when digging into this issue (issue135
I found one of the root cause is the lack of the implement to distribute the domains in parallel mode.
it is a really good time for us to implement it in C++.
However implementing it in cython could fix this issue in a short time and make us have a prototype to make sure the method work.
I propose we implement BOTH because
1. we have to migrate to c++ anyway in accordance with the earlier discussion and conclusion (issue153
). Yung-Yu is the best candidate to do that.
2. it is a good chance for me (Taihsiang) to get familiar with the Table structure by implementing the code relevant to it. I could implement the cython part.
3. the cython part could be a part of the unit test or function test for domain distribution in c++.
What do you think? Thanks.