TimeConstrained for parallel computations

Question

The Mathematica manual states

TimeConstrained takes account only of CPU time spent inside the main Mathematica kernel process; it does not include additional threads or processes.

However, my computation involves a parallel computation. As an example, consider the code

count = 0;
SetSharedVariable[count];
ProgressIndicator[Dynamic[count], {0, 500}]

f = Total[ParallelTable[count++; i^2, {i, 1, 500}]];

TimeConstrained[f, .1, "fail"]

(* 41791750 *)

How can I achieve that "fail" is returned. I can imagine that it could be complicated to compute the total CPU time of all sub-processes, but for my application it would be sufficient to consider the absolute amount of real-time that has passed as AbsoluteTiming[] does.

Any suggestions?

Answer 1

You should change

f = Total[ParallelTable[count++; i^2, {i, 1, 500}]];

into

f := (count = 0; Total[ParallelTable[count++; i^2, {i, 1, 500}]]);

Then, at least in Mathematica v.9, it works.

The fact is that f must be calculated only inside TimeConstrained and not before. The counter must be zeroed before any parallel evaluation