Mid-term pre-post.

Hi all,

This is a short post just to clarify my state at midterm. As I have arranged with Kai, at the beginning of the next week I will write a verbose post to explain all the details related to the development of "ilu" function. I need a couple of days to tidy up all the code and make it presentable.

The state of the code now is functional. It lacks of tests and documentation and need a bit of re-factorization I will do this weekend. I will list several relevant points about the implementation of the function.

• It is programmed to output the same as Matlab.

• My version is at least as fast as Matlab's, outperforming by small amounts of time in large output cases.

• I found a bug. At least on Matlab2013a regarding "ilutp" option when using milu="col". The col sum is not preserved in at least one case so I found in my testing cases that my function does not output the same. I will explain with more detail that issue next week.

 In the end I am happy with the performance of the function and that I'm meeting the time-line proposed at the beginning of GSOC (with a few days of delay maybe).

Future work: The second part of the GSOC I will implement "ichol" function. There are several points to discuss about its development with Kai because he implemented the code last year but there were some kind of issues with the licensing of it. This period is a bit longer and I will have no classes nor exams. Because of that, if I have time remaining at the end, I can start implementing "minres" or "lsqr" algorithms that Octave lacks of too. So there will be no time wasted.

See you,

Eduardo