ldlt functions

[Daniel Lee]

Marcus, Ben, or someone else, can you take care of the final few things so we can get this merged? I'm looking for a response with an estimate of time to completion (on the resolution of days).


From https://github.com/stan-dev/stan/pull/419#issuecomment-30100678:

Before this pull request gets in, these are the final things that need to be done:
- [ ] create a `check_ldlt_factor()` validation function or something like it.
- [ ] The distributions that create and use ldlt factor should use the check function and not throw directly from the distribution code itself. These are the ones I found:
    - src/stan/prob/distributions/multivariate/continuous/inv_wishart.hpp
    - src/stan/prob/distributions/multivariate/continuous/matrix_normal.hpp
    - src/stan/prob/distributions/multivariate/continuous/multi_gp.hpp
    - src/stan/prob/distributions/multivariate/continuous/multi_normal.hpp
    - src/stan/prob/distributions/multivariate/continuous/multi_normal_prec.hpp
    - src/stan/prob/distributions/multivariate/continuous/multi_student_t.hpp
    - src/stan/prob/distributions/multivariate/continuous/wishart.hpp
- [ ] For the unit tests written, make sure an `ASSERT_TRUE` is called with the `LDLT_factor::success` method so someone else writing additional functions with the LDLT_factor can use that as an example.
- [ ] Write at least one unit test for each of LDLT_factor and LDLT_alloc. Having an example in code of how to instantiate it will help immensely if we ever need to fix any of the code.
- [ ] remove any compiler warnings that are introduced in this branch (they're usually comparison between int and unsigned int warnings -- they show up in g++, but not clang++)

A lot of the headers and tests are getting moved, so it would be a lot less work to get this in soon.

Daniel

[Ben Goodrich]

I have this 90% done. -- Ben

[Marcus Brubaker]

Thanks Ben.  Sorry Daniel for not commenting on the PR sooner, I've been swamped with other projects the past few days and haven't had the time.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Daniel Lee]

Thanks. I'm glad it's still getting worked on. I don't want it to get lost -- it'll speed up a lot of models, but I want to make sure someone can maintain it going forward.