- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi
I am running a diffusion finite element program which has approximately 70,000 unknowns. The sparse coefficient array is built in double precision and solved hundreds of times over the duration of a transient. We would like to minimize the run time as there will be thousands of cases to simulate.
1) Any suggestions for which routines would best meet this requirement?
2) Suggested hardware (CPU, GPU, etc.)
3) Can you please provide sample coding for implementing the suggested solution routine?
Thank you!
Dennis Kawa
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Wouldn't you rather talk to the MKL team who have Pardiso in their library? Tuned and ready to go.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
If you do a search on this forum and the mkl forum for PARDISO, you will find a set of long posts, mostly by the expert @mecej4, where he, sorry I have always assumed this person is a he, but the name gives no clues, but the likely age is a male old school Fortran programmer, so continuing @mecej4 has shown the code to develop the input and output to the type of problem you are concerned about. There needs to be a word that is gender neutral and singular.
There is code etc... You will just have to hunt.
@mecej4 is very good and will often answer interesting questions, although he and I have some differing ideas of interesting, lol, he is much more selective.

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page