Determinants for matrices with >4 rows and columns

[thosvarley]

Concisely describe the proposed feature The restriction on the size of matrices that Taichi can compute the determinant of is extremely limiting. The matrix determinant (and the log-determinant) are fundamental operations in linear algebra. It would be really helpful to be able to work on arbitrarily sized matrices (or even just matrices up to 10x10 if arbitrary sizes are too complex).

Describe the solution you'd like (if any) A generalized determinant function. If the code for the small matrices is extremely well-optimized, maybe two options: a small matrix, and then something like an LU determinant for larger matrices?

Additional comments I have tried repeatedly to roll my own LU decomposition algorithm for determinants, but it seems to break the autodiff in strange ways, so I'm turning to the experts.