]*uVec[t]]) /. bCoef][[1]];
fxx = D[f, x, x];
fxxx = D[f, x, x, x];

(*The inner product function*)
inner[{f_, j_}] :=
    If[j == 0,
      Integrate[((1/(Pi)) ChebyshevT[j, x]*(f)/(Sqrt[1 - x^(2)])), {x, -1, 1}],
      Integrate[((2/(Pi)) ChebyshevT[j, x]*(f)/(Sqrt[1 - x^(2)])), {x, -1, 1}]]

(* The inner product of f *)

I have used monitor to determine where the code is at in its execution. This ParallelTable executes all values of j and then returns the expected result.

Monitor[z = 0; 
    fExp = ParallelTable[z += 1;
        inner[{Table[ CoefficientList[f, Subscript[u, i][t]][[2]], {i, 0, n}], j}],
    {j, 0, n}], z];

This ParallelTable computes inner for all values of j but for large n greater than about 50, fails to return any result and remains stuck in running mode.

Monitor[z = 0; expr = ParallelTable[z += 1; inner[{fxx, j}], {j, 0, n}], z];

I am hoping for any answers as to why Mathematica is able to complete the computation but fails to return the resulting table.

  • $\begingroup$ After more debugging, I have monitored the bytecount of expr during the unending function. It seems to be growing steadily. I suspect the total computation has been completed on each kernel and now it is compiling the result into the final table. What can I do to speed up this process? expr = ConstantArray[0, n + 1]; SetSharedVariable[expr] Monitor[ParallelDo[expr[[j + 1]] = inner[fxx, j], {j, 0, n}];, ByteCount[expr]] $\endgroup$
    – user62522
    Jan 24, 2019 at 15:21
  • $\begingroup$ Ok with enough patience it seems to compile the results into the full table. Now that I have discovered more information I will post a new question. $\endgroup$
    – user62522
    Jan 24, 2019 at 15:28
  • $\begingroup$ The answer to this problem can be explained here. mathematica.stackexchange.com/questions/181265/… $\endgroup$
    – user62522
    Jan 24, 2019 at 17:23


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.