Add mps_add function

[MatthiasReumann]

Changelog

• Add mps_add function (More or less a 1:1 copy of the PyTeNet implementation)
• Add test_mps_add test function (Generate Random MPS ➡ mps_add ➡ Compare state-vectors)
• Update README.md (Add dependency installation instructions for MacOS)
• Add .vscode/ to .gitignore

[cmendl]

Thanks for the nice work! Could you modify the pull request by additionally taking these suggestions into account?

• The data type of the logical output MPS should also be struct mps* (instead of a void pointer). Use allocate_empty_mps to allocate it inside mps_add.
• Try to rebase the new branch to the latest version of the main branch.
• For consistency, please start comments with lower-case and without a final dot.
• In the test, use a literal number instead of "site_array_len" (since "int site_array[site_array_len]" is not allowed in standard C).
• The test function "char* test_mps_add()" should come directly after char* test_mps_vdot() (to follow the same order as in mps.c)

[MatthiasReumann]

Thanks for the nice work! Could you modify the pull request by additionally taking these suggestions into account?

• The data type of the logical output MPS should also be struct mps* (instead of a void pointer). Use allocate_empty_mps to allocate it inside mps_add.
• Try to rebase the new branch to the latest version of the main branch.
• For consistency, please start comments with lower-case and without a final dot.
• In the test, use a literal number instead of "site_array_len" (since "int site_array[site_array_len]" is not allowed in standard C).
• The test function "char* test_mps_add()" should come directly after char* test_mps_vdot() (to follow the same order as in mps.c)

Thanks for the feedback! I've updated the code accordingly.

[cmendl]

Thanks again! Minor additional point: the local variable qsite is allocated but not freed; actually, instead of allocating this array, one can directly pass psi->qsite to allocate_empty_mps (since it creates its own copy internally).

[MatthiasReumann]

Thanks again! Minor additional point: the local variable qsite is allocated but not freed; actually, instead of allocating this array, one can directly pass psi->qsite to allocate_empty_mps (since it creates its own copy internally).

I've removed the qsite copying now, thanks!

[cmendl]

Final minor point: currently the automatic unit test does not run due to const qnumber qsite_chi[d] in the unit test; please use an explicit integer here as well, i.e., const qnumber qsite_chi[2]

[MatthiasReumann]

Final minor point: currently the automatic unit test does not run due to const qnumber qsite_chi[d] in the unit test; please use an explicit integer here as well, i.e., const qnumber qsite_chi[2]

Ah yes, sorry for that.

[MatthiasReumann]

Interesting. Locally ctest works without any troubles. I will build and run the tests on lxhalle (or similar) soon and try to fix the unit test.

[cmendl]

There is a problem in the "qnumber_all_equal" assertions: the first argument should be the logical dimension instead of "ndim", for example, assert(qnumber_all_equal(chi_a.dim_logical[0], ...)). Also, in the unit test, the quantum number sector qnum_sector = encode_quantum_number_pair(q_pnum, q_spin) leads to MPS which are logical zero vectors (since the quantum number restriction cannot be satisfied); instead, you can set qnum_sector to a small integer.

[cmendl]

Thanks! I've now squash-merged the pull request into the main branch.

[MatthiasReumann]

Great - thanks for the patience and sorry for the hiccups!