I have had issues computing L2-errors. Namely the issue lies in "para_sum" in messenger_elem.F. Namely, using the tracer test case from the branch "conv_test_example", and the sediment test from the test suite, I have found that the L2-error depends on the processor count. I believe this has to do with the fact that since ghost cells are instantiated on both domains, we are in effect double counting these cells (or triple counting them).
l2errorissue.xlsx
.
I have had issues computing L2-errors. Namely the issue lies in "para_sum" in messenger_elem.F. Namely, using the tracer test case from the branch "conv_test_example", and the sediment test from the test suite, I have found that the L2-error depends on the processor count. I believe this has to do with the fact that since ghost cells are instantiated on both domains, we are in effect double counting these cells (or triple counting them).
l2errorissue.xlsx
.