CPU Core Utilization in HPCGAP/WSL

[jaslamMiktex]

I have a working program on HPCGAP (using Ryzen 7- 8-core CPU), using WSL/HPCGAP build
But I do not see more than 2 cores being used in the WSL.
Here is the prototype template of my real application (real app not shown):

############################
createTasks := function(n)
local i, t, tr, sum, sumx;
   t := [];
   arrayDB[1] := [1 .. n];
   sumx := [];
   for i in [1 ..n ] do
      t[i] := RunTask(taskFunction,  i);
   od;
  sum := [];

   for i in [1 .. n] do
     WaitTask(t[i]);
     Add(sum, arrayDB[i]);
   od;
 return sum[Length(sum)];
 end;
#-------------------
taskFunction := function(r)
 local k, arx;
 arx := arrayDB[r];
 arrayDB[r+1] := List(arx, x->x+1);
# return arrayDB[r+1];
 return;
 end;
##############################

Any ideas will be highly appreciated.

Thank you

[fingolfin]

What does GAPInfo.KernelInfo.NUM_CPUS report? It may have not recognized the number of CPU cores correctly.

You could use use SetMaxTaskWorkers to get it use more worker threads, e.g. SetMaxTaskWorkers(12) to use up to 12 worker threads.

[jaslamMiktex]

Thank you Max. Will try that.

[jaslamMiktex]

Latest trial:
Finally, I was able to run this HPC-GAP application with 100% CPU utilization on all the cores, giving a 3X improvement
in the running time.
SetMaxTaskWorkers (4) or SetMaxTaskWorkers(16) does not make more 10% difference in the performance (running time).
At the startup, HPC GAP declares that only "4 worker threads are being used."
PS: With all the (hopefully) correct settings like ShareObj(), etc, I continued into the "read-access" violation problem.
The only resolution I found was using DISABLE_GUARDS(2);
<--------------------------------------------------------- Thank you

[ChrisJefferson]

I just want to point out that disabling guards turns off various checks. While your calculation may produce the correct answer, there is also a chance you might get very wrong answers -- I would not trust any generated result unless you have an independent way to verify it (having some way of checking your answer is of course always a good idea when possible, but particularly in this case).

[jaslamMiktex]

Yes, I do have a sequential version (using GAP 4.12.2) as well to verify.
There is no shared write area, so I do not see why would that be a problem. Those read-only data structures remain read-only-- never modified. Almost like a software ROM module.