This fix seems to work. The free memory error is not due to the wavelet operator. It is due to the communicator going out of scope each time the benchmark is run.
Me and @dpshelio found a similar error during the development main.cc, which we fixed. The fix also consistent with the other MPI benchmarks, which have the communicator in the fixture.
This fix seems to work. The free memory error is not due to the wavelet operator. It is due to the communicator going out of scope each time the benchmark is run.
Me and @dpshelio found a similar error during the development main.cc, which we fixed. The fix also consistent with the other MPI benchmarks, which have the communicator in the fixture.