Construct L and U from dgbtrf

[cornholio]

Hi,

I am using the LAPACK function dgbtrf to factorize a band matrix. I know it is usual to just hand over the result matrix to a solver (dgbtrs), but I need to obtain two separate matrices L and U. Unfortunately I don't understand in which way those matrices are stored in the output matrix of dgbtrf. U is quite obvious, but how can I derive L from those strange factors m(i,j)?

Thanks!

[kdallen]

I had a similar question about DSYTRF. The following link might help:

viewtopic.php?f=2&t=667